* [PATCH v4 01/14] drm/panthor: Add uAPI
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-02-08 14:29 ` Liviu Dudau
2024-01-22 16:30 ` [PATCH v4 02/14] drm/panthor: Add GPU register definitions Boris Brezillon
` (16 subsequent siblings)
17 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
Panthor follows the lead of other recently submitted drivers with
ioctls allowing us to support modern Vulkan features, like sparse memory
binding:
- Pretty standard GEM management ioctls (BO_CREATE and BO_MMAP_OFFSET),
with the 'exclusive-VM' bit to speed-up BO reservation on job submission
- VM management ioctls (VM_CREATE, VM_DESTROY and VM_BIND). The VM_BIND
ioctl is loosely based on the Xe model, and can handle both
asynchronous and synchronous requests
- GPU execution context creation/destruction, tiler heap context creation
and job submission. Those ioctls reflect how the hardware/scheduler
works and are thus driver specific.
We also have a way to expose IO regions, such that the usermode driver
can directly access specific/well-isolate registers, like the
LATEST_FLUSH register used to implement cache-flush reduction.
This uAPI intentionally keeps usermode queues out of the scope, which
explains why doorbell registers and command stream ring-buffers are not
directly exposed to userspace.
v4:
- Add a VM_GET_STATE ioctl
- Fix doc
- Expose the CORE_FEATURES register so we can deal with variants in the
UMD
- Add Steve's R-b
v3:
- Add the concept of sync-only VM operation
- Fix support for 32-bit userspace
- Rework drm_panthor_vm_create to pass the user VA size instead of
the kernel VA size (suggested by Robin Murphy)
- Typo fixes
- Explicitly cast enums with top bit set to avoid compiler warnings in
-pedantic mode.
- Drop property core_group_count as it can be easily calculated by the
number of bits set in l2_present.
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
---
Documentation/gpu/driver-uapi.rst | 5 +
include/uapi/drm/panthor_drm.h | 945 ++++++++++++++++++++++++++++++
2 files changed, 950 insertions(+)
create mode 100644 include/uapi/drm/panthor_drm.h
diff --git a/Documentation/gpu/driver-uapi.rst b/Documentation/gpu/driver-uapi.rst
index c08bcbb95fb3..7a667901830f 100644
--- a/Documentation/gpu/driver-uapi.rst
+++ b/Documentation/gpu/driver-uapi.rst
@@ -17,3 +17,8 @@ VM_BIND / EXEC uAPI
:doc: Overview
.. kernel-doc:: include/uapi/drm/nouveau_drm.h
+
+drm/panthor uAPI
+================
+
+.. kernel-doc:: include/uapi/drm/panthor_drm.h
diff --git a/include/uapi/drm/panthor_drm.h b/include/uapi/drm/panthor_drm.h
new file mode 100644
index 000000000000..47376cfaa949
--- /dev/null
+++ b/include/uapi/drm/panthor_drm.h
@@ -0,0 +1,945 @@
+/* SPDX-License-Identifier: MIT */
+/* Copyright (C) 2023 Collabora ltd. */
+#ifndef _PANTHOR_DRM_H_
+#define _PANTHOR_DRM_H_
+
+#include "drm.h"
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+/**
+ * DOC: Introduction
+ *
+ * This documentation describes the Panthor IOCTLs.
+ *
+ * Just a few generic rules about the data passed to the Panthor IOCTLs:
+ *
+ * - Structures must be aligned on 64-bit/8-byte. If the object is not
+ * naturally aligned, a padding field must be added.
+ * - Fields must be explicitly aligned to their natural type alignment with
+ * pad[0..N] fields.
+ * - All padding fields will be checked by the driver to make sure they are
+ * zeroed.
+ * - Flags can be added, but not removed/replaced.
+ * - New fields can be added to the main structures (the structures
+ * directly passed to the ioctl). Those fields can be added at the end of
+ * the structure, or replace existing padding fields. Any new field being
+ * added must preserve the behavior that existed before those fields were
+ * added when a value of zero is passed.
+ * - New fields can be added to indirect objects (objects pointed by the
+ * main structure), iff those objects are passed a size to reflect the
+ * size known by the userspace driver (see drm_panthor_obj_array::stride
+ * or drm_panthor_dev_query::size).
+ * - If the kernel driver is too old to know some fields, those will be
+ * ignored if zero, and otherwise rejected (and so will be zero on output).
+ * - If userspace is too old to know some fields, those will be zeroed
+ * (input) before the structure is parsed by the kernel driver.
+ * - Each new flag/field addition must come with a driver version update so
+ * the userspace driver doesn't have to trial and error to know which
+ * flags are supported.
+ * - Structures should not contain unions, as this would defeat the
+ * extensibility of such structures.
+ * - IOCTLs can't be removed or replaced. New IOCTL IDs should be placed
+ * at the end of the drm_panthor_ioctl_id enum.
+ */
+
+/**
+ * DOC: MMIO regions exposed to userspace.
+ *
+ * .. c:macro:: DRM_PANTHOR_USER_MMIO_OFFSET
+ *
+ * File offset for all MMIO regions being exposed to userspace. Don't use
+ * this value directly, use DRM_PANTHOR_USER_<name>_OFFSET values instead.
+ * pgoffset passed to mmap2() is an unsigned long, which forces us to use a
+ * different offset on 32-bit and 64-bit systems.
+ *
+ * .. c:macro:: DRM_PANTHOR_USER_FLUSH_ID_MMIO_OFFSET
+ *
+ * File offset for the LATEST_FLUSH_ID register. The Userspace driver controls
+ * GPU cache flushing through CS instructions, but the flush reduction
+ * mechanism requires a flush_id. This flush_id could be queried with an
+ * ioctl, but Arm provides a well-isolated register page containing only this
+ * read-only register, so let's expose this page through a static mmap offset
+ * and allow direct mapping of this MMIO region so we can avoid the
+ * user <-> kernel round-trip.
+ */
+#define DRM_PANTHOR_USER_MMIO_OFFSET_32BIT (1ull << 43)
+#define DRM_PANTHOR_USER_MMIO_OFFSET_64BIT (1ull << 56)
+#define DRM_PANTHOR_USER_MMIO_OFFSET (sizeof(unsigned long) < 8 ? \
+ DRM_PANTHOR_USER_MMIO_OFFSET_32BIT : \
+ DRM_PANTHOR_USER_MMIO_OFFSET_64BIT)
+#define DRM_PANTHOR_USER_FLUSH_ID_MMIO_OFFSET (DRM_PANTHOR_USER_MMIO_OFFSET | 0)
+
+/**
+ * DOC: IOCTL IDs
+ *
+ * enum drm_panthor_ioctl_id - IOCTL IDs
+ *
+ * Place new ioctls at the end, don't re-order, don't replace or remove entries.
+ *
+ * These IDs are not meant to be used directly. Use the DRM_IOCTL_PANTHOR_xxx
+ * definitions instead.
+ */
+enum drm_panthor_ioctl_id {
+ /** @DRM_PANTHOR_DEV_QUERY: Query device information. */
+ DRM_PANTHOR_DEV_QUERY = 0,
+
+ /** @DRM_PANTHOR_VM_CREATE: Create a VM. */
+ DRM_PANTHOR_VM_CREATE,
+
+ /** @DRM_PANTHOR_VM_DESTROY: Destroy a VM. */
+ DRM_PANTHOR_VM_DESTROY,
+
+ /** @DRM_PANTHOR_VM_BIND: Bind/unbind memory to a VM. */
+ DRM_PANTHOR_VM_BIND,
+
+ /** @DRM_PANTHOR_VM_GET_STATE: Get VM state. */
+ DRM_PANTHOR_VM_GET_STATE,
+
+ /** @DRM_PANTHOR_BO_CREATE: Create a buffer object. */
+ DRM_PANTHOR_BO_CREATE,
+
+ /**
+ * @DRM_PANTHOR_BO_MMAP_OFFSET: Get the file offset to pass to
+ * mmap to map a GEM object.
+ */
+ DRM_PANTHOR_BO_MMAP_OFFSET,
+
+ /** @DRM_PANTHOR_GROUP_CREATE: Create a scheduling group. */
+ DRM_PANTHOR_GROUP_CREATE,
+
+ /** @DRM_PANTHOR_GROUP_DESTROY: Destroy a scheduling group. */
+ DRM_PANTHOR_GROUP_DESTROY,
+
+ /**
+ * @DRM_PANTHOR_GROUP_SUBMIT: Submit jobs to queues belonging
+ * to a specific scheduling group.
+ */
+ DRM_PANTHOR_GROUP_SUBMIT,
+
+ /** @DRM_PANTHOR_GROUP_GET_STATE: Get the state of a scheduling group. */
+ DRM_PANTHOR_GROUP_GET_STATE,
+
+ /** @DRM_PANTHOR_TILER_HEAP_CREATE: Create a tiler heap. */
+ DRM_PANTHOR_TILER_HEAP_CREATE,
+
+ /** @DRM_PANTHOR_TILER_HEAP_DESTROY: Destroy a tiler heap. */
+ DRM_PANTHOR_TILER_HEAP_DESTROY,
+};
+
+/**
+ * DRM_IOCTL_PANTHOR() - Build a Panthor IOCTL number
+ * @__access: Access type. Must be R, W or RW.
+ * @__id: One of the DRM_PANTHOR_xxx id.
+ * @__type: Suffix of the type being passed to the IOCTL.
+ *
+ * Don't use this macro directly, use the DRM_IOCTL_PANTHOR_xxx
+ * values instead.
+ *
+ * Return: An IOCTL number to be passed to ioctl() from userspace.
+ */
+#define DRM_IOCTL_PANTHOR(__access, __id, __type) \
+ DRM_IO ## __access(DRM_COMMAND_BASE + DRM_PANTHOR_ ## __id, \
+ struct drm_panthor_ ## __type)
+
+#define DRM_IOCTL_PANTHOR_DEV_QUERY \
+ DRM_IOCTL_PANTHOR(WR, DEV_QUERY, dev_query)
+#define DRM_IOCTL_PANTHOR_VM_CREATE \
+ DRM_IOCTL_PANTHOR(WR, VM_CREATE, vm_create)
+#define DRM_IOCTL_PANTHOR_VM_DESTROY \
+ DRM_IOCTL_PANTHOR(WR, VM_DESTROY, vm_destroy)
+#define DRM_IOCTL_PANTHOR_VM_BIND \
+ DRM_IOCTL_PANTHOR(WR, VM_BIND, vm_bind)
+#define DRM_IOCTL_PANTHOR_VM_GET_STATE \
+ DRM_IOCTL_PANTHOR(WR, VM_GET_STATE, vm_get_state)
+#define DRM_IOCTL_PANTHOR_BO_CREATE \
+ DRM_IOCTL_PANTHOR(WR, BO_CREATE, bo_create)
+#define DRM_IOCTL_PANTHOR_BO_MMAP_OFFSET \
+ DRM_IOCTL_PANTHOR(WR, BO_MMAP_OFFSET, bo_mmap_offset)
+#define DRM_IOCTL_PANTHOR_GROUP_CREATE \
+ DRM_IOCTL_PANTHOR(WR, GROUP_CREATE, group_create)
+#define DRM_IOCTL_PANTHOR_GROUP_DESTROY \
+ DRM_IOCTL_PANTHOR(WR, GROUP_DESTROY, group_destroy)
+#define DRM_IOCTL_PANTHOR_GROUP_SUBMIT \
+ DRM_IOCTL_PANTHOR(WR, GROUP_SUBMIT, group_submit)
+#define DRM_IOCTL_PANTHOR_GROUP_GET_STATE \
+ DRM_IOCTL_PANTHOR(WR, GROUP_GET_STATE, group_get_state)
+#define DRM_IOCTL_PANTHOR_TILER_HEAP_CREATE \
+ DRM_IOCTL_PANTHOR(WR, TILER_HEAP_CREATE, tiler_heap_create)
+#define DRM_IOCTL_PANTHOR_TILER_HEAP_DESTROY \
+ DRM_IOCTL_PANTHOR(WR, TILER_HEAP_DESTROY, tiler_heap_destroy)
+
+/**
+ * DOC: IOCTL arguments
+ */
+
+/**
+ * struct drm_panthor_obj_array - Object array.
+ *
+ * This object is used to pass an array of objects whose size is subject to changes in
+ * future versions of the driver. In order to support this mutability, we pass a stride
+ * describing the size of the object as known by userspace.
+ *
+ * You shouldn't fill drm_panthor_obj_array fields directly. You should instead use
+ * the DRM_PANTHOR_OBJ_ARRAY() macro that takes care of initializing the stride to
+ * the object size.
+ */
+struct drm_panthor_obj_array {
+ /** @stride: Stride of object struct. Used for versioning. */
+ __u32 stride;
+
+ /** @count: Number of objects in the array. */
+ __u32 count;
+
+ /** @array: User pointer to an array of objects. */
+ __u64 array;
+};
+
+/**
+ * DRM_PANTHOR_OBJ_ARRAY() - Initialize a drm_panthor_obj_array field.
+ * @cnt: Number of elements in the array.
+ * @ptr: Pointer to the array to pass to the kernel.
+ *
+ * Macro initializing a drm_panthor_obj_array based on the object size as known
+ * by userspace.
+ */
+#define DRM_PANTHOR_OBJ_ARRAY(cnt, ptr) \
+ { .stride = sizeof((ptr)[0]), .count = (cnt), .array = (__u64)(uintptr_t)(ptr) }
+
+/**
+ * enum drm_panthor_sync_op_flags - Synchronization operation flags.
+ */
+enum drm_panthor_sync_op_flags {
+ /** @DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_MASK: Synchronization handle type mask. */
+ DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_MASK = 0xff,
+
+ /** @DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_SYNCOBJ: Synchronization object type. */
+ DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_SYNCOBJ = 0,
+
+ /**
+ * @DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_TIMELINE_SYNCOBJ: Timeline synchronization
+ * object type.
+ */
+ DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_TIMELINE_SYNCOBJ = 1,
+
+ /** @DRM_PANTHOR_SYNC_OP_WAIT: Wait operation. */
+ DRM_PANTHOR_SYNC_OP_WAIT = 0 << 31,
+
+ /** @DRM_PANTHOR_SYNC_OP_SIGNAL: Signal operation. */
+ DRM_PANTHOR_SYNC_OP_SIGNAL = (int)(1u << 31),
+};
+
+/**
+ * struct drm_panthor_sync_op - Synchronization operation.
+ */
+struct drm_panthor_sync_op {
+ /** @flags: Synchronization operation flags. Combination of DRM_PANTHOR_SYNC_OP values. */
+ __u32 flags;
+
+ /** @handle: Sync handle. */
+ __u32 handle;
+
+ /**
+ * @timeline_value: MBZ if
+ * (flags & DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_MASK) !=
+ * DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_TIMELINE_SYNCOBJ.
+ */
+ __u64 timeline_value;
+};
+
+/**
+ * enum drm_panthor_dev_query_type - Query type
+ *
+ * Place new types at the end, don't re-order, don't remove or replace.
+ */
+enum drm_panthor_dev_query_type {
+ /** @DRM_PANTHOR_DEV_QUERY_GPU_INFO: Query GPU information. */
+ DRM_PANTHOR_DEV_QUERY_GPU_INFO = 0,
+
+ /** @DRM_PANTHOR_DEV_QUERY_CSIF_INFO: Query command-stream interface information. */
+ DRM_PANTHOR_DEV_QUERY_CSIF_INFO,
+};
+
+/**
+ * struct drm_panthor_gpu_info - GPU information
+ *
+ * Structure grouping all queryable information relating to the GPU.
+ */
+struct drm_panthor_gpu_info {
+ /** @gpu_id : GPU ID. */
+ __u32 gpu_id;
+#define DRM_PANTHOR_ARCH_MAJOR(x) ((x) >> 28)
+#define DRM_PANTHOR_ARCH_MINOR(x) (((x) >> 24) & 0xf)
+#define DRM_PANTHOR_ARCH_REV(x) (((x) >> 20) & 0xf)
+#define DRM_PANTHOR_PRODUCT_MAJOR(x) (((x) >> 16) & 0xf)
+#define DRM_PANTHOR_VERSION_MAJOR(x) (((x) >> 12) & 0xf)
+#define DRM_PANTHOR_VERSION_MINOR(x) (((x) >> 4) & 0xff)
+#define DRM_PANTHOR_VERSION_STATUS(x) ((x) & 0xf)
+
+ /** @gpu_rev: GPU revision. */
+ __u32 gpu_rev;
+
+ /** @csf_id: Command stream frontend ID. */
+ __u32 csf_id;
+#define DRM_PANTHOR_CSHW_MAJOR(x) (((x) >> 26) & 0x3f)
+#define DRM_PANTHOR_CSHW_MINOR(x) (((x) >> 20) & 0x3f)
+#define DRM_PANTHOR_CSHW_REV(x) (((x) >> 16) & 0xf)
+#define DRM_PANTHOR_MCU_MAJOR(x) (((x) >> 10) & 0x3f)
+#define DRM_PANTHOR_MCU_MINOR(x) (((x) >> 4) & 0x3f)
+#define DRM_PANTHOR_MCU_REV(x) ((x) & 0xf)
+
+ /** @l2_features: L2-cache features. */
+ __u32 l2_features;
+
+ /** @tiler_features: Tiler features. */
+ __u32 tiler_features;
+
+ /** @mem_features: Memory features. */
+ __u32 mem_features;
+
+ /** @mmu_features: MMU features. */
+ __u32 mmu_features;
+#define DRM_PANTHOR_MMU_VA_BITS(x) ((x) & 0xff)
+
+ /** @thread_features: Thread features. */
+ __u32 thread_features;
+
+ /** @max_threads: Maximum number of threads. */
+ __u32 max_threads;
+
+ /** @thread_max_workgroup_size: Maximum workgroup size. */
+ __u32 thread_max_workgroup_size;
+
+ /**
+ * @thread_max_barrier_size: Maximum number of threads that can wait
+ * simultaneously on a barrier.
+ */
+ __u32 thread_max_barrier_size;
+
+ /** @coherency_features: Coherency features. */
+ __u32 coherency_features;
+
+ /** @texture_features: Texture features. */
+ __u32 texture_features[4];
+
+ /** @as_present: Bitmask encoding the number of address-space exposed by the MMU. */
+ __u32 as_present;
+
+ /** @shader_present: Bitmask encoding the shader cores exposed by the GPU. */
+ __u64 shader_present;
+
+ /** @l2_present: Bitmask encoding the L2 caches exposed by the GPU. */
+ __u64 l2_present;
+
+ /** @tiler_present: Bitmask encoding the tiler units exposed by the GPU. */
+ __u64 tiler_present;
+
+ /* @core_features: Used to discriminate core variants when they exist. */
+ __u32 core_features;
+
+ /* @pad: MBZ. */
+ __u32 pad;
+};
+
+/**
+ * struct drm_panthor_csif_info - Command stream interface information
+ *
+ * Structure grouping all queryable information relating to the command stream interface.
+ */
+struct drm_panthor_csif_info {
+ /** @csg_slot_count: Number of command stream group slots exposed by the firmware. */
+ __u32 csg_slot_count;
+
+ /** @cs_slot_count: Number of command stream slots per group. */
+ __u32 cs_slot_count;
+
+ /** @cs_reg_count: Number of command stream registers. */
+ __u32 cs_reg_count;
+
+ /** @scoreboard_slot_count: Number of scoreboard slots. */
+ __u32 scoreboard_slot_count;
+
+ /**
+ * @unpreserved_cs_reg_count: Number of command stream registers reserved by
+ * the kernel driver to call a userspace command stream.
+ *
+ * All registers can be used by a userspace command stream, but the
+ * [cs_slot_count - unpreserved_cs_reg_count .. cs_slot_count] registers are
+ * used by the kernel when DRM_PANTHOR_IOCTL_GROUP_SUBMIT is called.
+ */
+ __u32 unpreserved_cs_reg_count;
+
+ /**
+ * @pad: Padding field, set to zero.
+ */
+ __u32 pad;
+};
+
+/**
+ * struct drm_panthor_dev_query - Arguments passed to DRM_PANTHOR_IOCTL_DEV_QUERY
+ */
+struct drm_panthor_dev_query {
+ /** @type: the query type (see drm_panthor_dev_query_type). */
+ __u32 type;
+
+ /**
+ * @size: size of the type being queried.
+ *
+ * If pointer is NULL, size is updated by the driver to provide the
+ * output structure size. If pointer is not NULL, the driver will
+ * only copy min(size, actual_structure_size) bytes to the pointer,
+ * and update the size accordingly. This allows us to extend query
+ * types without breaking userspace.
+ */
+ __u32 size;
+
+ /**
+ * @pointer: user pointer to a query type struct.
+ *
+ * Pointer can be NULL, in which case, nothing is copied, but the
+ * actual structure size is returned. If not NULL, it must point to
+ * a location that's large enough to hold size bytes.
+ */
+ __u64 pointer;
+};
+
+/**
+ * struct drm_panthor_vm_create - Arguments passed to DRM_PANTHOR_IOCTL_VM_CREATE
+ */
+struct drm_panthor_vm_create {
+ /** @flags: VM flags, MBZ. */
+ __u32 flags;
+
+ /** @id: Returned VM ID. */
+ __u32 id;
+
+ /**
+ * @user_va_range: Size of the VA space reserved for user objects.
+ *
+ * The kernel will pick the remaining space to map kernel-only objects to the
+ * VM (heap chunks, heap context, ring buffers, kernel synchronization objects,
+ * ...). If the space left for kernel objects is too small, kernel object
+ * allocation will fail further down the road. One can use
+ * drm_panthor_gpu_info::mmu_features to extract the total virtual address
+ * range, and chose a user_va_range that leaves some space to the kernel.
+ *
+ * If user_va_range is zero, the kernel will pick a sensible value based on
+ * TASK_SIZE and the virtual range supported by the GPU MMU (the kernel/user
+ * split should leave enough VA space for userspace processes to support SVM,
+ * while still allowing the kernel to map some amount of kernel objects in
+ * the kernel VA range). The value chosen by the driver will be returned in
+ * @user_va_range.
+ *
+ * User VA space always starts at 0x0, kernel VA space is always placed after
+ * the user VA range.
+ */
+ __u64 user_va_range;
+};
+
+/**
+ * struct drm_panthor_vm_destroy - Arguments passed to DRM_PANTHOR_IOCTL_VM_DESTROY
+ */
+struct drm_panthor_vm_destroy {
+ /** @id: ID of the VM to destroy. */
+ __u32 id;
+
+ /** @pad: MBZ. */
+ __u32 pad;
+};
+
+/**
+ * enum drm_panthor_vm_bind_op_flags - VM bind operation flags
+ */
+enum drm_panthor_vm_bind_op_flags {
+ /**
+ * @DRM_PANTHOR_VM_BIND_OP_MAP_READONLY: Map the memory read-only.
+ *
+ * Only valid with DRM_PANTHOR_VM_BIND_OP_TYPE_MAP.
+ */
+ DRM_PANTHOR_VM_BIND_OP_MAP_READONLY = 1 << 0,
+
+ /**
+ * @DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC: Map the memory not-executable.
+ *
+ * Only valid with DRM_PANTHOR_VM_BIND_OP_TYPE_MAP.
+ */
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC = 1 << 1,
+
+ /**
+ * @DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED: Map the memory uncached.
+ *
+ * Only valid with DRM_PANTHOR_VM_BIND_OP_TYPE_MAP.
+ */
+ DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED = 1 << 2,
+
+ /**
+ * @DRM_PANTHOR_VM_BIND_OP_TYPE_MASK: Mask used to determine the type of operation.
+ */
+ DRM_PANTHOR_VM_BIND_OP_TYPE_MASK = (int)(0xfu << 28),
+
+ /** @DRM_PANTHOR_VM_BIND_OP_TYPE_MAP: Map operation. */
+ DRM_PANTHOR_VM_BIND_OP_TYPE_MAP = 0 << 28,
+
+ /** @DRM_PANTHOR_VM_BIND_OP_TYPE_UNMAP: Unmap operation. */
+ DRM_PANTHOR_VM_BIND_OP_TYPE_UNMAP = 1 << 28,
+
+ /**
+ * @DRM_PANTHOR_VM_BIND_OP_TYPE_SYNC_ONLY: No VM operation.
+ *
+ * Just serves as a synchronization point on a VM queue.
+ *
+ * Only valid if %DRM_PANTHOR_VM_BIND_ASYNC is set in drm_panthor_vm_bind::flags,
+ * and drm_panthor_vm_bind_op::syncs contains at least one element.
+ */
+ DRM_PANTHOR_VM_BIND_OP_TYPE_SYNC_ONLY = 2 << 28,
+};
+
+/**
+ * struct drm_panthor_vm_bind_op - VM bind operation
+ */
+struct drm_panthor_vm_bind_op {
+ /** @flags: Combination of drm_panthor_vm_bind_op_flags flags. */
+ __u32 flags;
+
+ /**
+ * @bo_handle: Handle of the buffer object to map.
+ * MBZ for unmap or sync-only operations.
+ */
+ __u32 bo_handle;
+
+ /**
+ * @bo_offset: Buffer object offset.
+ * MBZ for unmap or sync-only operations.
+ */
+ __u64 bo_offset;
+
+ /**
+ * @va: Virtual address to map/unmap.
+ * MBZ for sync-only operations.
+ */
+ __u64 va;
+
+ /**
+ * @size: Size to map/unmap.
+ * MBZ for sync-only operations.
+ */
+ __u64 size;
+
+ /**
+ * @syncs: Array of struct drm_panthor_sync_op synchronization
+ * operations.
+ *
+ * This array must be empty if %DRM_PANTHOR_VM_BIND_ASYNC is not set on
+ * the drm_panthor_vm_bind object containing this VM bind operation.
+ *
+ * This array shall not be empty for sync-only operations.
+ */
+ struct drm_panthor_obj_array syncs;
+
+};
+
+/**
+ * enum drm_panthor_vm_bind_flags - VM bind flags
+ */
+enum drm_panthor_vm_bind_flags {
+ /**
+ * @DRM_PANTHOR_VM_BIND_ASYNC: VM bind operations are queued to the VM
+ * queue instead of being executed synchronously.
+ */
+ DRM_PANTHOR_VM_BIND_ASYNC = 1 << 0,
+};
+
+/**
+ * struct drm_panthor_vm_bind - Arguments passed to DRM_IOCTL_PANTHOR_VM_BIND
+ */
+struct drm_panthor_vm_bind {
+ /** @vm_id: VM targeted by the bind request. */
+ __u32 vm_id;
+
+ /** @flags: Combination of drm_panthor_vm_bind_flags flags. */
+ __u32 flags;
+
+ /** @ops: Array of struct drm_panthor_vm_bind_op bind operations. */
+ struct drm_panthor_obj_array ops;
+};
+
+/**
+ * enum drm_panthor_vm_state - VM states.
+ */
+enum drm_panthor_vm_state {
+ /**
+ * @DRM_PANTHOR_VM_STATE_USABLE: VM is usable.
+ *
+ * New VM operations will be accepted on this VM.
+ */
+ DRM_PANTHOR_VM_STATE_USABLE,
+
+ /**
+ * @DRM_PANTHOR_VM_STATE_UNSABLE: VM is unsable.
+ *
+ * Something put the VM in an unusable state (like an asynchronous
+ * VM_BIND request failing for any reason).
+ *
+ * Once the VM is in this state, all new MAP operations will be
+ * rejected, and any GPU job targeting this VM will fail.
+ * UNMAP operations are still accepted.
+ *
+ * The only way to recover from an unusable VM is to create a new
+ * VM, and destroy the old one.
+ */
+ DRM_PANTHOR_VM_STATE_UNUSABLE,
+};
+
+/**
+ * struct drm_panthor_vm_get_state - Get VM state.
+ */
+struct drm_panthor_vm_get_state {
+ /** @vm_id: VM targeted by the get_state request. */
+ __u32 vm_id;
+
+ /**
+ * @state: state returned by the driver.
+ *
+ * Must be one of the enum drm_panthor_vm_state values.
+ */
+ __u32 state;
+};
+
+/**
+ * enum drm_panthor_bo_flags - Buffer object flags, passed at creation time.
+ */
+enum drm_panthor_bo_flags {
+ /** @DRM_PANTHOR_BO_NO_MMAP: The buffer object will never be CPU-mapped in userspace. */
+ DRM_PANTHOR_BO_NO_MMAP = (1 << 0),
+};
+
+/**
+ * struct drm_panthor_bo_create - Arguments passed to DRM_IOCTL_PANTHOR_BO_CREATE.
+ */
+struct drm_panthor_bo_create {
+ /**
+ * @size: Requested size for the object
+ *
+ * The (page-aligned) allocated size for the object will be returned.
+ */
+ __u64 size;
+
+ /**
+ * @flags: Flags. Must be a combination of drm_panthor_bo_flags flags.
+ */
+ __u32 flags;
+
+ /**
+ * @exclusive_vm_id: Exclusive VM this buffer object will be mapped to.
+ *
+ * If not zero, the field must refer to a valid VM ID, and implies that:
+ * - the buffer object will only ever be bound to that VM
+ * - cannot be exported as a PRIME fd
+ */
+ __u32 exclusive_vm_id;
+
+ /**
+ * @handle: Returned handle for the object.
+ *
+ * Object handles are nonzero.
+ */
+ __u32 handle;
+
+ /** @pad: MBZ. */
+ __u32 pad;
+};
+
+/**
+ * struct drm_panthor_bo_mmap_offset - Arguments passed to DRM_IOCTL_PANTHOR_BO_MMAP_OFFSET.
+ */
+struct drm_panthor_bo_mmap_offset {
+ /** @handle: Handle of the object we want an mmap offset for. */
+ __u32 handle;
+
+ /** @pad: MBZ. */
+ __u32 pad;
+
+ /** @offset: The fake offset to use for subsequent mmap calls. */
+ __u64 offset;
+};
+
+/**
+ * struct drm_panthor_queue_create - Queue creation arguments.
+ */
+struct drm_panthor_queue_create {
+ /**
+ * @priority: Defines the priority of queues inside a group. Goes from 0 to 15,
+ * 15 being the highest priority.
+ */
+ __u8 priority;
+
+ /** @pad: Padding fields, MBZ. */
+ __u8 pad[3];
+
+ /** @ringbuf_size: Size of the ring buffer to allocate to this queue. */
+ __u32 ringbuf_size;
+};
+
+/**
+ * enum drm_panthor_group_priority - Scheduling group priority
+ */
+enum drm_panthor_group_priority {
+ /** @PANTHOR_GROUP_PRIORITY_LOW: Low priority group. */
+ PANTHOR_GROUP_PRIORITY_LOW = 0,
+
+ /** @PANTHOR_GROUP_PRIORITY_MEDIUM: Medium priority group. */
+ PANTHOR_GROUP_PRIORITY_MEDIUM,
+
+ /** @PANTHOR_GROUP_PRIORITY_HIGH: High priority group. */
+ PANTHOR_GROUP_PRIORITY_HIGH,
+};
+
+/**
+ * struct drm_panthor_group_create - Arguments passed to DRM_IOCTL_PANTHOR_GROUP_CREATE
+ */
+struct drm_panthor_group_create {
+ /** @queues: Array of drm_panthor_queue_create elements. */
+ struct drm_panthor_obj_array queues;
+
+ /**
+ * @max_compute_cores: Maximum number of cores that can be used by compute
+ * jobs across CS queues bound to this group.
+ *
+ * Must be less or equal to the number of bits set in @compute_core_mask.
+ */
+ __u8 max_compute_cores;
+
+ /**
+ * @max_fragment_cores: Maximum number of cores that can be used by fragment
+ * jobs across CS queues bound to this group.
+ *
+ * Must be less or equal to the number of bits set in @fragment_core_mask.
+ */
+ __u8 max_fragment_cores;
+
+ /**
+ * @max_tiler_cores: Maximum number of tilers that can be used by tiler jobs
+ * across CS queues bound to this group.
+ *
+ * Must be less or equal to the number of bits set in @tiler_core_mask.
+ */
+ __u8 max_tiler_cores;
+
+ /** @priority: Group priority (see enum drm_panthor_group_priority). */
+ __u8 priority;
+
+ /** @pad: Padding field, MBZ. */
+ __u32 pad;
+
+ /**
+ * @compute_core_mask: Mask encoding cores that can be used for compute jobs.
+ *
+ * This field must have at least @max_compute_cores bits set.
+ *
+ * The bits set here should also be set in drm_panthor_gpu_info::shader_present.
+ */
+ __u64 compute_core_mask;
+
+ /**
+ * @fragment_core_mask: Mask encoding cores that can be used for fragment jobs.
+ *
+ * This field must have at least @max_fragment_cores bits set.
+ *
+ * The bits set here should also be set in drm_panthor_gpu_info::shader_present.
+ */
+ __u64 fragment_core_mask;
+
+ /**
+ * @tiler_core_mask: Mask encoding cores that can be used for tiler jobs.
+ *
+ * This field must have at least @max_tiler_cores bits set.
+ *
+ * The bits set here should also be set in drm_panthor_gpu_info::tiler_present.
+ */
+ __u64 tiler_core_mask;
+
+ /**
+ * @vm_id: VM ID to bind this group to.
+ *
+ * All submission to queues bound to this group will use this VM.
+ */
+ __u32 vm_id;
+
+ /**
+ * @group_handle: Returned group handle. Passed back when submitting jobs or
+ * destroying a group.
+ */
+ __u32 group_handle;
+};
+
+/**
+ * struct drm_panthor_group_destroy - Arguments passed to DRM_IOCTL_PANTHOR_GROUP_DESTROY
+ */
+struct drm_panthor_group_destroy {
+ /** @group_handle: Group to destroy */
+ __u32 group_handle;
+
+ /** @pad: Padding field, MBZ. */
+ __u32 pad;
+};
+
+/**
+ * struct drm_panthor_queue_submit - Job submission arguments.
+ *
+ * This is describing the userspace command stream to call from the kernel
+ * command stream ring-buffer. Queue submission is always part of a group
+ * submission, taking one or more jobs to submit to the underlying queues.
+ */
+struct drm_panthor_queue_submit {
+ /** @queue_index: Index of the queue inside a group. */
+ __u32 queue_index;
+
+ /**
+ * @stream_size: Size of the command stream to execute.
+ *
+ * Must be 64-bit/8-byte aligned (the size of a CS instruction)
+ *
+ * Can be zero if stream_addr is zero too.
+ */
+ __u32 stream_size;
+
+ /**
+ * @stream_addr: GPU address of the command stream to execute.
+ *
+ * Must be aligned on 64-byte.
+ *
+ * Can be zero is stream_size is zero too.
+ */
+ __u64 stream_addr;
+
+ /**
+ * @latest_flush: FLUSH_ID read at the time the stream was built.
+ *
+ * This allows cache flush elimination for the automatic
+ * flush+invalidate(all) done at submission time, which is needed to
+ * ensure the GPU doesn't get garbage when reading the indirect command
+ * stream buffers. If you want the cache flush to happen
+ * unconditionally, pass a zero here.
+ */
+ __u32 latest_flush;
+
+ /** @pad: MBZ. */
+ __u32 pad;
+
+ /** @syncs: Array of struct drm_panthor_sync_op sync operations. */
+ struct drm_panthor_obj_array syncs;
+};
+
+/**
+ * struct drm_panthor_group_submit - Arguments passed to DRM_IOCTL_PANTHOR_GROUP_SUBMIT
+ */
+struct drm_panthor_group_submit {
+ /** @group_handle: Handle of the group to queue jobs to. */
+ __u32 group_handle;
+
+ /** @pad: MBZ. */
+ __u32 pad;
+
+ /** @queue_submits: Array of drm_panthor_queue_submit objects. */
+ struct drm_panthor_obj_array queue_submits;
+};
+
+/**
+ * enum drm_panthor_group_state_flags - Group state flags
+ */
+enum drm_panthor_group_state_flags {
+ /**
+ * @DRM_PANTHOR_GROUP_STATE_TIMEDOUT: Group had unfinished jobs.
+ *
+ * When a group ends up with this flag set, no jobs can be submitted to its queues.
+ */
+ DRM_PANTHOR_GROUP_STATE_TIMEDOUT = 1 << 0,
+
+ /**
+ * @DRM_PANTHOR_GROUP_STATE_FATAL_FAULT: Group had fatal faults.
+ *
+ * When a group ends up with this flag set, no jobs can be submitted to its queues.
+ */
+ DRM_PANTHOR_GROUP_STATE_FATAL_FAULT = 1 << 1,
+};
+
+/**
+ * struct drm_panthor_group_get_state - Arguments passed to DRM_IOCTL_PANTHOR_GROUP_GET_STATE
+ *
+ * Used to query the state of a group and decide whether a new group should be created to
+ * replace it.
+ */
+struct drm_panthor_group_get_state {
+ /** @group_handle: Handle of the group to query state on */
+ __u32 group_handle;
+
+ /**
+ * @state: Combination of DRM_PANTHOR_GROUP_STATE_* flags encoding the
+ * group state.
+ */
+ __u32 state;
+
+ /** @fatal_queues: Bitmask of queues that faced fatal faults. */
+ __u32 fatal_queues;
+
+ /** @pad: MBZ */
+ __u32 pad;
+};
+
+/**
+ * struct drm_panthor_tiler_heap_create - Arguments passed to DRM_IOCTL_PANTHOR_TILER_HEAP_CREATE
+ */
+struct drm_panthor_tiler_heap_create {
+ /** @vm_id: VM ID the tiler heap should be mapped to */
+ __u32 vm_id;
+
+ /** @initial_chunk_count: Initial number of chunks to allocate. */
+ __u32 initial_chunk_count;
+
+ /** @chunk_size: Chunk size. Must be a power of two at least 256KB large. */
+ __u32 chunk_size;
+
+ /** @max_chunks: Maximum number of chunks that can be allocated. */
+ __u32 max_chunks;
+
+ /**
+ * @target_in_flight: Maximum number of in-flight render passes.
+ *
+ * If the heap has more than tiler jobs in-flight, the FW will wait for render
+ * passes to finish before queuing new tiler jobs.
+ */
+ __u32 target_in_flight;
+
+ /** @handle: Returned heap handle. Passed back to DESTROY_TILER_HEAP. */
+ __u32 handle;
+
+ /** @tiler_heap_ctx_gpu_va: Returned heap GPU virtual address returned */
+ __u64 tiler_heap_ctx_gpu_va;
+
+ /**
+ * @first_heap_chunk_gpu_va: First heap chunk.
+ *
+ * The tiler heap is formed of heap chunks forming a single-link list. This
+ * is the first element in the list.
+ */
+ __u64 first_heap_chunk_gpu_va;
+};
+
+/**
+ * struct drm_panthor_tiler_heap_destroy - Arguments passed to DRM_IOCTL_PANTHOR_TILER_HEAP_DESTROY
+ */
+struct drm_panthor_tiler_heap_destroy {
+ /** @handle: Handle of the tiler heap to destroy */
+ __u32 handle;
+
+ /** @pad: Padding field, MBZ. */
+ __u32 pad;
+};
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif /* _PANTHOR_DRM_H_ */
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 01/14] drm/panthor: Add uAPI
2024-01-22 16:30 ` [PATCH v4 01/14] drm/panthor: Add uAPI Boris Brezillon
@ 2024-02-08 14:29 ` Liviu Dudau
0 siblings, 0 replies; 51+ messages in thread
From: Liviu Dudau @ 2024-02-08 14:29 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen
Hi Boris,
Minor spelling issues to raise in the documentation text uAPI:
On Mon, Jan 22, 2024 at 05:30:32PM +0100, Boris Brezillon wrote:
> Panthor follows the lead of other recently submitted drivers with
> ioctls allowing us to support modern Vulkan features, like sparse memory
> binding:
>
> - Pretty standard GEM management ioctls (BO_CREATE and BO_MMAP_OFFSET),
> with the 'exclusive-VM' bit to speed-up BO reservation on job submission
> - VM management ioctls (VM_CREATE, VM_DESTROY and VM_BIND). The VM_BIND
> ioctl is loosely based on the Xe model, and can handle both
> asynchronous and synchronous requests
> - GPU execution context creation/destruction, tiler heap context creation
> and job submission. Those ioctls reflect how the hardware/scheduler
> works and are thus driver specific.
>
> We also have a way to expose IO regions, such that the usermode driver
> can directly access specific/well-isolate registers, like the
> LATEST_FLUSH register used to implement cache-flush reduction.
>
> This uAPI intentionally keeps usermode queues out of the scope, which
> explains why doorbell registers and command stream ring-buffers are not
> directly exposed to userspace.
>
> v4:
> - Add a VM_GET_STATE ioctl
> - Fix doc
> - Expose the CORE_FEATURES register so we can deal with variants in the
> UMD
> - Add Steve's R-b
>
[snip]
> +
> +/**
> + * enum drm_panthor_vm_state - VM states.
> + */
> +enum drm_panthor_vm_state {
> + /**
> + * @DRM_PANTHOR_VM_STATE_USABLE: VM is usable.
> + *
> + * New VM operations will be accepted on this VM.
> + */
> + DRM_PANTHOR_VM_STATE_USABLE,
> +
> + /**
> + * @DRM_PANTHOR_VM_STATE_UNSABLE: VM is unsable.
s/UNSABLE/UNUSABLE/ and s/unsable/unusable/
> + *
> + * Something put the VM in an unusable state (like an asynchronous
> + * VM_BIND request failing for any reason).
> + *
> + * Once the VM is in this state, all new MAP operations will be
> + * rejected, and any GPU job targeting this VM will fail.
> + * UNMAP operations are still accepted.
> + *
> + * The only way to recover from an unusable VM is to create a new
> + * VM, and destroy the old one.
> + */
> + DRM_PANTHOR_VM_STATE_UNUSABLE,
> +};
> +
[snip]
The rest of the patch looks good to me!
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
Best regards,
Liviu
--
====================
| I would like to |
| fix the world, |
| but they're not |
| giving me the |
\ source code! /
---------------
¯\_(ツ)_/¯
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 02/14] drm/panthor: Add GPU register definitions
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
2024-01-22 16:30 ` [PATCH v4 01/14] drm/panthor: Add uAPI Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-01-22 16:30 ` [PATCH v4 03/14] drm/panthor: Add the device logical block Boris Brezillon
` (15 subsequent siblings)
17 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
Those are the registers directly accessible through the MMIO range.
FW registers are exposed in panthor_fw.h.
v4:
- Add the CORE_FEATURES register (needed for GPU variants)
- Add Steve's R-b
v3:
- Add macros to extract GPU ID info
- Formatting changes
- Remove AS_TRANSCFG_ADRMODE_LEGACY - it doesn't exist post-CSF
- Remove CSF_GPU_LATEST_FLUSH_ID_DEFAULT
- Add GPU_L2_FEATURES_LINE_SIZE for extracting the GPU cache line size
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
Reviewed-by: Steven Price <steven.price@arm.com>
---
drivers/gpu/drm/panthor/panthor_regs.h | 239 +++++++++++++++++++++++++
1 file changed, 239 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_regs.h
diff --git a/drivers/gpu/drm/panthor/panthor_regs.h b/drivers/gpu/drm/panthor/panthor_regs.h
new file mode 100644
index 000000000000..b7b3b3add166
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_regs.h
@@ -0,0 +1,239 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
+/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
+/* Copyright 2023 Collabora ltd. */
+/*
+ * Register definitions based on mali_kbase_gpu_regmap.h and
+ * mali_kbase_gpu_regmap_csf.h
+ * (C) COPYRIGHT 2010-2022 ARM Limited. All rights reserved.
+ */
+#ifndef __PANTHOR_REGS_H__
+#define __PANTHOR_REGS_H__
+
+#define GPU_ID 0x0
+#define GPU_ARCH_MAJOR(x) ((x) >> 28)
+#define GPU_ARCH_MINOR(x) (((x) & GENMASK(27, 24)) >> 24)
+#define GPU_ARCH_REV(x) (((x) & GENMASK(23, 20)) >> 20)
+#define GPU_PROD_MAJOR(x) (((x) & GENMASK(19, 16)) >> 16)
+#define GPU_VER_MAJOR(x) (((x) & GENMASK(15, 12)) >> 12)
+#define GPU_VER_MINOR(x) (((x) & GENMASK(11, 4)) >> 4)
+#define GPU_VER_STATUS(x) ((x) & GENMASK(3, 0))
+
+#define GPU_L2_FEATURES 0x4
+#define GPU_L2_FEATURES_LINE_SIZE(x) (1 << ((x) & GENMASK(7, 0)))
+
+#define GPU_CORE_FEATURES 0x8
+
+#define GPU_TILER_FEATURES 0xC
+#define GPU_MEM_FEATURES 0x10
+#define GROUPS_L2_COHERENT BIT(0)
+
+#define GPU_MMU_FEATURES 0x14
+#define GPU_MMU_FEATURES_VA_BITS(x) ((x) & GENMASK(7, 0))
+#define GPU_MMU_FEATURES_PA_BITS(x) (((x) >> 8) & GENMASK(7, 0))
+#define GPU_AS_PRESENT 0x18
+#define GPU_CSF_ID 0x1C
+
+#define GPU_INT_RAWSTAT 0x20
+#define GPU_INT_CLEAR 0x24
+#define GPU_INT_MASK 0x28
+#define GPU_INT_STAT 0x2c
+#define GPU_IRQ_FAULT BIT(0)
+#define GPU_IRQ_PROTM_FAULT BIT(1)
+#define GPU_IRQ_RESET_COMPLETED BIT(8)
+#define GPU_IRQ_POWER_CHANGED BIT(9)
+#define GPU_IRQ_POWER_CHANGED_ALL BIT(10)
+#define GPU_IRQ_CLEAN_CACHES_COMPLETED BIT(17)
+#define GPU_IRQ_DOORBELL_MIRROR BIT(18)
+#define GPU_IRQ_MCU_STATUS_CHANGED BIT(19)
+#define GPU_CMD 0x30
+#define GPU_CMD_DEF(type, payload) ((type) | ((payload) << 8))
+#define GPU_SOFT_RESET GPU_CMD_DEF(1, 1)
+#define GPU_HARD_RESET GPU_CMD_DEF(1, 2)
+#define CACHE_CLEAN BIT(0)
+#define CACHE_INV BIT(1)
+#define GPU_FLUSH_CACHES(l2, lsc, oth) \
+ GPU_CMD_DEF(4, ((l2) << 0) | ((lsc) << 4) | ((oth) << 8))
+
+#define GPU_STATUS 0x34
+#define GPU_STATUS_ACTIVE BIT(0)
+#define GPU_STATUS_PWR_ACTIVE BIT(1)
+#define GPU_STATUS_PAGE_FAULT BIT(4)
+#define GPU_STATUS_PROTM_ACTIVE BIT(7)
+#define GPU_STATUS_DBG_ENABLED BIT(8)
+
+#define GPU_FAULT_STATUS 0x3C
+#define GPU_FAULT_ADDR_LO 0x40
+#define GPU_FAULT_ADDR_HI 0x44
+
+#define GPU_PWR_KEY 0x50
+#define GPU_PWR_KEY_UNLOCK 0x2968A819
+#define GPU_PWR_OVERRIDE0 0x54
+#define GPU_PWR_OVERRIDE1 0x58
+
+#define GPU_TIMESTAMP_OFFSET_LO 0x88
+#define GPU_TIMESTAMP_OFFSET_HI 0x8C
+#define GPU_CYCLE_COUNT_LO 0x90
+#define GPU_CYCLE_COUNT_HI 0x94
+#define GPU_TIMESTAMP_LO 0x98
+#define GPU_TIMESTAMP_HI 0x9C
+
+#define GPU_THREAD_MAX_THREADS 0xA0
+#define GPU_THREAD_MAX_WORKGROUP_SIZE 0xA4
+#define GPU_THREAD_MAX_BARRIER_SIZE 0xA8
+#define GPU_THREAD_FEATURES 0xAC
+
+#define GPU_TEXTURE_FEATURES(n) (0xB0 + ((n) * 4))
+
+#define GPU_SHADER_PRESENT_LO 0x100
+#define GPU_SHADER_PRESENT_HI 0x104
+#define GPU_TILER_PRESENT_LO 0x110
+#define GPU_TILER_PRESENT_HI 0x114
+#define GPU_L2_PRESENT_LO 0x120
+#define GPU_L2_PRESENT_HI 0x124
+
+#define SHADER_READY_LO 0x140
+#define SHADER_READY_HI 0x144
+#define TILER_READY_LO 0x150
+#define TILER_READY_HI 0x154
+#define L2_READY_LO 0x160
+#define L2_READY_HI 0x164
+
+#define SHADER_PWRON_LO 0x180
+#define SHADER_PWRON_HI 0x184
+#define TILER_PWRON_LO 0x190
+#define TILER_PWRON_HI 0x194
+#define L2_PWRON_LO 0x1A0
+#define L2_PWRON_HI 0x1A4
+
+#define SHADER_PWROFF_LO 0x1C0
+#define SHADER_PWROFF_HI 0x1C4
+#define TILER_PWROFF_LO 0x1D0
+#define TILER_PWROFF_HI 0x1D4
+#define L2_PWROFF_LO 0x1E0
+#define L2_PWROFF_HI 0x1E4
+
+#define SHADER_PWRTRANS_LO 0x200
+#define SHADER_PWRTRANS_HI 0x204
+#define TILER_PWRTRANS_LO 0x210
+#define TILER_PWRTRANS_HI 0x214
+#define L2_PWRTRANS_LO 0x220
+#define L2_PWRTRANS_HI 0x224
+
+#define SHADER_PWRACTIVE_LO 0x240
+#define SHADER_PWRACTIVE_HI 0x244
+#define TILER_PWRACTIVE_LO 0x250
+#define TILER_PWRACTIVE_HI 0x254
+#define L2_PWRACTIVE_LO 0x260
+#define L2_PWRACTIVE_HI 0x264
+
+#define GPU_REVID 0x280
+
+#define GPU_COHERENCY_FEATURES 0x300
+#define GPU_COHERENCY_PROT_BIT(name) BIT(GPU_COHERENCY_ ## name)
+
+#define GPU_COHERENCY_PROTOCOL 0x304
+#define GPU_COHERENCY_ACE 0
+#define GPU_COHERENCY_ACE_LITE 1
+#define GPU_COHERENCY_NONE 31
+
+#define MCU_CONTROL 0x700
+#define MCU_CONTROL_ENABLE 1
+#define MCU_CONTROL_AUTO 2
+#define MCU_CONTROL_DISABLE 0
+
+#define MCU_STATUS 0x704
+#define MCU_STATUS_DISABLED 0
+#define MCU_STATUS_ENABLED 1
+#define MCU_STATUS_HALT 2
+#define MCU_STATUS_FATAL 3
+
+/* Job Control regs */
+#define JOB_INT_RAWSTAT 0x1000
+#define JOB_INT_CLEAR 0x1004
+#define JOB_INT_MASK 0x1008
+#define JOB_INT_STAT 0x100c
+#define JOB_INT_GLOBAL_IF BIT(31)
+#define JOB_INT_CSG_IF(x) BIT(x)
+
+/* MMU regs */
+#define MMU_INT_RAWSTAT 0x2000
+#define MMU_INT_CLEAR 0x2004
+#define MMU_INT_MASK 0x2008
+#define MMU_INT_STAT 0x200c
+
+/* AS_COMMAND register commands */
+
+#define MMU_BASE 0x2400
+#define MMU_AS_SHIFT 6
+#define MMU_AS(as) (MMU_BASE + ((as) << MMU_AS_SHIFT))
+
+#define AS_TRANSTAB_LO(as) (MMU_AS(as) + 0x0)
+#define AS_TRANSTAB_HI(as) (MMU_AS(as) + 0x4)
+#define AS_MEMATTR_LO(as) (MMU_AS(as) + 0x8)
+#define AS_MEMATTR_HI(as) (MMU_AS(as) + 0xC)
+#define AS_MEMATTR_AARCH64_INNER_ALLOC_IMPL (2 << 2)
+#define AS_MEMATTR_AARCH64_INNER_ALLOC_EXPL(w, r) ((3 << 2) | \
+ ((w) ? BIT(0) : 0) | \
+ ((r) ? BIT(1) : 0))
+#define AS_MEMATTR_AARCH64_SH_MIDGARD_INNER (0 << 4)
+#define AS_MEMATTR_AARCH64_SH_CPU_INNER (1 << 4)
+#define AS_MEMATTR_AARCH64_SH_CPU_INNER_SHADER_COH (2 << 4)
+#define AS_MEMATTR_AARCH64_SHARED (0 << 6)
+#define AS_MEMATTR_AARCH64_INNER_OUTER_NC (1 << 6)
+#define AS_MEMATTR_AARCH64_INNER_OUTER_WB (2 << 6)
+#define AS_MEMATTR_AARCH64_FAULT (3 << 6)
+#define AS_LOCKADDR_LO(as) (MMU_AS(as) + 0x10)
+#define AS_LOCKADDR_HI(as) (MMU_AS(as) + 0x14)
+#define AS_COMMAND(as) (MMU_AS(as) + 0x18)
+#define AS_COMMAND_NOP 0
+#define AS_COMMAND_UPDATE 1
+#define AS_COMMAND_LOCK 2
+#define AS_COMMAND_UNLOCK 3
+#define AS_COMMAND_FLUSH_PT 4
+#define AS_COMMAND_FLUSH_MEM 5
+#define AS_LOCK_REGION_MIN_SIZE (1ULL << 15)
+#define AS_FAULTSTATUS(as) (MMU_AS(as) + 0x1C)
+#define AS_FAULTSTATUS_ACCESS_TYPE_MASK (0x3 << 8)
+#define AS_FAULTSTATUS_ACCESS_TYPE_ATOMIC (0x0 << 8)
+#define AS_FAULTSTATUS_ACCESS_TYPE_EX (0x1 << 8)
+#define AS_FAULTSTATUS_ACCESS_TYPE_READ (0x2 << 8)
+#define AS_FAULTSTATUS_ACCESS_TYPE_WRITE (0x3 << 8)
+#define AS_FAULTADDRESS_LO(as) (MMU_AS(as) + 0x20)
+#define AS_FAULTADDRESS_HI(as) (MMU_AS(as) + 0x24)
+#define AS_STATUS(as) (MMU_AS(as) + 0x28)
+#define AS_STATUS_AS_ACTIVE BIT(0)
+#define AS_TRANSCFG_LO(as) (MMU_AS(as) + 0x30)
+#define AS_TRANSCFG_HI(as) (MMU_AS(as) + 0x34)
+#define AS_TRANSCFG_ADRMODE_UNMAPPED (1 << 0)
+#define AS_TRANSCFG_ADRMODE_IDENTITY (2 << 0)
+#define AS_TRANSCFG_ADRMODE_AARCH64_4K (6 << 0)
+#define AS_TRANSCFG_ADRMODE_AARCH64_64K (8 << 0)
+#define AS_TRANSCFG_INA_BITS(x) ((x) << 6)
+#define AS_TRANSCFG_OUTA_BITS(x) ((x) << 14)
+#define AS_TRANSCFG_SL_CONCAT BIT(22)
+#define AS_TRANSCFG_PTW_MEMATTR_NC (1 << 24)
+#define AS_TRANSCFG_PTW_MEMATTR_WB (2 << 24)
+#define AS_TRANSCFG_PTW_SH_NS (0 << 28)
+#define AS_TRANSCFG_PTW_SH_OS (2 << 28)
+#define AS_TRANSCFG_PTW_SH_IS (3 << 28)
+#define AS_TRANSCFG_PTW_RA BIT(30)
+#define AS_TRANSCFG_DISABLE_HIER_AP BIT(33)
+#define AS_TRANSCFG_DISABLE_AF_FAULT BIT(34)
+#define AS_TRANSCFG_WXN BIT(35)
+#define AS_TRANSCFG_XREADABLE BIT(36)
+#define AS_FAULTEXTRA_LO(as) (MMU_AS(as) + 0x38)
+#define AS_FAULTEXTRA_HI(as) (MMU_AS(as) + 0x3C)
+
+#define CSF_GPU_LATEST_FLUSH_ID 0x10000
+
+#define CSF_DOORBELL(i) (0x80000 + ((i) * 0x10000))
+#define CSF_GLB_DOORBELL_ID 0
+
+#define gpu_write(dev, reg, data) \
+ writel(data, (dev)->iomem + (reg))
+
+#define gpu_read(dev, reg) \
+ readl((dev)->iomem + (reg))
+
+#endif
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* [PATCH v4 03/14] drm/panthor: Add the device logical block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
2024-01-22 16:30 ` [PATCH v4 01/14] drm/panthor: Add uAPI Boris Brezillon
2024-01-22 16:30 ` [PATCH v4 02/14] drm/panthor: Add GPU register definitions Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-02-08 14:30 ` Liviu Dudau
2024-01-22 16:30 ` [PATCH v4 04/14] drm/panthor: Add the GPU " Boris Brezillon
` (14 subsequent siblings)
17 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
The panthor driver is designed in a modular way, where each logical
block is dealing with a specific HW-block or software feature. In order
for those blocks to communicate with each other, we need a central
panthor_device collecting all the blocks, and exposing some common
features, like interrupt handling, power management, reset, ...
This what this panthor_device logical block is about.
v4:
- Check drmm_mutex_init() return code
- Fix panthor_device_reset_work() out path
- Fix the race in the unplug logic
- Fix typos
- Unplug blocks when something fails in panthor_device_init()
- Add Steve's R-b
v3:
- Add acks for the MIT+GPL2 relicensing
- Fix 32-bit support
- Shorten the sections protected by panthor_device::pm::mmio_lock to fix
lock ordering issues.
- Rename panthor_device::pm::lock into panthor_device::pm::mmio_lock to
better reflect what this lock is protecting
- Use dev_err_probe()
- Make sure we call drm_dev_exit() when something fails half-way in
panthor_device_reset_work()
- Replace CSF_GPU_LATEST_FLUSH_ID_DEFAULT with a constant '1' and a
comment to explain. Also remove setting the dummy flush ID on suspend.
- Remove drm_WARN_ON() in panthor_exception_name()
- Check pirq->suspended in panthor_xxx_irq_raw_handler()
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
Reviewed-by: Steven Price <steven.price@arm.com>
---
drivers/gpu/drm/panthor/panthor_device.c | 544 +++++++++++++++++++++++
drivers/gpu/drm/panthor/panthor_device.h | 393 ++++++++++++++++
2 files changed, 937 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_device.c
create mode 100644 drivers/gpu/drm/panthor/panthor_device.h
diff --git a/drivers/gpu/drm/panthor/panthor_device.c b/drivers/gpu/drm/panthor/panthor_device.c
new file mode 100644
index 000000000000..dda41a1dd47e
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_device.c
@@ -0,0 +1,544 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
+/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
+/* Copyright 2023 Collabora ltd. */
+
+#include <linux/clk.h>
+#include <linux/reset.h>
+#include <linux/platform_device.h>
+#include <linux/pm_domain.h>
+#include <linux/pm_runtime.h>
+#include <linux/regulator/consumer.h>
+
+#include <drm/drm_drv.h>
+#include <drm/drm_managed.h>
+
+#include "panthor_sched.h"
+#include "panthor_device.h"
+#include "panthor_devfreq.h"
+#include "panthor_gpu.h"
+#include "panthor_fw.h"
+#include "panthor_mmu.h"
+#include "panthor_regs.h"
+
+static int panthor_clk_init(struct panthor_device *ptdev)
+{
+ ptdev->clks.core = devm_clk_get(ptdev->base.dev, NULL);
+ if (IS_ERR(ptdev->clks.core))
+ return dev_err_probe(ptdev->base.dev,
+ PTR_ERR(ptdev->clks.core),
+ "get 'core' clock failed");
+
+ ptdev->clks.stacks = devm_clk_get_optional(ptdev->base.dev, "stacks");
+ if (IS_ERR(ptdev->clks.stacks))
+ return dev_err_probe(ptdev->base.dev,
+ PTR_ERR(ptdev->clks.stacks),
+ "get 'stacks' clock failed");
+
+ ptdev->clks.coregroup = devm_clk_get_optional(ptdev->base.dev, "coregroup");
+ if (IS_ERR(ptdev->clks.coregroup))
+ return dev_err_probe(ptdev->base.dev,
+ PTR_ERR(ptdev->clks.coregroup),
+ "get 'coregroup' clock failed");
+
+ drm_info(&ptdev->base, "clock rate = %lu\n", clk_get_rate(ptdev->clks.core));
+ return 0;
+}
+
+void panthor_device_unplug(struct panthor_device *ptdev)
+{
+ /* This function can be called from two different path: the reset work
+ * and the platform device remove callback. drm_dev_unplug() doesn't
+ * deal with concurrent callers, so we have to protect drm_dev_unplug()
+ * calls with our own lock, and bail out if the device is already
+ * unplugged.
+ */
+ mutex_lock(&ptdev->unplug.lock);
+ if (drm_dev_is_unplugged(&ptdev->base)) {
+ /* Someone beat us, release the lock and wait for the unplug
+ * operation to be reported as done.
+ **/
+ mutex_unlock(&ptdev->unplug.lock);
+ wait_for_completion(&ptdev->unplug.done);
+ return;
+ }
+
+ /* Call drm_dev_unplug() so any access to HW blocks happening after
+ * that point get rejected.
+ */
+ drm_dev_unplug(&ptdev->base);
+
+ /* We do the rest of the unplug with the unplug lock released,
+ * future callers will wait on ptdev->unplug.done anyway.
+ */
+ mutex_unlock(&ptdev->unplug.lock);
+
+ drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
+
+ /* Now, try to cleanly shutdown the GPU before the device resources
+ * get reclaimed.
+ */
+ panthor_sched_unplug(ptdev);
+ panthor_fw_unplug(ptdev);
+ panthor_mmu_unplug(ptdev);
+ panthor_gpu_unplug(ptdev);
+
+ pm_runtime_dont_use_autosuspend(ptdev->base.dev);
+ pm_runtime_put_sync_suspend(ptdev->base.dev);
+
+ /* Report the unplug operation as done to unblock concurrent
+ * panthor_device_unplug() callers.
+ */
+ complete_all(&ptdev->unplug.done);
+}
+
+static void panthor_device_reset_cleanup(struct drm_device *ddev, void *data)
+{
+ struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
+
+ cancel_work_sync(&ptdev->reset.work);
+ destroy_workqueue(ptdev->reset.wq);
+}
+
+static void panthor_device_reset_work(struct work_struct *work)
+{
+ struct panthor_device *ptdev = container_of(work, struct panthor_device, reset.work);
+ int ret = 0, cookie;
+
+ if (atomic_read(&ptdev->pm.state) != PANTHOR_DEVICE_PM_STATE_ACTIVE) {
+ /*
+ * No need for a reset as the device has been (or will be)
+ * powered down
+ */
+ atomic_set(&ptdev->reset.pending, 0);
+ return;
+ }
+
+ if (!drm_dev_enter(&ptdev->base, &cookie))
+ return;
+
+ panthor_sched_pre_reset(ptdev);
+ panthor_fw_pre_reset(ptdev, true);
+ panthor_mmu_pre_reset(ptdev);
+ panthor_gpu_soft_reset(ptdev);
+ panthor_gpu_l2_power_on(ptdev);
+ panthor_mmu_post_reset(ptdev);
+ ret = panthor_fw_post_reset(ptdev);
+ if (ret)
+ goto out_dev_exit;
+
+ atomic_set(&ptdev->reset.pending, 0);
+ panthor_sched_post_reset(ptdev);
+
+out_dev_exit:
+ drm_dev_exit(cookie);
+
+ if (ret) {
+ panthor_device_unplug(ptdev);
+ drm_err(&ptdev->base, "Failed to boot MCU after reset, making device unusable.");
+ }
+}
+
+static bool panthor_device_is_initialized(struct panthor_device *ptdev)
+{
+ return !!ptdev->scheduler;
+}
+
+static void panthor_device_free_page(struct drm_device *ddev, void *data)
+{
+ free_page((unsigned long)data);
+}
+
+int panthor_device_init(struct panthor_device *ptdev)
+{
+ struct resource *res;
+ struct page *p;
+ int ret;
+
+ ptdev->coherent = device_get_dma_attr(ptdev->base.dev) == DEV_DMA_COHERENT;
+
+ init_completion(&ptdev->unplug.done);
+ ret = drmm_mutex_init(&ptdev->base, &ptdev->unplug.lock);
+ if (ret)
+ return ret;
+
+ ret = drmm_mutex_init(&ptdev->base, &ptdev->pm.mmio_lock);
+ if (ret)
+ return ret;
+
+ atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
+ p = alloc_page(GFP_KERNEL | __GFP_ZERO);
+ if (!p)
+ return -ENOMEM;
+
+ ptdev->pm.dummy_latest_flush = page_address(p);
+ ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_free_page,
+ ptdev->pm.dummy_latest_flush);
+ if (ret)
+ return ret;
+
+ /*
+ * Set the dummy page holding the latest flush to 1. This will cause the
+ * flush to avoided as we know it isn't necessary if the submission
+ * happens while the dummy page is mapped. Zero cannot be used because
+ * that means 'always flush'.
+ */
+ *ptdev->pm.dummy_latest_flush = 1;
+
+ INIT_WORK(&ptdev->reset.work, panthor_device_reset_work);
+ ptdev->reset.wq = alloc_ordered_workqueue("panthor-reset-wq", 0);
+ if (!ptdev->reset.wq)
+ return -ENOMEM;
+
+ ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_reset_cleanup, NULL);
+ if (ret)
+ return ret;
+
+ ret = panthor_clk_init(ptdev);
+ if (ret)
+ return ret;
+
+ ret = panthor_devfreq_init(ptdev);
+ if (ret)
+ return ret;
+
+ ptdev->iomem = devm_platform_get_and_ioremap_resource(to_platform_device(ptdev->base.dev),
+ 0, &res);
+ if (IS_ERR(ptdev->iomem))
+ return PTR_ERR(ptdev->iomem);
+
+ ptdev->phys_addr = res->start;
+
+ ret = devm_pm_runtime_enable(ptdev->base.dev);
+ if (ret)
+ return ret;
+
+ ret = pm_runtime_resume_and_get(ptdev->base.dev);
+ if (ret)
+ return ret;
+
+ ret = panthor_gpu_init(ptdev);
+ if (ret)
+ goto err_rpm_put;
+
+ ret = panthor_mmu_init(ptdev);
+ if (ret)
+ goto err_unplug_gpu;
+
+ ret = panthor_fw_init(ptdev);
+ if (ret)
+ goto err_unplug_mmu;
+
+ ret = panthor_sched_init(ptdev);
+ if (ret)
+ goto err_unplug_fw;
+
+ /* ~3 frames */
+ pm_runtime_set_autosuspend_delay(ptdev->base.dev, 50);
+
+ ret = drm_dev_register(&ptdev->base, 0);
+ if (ret)
+ goto err_unplug_sched;
+
+ pm_runtime_use_autosuspend(ptdev->base.dev);
+ pm_runtime_put_autosuspend(ptdev->base.dev);
+ return 0;
+
+err_unplug_sched:
+ panthor_sched_unplug(ptdev);
+
+err_unplug_fw:
+ panthor_fw_unplug(ptdev);
+
+err_unplug_mmu:
+ panthor_mmu_unplug(ptdev);
+
+err_unplug_gpu:
+ panthor_gpu_unplug(ptdev);
+
+err_rpm_put:
+ pm_runtime_put_sync_suspend(ptdev->base.dev);
+ return ret;
+}
+
+#define PANTHOR_EXCEPTION(id) \
+ [DRM_PANTHOR_EXCEPTION_ ## id] = { \
+ .name = #id, \
+ }
+
+struct panthor_exception_info {
+ const char *name;
+};
+
+static const struct panthor_exception_info panthor_exception_infos[] = {
+ PANTHOR_EXCEPTION(OK),
+ PANTHOR_EXCEPTION(TERMINATED),
+ PANTHOR_EXCEPTION(KABOOM),
+ PANTHOR_EXCEPTION(EUREKA),
+ PANTHOR_EXCEPTION(ACTIVE),
+ PANTHOR_EXCEPTION(CS_RES_TERM),
+ PANTHOR_EXCEPTION(CS_CONFIG_FAULT),
+ PANTHOR_EXCEPTION(CS_ENDPOINT_FAULT),
+ PANTHOR_EXCEPTION(CS_BUS_FAULT),
+ PANTHOR_EXCEPTION(CS_INSTR_INVALID),
+ PANTHOR_EXCEPTION(CS_CALL_STACK_OVERFLOW),
+ PANTHOR_EXCEPTION(CS_INHERIT_FAULT),
+ PANTHOR_EXCEPTION(INSTR_INVALID_PC),
+ PANTHOR_EXCEPTION(INSTR_INVALID_ENC),
+ PANTHOR_EXCEPTION(INSTR_BARRIER_FAULT),
+ PANTHOR_EXCEPTION(DATA_INVALID_FAULT),
+ PANTHOR_EXCEPTION(TILE_RANGE_FAULT),
+ PANTHOR_EXCEPTION(ADDR_RANGE_FAULT),
+ PANTHOR_EXCEPTION(IMPRECISE_FAULT),
+ PANTHOR_EXCEPTION(OOM),
+ PANTHOR_EXCEPTION(CSF_FW_INTERNAL_ERROR),
+ PANTHOR_EXCEPTION(CSF_RES_EVICTION_TIMEOUT),
+ PANTHOR_EXCEPTION(GPU_BUS_FAULT),
+ PANTHOR_EXCEPTION(GPU_SHAREABILITY_FAULT),
+ PANTHOR_EXCEPTION(SYS_SHAREABILITY_FAULT),
+ PANTHOR_EXCEPTION(GPU_CACHEABILITY_FAULT),
+ PANTHOR_EXCEPTION(TRANSLATION_FAULT_0),
+ PANTHOR_EXCEPTION(TRANSLATION_FAULT_1),
+ PANTHOR_EXCEPTION(TRANSLATION_FAULT_2),
+ PANTHOR_EXCEPTION(TRANSLATION_FAULT_3),
+ PANTHOR_EXCEPTION(TRANSLATION_FAULT_4),
+ PANTHOR_EXCEPTION(PERM_FAULT_0),
+ PANTHOR_EXCEPTION(PERM_FAULT_1),
+ PANTHOR_EXCEPTION(PERM_FAULT_2),
+ PANTHOR_EXCEPTION(PERM_FAULT_3),
+ PANTHOR_EXCEPTION(ACCESS_FLAG_1),
+ PANTHOR_EXCEPTION(ACCESS_FLAG_2),
+ PANTHOR_EXCEPTION(ACCESS_FLAG_3),
+ PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_IN),
+ PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_OUT0),
+ PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_OUT1),
+ PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_OUT2),
+ PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_OUT3),
+ PANTHOR_EXCEPTION(MEM_ATTR_FAULT_0),
+ PANTHOR_EXCEPTION(MEM_ATTR_FAULT_1),
+ PANTHOR_EXCEPTION(MEM_ATTR_FAULT_2),
+ PANTHOR_EXCEPTION(MEM_ATTR_FAULT_3),
+};
+
+const char *panthor_exception_name(struct panthor_device *ptdev, u32 exception_code)
+{
+ if (exception_code >= ARRAY_SIZE(panthor_exception_infos) ||
+ !panthor_exception_infos[exception_code].name)
+ return "Unknown exception type";
+
+ return panthor_exception_infos[exception_code].name;
+}
+
+static vm_fault_t panthor_mmio_vm_fault(struct vm_fault *vmf)
+{
+ struct vm_area_struct *vma = vmf->vma;
+ struct panthor_device *ptdev = vma->vm_private_data;
+ u64 id = (u64)vma->vm_pgoff << PAGE_SHIFT;
+ unsigned long pfn;
+ pgprot_t pgprot;
+ vm_fault_t ret;
+ bool active;
+ int cookie;
+
+ if (!drm_dev_enter(&ptdev->base, &cookie))
+ return VM_FAULT_SIGBUS;
+
+ mutex_lock(&ptdev->pm.mmio_lock);
+ active = atomic_read(&ptdev->pm.state) == PANTHOR_DEVICE_PM_STATE_ACTIVE;
+
+ switch (panthor_device_mmio_offset(id)) {
+ case DRM_PANTHOR_USER_FLUSH_ID_MMIO_OFFSET:
+ if (active)
+ pfn = __phys_to_pfn(ptdev->phys_addr + CSF_GPU_LATEST_FLUSH_ID);
+ else
+ pfn = virt_to_pfn(ptdev->pm.dummy_latest_flush);
+ break;
+
+ default:
+ ret = VM_FAULT_SIGBUS;
+ goto out_unlock;
+ }
+
+ pgprot = vma->vm_page_prot;
+ if (active)
+ pgprot = pgprot_noncached(pgprot);
+
+ ret = vmf_insert_pfn_prot(vma, vmf->address, pfn, pgprot);
+
+out_unlock:
+ mutex_unlock(&ptdev->pm.mmio_lock);
+ drm_dev_exit(cookie);
+ return ret;
+}
+
+static const struct vm_operations_struct panthor_mmio_vm_ops = {
+ .fault = panthor_mmio_vm_fault,
+};
+
+int panthor_device_mmap_io(struct panthor_device *ptdev, struct vm_area_struct *vma)
+{
+ u64 id = (u64)vma->vm_pgoff << PAGE_SHIFT;
+
+ switch (panthor_device_mmio_offset(id)) {
+ case DRM_PANTHOR_USER_FLUSH_ID_MMIO_OFFSET:
+ if (vma->vm_end - vma->vm_start != PAGE_SIZE ||
+ (vma->vm_flags & (VM_WRITE | VM_EXEC)))
+ return -EINVAL;
+
+ break;
+
+ default:
+ return -EINVAL;
+ }
+
+ /* Defer actual mapping to the fault handler. */
+ vma->vm_private_data = ptdev;
+ vma->vm_ops = &panthor_mmio_vm_ops;
+ vm_flags_set(vma,
+ VM_IO | VM_DONTCOPY | VM_DONTEXPAND |
+ VM_NORESERVE | VM_DONTDUMP | VM_PFNMAP);
+ return 0;
+}
+
+#ifdef CONFIG_PM
+int panthor_device_resume(struct device *dev)
+{
+ struct panthor_device *ptdev = dev_get_drvdata(dev);
+ int ret, cookie;
+
+ if (atomic_read(&ptdev->pm.state) != PANTHOR_DEVICE_PM_STATE_SUSPENDED)
+ return -EINVAL;
+
+ atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_RESUMING);
+
+ ret = clk_prepare_enable(ptdev->clks.core);
+ if (ret)
+ goto err_set_suspended;
+
+ ret = clk_prepare_enable(ptdev->clks.stacks);
+ if (ret)
+ goto err_disable_core_clk;
+
+ ret = clk_prepare_enable(ptdev->clks.coregroup);
+ if (ret)
+ goto err_disable_stacks_clk;
+
+ ret = panthor_devfreq_resume(ptdev);
+ if (ret)
+ goto err_disable_coregroup_clk;
+
+ if (panthor_device_is_initialized(ptdev) &&
+ drm_dev_enter(&ptdev->base, &cookie)) {
+ panthor_gpu_resume(ptdev);
+ panthor_mmu_resume(ptdev);
+ ret = drm_WARN_ON(&ptdev->base, panthor_fw_resume(ptdev));
+ if (!ret)
+ panthor_sched_resume(ptdev);
+
+ drm_dev_exit(cookie);
+
+ if (ret)
+ goto err_devfreq_suspend;
+ }
+
+ if (atomic_read(&ptdev->reset.pending))
+ queue_work(ptdev->reset.wq, &ptdev->reset.work);
+
+ /* Clear all IOMEM mappings pointing to this device after we've
+ * resumed. This way the fake mappings pointing to the dummy pages
+ * are removed and the real iomem mapping will be restored on next
+ * access.
+ */
+ mutex_lock(&ptdev->pm.mmio_lock);
+ unmap_mapping_range(ptdev->base.anon_inode->i_mapping,
+ DRM_PANTHOR_USER_MMIO_OFFSET, 0, 1);
+ atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_ACTIVE);
+ mutex_unlock(&ptdev->pm.mmio_lock);
+ return 0;
+
+err_devfreq_suspend:
+ panthor_devfreq_suspend(ptdev);
+
+err_disable_coregroup_clk:
+ clk_disable_unprepare(ptdev->clks.coregroup);
+
+err_disable_stacks_clk:
+ clk_disable_unprepare(ptdev->clks.stacks);
+
+err_disable_core_clk:
+ clk_disable_unprepare(ptdev->clks.core);
+
+err_set_suspended:
+ atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
+ return ret;
+}
+
+int panthor_device_suspend(struct device *dev)
+{
+ struct panthor_device *ptdev = dev_get_drvdata(dev);
+ int ret, cookie;
+
+ if (atomic_read(&ptdev->pm.state) != PANTHOR_DEVICE_PM_STATE_ACTIVE)
+ return -EINVAL;
+
+ /* Clear all IOMEM mappings pointing to this device before we
+ * shutdown the power-domain and clocks. Failing to do that results
+ * in external aborts when the process accesses the iomem region.
+ * We change the state and call unmap_mapping_range() with the
+ * mmio_lock held to make sure the vm_fault handler won't set up
+ * invalid mappings.
+ */
+ mutex_lock(&ptdev->pm.mmio_lock);
+ atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDING);
+ unmap_mapping_range(ptdev->base.anon_inode->i_mapping,
+ DRM_PANTHOR_USER_MMIO_OFFSET, 0, 1);
+ mutex_unlock(&ptdev->pm.mmio_lock);
+
+ if (panthor_device_is_initialized(ptdev) &&
+ drm_dev_enter(&ptdev->base, &cookie)) {
+ cancel_work_sync(&ptdev->reset.work);
+
+ /* We prepare everything as if we were resetting the GPU.
+ * The end of the reset will happen in the resume path though.
+ */
+ panthor_sched_suspend(ptdev);
+ panthor_fw_suspend(ptdev);
+ panthor_mmu_suspend(ptdev);
+ panthor_gpu_suspend(ptdev);
+ drm_dev_exit(cookie);
+ }
+
+ ret = panthor_devfreq_suspend(ptdev);
+ if (ret) {
+ if (panthor_device_is_initialized(ptdev) &&
+ drm_dev_enter(&ptdev->base, &cookie)) {
+ panthor_gpu_resume(ptdev);
+ panthor_mmu_resume(ptdev);
+ drm_WARN_ON(&ptdev->base, panthor_fw_resume(ptdev));
+ panthor_sched_resume(ptdev);
+ drm_dev_exit(cookie);
+ }
+
+ goto err_set_active;
+ }
+
+ clk_disable_unprepare(ptdev->clks.coregroup);
+ clk_disable_unprepare(ptdev->clks.stacks);
+ clk_disable_unprepare(ptdev->clks.core);
+ atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
+ return 0;
+
+err_set_active:
+ /* If something failed and we have to revert back to an
+ * active state, we also need to clear the MMIO userspace
+ * mappings, so any dumb pages that were mapped while we
+ * were trying to suspend gets invalidated.
+ */
+ mutex_lock(&ptdev->pm.mmio_lock);
+ atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_ACTIVE);
+ unmap_mapping_range(ptdev->base.anon_inode->i_mapping,
+ DRM_PANTHOR_USER_MMIO_OFFSET, 0, 1);
+ mutex_unlock(&ptdev->pm.mmio_lock);
+ return ret;
+}
+#endif
diff --git a/drivers/gpu/drm/panthor/panthor_device.h b/drivers/gpu/drm/panthor/panthor_device.h
new file mode 100644
index 000000000000..7e385541b448
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_device.h
@@ -0,0 +1,393 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
+/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
+/* Copyright 2023 Collabora ltd. */
+
+#ifndef __PANTHOR_DEVICE_H__
+#define __PANTHOR_DEVICE_H__
+
+#include <linux/atomic.h>
+#include <linux/io-pgtable.h>
+#include <linux/regulator/consumer.h>
+#include <linux/sched.h>
+#include <linux/spinlock.h>
+#include <drm/drm_device.h>
+#include <drm/drm_mm.h>
+#include <drm/gpu_scheduler.h>
+#include <drm/panthor_drm.h>
+
+struct panthor_csf;
+struct panthor_csf_ctx;
+struct panthor_device;
+struct panthor_gpu;
+struct panthor_group_pool;
+struct panthor_heap_pool;
+struct panthor_job;
+struct panthor_mmu;
+struct panthor_fw;
+struct panthor_perfcnt;
+struct panthor_vm;
+struct panthor_vm_pool;
+
+/**
+ * enum panthor_device_pm_state - PM state
+ */
+enum panthor_device_pm_state {
+ /** @PANTHOR_DEVICE_PM_STATE_SUSPENDED: Device is suspended. */
+ PANTHOR_DEVICE_PM_STATE_SUSPENDED = 0,
+
+ /** @PANTHOR_DEVICE_PM_STATE_RESUMING: Device is being resumed. */
+ PANTHOR_DEVICE_PM_STATE_RESUMING,
+
+ /** @PANTHOR_DEVICE_PM_STATE_ACTIVE: Device is active. */
+ PANTHOR_DEVICE_PM_STATE_ACTIVE,
+
+ /** @PANTHOR_DEVICE_PM_STATE_SUSPENDING: Device is being suspended. */
+ PANTHOR_DEVICE_PM_STATE_SUSPENDING,
+};
+
+/**
+ * struct panthor_irq - IRQ data
+ *
+ * Used to automate IRQ handling for the 3 different IRQs we have in this driver.
+ */
+struct panthor_irq {
+ /** @ptdev: Panthor device */
+ struct panthor_device *ptdev;
+
+ /** @irq: IRQ number. */
+ int irq;
+
+ /** @mask: Current mask being applied to xxx_INT_MASK. */
+ u32 mask;
+
+ /** @suspended: Set to true when the IRQ is suspended. */
+ atomic_t suspended;
+};
+
+/**
+ * struct panthor_device - Panthor device
+ */
+struct panthor_device {
+ /** @base: Base drm_device. */
+ struct drm_device base;
+
+ /** @phys_addr: Physical address of the iomem region. */
+ phys_addr_t phys_addr;
+
+ /** @iomem: CPU mapping of the IOMEM region. */
+ void __iomem *iomem;
+
+ /** @clks: GPU clocks. */
+ struct {
+ /** @core: Core clock. */
+ struct clk *core;
+
+ /** @stacks: Stacks clock. This clock is optional. */
+ struct clk *stacks;
+
+ /** @coregroup: Core group clock. This clock is optional. */
+ struct clk *coregroup;
+ } clks;
+
+ /** @coherent: True if the CPU/GPU are memory coherent. */
+ bool coherent;
+
+ /** @gpu_info: GPU information. */
+ struct drm_panthor_gpu_info gpu_info;
+
+ /** @csif_info: Command stream interface information. */
+ struct drm_panthor_csif_info csif_info;
+
+ /** @gpu: GPU management data. */
+ struct panthor_gpu *gpu;
+
+ /** @fw: FW management data. */
+ struct panthor_fw *fw;
+
+ /** @mmu: MMU management data. */
+ struct panthor_mmu *mmu;
+
+ /** @scheduler: Scheduler management data. */
+ struct panthor_scheduler *scheduler;
+
+ /** @devfreq: Device frequency scaling management data. */
+ struct panthor_devfreq *devfreq;
+
+ /** @unplug: Device unplug related fields. */
+ struct {
+ /** @lock: Lock used to serialize unplug operations. */
+ struct mutex lock;
+
+ /**
+ * @done: Completion object signaled when the unplug
+ * operation is done.
+ */
+ struct completion done;
+ } unplug;
+
+ /** @reset: Reset related fields. */
+ struct {
+ /** @wq: Ordered worqueud used to schedule reset operations. */
+ struct workqueue_struct *wq;
+
+ /** @work: Reset work. */
+ struct work_struct work;
+
+ /** @pending: Set to true if a reset is pending. */
+ atomic_t pending;
+ } reset;
+
+ /** @pm: Power management related data. */
+ struct {
+ /** @state: Power state. */
+ atomic_t state;
+
+ /**
+ * @mmio_lock: Lock protecting MMIO userspace CPU mappings.
+ *
+ * This is needed to ensure we map the dummy IO pages when
+ * the device is being suspended, and the real IO pages when
+ * the device is being resumed. We can't just do with the
+ * state atomicity to deal with this race.
+ */
+ struct mutex mmio_lock;
+
+ /**
+ * @dummy_latest_flush: Dummy LATEST_FLUSH page.
+ *
+ * Used to replace the real LATEST_FLUSH page when the GPU
+ * is suspended.
+ */
+ u32 *dummy_latest_flush;
+ } pm;
+};
+
+/**
+ * struct panthor_file - Panthor file
+ */
+struct panthor_file {
+ /** @ptdev: Device attached to this file. */
+ struct panthor_device *ptdev;
+
+ /** @vms: VM pool attached to this file. */
+ struct panthor_vm_pool *vms;
+
+ /** @groups: Scheduling group pool attached to this file. */
+ struct panthor_group_pool *groups;
+};
+
+int panthor_device_init(struct panthor_device *ptdev);
+void panthor_device_unplug(struct panthor_device *ptdev);
+
+/**
+ * panthor_device_schedule_reset() - Schedules a reset operation
+ */
+static inline void panthor_device_schedule_reset(struct panthor_device *ptdev)
+{
+ if (!atomic_cmpxchg(&ptdev->reset.pending, 0, 1) &&
+ atomic_read(&ptdev->pm.state) == PANTHOR_DEVICE_PM_STATE_ACTIVE)
+ queue_work(ptdev->reset.wq, &ptdev->reset.work);
+}
+
+/**
+ * panthor_device_reset_is_pending() - Checks if a reset is pending.
+ *
+ * Return: true if a reset is pending, false otherwise.
+ */
+static inline bool panthor_device_reset_is_pending(struct panthor_device *ptdev)
+{
+ return atomic_read(&ptdev->reset.pending) != 0;
+}
+
+int panthor_device_mmap_io(struct panthor_device *ptdev,
+ struct vm_area_struct *vma);
+
+int panthor_device_resume(struct device *dev);
+int panthor_device_suspend(struct device *dev);
+
+enum drm_panthor_exception_type {
+ DRM_PANTHOR_EXCEPTION_OK = 0x00,
+ DRM_PANTHOR_EXCEPTION_TERMINATED = 0x04,
+ DRM_PANTHOR_EXCEPTION_KABOOM = 0x05,
+ DRM_PANTHOR_EXCEPTION_EUREKA = 0x06,
+ DRM_PANTHOR_EXCEPTION_ACTIVE = 0x08,
+ DRM_PANTHOR_EXCEPTION_CS_RES_TERM = 0x0f,
+ DRM_PANTHOR_EXCEPTION_MAX_NON_FAULT = 0x3f,
+ DRM_PANTHOR_EXCEPTION_CS_CONFIG_FAULT = 0x40,
+ DRM_PANTHOR_EXCEPTION_CS_ENDPOINT_FAULT = 0x44,
+ DRM_PANTHOR_EXCEPTION_CS_BUS_FAULT = 0x48,
+ DRM_PANTHOR_EXCEPTION_CS_INSTR_INVALID = 0x49,
+ DRM_PANTHOR_EXCEPTION_CS_CALL_STACK_OVERFLOW = 0x4a,
+ DRM_PANTHOR_EXCEPTION_CS_INHERIT_FAULT = 0x4b,
+ DRM_PANTHOR_EXCEPTION_INSTR_INVALID_PC = 0x50,
+ DRM_PANTHOR_EXCEPTION_INSTR_INVALID_ENC = 0x51,
+ DRM_PANTHOR_EXCEPTION_INSTR_BARRIER_FAULT = 0x55,
+ DRM_PANTHOR_EXCEPTION_DATA_INVALID_FAULT = 0x58,
+ DRM_PANTHOR_EXCEPTION_TILE_RANGE_FAULT = 0x59,
+ DRM_PANTHOR_EXCEPTION_ADDR_RANGE_FAULT = 0x5a,
+ DRM_PANTHOR_EXCEPTION_IMPRECISE_FAULT = 0x5b,
+ DRM_PANTHOR_EXCEPTION_OOM = 0x60,
+ DRM_PANTHOR_EXCEPTION_CSF_FW_INTERNAL_ERROR = 0x68,
+ DRM_PANTHOR_EXCEPTION_CSF_RES_EVICTION_TIMEOUT = 0x69,
+ DRM_PANTHOR_EXCEPTION_GPU_BUS_FAULT = 0x80,
+ DRM_PANTHOR_EXCEPTION_GPU_SHAREABILITY_FAULT = 0x88,
+ DRM_PANTHOR_EXCEPTION_SYS_SHAREABILITY_FAULT = 0x89,
+ DRM_PANTHOR_EXCEPTION_GPU_CACHEABILITY_FAULT = 0x8a,
+ DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_0 = 0xc0,
+ DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_1 = 0xc1,
+ DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_2 = 0xc2,
+ DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_3 = 0xc3,
+ DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_4 = 0xc4,
+ DRM_PANTHOR_EXCEPTION_PERM_FAULT_0 = 0xc8,
+ DRM_PANTHOR_EXCEPTION_PERM_FAULT_1 = 0xc9,
+ DRM_PANTHOR_EXCEPTION_PERM_FAULT_2 = 0xca,
+ DRM_PANTHOR_EXCEPTION_PERM_FAULT_3 = 0xcb,
+ DRM_PANTHOR_EXCEPTION_ACCESS_FLAG_1 = 0xd9,
+ DRM_PANTHOR_EXCEPTION_ACCESS_FLAG_2 = 0xda,
+ DRM_PANTHOR_EXCEPTION_ACCESS_FLAG_3 = 0xdb,
+ DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_IN = 0xe0,
+ DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_OUT0 = 0xe4,
+ DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_OUT1 = 0xe5,
+ DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_OUT2 = 0xe6,
+ DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_OUT3 = 0xe7,
+ DRM_PANTHOR_EXCEPTION_MEM_ATTR_FAULT_0 = 0xe8,
+ DRM_PANTHOR_EXCEPTION_MEM_ATTR_FAULT_1 = 0xe9,
+ DRM_PANTHOR_EXCEPTION_MEM_ATTR_FAULT_2 = 0xea,
+ DRM_PANTHOR_EXCEPTION_MEM_ATTR_FAULT_3 = 0xeb,
+};
+
+/**
+ * panthor_exception_is_fault() - Checks if an exception is a fault.
+ *
+ * Return: true if the exception is a fault, false otherwise.
+ */
+static inline bool
+panthor_exception_is_fault(u32 exception_code)
+{
+ return exception_code > DRM_PANTHOR_EXCEPTION_MAX_NON_FAULT;
+}
+
+const char *panthor_exception_name(struct panthor_device *ptdev,
+ u32 exception_code);
+
+/**
+ * PANTHOR_IRQ_HANDLER() - Define interrupt handlers and the interrupt
+ * registration function.
+ *
+ * The boiler-plate to gracefully deal with shared interrupts is
+ * auto-generated. All you have to do is call PANTHOR_IRQ_HANDLER()
+ * just after the actual handler. The handler prototype is:
+ *
+ * void (*handler)(struct panthor_device *, u32 status);
+ */
+#define PANTHOR_IRQ_HANDLER(__name, __reg_prefix, __handler) \
+static irqreturn_t panthor_ ## __name ## _irq_raw_handler(int irq, void *data) \
+{ \
+ struct panthor_irq *pirq = data; \
+ struct panthor_device *ptdev = pirq->ptdev; \
+ \
+ if (atomic_read(&pirq->suspended)) \
+ return IRQ_NONE; \
+ if (!gpu_read(ptdev, __reg_prefix ## _INT_STAT)) \
+ return IRQ_NONE; \
+ \
+ gpu_write(ptdev, __reg_prefix ## _INT_MASK, 0); \
+ return IRQ_WAKE_THREAD; \
+} \
+ \
+static irqreturn_t panthor_ ## __name ## _irq_threaded_handler(int irq, void *data) \
+{ \
+ struct panthor_irq *pirq = data; \
+ struct panthor_device *ptdev = pirq->ptdev; \
+ irqreturn_t ret = IRQ_NONE; \
+ \
+ while (true) { \
+ u32 status = gpu_read(ptdev, __reg_prefix ## _INT_RAWSTAT) & pirq->mask; \
+ \
+ if (!status) \
+ break; \
+ \
+ gpu_write(ptdev, __reg_prefix ## _INT_CLEAR, status); \
+ \
+ __handler(ptdev, status); \
+ ret = IRQ_HANDLED; \
+ } \
+ \
+ if (!atomic_read(&pirq->suspended)) \
+ gpu_write(ptdev, __reg_prefix ## _INT_MASK, pirq->mask); \
+ \
+ return ret; \
+} \
+ \
+static inline void panthor_ ## __name ## _irq_suspend(struct panthor_irq *pirq) \
+{ \
+ int cookie; \
+ \
+ atomic_set(&pirq->suspended, true); \
+ \
+ if (drm_dev_enter(&pirq->ptdev->base, &cookie)) { \
+ gpu_write(pirq->ptdev, __reg_prefix ## _INT_MASK, 0); \
+ synchronize_irq(pirq->irq); \
+ drm_dev_exit(cookie); \
+ } \
+ \
+ pirq->mask = 0; \
+} \
+ \
+static inline void panthor_ ## __name ## _irq_resume(struct panthor_irq *pirq, u32 mask) \
+{ \
+ int cookie; \
+ \
+ atomic_set(&pirq->suspended, false); \
+ pirq->mask = mask; \
+ \
+ if (drm_dev_enter(&pirq->ptdev->base, &cookie)) { \
+ gpu_write(pirq->ptdev, __reg_prefix ## _INT_CLEAR, mask); \
+ gpu_write(pirq->ptdev, __reg_prefix ## _INT_MASK, mask); \
+ drm_dev_exit(cookie); \
+ } \
+} \
+ \
+static int panthor_request_ ## __name ## _irq(struct panthor_device *ptdev, \
+ struct panthor_irq *pirq, \
+ int irq, u32 mask) \
+{ \
+ pirq->ptdev = ptdev; \
+ pirq->irq = irq; \
+ panthor_ ## __name ## _irq_resume(pirq, mask); \
+ \
+ return devm_request_threaded_irq(ptdev->base.dev, irq, \
+ panthor_ ## __name ## _irq_raw_handler, \
+ panthor_ ## __name ## _irq_threaded_handler, \
+ IRQF_SHARED, KBUILD_MODNAME "-" # __name, \
+ pirq); \
+}
+
+/**
+ * panthor_device_mmio_offset() - Turn a user MMIO offset into a kernel one
+ * @offset: Offset to convert.
+ *
+ * With 32-bit systems being limited by the 32-bit representation of mmap2's
+ * pgoffset field, we need to make the MMIO offset arch specific. This function
+ * converts a user MMIO offset into something the kernel driver understands.
+ *
+ * If the kernel and userspace architecture match, the offset is unchanged. If
+ * the kernel is 64-bit and userspace is 32-bit, the offset is adjusted to match
+ * 64-bit offsets. 32-bit kernel with 64-bit userspace is impossible.
+ *
+ * Return: Adjusted offset.
+ */
+static inline u64 panthor_device_mmio_offset(u64 offset)
+{
+#ifdef CONFIG_ARM64
+ if (test_tsk_thread_flag(current, TIF_32BIT))
+ offset += DRM_PANTHOR_USER_MMIO_OFFSET_64BIT - DRM_PANTHOR_USER_MMIO_OFFSET_32BIT;
+#endif
+
+ return offset;
+}
+
+extern struct workqueue_struct *panthor_cleanup_wq;
+
+#endif
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 03/14] drm/panthor: Add the device logical block
2024-01-22 16:30 ` [PATCH v4 03/14] drm/panthor: Add the device logical block Boris Brezillon
@ 2024-02-08 14:30 ` Liviu Dudau
2024-02-08 15:14 ` Boris Brezillon
2024-02-08 15:19 ` Boris Brezillon
0 siblings, 2 replies; 51+ messages in thread
From: Liviu Dudau @ 2024-02-08 14:30 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
Hi Boris,
On Mon, Jan 22, 2024 at 05:30:34PM +0100, Boris Brezillon wrote:
> The panthor driver is designed in a modular way, where each logical
> block is dealing with a specific HW-block or software feature. In order
> for those blocks to communicate with each other, we need a central
> panthor_device collecting all the blocks, and exposing some common
> features, like interrupt handling, power management, reset, ...
>
> This what this panthor_device logical block is about.
>
> v4:
> - Check drmm_mutex_init() return code
> - Fix panthor_device_reset_work() out path
> - Fix the race in the unplug logic
> - Fix typos
> - Unplug blocks when something fails in panthor_device_init()
> - Add Steve's R-b
>
> v3:
> - Add acks for the MIT+GPL2 relicensing
> - Fix 32-bit support
> - Shorten the sections protected by panthor_device::pm::mmio_lock to fix
> lock ordering issues.
> - Rename panthor_device::pm::lock into panthor_device::pm::mmio_lock to
> better reflect what this lock is protecting
> - Use dev_err_probe()
> - Make sure we call drm_dev_exit() when something fails half-way in
> panthor_device_reset_work()
> - Replace CSF_GPU_LATEST_FLUSH_ID_DEFAULT with a constant '1' and a
> comment to explain. Also remove setting the dummy flush ID on suspend.
> - Remove drm_WARN_ON() in panthor_exception_name()
> - Check pirq->suspended in panthor_xxx_irq_raw_handler()
>
> Co-developed-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
> Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
> Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
> Reviewed-by: Steven Price <steven.price@arm.com>
> ---
> drivers/gpu/drm/panthor/panthor_device.c | 544 +++++++++++++++++++++++
> drivers/gpu/drm/panthor/panthor_device.h | 393 ++++++++++++++++
> 2 files changed, 937 insertions(+)
> create mode 100644 drivers/gpu/drm/panthor/panthor_device.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_device.h
>
> diff --git a/drivers/gpu/drm/panthor/panthor_device.c b/drivers/gpu/drm/panthor/panthor_device.c
> new file mode 100644
> index 000000000000..dda41a1dd47e
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_device.c
> @@ -0,0 +1,544 @@
> +// SPDX-License-Identifier: GPL-2.0 or MIT
> +/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
> +/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
> +/* Copyright 2023 Collabora ltd. */
> +
> +#include <linux/clk.h>
> +#include <linux/reset.h>
> +#include <linux/platform_device.h>
> +#include <linux/pm_domain.h>
> +#include <linux/pm_runtime.h>
> +#include <linux/regulator/consumer.h>
> +
> +#include <drm/drm_drv.h>
> +#include <drm/drm_managed.h>
> +
> +#include "panthor_sched.h"
> +#include "panthor_device.h"
> +#include "panthor_devfreq.h"
> +#include "panthor_gpu.h"
> +#include "panthor_fw.h"
> +#include "panthor_mmu.h"
> +#include "panthor_regs.h"
> +
> +static int panthor_clk_init(struct panthor_device *ptdev)
> +{
> + ptdev->clks.core = devm_clk_get(ptdev->base.dev, NULL);
> + if (IS_ERR(ptdev->clks.core))
> + return dev_err_probe(ptdev->base.dev,
> + PTR_ERR(ptdev->clks.core),
> + "get 'core' clock failed");
> +
> + ptdev->clks.stacks = devm_clk_get_optional(ptdev->base.dev, "stacks");
> + if (IS_ERR(ptdev->clks.stacks))
> + return dev_err_probe(ptdev->base.dev,
> + PTR_ERR(ptdev->clks.stacks),
> + "get 'stacks' clock failed");
> +
> + ptdev->clks.coregroup = devm_clk_get_optional(ptdev->base.dev, "coregroup");
> + if (IS_ERR(ptdev->clks.coregroup))
> + return dev_err_probe(ptdev->base.dev,
> + PTR_ERR(ptdev->clks.coregroup),
> + "get 'coregroup' clock failed");
> +
> + drm_info(&ptdev->base, "clock rate = %lu\n", clk_get_rate(ptdev->clks.core));
> + return 0;
> +}
> +
> +void panthor_device_unplug(struct panthor_device *ptdev)
> +{
> + /* This function can be called from two different path: the reset work
> + * and the platform device remove callback. drm_dev_unplug() doesn't
> + * deal with concurrent callers, so we have to protect drm_dev_unplug()
> + * calls with our own lock, and bail out if the device is already
> + * unplugged.
> + */
> + mutex_lock(&ptdev->unplug.lock);
> + if (drm_dev_is_unplugged(&ptdev->base)) {
> + /* Someone beat us, release the lock and wait for the unplug
> + * operation to be reported as done.
> + **/
> + mutex_unlock(&ptdev->unplug.lock);
> + wait_for_completion(&ptdev->unplug.done);
> + return;
> + }
> +
> + /* Call drm_dev_unplug() so any access to HW blocks happening after
> + * that point get rejected.
> + */
> + drm_dev_unplug(&ptdev->base);
> +
> + /* We do the rest of the unplug with the unplug lock released,
> + * future callers will wait on ptdev->unplug.done anyway.
> + */
> + mutex_unlock(&ptdev->unplug.lock);
> +
> + drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> +
> + /* Now, try to cleanly shutdown the GPU before the device resources
> + * get reclaimed.
> + */
> + panthor_sched_unplug(ptdev);
> + panthor_fw_unplug(ptdev);
> + panthor_mmu_unplug(ptdev);
> + panthor_gpu_unplug(ptdev);
> +
> + pm_runtime_dont_use_autosuspend(ptdev->base.dev);
> + pm_runtime_put_sync_suspend(ptdev->base.dev);
> +
> + /* Report the unplug operation as done to unblock concurrent
> + * panthor_device_unplug() callers.
> + */
> + complete_all(&ptdev->unplug.done);
> +}
> +
> +static void panthor_device_reset_cleanup(struct drm_device *ddev, void *data)
> +{
> + struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
> +
> + cancel_work_sync(&ptdev->reset.work);
> + destroy_workqueue(ptdev->reset.wq);
> +}
> +
> +static void panthor_device_reset_work(struct work_struct *work)
> +{
> + struct panthor_device *ptdev = container_of(work, struct panthor_device, reset.work);
> + int ret = 0, cookie;
> +
> + if (atomic_read(&ptdev->pm.state) != PANTHOR_DEVICE_PM_STATE_ACTIVE) {
> + /*
> + * No need for a reset as the device has been (or will be)
> + * powered down
> + */
> + atomic_set(&ptdev->reset.pending, 0);
> + return;
> + }
> +
> + if (!drm_dev_enter(&ptdev->base, &cookie))
> + return;
> +
> + panthor_sched_pre_reset(ptdev);
> + panthor_fw_pre_reset(ptdev, true);
> + panthor_mmu_pre_reset(ptdev);
> + panthor_gpu_soft_reset(ptdev);
> + panthor_gpu_l2_power_on(ptdev);
> + panthor_mmu_post_reset(ptdev);
> + ret = panthor_fw_post_reset(ptdev);
> + if (ret)
> + goto out_dev_exit;
> +
> + atomic_set(&ptdev->reset.pending, 0);
> + panthor_sched_post_reset(ptdev);
> +
> +out_dev_exit:
> + drm_dev_exit(cookie);
> +
> + if (ret) {
> + panthor_device_unplug(ptdev);
> + drm_err(&ptdev->base, "Failed to boot MCU after reset, making device unusable.");
> + }
> +}
> +
> +static bool panthor_device_is_initialized(struct panthor_device *ptdev)
> +{
> + return !!ptdev->scheduler;
> +}
> +
> +static void panthor_device_free_page(struct drm_device *ddev, void *data)
> +{
> + free_page((unsigned long)data);
> +}
> +
> +int panthor_device_init(struct panthor_device *ptdev)
> +{
> + struct resource *res;
> + struct page *p;
> + int ret;
> +
> + ptdev->coherent = device_get_dma_attr(ptdev->base.dev) == DEV_DMA_COHERENT;
> +
> + init_completion(&ptdev->unplug.done);
> + ret = drmm_mutex_init(&ptdev->base, &ptdev->unplug.lock);
> + if (ret)
> + return ret;
> +
> + ret = drmm_mutex_init(&ptdev->base, &ptdev->pm.mmio_lock);
> + if (ret)
> + return ret;
> +
> + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> + p = alloc_page(GFP_KERNEL | __GFP_ZERO);
> + if (!p)
> + return -ENOMEM;
> +
> + ptdev->pm.dummy_latest_flush = page_address(p);
> + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_free_page,
> + ptdev->pm.dummy_latest_flush);
> + if (ret)
> + return ret;
> +
> + /*
> + * Set the dummy page holding the latest flush to 1. This will cause the
> + * flush to avoided as we know it isn't necessary if the submission
> + * happens while the dummy page is mapped. Zero cannot be used because
> + * that means 'always flush'.
> + */
> + *ptdev->pm.dummy_latest_flush = 1;
> +
> + INIT_WORK(&ptdev->reset.work, panthor_device_reset_work);
> + ptdev->reset.wq = alloc_ordered_workqueue("panthor-reset-wq", 0);
> + if (!ptdev->reset.wq)
> + return -ENOMEM;
> +
> + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_reset_cleanup, NULL);
> + if (ret)
> + return ret;
> +
> + ret = panthor_clk_init(ptdev);
> + if (ret)
> + return ret;
> +
> + ret = panthor_devfreq_init(ptdev);
> + if (ret)
> + return ret;
> +
> + ptdev->iomem = devm_platform_get_and_ioremap_resource(to_platform_device(ptdev->base.dev),
> + 0, &res);
> + if (IS_ERR(ptdev->iomem))
> + return PTR_ERR(ptdev->iomem);
> +
> + ptdev->phys_addr = res->start;
> +
> + ret = devm_pm_runtime_enable(ptdev->base.dev);
> + if (ret)
> + return ret;
> +
> + ret = pm_runtime_resume_and_get(ptdev->base.dev);
> + if (ret)
> + return ret;
> +
> + ret = panthor_gpu_init(ptdev);
> + if (ret)
> + goto err_rpm_put;
> +
> + ret = panthor_mmu_init(ptdev);
> + if (ret)
> + goto err_unplug_gpu;
> +
> + ret = panthor_fw_init(ptdev);
> + if (ret)
> + goto err_unplug_mmu;
> +
> + ret = panthor_sched_init(ptdev);
> + if (ret)
> + goto err_unplug_fw;
> +
> + /* ~3 frames */
> + pm_runtime_set_autosuspend_delay(ptdev->base.dev, 50);
> +
> + ret = drm_dev_register(&ptdev->base, 0);
> + if (ret)
> + goto err_unplug_sched;
For sake of replicating the panthor_device_unplus() calls, should we do
here:
if (ret) {
pm_runtime_dont_use_autosuspend(ptdev->base.dev);
goto err_unplug_sched;
}
> +
> + pm_runtime_use_autosuspend(ptdev->base.dev);
> + pm_runtime_put_autosuspend(ptdev->base.dev);
> + return 0;
> +
> +err_unplug_sched:
> + panthor_sched_unplug(ptdev);
> +
> +err_unplug_fw:
> + panthor_fw_unplug(ptdev);
> +
> +err_unplug_mmu:
> + panthor_mmu_unplug(ptdev);
> +
> +err_unplug_gpu:
> + panthor_gpu_unplug(ptdev);
> +
> +err_rpm_put:
> + pm_runtime_put_sync_suspend(ptdev->base.dev);
> + return ret;
> +}
> +
> +#define PANTHOR_EXCEPTION(id) \
> + [DRM_PANTHOR_EXCEPTION_ ## id] = { \
> + .name = #id, \
> + }
> +
> +struct panthor_exception_info {
> + const char *name;
> +};
> +
> +static const struct panthor_exception_info panthor_exception_infos[] = {
> + PANTHOR_EXCEPTION(OK),
> + PANTHOR_EXCEPTION(TERMINATED),
> + PANTHOR_EXCEPTION(KABOOM),
> + PANTHOR_EXCEPTION(EUREKA),
> + PANTHOR_EXCEPTION(ACTIVE),
> + PANTHOR_EXCEPTION(CS_RES_TERM),
> + PANTHOR_EXCEPTION(CS_CONFIG_FAULT),
> + PANTHOR_EXCEPTION(CS_ENDPOINT_FAULT),
> + PANTHOR_EXCEPTION(CS_BUS_FAULT),
> + PANTHOR_EXCEPTION(CS_INSTR_INVALID),
> + PANTHOR_EXCEPTION(CS_CALL_STACK_OVERFLOW),
> + PANTHOR_EXCEPTION(CS_INHERIT_FAULT),
> + PANTHOR_EXCEPTION(INSTR_INVALID_PC),
> + PANTHOR_EXCEPTION(INSTR_INVALID_ENC),
> + PANTHOR_EXCEPTION(INSTR_BARRIER_FAULT),
> + PANTHOR_EXCEPTION(DATA_INVALID_FAULT),
> + PANTHOR_EXCEPTION(TILE_RANGE_FAULT),
> + PANTHOR_EXCEPTION(ADDR_RANGE_FAULT),
> + PANTHOR_EXCEPTION(IMPRECISE_FAULT),
> + PANTHOR_EXCEPTION(OOM),
> + PANTHOR_EXCEPTION(CSF_FW_INTERNAL_ERROR),
> + PANTHOR_EXCEPTION(CSF_RES_EVICTION_TIMEOUT),
> + PANTHOR_EXCEPTION(GPU_BUS_FAULT),
> + PANTHOR_EXCEPTION(GPU_SHAREABILITY_FAULT),
> + PANTHOR_EXCEPTION(SYS_SHAREABILITY_FAULT),
> + PANTHOR_EXCEPTION(GPU_CACHEABILITY_FAULT),
> + PANTHOR_EXCEPTION(TRANSLATION_FAULT_0),
> + PANTHOR_EXCEPTION(TRANSLATION_FAULT_1),
> + PANTHOR_EXCEPTION(TRANSLATION_FAULT_2),
> + PANTHOR_EXCEPTION(TRANSLATION_FAULT_3),
> + PANTHOR_EXCEPTION(TRANSLATION_FAULT_4),
> + PANTHOR_EXCEPTION(PERM_FAULT_0),
> + PANTHOR_EXCEPTION(PERM_FAULT_1),
> + PANTHOR_EXCEPTION(PERM_FAULT_2),
> + PANTHOR_EXCEPTION(PERM_FAULT_3),
> + PANTHOR_EXCEPTION(ACCESS_FLAG_1),
> + PANTHOR_EXCEPTION(ACCESS_FLAG_2),
> + PANTHOR_EXCEPTION(ACCESS_FLAG_3),
> + PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_IN),
> + PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_OUT0),
> + PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_OUT1),
> + PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_OUT2),
> + PANTHOR_EXCEPTION(ADDR_SIZE_FAULT_OUT3),
> + PANTHOR_EXCEPTION(MEM_ATTR_FAULT_0),
> + PANTHOR_EXCEPTION(MEM_ATTR_FAULT_1),
> + PANTHOR_EXCEPTION(MEM_ATTR_FAULT_2),
> + PANTHOR_EXCEPTION(MEM_ATTR_FAULT_3),
> +};
> +
> +const char *panthor_exception_name(struct panthor_device *ptdev, u32 exception_code)
> +{
> + if (exception_code >= ARRAY_SIZE(panthor_exception_infos) ||
> + !panthor_exception_infos[exception_code].name)
> + return "Unknown exception type";
> +
> + return panthor_exception_infos[exception_code].name;
> +}
> +
> +static vm_fault_t panthor_mmio_vm_fault(struct vm_fault *vmf)
> +{
> + struct vm_area_struct *vma = vmf->vma;
> + struct panthor_device *ptdev = vma->vm_private_data;
> + u64 id = (u64)vma->vm_pgoff << PAGE_SHIFT;
> + unsigned long pfn;
> + pgprot_t pgprot;
> + vm_fault_t ret;
> + bool active;
> + int cookie;
> +
> + if (!drm_dev_enter(&ptdev->base, &cookie))
> + return VM_FAULT_SIGBUS;
> +
> + mutex_lock(&ptdev->pm.mmio_lock);
> + active = atomic_read(&ptdev->pm.state) == PANTHOR_DEVICE_PM_STATE_ACTIVE;
> +
> + switch (panthor_device_mmio_offset(id)) {
> + case DRM_PANTHOR_USER_FLUSH_ID_MMIO_OFFSET:
> + if (active)
> + pfn = __phys_to_pfn(ptdev->phys_addr + CSF_GPU_LATEST_FLUSH_ID);
> + else
> + pfn = virt_to_pfn(ptdev->pm.dummy_latest_flush);
> + break;
> +
> + default:
> + ret = VM_FAULT_SIGBUS;
> + goto out_unlock;
> + }
> +
> + pgprot = vma->vm_page_prot;
> + if (active)
> + pgprot = pgprot_noncached(pgprot);
> +
> + ret = vmf_insert_pfn_prot(vma, vmf->address, pfn, pgprot);
> +
> +out_unlock:
> + mutex_unlock(&ptdev->pm.mmio_lock);
> + drm_dev_exit(cookie);
> + return ret;
> +}
> +
> +static const struct vm_operations_struct panthor_mmio_vm_ops = {
> + .fault = panthor_mmio_vm_fault,
> +};
> +
> +int panthor_device_mmap_io(struct panthor_device *ptdev, struct vm_area_struct *vma)
> +{
> + u64 id = (u64)vma->vm_pgoff << PAGE_SHIFT;
> +
> + switch (panthor_device_mmio_offset(id)) {
> + case DRM_PANTHOR_USER_FLUSH_ID_MMIO_OFFSET:
> + if (vma->vm_end - vma->vm_start != PAGE_SIZE ||
> + (vma->vm_flags & (VM_WRITE | VM_EXEC)))
> + return -EINVAL;
> +
> + break;
> +
> + default:
> + return -EINVAL;
> + }
> +
> + /* Defer actual mapping to the fault handler. */
> + vma->vm_private_data = ptdev;
> + vma->vm_ops = &panthor_mmio_vm_ops;
> + vm_flags_set(vma,
> + VM_IO | VM_DONTCOPY | VM_DONTEXPAND |
> + VM_NORESERVE | VM_DONTDUMP | VM_PFNMAP);
> + return 0;
> +}
> +
> +#ifdef CONFIG_PM
> +int panthor_device_resume(struct device *dev)
> +{
> + struct panthor_device *ptdev = dev_get_drvdata(dev);
> + int ret, cookie;
> +
> + if (atomic_read(&ptdev->pm.state) != PANTHOR_DEVICE_PM_STATE_SUSPENDED)
> + return -EINVAL;
> +
> + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_RESUMING);
> +
> + ret = clk_prepare_enable(ptdev->clks.core);
> + if (ret)
> + goto err_set_suspended;
> +
> + ret = clk_prepare_enable(ptdev->clks.stacks);
> + if (ret)
> + goto err_disable_core_clk;
> +
> + ret = clk_prepare_enable(ptdev->clks.coregroup);
> + if (ret)
> + goto err_disable_stacks_clk;
> +
> + ret = panthor_devfreq_resume(ptdev);
> + if (ret)
> + goto err_disable_coregroup_clk;
> +
> + if (panthor_device_is_initialized(ptdev) &&
> + drm_dev_enter(&ptdev->base, &cookie)) {
> + panthor_gpu_resume(ptdev);
> + panthor_mmu_resume(ptdev);
> + ret = drm_WARN_ON(&ptdev->base, panthor_fw_resume(ptdev));
> + if (!ret)
> + panthor_sched_resume(ptdev);
> +
> + drm_dev_exit(cookie);
> +
> + if (ret)
> + goto err_devfreq_suspend;
> + }
> +
> + if (atomic_read(&ptdev->reset.pending))
> + queue_work(ptdev->reset.wq, &ptdev->reset.work);
> +
> + /* Clear all IOMEM mappings pointing to this device after we've
> + * resumed. This way the fake mappings pointing to the dummy pages
> + * are removed and the real iomem mapping will be restored on next
> + * access.
> + */
> + mutex_lock(&ptdev->pm.mmio_lock);
> + unmap_mapping_range(ptdev->base.anon_inode->i_mapping,
> + DRM_PANTHOR_USER_MMIO_OFFSET, 0, 1);
> + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_ACTIVE);
> + mutex_unlock(&ptdev->pm.mmio_lock);
> + return 0;
> +
> +err_devfreq_suspend:
> + panthor_devfreq_suspend(ptdev);
Should we also call panthor_mmu_suspend(ptdev) and panthor_gpu_suspend(ptdev)
before panthor_devfreq_suspend()? Otherwise those blocks will be left in an
inconsistent state.
With this addressed, Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
Best regards,
Liviu
> +
> +err_disable_coregroup_clk:
> + clk_disable_unprepare(ptdev->clks.coregroup);
> +
> +err_disable_stacks_clk:
> + clk_disable_unprepare(ptdev->clks.stacks);
> +
> +err_disable_core_clk:
> + clk_disable_unprepare(ptdev->clks.core);
> +
> +err_set_suspended:
> + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> + return ret;
> +}
> +
> +int panthor_device_suspend(struct device *dev)
> +{
> + struct panthor_device *ptdev = dev_get_drvdata(dev);
> + int ret, cookie;
> +
> + if (atomic_read(&ptdev->pm.state) != PANTHOR_DEVICE_PM_STATE_ACTIVE)
> + return -EINVAL;
> +
> + /* Clear all IOMEM mappings pointing to this device before we
> + * shutdown the power-domain and clocks. Failing to do that results
> + * in external aborts when the process accesses the iomem region.
> + * We change the state and call unmap_mapping_range() with the
> + * mmio_lock held to make sure the vm_fault handler won't set up
> + * invalid mappings.
> + */
> + mutex_lock(&ptdev->pm.mmio_lock);
> + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDING);
> + unmap_mapping_range(ptdev->base.anon_inode->i_mapping,
> + DRM_PANTHOR_USER_MMIO_OFFSET, 0, 1);
> + mutex_unlock(&ptdev->pm.mmio_lock);
> +
> + if (panthor_device_is_initialized(ptdev) &&
> + drm_dev_enter(&ptdev->base, &cookie)) {
> + cancel_work_sync(&ptdev->reset.work);
> +
> + /* We prepare everything as if we were resetting the GPU.
> + * The end of the reset will happen in the resume path though.
> + */
> + panthor_sched_suspend(ptdev);
> + panthor_fw_suspend(ptdev);
> + panthor_mmu_suspend(ptdev);
> + panthor_gpu_suspend(ptdev);
> + drm_dev_exit(cookie);
> + }
> +
> + ret = panthor_devfreq_suspend(ptdev);
> + if (ret) {
> + if (panthor_device_is_initialized(ptdev) &&
> + drm_dev_enter(&ptdev->base, &cookie)) {
> + panthor_gpu_resume(ptdev);
> + panthor_mmu_resume(ptdev);
> + drm_WARN_ON(&ptdev->base, panthor_fw_resume(ptdev));
> + panthor_sched_resume(ptdev);
> + drm_dev_exit(cookie);
> + }
> +
> + goto err_set_active;
> + }
> +
> + clk_disable_unprepare(ptdev->clks.coregroup);
> + clk_disable_unprepare(ptdev->clks.stacks);
> + clk_disable_unprepare(ptdev->clks.core);
> + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> + return 0;
> +
> +err_set_active:
> + /* If something failed and we have to revert back to an
> + * active state, we also need to clear the MMIO userspace
> + * mappings, so any dumb pages that were mapped while we
> + * were trying to suspend gets invalidated.
> + */
> + mutex_lock(&ptdev->pm.mmio_lock);
> + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_ACTIVE);
> + unmap_mapping_range(ptdev->base.anon_inode->i_mapping,
> + DRM_PANTHOR_USER_MMIO_OFFSET, 0, 1);
> + mutex_unlock(&ptdev->pm.mmio_lock);
> + return ret;
> +}
> +#endif
> diff --git a/drivers/gpu/drm/panthor/panthor_device.h b/drivers/gpu/drm/panthor/panthor_device.h
> new file mode 100644
> index 000000000000..7e385541b448
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_device.h
> @@ -0,0 +1,393 @@
> +/* SPDX-License-Identifier: GPL-2.0 or MIT */
> +/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
> +/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
> +/* Copyright 2023 Collabora ltd. */
> +
> +#ifndef __PANTHOR_DEVICE_H__
> +#define __PANTHOR_DEVICE_H__
> +
> +#include <linux/atomic.h>
> +#include <linux/io-pgtable.h>
> +#include <linux/regulator/consumer.h>
> +#include <linux/sched.h>
> +#include <linux/spinlock.h>
> +#include <drm/drm_device.h>
> +#include <drm/drm_mm.h>
> +#include <drm/gpu_scheduler.h>
> +#include <drm/panthor_drm.h>
> +
> +struct panthor_csf;
> +struct panthor_csf_ctx;
> +struct panthor_device;
> +struct panthor_gpu;
> +struct panthor_group_pool;
> +struct panthor_heap_pool;
> +struct panthor_job;
> +struct panthor_mmu;
> +struct panthor_fw;
> +struct panthor_perfcnt;
> +struct panthor_vm;
> +struct panthor_vm_pool;
> +
> +/**
> + * enum panthor_device_pm_state - PM state
> + */
> +enum panthor_device_pm_state {
> + /** @PANTHOR_DEVICE_PM_STATE_SUSPENDED: Device is suspended. */
> + PANTHOR_DEVICE_PM_STATE_SUSPENDED = 0,
> +
> + /** @PANTHOR_DEVICE_PM_STATE_RESUMING: Device is being resumed. */
> + PANTHOR_DEVICE_PM_STATE_RESUMING,
> +
> + /** @PANTHOR_DEVICE_PM_STATE_ACTIVE: Device is active. */
> + PANTHOR_DEVICE_PM_STATE_ACTIVE,
> +
> + /** @PANTHOR_DEVICE_PM_STATE_SUSPENDING: Device is being suspended. */
> + PANTHOR_DEVICE_PM_STATE_SUSPENDING,
> +};
> +
> +/**
> + * struct panthor_irq - IRQ data
> + *
> + * Used to automate IRQ handling for the 3 different IRQs we have in this driver.
> + */
> +struct panthor_irq {
> + /** @ptdev: Panthor device */
> + struct panthor_device *ptdev;
> +
> + /** @irq: IRQ number. */
> + int irq;
> +
> + /** @mask: Current mask being applied to xxx_INT_MASK. */
> + u32 mask;
> +
> + /** @suspended: Set to true when the IRQ is suspended. */
> + atomic_t suspended;
> +};
> +
> +/**
> + * struct panthor_device - Panthor device
> + */
> +struct panthor_device {
> + /** @base: Base drm_device. */
> + struct drm_device base;
> +
> + /** @phys_addr: Physical address of the iomem region. */
> + phys_addr_t phys_addr;
> +
> + /** @iomem: CPU mapping of the IOMEM region. */
> + void __iomem *iomem;
> +
> + /** @clks: GPU clocks. */
> + struct {
> + /** @core: Core clock. */
> + struct clk *core;
> +
> + /** @stacks: Stacks clock. This clock is optional. */
> + struct clk *stacks;
> +
> + /** @coregroup: Core group clock. This clock is optional. */
> + struct clk *coregroup;
> + } clks;
> +
> + /** @coherent: True if the CPU/GPU are memory coherent. */
> + bool coherent;
> +
> + /** @gpu_info: GPU information. */
> + struct drm_panthor_gpu_info gpu_info;
> +
> + /** @csif_info: Command stream interface information. */
> + struct drm_panthor_csif_info csif_info;
> +
> + /** @gpu: GPU management data. */
> + struct panthor_gpu *gpu;
> +
> + /** @fw: FW management data. */
> + struct panthor_fw *fw;
> +
> + /** @mmu: MMU management data. */
> + struct panthor_mmu *mmu;
> +
> + /** @scheduler: Scheduler management data. */
> + struct panthor_scheduler *scheduler;
> +
> + /** @devfreq: Device frequency scaling management data. */
> + struct panthor_devfreq *devfreq;
> +
> + /** @unplug: Device unplug related fields. */
> + struct {
> + /** @lock: Lock used to serialize unplug operations. */
> + struct mutex lock;
> +
> + /**
> + * @done: Completion object signaled when the unplug
> + * operation is done.
> + */
> + struct completion done;
> + } unplug;
> +
> + /** @reset: Reset related fields. */
> + struct {
> + /** @wq: Ordered worqueud used to schedule reset operations. */
> + struct workqueue_struct *wq;
> +
> + /** @work: Reset work. */
> + struct work_struct work;
> +
> + /** @pending: Set to true if a reset is pending. */
> + atomic_t pending;
> + } reset;
> +
> + /** @pm: Power management related data. */
> + struct {
> + /** @state: Power state. */
> + atomic_t state;
> +
> + /**
> + * @mmio_lock: Lock protecting MMIO userspace CPU mappings.
> + *
> + * This is needed to ensure we map the dummy IO pages when
> + * the device is being suspended, and the real IO pages when
> + * the device is being resumed. We can't just do with the
> + * state atomicity to deal with this race.
> + */
> + struct mutex mmio_lock;
> +
> + /**
> + * @dummy_latest_flush: Dummy LATEST_FLUSH page.
> + *
> + * Used to replace the real LATEST_FLUSH page when the GPU
> + * is suspended.
> + */
> + u32 *dummy_latest_flush;
> + } pm;
> +};
> +
> +/**
> + * struct panthor_file - Panthor file
> + */
> +struct panthor_file {
> + /** @ptdev: Device attached to this file. */
> + struct panthor_device *ptdev;
> +
> + /** @vms: VM pool attached to this file. */
> + struct panthor_vm_pool *vms;
> +
> + /** @groups: Scheduling group pool attached to this file. */
> + struct panthor_group_pool *groups;
> +};
> +
> +int panthor_device_init(struct panthor_device *ptdev);
> +void panthor_device_unplug(struct panthor_device *ptdev);
> +
> +/**
> + * panthor_device_schedule_reset() - Schedules a reset operation
> + */
> +static inline void panthor_device_schedule_reset(struct panthor_device *ptdev)
> +{
> + if (!atomic_cmpxchg(&ptdev->reset.pending, 0, 1) &&
> + atomic_read(&ptdev->pm.state) == PANTHOR_DEVICE_PM_STATE_ACTIVE)
> + queue_work(ptdev->reset.wq, &ptdev->reset.work);
> +}
> +
> +/**
> + * panthor_device_reset_is_pending() - Checks if a reset is pending.
> + *
> + * Return: true if a reset is pending, false otherwise.
> + */
> +static inline bool panthor_device_reset_is_pending(struct panthor_device *ptdev)
> +{
> + return atomic_read(&ptdev->reset.pending) != 0;
> +}
> +
> +int panthor_device_mmap_io(struct panthor_device *ptdev,
> + struct vm_area_struct *vma);
> +
> +int panthor_device_resume(struct device *dev);
> +int panthor_device_suspend(struct device *dev);
> +
> +enum drm_panthor_exception_type {
> + DRM_PANTHOR_EXCEPTION_OK = 0x00,
> + DRM_PANTHOR_EXCEPTION_TERMINATED = 0x04,
> + DRM_PANTHOR_EXCEPTION_KABOOM = 0x05,
> + DRM_PANTHOR_EXCEPTION_EUREKA = 0x06,
> + DRM_PANTHOR_EXCEPTION_ACTIVE = 0x08,
> + DRM_PANTHOR_EXCEPTION_CS_RES_TERM = 0x0f,
> + DRM_PANTHOR_EXCEPTION_MAX_NON_FAULT = 0x3f,
> + DRM_PANTHOR_EXCEPTION_CS_CONFIG_FAULT = 0x40,
> + DRM_PANTHOR_EXCEPTION_CS_ENDPOINT_FAULT = 0x44,
> + DRM_PANTHOR_EXCEPTION_CS_BUS_FAULT = 0x48,
> + DRM_PANTHOR_EXCEPTION_CS_INSTR_INVALID = 0x49,
> + DRM_PANTHOR_EXCEPTION_CS_CALL_STACK_OVERFLOW = 0x4a,
> + DRM_PANTHOR_EXCEPTION_CS_INHERIT_FAULT = 0x4b,
> + DRM_PANTHOR_EXCEPTION_INSTR_INVALID_PC = 0x50,
> + DRM_PANTHOR_EXCEPTION_INSTR_INVALID_ENC = 0x51,
> + DRM_PANTHOR_EXCEPTION_INSTR_BARRIER_FAULT = 0x55,
> + DRM_PANTHOR_EXCEPTION_DATA_INVALID_FAULT = 0x58,
> + DRM_PANTHOR_EXCEPTION_TILE_RANGE_FAULT = 0x59,
> + DRM_PANTHOR_EXCEPTION_ADDR_RANGE_FAULT = 0x5a,
> + DRM_PANTHOR_EXCEPTION_IMPRECISE_FAULT = 0x5b,
> + DRM_PANTHOR_EXCEPTION_OOM = 0x60,
> + DRM_PANTHOR_EXCEPTION_CSF_FW_INTERNAL_ERROR = 0x68,
> + DRM_PANTHOR_EXCEPTION_CSF_RES_EVICTION_TIMEOUT = 0x69,
> + DRM_PANTHOR_EXCEPTION_GPU_BUS_FAULT = 0x80,
> + DRM_PANTHOR_EXCEPTION_GPU_SHAREABILITY_FAULT = 0x88,
> + DRM_PANTHOR_EXCEPTION_SYS_SHAREABILITY_FAULT = 0x89,
> + DRM_PANTHOR_EXCEPTION_GPU_CACHEABILITY_FAULT = 0x8a,
> + DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_0 = 0xc0,
> + DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_1 = 0xc1,
> + DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_2 = 0xc2,
> + DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_3 = 0xc3,
> + DRM_PANTHOR_EXCEPTION_TRANSLATION_FAULT_4 = 0xc4,
> + DRM_PANTHOR_EXCEPTION_PERM_FAULT_0 = 0xc8,
> + DRM_PANTHOR_EXCEPTION_PERM_FAULT_1 = 0xc9,
> + DRM_PANTHOR_EXCEPTION_PERM_FAULT_2 = 0xca,
> + DRM_PANTHOR_EXCEPTION_PERM_FAULT_3 = 0xcb,
> + DRM_PANTHOR_EXCEPTION_ACCESS_FLAG_1 = 0xd9,
> + DRM_PANTHOR_EXCEPTION_ACCESS_FLAG_2 = 0xda,
> + DRM_PANTHOR_EXCEPTION_ACCESS_FLAG_3 = 0xdb,
> + DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_IN = 0xe0,
> + DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_OUT0 = 0xe4,
> + DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_OUT1 = 0xe5,
> + DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_OUT2 = 0xe6,
> + DRM_PANTHOR_EXCEPTION_ADDR_SIZE_FAULT_OUT3 = 0xe7,
> + DRM_PANTHOR_EXCEPTION_MEM_ATTR_FAULT_0 = 0xe8,
> + DRM_PANTHOR_EXCEPTION_MEM_ATTR_FAULT_1 = 0xe9,
> + DRM_PANTHOR_EXCEPTION_MEM_ATTR_FAULT_2 = 0xea,
> + DRM_PANTHOR_EXCEPTION_MEM_ATTR_FAULT_3 = 0xeb,
> +};
> +
> +/**
> + * panthor_exception_is_fault() - Checks if an exception is a fault.
> + *
> + * Return: true if the exception is a fault, false otherwise.
> + */
> +static inline bool
> +panthor_exception_is_fault(u32 exception_code)
> +{
> + return exception_code > DRM_PANTHOR_EXCEPTION_MAX_NON_FAULT;
> +}
> +
> +const char *panthor_exception_name(struct panthor_device *ptdev,
> + u32 exception_code);
> +
> +/**
> + * PANTHOR_IRQ_HANDLER() - Define interrupt handlers and the interrupt
> + * registration function.
> + *
> + * The boiler-plate to gracefully deal with shared interrupts is
> + * auto-generated. All you have to do is call PANTHOR_IRQ_HANDLER()
> + * just after the actual handler. The handler prototype is:
> + *
> + * void (*handler)(struct panthor_device *, u32 status);
> + */
> +#define PANTHOR_IRQ_HANDLER(__name, __reg_prefix, __handler) \
> +static irqreturn_t panthor_ ## __name ## _irq_raw_handler(int irq, void *data) \
> +{ \
> + struct panthor_irq *pirq = data; \
> + struct panthor_device *ptdev = pirq->ptdev; \
> + \
> + if (atomic_read(&pirq->suspended)) \
> + return IRQ_NONE; \
> + if (!gpu_read(ptdev, __reg_prefix ## _INT_STAT)) \
> + return IRQ_NONE; \
> + \
> + gpu_write(ptdev, __reg_prefix ## _INT_MASK, 0); \
> + return IRQ_WAKE_THREAD; \
> +} \
> + \
> +static irqreturn_t panthor_ ## __name ## _irq_threaded_handler(int irq, void *data) \
> +{ \
> + struct panthor_irq *pirq = data; \
> + struct panthor_device *ptdev = pirq->ptdev; \
> + irqreturn_t ret = IRQ_NONE; \
> + \
> + while (true) { \
> + u32 status = gpu_read(ptdev, __reg_prefix ## _INT_RAWSTAT) & pirq->mask; \
> + \
> + if (!status) \
> + break; \
> + \
> + gpu_write(ptdev, __reg_prefix ## _INT_CLEAR, status); \
> + \
> + __handler(ptdev, status); \
> + ret = IRQ_HANDLED; \
> + } \
> + \
> + if (!atomic_read(&pirq->suspended)) \
> + gpu_write(ptdev, __reg_prefix ## _INT_MASK, pirq->mask); \
> + \
> + return ret; \
> +} \
> + \
> +static inline void panthor_ ## __name ## _irq_suspend(struct panthor_irq *pirq) \
> +{ \
> + int cookie; \
> + \
> + atomic_set(&pirq->suspended, true); \
> + \
> + if (drm_dev_enter(&pirq->ptdev->base, &cookie)) { \
> + gpu_write(pirq->ptdev, __reg_prefix ## _INT_MASK, 0); \
> + synchronize_irq(pirq->irq); \
> + drm_dev_exit(cookie); \
> + } \
> + \
> + pirq->mask = 0; \
> +} \
> + \
> +static inline void panthor_ ## __name ## _irq_resume(struct panthor_irq *pirq, u32 mask) \
> +{ \
> + int cookie; \
> + \
> + atomic_set(&pirq->suspended, false); \
> + pirq->mask = mask; \
> + \
> + if (drm_dev_enter(&pirq->ptdev->base, &cookie)) { \
> + gpu_write(pirq->ptdev, __reg_prefix ## _INT_CLEAR, mask); \
> + gpu_write(pirq->ptdev, __reg_prefix ## _INT_MASK, mask); \
> + drm_dev_exit(cookie); \
> + } \
> +} \
> + \
> +static int panthor_request_ ## __name ## _irq(struct panthor_device *ptdev, \
> + struct panthor_irq *pirq, \
> + int irq, u32 mask) \
> +{ \
> + pirq->ptdev = ptdev; \
> + pirq->irq = irq; \
> + panthor_ ## __name ## _irq_resume(pirq, mask); \
> + \
> + return devm_request_threaded_irq(ptdev->base.dev, irq, \
> + panthor_ ## __name ## _irq_raw_handler, \
> + panthor_ ## __name ## _irq_threaded_handler, \
> + IRQF_SHARED, KBUILD_MODNAME "-" # __name, \
> + pirq); \
> +}
> +
> +/**
> + * panthor_device_mmio_offset() - Turn a user MMIO offset into a kernel one
> + * @offset: Offset to convert.
> + *
> + * With 32-bit systems being limited by the 32-bit representation of mmap2's
> + * pgoffset field, we need to make the MMIO offset arch specific. This function
> + * converts a user MMIO offset into something the kernel driver understands.
> + *
> + * If the kernel and userspace architecture match, the offset is unchanged. If
> + * the kernel is 64-bit and userspace is 32-bit, the offset is adjusted to match
> + * 64-bit offsets. 32-bit kernel with 64-bit userspace is impossible.
> + *
> + * Return: Adjusted offset.
> + */
> +static inline u64 panthor_device_mmio_offset(u64 offset)
> +{
> +#ifdef CONFIG_ARM64
> + if (test_tsk_thread_flag(current, TIF_32BIT))
> + offset += DRM_PANTHOR_USER_MMIO_OFFSET_64BIT - DRM_PANTHOR_USER_MMIO_OFFSET_32BIT;
> +#endif
> +
> + return offset;
> +}
> +
> +extern struct workqueue_struct *panthor_cleanup_wq;
> +
> +#endif
> --
> 2.43.0
>
--
====================
| I would like to |
| fix the world, |
| but they're not |
| giving me the |
\ source code! /
---------------
¯\_(ツ)_/¯
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 03/14] drm/panthor: Add the device logical block
2024-02-08 14:30 ` Liviu Dudau
@ 2024-02-08 15:14 ` Boris Brezillon
2024-02-08 15:55 ` Liviu Dudau
2024-02-08 15:19 ` Boris Brezillon
1 sibling, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-02-08 15:14 UTC (permalink / raw)
To: Liviu Dudau
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On Thu, 8 Feb 2024 14:30:02 +0000
Liviu Dudau <Liviu.Dudau@arm.com> wrote:
> > +int panthor_device_init(struct panthor_device *ptdev)
> > +{
> > + struct resource *res;
> > + struct page *p;
> > + int ret;
> > +
> > + ptdev->coherent = device_get_dma_attr(ptdev->base.dev) == DEV_DMA_COHERENT;
> > +
> > + init_completion(&ptdev->unplug.done);
> > + ret = drmm_mutex_init(&ptdev->base, &ptdev->unplug.lock);
> > + if (ret)
> > + return ret;
> > +
> > + ret = drmm_mutex_init(&ptdev->base, &ptdev->pm.mmio_lock);
> > + if (ret)
> > + return ret;
> > +
> > + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> > + p = alloc_page(GFP_KERNEL | __GFP_ZERO);
> > + if (!p)
> > + return -ENOMEM;
> > +
> > + ptdev->pm.dummy_latest_flush = page_address(p);
> > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_free_page,
> > + ptdev->pm.dummy_latest_flush);
> > + if (ret)
> > + return ret;
> > +
> > + /*
> > + * Set the dummy page holding the latest flush to 1. This will cause the
> > + * flush to avoided as we know it isn't necessary if the submission
> > + * happens while the dummy page is mapped. Zero cannot be used because
> > + * that means 'always flush'.
> > + */
> > + *ptdev->pm.dummy_latest_flush = 1;
> > +
> > + INIT_WORK(&ptdev->reset.work, panthor_device_reset_work);
> > + ptdev->reset.wq = alloc_ordered_workqueue("panthor-reset-wq", 0);
> > + if (!ptdev->reset.wq)
> > + return -ENOMEM;
> > +
> > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_reset_cleanup, NULL);
> > + if (ret)
> > + return ret;
> > +
> > + ret = panthor_clk_init(ptdev);
> > + if (ret)
> > + return ret;
> > +
> > + ret = panthor_devfreq_init(ptdev);
> > + if (ret)
> > + return ret;
> > +
> > + ptdev->iomem = devm_platform_get_and_ioremap_resource(to_platform_device(ptdev->base.dev),
> > + 0, &res);
> > + if (IS_ERR(ptdev->iomem))
> > + return PTR_ERR(ptdev->iomem);
> > +
> > + ptdev->phys_addr = res->start;
> > +
> > + ret = devm_pm_runtime_enable(ptdev->base.dev);
> > + if (ret)
> > + return ret;
> > +
> > + ret = pm_runtime_resume_and_get(ptdev->base.dev);
> > + if (ret)
> > + return ret;
> > +
> > + ret = panthor_gpu_init(ptdev);
> > + if (ret)
> > + goto err_rpm_put;
> > +
> > + ret = panthor_mmu_init(ptdev);
> > + if (ret)
> > + goto err_unplug_gpu;
> > +
> > + ret = panthor_fw_init(ptdev);
> > + if (ret)
> > + goto err_unplug_mmu;
> > +
> > + ret = panthor_sched_init(ptdev);
> > + if (ret)
> > + goto err_unplug_fw;
> > +
> > + /* ~3 frames */
> > + pm_runtime_set_autosuspend_delay(ptdev->base.dev, 50);
> > +
> > + ret = drm_dev_register(&ptdev->base, 0);
> > + if (ret)
> > + goto err_unplug_sched;
>
> For sake of replicating the panthor_device_unplus() calls, should we do
> here:
>
> if (ret) {
> pm_runtime_dont_use_autosuspend(ptdev->base.dev);
But pm_runtime_use_autosuspend() is called after that, why do we need
to call pm_runtime_dont_use_autosuspend() here?
> goto err_unplug_sched;
> }
>
>
> > +
> > + pm_runtime_use_autosuspend(ptdev->base.dev);
> > + pm_runtime_put_autosuspend(ptdev->base.dev);
> > + return 0;
> > +
> > +err_unplug_sched:
> > + panthor_sched_unplug(ptdev);
> > +
> > +err_unplug_fw:
> > + panthor_fw_unplug(ptdev);
> > +
> > +err_unplug_mmu:
> > + panthor_mmu_unplug(ptdev);
> > +
> > +err_unplug_gpu:
> > + panthor_gpu_unplug(ptdev);
> > +
> > +err_rpm_put:
> > + pm_runtime_put_sync_suspend(ptdev->base.dev);
> > + return ret;
> > +}
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 03/14] drm/panthor: Add the device logical block
2024-02-08 15:14 ` Boris Brezillon
@ 2024-02-08 15:55 ` Liviu Dudau
2024-02-08 16:00 ` Boris Brezillon
0 siblings, 1 reply; 51+ messages in thread
From: Liviu Dudau @ 2024-02-08 15:55 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On Thu, Feb 08, 2024 at 04:14:59PM +0100, Boris Brezillon wrote:
> On Thu, 8 Feb 2024 14:30:02 +0000
> Liviu Dudau <Liviu.Dudau@arm.com> wrote:
>
> > > +int panthor_device_init(struct panthor_device *ptdev)
> > > +{
> > > + struct resource *res;
> > > + struct page *p;
> > > + int ret;
> > > +
> > > + ptdev->coherent = device_get_dma_attr(ptdev->base.dev) == DEV_DMA_COHERENT;
> > > +
> > > + init_completion(&ptdev->unplug.done);
> > > + ret = drmm_mutex_init(&ptdev->base, &ptdev->unplug.lock);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + ret = drmm_mutex_init(&ptdev->base, &ptdev->pm.mmio_lock);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> > > + p = alloc_page(GFP_KERNEL | __GFP_ZERO);
> > > + if (!p)
> > > + return -ENOMEM;
> > > +
> > > + ptdev->pm.dummy_latest_flush = page_address(p);
> > > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_free_page,
> > > + ptdev->pm.dummy_latest_flush);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + /*
> > > + * Set the dummy page holding the latest flush to 1. This will cause the
> > > + * flush to avoided as we know it isn't necessary if the submission
> > > + * happens while the dummy page is mapped. Zero cannot be used because
> > > + * that means 'always flush'.
> > > + */
> > > + *ptdev->pm.dummy_latest_flush = 1;
> > > +
> > > + INIT_WORK(&ptdev->reset.work, panthor_device_reset_work);
> > > + ptdev->reset.wq = alloc_ordered_workqueue("panthor-reset-wq", 0);
> > > + if (!ptdev->reset.wq)
> > > + return -ENOMEM;
> > > +
> > > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_reset_cleanup, NULL);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + ret = panthor_clk_init(ptdev);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + ret = panthor_devfreq_init(ptdev);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + ptdev->iomem = devm_platform_get_and_ioremap_resource(to_platform_device(ptdev->base.dev),
> > > + 0, &res);
> > > + if (IS_ERR(ptdev->iomem))
> > > + return PTR_ERR(ptdev->iomem);
> > > +
> > > + ptdev->phys_addr = res->start;
> > > +
> > > + ret = devm_pm_runtime_enable(ptdev->base.dev);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + ret = pm_runtime_resume_and_get(ptdev->base.dev);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + ret = panthor_gpu_init(ptdev);
> > > + if (ret)
> > > + goto err_rpm_put;
> > > +
> > > + ret = panthor_mmu_init(ptdev);
> > > + if (ret)
> > > + goto err_unplug_gpu;
> > > +
> > > + ret = panthor_fw_init(ptdev);
> > > + if (ret)
> > > + goto err_unplug_mmu;
> > > +
> > > + ret = panthor_sched_init(ptdev);
> > > + if (ret)
> > > + goto err_unplug_fw;
> > > +
> > > + /* ~3 frames */
> > > + pm_runtime_set_autosuspend_delay(ptdev->base.dev, 50);
> > > +
> > > + ret = drm_dev_register(&ptdev->base, 0);
> > > + if (ret)
> > > + goto err_unplug_sched;
> >
> > For sake of replicating the panthor_device_unplus() calls, should we do
> > here:
> >
> > if (ret) {
> > pm_runtime_dont_use_autosuspend(ptdev->base.dev);
>
> But pm_runtime_use_autosuspend() is called after that, why do we need
> to call pm_runtime_dont_use_autosuspend() here?
This is in the case where ret != 0, so we're going to skip over
pm_runtime_use_autosuspend(). We've just called
pm_runtime_set_autosuspend_delay() which by my reading also enables
runtime PM when it calls update_autosuspend(), so this is needed.
Best regards,
Liviu
>
> > goto err_unplug_sched;
> > }
> >
> >
> > > +
> > > + pm_runtime_use_autosuspend(ptdev->base.dev);
> > > + pm_runtime_put_autosuspend(ptdev->base.dev);
> > > + return 0;
> > > +
> > > +err_unplug_sched:
> > > + panthor_sched_unplug(ptdev);
> > > +
> > > +err_unplug_fw:
> > > + panthor_fw_unplug(ptdev);
> > > +
> > > +err_unplug_mmu:
> > > + panthor_mmu_unplug(ptdev);
> > > +
> > > +err_unplug_gpu:
> > > + panthor_gpu_unplug(ptdev);
> > > +
> > > +err_rpm_put:
> > > + pm_runtime_put_sync_suspend(ptdev->base.dev);
> > > + return ret;
> > > +}
--
====================
| I would like to |
| fix the world, |
| but they're not |
| giving me the |
\ source code! /
---------------
¯\_(ツ)_/¯
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 03/14] drm/panthor: Add the device logical block
2024-02-08 15:55 ` Liviu Dudau
@ 2024-02-08 16:00 ` Boris Brezillon
2024-02-08 16:18 ` Liviu Dudau
0 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-02-08 16:00 UTC (permalink / raw)
To: Liviu Dudau
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On Thu, 8 Feb 2024 15:55:36 +0000
Liviu Dudau <Liviu.Dudau@arm.com> wrote:
> On Thu, Feb 08, 2024 at 04:14:59PM +0100, Boris Brezillon wrote:
> > On Thu, 8 Feb 2024 14:30:02 +0000
> > Liviu Dudau <Liviu.Dudau@arm.com> wrote:
> >
> > > > +int panthor_device_init(struct panthor_device *ptdev)
> > > > +{
> > > > + struct resource *res;
> > > > + struct page *p;
> > > > + int ret;
> > > > +
> > > > + ptdev->coherent = device_get_dma_attr(ptdev->base.dev) == DEV_DMA_COHERENT;
> > > > +
> > > > + init_completion(&ptdev->unplug.done);
> > > > + ret = drmm_mutex_init(&ptdev->base, &ptdev->unplug.lock);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + ret = drmm_mutex_init(&ptdev->base, &ptdev->pm.mmio_lock);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> > > > + p = alloc_page(GFP_KERNEL | __GFP_ZERO);
> > > > + if (!p)
> > > > + return -ENOMEM;
> > > > +
> > > > + ptdev->pm.dummy_latest_flush = page_address(p);
> > > > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_free_page,
> > > > + ptdev->pm.dummy_latest_flush);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + /*
> > > > + * Set the dummy page holding the latest flush to 1. This will cause the
> > > > + * flush to avoided as we know it isn't necessary if the submission
> > > > + * happens while the dummy page is mapped. Zero cannot be used because
> > > > + * that means 'always flush'.
> > > > + */
> > > > + *ptdev->pm.dummy_latest_flush = 1;
> > > > +
> > > > + INIT_WORK(&ptdev->reset.work, panthor_device_reset_work);
> > > > + ptdev->reset.wq = alloc_ordered_workqueue("panthor-reset-wq", 0);
> > > > + if (!ptdev->reset.wq)
> > > > + return -ENOMEM;
> > > > +
> > > > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_reset_cleanup, NULL);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + ret = panthor_clk_init(ptdev);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + ret = panthor_devfreq_init(ptdev);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + ptdev->iomem = devm_platform_get_and_ioremap_resource(to_platform_device(ptdev->base.dev),
> > > > + 0, &res);
> > > > + if (IS_ERR(ptdev->iomem))
> > > > + return PTR_ERR(ptdev->iomem);
> > > > +
> > > > + ptdev->phys_addr = res->start;
> > > > +
> > > > + ret = devm_pm_runtime_enable(ptdev->base.dev);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + ret = pm_runtime_resume_and_get(ptdev->base.dev);
> > > > + if (ret)
> > > > + return ret;
> > > > +
> > > > + ret = panthor_gpu_init(ptdev);
> > > > + if (ret)
> > > > + goto err_rpm_put;
> > > > +
> > > > + ret = panthor_mmu_init(ptdev);
> > > > + if (ret)
> > > > + goto err_unplug_gpu;
> > > > +
> > > > + ret = panthor_fw_init(ptdev);
> > > > + if (ret)
> > > > + goto err_unplug_mmu;
> > > > +
> > > > + ret = panthor_sched_init(ptdev);
> > > > + if (ret)
> > > > + goto err_unplug_fw;
> > > > +
> > > > + /* ~3 frames */
> > > > + pm_runtime_set_autosuspend_delay(ptdev->base.dev, 50);
> > > > +
> > > > + ret = drm_dev_register(&ptdev->base, 0);
> > > > + if (ret)
> > > > + goto err_unplug_sched;
> > >
> > > For sake of replicating the panthor_device_unplus() calls, should we do
> > > here:
> > >
> > > if (ret) {
> > > pm_runtime_dont_use_autosuspend(ptdev->base.dev);
> >
> > But pm_runtime_use_autosuspend() is called after that, why do we need
> > to call pm_runtime_dont_use_autosuspend() here?
>
> This is in the case where ret != 0, so we're going to skip over
> pm_runtime_use_autosuspend(). We've just called
> pm_runtime_set_autosuspend_delay() which by my reading also enables
> runtime PM when it calls update_autosuspend(), so this is needed.
That's not how I understand it. To me,
pm_runtime_set_autosuspend_delay() just updates the delay, but doesn't
change the autosuspend status, and update_autosuspend() doesn't seem to
change it either.
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 03/14] drm/panthor: Add the device logical block
2024-02-08 16:00 ` Boris Brezillon
@ 2024-02-08 16:18 ` Liviu Dudau
2024-02-08 16:54 ` Boris Brezillon
0 siblings, 1 reply; 51+ messages in thread
From: Liviu Dudau @ 2024-02-08 16:18 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On Thu, Feb 08, 2024 at 05:00:23PM +0100, Boris Brezillon wrote:
> On Thu, 8 Feb 2024 15:55:36 +0000
> Liviu Dudau <Liviu.Dudau@arm.com> wrote:
>
> > On Thu, Feb 08, 2024 at 04:14:59PM +0100, Boris Brezillon wrote:
> > > On Thu, 8 Feb 2024 14:30:02 +0000
> > > Liviu Dudau <Liviu.Dudau@arm.com> wrote:
> > >
> > > > > +int panthor_device_init(struct panthor_device *ptdev)
> > > > > +{
> > > > > + struct resource *res;
> > > > > + struct page *p;
> > > > > + int ret;
> > > > > +
> > > > > + ptdev->coherent = device_get_dma_attr(ptdev->base.dev) == DEV_DMA_COHERENT;
> > > > > +
> > > > > + init_completion(&ptdev->unplug.done);
> > > > > + ret = drmm_mutex_init(&ptdev->base, &ptdev->unplug.lock);
> > > > > + if (ret)
> > > > > + return ret;
> > > > > +
> > > > > + ret = drmm_mutex_init(&ptdev->base, &ptdev->pm.mmio_lock);
> > > > > + if (ret)
> > > > > + return ret;
> > > > > +
> > > > > + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> > > > > + p = alloc_page(GFP_KERNEL | __GFP_ZERO);
> > > > > + if (!p)
> > > > > + return -ENOMEM;
> > > > > +
> > > > > + ptdev->pm.dummy_latest_flush = page_address(p);
> > > > > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_free_page,
> > > > > + ptdev->pm.dummy_latest_flush);
> > > > > + if (ret)
> > > > > + return ret;
> > > > > +
> > > > > + /*
> > > > > + * Set the dummy page holding the latest flush to 1. This will cause the
> > > > > + * flush to avoided as we know it isn't necessary if the submission
> > > > > + * happens while the dummy page is mapped. Zero cannot be used because
> > > > > + * that means 'always flush'.
> > > > > + */
> > > > > + *ptdev->pm.dummy_latest_flush = 1;
> > > > > +
> > > > > + INIT_WORK(&ptdev->reset.work, panthor_device_reset_work);
> > > > > + ptdev->reset.wq = alloc_ordered_workqueue("panthor-reset-wq", 0);
> > > > > + if (!ptdev->reset.wq)
> > > > > + return -ENOMEM;
> > > > > +
> > > > > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_reset_cleanup, NULL);
> > > > > + if (ret)
> > > > > + return ret;
> > > > > +
> > > > > + ret = panthor_clk_init(ptdev);
> > > > > + if (ret)
> > > > > + return ret;
> > > > > +
> > > > > + ret = panthor_devfreq_init(ptdev);
> > > > > + if (ret)
> > > > > + return ret;
> > > > > +
> > > > > + ptdev->iomem = devm_platform_get_and_ioremap_resource(to_platform_device(ptdev->base.dev),
> > > > > + 0, &res);
> > > > > + if (IS_ERR(ptdev->iomem))
> > > > > + return PTR_ERR(ptdev->iomem);
> > > > > +
> > > > > + ptdev->phys_addr = res->start;
> > > > > +
> > > > > + ret = devm_pm_runtime_enable(ptdev->base.dev);
> > > > > + if (ret)
> > > > > + return ret;
> > > > > +
> > > > > + ret = pm_runtime_resume_and_get(ptdev->base.dev);
> > > > > + if (ret)
> > > > > + return ret;
> > > > > +
> > > > > + ret = panthor_gpu_init(ptdev);
> > > > > + if (ret)
> > > > > + goto err_rpm_put;
> > > > > +
> > > > > + ret = panthor_mmu_init(ptdev);
> > > > > + if (ret)
> > > > > + goto err_unplug_gpu;
> > > > > +
> > > > > + ret = panthor_fw_init(ptdev);
> > > > > + if (ret)
> > > > > + goto err_unplug_mmu;
> > > > > +
> > > > > + ret = panthor_sched_init(ptdev);
> > > > > + if (ret)
> > > > > + goto err_unplug_fw;
> > > > > +
> > > > > + /* ~3 frames */
> > > > > + pm_runtime_set_autosuspend_delay(ptdev->base.dev, 50);
> > > > > +
> > > > > + ret = drm_dev_register(&ptdev->base, 0);
> > > > > + if (ret)
> > > > > + goto err_unplug_sched;
> > > >
> > > > For sake of replicating the panthor_device_unplus() calls, should we do
> > > > here:
> > > >
> > > > if (ret) {
> > > > pm_runtime_dont_use_autosuspend(ptdev->base.dev);
> > >
> > > But pm_runtime_use_autosuspend() is called after that, why do we need
> > > to call pm_runtime_dont_use_autosuspend() here?
> >
> > This is in the case where ret != 0, so we're going to skip over
> > pm_runtime_use_autosuspend(). We've just called
> > pm_runtime_set_autosuspend_delay() which by my reading also enables
> > runtime PM when it calls update_autosuspend(), so this is needed.
>
> That's not how I understand it. To me,
> pm_runtime_set_autosuspend_delay() just updates the delay, but doesn't
> change the autosuspend status, and update_autosuspend() doesn't seem to
> change it either.
That depends on how you interpret the vague: "If it changes the other way
[read: not negative delay and (or?) power.use_autosuspend set], allow runtime
suspends".
The "else" branch in update_autosuspend() is taken when either delay >= 0 *or*
dev->power.use_autosuspend is zero. rpm_idle() will then be called regardless
of the values for old_delay and old_use.
Maybe we have uncovered a bug?
Best regards,
Liviu
--
====================
| I would like to |
| fix the world, |
| but they're not |
| giving me the |
\ source code! /
---------------
¯\_(ツ)_/¯
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 03/14] drm/panthor: Add the device logical block
2024-02-08 16:18 ` Liviu Dudau
@ 2024-02-08 16:54 ` Boris Brezillon
0 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-02-08 16:54 UTC (permalink / raw)
To: Liviu Dudau
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On Thu, 8 Feb 2024 16:18:48 +0000
Liviu Dudau <Liviu.Dudau@arm.com> wrote:
> On Thu, Feb 08, 2024 at 05:00:23PM +0100, Boris Brezillon wrote:
> > On Thu, 8 Feb 2024 15:55:36 +0000
> > Liviu Dudau <Liviu.Dudau@arm.com> wrote:
> >
> > > On Thu, Feb 08, 2024 at 04:14:59PM +0100, Boris Brezillon wrote:
> > > > On Thu, 8 Feb 2024 14:30:02 +0000
> > > > Liviu Dudau <Liviu.Dudau@arm.com> wrote:
> > > >
> > > > > > +int panthor_device_init(struct panthor_device *ptdev)
> > > > > > +{
> > > > > > + struct resource *res;
> > > > > > + struct page *p;
> > > > > > + int ret;
> > > > > > +
> > > > > > + ptdev->coherent = device_get_dma_attr(ptdev->base.dev) == DEV_DMA_COHERENT;
> > > > > > +
> > > > > > + init_completion(&ptdev->unplug.done);
> > > > > > + ret = drmm_mutex_init(&ptdev->base, &ptdev->unplug.lock);
> > > > > > + if (ret)
> > > > > > + return ret;
> > > > > > +
> > > > > > + ret = drmm_mutex_init(&ptdev->base, &ptdev->pm.mmio_lock);
> > > > > > + if (ret)
> > > > > > + return ret;
> > > > > > +
> > > > > > + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> > > > > > + p = alloc_page(GFP_KERNEL | __GFP_ZERO);
> > > > > > + if (!p)
> > > > > > + return -ENOMEM;
> > > > > > +
> > > > > > + ptdev->pm.dummy_latest_flush = page_address(p);
> > > > > > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_free_page,
> > > > > > + ptdev->pm.dummy_latest_flush);
> > > > > > + if (ret)
> > > > > > + return ret;
> > > > > > +
> > > > > > + /*
> > > > > > + * Set the dummy page holding the latest flush to 1. This will cause the
> > > > > > + * flush to avoided as we know it isn't necessary if the submission
> > > > > > + * happens while the dummy page is mapped. Zero cannot be used because
> > > > > > + * that means 'always flush'.
> > > > > > + */
> > > > > > + *ptdev->pm.dummy_latest_flush = 1;
> > > > > > +
> > > > > > + INIT_WORK(&ptdev->reset.work, panthor_device_reset_work);
> > > > > > + ptdev->reset.wq = alloc_ordered_workqueue("panthor-reset-wq", 0);
> > > > > > + if (!ptdev->reset.wq)
> > > > > > + return -ENOMEM;
> > > > > > +
> > > > > > + ret = drmm_add_action_or_reset(&ptdev->base, panthor_device_reset_cleanup, NULL);
> > > > > > + if (ret)
> > > > > > + return ret;
> > > > > > +
> > > > > > + ret = panthor_clk_init(ptdev);
> > > > > > + if (ret)
> > > > > > + return ret;
> > > > > > +
> > > > > > + ret = panthor_devfreq_init(ptdev);
> > > > > > + if (ret)
> > > > > > + return ret;
> > > > > > +
> > > > > > + ptdev->iomem = devm_platform_get_and_ioremap_resource(to_platform_device(ptdev->base.dev),
> > > > > > + 0, &res);
> > > > > > + if (IS_ERR(ptdev->iomem))
> > > > > > + return PTR_ERR(ptdev->iomem);
> > > > > > +
> > > > > > + ptdev->phys_addr = res->start;
> > > > > > +
> > > > > > + ret = devm_pm_runtime_enable(ptdev->base.dev);
> > > > > > + if (ret)
> > > > > > + return ret;
> > > > > > +
> > > > > > + ret = pm_runtime_resume_and_get(ptdev->base.dev);
> > > > > > + if (ret)
> > > > > > + return ret;
> > > > > > +
> > > > > > + ret = panthor_gpu_init(ptdev);
> > > > > > + if (ret)
> > > > > > + goto err_rpm_put;
> > > > > > +
> > > > > > + ret = panthor_mmu_init(ptdev);
> > > > > > + if (ret)
> > > > > > + goto err_unplug_gpu;
> > > > > > +
> > > > > > + ret = panthor_fw_init(ptdev);
> > > > > > + if (ret)
> > > > > > + goto err_unplug_mmu;
> > > > > > +
> > > > > > + ret = panthor_sched_init(ptdev);
> > > > > > + if (ret)
> > > > > > + goto err_unplug_fw;
> > > > > > +
> > > > > > + /* ~3 frames */
> > > > > > + pm_runtime_set_autosuspend_delay(ptdev->base.dev, 50);
> > > > > > +
> > > > > > + ret = drm_dev_register(&ptdev->base, 0);
> > > > > > + if (ret)
> > > > > > + goto err_unplug_sched;
> > > > >
> > > > > For sake of replicating the panthor_device_unplus() calls, should we do
> > > > > here:
> > > > >
> > > > > if (ret) {
> > > > > pm_runtime_dont_use_autosuspend(ptdev->base.dev);
> > > >
> > > > But pm_runtime_use_autosuspend() is called after that, why do we need
> > > > to call pm_runtime_dont_use_autosuspend() here?
> > >
> > > This is in the case where ret != 0, so we're going to skip over
> > > pm_runtime_use_autosuspend(). We've just called
> > > pm_runtime_set_autosuspend_delay() which by my reading also enables
> > > runtime PM when it calls update_autosuspend(), so this is needed.
> >
> > That's not how I understand it. To me,
> > pm_runtime_set_autosuspend_delay() just updates the delay, but doesn't
> > change the autosuspend status, and update_autosuspend() doesn't seem to
> > change it either.
>
> That depends on how you interpret the vague: "If it changes the other way
> [read: not negative delay and (or?) power.use_autosuspend set], allow runtime
> suspends".
>
> The "else" branch in update_autosuspend() is taken when either delay >= 0 *or*
> dev->power.use_autosuspend is zero. rpm_idle() will then be called regardless
> of the values for old_delay and old_use.
>
> Maybe we have uncovered a bug?
Nah, you're right, but that's confusing. I thought by moving
pm_runtime_set_autosuspend_delay() before the drm_dev_register() we
wouldn't change the autosuspend state, and at the same time preset a
default delay that can be changed by userland as soon as the device is
exposed (when drm_dev_register() is called). If we can't do that,
there's no point splitting the set_autosuspend_delay() and
use_autosuspend() calls, we can just move them before
drm_dev_register() and call dont_use_autosuspend() is something fails
(as you suggested).
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 03/14] drm/panthor: Add the device logical block
2024-02-08 14:30 ` Liviu Dudau
2024-02-08 15:14 ` Boris Brezillon
@ 2024-02-08 15:19 ` Boris Brezillon
1 sibling, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-02-08 15:19 UTC (permalink / raw)
To: Liviu Dudau
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On Thu, 8 Feb 2024 14:30:02 +0000
Liviu Dudau <Liviu.Dudau@arm.com> wrote:
> > +#ifdef CONFIG_PM
> > +int panthor_device_resume(struct device *dev)
> > +{
> > + struct panthor_device *ptdev = dev_get_drvdata(dev);
> > + int ret, cookie;
> > +
> > + if (atomic_read(&ptdev->pm.state) != PANTHOR_DEVICE_PM_STATE_SUSPENDED)
> > + return -EINVAL;
> > +
> > + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_RESUMING);
> > +
> > + ret = clk_prepare_enable(ptdev->clks.core);
> > + if (ret)
> > + goto err_set_suspended;
> > +
> > + ret = clk_prepare_enable(ptdev->clks.stacks);
> > + if (ret)
> > + goto err_disable_core_clk;
> > +
> > + ret = clk_prepare_enable(ptdev->clks.coregroup);
> > + if (ret)
> > + goto err_disable_stacks_clk;
> > +
> > + ret = panthor_devfreq_resume(ptdev);
> > + if (ret)
> > + goto err_disable_coregroup_clk;
> > +
> > + if (panthor_device_is_initialized(ptdev) &&
> > + drm_dev_enter(&ptdev->base, &cookie)) {
> > + panthor_gpu_resume(ptdev);
> > + panthor_mmu_resume(ptdev);
> > + ret = drm_WARN_ON(&ptdev->base, panthor_fw_resume(ptdev));
> > + if (!ret)
> > + panthor_sched_resume(ptdev);
> > +
> > + drm_dev_exit(cookie);
> > +
> > + if (ret)
> > + goto err_devfreq_suspend;
> > + }
> > +
> > + if (atomic_read(&ptdev->reset.pending))
> > + queue_work(ptdev->reset.wq, &ptdev->reset.work);
> > +
> > + /* Clear all IOMEM mappings pointing to this device after we've
> > + * resumed. This way the fake mappings pointing to the dummy pages
> > + * are removed and the real iomem mapping will be restored on next
> > + * access.
> > + */
> > + mutex_lock(&ptdev->pm.mmio_lock);
> > + unmap_mapping_range(ptdev->base.anon_inode->i_mapping,
> > + DRM_PANTHOR_USER_MMIO_OFFSET, 0, 1);
> > + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_ACTIVE);
> > + mutex_unlock(&ptdev->pm.mmio_lock);
> > + return 0;
> > +
> > +err_devfreq_suspend:
> > + panthor_devfreq_suspend(ptdev);
>
> Should we also call panthor_mmu_suspend(ptdev) and panthor_gpu_suspend(ptdev)
> before panthor_devfreq_suspend()? Otherwise those blocks will be left in an
> inconsistent state.
Yep, it's been fixed in panthor-next (see [1]). Note that I didn't add
it to the error path, to keep the panthor_{mmu,gpu}_suspend() under the
drm_dev_enter/exit() section.
> > +
> > +err_disable_coregroup_clk:
> > + clk_disable_unprepare(ptdev->clks.coregroup);
> > +
> > +err_disable_stacks_clk:
> > + clk_disable_unprepare(ptdev->clks.stacks);
> > +
> > +err_disable_core_clk:
> > + clk_disable_unprepare(ptdev->clks.core);
> > +
> > +err_set_suspended:
> > + atomic_set(&ptdev->pm.state, PANTHOR_DEVICE_PM_STATE_SUSPENDED);
> > + return ret;
> > +}
[1]https://gitlab.freedesktop.org/panfrost/linux/-/commit/d9eccd669206ffd9383b955bffba93426ebea40a
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 04/14] drm/panthor: Add the GPU logical block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (2 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 03/14] drm/panthor: Add the device logical block Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-02-08 16:09 ` Steven Price
2024-01-22 16:30 ` [PATCH v4 05/14] drm/panthor: Add GEM " Boris Brezillon
` (13 subsequent siblings)
17 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Alexey Sheplyakov, Boris Brezillon, Clément Péron,
Daniel Vetter, Chris Diamand, Grant Likely, Marty E . Plummer,
Robin Murphy, Faith Ekstrand
Handles everything that's not related to the FW, the MMU or the
scheduler. This is the block dealing with the GPU property retrieval,
the GPU block power on/off logic, and some global operations, like
global cache flushing.
v4:
- Expose CORE_FEATURES through DEV_QUERY
v3:
- Add acks for the MIT/GPL2 relicensing
- Use macros to extract GPU ID info
- Make sure we reset clear pending_reqs bits when wait_event_timeout()
times out but the corresponding bit is cleared in GPU_INT_RAWSTAT
(can happen if the IRQ is masked or HW takes to long to call the IRQ
handler)
- GPU_MODEL now takes separate arch and product majors to be more
readable.
- Drop GPU_IRQ_MCU_STATUS_CHANGED from interrupt mask.
- Handle GPU_IRQ_PROTM_FAULT correctly (don't output registers that are
not updated for protected interrupts).
- Minor code tidy ups
Cc: Alexey Sheplyakov <asheplyakov@basealt.ru> # MIT+GPL2 relicensing
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
---
drivers/gpu/drm/panthor/panthor_gpu.c | 482 ++++++++++++++++++++++++++
drivers/gpu/drm/panthor/panthor_gpu.h | 52 +++
2 files changed, 534 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.c
create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.h
diff --git a/drivers/gpu/drm/panthor/panthor_gpu.c b/drivers/gpu/drm/panthor/panthor_gpu.c
new file mode 100644
index 000000000000..014e0f7c1b13
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_gpu.c
@@ -0,0 +1,482 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
+/* Copyright 2019 Linaro, Ltd., Rob Herring <robh@kernel.org> */
+/* Copyright 2019 Collabora ltd. */
+
+#include <linux/bitfield.h>
+#include <linux/bitmap.h>
+#include <linux/delay.h>
+#include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/iopoll.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+
+#include <drm/drm_drv.h>
+#include <drm/drm_managed.h>
+
+#include "panthor_device.h"
+#include "panthor_gpu.h"
+#include "panthor_regs.h"
+
+/**
+ * struct panthor_gpu - GPU block management data.
+ */
+struct panthor_gpu {
+ /** @irq: GPU irq. */
+ struct panthor_irq irq;
+
+ /** @reqs_lock: Lock protecting access to pending_reqs. */
+ spinlock_t reqs_lock;
+
+ /** @pending_reqs: Pending GPU requests. */
+ u32 pending_reqs;
+
+ /** @reqs_acked: GPU request wait queue. */
+ wait_queue_head_t reqs_acked;
+};
+
+/**
+ * struct panthor_model - GPU model description
+ */
+struct panthor_model {
+ /** @name: Model name. */
+ const char *name;
+
+ /** @arch_major: Major version number of architecture. */
+ u8 arch_major;
+
+ /** @product_major: Major version number of product. */
+ u8 product_major;
+};
+
+/**
+ * GPU_MODEL() - Define a GPU model. A GPU product can be uniquely identified
+ * by a combination of the major architecture version and the major product
+ * version.
+ * @name: Name for the GPU model.
+ * @_arch_major: Architecture major.
+ * @_product_major: Product major.
+ */
+#define GPU_MODEL(_name, _arch_major, _product_major) \
+{\
+ .name = __stringify(_name), \
+ .arch_major = _arch_major, \
+ .product_major = _product_major, \
+}
+
+static const struct panthor_model gpu_models[] = {
+ GPU_MODEL(g610, 10, 7),
+ {},
+};
+
+#define GPU_INTERRUPTS_MASK \
+ (GPU_IRQ_FAULT | \
+ GPU_IRQ_PROTM_FAULT | \
+ GPU_IRQ_RESET_COMPLETED | \
+ GPU_IRQ_CLEAN_CACHES_COMPLETED)
+
+static void panthor_gpu_init_info(struct panthor_device *ptdev)
+{
+ const struct panthor_model *model;
+ u32 arch_major, product_major;
+ u32 major, minor, status;
+ unsigned int i;
+
+ ptdev->gpu_info.gpu_id = gpu_read(ptdev, GPU_ID);
+ ptdev->gpu_info.csf_id = gpu_read(ptdev, GPU_CSF_ID);
+ ptdev->gpu_info.gpu_rev = gpu_read(ptdev, GPU_REVID);
+ ptdev->gpu_info.core_features = gpu_read(ptdev, GPU_CORE_FEATURES);
+ ptdev->gpu_info.l2_features = gpu_read(ptdev, GPU_L2_FEATURES);
+ ptdev->gpu_info.tiler_features = gpu_read(ptdev, GPU_TILER_FEATURES);
+ ptdev->gpu_info.mem_features = gpu_read(ptdev, GPU_MEM_FEATURES);
+ ptdev->gpu_info.mmu_features = gpu_read(ptdev, GPU_MMU_FEATURES);
+ ptdev->gpu_info.thread_features = gpu_read(ptdev, GPU_THREAD_FEATURES);
+ ptdev->gpu_info.max_threads = gpu_read(ptdev, GPU_THREAD_MAX_THREADS);
+ ptdev->gpu_info.thread_max_workgroup_size = gpu_read(ptdev, GPU_THREAD_MAX_WORKGROUP_SIZE);
+ ptdev->gpu_info.thread_max_barrier_size = gpu_read(ptdev, GPU_THREAD_MAX_BARRIER_SIZE);
+ ptdev->gpu_info.coherency_features = gpu_read(ptdev, GPU_COHERENCY_FEATURES);
+ for (i = 0; i < 4; i++)
+ ptdev->gpu_info.texture_features[i] = gpu_read(ptdev, GPU_TEXTURE_FEATURES(i));
+
+ ptdev->gpu_info.as_present = gpu_read(ptdev, GPU_AS_PRESENT);
+
+ ptdev->gpu_info.shader_present = gpu_read(ptdev, GPU_SHADER_PRESENT_LO);
+ ptdev->gpu_info.shader_present |= (u64)gpu_read(ptdev, GPU_SHADER_PRESENT_HI) << 32;
+
+ ptdev->gpu_info.tiler_present = gpu_read(ptdev, GPU_TILER_PRESENT_LO);
+ ptdev->gpu_info.tiler_present |= (u64)gpu_read(ptdev, GPU_TILER_PRESENT_HI) << 32;
+
+ ptdev->gpu_info.l2_present = gpu_read(ptdev, GPU_L2_PRESENT_LO);
+ ptdev->gpu_info.l2_present |= (u64)gpu_read(ptdev, GPU_L2_PRESENT_HI) << 32;
+
+ arch_major = GPU_ARCH_MAJOR(ptdev->gpu_info.gpu_id);
+ product_major = GPU_PROD_MAJOR(ptdev->gpu_info.gpu_id);
+ major = GPU_VER_MAJOR(ptdev->gpu_info.gpu_id);
+ minor = GPU_VER_MINOR(ptdev->gpu_info.gpu_id);
+ status = GPU_VER_STATUS(ptdev->gpu_info.gpu_id);
+
+ for (model = gpu_models; model->name; model++) {
+ if (model->arch_major == arch_major &&
+ model->product_major == product_major)
+ break;
+ }
+
+ drm_info(&ptdev->base,
+ "mali-%s id 0x%x major 0x%x minor 0x%x status 0x%x",
+ model->name ?: "unknown", ptdev->gpu_info.gpu_id >> 16,
+ major, minor, status);
+
+ drm_info(&ptdev->base,
+ "Features: L2:%#x Tiler:%#x Mem:%#x MMU:%#x AS:%#x",
+ ptdev->gpu_info.l2_features,
+ ptdev->gpu_info.tiler_features,
+ ptdev->gpu_info.mem_features,
+ ptdev->gpu_info.mmu_features,
+ ptdev->gpu_info.as_present);
+
+ drm_info(&ptdev->base,
+ "shader_present=0x%0llx l2_present=0x%0llx tiler_present=0x%0llx",
+ ptdev->gpu_info.shader_present, ptdev->gpu_info.l2_present,
+ ptdev->gpu_info.tiler_present);
+}
+
+static void panthor_gpu_irq_handler(struct panthor_device *ptdev, u32 status)
+{
+ if (status & GPU_IRQ_FAULT) {
+ u32 fault_status = gpu_read(ptdev, GPU_FAULT_STATUS);
+ u64 address = ((u64)gpu_read(ptdev, GPU_FAULT_ADDR_HI) << 32) |
+ gpu_read(ptdev, GPU_FAULT_ADDR_LO);
+
+ drm_warn(&ptdev->base, "GPU Fault 0x%08x (%s) at 0x%016llx\n",
+ fault_status, panthor_exception_name(ptdev, fault_status & 0xFF),
+ address);
+ }
+ if (status & GPU_IRQ_PROTM_FAULT)
+ drm_warn(&ptdev->base, "GPU Fault in protected mode\n");
+
+ spin_lock(&ptdev->gpu->reqs_lock);
+ if (status & ptdev->gpu->pending_reqs) {
+ ptdev->gpu->pending_reqs &= ~status;
+ wake_up_all(&ptdev->gpu->reqs_acked);
+ }
+ spin_unlock(&ptdev->gpu->reqs_lock);
+}
+PANTHOR_IRQ_HANDLER(gpu, GPU, panthor_gpu_irq_handler);
+
+/**
+ * panthor_gpu_unplug() - Called when the GPU is unplugged.
+ * @ptdev: Device to unplug.
+ */
+void panthor_gpu_unplug(struct panthor_device *ptdev)
+{
+ unsigned long flags;
+
+ /* Make sure the IRQ handler is not running after that point. */
+ panthor_gpu_irq_suspend(&ptdev->gpu->irq);
+
+ /* Wake-up all waiters. */
+ spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
+ ptdev->gpu->pending_reqs = 0;
+ wake_up_all(&ptdev->gpu->reqs_acked);
+ spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
+}
+
+/**
+ * panthor_gpu_init() - Initialize the GPU block
+ * @ptdev: Device.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_gpu_init(struct panthor_device *ptdev)
+{
+ struct panthor_gpu *gpu;
+ u32 pa_bits;
+ int ret, irq;
+
+ gpu = drmm_kzalloc(&ptdev->base, sizeof(*gpu), GFP_KERNEL);
+ if (!gpu)
+ return -ENOMEM;
+
+ spin_lock_init(&gpu->reqs_lock);
+ init_waitqueue_head(&gpu->reqs_acked);
+ ptdev->gpu = gpu;
+ panthor_gpu_init_info(ptdev);
+
+ dma_set_max_seg_size(ptdev->base.dev, UINT_MAX);
+ pa_bits = GPU_MMU_FEATURES_PA_BITS(ptdev->gpu_info.mmu_features);
+ ret = dma_set_mask_and_coherent(ptdev->base.dev, DMA_BIT_MASK(pa_bits));
+ if (ret)
+ return ret;
+
+ irq = platform_get_irq_byname(to_platform_device(ptdev->base.dev), "gpu");
+ if (irq <= 0)
+ return ret;
+
+ ret = panthor_request_gpu_irq(ptdev, &ptdev->gpu->irq, irq, GPU_INTERRUPTS_MASK);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
+/**
+ * panthor_gpu_block_power_off() - Power-off a specific block of the GPU
+ * @ptdev: Device.
+ * @blk_name: Block name.
+ * @pwroff_reg: Power-off register for this block.
+ * @pwrtrans_reg: Power transition register for this block.
+ * @mask: Sub-elements to power-off.
+ * @timeout_us: Timeout in microseconds.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_gpu_block_power_off(struct panthor_device *ptdev,
+ const char *blk_name,
+ u32 pwroff_reg, u32 pwrtrans_reg,
+ u64 mask, u32 timeout_us)
+{
+ u32 val, i;
+ int ret;
+
+ for (i = 0; i < 2; i++) {
+ u32 mask32 = mask >> (i * 32);
+
+ if (!mask32)
+ continue;
+
+ ret = readl_relaxed_poll_timeout(ptdev->iomem + pwrtrans_reg + (i * 4),
+ val, !(mask32 & val),
+ 100, timeout_us);
+ if (ret) {
+ drm_err(&ptdev->base, "timeout waiting on %s:%llx power transition",
+ blk_name, mask);
+ return ret;
+ }
+ }
+
+ if (mask & GENMASK(31, 0))
+ gpu_write(ptdev, pwroff_reg, mask);
+
+ if (mask >> 32)
+ gpu_write(ptdev, pwroff_reg + 4, mask >> 32);
+
+ for (i = 0; i < 2; i++) {
+ u32 mask32 = mask >> (i * 32);
+
+ if (!mask32)
+ continue;
+
+ ret = readl_relaxed_poll_timeout(ptdev->iomem + pwrtrans_reg + (i * 4),
+ val, !(mask & val),
+ 100, timeout_us);
+ if (ret) {
+ drm_err(&ptdev->base, "timeout waiting on %s:%llx power transition",
+ blk_name, mask);
+ return ret;
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_gpu_block_power_on() - Power-on a specific block of the GPU
+ * @ptdev: Device.
+ * @blk_name: Block name.
+ * @pwron_reg: Power-on register for this block.
+ * @pwrtrans_reg: Power transition register for this block.
+ * @rdy_reg: Power transition ready register.
+ * @mask: Sub-elements to power-on.
+ * @timeout_us: Timeout in microseconds.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_gpu_block_power_on(struct panthor_device *ptdev,
+ const char *blk_name,
+ u32 pwron_reg, u32 pwrtrans_reg,
+ u32 rdy_reg, u64 mask, u32 timeout_us)
+{
+ u32 val, i;
+ int ret;
+
+ for (i = 0; i < 2; i++) {
+ u32 mask32 = mask >> (i * 32);
+
+ if (!mask32)
+ continue;
+
+ ret = readl_relaxed_poll_timeout(ptdev->iomem + pwrtrans_reg + (i * 4),
+ val, !(mask32 & val),
+ 100, timeout_us);
+ if (ret) {
+ drm_err(&ptdev->base, "timeout waiting on %s:%llx power transition",
+ blk_name, mask);
+ return ret;
+ }
+ }
+
+ if (mask & GENMASK(31, 0))
+ gpu_write(ptdev, pwron_reg, mask);
+
+ if (mask >> 32)
+ gpu_write(ptdev, pwron_reg + 4, mask >> 32);
+
+ for (i = 0; i < 2; i++) {
+ u32 mask32 = mask >> (i * 32);
+
+ if (!mask32)
+ continue;
+
+ ret = readl_relaxed_poll_timeout(ptdev->iomem + rdy_reg + (i * 4),
+ val, (mask32 & val) == mask32,
+ 100, timeout_us);
+ if (ret) {
+ drm_err(&ptdev->base, "timeout waiting on %s:%llx readyness",
+ blk_name, mask);
+ return ret;
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_gpu_l2_power_on() - Power-on the L2-cache
+ * @ptdev: Device.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_gpu_l2_power_on(struct panthor_device *ptdev)
+{
+ if (ptdev->gpu_info.l2_present != 1) {
+ /*
+ * Only support one core group now.
+ * ~(l2_present - 1) unsets all bits in l2_present except
+ * the bottom bit. (l2_present - 2) has all the bits in
+ * the first core group set. AND them together to generate
+ * a mask of cores in the first core group.
+ */
+ u64 core_mask = ~(ptdev->gpu_info.l2_present - 1) &
+ (ptdev->gpu_info.l2_present - 2);
+ drm_info_once(&ptdev->base, "using only 1st core group (%lu cores from %lu)\n",
+ hweight64(core_mask),
+ hweight64(ptdev->gpu_info.shader_present));
+ }
+
+ return panthor_gpu_power_on(ptdev, L2, 1, 20000);
+}
+
+/**
+ * panthor_gpu_flush_caches() - Flush caches
+ * @ptdev: Device.
+ * @l2: L2 flush type.
+ * @lsc: LSC flush type.
+ * @other: Other flush type.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_gpu_flush_caches(struct panthor_device *ptdev,
+ u32 l2, u32 lsc, u32 other)
+{
+ bool timedout = false;
+ unsigned long flags;
+
+ spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
+ if (!drm_WARN_ON(&ptdev->base,
+ ptdev->gpu->pending_reqs & GPU_IRQ_CLEAN_CACHES_COMPLETED)) {
+ ptdev->gpu->pending_reqs |= GPU_IRQ_CLEAN_CACHES_COMPLETED;
+ gpu_write(ptdev, GPU_CMD, GPU_FLUSH_CACHES(l2, lsc, other));
+ }
+ spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
+
+ if (!wait_event_timeout(ptdev->gpu->reqs_acked,
+ !(ptdev->gpu->pending_reqs & GPU_IRQ_CLEAN_CACHES_COMPLETED),
+ msecs_to_jiffies(100))) {
+ spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
+ if ((ptdev->gpu->pending_reqs & GPU_IRQ_CLEAN_CACHES_COMPLETED) != 0 &&
+ !(gpu_read(ptdev, GPU_INT_RAWSTAT) & GPU_IRQ_CLEAN_CACHES_COMPLETED))
+ timedout = true;
+ else
+ ptdev->gpu->pending_reqs &= ~GPU_IRQ_CLEAN_CACHES_COMPLETED;
+ spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
+ }
+
+ if (timedout) {
+ drm_err(&ptdev->base, "Flush caches timeout");
+ return -ETIMEDOUT;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_gpu_soft_reset() - Issue a soft-reset
+ * @ptdev: Device.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_gpu_soft_reset(struct panthor_device *ptdev)
+{
+ bool timedout = false;
+ unsigned long flags;
+
+ spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
+ if (!drm_WARN_ON(&ptdev->base,
+ ptdev->gpu->pending_reqs & GPU_IRQ_RESET_COMPLETED)) {
+ ptdev->gpu->pending_reqs |= GPU_IRQ_RESET_COMPLETED;
+ gpu_write(ptdev, GPU_INT_CLEAR, GPU_IRQ_RESET_COMPLETED);
+ gpu_write(ptdev, GPU_CMD, GPU_SOFT_RESET);
+ }
+ spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
+
+ if (!wait_event_timeout(ptdev->gpu->reqs_acked,
+ !(ptdev->gpu->pending_reqs & GPU_IRQ_RESET_COMPLETED),
+ msecs_to_jiffies(100))) {
+ spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
+ if ((ptdev->gpu->pending_reqs & GPU_IRQ_RESET_COMPLETED) != 0 &&
+ !(gpu_read(ptdev, GPU_INT_RAWSTAT) & GPU_IRQ_RESET_COMPLETED))
+ timedout = true;
+ else
+ ptdev->gpu->pending_reqs &= ~GPU_IRQ_RESET_COMPLETED;
+ spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
+ }
+
+ if (timedout) {
+ drm_err(&ptdev->base, "Soft reset timeout");
+ return -ETIMEDOUT;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_gpu_suspend() - Suspend the GPU block.
+ * @ptdev: Device.
+ *
+ * Suspend the GPU irq. This should be called last in the suspend procedure,
+ * after all other blocks have been suspented.
+ */
+void panthor_gpu_suspend(struct panthor_device *ptdev)
+{
+ /*
+ * It may be preferable to simply power down the L2, but for now just
+ * soft-reset which will leave the L2 powered down.
+ */
+ panthor_gpu_soft_reset(ptdev);
+ panthor_gpu_irq_suspend(&ptdev->gpu->irq);
+}
+
+/**
+ * panthor_gpu_resume() - Resume the GPU block.
+ * @ptdev: Device.
+ *
+ * Resume the IRQ handler and power-on the L2-cache.
+ * The FW takes care of powering the other blocks.
+ */
+void panthor_gpu_resume(struct panthor_device *ptdev)
+{
+ panthor_gpu_irq_resume(&ptdev->gpu->irq, GPU_INTERRUPTS_MASK);
+ panthor_gpu_l2_power_on(ptdev);
+}
diff --git a/drivers/gpu/drm/panthor/panthor_gpu.h b/drivers/gpu/drm/panthor/panthor_gpu.h
new file mode 100644
index 000000000000..bba7555dd3c6
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_gpu.h
@@ -0,0 +1,52 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
+/* Copyright 2019 Collabora ltd. */
+
+#ifndef __PANTHOR_GPU_H__
+#define __PANTHOR_GPU_H__
+
+struct panthor_device;
+
+int panthor_gpu_init(struct panthor_device *ptdev);
+void panthor_gpu_unplug(struct panthor_device *ptdev);
+void panthor_gpu_suspend(struct panthor_device *ptdev);
+void panthor_gpu_resume(struct panthor_device *ptdev);
+
+int panthor_gpu_block_power_on(struct panthor_device *ptdev,
+ const char *blk_name,
+ u32 pwron_reg, u32 pwrtrans_reg,
+ u32 rdy_reg, u64 mask, u32 timeout_us);
+int panthor_gpu_block_power_off(struct panthor_device *ptdev,
+ const char *blk_name,
+ u32 pwroff_reg, u32 pwrtrans_reg,
+ u64 mask, u32 timeout_us);
+
+/**
+ * panthor_gpu_power_on() - Power on the GPU block.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+#define panthor_gpu_power_on(ptdev, type, mask, timeout_us) \
+ panthor_gpu_block_power_on(ptdev, #type, \
+ type ## _PWRON_LO, \
+ type ## _PWRTRANS_LO, \
+ type ## _READY_LO, \
+ mask, timeout_us)
+
+/**
+ * panthor_gpu_power_off() - Power off the GPU block.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+#define panthor_gpu_power_off(ptdev, type, mask, timeout_us) \
+ panthor_gpu_block_power_off(ptdev, #type, \
+ type ## _PWROFF_LO, \
+ type ## _PWRTRANS_LO, \
+ mask, timeout_us)
+
+int panthor_gpu_l2_power_on(struct panthor_device *ptdev);
+int panthor_gpu_flush_caches(struct panthor_device *ptdev,
+ u32 l2, u32 lsc, u32 other);
+int panthor_gpu_soft_reset(struct panthor_device *ptdev);
+
+#endif
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 04/14] drm/panthor: Add the GPU logical block
2024-01-22 16:30 ` [PATCH v4 04/14] drm/panthor: Add the GPU " Boris Brezillon
@ 2024-02-08 16:09 ` Steven Price
0 siblings, 0 replies; 51+ messages in thread
From: Steven Price @ 2024-02-08 16:09 UTC (permalink / raw)
To: Boris Brezillon, dri-devel
Cc: Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Alexey Sheplyakov, Grant Likely
On 22/01/2024 16:30, Boris Brezillon wrote:
> Handles everything that's not related to the FW, the MMU or the
> scheduler. This is the block dealing with the GPU property retrieval,
> the GPU block power on/off logic, and some global operations, like
> global cache flushing.
>
> v4:
> - Expose CORE_FEATURES through DEV_QUERY
>
> v3:
> - Add acks for the MIT/GPL2 relicensing
> - Use macros to extract GPU ID info
> - Make sure we reset clear pending_reqs bits when wait_event_timeout()
> times out but the corresponding bit is cleared in GPU_INT_RAWSTAT
> (can happen if the IRQ is masked or HW takes to long to call the IRQ
> handler)
> - GPU_MODEL now takes separate arch and product majors to be more
> readable.
> - Drop GPU_IRQ_MCU_STATUS_CHANGED from interrupt mask.
> - Handle GPU_IRQ_PROTM_FAULT correctly (don't output registers that are
> not updated for protected interrupts).
> - Minor code tidy ups
>
> Cc: Alexey Sheplyakov <asheplyakov@basealt.ru> # MIT+GPL2 relicensing
> Co-developed-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
> Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
> Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
Reviewed-by: Steven Price <steven.price@arm.com>
But two minor nits below.
> ---
> drivers/gpu/drm/panthor/panthor_gpu.c | 482 ++++++++++++++++++++++++++
> drivers/gpu/drm/panthor/panthor_gpu.h | 52 +++
> 2 files changed, 534 insertions(+)
> create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.h
>
> diff --git a/drivers/gpu/drm/panthor/panthor_gpu.c b/drivers/gpu/drm/panthor/panthor_gpu.c
> new file mode 100644
> index 000000000000..014e0f7c1b13
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_gpu.c
> @@ -0,0 +1,482 @@
> +// SPDX-License-Identifier: GPL-2.0 or MIT
> +/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
> +/* Copyright 2019 Linaro, Ltd., Rob Herring <robh@kernel.org> */
> +/* Copyright 2019 Collabora ltd. */
> +
> +#include <linux/bitfield.h>
> +#include <linux/bitmap.h>
> +#include <linux/delay.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/interrupt.h>
> +#include <linux/io.h>
> +#include <linux/iopoll.h>
> +#include <linux/platform_device.h>
> +#include <linux/pm_runtime.h>
> +
> +#include <drm/drm_drv.h>
> +#include <drm/drm_managed.h>
> +
> +#include "panthor_device.h"
> +#include "panthor_gpu.h"
> +#include "panthor_regs.h"
> +
> +/**
> + * struct panthor_gpu - GPU block management data.
> + */
> +struct panthor_gpu {
> + /** @irq: GPU irq. */
> + struct panthor_irq irq;
> +
> + /** @reqs_lock: Lock protecting access to pending_reqs. */
> + spinlock_t reqs_lock;
> +
> + /** @pending_reqs: Pending GPU requests. */
> + u32 pending_reqs;
> +
> + /** @reqs_acked: GPU request wait queue. */
> + wait_queue_head_t reqs_acked;
> +};
> +
> +/**
> + * struct panthor_model - GPU model description
> + */
> +struct panthor_model {
> + /** @name: Model name. */
> + const char *name;
> +
> + /** @arch_major: Major version number of architecture. */
> + u8 arch_major;
> +
> + /** @product_major: Major version number of product. */
> + u8 product_major;
> +};
> +
> +/**
> + * GPU_MODEL() - Define a GPU model. A GPU product can be uniquely identified
> + * by a combination of the major architecture version and the major product
> + * version.
> + * @name: Name for the GPU model.
NIT: s/@name/@_name/
> + * @_arch_major: Architecture major.
> + * @_product_major: Product major.
> + */
> +#define GPU_MODEL(_name, _arch_major, _product_major) \
> +{\
> + .name = __stringify(_name), \
> + .arch_major = _arch_major, \
> + .product_major = _product_major, \
> +}
> +
> +static const struct panthor_model gpu_models[] = {
> + GPU_MODEL(g610, 10, 7),
> + {},
> +};
> +
> +#define GPU_INTERRUPTS_MASK \
> + (GPU_IRQ_FAULT | \
> + GPU_IRQ_PROTM_FAULT | \
> + GPU_IRQ_RESET_COMPLETED | \
> + GPU_IRQ_CLEAN_CACHES_COMPLETED)
> +
> +static void panthor_gpu_init_info(struct panthor_device *ptdev)
> +{
> + const struct panthor_model *model;
> + u32 arch_major, product_major;
> + u32 major, minor, status;
> + unsigned int i;
> +
> + ptdev->gpu_info.gpu_id = gpu_read(ptdev, GPU_ID);
> + ptdev->gpu_info.csf_id = gpu_read(ptdev, GPU_CSF_ID);
> + ptdev->gpu_info.gpu_rev = gpu_read(ptdev, GPU_REVID);
> + ptdev->gpu_info.core_features = gpu_read(ptdev, GPU_CORE_FEATURES);
> + ptdev->gpu_info.l2_features = gpu_read(ptdev, GPU_L2_FEATURES);
> + ptdev->gpu_info.tiler_features = gpu_read(ptdev, GPU_TILER_FEATURES);
> + ptdev->gpu_info.mem_features = gpu_read(ptdev, GPU_MEM_FEATURES);
> + ptdev->gpu_info.mmu_features = gpu_read(ptdev, GPU_MMU_FEATURES);
> + ptdev->gpu_info.thread_features = gpu_read(ptdev, GPU_THREAD_FEATURES);
> + ptdev->gpu_info.max_threads = gpu_read(ptdev, GPU_THREAD_MAX_THREADS);
> + ptdev->gpu_info.thread_max_workgroup_size = gpu_read(ptdev, GPU_THREAD_MAX_WORKGROUP_SIZE);
> + ptdev->gpu_info.thread_max_barrier_size = gpu_read(ptdev, GPU_THREAD_MAX_BARRIER_SIZE);
> + ptdev->gpu_info.coherency_features = gpu_read(ptdev, GPU_COHERENCY_FEATURES);
> + for (i = 0; i < 4; i++)
> + ptdev->gpu_info.texture_features[i] = gpu_read(ptdev, GPU_TEXTURE_FEATURES(i));
> +
> + ptdev->gpu_info.as_present = gpu_read(ptdev, GPU_AS_PRESENT);
> +
> + ptdev->gpu_info.shader_present = gpu_read(ptdev, GPU_SHADER_PRESENT_LO);
> + ptdev->gpu_info.shader_present |= (u64)gpu_read(ptdev, GPU_SHADER_PRESENT_HI) << 32;
> +
> + ptdev->gpu_info.tiler_present = gpu_read(ptdev, GPU_TILER_PRESENT_LO);
> + ptdev->gpu_info.tiler_present |= (u64)gpu_read(ptdev, GPU_TILER_PRESENT_HI) << 32;
> +
> + ptdev->gpu_info.l2_present = gpu_read(ptdev, GPU_L2_PRESENT_LO);
> + ptdev->gpu_info.l2_present |= (u64)gpu_read(ptdev, GPU_L2_PRESENT_HI) << 32;
> +
> + arch_major = GPU_ARCH_MAJOR(ptdev->gpu_info.gpu_id);
> + product_major = GPU_PROD_MAJOR(ptdev->gpu_info.gpu_id);
> + major = GPU_VER_MAJOR(ptdev->gpu_info.gpu_id);
> + minor = GPU_VER_MINOR(ptdev->gpu_info.gpu_id);
> + status = GPU_VER_STATUS(ptdev->gpu_info.gpu_id);
> +
> + for (model = gpu_models; model->name; model++) {
> + if (model->arch_major == arch_major &&
> + model->product_major == product_major)
> + break;
> + }
> +
> + drm_info(&ptdev->base,
> + "mali-%s id 0x%x major 0x%x minor 0x%x status 0x%x",
> + model->name ?: "unknown", ptdev->gpu_info.gpu_id >> 16,
> + major, minor, status);
> +
> + drm_info(&ptdev->base,
> + "Features: L2:%#x Tiler:%#x Mem:%#x MMU:%#x AS:%#x",
> + ptdev->gpu_info.l2_features,
> + ptdev->gpu_info.tiler_features,
> + ptdev->gpu_info.mem_features,
> + ptdev->gpu_info.mmu_features,
> + ptdev->gpu_info.as_present);
> +
> + drm_info(&ptdev->base,
> + "shader_present=0x%0llx l2_present=0x%0llx tiler_present=0x%0llx",
> + ptdev->gpu_info.shader_present, ptdev->gpu_info.l2_present,
> + ptdev->gpu_info.tiler_present);
> +}
> +
> +static void panthor_gpu_irq_handler(struct panthor_device *ptdev, u32 status)
> +{
> + if (status & GPU_IRQ_FAULT) {
> + u32 fault_status = gpu_read(ptdev, GPU_FAULT_STATUS);
> + u64 address = ((u64)gpu_read(ptdev, GPU_FAULT_ADDR_HI) << 32) |
> + gpu_read(ptdev, GPU_FAULT_ADDR_LO);
> +
> + drm_warn(&ptdev->base, "GPU Fault 0x%08x (%s) at 0x%016llx\n",
> + fault_status, panthor_exception_name(ptdev, fault_status & 0xFF),
> + address);
> + }
> + if (status & GPU_IRQ_PROTM_FAULT)
> + drm_warn(&ptdev->base, "GPU Fault in protected mode\n");
> +
> + spin_lock(&ptdev->gpu->reqs_lock);
> + if (status & ptdev->gpu->pending_reqs) {
> + ptdev->gpu->pending_reqs &= ~status;
> + wake_up_all(&ptdev->gpu->reqs_acked);
> + }
> + spin_unlock(&ptdev->gpu->reqs_lock);
> +}
> +PANTHOR_IRQ_HANDLER(gpu, GPU, panthor_gpu_irq_handler);
> +
> +/**
> + * panthor_gpu_unplug() - Called when the GPU is unplugged.
> + * @ptdev: Device to unplug.
> + */
> +void panthor_gpu_unplug(struct panthor_device *ptdev)
> +{
> + unsigned long flags;
> +
> + /* Make sure the IRQ handler is not running after that point. */
> + panthor_gpu_irq_suspend(&ptdev->gpu->irq);
> +
> + /* Wake-up all waiters. */
> + spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
> + ptdev->gpu->pending_reqs = 0;
> + wake_up_all(&ptdev->gpu->reqs_acked);
> + spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
> +}
> +
> +/**
> + * panthor_gpu_init() - Initialize the GPU block
> + * @ptdev: Device.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +int panthor_gpu_init(struct panthor_device *ptdev)
> +{
> + struct panthor_gpu *gpu;
> + u32 pa_bits;
> + int ret, irq;
> +
> + gpu = drmm_kzalloc(&ptdev->base, sizeof(*gpu), GFP_KERNEL);
> + if (!gpu)
> + return -ENOMEM;
> +
> + spin_lock_init(&gpu->reqs_lock);
> + init_waitqueue_head(&gpu->reqs_acked);
> + ptdev->gpu = gpu;
> + panthor_gpu_init_info(ptdev);
> +
> + dma_set_max_seg_size(ptdev->base.dev, UINT_MAX);
> + pa_bits = GPU_MMU_FEATURES_PA_BITS(ptdev->gpu_info.mmu_features);
> + ret = dma_set_mask_and_coherent(ptdev->base.dev, DMA_BIT_MASK(pa_bits));
> + if (ret)
> + return ret;
> +
> + irq = platform_get_irq_byname(to_platform_device(ptdev->base.dev), "gpu");
> + if (irq <= 0)
> + return ret;
> +
> + ret = panthor_request_gpu_irq(ptdev, &ptdev->gpu->irq, irq, GPU_INTERRUPTS_MASK);
> + if (ret)
> + return ret;
> +
> + return 0;
> +}
> +
> +/**
> + * panthor_gpu_block_power_off() - Power-off a specific block of the GPU
> + * @ptdev: Device.
> + * @blk_name: Block name.
> + * @pwroff_reg: Power-off register for this block.
> + * @pwrtrans_reg: Power transition register for this block.
> + * @mask: Sub-elements to power-off.
> + * @timeout_us: Timeout in microseconds.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +int panthor_gpu_block_power_off(struct panthor_device *ptdev,
> + const char *blk_name,
> + u32 pwroff_reg, u32 pwrtrans_reg,
> + u64 mask, u32 timeout_us)
> +{
> + u32 val, i;
> + int ret;
> +
> + for (i = 0; i < 2; i++) {
> + u32 mask32 = mask >> (i * 32);
> +
> + if (!mask32)
> + continue;
> +
> + ret = readl_relaxed_poll_timeout(ptdev->iomem + pwrtrans_reg + (i * 4),
> + val, !(mask32 & val),
> + 100, timeout_us);
> + if (ret) {
> + drm_err(&ptdev->base, "timeout waiting on %s:%llx power transition",
> + blk_name, mask);
> + return ret;
> + }
> + }
> +
> + if (mask & GENMASK(31, 0))
> + gpu_write(ptdev, pwroff_reg, mask);
> +
> + if (mask >> 32)
> + gpu_write(ptdev, pwroff_reg + 4, mask >> 32);
> +
> + for (i = 0; i < 2; i++) {
> + u32 mask32 = mask >> (i * 32);
> +
> + if (!mask32)
> + continue;
> +
> + ret = readl_relaxed_poll_timeout(ptdev->iomem + pwrtrans_reg + (i * 4),
> + val, !(mask & val),
Should be mask32 not mask. (Not that I think anything uses the upper
register yet)
Steve
> + 100, timeout_us);
> + if (ret) {
> + drm_err(&ptdev->base, "timeout waiting on %s:%llx power transition",
> + blk_name, mask);
> + return ret;
> + }
> + }
> +
> + return 0;
> +}
> +
> +/**
> + * panthor_gpu_block_power_on() - Power-on a specific block of the GPU
> + * @ptdev: Device.
> + * @blk_name: Block name.
> + * @pwron_reg: Power-on register for this block.
> + * @pwrtrans_reg: Power transition register for this block.
> + * @rdy_reg: Power transition ready register.
> + * @mask: Sub-elements to power-on.
> + * @timeout_us: Timeout in microseconds.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +int panthor_gpu_block_power_on(struct panthor_device *ptdev,
> + const char *blk_name,
> + u32 pwron_reg, u32 pwrtrans_reg,
> + u32 rdy_reg, u64 mask, u32 timeout_us)
> +{
> + u32 val, i;
> + int ret;
> +
> + for (i = 0; i < 2; i++) {
> + u32 mask32 = mask >> (i * 32);
> +
> + if (!mask32)
> + continue;
> +
> + ret = readl_relaxed_poll_timeout(ptdev->iomem + pwrtrans_reg + (i * 4),
> + val, !(mask32 & val),
> + 100, timeout_us);
> + if (ret) {
> + drm_err(&ptdev->base, "timeout waiting on %s:%llx power transition",
> + blk_name, mask);
> + return ret;
> + }
> + }
> +
> + if (mask & GENMASK(31, 0))
> + gpu_write(ptdev, pwron_reg, mask);
> +
> + if (mask >> 32)
> + gpu_write(ptdev, pwron_reg + 4, mask >> 32);
> +
> + for (i = 0; i < 2; i++) {
> + u32 mask32 = mask >> (i * 32);
> +
> + if (!mask32)
> + continue;
> +
> + ret = readl_relaxed_poll_timeout(ptdev->iomem + rdy_reg + (i * 4),
> + val, (mask32 & val) == mask32,
> + 100, timeout_us);
> + if (ret) {
> + drm_err(&ptdev->base, "timeout waiting on %s:%llx readyness",
> + blk_name, mask);
> + return ret;
> + }
> + }
> +
> + return 0;
> +}
> +
> +/**
> + * panthor_gpu_l2_power_on() - Power-on the L2-cache
> + * @ptdev: Device.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +int panthor_gpu_l2_power_on(struct panthor_device *ptdev)
> +{
> + if (ptdev->gpu_info.l2_present != 1) {
> + /*
> + * Only support one core group now.
> + * ~(l2_present - 1) unsets all bits in l2_present except
> + * the bottom bit. (l2_present - 2) has all the bits in
> + * the first core group set. AND them together to generate
> + * a mask of cores in the first core group.
> + */
> + u64 core_mask = ~(ptdev->gpu_info.l2_present - 1) &
> + (ptdev->gpu_info.l2_present - 2);
> + drm_info_once(&ptdev->base, "using only 1st core group (%lu cores from %lu)\n",
> + hweight64(core_mask),
> + hweight64(ptdev->gpu_info.shader_present));
> + }
> +
> + return panthor_gpu_power_on(ptdev, L2, 1, 20000);
> +}
> +
> +/**
> + * panthor_gpu_flush_caches() - Flush caches
> + * @ptdev: Device.
> + * @l2: L2 flush type.
> + * @lsc: LSC flush type.
> + * @other: Other flush type.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +int panthor_gpu_flush_caches(struct panthor_device *ptdev,
> + u32 l2, u32 lsc, u32 other)
> +{
> + bool timedout = false;
> + unsigned long flags;
> +
> + spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
> + if (!drm_WARN_ON(&ptdev->base,
> + ptdev->gpu->pending_reqs & GPU_IRQ_CLEAN_CACHES_COMPLETED)) {
> + ptdev->gpu->pending_reqs |= GPU_IRQ_CLEAN_CACHES_COMPLETED;
> + gpu_write(ptdev, GPU_CMD, GPU_FLUSH_CACHES(l2, lsc, other));
> + }
> + spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
> +
> + if (!wait_event_timeout(ptdev->gpu->reqs_acked,
> + !(ptdev->gpu->pending_reqs & GPU_IRQ_CLEAN_CACHES_COMPLETED),
> + msecs_to_jiffies(100))) {
> + spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
> + if ((ptdev->gpu->pending_reqs & GPU_IRQ_CLEAN_CACHES_COMPLETED) != 0 &&
> + !(gpu_read(ptdev, GPU_INT_RAWSTAT) & GPU_IRQ_CLEAN_CACHES_COMPLETED))
> + timedout = true;
> + else
> + ptdev->gpu->pending_reqs &= ~GPU_IRQ_CLEAN_CACHES_COMPLETED;
> + spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
> + }
> +
> + if (timedout) {
> + drm_err(&ptdev->base, "Flush caches timeout");
> + return -ETIMEDOUT;
> + }
> +
> + return 0;
> +}
> +
> +/**
> + * panthor_gpu_soft_reset() - Issue a soft-reset
> + * @ptdev: Device.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +int panthor_gpu_soft_reset(struct panthor_device *ptdev)
> +{
> + bool timedout = false;
> + unsigned long flags;
> +
> + spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
> + if (!drm_WARN_ON(&ptdev->base,
> + ptdev->gpu->pending_reqs & GPU_IRQ_RESET_COMPLETED)) {
> + ptdev->gpu->pending_reqs |= GPU_IRQ_RESET_COMPLETED;
> + gpu_write(ptdev, GPU_INT_CLEAR, GPU_IRQ_RESET_COMPLETED);
> + gpu_write(ptdev, GPU_CMD, GPU_SOFT_RESET);
> + }
> + spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
> +
> + if (!wait_event_timeout(ptdev->gpu->reqs_acked,
> + !(ptdev->gpu->pending_reqs & GPU_IRQ_RESET_COMPLETED),
> + msecs_to_jiffies(100))) {
> + spin_lock_irqsave(&ptdev->gpu->reqs_lock, flags);
> + if ((ptdev->gpu->pending_reqs & GPU_IRQ_RESET_COMPLETED) != 0 &&
> + !(gpu_read(ptdev, GPU_INT_RAWSTAT) & GPU_IRQ_RESET_COMPLETED))
> + timedout = true;
> + else
> + ptdev->gpu->pending_reqs &= ~GPU_IRQ_RESET_COMPLETED;
> + spin_unlock_irqrestore(&ptdev->gpu->reqs_lock, flags);
> + }
> +
> + if (timedout) {
> + drm_err(&ptdev->base, "Soft reset timeout");
> + return -ETIMEDOUT;
> + }
> +
> + return 0;
> +}
> +
> +/**
> + * panthor_gpu_suspend() - Suspend the GPU block.
> + * @ptdev: Device.
> + *
> + * Suspend the GPU irq. This should be called last in the suspend procedure,
> + * after all other blocks have been suspented.
> + */
> +void panthor_gpu_suspend(struct panthor_device *ptdev)
> +{
> + /*
> + * It may be preferable to simply power down the L2, but for now just
> + * soft-reset which will leave the L2 powered down.
> + */
> + panthor_gpu_soft_reset(ptdev);
> + panthor_gpu_irq_suspend(&ptdev->gpu->irq);
> +}
> +
> +/**
> + * panthor_gpu_resume() - Resume the GPU block.
> + * @ptdev: Device.
> + *
> + * Resume the IRQ handler and power-on the L2-cache.
> + * The FW takes care of powering the other blocks.
> + */
> +void panthor_gpu_resume(struct panthor_device *ptdev)
> +{
> + panthor_gpu_irq_resume(&ptdev->gpu->irq, GPU_INTERRUPTS_MASK);
> + panthor_gpu_l2_power_on(ptdev);
> +}
> diff --git a/drivers/gpu/drm/panthor/panthor_gpu.h b/drivers/gpu/drm/panthor/panthor_gpu.h
> new file mode 100644
> index 000000000000..bba7555dd3c6
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_gpu.h
> @@ -0,0 +1,52 @@
> +/* SPDX-License-Identifier: GPL-2.0 or MIT */
> +/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
> +/* Copyright 2019 Collabora ltd. */
> +
> +#ifndef __PANTHOR_GPU_H__
> +#define __PANTHOR_GPU_H__
> +
> +struct panthor_device;
> +
> +int panthor_gpu_init(struct panthor_device *ptdev);
> +void panthor_gpu_unplug(struct panthor_device *ptdev);
> +void panthor_gpu_suspend(struct panthor_device *ptdev);
> +void panthor_gpu_resume(struct panthor_device *ptdev);
> +
> +int panthor_gpu_block_power_on(struct panthor_device *ptdev,
> + const char *blk_name,
> + u32 pwron_reg, u32 pwrtrans_reg,
> + u32 rdy_reg, u64 mask, u32 timeout_us);
> +int panthor_gpu_block_power_off(struct panthor_device *ptdev,
> + const char *blk_name,
> + u32 pwroff_reg, u32 pwrtrans_reg,
> + u64 mask, u32 timeout_us);
> +
> +/**
> + * panthor_gpu_power_on() - Power on the GPU block.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +#define panthor_gpu_power_on(ptdev, type, mask, timeout_us) \
> + panthor_gpu_block_power_on(ptdev, #type, \
> + type ## _PWRON_LO, \
> + type ## _PWRTRANS_LO, \
> + type ## _READY_LO, \
> + mask, timeout_us)
> +
> +/**
> + * panthor_gpu_power_off() - Power off the GPU block.
> + *
> + * Return: 0 on success, a negative error code otherwise.
> + */
> +#define panthor_gpu_power_off(ptdev, type, mask, timeout_us) \
> + panthor_gpu_block_power_off(ptdev, #type, \
> + type ## _PWROFF_LO, \
> + type ## _PWRTRANS_LO, \
> + mask, timeout_us)
> +
> +int panthor_gpu_l2_power_on(struct panthor_device *ptdev);
> +int panthor_gpu_flush_caches(struct panthor_device *ptdev,
> + u32 l2, u32 lsc, u32 other);
> +int panthor_gpu_soft_reset(struct panthor_device *ptdev);
> +
> +#endif
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 05/14] drm/panthor: Add GEM logical block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (3 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 04/14] drm/panthor: Add the GPU " Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-02-09 15:58 ` Steven Price
2024-02-12 12:31 ` Liviu Dudau
2024-01-22 16:30 ` [PATCH v4 06/14] drm/panthor: Add the devfreq " Boris Brezillon
` (12 subsequent siblings)
17 siblings, 2 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
Anything relating to GEM object management is placed here. Nothing
particularly interesting here, given the implementation is based on
drm_gem_shmem_object, which is doing most of the work.
v4:
- Force kernel BOs to be GPU mapped
- Make panthor_kernel_bo_destroy() robust against ERR/NULL BO pointers
to simplify the call sites
v3:
- Add acks for the MIT/GPL2 relicensing
- Provide a panthor_kernel_bo abstraction for buffer objects managed by
the kernel (will replace panthor_fw_mem and be used everywhere we were
using panthor_gem_create_and_map() before)
- Adjust things to match drm_gpuvm changes
- Change return of panthor_gem_create_with_handle() to int
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
---
drivers/gpu/drm/panthor/panthor_gem.c | 228 ++++++++++++++++++++++++++
drivers/gpu/drm/panthor/panthor_gem.h | 144 ++++++++++++++++
2 files changed, 372 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_gem.c
create mode 100644 drivers/gpu/drm/panthor/panthor_gem.h
diff --git a/drivers/gpu/drm/panthor/panthor_gem.c b/drivers/gpu/drm/panthor/panthor_gem.c
new file mode 100644
index 000000000000..42e342fcad19
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_gem.c
@@ -0,0 +1,228 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
+/* Copyright 2023 Collabora ltd. */
+
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/dma-buf.h>
+#include <linux/dma-mapping.h>
+
+#include <drm/panthor_drm.h>
+
+#include "panthor_device.h"
+#include "panthor_gem.h"
+#include "panthor_mmu.h"
+
+static void panthor_gem_free_object(struct drm_gem_object *obj)
+{
+ struct panthor_gem_object *bo = to_panthor_bo(obj);
+ struct drm_gem_object *vm_root_gem = bo->exclusive_vm_root_gem;
+
+ drm_gem_free_mmap_offset(&bo->base.base);
+ mutex_destroy(&bo->gpuva_list_lock);
+ drm_gem_shmem_free(&bo->base);
+ drm_gem_object_put(vm_root_gem);
+}
+
+/**
+ * panthor_kernel_bo_destroy() - Destroy a kernel buffer object
+ * @vm: The VM this BO was mapped to.
+ * @bo: Kernel buffer object to destroy. If NULL or an ERR_PTR(), the destruction
+ * is skipped.
+ */
+void panthor_kernel_bo_destroy(struct panthor_vm *vm,
+ struct panthor_kernel_bo *bo)
+{
+ int ret;
+
+ if (IS_ERR_OR_NULL(bo))
+ return;
+
+ panthor_kernel_bo_vunmap(bo);
+
+ if (drm_WARN_ON(bo->obj->dev,
+ to_panthor_bo(bo->obj)->exclusive_vm_root_gem != panthor_vm_root_gem(vm)))
+ goto out_free_bo;
+
+ ret = panthor_vm_unmap_range(vm, bo->va_node.start,
+ panthor_kernel_bo_size(bo));
+ if (ret)
+ goto out_free_bo;
+
+ panthor_vm_free_va(vm, &bo->va_node);
+ drm_gem_object_put(bo->obj);
+
+out_free_bo:
+ kfree(bo);
+}
+
+/**
+ * panthor_kernel_bo_create() - Create and map a GEM object to a VM
+ * @ptdev: Device.
+ * @vm: VM to map the GEM to. If NULL, the kernel object is not GPU mapped.
+ * @size: Size of the buffer object.
+ * @bo_flags: Combination of drm_panthor_bo_flags flags.
+ * @vm_map_flags: Combination of drm_panthor_vm_bind_op_flags (only those
+ * that are related to map operations).
+ * @gpu_va: GPU address assigned when mapping to the VM.
+ * If gpu_va == PANTHOR_VM_KERNEL_AUTO_VA, the virtual address will be
+ * automatically allocated.
+ *
+ * Return: A valid pointer in case of success, an ERR_PTR() otherwise.
+ */
+struct panthor_kernel_bo *
+panthor_kernel_bo_create(struct panthor_device *ptdev, struct panthor_vm *vm,
+ size_t size, u32 bo_flags, u32 vm_map_flags,
+ u64 gpu_va)
+{
+ struct drm_gem_shmem_object *obj;
+ struct panthor_kernel_bo *kbo;
+ struct panthor_gem_object *bo;
+ int ret;
+
+ if (drm_WARN_ON(&ptdev->base, !vm))
+ return ERR_PTR(-EINVAL);
+
+ kbo = kzalloc(sizeof(*kbo), GFP_KERNEL);
+ if (!kbo)
+ return ERR_PTR(-ENOMEM);
+
+ obj = drm_gem_shmem_create(&ptdev->base, size);
+ if (IS_ERR(obj)) {
+ ret = PTR_ERR(obj);
+ goto err_free_bo;
+ }
+
+ bo = to_panthor_bo(&obj->base);
+ size = obj->base.size;
+ kbo->obj = &obj->base;
+ bo->flags = bo_flags;
+
+ ret = panthor_vm_alloc_va(vm, gpu_va, size, &kbo->va_node);
+ if (ret)
+ goto err_put_obj;
+
+ ret = panthor_vm_map_bo_range(vm, bo, 0, size, kbo->va_node.start, vm_map_flags);
+ if (ret)
+ goto err_free_va;
+
+ bo->exclusive_vm_root_gem = panthor_vm_root_gem(vm);
+ drm_gem_object_get(bo->exclusive_vm_root_gem);
+ bo->base.base.resv = bo->exclusive_vm_root_gem->resv;
+ return kbo;
+
+err_free_va:
+ panthor_vm_free_va(vm, &kbo->va_node);
+
+err_put_obj:
+ drm_gem_object_put(&obj->base);
+
+err_free_bo:
+ kfree(kbo);
+ return ERR_PTR(ret);
+}
+
+static int panthor_gem_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
+{
+ struct panthor_gem_object *bo = to_panthor_bo(obj);
+
+ /* Don't allow mmap on objects that have the NO_MMAP flag set. */
+ if (bo->flags & DRM_PANTHOR_BO_NO_MMAP)
+ return -EINVAL;
+
+ return drm_gem_shmem_object_mmap(obj, vma);
+}
+
+static struct dma_buf *
+panthor_gem_prime_export(struct drm_gem_object *obj, int flags)
+{
+ /* We can't export GEMs that have an exclusive VM. */
+ if (to_panthor_bo(obj)->exclusive_vm_root_gem)
+ return ERR_PTR(-EINVAL);
+
+ return drm_gem_prime_export(obj, flags);
+}
+
+static const struct drm_gem_object_funcs panthor_gem_funcs = {
+ .free = panthor_gem_free_object,
+ .print_info = drm_gem_shmem_object_print_info,
+ .pin = drm_gem_shmem_object_pin,
+ .unpin = drm_gem_shmem_object_unpin,
+ .get_sg_table = drm_gem_shmem_object_get_sg_table,
+ .vmap = drm_gem_shmem_object_vmap,
+ .vunmap = drm_gem_shmem_object_vunmap,
+ .mmap = panthor_gem_mmap,
+ .export = panthor_gem_prime_export,
+ .vm_ops = &drm_gem_shmem_vm_ops,
+};
+
+/**
+ * panthor_gem_create_object - Implementation of driver->gem_create_object.
+ * @ddev: DRM device
+ * @size: Size in bytes of the memory the object will reference
+ *
+ * This lets the GEM helpers allocate object structs for us, and keep
+ * our BO stats correct.
+ */
+struct drm_gem_object *panthor_gem_create_object(struct drm_device *ddev, size_t size)
+{
+ struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
+ struct panthor_gem_object *obj;
+
+ obj = kzalloc(sizeof(*obj), GFP_KERNEL);
+ if (!obj)
+ return ERR_PTR(-ENOMEM);
+
+ obj->base.base.funcs = &panthor_gem_funcs;
+ obj->base.map_wc = !ptdev->coherent;
+ mutex_init(&obj->gpuva_list_lock);
+ drm_gem_gpuva_set_lock(&obj->base.base, &obj->gpuva_list_lock);
+
+ return &obj->base.base;
+}
+
+/**
+ * panthor_gem_create_with_handle() - Create a GEM object and attach it to a handle.
+ * @file: DRM file.
+ * @ddev: DRM device.
+ * @exclusive_vm: Exclusive VM. Not NULL if the GEM object can't be shared.
+ * @size: Size of the GEM object to allocate.
+ * @flags: Combination of drm_panthor_bo_flags flags.
+ * @handle: Pointer holding the handle pointing to the new GEM object.
+ *
+ * Return: Zero on success
+ */
+int
+panthor_gem_create_with_handle(struct drm_file *file,
+ struct drm_device *ddev,
+ struct panthor_vm *exclusive_vm,
+ size_t size,
+ u32 flags, u32 *handle)
+{
+ int ret;
+ struct drm_gem_shmem_object *shmem;
+ struct panthor_gem_object *bo;
+
+ shmem = drm_gem_shmem_create(ddev, size);
+ if (IS_ERR(shmem))
+ return PTR_ERR(shmem);
+
+ bo = to_panthor_bo(&shmem->base);
+ bo->flags = flags;
+
+ if (exclusive_vm) {
+ bo->exclusive_vm_root_gem = panthor_vm_root_gem(exclusive_vm);
+ drm_gem_object_get(bo->exclusive_vm_root_gem);
+ bo->base.base.resv = bo->exclusive_vm_root_gem->resv;
+ }
+
+ /*
+ * Allocate an id of idr table where the obj is registered
+ * and handle has the id what user can see.
+ */
+ ret = drm_gem_handle_create(file, &shmem->base, handle);
+ /* drop reference from allocate - handle holds it now. */
+ drm_gem_object_put(&shmem->base);
+
+ return ret;
+}
diff --git a/drivers/gpu/drm/panthor/panthor_gem.h b/drivers/gpu/drm/panthor/panthor_gem.h
new file mode 100644
index 000000000000..6c8010ceb641
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_gem.h
@@ -0,0 +1,144 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
+/* Copyright 2023 Collabora ltd. */
+
+#ifndef __PANTHOR_GEM_H__
+#define __PANTHOR_GEM_H__
+
+#include <drm/drm_gem_shmem_helper.h>
+#include <drm/drm_mm.h>
+
+#include <linux/iosys-map.h>
+#include <linux/rwsem.h>
+
+struct panthor_vm;
+
+/**
+ * struct panthor_gem_object - Driver specific GEM object.
+ */
+struct panthor_gem_object {
+ /** @base: Inherit from drm_gem_shmem_object. */
+ struct drm_gem_shmem_object base;
+
+ /**
+ * @exclusive_vm_root_gem: Root GEM of the exclusive VM this GEM object
+ * is attached to.
+ *
+ * If @exclusive_vm_root_gem != NULL, any attempt to bind the GEM to a
+ * different VM will fail.
+ *
+ * All FW memory objects have this field set to the root GEM of the MCU
+ * VM.
+ */
+ struct drm_gem_object *exclusive_vm_root_gem;
+
+ /**
+ * @gpuva_list_lock: Custom GPUVA lock.
+ *
+ * Used to protect insertion of drm_gpuva elements to the
+ * drm_gem_object.gpuva.list list.
+ *
+ * We can't use the GEM resv for that, because drm_gpuva_link() is
+ * called in a dma-signaling path, where we're not allowed to take
+ * resv locks.
+ */
+ struct mutex gpuva_list_lock;
+
+ /** @flags: Combination of drm_panthor_bo_flags flags. */
+ u32 flags;
+};
+
+/**
+ * struct panthor_kernel_bo - Kernel buffer object.
+ *
+ * These objects are only manipulated by the kernel driver and not
+ * directly exposed to the userspace. The GPU address of a kernel
+ * BO might be passed to userspace though.
+ */
+struct panthor_kernel_bo {
+ /**
+ * @obj: The GEM object backing this kernel buffer object.
+ */
+ struct drm_gem_object *obj;
+
+ /**
+ * @va_node: VA space allocated to this GEM.
+ */
+ struct drm_mm_node va_node;
+
+ /**
+ * @kmap: Kernel CPU mapping of @gem.
+ */
+ void *kmap;
+};
+
+static inline
+struct panthor_gem_object *to_panthor_bo(struct drm_gem_object *obj)
+{
+ return container_of(to_drm_gem_shmem_obj(obj), struct panthor_gem_object, base);
+}
+
+struct drm_gem_object *panthor_gem_create_object(struct drm_device *ddev, size_t size);
+
+struct drm_gem_object *
+panthor_gem_prime_import_sg_table(struct drm_device *ddev,
+ struct dma_buf_attachment *attach,
+ struct sg_table *sgt);
+
+int
+panthor_gem_create_with_handle(struct drm_file *file,
+ struct drm_device *ddev,
+ struct panthor_vm *exclusive_vm,
+ size_t size,
+ u32 flags,
+ uint32_t *handle);
+
+static inline u64
+panthor_kernel_bo_gpuva(struct panthor_kernel_bo *bo)
+{
+ return bo->va_node.start;
+}
+
+static inline size_t
+panthor_kernel_bo_size(struct panthor_kernel_bo *bo)
+{
+ return bo->obj->size;
+}
+
+static inline int
+panthor_kernel_bo_vmap(struct panthor_kernel_bo *bo)
+{
+ struct iosys_map map;
+ int ret;
+
+ if (bo->kmap)
+ return 0;
+
+ ret = drm_gem_vmap_unlocked(bo->obj, &map);
+ if (ret)
+ return ret;
+
+ bo->kmap = map.vaddr;
+ return 0;
+}
+
+static inline void
+panthor_kernel_bo_vunmap(struct panthor_kernel_bo *bo)
+{
+ if (bo->kmap) {
+ struct iosys_map map = IOSYS_MAP_INIT_VADDR(bo->kmap);
+
+ drm_gem_vunmap_unlocked(bo->obj, &map);
+ bo->kmap = NULL;
+ }
+}
+
+struct panthor_kernel_bo *
+panthor_kernel_bo_create(struct panthor_device *ptdev, struct panthor_vm *vm,
+ size_t size, u32 bo_flags, u32 vm_map_flags,
+ u64 gpu_va);
+
+void panthor_kernel_bo_destroy(struct panthor_vm *vm,
+ struct panthor_kernel_bo *bo);
+
+#endif /* __PANTHOR_GEM_H__ */
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 05/14] drm/panthor: Add GEM logical block
2024-01-22 16:30 ` [PATCH v4 05/14] drm/panthor: Add GEM " Boris Brezillon
@ 2024-02-09 15:58 ` Steven Price
2024-02-12 12:31 ` Liviu Dudau
1 sibling, 0 replies; 51+ messages in thread
From: Steven Price @ 2024-02-09 15:58 UTC (permalink / raw)
To: Boris Brezillon, dri-devel
Cc: Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On 22/01/2024 16:30, Boris Brezillon wrote:
> Anything relating to GEM object management is placed here. Nothing
> particularly interesting here, given the implementation is based on
> drm_gem_shmem_object, which is doing most of the work.
>
> v4:
> - Force kernel BOs to be GPU mapped
> - Make panthor_kernel_bo_destroy() robust against ERR/NULL BO pointers
> to simplify the call sites
>
> v3:
> - Add acks for the MIT/GPL2 relicensing
> - Provide a panthor_kernel_bo abstraction for buffer objects managed by
> the kernel (will replace panthor_fw_mem and be used everywhere we were
> using panthor_gem_create_and_map() before)
> - Adjust things to match drm_gpuvm changes
> - Change return of panthor_gem_create_with_handle() to int
>
> Co-developed-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
> Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
> Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
Reviewed-by: Steven Price <steven.price@arm.com>
> ---
> drivers/gpu/drm/panthor/panthor_gem.c | 228 ++++++++++++++++++++++++++
> drivers/gpu/drm/panthor/panthor_gem.h | 144 ++++++++++++++++
> 2 files changed, 372 insertions(+)
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.h
>
> diff --git a/drivers/gpu/drm/panthor/panthor_gem.c b/drivers/gpu/drm/panthor/panthor_gem.c
> new file mode 100644
> index 000000000000..42e342fcad19
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_gem.c
> @@ -0,0 +1,228 @@
> +// SPDX-License-Identifier: GPL-2.0 or MIT
> +/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
> +/* Copyright 2023 Collabora ltd. */
> +
> +#include <linux/err.h>
> +#include <linux/slab.h>
> +#include <linux/dma-buf.h>
> +#include <linux/dma-mapping.h>
> +
> +#include <drm/panthor_drm.h>
> +
> +#include "panthor_device.h"
> +#include "panthor_gem.h"
> +#include "panthor_mmu.h"
> +
> +static void panthor_gem_free_object(struct drm_gem_object *obj)
> +{
> + struct panthor_gem_object *bo = to_panthor_bo(obj);
> + struct drm_gem_object *vm_root_gem = bo->exclusive_vm_root_gem;
> +
> + drm_gem_free_mmap_offset(&bo->base.base);
> + mutex_destroy(&bo->gpuva_list_lock);
> + drm_gem_shmem_free(&bo->base);
> + drm_gem_object_put(vm_root_gem);
> +}
> +
> +/**
> + * panthor_kernel_bo_destroy() - Destroy a kernel buffer object
> + * @vm: The VM this BO was mapped to.
> + * @bo: Kernel buffer object to destroy. If NULL or an ERR_PTR(), the destruction
> + * is skipped.
> + */
> +void panthor_kernel_bo_destroy(struct panthor_vm *vm,
> + struct panthor_kernel_bo *bo)
> +{
> + int ret;
> +
> + if (IS_ERR_OR_NULL(bo))
> + return;
> +
> + panthor_kernel_bo_vunmap(bo);
> +
> + if (drm_WARN_ON(bo->obj->dev,
> + to_panthor_bo(bo->obj)->exclusive_vm_root_gem != panthor_vm_root_gem(vm)))
> + goto out_free_bo;
> +
> + ret = panthor_vm_unmap_range(vm, bo->va_node.start,
> + panthor_kernel_bo_size(bo));
> + if (ret)
> + goto out_free_bo;
> +
> + panthor_vm_free_va(vm, &bo->va_node);
> + drm_gem_object_put(bo->obj);
> +
> +out_free_bo:
> + kfree(bo);
> +}
> +
> +/**
> + * panthor_kernel_bo_create() - Create and map a GEM object to a VM
> + * @ptdev: Device.
> + * @vm: VM to map the GEM to. If NULL, the kernel object is not GPU mapped.
> + * @size: Size of the buffer object.
> + * @bo_flags: Combination of drm_panthor_bo_flags flags.
> + * @vm_map_flags: Combination of drm_panthor_vm_bind_op_flags (only those
> + * that are related to map operations).
> + * @gpu_va: GPU address assigned when mapping to the VM.
> + * If gpu_va == PANTHOR_VM_KERNEL_AUTO_VA, the virtual address will be
> + * automatically allocated.
> + *
> + * Return: A valid pointer in case of success, an ERR_PTR() otherwise.
> + */
> +struct panthor_kernel_bo *
> +panthor_kernel_bo_create(struct panthor_device *ptdev, struct panthor_vm *vm,
> + size_t size, u32 bo_flags, u32 vm_map_flags,
> + u64 gpu_va)
> +{
> + struct drm_gem_shmem_object *obj;
> + struct panthor_kernel_bo *kbo;
> + struct panthor_gem_object *bo;
> + int ret;
> +
> + if (drm_WARN_ON(&ptdev->base, !vm))
> + return ERR_PTR(-EINVAL);
> +
> + kbo = kzalloc(sizeof(*kbo), GFP_KERNEL);
> + if (!kbo)
> + return ERR_PTR(-ENOMEM);
> +
> + obj = drm_gem_shmem_create(&ptdev->base, size);
> + if (IS_ERR(obj)) {
> + ret = PTR_ERR(obj);
> + goto err_free_bo;
> + }
> +
> + bo = to_panthor_bo(&obj->base);
> + size = obj->base.size;
> + kbo->obj = &obj->base;
> + bo->flags = bo_flags;
> +
> + ret = panthor_vm_alloc_va(vm, gpu_va, size, &kbo->va_node);
> + if (ret)
> + goto err_put_obj;
> +
> + ret = panthor_vm_map_bo_range(vm, bo, 0, size, kbo->va_node.start, vm_map_flags);
> + if (ret)
> + goto err_free_va;
> +
> + bo->exclusive_vm_root_gem = panthor_vm_root_gem(vm);
> + drm_gem_object_get(bo->exclusive_vm_root_gem);
> + bo->base.base.resv = bo->exclusive_vm_root_gem->resv;
> + return kbo;
> +
> +err_free_va:
> + panthor_vm_free_va(vm, &kbo->va_node);
> +
> +err_put_obj:
> + drm_gem_object_put(&obj->base);
> +
> +err_free_bo:
> + kfree(kbo);
> + return ERR_PTR(ret);
> +}
> +
> +static int panthor_gem_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
> +{
> + struct panthor_gem_object *bo = to_panthor_bo(obj);
> +
> + /* Don't allow mmap on objects that have the NO_MMAP flag set. */
> + if (bo->flags & DRM_PANTHOR_BO_NO_MMAP)
> + return -EINVAL;
> +
> + return drm_gem_shmem_object_mmap(obj, vma);
> +}
> +
> +static struct dma_buf *
> +panthor_gem_prime_export(struct drm_gem_object *obj, int flags)
> +{
> + /* We can't export GEMs that have an exclusive VM. */
> + if (to_panthor_bo(obj)->exclusive_vm_root_gem)
> + return ERR_PTR(-EINVAL);
> +
> + return drm_gem_prime_export(obj, flags);
> +}
> +
> +static const struct drm_gem_object_funcs panthor_gem_funcs = {
> + .free = panthor_gem_free_object,
> + .print_info = drm_gem_shmem_object_print_info,
> + .pin = drm_gem_shmem_object_pin,
> + .unpin = drm_gem_shmem_object_unpin,
> + .get_sg_table = drm_gem_shmem_object_get_sg_table,
> + .vmap = drm_gem_shmem_object_vmap,
> + .vunmap = drm_gem_shmem_object_vunmap,
> + .mmap = panthor_gem_mmap,
> + .export = panthor_gem_prime_export,
> + .vm_ops = &drm_gem_shmem_vm_ops,
> +};
> +
> +/**
> + * panthor_gem_create_object - Implementation of driver->gem_create_object.
> + * @ddev: DRM device
> + * @size: Size in bytes of the memory the object will reference
> + *
> + * This lets the GEM helpers allocate object structs for us, and keep
> + * our BO stats correct.
> + */
> +struct drm_gem_object *panthor_gem_create_object(struct drm_device *ddev, size_t size)
> +{
> + struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
> + struct panthor_gem_object *obj;
> +
> + obj = kzalloc(sizeof(*obj), GFP_KERNEL);
> + if (!obj)
> + return ERR_PTR(-ENOMEM);
> +
> + obj->base.base.funcs = &panthor_gem_funcs;
> + obj->base.map_wc = !ptdev->coherent;
> + mutex_init(&obj->gpuva_list_lock);
> + drm_gem_gpuva_set_lock(&obj->base.base, &obj->gpuva_list_lock);
> +
> + return &obj->base.base;
> +}
> +
> +/**
> + * panthor_gem_create_with_handle() - Create a GEM object and attach it to a handle.
> + * @file: DRM file.
> + * @ddev: DRM device.
> + * @exclusive_vm: Exclusive VM. Not NULL if the GEM object can't be shared.
> + * @size: Size of the GEM object to allocate.
> + * @flags: Combination of drm_panthor_bo_flags flags.
> + * @handle: Pointer holding the handle pointing to the new GEM object.
> + *
> + * Return: Zero on success
> + */
> +int
> +panthor_gem_create_with_handle(struct drm_file *file,
> + struct drm_device *ddev,
> + struct panthor_vm *exclusive_vm,
> + size_t size,
> + u32 flags, u32 *handle)
> +{
> + int ret;
> + struct drm_gem_shmem_object *shmem;
> + struct panthor_gem_object *bo;
> +
> + shmem = drm_gem_shmem_create(ddev, size);
> + if (IS_ERR(shmem))
> + return PTR_ERR(shmem);
> +
> + bo = to_panthor_bo(&shmem->base);
> + bo->flags = flags;
> +
> + if (exclusive_vm) {
> + bo->exclusive_vm_root_gem = panthor_vm_root_gem(exclusive_vm);
> + drm_gem_object_get(bo->exclusive_vm_root_gem);
> + bo->base.base.resv = bo->exclusive_vm_root_gem->resv;
> + }
> +
> + /*
> + * Allocate an id of idr table where the obj is registered
> + * and handle has the id what user can see.
> + */
> + ret = drm_gem_handle_create(file, &shmem->base, handle);
> + /* drop reference from allocate - handle holds it now. */
> + drm_gem_object_put(&shmem->base);
> +
> + return ret;
> +}
> diff --git a/drivers/gpu/drm/panthor/panthor_gem.h b/drivers/gpu/drm/panthor/panthor_gem.h
> new file mode 100644
> index 000000000000..6c8010ceb641
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_gem.h
> @@ -0,0 +1,144 @@
> +/* SPDX-License-Identifier: GPL-2.0 or MIT */
> +/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
> +/* Copyright 2023 Collabora ltd. */
> +
> +#ifndef __PANTHOR_GEM_H__
> +#define __PANTHOR_GEM_H__
> +
> +#include <drm/drm_gem_shmem_helper.h>
> +#include <drm/drm_mm.h>
> +
> +#include <linux/iosys-map.h>
> +#include <linux/rwsem.h>
> +
> +struct panthor_vm;
> +
> +/**
> + * struct panthor_gem_object - Driver specific GEM object.
> + */
> +struct panthor_gem_object {
> + /** @base: Inherit from drm_gem_shmem_object. */
> + struct drm_gem_shmem_object base;
> +
> + /**
> + * @exclusive_vm_root_gem: Root GEM of the exclusive VM this GEM object
> + * is attached to.
> + *
> + * If @exclusive_vm_root_gem != NULL, any attempt to bind the GEM to a
> + * different VM will fail.
> + *
> + * All FW memory objects have this field set to the root GEM of the MCU
> + * VM.
> + */
> + struct drm_gem_object *exclusive_vm_root_gem;
> +
> + /**
> + * @gpuva_list_lock: Custom GPUVA lock.
> + *
> + * Used to protect insertion of drm_gpuva elements to the
> + * drm_gem_object.gpuva.list list.
> + *
> + * We can't use the GEM resv for that, because drm_gpuva_link() is
> + * called in a dma-signaling path, where we're not allowed to take
> + * resv locks.
> + */
> + struct mutex gpuva_list_lock;
> +
> + /** @flags: Combination of drm_panthor_bo_flags flags. */
> + u32 flags;
> +};
> +
> +/**
> + * struct panthor_kernel_bo - Kernel buffer object.
> + *
> + * These objects are only manipulated by the kernel driver and not
> + * directly exposed to the userspace. The GPU address of a kernel
> + * BO might be passed to userspace though.
> + */
> +struct panthor_kernel_bo {
> + /**
> + * @obj: The GEM object backing this kernel buffer object.
> + */
> + struct drm_gem_object *obj;
> +
> + /**
> + * @va_node: VA space allocated to this GEM.
> + */
> + struct drm_mm_node va_node;
> +
> + /**
> + * @kmap: Kernel CPU mapping of @gem.
> + */
> + void *kmap;
> +};
> +
> +static inline
> +struct panthor_gem_object *to_panthor_bo(struct drm_gem_object *obj)
> +{
> + return container_of(to_drm_gem_shmem_obj(obj), struct panthor_gem_object, base);
> +}
> +
> +struct drm_gem_object *panthor_gem_create_object(struct drm_device *ddev, size_t size);
> +
> +struct drm_gem_object *
> +panthor_gem_prime_import_sg_table(struct drm_device *ddev,
> + struct dma_buf_attachment *attach,
> + struct sg_table *sgt);
> +
> +int
> +panthor_gem_create_with_handle(struct drm_file *file,
> + struct drm_device *ddev,
> + struct panthor_vm *exclusive_vm,
> + size_t size,
> + u32 flags,
> + uint32_t *handle);
> +
> +static inline u64
> +panthor_kernel_bo_gpuva(struct panthor_kernel_bo *bo)
> +{
> + return bo->va_node.start;
> +}
> +
> +static inline size_t
> +panthor_kernel_bo_size(struct panthor_kernel_bo *bo)
> +{
> + return bo->obj->size;
> +}
> +
> +static inline int
> +panthor_kernel_bo_vmap(struct panthor_kernel_bo *bo)
> +{
> + struct iosys_map map;
> + int ret;
> +
> + if (bo->kmap)
> + return 0;
> +
> + ret = drm_gem_vmap_unlocked(bo->obj, &map);
> + if (ret)
> + return ret;
> +
> + bo->kmap = map.vaddr;
> + return 0;
> +}
> +
> +static inline void
> +panthor_kernel_bo_vunmap(struct panthor_kernel_bo *bo)
> +{
> + if (bo->kmap) {
> + struct iosys_map map = IOSYS_MAP_INIT_VADDR(bo->kmap);
> +
> + drm_gem_vunmap_unlocked(bo->obj, &map);
> + bo->kmap = NULL;
> + }
> +}
> +
> +struct panthor_kernel_bo *
> +panthor_kernel_bo_create(struct panthor_device *ptdev, struct panthor_vm *vm,
> + size_t size, u32 bo_flags, u32 vm_map_flags,
> + u64 gpu_va);
> +
> +void panthor_kernel_bo_destroy(struct panthor_vm *vm,
> + struct panthor_kernel_bo *bo);
> +
> +#endif /* __PANTHOR_GEM_H__ */
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 05/14] drm/panthor: Add GEM logical block
2024-01-22 16:30 ` [PATCH v4 05/14] drm/panthor: Add GEM " Boris Brezillon
2024-02-09 15:58 ` Steven Price
@ 2024-02-12 12:31 ` Liviu Dudau
1 sibling, 0 replies; 51+ messages in thread
From: Liviu Dudau @ 2024-02-12 12:31 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On Mon, Jan 22, 2024 at 05:30:36PM +0100, Boris Brezillon wrote:
> Anything relating to GEM object management is placed here. Nothing
> particularly interesting here, given the implementation is based on
> drm_gem_shmem_object, which is doing most of the work.
>
> v4:
> - Force kernel BOs to be GPU mapped
> - Make panthor_kernel_bo_destroy() robust against ERR/NULL BO pointers
> to simplify the call sites
>
> v3:
> - Add acks for the MIT/GPL2 relicensing
> - Provide a panthor_kernel_bo abstraction for buffer objects managed by
> the kernel (will replace panthor_fw_mem and be used everywhere we were
> using panthor_gem_create_and_map() before)
> - Adjust things to match drm_gpuvm changes
> - Change return of panthor_gem_create_with_handle() to int
>
> Co-developed-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
> Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
> Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
> ---
> drivers/gpu/drm/panthor/panthor_gem.c | 228 ++++++++++++++++++++++++++
> drivers/gpu/drm/panthor/panthor_gem.h | 144 ++++++++++++++++
> 2 files changed, 372 insertions(+)
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.h
>
> diff --git a/drivers/gpu/drm/panthor/panthor_gem.c b/drivers/gpu/drm/panthor/panthor_gem.c
> new file mode 100644
> index 000000000000..42e342fcad19
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_gem.c
> @@ -0,0 +1,228 @@
> +// SPDX-License-Identifier: GPL-2.0 or MIT
> +/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
> +/* Copyright 2023 Collabora ltd. */
> +
> +#include <linux/err.h>
> +#include <linux/slab.h>
> +#include <linux/dma-buf.h>
> +#include <linux/dma-mapping.h>
> +
> +#include <drm/panthor_drm.h>
> +
> +#include "panthor_device.h"
> +#include "panthor_gem.h"
> +#include "panthor_mmu.h"
> +
> +static void panthor_gem_free_object(struct drm_gem_object *obj)
> +{
> + struct panthor_gem_object *bo = to_panthor_bo(obj);
> + struct drm_gem_object *vm_root_gem = bo->exclusive_vm_root_gem;
> +
> + drm_gem_free_mmap_offset(&bo->base.base);
> + mutex_destroy(&bo->gpuva_list_lock);
> + drm_gem_shmem_free(&bo->base);
> + drm_gem_object_put(vm_root_gem);
> +}
> +
> +/**
> + * panthor_kernel_bo_destroy() - Destroy a kernel buffer object
> + * @vm: The VM this BO was mapped to.
> + * @bo: Kernel buffer object to destroy. If NULL or an ERR_PTR(), the destruction
> + * is skipped.
> + */
> +void panthor_kernel_bo_destroy(struct panthor_vm *vm,
> + struct panthor_kernel_bo *bo)
> +{
> + int ret;
> +
> + if (IS_ERR_OR_NULL(bo))
> + return;
> +
> + panthor_kernel_bo_vunmap(bo);
> +
> + if (drm_WARN_ON(bo->obj->dev,
> + to_panthor_bo(bo->obj)->exclusive_vm_root_gem != panthor_vm_root_gem(vm)))
> + goto out_free_bo;
> +
> + ret = panthor_vm_unmap_range(vm, bo->va_node.start,
> + panthor_kernel_bo_size(bo));
> + if (ret)
> + goto out_free_bo;
> +
> + panthor_vm_free_va(vm, &bo->va_node);
> + drm_gem_object_put(bo->obj);
> +
> +out_free_bo:
> + kfree(bo);
> +}
> +
> +/**
> + * panthor_kernel_bo_create() - Create and map a GEM object to a VM
> + * @ptdev: Device.
> + * @vm: VM to map the GEM to. If NULL, the kernel object is not GPU mapped.
> + * @size: Size of the buffer object.
> + * @bo_flags: Combination of drm_panthor_bo_flags flags.
> + * @vm_map_flags: Combination of drm_panthor_vm_bind_op_flags (only those
> + * that are related to map operations).
> + * @gpu_va: GPU address assigned when mapping to the VM.
> + * If gpu_va == PANTHOR_VM_KERNEL_AUTO_VA, the virtual address will be
> + * automatically allocated.
> + *
> + * Return: A valid pointer in case of success, an ERR_PTR() otherwise.
> + */
> +struct panthor_kernel_bo *
> +panthor_kernel_bo_create(struct panthor_device *ptdev, struct panthor_vm *vm,
> + size_t size, u32 bo_flags, u32 vm_map_flags,
> + u64 gpu_va)
> +{
> + struct drm_gem_shmem_object *obj;
> + struct panthor_kernel_bo *kbo;
> + struct panthor_gem_object *bo;
> + int ret;
> +
> + if (drm_WARN_ON(&ptdev->base, !vm))
> + return ERR_PTR(-EINVAL);
> +
> + kbo = kzalloc(sizeof(*kbo), GFP_KERNEL);
> + if (!kbo)
> + return ERR_PTR(-ENOMEM);
> +
> + obj = drm_gem_shmem_create(&ptdev->base, size);
> + if (IS_ERR(obj)) {
> + ret = PTR_ERR(obj);
> + goto err_free_bo;
> + }
> +
> + bo = to_panthor_bo(&obj->base);
> + size = obj->base.size;
> + kbo->obj = &obj->base;
> + bo->flags = bo_flags;
> +
> + ret = panthor_vm_alloc_va(vm, gpu_va, size, &kbo->va_node);
> + if (ret)
> + goto err_put_obj;
> +
> + ret = panthor_vm_map_bo_range(vm, bo, 0, size, kbo->va_node.start, vm_map_flags);
> + if (ret)
> + goto err_free_va;
> +
> + bo->exclusive_vm_root_gem = panthor_vm_root_gem(vm);
> + drm_gem_object_get(bo->exclusive_vm_root_gem);
> + bo->base.base.resv = bo->exclusive_vm_root_gem->resv;
> + return kbo;
> +
> +err_free_va:
> + panthor_vm_free_va(vm, &kbo->va_node);
> +
> +err_put_obj:
> + drm_gem_object_put(&obj->base);
> +
> +err_free_bo:
> + kfree(kbo);
> + return ERR_PTR(ret);
> +}
> +
> +static int panthor_gem_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
> +{
> + struct panthor_gem_object *bo = to_panthor_bo(obj);
> +
> + /* Don't allow mmap on objects that have the NO_MMAP flag set. */
> + if (bo->flags & DRM_PANTHOR_BO_NO_MMAP)
> + return -EINVAL;
> +
> + return drm_gem_shmem_object_mmap(obj, vma);
> +}
> +
> +static struct dma_buf *
> +panthor_gem_prime_export(struct drm_gem_object *obj, int flags)
> +{
> + /* We can't export GEMs that have an exclusive VM. */
> + if (to_panthor_bo(obj)->exclusive_vm_root_gem)
> + return ERR_PTR(-EINVAL);
> +
> + return drm_gem_prime_export(obj, flags);
> +}
> +
> +static const struct drm_gem_object_funcs panthor_gem_funcs = {
> + .free = panthor_gem_free_object,
> + .print_info = drm_gem_shmem_object_print_info,
> + .pin = drm_gem_shmem_object_pin,
> + .unpin = drm_gem_shmem_object_unpin,
> + .get_sg_table = drm_gem_shmem_object_get_sg_table,
> + .vmap = drm_gem_shmem_object_vmap,
> + .vunmap = drm_gem_shmem_object_vunmap,
> + .mmap = panthor_gem_mmap,
> + .export = panthor_gem_prime_export,
> + .vm_ops = &drm_gem_shmem_vm_ops,
> +};
> +
> +/**
> + * panthor_gem_create_object - Implementation of driver->gem_create_object.
> + * @ddev: DRM device
> + * @size: Size in bytes of the memory the object will reference
> + *
> + * This lets the GEM helpers allocate object structs for us, and keep
> + * our BO stats correct.
> + */
> +struct drm_gem_object *panthor_gem_create_object(struct drm_device *ddev, size_t size)
> +{
> + struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
> + struct panthor_gem_object *obj;
> +
> + obj = kzalloc(sizeof(*obj), GFP_KERNEL);
> + if (!obj)
> + return ERR_PTR(-ENOMEM);
> +
> + obj->base.base.funcs = &panthor_gem_funcs;
> + obj->base.map_wc = !ptdev->coherent;
> + mutex_init(&obj->gpuva_list_lock);
> + drm_gem_gpuva_set_lock(&obj->base.base, &obj->gpuva_list_lock);
> +
> + return &obj->base.base;
> +}
> +
> +/**
> + * panthor_gem_create_with_handle() - Create a GEM object and attach it to a handle.
> + * @file: DRM file.
> + * @ddev: DRM device.
> + * @exclusive_vm: Exclusive VM. Not NULL if the GEM object can't be shared.
> + * @size: Size of the GEM object to allocate.
> + * @flags: Combination of drm_panthor_bo_flags flags.
> + * @handle: Pointer holding the handle pointing to the new GEM object.
> + *
> + * Return: Zero on success
> + */
> +int
> +panthor_gem_create_with_handle(struct drm_file *file,
> + struct drm_device *ddev,
> + struct panthor_vm *exclusive_vm,
> + size_t size,
> + u32 flags, u32 *handle)
> +{
> + int ret;
> + struct drm_gem_shmem_object *shmem;
> + struct panthor_gem_object *bo;
> +
> + shmem = drm_gem_shmem_create(ddev, size);
> + if (IS_ERR(shmem))
> + return PTR_ERR(shmem);
> +
> + bo = to_panthor_bo(&shmem->base);
> + bo->flags = flags;
> +
> + if (exclusive_vm) {
> + bo->exclusive_vm_root_gem = panthor_vm_root_gem(exclusive_vm);
> + drm_gem_object_get(bo->exclusive_vm_root_gem);
> + bo->base.base.resv = bo->exclusive_vm_root_gem->resv;
> + }
> +
> + /*
> + * Allocate an id of idr table where the obj is registered
> + * and handle has the id what user can see.
> + */
> + ret = drm_gem_handle_create(file, &shmem->base, handle);
> + /* drop reference from allocate - handle holds it now. */
> + drm_gem_object_put(&shmem->base);
> +
> + return ret;
> +}
> diff --git a/drivers/gpu/drm/panthor/panthor_gem.h b/drivers/gpu/drm/panthor/panthor_gem.h
> new file mode 100644
> index 000000000000..6c8010ceb641
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_gem.h
> @@ -0,0 +1,144 @@
> +/* SPDX-License-Identifier: GPL-2.0 or MIT */
> +/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
> +/* Copyright 2023 Collabora ltd. */
> +
> +#ifndef __PANTHOR_GEM_H__
> +#define __PANTHOR_GEM_H__
> +
> +#include <drm/drm_gem_shmem_helper.h>
> +#include <drm/drm_mm.h>
> +
> +#include <linux/iosys-map.h>
> +#include <linux/rwsem.h>
> +
> +struct panthor_vm;
> +
> +/**
> + * struct panthor_gem_object - Driver specific GEM object.
> + */
> +struct panthor_gem_object {
> + /** @base: Inherit from drm_gem_shmem_object. */
> + struct drm_gem_shmem_object base;
> +
> + /**
> + * @exclusive_vm_root_gem: Root GEM of the exclusive VM this GEM object
> + * is attached to.
> + *
> + * If @exclusive_vm_root_gem != NULL, any attempt to bind the GEM to a
> + * different VM will fail.
> + *
> + * All FW memory objects have this field set to the root GEM of the MCU
> + * VM.
> + */
> + struct drm_gem_object *exclusive_vm_root_gem;
> +
> + /**
> + * @gpuva_list_lock: Custom GPUVA lock.
> + *
> + * Used to protect insertion of drm_gpuva elements to the
> + * drm_gem_object.gpuva.list list.
> + *
> + * We can't use the GEM resv for that, because drm_gpuva_link() is
> + * called in a dma-signaling path, where we're not allowed to take
> + * resv locks.
> + */
> + struct mutex gpuva_list_lock;
> +
> + /** @flags: Combination of drm_panthor_bo_flags flags. */
> + u32 flags;
> +};
> +
> +/**
> + * struct panthor_kernel_bo - Kernel buffer object.
> + *
> + * These objects are only manipulated by the kernel driver and not
> + * directly exposed to the userspace. The GPU address of a kernel
> + * BO might be passed to userspace though.
> + */
> +struct panthor_kernel_bo {
> + /**
> + * @obj: The GEM object backing this kernel buffer object.
> + */
> + struct drm_gem_object *obj;
> +
> + /**
> + * @va_node: VA space allocated to this GEM.
> + */
> + struct drm_mm_node va_node;
> +
> + /**
> + * @kmap: Kernel CPU mapping of @gem.
> + */
> + void *kmap;
> +};
> +
> +static inline
> +struct panthor_gem_object *to_panthor_bo(struct drm_gem_object *obj)
> +{
> + return container_of(to_drm_gem_shmem_obj(obj), struct panthor_gem_object, base);
> +}
> +
> +struct drm_gem_object *panthor_gem_create_object(struct drm_device *ddev, size_t size);
> +
> +struct drm_gem_object *
> +panthor_gem_prime_import_sg_table(struct drm_device *ddev,
> + struct dma_buf_attachment *attach,
> + struct sg_table *sgt);
> +
> +int
> +panthor_gem_create_with_handle(struct drm_file *file,
> + struct drm_device *ddev,
> + struct panthor_vm *exclusive_vm,
> + size_t size,
> + u32 flags,
> + uint32_t *handle);
> +
> +static inline u64
> +panthor_kernel_bo_gpuva(struct panthor_kernel_bo *bo)
> +{
> + return bo->va_node.start;
> +}
> +
> +static inline size_t
> +panthor_kernel_bo_size(struct panthor_kernel_bo *bo)
> +{
> + return bo->obj->size;
> +}
> +
> +static inline int
> +panthor_kernel_bo_vmap(struct panthor_kernel_bo *bo)
> +{
> + struct iosys_map map;
> + int ret;
> +
> + if (bo->kmap)
> + return 0;
> +
> + ret = drm_gem_vmap_unlocked(bo->obj, &map);
> + if (ret)
> + return ret;
> +
> + bo->kmap = map.vaddr;
> + return 0;
> +}
> +
> +static inline void
> +panthor_kernel_bo_vunmap(struct panthor_kernel_bo *bo)
> +{
> + if (bo->kmap) {
> + struct iosys_map map = IOSYS_MAP_INIT_VADDR(bo->kmap);
> +
> + drm_gem_vunmap_unlocked(bo->obj, &map);
> + bo->kmap = NULL;
> + }
> +}
> +
> +struct panthor_kernel_bo *
> +panthor_kernel_bo_create(struct panthor_device *ptdev, struct panthor_vm *vm,
> + size_t size, u32 bo_flags, u32 vm_map_flags,
> + u64 gpu_va);
> +
> +void panthor_kernel_bo_destroy(struct panthor_vm *vm,
> + struct panthor_kernel_bo *bo);
> +
> +#endif /* __PANTHOR_GEM_H__ */
> --
> 2.43.0
>
--
====================
| I would like to |
| fix the world, |
| but they're not |
| giving me the |
\ source code! /
---------------
¯\_(ツ)_/¯
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 06/14] drm/panthor: Add the devfreq logical block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (4 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 05/14] drm/panthor: Add GEM " Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-01-22 16:30 ` [PATCH v4 07/14] drm/panthor: Add the MMU/VM " Boris Brezillon
` (11 subsequent siblings)
17 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
Every thing related to devfreq in placed in panthor_devfreq.c, and
helpers that can be called by other logical blocks are exposed through
panthor_devfreq.h.
This implementation is loosely based on the panfrost implementation,
the only difference being that we don't count device users, because
the idle/active state will be managed by the scheduler logic.
v4:
- Add Clément's A-b for the relicensing
v3:
- Add acks for the MIT/GPL2 relicensing
v2:
- Added in v2
Cc: Clément Péron <peron.clem@gmail.com> # MIT+GPL2 relicensing
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
Acked-by: Clément Péron <peron.clem@gmail.com> # MIT+GPL2 relicensing
---
drivers/gpu/drm/panthor/panthor_devfreq.c | 283 ++++++++++++++++++++++
drivers/gpu/drm/panthor/panthor_devfreq.h | 25 ++
2 files changed, 308 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_devfreq.c
create mode 100644 drivers/gpu/drm/panthor/panthor_devfreq.h
diff --git a/drivers/gpu/drm/panthor/panthor_devfreq.c b/drivers/gpu/drm/panthor/panthor_devfreq.c
new file mode 100644
index 000000000000..dd28b15337d4
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_devfreq.c
@@ -0,0 +1,283 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2019 Collabora ltd. */
+
+#include <linux/clk.h>
+#include <linux/devfreq.h>
+#include <linux/devfreq_cooling.h>
+#include <linux/platform_device.h>
+#include <linux/pm_opp.h>
+
+#include <drm/drm_managed.h>
+
+#include "panthor_device.h"
+#include "panthor_devfreq.h"
+
+/**
+ * struct panthor_devfreq - Device frequency management
+ */
+struct panthor_devfreq {
+ /** @devfreq: devfreq device. */
+ struct devfreq *devfreq;
+
+ /** @gov_data: Governor data. */
+ struct devfreq_simple_ondemand_data gov_data;
+
+ /** @busy_time: Busy time. */
+ ktime_t busy_time;
+
+ /** @idle_time: Idle time. */
+ ktime_t idle_time;
+
+ /** @time_last_update: Last update time. */
+ ktime_t time_last_update;
+
+ /** @last_busy_state: True if the GPU was busy last time we updated the state. */
+ bool last_busy_state;
+
+ /*
+ * @lock: Lock used to protect busy_time, idle_time, time_last_update and
+ * last_busy_state.
+ *
+ * These fields can be accessed concurrently by panthor_devfreq_get_dev_status()
+ * and panthor_devfreq_record_{busy,idle}().
+ */
+ spinlock_t lock;
+};
+
+static void panthor_devfreq_update_utilization(struct panthor_devfreq *pdevfreq)
+{
+ ktime_t now, last;
+
+ now = ktime_get();
+ last = pdevfreq->time_last_update;
+
+ if (pdevfreq->last_busy_state)
+ pdevfreq->busy_time += ktime_sub(now, last);
+ else
+ pdevfreq->idle_time += ktime_sub(now, last);
+
+ pdevfreq->time_last_update = now;
+}
+
+static int panthor_devfreq_target(struct device *dev, unsigned long *freq,
+ u32 flags)
+{
+ struct dev_pm_opp *opp;
+
+ opp = devfreq_recommended_opp(dev, freq, flags);
+ if (IS_ERR(opp))
+ return PTR_ERR(opp);
+ dev_pm_opp_put(opp);
+
+ return dev_pm_opp_set_rate(dev, *freq);
+}
+
+static void panthor_devfreq_reset(struct panthor_devfreq *pdevfreq)
+{
+ pdevfreq->busy_time = 0;
+ pdevfreq->idle_time = 0;
+ pdevfreq->time_last_update = ktime_get();
+}
+
+static int panthor_devfreq_get_dev_status(struct device *dev,
+ struct devfreq_dev_status *status)
+{
+ struct panthor_device *ptdev = dev_get_drvdata(dev);
+ struct panthor_devfreq *pdevfreq = ptdev->devfreq;
+ unsigned long irqflags;
+
+ status->current_frequency = clk_get_rate(ptdev->clks.core);
+
+ spin_lock_irqsave(&pdevfreq->lock, irqflags);
+
+ panthor_devfreq_update_utilization(pdevfreq);
+
+ status->total_time = ktime_to_ns(ktime_add(pdevfreq->busy_time,
+ pdevfreq->idle_time));
+
+ status->busy_time = ktime_to_ns(pdevfreq->busy_time);
+
+ panthor_devfreq_reset(pdevfreq);
+
+ spin_unlock_irqrestore(&pdevfreq->lock, irqflags);
+
+ drm_dbg(&ptdev->base, "busy %lu total %lu %lu %% freq %lu MHz\n",
+ status->busy_time, status->total_time,
+ status->busy_time / (status->total_time / 100),
+ status->current_frequency / 1000 / 1000);
+
+ return 0;
+}
+
+static struct devfreq_dev_profile panthor_devfreq_profile = {
+ .timer = DEVFREQ_TIMER_DELAYED,
+ .polling_ms = 50, /* ~3 frames */
+ .target = panthor_devfreq_target,
+ .get_dev_status = panthor_devfreq_get_dev_status,
+};
+
+int panthor_devfreq_init(struct panthor_device *ptdev)
+{
+ /* There's actually 2 regulators (mali and sram), but the OPP core only
+ * supports one.
+ *
+ * We assume the sram regulator is coupled with the mali one and let
+ * the coupling logic deal with voltage updates.
+ */
+ static const char * const reg_names[] = { "mali", NULL };
+ struct thermal_cooling_device *cooling;
+ struct device *dev = ptdev->base.dev;
+ struct panthor_devfreq *pdevfreq;
+ struct dev_pm_opp *opp;
+ unsigned long cur_freq;
+ int ret;
+
+ pdevfreq = drmm_kzalloc(&ptdev->base, sizeof(*ptdev->devfreq), GFP_KERNEL);
+ if (!pdevfreq)
+ return -ENOMEM;
+
+ ptdev->devfreq = pdevfreq;
+
+ ret = devm_pm_opp_set_regulators(dev, reg_names);
+ if (ret) {
+ if (ret != -EPROBE_DEFER)
+ DRM_DEV_ERROR(dev, "Couldn't set OPP regulators\n");
+
+ return ret;
+ }
+
+ ret = devm_pm_opp_of_add_table(dev);
+ if (ret)
+ return ret;
+
+ spin_lock_init(&pdevfreq->lock);
+
+ panthor_devfreq_reset(pdevfreq);
+
+ cur_freq = clk_get_rate(ptdev->clks.core);
+
+ opp = devfreq_recommended_opp(dev, &cur_freq, 0);
+ if (IS_ERR(opp))
+ return PTR_ERR(opp);
+
+ panthor_devfreq_profile.initial_freq = cur_freq;
+
+ /* Regulator coupling only takes care of synchronizing/balancing voltage
+ * updates, but the coupled regulator needs to be enabled manually.
+ *
+ * We use devm_regulator_get_enable_optional() and keep the sram supply
+ * enabled until the device is removed, just like we do for the mali
+ * supply, which is enabled when dev_pm_opp_set_opp(dev, opp) is called,
+ * and disabled when the opp_table is torn down, using the devm action.
+ *
+ * If we really care about disabling regulators on suspend, we should:
+ * - use devm_regulator_get_optional() here
+ * - call dev_pm_opp_set_opp(dev, NULL) before leaving this function
+ * (this disables the regulator passed to the OPP layer)
+ * - call dev_pm_opp_set_opp(dev, NULL) and
+ * regulator_disable(ptdev->regulators.sram) in
+ * panthor_devfreq_suspend()
+ * - call dev_pm_opp_set_opp(dev, default_opp) and
+ * regulator_enable(ptdev->regulators.sram) in
+ * panthor_devfreq_resume()
+ *
+ * But without knowing if it's beneficial or not (in term of power
+ * consumption), or how much it slows down the suspend/resume steps,
+ * let's just keep regulators enabled for the device lifetime.
+ */
+ ret = devm_regulator_get_enable_optional(dev, "sram");
+ if (ret && ret != -ENODEV) {
+ if (ret != -EPROBE_DEFER)
+ DRM_DEV_ERROR(dev, "Couldn't retrieve/enable sram supply\n");
+ return ret;
+ }
+
+ /*
+ * Set the recommend OPP this will enable and configure the regulator
+ * if any and will avoid a switch off by regulator_late_cleanup()
+ */
+ ret = dev_pm_opp_set_opp(dev, opp);
+ if (ret) {
+ DRM_DEV_ERROR(dev, "Couldn't set recommended OPP\n");
+ return ret;
+ }
+
+ dev_pm_opp_put(opp);
+
+ /*
+ * Setup default thresholds for the simple_ondemand governor.
+ * The values are chosen based on experiments.
+ */
+ pdevfreq->gov_data.upthreshold = 45;
+ pdevfreq->gov_data.downdifferential = 5;
+
+ pdevfreq->devfreq = devm_devfreq_add_device(dev, &panthor_devfreq_profile,
+ DEVFREQ_GOV_SIMPLE_ONDEMAND,
+ &pdevfreq->gov_data);
+ if (IS_ERR(pdevfreq->devfreq)) {
+ DRM_DEV_ERROR(dev, "Couldn't initialize GPU devfreq\n");
+ ret = PTR_ERR(pdevfreq->devfreq);
+ pdevfreq->devfreq = NULL;
+ return ret;
+ }
+
+ cooling = devfreq_cooling_em_register(pdevfreq->devfreq, NULL);
+ if (IS_ERR(cooling))
+ DRM_DEV_INFO(dev, "Failed to register cooling device\n");
+
+ return 0;
+}
+
+int panthor_devfreq_resume(struct panthor_device *ptdev)
+{
+ struct panthor_devfreq *pdevfreq = ptdev->devfreq;
+
+ if (!pdevfreq->devfreq)
+ return 0;
+
+ panthor_devfreq_reset(pdevfreq);
+
+ return devfreq_resume_device(pdevfreq->devfreq);
+}
+
+int panthor_devfreq_suspend(struct panthor_device *ptdev)
+{
+ struct panthor_devfreq *pdevfreq = ptdev->devfreq;
+
+ if (!pdevfreq->devfreq)
+ return 0;
+
+ return devfreq_suspend_device(pdevfreq->devfreq);
+}
+
+void panthor_devfreq_record_busy(struct panthor_device *ptdev)
+{
+ struct panthor_devfreq *pdevfreq = ptdev->devfreq;
+ unsigned long irqflags;
+
+ if (!pdevfreq->devfreq)
+ return;
+
+ spin_lock_irqsave(&pdevfreq->lock, irqflags);
+
+ panthor_devfreq_update_utilization(pdevfreq);
+ pdevfreq->last_busy_state = true;
+
+ spin_unlock_irqrestore(&pdevfreq->lock, irqflags);
+}
+
+void panthor_devfreq_record_idle(struct panthor_device *ptdev)
+{
+ struct panthor_devfreq *pdevfreq = ptdev->devfreq;
+ unsigned long irqflags;
+
+ if (!pdevfreq->devfreq)
+ return;
+
+ spin_lock_irqsave(&pdevfreq->lock, irqflags);
+
+ panthor_devfreq_update_utilization(pdevfreq);
+ pdevfreq->last_busy_state = false;
+
+ spin_unlock_irqrestore(&pdevfreq->lock, irqflags);
+}
diff --git a/drivers/gpu/drm/panthor/panthor_devfreq.h b/drivers/gpu/drm/panthor/panthor_devfreq.h
new file mode 100644
index 000000000000..875fbb5a1c1b
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_devfreq.h
@@ -0,0 +1,25 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2019 Collabora ltd. */
+
+#ifndef __PANTHOR_DEVFREQ_H__
+#define __PANTHOR_DEVFREQ_H__
+
+#include <linux/devfreq.h>
+#include <linux/spinlock.h>
+#include <linux/ktime.h>
+
+struct devfreq;
+struct thermal_cooling_device;
+
+struct panthor_device;
+struct panthor_devfreq;
+
+int panthor_devfreq_init(struct panthor_device *ptdev);
+
+int panthor_devfreq_resume(struct panthor_device *ptdev);
+int panthor_devfreq_suspend(struct panthor_device *ptdev);
+
+void panthor_devfreq_record_busy(struct panthor_device *ptdev);
+void panthor_devfreq_record_idle(struct panthor_device *ptdev);
+
+#endif /* __PANTHOR_DEVFREQ_H__ */
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* [PATCH v4 07/14] drm/panthor: Add the MMU/VM logical block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (5 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 06/14] drm/panthor: Add the devfreq " Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-02-09 16:51 ` Steven Price
2024-01-22 16:30 ` [PATCH v4 08/14] drm/panthor: Add the FW " Boris Brezillon
` (10 subsequent siblings)
17 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
MMU and VM management is related and placed in the same source file.
Page table updates are delegated to the io-pgtable-arm driver that's in
the iommu subsystem.
The VM management logic is based on drm_gpuva_mgr, and is assuming the
VA space is mostly managed by the usermode driver, except for a reserved
portion of this VA-space that's used for kernel objects (like the heap
contexts/chunks).
Both asynchronous and synchronous VM operations are supported, and
internal helpers are exposed to allow other logical blocks to map their
buffers in the GPU VA space.
There's one VM_BIND queue per-VM (meaning the Vulkan driver can only
expose one sparse-binding queue), and this bind queue is managed with
a 1:1 drm_sched_entity:drm_gpu_scheduler, such that each VM gets its own
independent execution queue, avoiding VM operation serialization at the
device level (things are still serialized at the VM level).
The rest is just implementation details that are hopefully well explained
in the documentation.
v4:
- Add an helper to return the VM state
- Check drmm_mutex_init() return code
- Remove the VM from the AS reclaim list when panthor_vm_active() is
called
- Count the number of active VM users instead of considering there's
at most one user (several scheduling groups can point to the same
vM)
- Pre-allocate a VMA object for unmap operations (unmaps can trigger
a sm_step_remap() call)
- Check vm->root_page_table instead of vm->pgtbl_ops to detect if
the io-pgtable is trying to allocate the root page table
- Don't memset() the va_node in panthor_vm_alloc_va(), make it a
caller requirement
- Fix the kernel doc in a few places
- Drop the panthor_vm::base offset constraint and modify
panthor_vm_put() to explicitly check for a NULL value
- Fix unbalanced vm_bo refcount in panthor_gpuva_sm_step_remap()
- Drop stale comments about the shared_bos list
- Patch mmu_features::va_bits on 32-bit builds to reflect the
io_pgtable limitation and let the UMD know about it
v3:
- Add acks for the MIT/GPL2 relicensing
- Propagate MMU faults to the scheduler
- Move pages pinning/unpinning out of the dma_signalling path
- Fix 32-bit support
- Rework the user/kernel VA range calculation
- Make the auto-VA range explicit (auto-VA range doesn't cover the full
kernel-VA range on the MCU VM)
- Let callers of panthor_vm_alloc_va() allocate the drm_mm_node
(embedded in panthor_kernel_bo now)
- Adjust things to match the latest drm_gpuvm changes (extobj tracking,
resv prep and more)
- Drop the per-AS lock and use slots_lock (fixes a race on vm->as.id)
- Set as.id to -1 when reusing an address space from the LRU list
- Drop misleading comment about page faults
- Remove check for irq being assigned in panthor_mmu_unplug()
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
---
drivers/gpu/drm/panthor/panthor_mmu.c | 2760 +++++++++++++++++++++++++
drivers/gpu/drm/panthor/panthor_mmu.h | 102 +
2 files changed, 2862 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.c
create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.h
diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
new file mode 100644
index 000000000000..d3ce29cd0662
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_mmu.c
@@ -0,0 +1,2760 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
+/* Copyright 2023 Collabora ltd. */
+
+#include <drm/drm_debugfs.h>
+#include <drm/drm_drv.h>
+#include <drm/drm_exec.h>
+#include <drm/drm_gpuvm.h>
+#include <drm/drm_managed.h>
+#include <drm/gpu_scheduler.h>
+#include <drm/panthor_drm.h>
+
+#include <linux/atomic.h>
+#include <linux/bitfield.h>
+#include <linux/delay.h>
+#include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/iopoll.h>
+#include <linux/io-pgtable.h>
+#include <linux/iommu.h>
+#include <linux/kmemleak.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+#include <linux/rwsem.h>
+#include <linux/sched.h>
+#include <linux/shmem_fs.h>
+#include <linux/sizes.h>
+
+#include "panthor_device.h"
+#include "panthor_heap.h"
+#include "panthor_mmu.h"
+#include "panthor_sched.h"
+#include "panthor_gem.h"
+#include "panthor_regs.h"
+
+#define MAX_AS_SLOTS 32
+
+struct panthor_vm;
+
+/**
+ * struct panthor_as_slot - Address space slot
+ */
+struct panthor_as_slot {
+ /** @vm: VM bound to this slot. NULL is no VM is bound. */
+ struct panthor_vm *vm;
+};
+
+/**
+ * struct panthor_mmu - MMU related data
+ */
+struct panthor_mmu {
+ /** @irq: The MMU irq. */
+ struct panthor_irq irq;
+
+ /** @as: Address space related fields.
+ *
+ * The GPU has a limited number of address spaces (AS) slots, forcing
+ * us to re-assign them to re-assign slots on-demand.
+ */
+ struct {
+ /** @slots_lock: Lock protecting access to all other AS fields. */
+ struct mutex slots_lock;
+
+ /** @alloc_mask: Bitmask encoding the allocated slots. */
+ unsigned long alloc_mask;
+
+ /** @faulty_mask: Bitmask encoding the faulty slots. */
+ unsigned long faulty_mask;
+
+ /** @slots: VMs currently bound to the AS slots. */
+ struct panthor_as_slot slots[MAX_AS_SLOTS];
+
+ /**
+ * @lru_list: List of least recently used VMs.
+ *
+ * We use this list to pick a VM to evict when all slots are
+ * used.
+ *
+ * There should be no more active VMs than there are AS slots,
+ * so this LRU is just here to keep VMs bound until there's
+ * a need to release a slot, thus avoid unnecessary TLB/cache
+ * flushes.
+ */
+ struct list_head lru_list;
+ } as;
+
+ /** @vm: VMs management fields */
+ struct {
+ /** @lock: Lock protecting access to list. */
+ struct mutex lock;
+
+ /** @list: List containing all VMs. */
+ struct list_head list;
+
+ /** @reset_in_progress: True if a reset is in progress. */
+ bool reset_in_progress;
+
+ /** @wq: Workqueue used for the VM_BIND queues. */
+ struct workqueue_struct *wq;
+ } vm;
+};
+
+/**
+ * struct panthor_vm_pool - VM pool object
+ */
+struct panthor_vm_pool {
+ /** @xa: Array used for VM handle tracking. */
+ struct xarray xa;
+};
+
+/**
+ * struct panthor_vma - GPU mapping object
+ *
+ * This is used to track GEM mappings in GPU space.
+ */
+struct panthor_vma {
+ /** @base: Inherits from drm_gpuva. */
+ struct drm_gpuva base;
+
+ /** @node: Used to implement deferred release of VMAs. */
+ struct list_head node;
+
+ /**
+ * @flags: Combination of drm_panthor_vm_bind_op_flags.
+ *
+ * Only map related flags are accepted.
+ */
+ u32 flags;
+};
+
+/**
+ * struct panthor_vm_op_ctx - VM operation context
+ *
+ * With VM operations potentially taking place in a dma-signaling path, we
+ * need to make sure everything that might require resource allocation is
+ * pre-allocated upfront. This is what this operation context is far.
+ *
+ * We also collect resources that have been freed, so we can release them
+ * asynchronously, and let the VM_BIND scheduler process the next VM_BIND
+ * request.
+ */
+struct panthor_vm_op_ctx {
+ /** @rsvd_page_tables: Pages reserved for the MMU page table update. */
+ struct {
+ /** @count: Number of pages reserved. */
+ u32 count;
+
+ /** @ptr: Point to the first unused page in the @pages table. */
+ u32 ptr;
+
+ /**
+ * @page: Array of pages that can be used for an MMU page table update.
+ *
+ * After an VM operation, there might be free pages left in this array.
+ * They should be returned to the pt_cache as part of the op_ctx cleanup.
+ */
+ void **pages;
+ } rsvd_page_tables;
+
+ /**
+ * @preallocated_vmas: Pre-allocated VMAs to handle the remap case.
+ *
+ * Partial unmap requests or map requests overlapping existing mappings will
+ * trigger a remap call, which need to register up to three panthor_vma objects
+ * (one for the new mapping, and two for the previous and next mappings).
+ */
+ struct panthor_vma *preallocated_vmas[3];
+
+ /** @flags: Combination of drm_panthor_vm_bind_op_flags. */
+ u32 flags;
+
+ /** @va: Virtual range targeted by the VM operation. */
+ struct {
+ /** @addr: Start address. */
+ u64 addr;
+
+ /** @range: Range size. */
+ u64 range;
+ } va;
+
+ /**
+ * @returned_vmas: List of panthor_vma objects returned after a VM operation.
+ *
+ * For unmap operations, this will contain all VMAs that were covered by the
+ * specified VA range.
+ *
+ * For map operations, this will contain all VMAs that previously mapped to
+ * the specified VA range.
+ *
+ * Those VMAs, and the resources they point to will be released as part of
+ * the op_ctx cleanup operation.
+ */
+ struct list_head returned_vmas;
+
+ /** @map: Fields specific to a map operation. */
+ struct {
+ /** @vm_bo: Buffer object to map. */
+ struct drm_gpuvm_bo *vm_bo;
+
+ /** @bo_offset: Offset in the buffer object. */
+ u64 bo_offset;
+
+ /**
+ * @sgt: sg-table pointing to pages backing the GEM object.
+ *
+ * This is gathered at job creation time, such that we don't have
+ * to allocate in ::run_job().
+ */
+ struct sg_table *sgt;
+
+ /**
+ * @new_vma: The new VMA object that will be inserted to the VA tree.
+ */
+ struct panthor_vma *new_vma;
+ } map;
+};
+
+/**
+ * struct panthor_vm - VM object
+ *
+ * A VM is an object representing a GPU (or MCU) virtual address space.
+ * It embeds the MMU page table for this address space, a tree containing
+ * all the virtual mappings of GEM objects, and other things needed to manage
+ * the VM.
+ *
+ * Except for the MCU VM, which is managed by the kernel, all other VMs are
+ * created by userspace and mostly managed by userspace, using the
+ * %DRM_IOCTL_PANTHOR_VM_BIND ioctl.
+ *
+ * A portion of the virtual address space is reserved for kernel objects,
+ * like heap chunks, and userspace gets to decide how much of the virtual
+ * address space is left to the kernel (half of the virtual address space
+ * by default).
+ */
+struct panthor_vm {
+ /**
+ * @base: Inherit from drm_gpuvm.
+ *
+ * We delegate all the VA management to the common drm_gpuvm framework
+ * and only implement hooks to update the MMU page table.
+ */
+ struct drm_gpuvm base;
+
+ /**
+ * @sched: Scheduler used for asynchronous VM_BIND request.
+ *
+ * We use a 1:1 scheduler here.
+ */
+ struct drm_gpu_scheduler sched;
+
+ /**
+ * @entity: Scheduling entity representing the VM_BIND queue.
+ *
+ * There's currently one bind queue per VM. It doesn't make sense to
+ * allow more given the VM operations are serialized anyway.
+ */
+ struct drm_sched_entity entity;
+
+ /** @ptdev: Device. */
+ struct panthor_device *ptdev;
+
+ /** @memattr: Value to program to the AS_MEMATTR register. */
+ u64 memattr;
+
+ /** @pgtbl_ops: Page table operations. */
+ struct io_pgtable_ops *pgtbl_ops;
+
+ /** @root_page_table: Stores the root page table pointer. */
+ void *root_page_table;
+
+ /**
+ * @op_lock: Lock used to serialize operations on a VM.
+ *
+ * The serialization of jobs queued to the VM_BIND queue is already
+ * taken care of by drm_sched, but we need to serialize synchronous
+ * and asynchronous VM_BIND request. This is what this lock is for.
+ */
+ struct mutex op_lock;
+
+ /**
+ * @op_ctx: The context attached to the currently executing VM operation.
+ *
+ * NULL when no operation is in progress.
+ */
+ struct panthor_vm_op_ctx *op_ctx;
+
+ /**
+ * @mm: Memory management object representing the auto-VA/kernel-VA.
+ *
+ * Used to auto-allocate VA space for kernel-managed objects (tiler
+ * heaps, ...).
+ *
+ * For the MCU VM, this is managing the VA range that's used to map
+ * all shared interfaces.
+ *
+ * For user VMs, the range is specified by userspace, and must not
+ * exceed half of the VA space addressable.
+ */
+ struct drm_mm mm;
+
+ /** @mm_lock: Lock protecting the @mm field. */
+ struct mutex mm_lock;
+
+ /** @kernel_auto_va: Automatic VA-range for kernel BOs. */
+ struct {
+ /** @start: Start of the automatic VA-range for kernel BOs. */
+ u64 start;
+
+ /** @size: Size of the automatic VA-range for kernel BOs. */
+ u64 end;
+ } kernel_auto_va;
+
+ /** @as: Address space related fields. */
+ struct {
+ /**
+ * @id: ID of the address space this VM is bound to.
+ *
+ * A value of -1 means the VM is inactive/not bound.
+ */
+ int id;
+
+ /** @active_cnt: Number of active users of this VM. */
+ refcount_t active_cnt;
+
+ /**
+ * @lru_node: Used to instead the VM in the panthor_mmu::as::lru_list.
+ *
+ * Active VMs should not be inserted in the LRU list.
+ */
+ struct list_head lru_node;
+ } as;
+
+ /**
+ * @heaps: Tiler heap related fields.
+ */
+ struct {
+ /**
+ * @pool: The heap pool attached to this VM.
+ *
+ * Will stay NULL until someone creates a heap context on this VM.
+ */
+ struct panthor_heap_pool *pool;
+
+ /** @lock: Lock used to protect access to @pool. */
+ struct mutex lock;
+ } heaps;
+
+ /** @node: Used to insert the VM in the panthor_mmu::vm::list. */
+ struct list_head node;
+
+ /** @for_mcu: True if this is the MCU VM. */
+ bool for_mcu;
+
+ /**
+ * @destroyed: True if the VM was destroyed.
+ *
+ * No further bind requests should be queued to a destroyed VM.
+ */
+ bool destroyed;
+
+ /**
+ * @unusable: True if the VM has turned unusable because something
+ * bad happened during an asynchronous request.
+ *
+ * We don't try to recover from such failures, because this implies
+ * informing userspace about the specific operation that failed, and
+ * hoping the userspace driver can replay things from there. This all
+ * sounds very complicated for little gain.
+ *
+ * Instead, we should just flag the VM as unusable, and fail any
+ * further request targeting this VM.
+ *
+ * We also provide a way to query a VM state, so userspace can destroy
+ * it and create a new one.
+ *
+ * As an analogy, this would be mapped to a VK_ERROR_DEVICE_LOST
+ * situation, where the logical device needs to be re-created.
+ */
+ bool unusable;
+
+ /**
+ * @unhandled_fault: Unhandled fault happened.
+ *
+ * This should be reported to the scheduler, and the queue/group be
+ * flagged as faulty as a result.
+ */
+ bool unhandled_fault;
+};
+
+/**
+ * struct panthor_vm_bind_job - VM bind job
+ */
+struct panthor_vm_bind_job {
+ /** @base: Inherit from drm_sched_job. */
+ struct drm_sched_job base;
+
+ /** @refcount: Reference count. */
+ struct kref refcount;
+
+ /** @cleanup_op_ctx_work: Work used to cleanup the VM operation context. */
+ struct work_struct cleanup_op_ctx_work;
+
+ /** @vm: VM targeted by the VM operation. */
+ struct panthor_vm *vm;
+
+ /** @ctx: Operation context. */
+ struct panthor_vm_op_ctx ctx;
+};
+
+/**
+ * @pt_cache: Cache used to allocate MMU page tables.
+ *
+ * The pre-allocation pattern forces us to over-allocate to plan for
+ * the worst case scenario, and return the pages we didn't use.
+ *
+ * Having a kmem_cache allows us to speed allocations.
+ */
+static struct kmem_cache *pt_cache;
+
+/**
+ * alloc_pt() - Custom page table allocator
+ * @cookie: Cookie passed at page table allocation time.
+ * @size: Size of the page table. This size should be fixed,
+ * and determined at creation time based on the granule size.
+ * @gfp: GFP flags.
+ *
+ * We want a custom allocator so we can use a cache for page table
+ * allocations and amortize the cost of the over-reservation that's
+ * done to allow asynchronous VM operations.
+ *
+ * Return: non-NULL on success, NULL if the allocation failed for any
+ * reason.
+ */
+static void *alloc_pt(void *cookie, size_t size, gfp_t gfp)
+{
+ struct panthor_vm *vm = cookie;
+ void *page;
+
+ /* Allocation of the root page table happening during init. */
+ if (unlikely(!vm->root_page_table)) {
+ struct page *p;
+
+ drm_WARN_ON(&vm->ptdev->base, vm->op_ctx);
+ p = alloc_pages_node(dev_to_node(vm->ptdev->base.dev),
+ gfp | __GFP_ZERO, get_order(size));
+ page = p ? page_address(p) : NULL;
+ vm->root_page_table = page;
+ return page;
+ }
+
+ /* We're not supposed to have anything bigger than 4k here, because we picked a
+ * 4k granule size at init time.
+ */
+ if (drm_WARN_ON(&vm->ptdev->base, size != SZ_4K))
+ return NULL;
+
+ /* We must have some op_ctx attached to the VM and it must have at least one
+ * free page.
+ */
+ if (drm_WARN_ON(&vm->ptdev->base, !vm->op_ctx) ||
+ drm_WARN_ON(&vm->ptdev->base,
+ vm->op_ctx->rsvd_page_tables.ptr >= vm->op_ctx->rsvd_page_tables.count))
+ return NULL;
+
+ page = vm->op_ctx->rsvd_page_tables.pages[vm->op_ctx->rsvd_page_tables.ptr++];
+ memset(page, 0, SZ_4K);
+
+ /* Page table entries don't use virtual addresses, which trips out
+ * kmemleak. kmemleak_alloc_phys() might work, but physical addresses
+ * are mixed with other fields, and I fear kmemleak won't detect that
+ * either.
+ *
+ * Let's just ignore memory passed to the page-table driver for now.
+ */
+ kmemleak_ignore(page);
+ return page;
+}
+
+/**
+ * @free_pt() - Custom page table free function
+ * @cookie: Cookie passed at page table allocation time.
+ * @data: Page table to free.
+ * @size: Size of the page table. This size should be fixed,
+ * and determined at creation time based on the granule size.
+ */
+static void free_pt(void *cookie, void *data, size_t size)
+{
+ struct panthor_vm *vm = cookie;
+
+ if (unlikely(vm->root_page_table == data)) {
+ free_pages((unsigned long)data, get_order(size));
+ vm->root_page_table = NULL;
+ return;
+ }
+
+ if (drm_WARN_ON(&vm->ptdev->base, size != SZ_4K))
+ return;
+
+ /* Return the page to the pt_cache. */
+ kmem_cache_free(pt_cache, data);
+}
+
+static int wait_ready(struct panthor_device *ptdev, u32 as_nr)
+{
+ int ret;
+ u32 val;
+
+ /* Wait for the MMU status to indicate there is no active command, in
+ * case one is pending.
+ */
+ ret = readl_relaxed_poll_timeout_atomic(ptdev->iomem + AS_STATUS(as_nr),
+ val, !(val & AS_STATUS_AS_ACTIVE),
+ 10, 100000);
+
+ if (ret) {
+ panthor_device_schedule_reset(ptdev);
+ drm_err(&ptdev->base, "AS_ACTIVE bit stuck\n");
+ }
+
+ return ret;
+}
+
+static int write_cmd(struct panthor_device *ptdev, u32 as_nr, u32 cmd)
+{
+ int status;
+
+ /* write AS_COMMAND when MMU is ready to accept another command */
+ status = wait_ready(ptdev, as_nr);
+ if (!status)
+ gpu_write(ptdev, AS_COMMAND(as_nr), cmd);
+
+ return status;
+}
+
+static void lock_region(struct panthor_device *ptdev, u32 as_nr,
+ u64 region_start, u64 size)
+{
+ u8 region_width;
+ u64 region;
+ u64 region_end = region_start + size;
+
+ if (!size)
+ return;
+
+ /*
+ * The locked region is a naturally aligned power of 2 block encoded as
+ * log2 minus(1).
+ * Calculate the desired start/end and look for the highest bit which
+ * differs. The smallest naturally aligned block must include this bit
+ * change, the desired region starts with this bit (and subsequent bits)
+ * zeroed and ends with the bit (and subsequent bits) set to one.
+ */
+ region_width = max(fls64(region_start ^ (region_end - 1)),
+ const_ilog2(AS_LOCK_REGION_MIN_SIZE)) - 1;
+
+ /*
+ * Mask off the low bits of region_start (which would be ignored by
+ * the hardware anyway)
+ */
+ region_start &= GENMASK_ULL(63, region_width);
+
+ region = region_width | region_start;
+
+ /* Lock the region that needs to be updated */
+ gpu_write(ptdev, AS_LOCKADDR_LO(as_nr), lower_32_bits(region));
+ gpu_write(ptdev, AS_LOCKADDR_HI(as_nr), upper_32_bits(region));
+ write_cmd(ptdev, as_nr, AS_COMMAND_LOCK);
+}
+
+static int mmu_hw_do_operation_locked(struct panthor_device *ptdev, int as_nr,
+ u64 iova, u64 size, u32 op)
+{
+ lockdep_assert_held(&ptdev->mmu->as.slots_lock);
+
+ if (as_nr < 0)
+ return 0;
+
+ if (op != AS_COMMAND_UNLOCK)
+ lock_region(ptdev, as_nr, iova, size);
+
+ /* Run the MMU operation */
+ write_cmd(ptdev, as_nr, op);
+
+ /* Wait for the flush to complete */
+ return wait_ready(ptdev, as_nr);
+}
+
+static int mmu_hw_do_operation(struct panthor_vm *vm,
+ u64 iova, u64 size, u32 op)
+{
+ struct panthor_device *ptdev = vm->ptdev;
+ int ret;
+
+ mutex_lock(&ptdev->mmu->as.slots_lock);
+ ret = mmu_hw_do_operation_locked(ptdev, vm->as.id, iova, size, op);
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+
+ return ret;
+}
+
+static int panthor_mmu_as_enable(struct panthor_device *ptdev, u32 as_nr,
+ u64 transtab, u64 transcfg, u64 memattr)
+{
+ int ret;
+
+ ret = mmu_hw_do_operation_locked(ptdev, as_nr, 0, ~0ULL, AS_COMMAND_FLUSH_MEM);
+ if (ret)
+ return ret;
+
+ gpu_write(ptdev, AS_TRANSTAB_LO(as_nr), lower_32_bits(transtab));
+ gpu_write(ptdev, AS_TRANSTAB_HI(as_nr), upper_32_bits(transtab));
+
+ gpu_write(ptdev, AS_MEMATTR_LO(as_nr), lower_32_bits(memattr));
+ gpu_write(ptdev, AS_MEMATTR_HI(as_nr), upper_32_bits(memattr));
+
+ gpu_write(ptdev, AS_TRANSCFG_LO(as_nr), lower_32_bits(transcfg));
+ gpu_write(ptdev, AS_TRANSCFG_HI(as_nr), upper_32_bits(transcfg));
+
+ return write_cmd(ptdev, as_nr, AS_COMMAND_UPDATE);
+}
+
+static int panthor_mmu_as_disable(struct panthor_device *ptdev, u32 as_nr)
+{
+ int ret;
+
+ ret = mmu_hw_do_operation_locked(ptdev, as_nr, 0, ~0ULL, AS_COMMAND_FLUSH_MEM);
+ if (ret)
+ return ret;
+
+ gpu_write(ptdev, AS_TRANSTAB_LO(as_nr), 0);
+ gpu_write(ptdev, AS_TRANSTAB_HI(as_nr), 0);
+
+ gpu_write(ptdev, AS_MEMATTR_LO(as_nr), 0);
+ gpu_write(ptdev, AS_MEMATTR_HI(as_nr), 0);
+
+ gpu_write(ptdev, AS_TRANSCFG_LO(as_nr), AS_TRANSCFG_ADRMODE_UNMAPPED);
+ gpu_write(ptdev, AS_TRANSCFG_HI(as_nr), 0);
+
+ return write_cmd(ptdev, as_nr, AS_COMMAND_UPDATE);
+}
+
+static u32 panthor_mmu_fault_mask(struct panthor_device *ptdev, u32 value)
+{
+ /* Bits 16 to 31 mean REQ_COMPLETE. */
+ return value & GENMASK(15, 0);
+}
+
+static u32 panthor_mmu_as_fault_mask(struct panthor_device *ptdev, u32 as)
+{
+ return BIT(as);
+}
+
+/**
+ * panthor_vm_has_unhandled_faults() - Check if a VM has unhandled faults
+ * @vm: VM to check.
+ *
+ * Return: true if the VM has unhandled faults, false otherwise.
+ */
+bool panthor_vm_has_unhandled_faults(struct panthor_vm *vm)
+{
+ return vm->unhandled_fault;
+}
+
+/**
+ * panthor_vm_is_unusable() - Check if the VM is still usable
+ * @vm: VM to check.
+ *
+ * Return: true if the VM is unusable, false otherwise.
+ */
+bool panthor_vm_is_unusable(struct panthor_vm *vm)
+{
+ return vm->unusable;
+}
+
+static void panthor_vm_release_as_locked(struct panthor_vm *vm)
+{
+ struct panthor_device *ptdev = vm->ptdev;
+
+ lockdep_assert_held(&ptdev->mmu->as.slots_lock);
+
+ if (drm_WARN_ON(&ptdev->base, vm->as.id < 0))
+ return;
+
+ ptdev->mmu->as.slots[vm->as.id].vm = NULL;
+ clear_bit(vm->as.id, &ptdev->mmu->as.alloc_mask);
+ refcount_set(&vm->as.active_cnt, 0);
+ list_del_init(&vm->as.lru_node);
+ vm->as.id = -1;
+}
+
+/**
+ * panthor_vm_active() - Flag a VM as active
+ * @VM: VM to flag as active.
+ *
+ * Assigns an address space to a VM so it can be used by the GPU/MCU.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_vm_active(struct panthor_vm *vm)
+{
+ struct panthor_device *ptdev = vm->ptdev;
+ struct io_pgtable_cfg *cfg = &io_pgtable_ops_to_pgtable(vm->pgtbl_ops)->cfg;
+ int ret = 0, as, cookie;
+ u64 transtab, transcfg;
+
+ if (!drm_dev_enter(&ptdev->base, &cookie))
+ return -ENODEV;
+
+ if (refcount_inc_not_zero(&vm->as.active_cnt))
+ goto out_dev_exit;
+
+ mutex_lock(&ptdev->mmu->as.slots_lock);
+
+ if (refcount_inc_not_zero(&vm->as.active_cnt))
+ goto out_unlock;
+
+ as = vm->as.id;
+ if (as >= 0) {
+ /* Unhandled pagefault on this AS, the MMU was disabled. We need to
+ * re-enable the MMU after clearing+unmasking the AS interrupts.
+ */
+ if (ptdev->mmu->as.faulty_mask & panthor_mmu_as_fault_mask(ptdev, as))
+ goto out_enable_as;
+
+ goto out_make_active;
+ }
+
+ /* Check for a free AS */
+ if (vm->for_mcu) {
+ drm_WARN_ON(&ptdev->base, ptdev->mmu->as.alloc_mask & BIT(0));
+ as = 0;
+ } else {
+ as = ffz(ptdev->mmu->as.alloc_mask | BIT(0));
+ }
+
+ if (!(BIT(as) & ptdev->gpu_info.as_present)) {
+ struct panthor_vm *lru_vm;
+
+ lru_vm = list_first_entry_or_null(&ptdev->mmu->as.lru_list,
+ struct panthor_vm,
+ as.lru_node);
+ if (drm_WARN_ON(&ptdev->base, !lru_vm)) {
+ ret = -EBUSY;
+ goto out_unlock;
+ }
+
+ drm_WARN_ON(&ptdev->base, refcount_read(&lru_vm->as.active_cnt));
+ as = lru_vm->as.id;
+ panthor_vm_release_as_locked(lru_vm);
+ }
+
+ /* Assign the free or reclaimed AS to the FD */
+ vm->as.id = as;
+ set_bit(as, &ptdev->mmu->as.alloc_mask);
+ ptdev->mmu->as.slots[as].vm = vm;
+
+out_enable_as:
+ transtab = cfg->arm_lpae_s1_cfg.ttbr;
+ transcfg = AS_TRANSCFG_PTW_MEMATTR_WB |
+ AS_TRANSCFG_PTW_RA |
+ AS_TRANSCFG_ADRMODE_AARCH64_4K;
+ if (ptdev->coherent)
+ transcfg |= AS_TRANSCFG_PTW_SH_OS;
+
+ /* If the VM is re-activated, we clear the fault. */
+ vm->unhandled_fault = false;
+
+ /* Unhandled pagefault on this AS, clear the fault and re-enable interrupts
+ * before enabling the AS.
+ */
+ if (ptdev->mmu->as.faulty_mask & panthor_mmu_as_fault_mask(ptdev, as)) {
+ gpu_write(ptdev, MMU_INT_CLEAR, panthor_mmu_as_fault_mask(ptdev, as));
+ ptdev->mmu->as.faulty_mask &= ~panthor_mmu_as_fault_mask(ptdev, as);
+ gpu_write(ptdev, MMU_INT_MASK, ~ptdev->mmu->as.faulty_mask);
+ }
+
+ ret = panthor_mmu_as_enable(vm->ptdev, vm->as.id, transtab, transcfg, vm->memattr);
+
+out_make_active:
+ if (!ret) {
+ refcount_set(&vm->as.active_cnt, 1);
+ list_del_init(&vm->as.lru_node);
+ }
+
+out_unlock:
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+
+out_dev_exit:
+ drm_dev_exit(cookie);
+ return ret;
+}
+
+/**
+ * panthor_vm_idle() - Flag a VM idle
+ * @VM: VM to flag as idle.
+ *
+ * When we know the GPU is done with the VM (no more jobs to process),
+ * we can relinquish the AS slot attached to this VM, if any.
+ *
+ * We don't release the slot immediately, but instead place the VM in
+ * the LRU list, so it can be evicted if another VM needs an AS slot.
+ * This way, VMs keep attached to the AS they were given until we run
+ * out of free slot, limiting the number of MMU operations (TLB flush
+ * and other AS updates).
+ */
+void panthor_vm_idle(struct panthor_vm *vm)
+{
+ struct panthor_device *ptdev = vm->ptdev;
+
+ if (!refcount_dec_and_mutex_lock(&vm->as.active_cnt, &ptdev->mmu->as.slots_lock))
+ return;
+
+ if (!drm_WARN_ON(&ptdev->base, vm->as.id == -1 || !list_empty(&vm->as.lru_node)))
+ list_add_tail(&vm->as.lru_node, &ptdev->mmu->as.lru_list);
+
+ refcount_set(&vm->as.active_cnt, 0);
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+}
+
+static void panthor_vm_stop(struct panthor_vm *vm)
+{
+ drm_sched_stop(&vm->sched, NULL);
+}
+
+static void panthor_vm_start(struct panthor_vm *vm)
+{
+ drm_sched_start(&vm->sched, true);
+}
+
+/**
+ * panthor_vm_as() - Get the AS slot attached to a VM
+ * @vm: VM to get the AS slot of.
+ *
+ * Return: -1 if the VM is not assigned an AS slot yet, >= 0 otherwise.
+ */
+int panthor_vm_as(struct panthor_vm *vm)
+{
+ return vm->as.id;
+}
+
+static size_t get_pgsize(u64 addr, size_t size, size_t *count)
+{
+ /*
+ * io-pgtable only operates on multiple pages within a single table
+ * entry, so we need to split at boundaries of the table size, i.e.
+ * the next block size up. The distance from address A to the next
+ * boundary of block size B is logically B - A % B, but in unsigned
+ * two's complement where B is a power of two we get the equivalence
+ * B - A % B == (B - A) % B == (n * B - A) % B, and choose n = 0 :)
+ */
+ size_t blk_offset = -addr % SZ_2M;
+
+ if (blk_offset || size < SZ_2M) {
+ *count = min_not_zero(blk_offset, size) / SZ_4K;
+ return SZ_4K;
+ }
+ blk_offset = -addr % SZ_1G ?: SZ_1G;
+ *count = min(blk_offset, size) / SZ_2M;
+ return SZ_2M;
+}
+
+static int panthor_vm_flush_range(struct panthor_vm *vm, u64 iova, u64 size)
+{
+ struct panthor_device *ptdev = vm->ptdev;
+ int ret = 0, cookie;
+
+ if (vm->as.id < 0)
+ return 0;
+
+ /* If the device is unplugged, we just silently skip the flush. */
+ if (!drm_dev_enter(&ptdev->base, &cookie))
+ return 0;
+
+ /* Flush the PTs only if we're already awake */
+ if (pm_runtime_active(ptdev->base.dev))
+ ret = mmu_hw_do_operation(vm, iova, size, AS_COMMAND_FLUSH_PT);
+
+ drm_dev_exit(cookie);
+ return ret;
+}
+
+static int panthor_vm_unmap_pages(struct panthor_vm *vm, u64 iova, u64 size)
+{
+ struct panthor_device *ptdev = vm->ptdev;
+ struct io_pgtable_ops *ops = vm->pgtbl_ops;
+ u64 offset = 0;
+
+ drm_dbg(&ptdev->base, "unmap: as=%d, iova=%llx, len=%llx", vm->as.id, iova, size);
+
+ while (offset < size) {
+ size_t unmapped_sz = 0, pgcount;
+ size_t pgsize = get_pgsize(iova + offset, size - offset, &pgcount);
+
+ unmapped_sz = ops->unmap_pages(ops, iova + offset, pgsize, pgcount, NULL);
+
+ if (drm_WARN_ON(&ptdev->base, unmapped_sz != pgsize * pgcount)) {
+ drm_err(&ptdev->base, "failed to unmap range %llx-%llx (requested range %llx-%llx)\n",
+ iova + offset + unmapped_sz,
+ iova + offset + pgsize * pgcount,
+ iova, iova + size);
+ panthor_vm_flush_range(vm, iova, offset + unmapped_sz);
+ return -EINVAL;
+ }
+ offset += unmapped_sz;
+ }
+
+ return panthor_vm_flush_range(vm, iova, size);
+}
+
+static int
+panthor_vm_map_pages(struct panthor_vm *vm, u64 iova, int prot,
+ struct sg_table *sgt, u64 offset, u64 size)
+{
+ struct panthor_device *ptdev = vm->ptdev;
+ unsigned int count;
+ struct scatterlist *sgl;
+ struct io_pgtable_ops *ops = vm->pgtbl_ops;
+ u64 start_iova = iova;
+ int ret;
+
+ if (!size)
+ return 0;
+
+ for_each_sgtable_dma_sg(sgt, sgl, count) {
+ dma_addr_t paddr = sg_dma_address(sgl);
+ size_t len = sg_dma_len(sgl);
+
+ if (len <= offset) {
+ offset -= len;
+ continue;
+ }
+
+ paddr -= offset;
+ len -= offset;
+
+ if (size >= 0) {
+ len = min_t(size_t, len, size);
+ size -= len;
+ }
+
+ drm_dbg(&ptdev->base, "map: as=%d, iova=%llx, paddr=%pad, len=%zx",
+ vm->as.id, iova, &paddr, len);
+
+ while (len) {
+ size_t pgcount, mapped = 0;
+ size_t pgsize = get_pgsize(iova | paddr, len, &pgcount);
+
+ ret = ops->map_pages(ops, iova, paddr, pgsize, pgcount, prot,
+ GFP_KERNEL, &mapped);
+ iova += mapped;
+ paddr += mapped;
+ len -= mapped;
+
+ if (drm_WARN_ON(&ptdev->base, !ret && !mapped))
+ ret = -ENOMEM;
+
+ if (ret) {
+ /* If something failed, unmap what we've already mapped before
+ * returning. The unmap call is not supposed to fail.
+ */
+ drm_WARN_ON(&ptdev->base,
+ panthor_vm_unmap_pages(vm, start_iova,
+ iova - start_iova));
+ return ret;
+ }
+ }
+
+ if (!size)
+ break;
+ }
+
+ return panthor_vm_flush_range(vm, start_iova, iova - start_iova);
+}
+
+static int flags_to_prot(u32 flags)
+{
+ int prot = 0;
+
+ if (flags & DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC)
+ prot |= IOMMU_NOEXEC;
+
+ if (!(flags & DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED))
+ prot |= IOMMU_CACHE;
+
+ if (flags & DRM_PANTHOR_VM_BIND_OP_MAP_READONLY)
+ prot |= IOMMU_READ;
+ else
+ prot |= IOMMU_READ | IOMMU_WRITE;
+
+ return prot;
+}
+
+/**
+ * panthor_vm_alloc_va() - Allocate a region in the auto-va space
+ * @VM: VM to allocate a region on.
+ * @va: start of the VA range. Can be PANTHOR_VM_KERNEL_AUTO_VA if the user
+ * wants the VA to be automatically allocated from the auto-VA range.
+ * @size: size of the VA range.
+ * @va_node: drm_mm_node to initialize. Must be zero-initialized.
+ *
+ * Some GPU objects, like heap chunks, are fully managed by the kernel and
+ * need to be mapped to the userspace VM, in the region reserved for kernel
+ * objects.
+ *
+ * This function takes care of allocating a region in the kernel auto-VA space.
+ *
+ * Return: 0 on success, an error code otherwise.
+ */
+int
+panthor_vm_alloc_va(struct panthor_vm *vm, u64 va, u64 size,
+ struct drm_mm_node *va_node)
+{
+ int ret;
+
+ if (!size || (size & ~PAGE_MASK))
+ return -EINVAL;
+
+ if (va != PANTHOR_VM_KERNEL_AUTO_VA && (va & ~PAGE_MASK))
+ return -EINVAL;
+
+ mutex_lock(&vm->mm_lock);
+ if (va != PANTHOR_VM_KERNEL_AUTO_VA) {
+ va_node->start = va;
+ va_node->size = size;
+ ret = drm_mm_reserve_node(&vm->mm, va_node);
+ } else {
+ ret = drm_mm_insert_node_in_range(&vm->mm, va_node, size,
+ size >= SZ_2M ? SZ_2M : SZ_4K,
+ 0, vm->kernel_auto_va.start,
+ vm->kernel_auto_va.end,
+ DRM_MM_INSERT_BEST);
+ }
+ mutex_unlock(&vm->mm_lock);
+
+ return ret;
+}
+
+/**
+ * panthor_vm_free_va() - Free a region allocated with panthor_vm_alloc_va()
+ * @VM: VM to free the region on.
+ * @va_node: Memory node representing the region to free.
+ */
+void panthor_vm_free_va(struct panthor_vm *vm, struct drm_mm_node *va_node)
+{
+ mutex_lock(&vm->mm_lock);
+ drm_mm_remove_node(va_node);
+ mutex_unlock(&vm->mm_lock);
+}
+
+static void panthor_vm_bo_put(struct drm_gpuvm_bo *vm_bo)
+{
+ struct panthor_gem_object *bo = to_panthor_bo(vm_bo->obj);
+ bool unpin;
+
+ /* We must retain the GEM before calling drm_gpuvm_bo_put(),
+ * otherwise the mutex might be destroyed while we hold it.
+ */
+ drm_gem_object_get(&bo->base.base);
+ mutex_lock(&bo->gpuva_list_lock);
+ unpin = drm_gpuvm_bo_put(vm_bo);
+ mutex_unlock(&bo->gpuva_list_lock);
+
+ /* If the vm_bo object was destroyed, release the pin reference that
+ * was hold by this object.
+ */
+ if (unpin && !bo->base.base.import_attach)
+ drm_gem_shmem_unpin(&bo->base);
+
+ drm_gem_object_put(&bo->base.base);
+}
+
+static void panthor_vm_cleanup_op_ctx(struct panthor_vm_op_ctx *op_ctx,
+ struct panthor_vm *vm)
+{
+ struct panthor_vma *vma, *tmp_vma;
+
+ u32 remaining_pt_count = op_ctx->rsvd_page_tables.count -
+ op_ctx->rsvd_page_tables.ptr;
+
+ if (remaining_pt_count) {
+ kmem_cache_free_bulk(pt_cache, remaining_pt_count,
+ op_ctx->rsvd_page_tables.pages +
+ op_ctx->rsvd_page_tables.ptr);
+ }
+
+ kfree(op_ctx->rsvd_page_tables.pages);
+ memset(&op_ctx->rsvd_page_tables, 0, sizeof(op_ctx->rsvd_page_tables));
+
+ if (op_ctx->map.vm_bo)
+ panthor_vm_bo_put(op_ctx->map.vm_bo);
+
+ memset(&op_ctx->map, 0, sizeof(op_ctx->map));
+
+ for (u32 i = 0; i < ARRAY_SIZE(op_ctx->preallocated_vmas); i++) {
+ kfree(op_ctx->preallocated_vmas[i]);
+ op_ctx->preallocated_vmas[i] = NULL;
+ }
+
+ list_for_each_entry_safe(vma, tmp_vma, &op_ctx->returned_vmas, node) {
+ list_del(&vma->node);
+ panthor_vm_bo_put(vma->base.vm_bo);
+ kfree(vma);
+ }
+}
+
+static struct panthor_vma *
+panthor_vm_op_ctx_get_vma(struct panthor_vm_op_ctx *op_ctx)
+{
+ for (u32 i = 0; i < ARRAY_SIZE(op_ctx->preallocated_vmas); i++) {
+ struct panthor_vma *vma = op_ctx->preallocated_vmas[i];
+
+ if (vma) {
+ op_ctx->preallocated_vmas[i] = NULL;
+ return vma;
+ }
+ }
+
+ return NULL;
+}
+
+static int
+panthor_vm_op_ctx_prealloc_vmas(struct panthor_vm_op_ctx *op_ctx)
+{
+ u32 vma_count;
+
+ switch (op_ctx->flags & DRM_PANTHOR_VM_BIND_OP_TYPE_MASK) {
+ case DRM_PANTHOR_VM_BIND_OP_TYPE_MAP:
+ /* One VMA for the new mapping, and two more VMAs for the remap case
+ * which might contain both a prev and next VA.
+ */
+ vma_count = 3;
+ break;
+
+ case DRM_PANTHOR_VM_BIND_OP_TYPE_UNMAP:
+ /* Partial unmaps might trigger a remap with either a prev or a next VA,
+ * but not both.
+ */
+ vma_count = 1;
+ break;
+
+ default:
+ return 0;
+ }
+
+ for (u32 i = 0; i < vma_count; i++) {
+ struct panthor_vma *vma = kzalloc(sizeof(*vma), GFP_KERNEL);
+
+ if (!vma)
+ return -ENOMEM;
+
+ op_ctx->preallocated_vmas[i] = vma;
+ }
+
+ return 0;
+}
+
+#define PANTHOR_VM_BIND_OP_MAP_FLAGS \
+ (DRM_PANTHOR_VM_BIND_OP_MAP_READONLY | \
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC | \
+ DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED | \
+ DRM_PANTHOR_VM_BIND_OP_TYPE_MASK)
+
+static int panthor_vm_prepare_map_op_ctx(struct panthor_vm_op_ctx *op_ctx,
+ struct panthor_vm *vm,
+ struct panthor_gem_object *bo,
+ u64 offset,
+ u64 size, u64 va,
+ u32 flags)
+{
+ struct drm_gpuvm_bo *preallocated_vm_bo;
+ struct sg_table *sgt = NULL;
+ u64 pt_count;
+ int ret;
+
+ if (!bo)
+ return -EINVAL;
+
+ if ((flags & ~PANTHOR_VM_BIND_OP_MAP_FLAGS) ||
+ (flags & DRM_PANTHOR_VM_BIND_OP_TYPE_MASK) != DRM_PANTHOR_VM_BIND_OP_TYPE_MAP)
+ return -EINVAL;
+
+ /* Make sure the VA and size are aligned and in-bounds. */
+ if (size > bo->base.base.size || offset > bo->base.base.size - size)
+ return -EINVAL;
+
+ /* If the BO has an exclusive VM attached, it can't be mapped to other VMs. */
+ if (bo->exclusive_vm_root_gem &&
+ bo->exclusive_vm_root_gem != panthor_vm_root_gem(vm))
+ return -EINVAL;
+
+ memset(op_ctx, 0, sizeof(*op_ctx));
+ INIT_LIST_HEAD(&op_ctx->returned_vmas);
+ op_ctx->flags = flags;
+ op_ctx->va.range = size;
+ op_ctx->va.addr = va;
+
+ ret = panthor_vm_op_ctx_prealloc_vmas(op_ctx);
+ if (ret)
+ goto err_cleanup;
+
+ if (!bo->base.base.import_attach) {
+ /* Pre-reserve the BO pages, so the map operation doesn't have to
+ * allocate.
+ */
+ ret = drm_gem_shmem_pin(&bo->base);
+ if (ret)
+ goto err_cleanup;
+ }
+
+ sgt = drm_gem_shmem_get_pages_sgt(&bo->base);
+ if (IS_ERR(sgt)) {
+ if (!bo->base.base.import_attach)
+ drm_gem_shmem_unpin(&bo->base);
+
+ ret = PTR_ERR(sgt);
+ goto err_cleanup;
+ }
+
+ op_ctx->map.sgt = sgt;
+
+ preallocated_vm_bo = drm_gpuvm_bo_create(&vm->base, &bo->base.base);
+ if (!preallocated_vm_bo) {
+ if (!bo->base.base.import_attach)
+ drm_gem_shmem_unpin(&bo->base);
+
+ ret = -ENOMEM;
+ goto err_cleanup;
+ }
+
+ mutex_lock(&bo->gpuva_list_lock);
+ op_ctx->map.vm_bo = drm_gpuvm_bo_obtain_prealloc(preallocated_vm_bo);
+ mutex_unlock(&bo->gpuva_list_lock);
+
+ /* If the a vm_bo for this <VM,BO> combination exists, it already
+ * retains a pin ref, and we can release the one we took earlier.
+ *
+ * If our pre-allocated vm_bo is picked, it now retains the pin ref,
+ * which will be released in panthor_vm_bo_put().
+ */
+ if (preallocated_vm_bo != op_ctx->map.vm_bo &&
+ !bo->base.base.import_attach)
+ drm_gem_shmem_unpin(&bo->base);
+
+ op_ctx->map.bo_offset = offset;
+
+ /* L1, L2 and L3 page tables.
+ * We could optimize L3 allocation by iterating over the sgt and merging
+ * 2M contiguous blocks, but it's simpler to over-provision and return
+ * the pages if they're not used.
+ */
+ pt_count = ((ALIGN(va + size, 1ull << 39) - ALIGN_DOWN(va, 1ull << 39)) >> 39) +
+ ((ALIGN(va + size, 1ull << 30) - ALIGN_DOWN(va, 1ull << 30)) >> 30) +
+ ((ALIGN(va + size, 1ull << 21) - ALIGN_DOWN(va, 1ull << 21)) >> 21);
+
+ op_ctx->rsvd_page_tables.pages = kcalloc(pt_count,
+ sizeof(*op_ctx->rsvd_page_tables.pages),
+ GFP_KERNEL);
+ if (!op_ctx->rsvd_page_tables.pages)
+ goto err_cleanup;
+
+ ret = kmem_cache_alloc_bulk(pt_cache, GFP_KERNEL, pt_count,
+ op_ctx->rsvd_page_tables.pages);
+ op_ctx->rsvd_page_tables.count = ret;
+ if (ret != pt_count) {
+ ret = -ENOMEM;
+ goto err_cleanup;
+ }
+
+ /* Insert BO into the extobj list last, when we know nothing can fail. */
+ drm_gpuvm_bo_extobj_add(op_ctx->map.vm_bo);
+
+ return 0;
+
+err_cleanup:
+ panthor_vm_cleanup_op_ctx(op_ctx, vm);
+ return ret;
+}
+
+static int panthor_vm_prepare_unmap_op_ctx(struct panthor_vm_op_ctx *op_ctx,
+ struct panthor_vm *vm,
+ u64 va, u64 size)
+{
+ u32 pt_count = 0;
+ int ret;
+
+ memset(op_ctx, 0, sizeof(*op_ctx));
+ INIT_LIST_HEAD(&op_ctx->returned_vmas);
+ op_ctx->va.range = size;
+ op_ctx->va.addr = va;
+ op_ctx->flags = DRM_PANTHOR_VM_BIND_OP_TYPE_UNMAP;
+
+ /* Pre-allocate L3 page tables to account for the split-2M-block
+ * situation on unmap.
+ */
+ if (va != ALIGN(va, SZ_2M))
+ pt_count++;
+
+ if (va + size != ALIGN(va + size, SZ_2M) &&
+ ALIGN(va + size, SZ_2M) != ALIGN(va, SZ_2M))
+ pt_count++;
+
+ ret = panthor_vm_op_ctx_prealloc_vmas(op_ctx);
+ if (ret)
+ goto err_cleanup;
+
+ if (pt_count) {
+ op_ctx->rsvd_page_tables.pages = kcalloc(pt_count,
+ sizeof(*op_ctx->rsvd_page_tables.pages),
+ GFP_KERNEL);
+ if (!op_ctx->rsvd_page_tables.pages)
+ goto err_cleanup;
+
+ ret = kmem_cache_alloc_bulk(pt_cache, GFP_KERNEL, pt_count,
+ op_ctx->rsvd_page_tables.pages);
+ if (ret != pt_count) {
+ ret = -ENOMEM;
+ goto err_cleanup;
+ }
+ op_ctx->rsvd_page_tables.count = pt_count;
+ }
+
+ return 0;
+
+err_cleanup:
+ panthor_vm_cleanup_op_ctx(op_ctx, vm);
+ return ret;
+}
+
+static void panthor_vm_prepare_sync_only_op_ctx(struct panthor_vm_op_ctx *op_ctx,
+ struct panthor_vm *vm)
+{
+ memset(op_ctx, 0, sizeof(*op_ctx));
+ INIT_LIST_HEAD(&op_ctx->returned_vmas);
+ op_ctx->flags = DRM_PANTHOR_VM_BIND_OP_TYPE_SYNC_ONLY;
+}
+
+/**
+ * panthor_vm_get_bo_for_va() - Get the GEM object mapped at a virtual address
+ * @vm: VM to look into.
+ * @va: Virtual address to search for.
+ * @bo_offset: Offset of the GEM object mapped at this virtual address.
+ * Only valid on success.
+ *
+ * The object returned by this function might no longer be mapped when the
+ * function returns. It's the caller responsibility to ensure there's no
+ * concurrent map/unmap operations making the returned value invalid, or
+ * make sure it doesn't matter if the object is no longer mapped.
+ *
+ * Return: A valid pointer on success, an ERR_PTR() otherwise.
+ */
+struct panthor_gem_object *
+panthor_vm_get_bo_for_va(struct panthor_vm *vm, u64 va, u64 *bo_offset)
+{
+ struct panthor_gem_object *bo = ERR_PTR(-ENOENT);
+ struct drm_gpuva *gpuva;
+ struct panthor_vma *vma;
+
+ /* Take the VM lock to prevent concurrent map/unmap operations. */
+ mutex_lock(&vm->op_lock);
+ gpuva = drm_gpuva_find_first(&vm->base, va, 1);
+ vma = gpuva ? container_of(gpuva, struct panthor_vma, base) : NULL;
+ if (vma && vma->base.gem.obj) {
+ drm_gem_object_get(vma->base.gem.obj);
+ bo = to_panthor_bo(vma->base.gem.obj);
+ *bo_offset = vma->base.gem.offset;
+ }
+ mutex_unlock(&vm->op_lock);
+
+ return bo;
+}
+
+#define PANTHOR_VM_MIN_KERNEL_VA_SIZE SZ_256M
+
+static u64
+panthor_vm_create_get_user_va_range(const struct drm_panthor_vm_create *args,
+ u64 full_va_range)
+{
+ u64 user_va_range;
+
+ /* Make sure we have a minimum amount of VA space for kernel objects. */
+ if (full_va_range < PANTHOR_VM_MIN_KERNEL_VA_SIZE)
+ return 0;
+
+ if (args->user_va_range) {
+ /* Use the user provided value if != 0. */
+ user_va_range = args->user_va_range;
+ } else if (TASK_SIZE_OF(current) < full_va_range) {
+ /* If the task VM size is smaller than the GPU VA range, pick this
+ * as our default user VA range, so userspace can CPU/GPU map buffers
+ * at the same address.
+ */
+ user_va_range = TASK_SIZE_OF(current);
+ } else {
+ /* If the GPU VA range is smaller than the task VM size, we
+ * just have to live with the fact we won't be able to map
+ * all buffers at the same GPU/CPU address.
+ *
+ * If the GPU VA range is bigger than 4G (more than 32-bit of
+ * VA), we split the range in two, and assign half of it to
+ * the user and the other half to the kernel, if it's not, we
+ * keep the kernel VA space as small as possible.
+ */
+ user_va_range = full_va_range > SZ_4G ?
+ full_va_range / 2 :
+ full_va_range - PANTHOR_VM_MIN_KERNEL_VA_SIZE;
+ }
+
+ if (full_va_range - PANTHOR_VM_MIN_KERNEL_VA_SIZE < user_va_range)
+ user_va_range = full_va_range - PANTHOR_VM_MIN_KERNEL_VA_SIZE;
+
+ return user_va_range;
+}
+
+#define PANTHOR_VM_CREATE_FLAGS 0
+
+static int
+panthor_vm_create_check_args(const struct panthor_device *ptdev,
+ const struct drm_panthor_vm_create *args,
+ u64 *kernel_va_start, u64 *kernel_va_range)
+{
+ u32 va_bits = GPU_MMU_FEATURES_VA_BITS(ptdev->gpu_info.mmu_features);
+ u64 full_va_range = 1ull << va_bits;
+ u64 user_va_range;
+
+ if (args->flags & ~PANTHOR_VM_CREATE_FLAGS)
+ return -EINVAL;
+
+ user_va_range = panthor_vm_create_get_user_va_range(args, full_va_range);
+ if (!user_va_range || (args->user_va_range && args->user_va_range > user_va_range))
+ return -EINVAL;
+
+ /* Pick a kernel VA range that's a power of two, to have a clear split. */
+ *kernel_va_range = rounddown_pow_of_two(full_va_range - user_va_range);
+ *kernel_va_start = full_va_range - *kernel_va_range;
+ return 0;
+}
+
+/*
+ * Only 32 VMs per open file. If that becomes a limiting factor, we can
+ * increase this number.
+ */
+#define PANTHOR_MAX_VMS_PER_FILE 32
+
+/**
+ * panthor_vm_pool_create_vm() - Create a VM
+ * @pool: The VM to create this VM on.
+ * @kernel_va_start: Start of the region reserved for kernel objects.
+ * @kernel_va_range: Size of the region reserved for kernel objects.
+ *
+ * Return: a positive VM ID on success, a negative error code otherwise.
+ */
+int panthor_vm_pool_create_vm(struct panthor_device *ptdev,
+ struct panthor_vm_pool *pool,
+ struct drm_panthor_vm_create *args)
+{
+ u64 kernel_va_start, kernel_va_range;
+ struct panthor_vm *vm;
+ int ret;
+ u32 id;
+
+ ret = panthor_vm_create_check_args(ptdev, args, &kernel_va_start, &kernel_va_range);
+ if (ret)
+ return ret;
+
+ vm = panthor_vm_create(ptdev, false, kernel_va_start, kernel_va_range,
+ kernel_va_start, kernel_va_range);
+ if (IS_ERR(vm))
+ return PTR_ERR(vm);
+
+ ret = xa_alloc(&pool->xa, &id, vm,
+ XA_LIMIT(1, PANTHOR_MAX_VMS_PER_FILE), GFP_KERNEL);
+
+ if (ret) {
+ panthor_vm_put(vm);
+ return ret;
+ }
+
+ args->user_va_range = kernel_va_start;
+ return id;
+}
+
+static void panthor_vm_destroy(struct panthor_vm *vm)
+{
+ if (!vm)
+ return;
+
+ vm->destroyed = true;
+
+ mutex_lock(&vm->heaps.lock);
+ panthor_heap_pool_destroy(vm->heaps.pool);
+ vm->heaps.pool = NULL;
+ mutex_unlock(&vm->heaps.lock);
+
+ drm_WARN_ON(&vm->ptdev->base,
+ panthor_vm_unmap_range(vm, vm->base.mm_start, vm->base.mm_range));
+ panthor_vm_put(vm);
+}
+
+/**
+ * panthor_vm_destroy() - Destroy a VM.
+ * @pool: VM pool.
+ * @handle: VM handle.
+ *
+ * This function doesn't free the VM object or its resources, it just kills
+ * all mappings, and makes sure nothing can be mapped after that point.
+ *
+ * If there was any active jobs at the time this function is called, these
+ * jobs should experience page faults and be killed as a result.
+ *
+ * The VM resources are freed when the last reference on the VM object is
+ * dropped.
+ */
+int panthor_vm_pool_destroy_vm(struct panthor_vm_pool *pool, u32 handle)
+{
+ struct panthor_vm *vm;
+
+ vm = xa_erase(&pool->xa, handle);
+
+ panthor_vm_destroy(vm);
+
+ return vm ? 0 : -EINVAL;
+}
+
+/**
+ * panthor_vm_pool_get_vm() - Retrieve VM object bound to a VM handle
+ * @pool: VM pool to check.
+ * @handle: Handle of the VM to retrieve.
+ *
+ * Return: A valid pointer if the VM exists, NULL otherwise.
+ */
+struct panthor_vm *
+panthor_vm_pool_get_vm(struct panthor_vm_pool *pool, u32 handle)
+{
+ struct panthor_vm *vm;
+
+ vm = panthor_vm_get(xa_load(&pool->xa, handle));
+
+ return vm;
+}
+
+/**
+ * panthor_vm_pool_destroy() - Destroy a VM pool.
+ * @pfile: File.
+ *
+ * Destroy all VMs in the pool, and release the pool resources.
+ *
+ * Note that VMs can outlive the pool they were created from if other
+ * objects hold a reference to there VMs.
+ */
+void panthor_vm_pool_destroy(struct panthor_file *pfile)
+{
+ struct panthor_vm *vm;
+ unsigned long i;
+
+ if (!pfile->vms)
+ return;
+
+ xa_for_each(&pfile->vms->xa, i, vm)
+ panthor_vm_destroy(vm);
+
+ xa_destroy(&pfile->vms->xa);
+ kfree(pfile->vms);
+}
+
+/**
+ * panthor_vm_pool_create() - Create a VM pool
+ * @pfile: File.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_vm_pool_create(struct panthor_file *pfile)
+{
+ pfile->vms = kzalloc(sizeof(*pfile->vms), GFP_KERNEL);
+ if (!pfile->vms)
+ return -ENOMEM;
+
+ xa_init_flags(&pfile->vms->xa, XA_FLAGS_ALLOC1);
+ return 0;
+}
+
+/* dummy TLB ops, the real TLB flush happens in panthor_vm_flush_range() */
+static void mmu_tlb_flush_all(void *cookie)
+{
+}
+
+static void mmu_tlb_flush_walk(unsigned long iova, size_t size, size_t granule, void *cookie)
+{
+}
+
+static const struct iommu_flush_ops mmu_tlb_ops = {
+ .tlb_flush_all = mmu_tlb_flush_all,
+ .tlb_flush_walk = mmu_tlb_flush_walk,
+};
+
+static const char *access_type_name(struct panthor_device *ptdev,
+ u32 fault_status)
+{
+ switch (fault_status & AS_FAULTSTATUS_ACCESS_TYPE_MASK) {
+ case AS_FAULTSTATUS_ACCESS_TYPE_ATOMIC:
+ return "ATOMIC";
+ case AS_FAULTSTATUS_ACCESS_TYPE_READ:
+ return "READ";
+ case AS_FAULTSTATUS_ACCESS_TYPE_WRITE:
+ return "WRITE";
+ case AS_FAULTSTATUS_ACCESS_TYPE_EX:
+ return "EXECUTE";
+ default:
+ drm_WARN_ON(&ptdev->base, 1);
+ return NULL;
+ }
+}
+
+static void panthor_mmu_irq_handler(struct panthor_device *ptdev, u32 status)
+{
+ bool has_unhandled_faults = false;
+
+ status = panthor_mmu_fault_mask(ptdev, status);
+ while (status) {
+ u32 as = ffs(status | (status >> 16)) - 1;
+ u32 mask = panthor_mmu_as_fault_mask(ptdev, as);
+ u32 new_int_mask;
+ u64 addr;
+ u32 fault_status;
+ u32 exception_type;
+ u32 access_type;
+ u32 source_id;
+
+ fault_status = gpu_read(ptdev, AS_FAULTSTATUS(as));
+ addr = gpu_read(ptdev, AS_FAULTADDRESS_LO(as));
+ addr |= (u64)gpu_read(ptdev, AS_FAULTADDRESS_HI(as)) << 32;
+
+ /* decode the fault status */
+ exception_type = fault_status & 0xFF;
+ access_type = (fault_status >> 8) & 0x3;
+ source_id = (fault_status >> 16);
+
+ mutex_lock(&ptdev->mmu->as.slots_lock);
+
+ ptdev->mmu->as.faulty_mask |= mask;
+ new_int_mask =
+ panthor_mmu_fault_mask(ptdev, ~ptdev->mmu->as.faulty_mask);
+
+ /* terminal fault, print info about the fault */
+ drm_err(&ptdev->base,
+ "Unhandled Page fault in AS%d at VA 0x%016llX\n"
+ "raw fault status: 0x%X\n"
+ "decoded fault status: %s\n"
+ "exception type 0x%X: %s\n"
+ "access type 0x%X: %s\n"
+ "source id 0x%X\n",
+ as, addr,
+ fault_status,
+ (fault_status & (1 << 10) ? "DECODER FAULT" : "SLAVE FAULT"),
+ exception_type, panthor_exception_name(ptdev, exception_type),
+ access_type, access_type_name(ptdev, fault_status),
+ source_id);
+
+ /* Ignore MMU interrupts on this AS until it's been
+ * re-enabled.
+ */
+ ptdev->mmu->irq.mask = new_int_mask;
+ gpu_write(ptdev, MMU_INT_MASK, new_int_mask);
+
+ if (ptdev->mmu->as.slots[as].vm)
+ ptdev->mmu->as.slots[as].vm->unhandled_fault = true;
+
+ /* Disable the MMU to kill jobs on this AS. */
+ panthor_mmu_as_disable(ptdev, as);
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+
+ status &= ~mask;
+ has_unhandled_faults = true;
+ }
+
+ if (has_unhandled_faults)
+ panthor_sched_report_mmu_fault(ptdev);
+}
+PANTHOR_IRQ_HANDLER(mmu, MMU, panthor_mmu_irq_handler);
+
+/**
+ * panthor_mmu_suspend() - Suspend the MMU logic
+ * @ptdev: Device.
+ *
+ * All we do here is de-assign the AS slots on all active VMs, so things
+ * get flushed to the main memory, and no further access to these VMs are
+ * possible.
+ *
+ * We also suspend the MMU IRQ.
+ */
+void panthor_mmu_suspend(struct panthor_device *ptdev)
+{
+ mutex_lock(&ptdev->mmu->as.slots_lock);
+ for (u32 i = 0; i < ARRAY_SIZE(ptdev->mmu->as.slots); i++) {
+ struct panthor_vm *vm = ptdev->mmu->as.slots[i].vm;
+
+ if (vm) {
+ drm_WARN_ON(&ptdev->base, panthor_mmu_as_disable(ptdev, i));
+ panthor_vm_release_as_locked(vm);
+ }
+ }
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+
+ panthor_mmu_irq_suspend(&ptdev->mmu->irq);
+}
+
+/**
+ * panthor_mmu_resume() - Resume the MMU logic
+ * @ptdev: Device.
+ *
+ * Resume the IRQ.
+ *
+ * We don't re-enable previously active VMs. We assume other parts of the
+ * driver will call panthor_vm_active() on the VMs they intend to use.
+ */
+void panthor_mmu_resume(struct panthor_device *ptdev)
+{
+ mutex_lock(&ptdev->mmu->as.slots_lock);
+ ptdev->mmu->as.alloc_mask = 0;
+ ptdev->mmu->as.faulty_mask = 0;
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+
+ panthor_mmu_irq_resume(&ptdev->mmu->irq, panthor_mmu_fault_mask(ptdev, ~0));
+}
+
+/**
+ * panthor_mmu_pre_reset() - Prepare for a reset
+ * @ptdev: Device.
+ *
+ * Suspend the IRQ, and make sure all VM_BIND queues are stopped, so we
+ * don't get asked to do a VM operation while the GPU is down.
+ *
+ * We don't cleanly shutdown the AS slots here, because the reset might
+ * come from an AS_ACTIVE_BIT stuck situation.
+ */
+void panthor_mmu_pre_reset(struct panthor_device *ptdev)
+{
+ struct panthor_vm *vm;
+
+ panthor_mmu_irq_suspend(&ptdev->mmu->irq);
+
+ mutex_lock(&ptdev->mmu->vm.lock);
+ ptdev->mmu->vm.reset_in_progress = true;
+ list_for_each_entry(vm, &ptdev->mmu->vm.list, node)
+ panthor_vm_stop(vm);
+ mutex_unlock(&ptdev->mmu->vm.lock);
+}
+
+/**
+ * panthor_mmu_post_reset() - Restore things after a reset
+ * @ptdev: Device.
+ *
+ * Put the MMU logic back in action after a reset. That implies resuming the
+ * IRQ and re-enabling the VM_BIND queues.
+ */
+void panthor_mmu_post_reset(struct panthor_device *ptdev)
+{
+ struct panthor_vm *vm;
+
+ mutex_lock(&ptdev->mmu->as.slots_lock);
+
+ /* Now that the reset is effective, we can assume that none of the
+ * AS slots are setup, and clear the faulty flags too.
+ */
+ ptdev->mmu->as.alloc_mask = 0;
+ ptdev->mmu->as.faulty_mask = 0;
+
+ for (u32 i = 0; i < ARRAY_SIZE(ptdev->mmu->as.slots); i++) {
+ struct panthor_vm *vm = ptdev->mmu->as.slots[i].vm;
+
+ if (vm)
+ panthor_vm_release_as_locked(vm);
+ }
+
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+
+ panthor_mmu_irq_resume(&ptdev->mmu->irq, panthor_mmu_fault_mask(ptdev, ~0));
+
+ /* Restart the VM_BIND queues. */
+ mutex_lock(&ptdev->mmu->vm.lock);
+ list_for_each_entry(vm, &ptdev->mmu->vm.list, node) {
+ panthor_vm_start(vm);
+ }
+ ptdev->mmu->vm.reset_in_progress = false;
+ mutex_unlock(&ptdev->mmu->vm.lock);
+}
+
+static void panthor_vm_free(struct drm_gpuvm *gpuvm)
+{
+ struct panthor_vm *vm = container_of(gpuvm, struct panthor_vm, base);
+ struct panthor_device *ptdev = vm->ptdev;
+
+ mutex_lock(&vm->heaps.lock);
+ if (drm_WARN_ON(&ptdev->base, vm->heaps.pool))
+ panthor_heap_pool_destroy(vm->heaps.pool);
+ mutex_unlock(&vm->heaps.lock);
+ mutex_destroy(&vm->heaps.lock);
+
+ mutex_lock(&ptdev->mmu->vm.lock);
+ list_del(&vm->node);
+ /* Restore the scheduler state so we can call drm_sched_entity_destroy()
+ * and drm_sched_fini(). If get there, that means we have no job left
+ * and no new jobs can be queued, so we can start the scheduler without
+ * risking interfering with the reset.
+ */
+ if (ptdev->mmu->vm.reset_in_progress)
+ panthor_vm_start(vm);
+ mutex_unlock(&ptdev->mmu->vm.lock);
+
+ drm_sched_entity_destroy(&vm->entity);
+ drm_sched_fini(&vm->sched);
+
+ mutex_lock(&ptdev->mmu->as.slots_lock);
+ if (vm->as.id >= 0) {
+ int cookie;
+
+ if (drm_dev_enter(&ptdev->base, &cookie)) {
+ panthor_mmu_as_disable(ptdev, vm->as.id);
+ drm_dev_exit(cookie);
+ }
+
+ ptdev->mmu->as.slots[vm->as.id].vm = NULL;
+ clear_bit(vm->as.id, &ptdev->mmu->as.alloc_mask);
+ list_del(&vm->as.lru_node);
+ }
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+
+ free_io_pgtable_ops(vm->pgtbl_ops);
+
+ drm_mm_takedown(&vm->mm);
+ kfree(vm);
+}
+
+/**
+ * panthor_vm_put() - Release a reference on a VM
+ * @vm: VM to release the reference on. Can be NULL.
+ */
+void panthor_vm_put(struct panthor_vm *vm)
+{
+ drm_gpuvm_put(vm ? &vm->base : NULL);
+}
+
+/**
+ * panthor_vm_get() - Get a VM reference
+ * @vm: VM to get the reference on. Can be NULL.
+ *
+ * Return: @vm value.
+ */
+struct panthor_vm *panthor_vm_get(struct panthor_vm *vm)
+{
+ if (vm)
+ drm_gpuvm_get(&vm->base);
+
+ return vm;
+}
+
+/**
+ * panthor_vm_get_heap_pool() - Get the heap pool attached to a VM
+ * @vm: VM to query the heap pool on.
+ * @create: True if the heap pool should be created when it doesn't exist.
+ *
+ * Heap pools are per-VM. This function allows one to retrieve the heap pool
+ * attached to a VM.
+ *
+ * If no heap pool exists yet, and @create is true, we create one.
+ *
+ * The returned panthor_heap_pool should be released with panthor_heap_pool_put().
+ *
+ * Return: A valid pointer on success, an ERR_PTR() otherwise.
+ */
+struct panthor_heap_pool *panthor_vm_get_heap_pool(struct panthor_vm *vm, bool create)
+{
+ struct panthor_heap_pool *pool;
+
+ mutex_lock(&vm->heaps.lock);
+ if (!vm->heaps.pool && create) {
+ if (vm->destroyed)
+ pool = ERR_PTR(-EINVAL);
+ else
+ pool = panthor_heap_pool_create(vm->ptdev, vm);
+
+ if (!IS_ERR(pool))
+ vm->heaps.pool = panthor_heap_pool_get(pool);
+ } else {
+ pool = panthor_heap_pool_get(vm->heaps.pool);
+ }
+ mutex_unlock(&vm->heaps.lock);
+
+ return pool;
+}
+
+static u64 mair_to_memattr(u64 mair)
+{
+ u64 memattr = 0;
+ u32 i;
+
+ for (i = 0; i < 8; i++) {
+ u8 in_attr = mair >> (8 * i), out_attr;
+ u8 outer = in_attr >> 4, inner = in_attr & 0xf;
+
+ /* For caching to be enabled, inner and outer caching policy
+ * have to be both write-back, if one of them is write-through
+ * or non-cacheable, we just choose non-cacheable. Device
+ * memory is also translated to non-cacheable.
+ */
+ if (!(outer & 3) || !(outer & 4) || !(inner & 4)) {
+ out_attr = AS_MEMATTR_AARCH64_INNER_OUTER_NC |
+ AS_MEMATTR_AARCH64_SH_MIDGARD_INNER |
+ AS_MEMATTR_AARCH64_INNER_ALLOC_EXPL(false, false);
+ } else {
+ /* Use SH_CPU_INNER mode so SH_IS, which is used when
+ * IOMMU_CACHE is set, actually maps to the standard
+ * definition of inner-shareable and not Mali's
+ * internal-shareable mode.
+ */
+ out_attr = AS_MEMATTR_AARCH64_INNER_OUTER_WB |
+ AS_MEMATTR_AARCH64_SH_CPU_INNER |
+ AS_MEMATTR_AARCH64_INNER_ALLOC_EXPL(inner & 1, inner & 2);
+ }
+
+ memattr |= (u64)out_attr << (8 * i);
+ }
+
+ return memattr;
+}
+
+static void panthor_vma_link(struct panthor_vm *vm,
+ struct panthor_vma *vma,
+ struct drm_gpuvm_bo *vm_bo)
+{
+ struct panthor_gem_object *bo = to_panthor_bo(vma->base.gem.obj);
+
+ mutex_lock(&bo->gpuva_list_lock);
+ drm_gpuva_link(&vma->base, vm_bo);
+ drm_gpuvm_bo_put(vm_bo);
+ mutex_unlock(&bo->gpuva_list_lock);
+}
+
+static void panthor_vma_unlink(struct panthor_vm *vm,
+ struct panthor_vma *vma)
+{
+ struct panthor_gem_object *bo = to_panthor_bo(vma->base.gem.obj);
+ struct drm_gpuvm_bo *vm_bo = drm_gpuvm_bo_get(vma->base.vm_bo);
+
+ mutex_lock(&bo->gpuva_list_lock);
+ drm_gpuva_unlink(&vma->base);
+ mutex_unlock(&bo->gpuva_list_lock);
+
+ /* drm_gpuva_unlink() release the vm_bo, but we manually retained it
+ * when entering this function, so we can implement deferred VMA
+ * destruction. Re-assign it here.
+ */
+ vma->base.vm_bo = vm_bo;
+ list_add_tail(&vma->node, &vm->op_ctx->returned_vmas);
+}
+
+static void panthor_vma_init(struct panthor_vma *vma, u32 flags)
+{
+ INIT_LIST_HEAD(&vma->node);
+ vma->flags = flags;
+}
+
+#define PANTHOR_VM_MAP_FLAGS \
+ (DRM_PANTHOR_VM_BIND_OP_MAP_READONLY | \
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC | \
+ DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED)
+
+static int panthor_gpuva_sm_step_map(struct drm_gpuva_op *op, void *priv)
+{
+ struct panthor_vm *vm = priv;
+ struct panthor_vm_op_ctx *op_ctx = vm->op_ctx;
+ struct panthor_vma *vma = panthor_vm_op_ctx_get_vma(op_ctx);
+ int ret;
+
+ if (!vma)
+ return -EINVAL;
+
+ panthor_vma_init(vma, op_ctx->flags & PANTHOR_VM_MAP_FLAGS);
+
+ ret = panthor_vm_map_pages(vm, op->map.va.addr, flags_to_prot(vma->flags),
+ op_ctx->map.sgt, op->map.gem.offset,
+ op->map.va.range);
+ if (ret)
+ return ret;
+
+ /* Ref owned by the mapping now, clear the obj field so we don't release the
+ * pinning/obj ref behind GPUVA's back.
+ */
+ drm_gpuva_map(&vm->base, &vma->base, &op->map);
+ panthor_vma_link(vm, vma, op_ctx->map.vm_bo);
+ op_ctx->map.vm_bo = NULL;
+ return 0;
+}
+
+static int panthor_gpuva_sm_step_remap(struct drm_gpuva_op *op,
+ void *priv)
+{
+ struct panthor_vma *unmap_vma = container_of(op->remap.unmap->va, struct panthor_vma, base);
+ struct panthor_vm *vm = priv;
+ struct panthor_vm_op_ctx *op_ctx = vm->op_ctx;
+ struct panthor_vma *prev_vma = NULL, *next_vma = NULL;
+ u64 unmap_start, unmap_range;
+ int ret;
+
+ drm_gpuva_op_remap_to_unmap_range(&op->remap, &unmap_start, &unmap_range);
+ ret = panthor_vm_unmap_pages(vm, unmap_start, unmap_range);
+ if (ret)
+ return ret;
+
+ if (op->remap.prev) {
+ prev_vma = panthor_vm_op_ctx_get_vma(op_ctx);
+ panthor_vma_init(prev_vma, unmap_vma->flags);
+ }
+
+ if (op->remap.next) {
+ next_vma = panthor_vm_op_ctx_get_vma(op_ctx);
+ panthor_vma_init(next_vma, unmap_vma->flags);
+ }
+
+ drm_gpuva_remap(prev_vma ? &prev_vma->base : NULL,
+ next_vma ? &next_vma->base : NULL,
+ &op->remap);
+
+ if (prev_vma) {
+ /* panthor_vma_link() transfers the vm_bo ownership to
+ * the VMA object. Since the vm_bo we're passing is still
+ * owned by the old mapping which will be released when this
+ * mapping is destroyed, we need to grab a ref here.
+ */
+ panthor_vma_link(vm, prev_vma,
+ drm_gpuvm_bo_get(op->remap.unmap->va->vm_bo));
+ }
+
+ if (next_vma) {
+ panthor_vma_link(vm, next_vma,
+ drm_gpuvm_bo_get(op->remap.unmap->va->vm_bo));
+ }
+
+ panthor_vma_unlink(vm, unmap_vma);
+ return 0;
+}
+
+static int panthor_gpuva_sm_step_unmap(struct drm_gpuva_op *op,
+ void *priv)
+{
+ struct panthor_vma *unmap_vma = container_of(op->unmap.va, struct panthor_vma, base);
+ struct panthor_vm *vm = priv;
+ int ret;
+
+ ret = panthor_vm_unmap_pages(vm, unmap_vma->base.va.addr,
+ unmap_vma->base.va.range);
+ if (drm_WARN_ON(&vm->ptdev->base, ret))
+ return ret;
+
+ drm_gpuva_unmap(&op->unmap);
+ panthor_vma_unlink(vm, unmap_vma);
+ return 0;
+}
+
+static const struct drm_gpuvm_ops panthor_gpuvm_ops = {
+ .vm_free = panthor_vm_free,
+ .sm_step_map = panthor_gpuva_sm_step_map,
+ .sm_step_remap = panthor_gpuva_sm_step_remap,
+ .sm_step_unmap = panthor_gpuva_sm_step_unmap,
+};
+
+/**
+ * panthor_vm_resv() - Get the dma_resv object attached to a VM.
+ * @vm: VM to get the dma_resv of.
+ *
+ * Return: A dma_resv object.
+ */
+struct dma_resv *panthor_vm_resv(struct panthor_vm *vm)
+{
+ return drm_gpuvm_resv(&vm->base);
+}
+
+struct drm_gem_object *panthor_vm_root_gem(struct panthor_vm *vm)
+{
+ if (!vm)
+ return NULL;
+
+ return vm->base.r_obj;
+}
+
+static int
+panthor_vm_exec_op(struct panthor_vm *vm, struct panthor_vm_op_ctx *op,
+ bool flag_vm_unusable_on_failure)
+{
+ u32 op_type = op->flags & DRM_PANTHOR_VM_BIND_OP_TYPE_MASK;
+ int ret;
+
+ if (op_type == DRM_PANTHOR_VM_BIND_OP_TYPE_SYNC_ONLY)
+ return 0;
+
+ mutex_lock(&vm->op_lock);
+ vm->op_ctx = op;
+ switch (op_type) {
+ case DRM_PANTHOR_VM_BIND_OP_TYPE_MAP:
+ if (vm->unusable) {
+ ret = -EINVAL;
+ break;
+ }
+
+ ret = drm_gpuvm_sm_map(&vm->base, vm, op->va.addr, op->va.range,
+ op->map.vm_bo->obj, op->map.bo_offset);
+ break;
+
+ case DRM_PANTHOR_VM_BIND_OP_TYPE_UNMAP:
+ ret = drm_gpuvm_sm_unmap(&vm->base, vm, op->va.addr, op->va.range);
+ break;
+
+ default:
+ ret = -EINVAL;
+ break;
+ }
+
+ if (ret && flag_vm_unusable_on_failure)
+ vm->unusable = true;
+
+ vm->op_ctx = NULL;
+ mutex_unlock(&vm->op_lock);
+
+ return ret;
+}
+
+static struct dma_fence *
+panthor_vm_bind_run_job(struct drm_sched_job *sched_job)
+{
+ struct panthor_vm_bind_job *job = container_of(sched_job, struct panthor_vm_bind_job, base);
+ bool cookie;
+ int ret;
+
+ /* Not only we report an error whose result is propagated to the
+ * drm_sched finished fence, but we also flag the VM as unusable, because
+ * a failure in the async VM_BIND results in an inconsistent state. VM needs
+ * to be destroyed and recreated.
+ */
+ cookie = dma_fence_begin_signalling();
+ ret = panthor_vm_exec_op(job->vm, &job->ctx, true);
+ dma_fence_end_signalling(cookie);
+
+ return ret ? ERR_PTR(ret) : NULL;
+}
+
+static void panthor_vm_bind_job_release(struct kref *kref)
+{
+ struct panthor_vm_bind_job *job = container_of(kref, struct panthor_vm_bind_job, refcount);
+
+ if (job->base.s_fence)
+ drm_sched_job_cleanup(&job->base);
+
+ panthor_vm_cleanup_op_ctx(&job->ctx, job->vm);
+ panthor_vm_put(job->vm);
+ kfree(job);
+}
+
+/**
+ * panthor_vm_bind_job_put() - Release a VM_BIND job reference
+ * @sched_job: Job to release the reference on.
+ */
+void panthor_vm_bind_job_put(struct drm_sched_job *sched_job)
+{
+ struct panthor_vm_bind_job *job =
+ container_of(sched_job, struct panthor_vm_bind_job, base);
+
+ if (sched_job)
+ kref_put(&job->refcount, panthor_vm_bind_job_release);
+}
+
+static void
+panthor_vm_bind_free_job(struct drm_sched_job *sched_job)
+{
+ struct panthor_vm_bind_job *job =
+ container_of(sched_job, struct panthor_vm_bind_job, base);
+
+ drm_sched_job_cleanup(sched_job);
+
+ /* Do the heavy cleanups asynchronously, so we're out of the
+ * dma-signaling path and can acquire dma-resv locks safely.
+ */
+ queue_work(panthor_cleanup_wq, &job->cleanup_op_ctx_work);
+}
+
+static enum drm_gpu_sched_stat
+panthor_vm_bind_timedout_job(struct drm_sched_job *sched_job)
+{
+ WARN(1, "VM_BIND ops are synchronous for now, there should be no timeout!");
+ return DRM_GPU_SCHED_STAT_NOMINAL;
+}
+
+static const struct drm_sched_backend_ops panthor_vm_bind_ops = {
+ .run_job = panthor_vm_bind_run_job,
+ .free_job = panthor_vm_bind_free_job,
+ .timedout_job = panthor_vm_bind_timedout_job,
+};
+
+/**
+ * panthor_vm_create() - Create a VM
+ * @ptdev: Device.
+ * @for_mcu: True if this is the FW MCU VM.
+ * @kernel_va_start: Start of the range reserved for kernel BO mapping.
+ * @kernel_va_size: Size of the range reserved for kernel BO mapping.
+ * @auto_kernel_va_start: Start of the auto-VA kernel range.
+ * @auto_kernel_va_size: Size of the auto-VA kernel range.
+ *
+ * Return: A valid pointer on success, an ERR_PTR() otherwise.
+ */
+struct panthor_vm *
+panthor_vm_create(struct panthor_device *ptdev, bool for_mcu,
+ u64 kernel_va_start, u64 kernel_va_size,
+ u64 auto_kernel_va_start, u64 auto_kernel_va_size)
+{
+ u32 va_bits = GPU_MMU_FEATURES_VA_BITS(ptdev->gpu_info.mmu_features);
+ u32 pa_bits = GPU_MMU_FEATURES_PA_BITS(ptdev->gpu_info.mmu_features);
+ u64 full_va_range = 1ull << va_bits;
+ struct drm_gem_object *dummy_gem;
+ struct drm_gpu_scheduler *sched;
+ struct io_pgtable_cfg pgtbl_cfg;
+ u64 mair, min_va, va_range;
+ struct panthor_vm *vm;
+ int ret;
+
+ vm = kzalloc(sizeof(*vm), GFP_KERNEL);
+ if (!vm)
+ return ERR_PTR(-ENOMEM);
+
+ /* We allocate a dummy GEM for the VM. */
+ dummy_gem = drm_gpuvm_resv_object_alloc(&ptdev->base);
+ if (!dummy_gem) {
+ ret = -ENOMEM;
+ goto err_free_vm;
+ }
+
+ mutex_init(&vm->heaps.lock);
+ vm->for_mcu = for_mcu;
+ vm->ptdev = ptdev;
+ mutex_init(&vm->op_lock);
+
+ if (for_mcu) {
+ /* CSF MCU is a cortex M7, and can only address 4G */
+ min_va = 0;
+ va_range = SZ_4G;
+ } else {
+ min_va = 0;
+ va_range = full_va_range;
+ }
+
+ mutex_init(&vm->mm_lock);
+ drm_mm_init(&vm->mm, kernel_va_start, kernel_va_size);
+ vm->kernel_auto_va.start = auto_kernel_va_start;
+ vm->kernel_auto_va.end = vm->kernel_auto_va.start + auto_kernel_va_size - 1;
+
+ INIT_LIST_HEAD(&vm->node);
+ INIT_LIST_HEAD(&vm->as.lru_node);
+ vm->as.id = -1;
+ refcount_set(&vm->as.active_cnt, 0);
+
+ pgtbl_cfg = (struct io_pgtable_cfg) {
+ .pgsize_bitmap = SZ_4K | SZ_2M,
+ .ias = va_bits,
+ .oas = pa_bits,
+ .coherent_walk = ptdev->coherent,
+ .tlb = &mmu_tlb_ops,
+ .iommu_dev = ptdev->base.dev,
+ .alloc = alloc_pt,
+ .free = free_pt,
+ };
+
+ vm->pgtbl_ops = alloc_io_pgtable_ops(ARM_64_LPAE_S1, &pgtbl_cfg, vm);
+ if (!vm->pgtbl_ops) {
+ ret = -EINVAL;
+ goto err_mm_takedown;
+ }
+
+ /* Bind operations are synchronous for now, no timeout needed. */
+ ret = drm_sched_init(&vm->sched, &panthor_vm_bind_ops, ptdev->mmu->vm.wq,
+ 1, 1, 0,
+ MAX_SCHEDULE_TIMEOUT, NULL, NULL,
+ "panthor-vm-bind", ptdev->base.dev);
+ if (ret)
+ goto err_free_io_pgtable;
+
+ sched = &vm->sched;
+ ret = drm_sched_entity_init(&vm->entity, 0, &sched, 1, NULL);
+ if (ret)
+ goto err_sched_fini;
+
+ mair = io_pgtable_ops_to_pgtable(vm->pgtbl_ops)->cfg.arm_lpae_s1_cfg.mair;
+ vm->memattr = mair_to_memattr(mair);
+
+ mutex_lock(&ptdev->mmu->vm.lock);
+ list_add_tail(&vm->node, &ptdev->mmu->vm.list);
+
+ /* If a reset is in progress, stop the scheduler. */
+ if (ptdev->mmu->vm.reset_in_progress)
+ panthor_vm_stop(vm);
+ mutex_unlock(&ptdev->mmu->vm.lock);
+
+ /* We intentionally leave the reserved range to zero, because we want kernel VMAs
+ * to be handled the same way user VMAs are.
+ */
+ drm_gpuvm_init(&vm->base,
+ for_mcu ? "panthor-MCU-VM" : "panthor-GPU-VM",
+ 0, &ptdev->base, dummy_gem, min_va, va_range, 0, 0,
+ &panthor_gpuvm_ops);
+ drm_gem_object_put(dummy_gem);
+ return vm;
+
+err_sched_fini:
+ drm_sched_fini(&vm->sched);
+
+err_free_io_pgtable:
+ free_io_pgtable_ops(vm->pgtbl_ops);
+
+err_mm_takedown:
+ drm_mm_takedown(&vm->mm);
+ drm_gem_object_put(dummy_gem);
+
+err_free_vm:
+ kfree(vm);
+ return ERR_PTR(ret);
+}
+
+static int
+panthor_vm_bind_prepare_op_ctx(struct drm_file *file,
+ struct panthor_vm *vm,
+ const struct drm_panthor_vm_bind_op *op,
+ struct panthor_vm_op_ctx *op_ctx)
+{
+ struct drm_gem_object *gem;
+ int ret;
+
+ /* Aligned on page size. */
+ if ((op->va | op->size) & ~PAGE_MASK)
+ return -EINVAL;
+
+ switch (op->flags & DRM_PANTHOR_VM_BIND_OP_TYPE_MASK) {
+ case DRM_PANTHOR_VM_BIND_OP_TYPE_MAP:
+ gem = drm_gem_object_lookup(file, op->bo_handle);
+ ret = panthor_vm_prepare_map_op_ctx(op_ctx, vm,
+ gem ? to_panthor_bo(gem) : NULL,
+ op->bo_offset,
+ op->size,
+ op->va,
+ op->flags);
+ drm_gem_object_put(gem);
+ return ret;
+
+ case DRM_PANTHOR_VM_BIND_OP_TYPE_UNMAP:
+ if (op->flags & ~DRM_PANTHOR_VM_BIND_OP_TYPE_MASK)
+ return -EINVAL;
+
+ if (op->bo_handle || op->bo_offset)
+ return -EINVAL;
+
+ return panthor_vm_prepare_unmap_op_ctx(op_ctx, vm, op->va, op->size);
+
+ case DRM_PANTHOR_VM_BIND_OP_TYPE_SYNC_ONLY:
+ if (op->flags & ~DRM_PANTHOR_VM_BIND_OP_TYPE_MASK)
+ return -EINVAL;
+
+ if (op->bo_handle || op->bo_offset)
+ return -EINVAL;
+
+ if (op->va || op->size)
+ return -EINVAL;
+
+ if (!op->syncs.count)
+ return -EINVAL;
+
+ panthor_vm_prepare_sync_only_op_ctx(op_ctx, vm);
+ return 0;
+
+ default:
+ return -EINVAL;
+ }
+}
+
+static void panthor_vm_bind_job_cleanup_op_ctx_work(struct work_struct *work)
+{
+ struct panthor_vm_bind_job *job =
+ container_of(work, struct panthor_vm_bind_job, cleanup_op_ctx_work);
+
+ panthor_vm_cleanup_op_ctx(&job->ctx, job->vm);
+ panthor_vm_bind_job_put(&job->base);
+}
+
+/**
+ * panthor_vm_bind_job_create() - Create a VM_BIND job
+ * @file: File.
+ * @vm: VM targeted by the VM_BIND job.
+ * @op: VM operation data.
+ *
+ * Return: A valid pointer on success, an ERR_PTR() otherwise.
+ */
+struct drm_sched_job *
+panthor_vm_bind_job_create(struct drm_file *file,
+ struct panthor_vm *vm,
+ const struct drm_panthor_vm_bind_op *op)
+{
+ struct panthor_vm_bind_job *job;
+ int ret;
+
+ if (!vm)
+ return ERR_PTR(-EINVAL);
+
+ if (vm->destroyed || vm->unusable)
+ return ERR_PTR(-EINVAL);
+
+ job = kzalloc(sizeof(*job), GFP_KERNEL);
+ if (!job)
+ return ERR_PTR(-ENOMEM);
+
+ ret = panthor_vm_bind_prepare_op_ctx(file, vm, op, &job->ctx);
+ if (ret) {
+ kfree(job);
+ return ERR_PTR(ret);
+ }
+
+ INIT_WORK(&job->cleanup_op_ctx_work, panthor_vm_bind_job_cleanup_op_ctx_work);
+ kref_init(&job->refcount);
+ job->vm = panthor_vm_get(vm);
+
+ ret = drm_sched_job_init(&job->base, &vm->entity, 1, vm);
+ if (ret)
+ goto err_put_job;
+
+ return &job->base;
+
+err_put_job:
+ panthor_vm_bind_job_put(&job->base);
+ return ERR_PTR(ret);
+}
+
+/**
+ * panthor_vm_bind_job_prepare_resvs() - Prepare VM_BIND job dma_resvs
+ * @exec: The locking/preparation context.
+ * @sched_job: The job to prepare resvs on.
+ *
+ * Locks and prepare the VM resv.
+ *
+ * If this is a map operation, locks and prepares the GEM resv.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_vm_bind_job_prepare_resvs(struct drm_exec *exec,
+ struct drm_sched_job *sched_job)
+{
+ struct panthor_vm_bind_job *job = container_of(sched_job, struct panthor_vm_bind_job, base);
+ int ret;
+
+ /* Acquire the VM lock an reserve a slot for this VM bind job. */
+ ret = drm_gpuvm_prepare_vm(&job->vm->base, exec, 1);
+ if (ret)
+ return ret;
+
+ if (job->ctx.map.vm_bo) {
+ /* Lock/prepare the GEM being mapped. */
+ ret = drm_exec_prepare_obj(exec, job->ctx.map.vm_bo->obj, 1);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_vm_bind_job_update_resvs() - Update the resv objects touched by a job
+ * @exec: drm_exec context.
+ * @sched_job: Job to update the resvs on.
+ */
+void panthor_vm_bind_job_update_resvs(struct drm_exec *exec,
+ struct drm_sched_job *sched_job)
+{
+ struct panthor_vm_bind_job *job = container_of(sched_job, struct panthor_vm_bind_job, base);
+
+ /* Explicit sync => we just register our job finished fence as bookkeep. */
+ drm_gpuvm_resv_add_fence(&job->vm->base, exec,
+ &sched_job->s_fence->finished,
+ DMA_RESV_USAGE_BOOKKEEP,
+ DMA_RESV_USAGE_BOOKKEEP);
+}
+
+void panthor_vm_update_resvs(struct panthor_vm *vm, struct drm_exec *exec,
+ struct dma_fence *fence,
+ enum dma_resv_usage private_usage,
+ enum dma_resv_usage extobj_usage)
+{
+ drm_gpuvm_resv_add_fence(&vm->base, exec, fence, private_usage, extobj_usage);
+}
+
+/**
+ * panthor_vm_bind_exec_sync_op() - Execute a VM_BIND operation synchronously.
+ * @file: File.
+ * @vm: VM targeted by the VM operation.
+ * @op: Data describing the VM operation.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_vm_bind_exec_sync_op(struct drm_file *file,
+ struct panthor_vm *vm,
+ struct drm_panthor_vm_bind_op *op)
+{
+ struct panthor_vm_op_ctx op_ctx;
+ int ret;
+
+ /* No sync objects allowed on synchronous operations. */
+ if (op->syncs.count)
+ return -EINVAL;
+
+ if (!op->size)
+ return 0;
+
+ ret = panthor_vm_bind_prepare_op_ctx(file, vm, op, &op_ctx);
+ if (ret)
+ return ret;
+
+ ret = panthor_vm_exec_op(vm, &op_ctx, false);
+ panthor_vm_cleanup_op_ctx(&op_ctx, vm);
+
+ return ret;
+}
+
+/**
+ * panthor_vm_map_bo_range() - Map a GEM object range to a VM
+ * @vm: VM to map the GEM to.
+ * @bo: GEM object to map.
+ * @offset: Offset in the GEM object.
+ * @size: Size to map.
+ * @va: Virtual address to map the object to.
+ * @flags: Combination of drm_panthor_vm_bind_op_flags flags.
+ * Only map-related flags are valid.
+ *
+ * Internal use only. For userspace requests, use
+ * panthor_vm_bind_exec_sync_op() instead.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_vm_map_bo_range(struct panthor_vm *vm, struct panthor_gem_object *bo,
+ u64 offset, u64 size, u64 va, u32 flags)
+{
+ struct panthor_vm_op_ctx op_ctx;
+ int ret;
+
+ ret = panthor_vm_prepare_map_op_ctx(&op_ctx, vm, bo, offset, size, va, flags);
+ if (ret)
+ return ret;
+
+ ret = panthor_vm_exec_op(vm, &op_ctx, false);
+ panthor_vm_cleanup_op_ctx(&op_ctx, vm);
+
+ return ret;
+}
+
+/**
+ * panthor_vm_unmap_range() - Unmap a portion of the VA space
+ * @vm: VM to unmap the region from.
+ * @va: Virtual address to unmap. Must be 4k aligned.
+ * @size: Size of the region to unmap. Must be 4k aligned.
+ *
+ * Internal use only. For userspace requests, use
+ * panthor_vm_bind_exec_sync_op() instead.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_vm_unmap_range(struct panthor_vm *vm, u64 va, u64 size)
+{
+ struct panthor_vm_op_ctx op_ctx;
+ int ret;
+
+ ret = panthor_vm_prepare_unmap_op_ctx(&op_ctx, vm, va, size);
+ if (ret)
+ return ret;
+
+ ret = panthor_vm_exec_op(vm, &op_ctx, false);
+ panthor_vm_cleanup_op_ctx(&op_ctx, vm);
+
+ return ret;
+}
+
+/**
+ * panthor_vm_prepare_mapped_bos_resvs() - Prepare resvs on VM BOs.
+ * @exec: Locking/preparation context.
+ * @vm: VM targeted by the GPU job.
+ * @slot_count: Number of slots to reserve.
+ *
+ * GPU jobs assume all BOs bound to the VM at the time the job is submitted
+ * are available when the job is executed. In order to guarantee that, we
+ * need to reserve a slot on all BOs mapped to a VM and update this slot with
+ * the job fence after its submission.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_vm_prepare_mapped_bos_resvs(struct drm_exec *exec, struct panthor_vm *vm,
+ u32 slot_count)
+{
+ int ret;
+
+ /* Acquire the VM lock and reserve a slot for this GPU job. */
+ ret = drm_gpuvm_prepare_vm(&vm->base, exec, slot_count);
+ if (ret)
+ return ret;
+
+ return drm_gpuvm_prepare_objects(&vm->base, exec, slot_count);
+}
+
+/**
+ * panthor_mmu_unplug() - Unplug the MMU logic
+ * @ptdev: Device.
+ *
+ * No access to the MMU regs should be done after this function is called.
+ * We suspend the IRQ and disable all VMs to guarantee that.
+ */
+void panthor_mmu_unplug(struct panthor_device *ptdev)
+{
+ panthor_mmu_irq_suspend(&ptdev->mmu->irq);
+
+ mutex_lock(&ptdev->mmu->as.slots_lock);
+ for (u32 i = 0; i < ARRAY_SIZE(ptdev->mmu->as.slots); i++) {
+ struct panthor_vm *vm = ptdev->mmu->as.slots[i].vm;
+
+ if (vm) {
+ drm_WARN_ON(&ptdev->base, panthor_mmu_as_disable(ptdev, i));
+ panthor_vm_release_as_locked(vm);
+ }
+ }
+ mutex_unlock(&ptdev->mmu->as.slots_lock);
+}
+
+static void panthor_mmu_release_wq(struct drm_device *ddev, void *res)
+{
+ destroy_workqueue(res);
+}
+
+/**
+ * panthor_mmu_init() - Initialize the MMU logic.
+ * @ptdev: Device.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_mmu_init(struct panthor_device *ptdev)
+{
+ u32 va_bits = GPU_MMU_FEATURES_VA_BITS(ptdev->gpu_info.mmu_features);
+ struct panthor_mmu *mmu;
+ int ret, irq;
+
+ mmu = drmm_kzalloc(&ptdev->base, sizeof(*mmu), GFP_KERNEL);
+ if (!mmu)
+ return -ENOMEM;
+
+ INIT_LIST_HEAD(&mmu->as.lru_list);
+
+ ret = drmm_mutex_init(&ptdev->base, &mmu->as.slots_lock);
+ if (ret)
+ return ret;
+
+ INIT_LIST_HEAD(&mmu->vm.list);
+ ret = drmm_mutex_init(&ptdev->base, &mmu->vm.lock);
+ if (ret)
+ return ret;
+
+ ptdev->mmu = mmu;
+
+ irq = platform_get_irq_byname(to_platform_device(ptdev->base.dev), "mmu");
+ if (irq <= 0)
+ return -ENODEV;
+
+ ret = panthor_request_mmu_irq(ptdev, &mmu->irq, irq,
+ panthor_mmu_fault_mask(ptdev, ~0));
+ if (ret)
+ return ret;
+
+ mmu->vm.wq = alloc_workqueue("panthor-vm-bind", WQ_UNBOUND, 0);
+ if (!mmu->vm.wq)
+ return -ENOMEM;
+
+ /* On 32-bit kernels, the VA space is limited by the io_pgtable_ops abstraction,
+ * which passes iova as an unsigned long. Patch the mmu_features to reflect this
+ * limitation.
+ */
+ if (sizeof(unsigned long) * 8 < va_bits) {
+ ptdev->gpu_info.mmu_features &= ~GENMASK(7, 0);
+ ptdev->gpu_info.mmu_features |= sizeof(unsigned long) * 8;
+ }
+
+ return drmm_add_action_or_reset(&ptdev->base, panthor_mmu_release_wq, mmu->vm.wq);
+}
+
+#ifdef CONFIG_DEBUG_FS
+static int show_vm_gpuvas(struct panthor_vm *vm, struct seq_file *m)
+{
+ int ret;
+
+ mutex_lock(&vm->op_lock);
+ ret = drm_debugfs_gpuva_info(m, &vm->base);
+ mutex_unlock(&vm->op_lock);
+
+ return ret;
+}
+
+static int show_each_vm(struct seq_file *m, void *arg)
+{
+ struct drm_info_node *node = (struct drm_info_node *)m->private;
+ struct drm_device *ddev = node->minor->dev;
+ struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
+ int (*show)(struct panthor_vm *, struct seq_file *) = node->info_ent->data;
+ struct panthor_vm *vm;
+ int ret = 0;
+
+ mutex_lock(&ptdev->mmu->vm.lock);
+ list_for_each_entry(vm, &ptdev->mmu->vm.list, node) {
+ ret = show(vm, m);
+ if (ret < 0)
+ break;
+
+ seq_puts(m, "\n");
+ }
+ mutex_unlock(&ptdev->mmu->vm.lock);
+
+ return ret;
+}
+
+static struct drm_info_list panthor_mmu_debugfs_list[] = {
+ DRM_DEBUGFS_GPUVA_INFO(show_each_vm, show_vm_gpuvas),
+};
+
+/**
+ * panthor_mmu_debugfs_init() - Initialize MMU debugfs entries
+ * @minor: Minor.
+ */
+void panthor_mmu_debugfs_init(struct drm_minor *minor)
+{
+ drm_debugfs_create_files(panthor_mmu_debugfs_list,
+ ARRAY_SIZE(panthor_mmu_debugfs_list),
+ minor->debugfs_root, minor);
+}
+#endif /* CONFIG_DEBUG_FS */
+
+/**
+ * panthor_mmu_pt_cache_init() - Initialize the page table cache.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_mmu_pt_cache_init(void)
+{
+ pt_cache = kmem_cache_create("panthor-mmu-pt", SZ_4K, SZ_4K, 0, NULL);
+ if (!pt_cache)
+ return -ENOMEM;
+
+ return 0;
+}
+
+/**
+ * panthor_mmu_pt_cache_fini() - Destroy the page table cache.
+ */
+void panthor_mmu_pt_cache_fini(void)
+{
+ kmem_cache_destroy(pt_cache);
+}
diff --git a/drivers/gpu/drm/panthor/panthor_mmu.h b/drivers/gpu/drm/panthor/panthor_mmu.h
new file mode 100644
index 000000000000..f3c1ed19f973
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_mmu.h
@@ -0,0 +1,102 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2019 Linaro, Ltd, Rob Herring <robh@kernel.org> */
+/* Copyright 2023 Collabora ltd. */
+
+#ifndef __PANTHOR_MMU_H__
+#define __PANTHOR_MMU_H__
+
+#include <linux/dma-resv.h>
+
+struct drm_exec;
+struct drm_sched_job;
+struct panthor_gem_object;
+struct panthor_heap_pool;
+struct panthor_vm;
+struct panthor_vma;
+struct panthor_mmu;
+
+int panthor_mmu_init(struct panthor_device *ptdev);
+void panthor_mmu_unplug(struct panthor_device *ptdev);
+void panthor_mmu_pre_reset(struct panthor_device *ptdev);
+void panthor_mmu_post_reset(struct panthor_device *ptdev);
+void panthor_mmu_suspend(struct panthor_device *ptdev);
+void panthor_mmu_resume(struct panthor_device *ptdev);
+
+int panthor_vm_map_bo_range(struct panthor_vm *vm, struct panthor_gem_object *bo,
+ u64 offset, u64 size, u64 va, u32 flags);
+int panthor_vm_unmap_range(struct panthor_vm *vm, u64 va, u64 size);
+struct panthor_gem_object *
+panthor_vm_get_bo_for_va(struct panthor_vm *vm, u64 va, u64 *bo_offset);
+
+int panthor_vm_active(struct panthor_vm *vm);
+void panthor_vm_idle(struct panthor_vm *vm);
+int panthor_vm_as(struct panthor_vm *vm);
+
+struct panthor_heap_pool *
+panthor_vm_get_heap_pool(struct panthor_vm *vm, bool create);
+
+struct panthor_vm *panthor_vm_get(struct panthor_vm *vm);
+void panthor_vm_put(struct panthor_vm *vm);
+struct panthor_vm *panthor_vm_create(struct panthor_device *ptdev, bool for_mcu,
+ u64 kernel_va_start, u64 kernel_va_size,
+ u64 kernel_auto_va_start,
+ u64 kernel_auto_va_size);
+
+int panthor_vm_prepare_mapped_bos_resvs(struct drm_exec *exec,
+ struct panthor_vm *vm,
+ u32 slot_count);
+int panthor_vm_add_bos_resvs_deps_to_job(struct panthor_vm *vm,
+ struct drm_sched_job *job);
+void panthor_vm_add_job_fence_to_bos_resvs(struct panthor_vm *vm,
+ struct drm_sched_job *job);
+
+struct dma_resv *panthor_vm_resv(struct panthor_vm *vm);
+struct drm_gem_object *panthor_vm_root_gem(struct panthor_vm *vm);
+
+void panthor_vm_pool_destroy(struct panthor_file *pfile);
+int panthor_vm_pool_create(struct panthor_file *pfile);
+int panthor_vm_pool_create_vm(struct panthor_device *ptdev,
+ struct panthor_vm_pool *pool,
+ struct drm_panthor_vm_create *args);
+int panthor_vm_pool_destroy_vm(struct panthor_vm_pool *pool, u32 handle);
+struct panthor_vm *panthor_vm_pool_get_vm(struct panthor_vm_pool *pool, u32 handle);
+
+bool panthor_vm_has_unhandled_faults(struct panthor_vm *vm);
+bool panthor_vm_is_unusable(struct panthor_vm *vm);
+
+/*
+ * PANTHOR_VM_KERNEL_AUTO_VA: Use this magic address when you want the GEM
+ * logic to auto-allocate the virtual address in the reserved kernel VA range.
+ */
+#define PANTHOR_VM_KERNEL_AUTO_VA ~0ull
+
+int panthor_vm_alloc_va(struct panthor_vm *vm, u64 va, u64 size,
+ struct drm_mm_node *va_node);
+void panthor_vm_free_va(struct panthor_vm *vm, struct drm_mm_node *va_node);
+
+int panthor_vm_bind_exec_sync_op(struct drm_file *file,
+ struct panthor_vm *vm,
+ struct drm_panthor_vm_bind_op *op);
+
+struct drm_sched_job *
+panthor_vm_bind_job_create(struct drm_file *file,
+ struct panthor_vm *vm,
+ const struct drm_panthor_vm_bind_op *op);
+void panthor_vm_bind_job_put(struct drm_sched_job *job);
+int panthor_vm_bind_job_prepare_resvs(struct drm_exec *exec,
+ struct drm_sched_job *job);
+void panthor_vm_bind_job_update_resvs(struct drm_exec *exec, struct drm_sched_job *job);
+
+void panthor_vm_update_resvs(struct panthor_vm *vm, struct drm_exec *exec,
+ struct dma_fence *fence,
+ enum dma_resv_usage private_usage,
+ enum dma_resv_usage extobj_usage);
+
+int panthor_mmu_pt_cache_init(void);
+void panthor_mmu_pt_cache_fini(void);
+
+#ifdef CONFIG_DEBUG_FS
+void panthor_mmu_debugfs_init(struct drm_minor *minor);
+#endif
+
+#endif
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 07/14] drm/panthor: Add the MMU/VM logical block
2024-01-22 16:30 ` [PATCH v4 07/14] drm/panthor: Add the MMU/VM " Boris Brezillon
@ 2024-02-09 16:51 ` Steven Price
2024-02-10 8:02 ` Boris Brezillon
0 siblings, 1 reply; 51+ messages in thread
From: Steven Price @ 2024-02-09 16:51 UTC (permalink / raw)
To: Boris Brezillon, dri-devel
Cc: Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
On 22/01/2024 16:30, Boris Brezillon wrote:
> MMU and VM management is related and placed in the same source file.
>
> Page table updates are delegated to the io-pgtable-arm driver that's in
> the iommu subsystem.
>
> The VM management logic is based on drm_gpuva_mgr, and is assuming the
> VA space is mostly managed by the usermode driver, except for a reserved
> portion of this VA-space that's used for kernel objects (like the heap
> contexts/chunks).
>
> Both asynchronous and synchronous VM operations are supported, and
> internal helpers are exposed to allow other logical blocks to map their
> buffers in the GPU VA space.
>
> There's one VM_BIND queue per-VM (meaning the Vulkan driver can only
> expose one sparse-binding queue), and this bind queue is managed with
> a 1:1 drm_sched_entity:drm_gpu_scheduler, such that each VM gets its own
> independent execution queue, avoiding VM operation serialization at the
> device level (things are still serialized at the VM level).
>
> The rest is just implementation details that are hopefully well explained
> in the documentation.
panthor_vm_map_pages() looks a bit suspect (see below) and there's a
kerneldoc header which needs updating. But generally it looks fine.
>
> v4:
> - Add an helper to return the VM state
> - Check drmm_mutex_init() return code
> - Remove the VM from the AS reclaim list when panthor_vm_active() is
> called
> - Count the number of active VM users instead of considering there's
> at most one user (several scheduling groups can point to the same
> vM)
> - Pre-allocate a VMA object for unmap operations (unmaps can trigger
> a sm_step_remap() call)
> - Check vm->root_page_table instead of vm->pgtbl_ops to detect if
> the io-pgtable is trying to allocate the root page table
> - Don't memset() the va_node in panthor_vm_alloc_va(), make it a
> caller requirement
> - Fix the kernel doc in a few places
> - Drop the panthor_vm::base offset constraint and modify
> panthor_vm_put() to explicitly check for a NULL value
> - Fix unbalanced vm_bo refcount in panthor_gpuva_sm_step_remap()
> - Drop stale comments about the shared_bos list
> - Patch mmu_features::va_bits on 32-bit builds to reflect the
> io_pgtable limitation and let the UMD know about it
>
> v3:
> - Add acks for the MIT/GPL2 relicensing
> - Propagate MMU faults to the scheduler
> - Move pages pinning/unpinning out of the dma_signalling path
> - Fix 32-bit support
> - Rework the user/kernel VA range calculation
> - Make the auto-VA range explicit (auto-VA range doesn't cover the full
> kernel-VA range on the MCU VM)
> - Let callers of panthor_vm_alloc_va() allocate the drm_mm_node
> (embedded in panthor_kernel_bo now)
> - Adjust things to match the latest drm_gpuvm changes (extobj tracking,
> resv prep and more)
> - Drop the per-AS lock and use slots_lock (fixes a race on vm->as.id)
> - Set as.id to -1 when reusing an address space from the LRU list
> - Drop misleading comment about page faults
> - Remove check for irq being assigned in panthor_mmu_unplug()
>
> Co-developed-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
> Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
> Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
> ---
> drivers/gpu/drm/panthor/panthor_mmu.c | 2760 +++++++++++++++++++++++++
> drivers/gpu/drm/panthor/panthor_mmu.h | 102 +
> 2 files changed, 2862 insertions(+)
> create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.h
>
> diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
> new file mode 100644
> index 000000000000..d3ce29cd0662
<snip>
> +static int
> +panthor_vm_map_pages(struct panthor_vm *vm, u64 iova, int prot,
> + struct sg_table *sgt, u64 offset, u64 size)
> +{
> + struct panthor_device *ptdev = vm->ptdev;
> + unsigned int count;
> + struct scatterlist *sgl;
> + struct io_pgtable_ops *ops = vm->pgtbl_ops;
> + u64 start_iova = iova;
> + int ret;
> +
> + if (!size)
> + return 0;
> +
> + for_each_sgtable_dma_sg(sgt, sgl, count) {
> + dma_addr_t paddr = sg_dma_address(sgl);
> + size_t len = sg_dma_len(sgl);
> +
> + if (len <= offset) {
> + offset -= len;
> + continue;
> + }
> +
> + paddr -= offset;
Shouldn't that be "+="?
> + len -= offset;
> +
> + if (size >= 0) {
size is unsigned... so this is always true!
> + len = min_t(size_t, len, size);
> + size -= len;
> + }
> +
> + drm_dbg(&ptdev->base, "map: as=%d, iova=%llx, paddr=%pad, len=%zx",
> + vm->as.id, iova, &paddr, len);
> +
> + while (len) {
> + size_t pgcount, mapped = 0;
> + size_t pgsize = get_pgsize(iova | paddr, len, &pgcount);
> +
> + ret = ops->map_pages(ops, iova, paddr, pgsize, pgcount, prot,
> + GFP_KERNEL, &mapped);
> + iova += mapped;
> + paddr += mapped;
> + len -= mapped;
> +
> + if (drm_WARN_ON(&ptdev->base, !ret && !mapped))
> + ret = -ENOMEM;
> +
> + if (ret) {
> + /* If something failed, unmap what we've already mapped before
> + * returning. The unmap call is not supposed to fail.
> + */
> + drm_WARN_ON(&ptdev->base,
> + panthor_vm_unmap_pages(vm, start_iova,
> + iova - start_iova));
> + return ret;
> + }
> + }
> +
> + if (!size)
> + break;
> + }
> +
> + return panthor_vm_flush_range(vm, start_iova, iova - start_iova);
> +}
> +
<snip>
> +static void panthor_vm_destroy(struct panthor_vm *vm)
> +{
> + if (!vm)
> + return;
> +
> + vm->destroyed = true;
> +
> + mutex_lock(&vm->heaps.lock);
> + panthor_heap_pool_destroy(vm->heaps.pool);
> + vm->heaps.pool = NULL;
> + mutex_unlock(&vm->heaps.lock);
> +
> + drm_WARN_ON(&vm->ptdev->base,
> + panthor_vm_unmap_range(vm, vm->base.mm_start, vm->base.mm_range));
> + panthor_vm_put(vm);
> +}
> +
> +/**
> + * panthor_vm_destroy() - Destroy a VM.
^^^^^^^^^^^^^^^^^^ needs updating to the new function name.
Steve
> + * @pool: VM pool.
> + * @handle: VM handle.
> + *
> + * This function doesn't free the VM object or its resources, it just kills
> + * all mappings, and makes sure nothing can be mapped after that point.
> + *
> + * If there was any active jobs at the time this function is called, these
> + * jobs should experience page faults and be killed as a result.
> + *
> + * The VM resources are freed when the last reference on the VM object is
> + * dropped.
> + */
> +int panthor_vm_pool_destroy_vm(struct panthor_vm_pool *pool, u32 handle)
> +{
> + struct panthor_vm *vm;
> +
> + vm = xa_erase(&pool->xa, handle);
> +
> + panthor_vm_destroy(vm);
> +
> + return vm ? 0 : -EINVAL;
> +}
<snip>
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 07/14] drm/panthor: Add the MMU/VM logical block
2024-02-09 16:51 ` Steven Price
@ 2024-02-10 8:02 ` Boris Brezillon
0 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-02-10 8:02 UTC (permalink / raw)
To: Steven Price
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen,
Grant Likely
Hi Steve,
On Fri, 9 Feb 2024 16:51:46 +0000
Steven Price <steven.price@arm.com> wrote:
> On 22/01/2024 16:30, Boris Brezillon wrote:
> > MMU and VM management is related and placed in the same source file.
> >
> > Page table updates are delegated to the io-pgtable-arm driver that's in
> > the iommu subsystem.
> >
> > The VM management logic is based on drm_gpuva_mgr, and is assuming the
> > VA space is mostly managed by the usermode driver, except for a reserved
> > portion of this VA-space that's used for kernel objects (like the heap
> > contexts/chunks).
> >
> > Both asynchronous and synchronous VM operations are supported, and
> > internal helpers are exposed to allow other logical blocks to map their
> > buffers in the GPU VA space.
> >
> > There's one VM_BIND queue per-VM (meaning the Vulkan driver can only
> > expose one sparse-binding queue), and this bind queue is managed with
> > a 1:1 drm_sched_entity:drm_gpu_scheduler, such that each VM gets its own
> > independent execution queue, avoiding VM operation serialization at the
> > device level (things are still serialized at the VM level).
> >
> > The rest is just implementation details that are hopefully well explained
> > in the documentation.
>
> panthor_vm_map_pages() looks a bit suspect (see below) and there's a
> kerneldoc header which needs updating. But generally it looks fine.
>
> >
> > v4:
> > - Add an helper to return the VM state
> > - Check drmm_mutex_init() return code
> > - Remove the VM from the AS reclaim list when panthor_vm_active() is
> > called
> > - Count the number of active VM users instead of considering there's
> > at most one user (several scheduling groups can point to the same
> > vM)
> > - Pre-allocate a VMA object for unmap operations (unmaps can trigger
> > a sm_step_remap() call)
> > - Check vm->root_page_table instead of vm->pgtbl_ops to detect if
> > the io-pgtable is trying to allocate the root page table
> > - Don't memset() the va_node in panthor_vm_alloc_va(), make it a
> > caller requirement
> > - Fix the kernel doc in a few places
> > - Drop the panthor_vm::base offset constraint and modify
> > panthor_vm_put() to explicitly check for a NULL value
> > - Fix unbalanced vm_bo refcount in panthor_gpuva_sm_step_remap()
> > - Drop stale comments about the shared_bos list
> > - Patch mmu_features::va_bits on 32-bit builds to reflect the
> > io_pgtable limitation and let the UMD know about it
> >
> > v3:
> > - Add acks for the MIT/GPL2 relicensing
> > - Propagate MMU faults to the scheduler
> > - Move pages pinning/unpinning out of the dma_signalling path
> > - Fix 32-bit support
> > - Rework the user/kernel VA range calculation
> > - Make the auto-VA range explicit (auto-VA range doesn't cover the full
> > kernel-VA range on the MCU VM)
> > - Let callers of panthor_vm_alloc_va() allocate the drm_mm_node
> > (embedded in panthor_kernel_bo now)
> > - Adjust things to match the latest drm_gpuvm changes (extobj tracking,
> > resv prep and more)
> > - Drop the per-AS lock and use slots_lock (fixes a race on vm->as.id)
> > - Set as.id to -1 when reusing an address space from the LRU list
> > - Drop misleading comment about page faults
> > - Remove check for irq being assigned in panthor_mmu_unplug()
> >
> > Co-developed-by: Steven Price <steven.price@arm.com>
> > Signed-off-by: Steven Price <steven.price@arm.com>
> > Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> > Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
> > Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
> > Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
> > ---
> > drivers/gpu/drm/panthor/panthor_mmu.c | 2760 +++++++++++++++++++++++++
> > drivers/gpu/drm/panthor/panthor_mmu.h | 102 +
> > 2 files changed, 2862 insertions(+)
> > create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.c
> > create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.h
> >
> > diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
> > new file mode 100644
> > index 000000000000..d3ce29cd0662
>
> <snip>
>
> > +static int
> > +panthor_vm_map_pages(struct panthor_vm *vm, u64 iova, int prot,
> > + struct sg_table *sgt, u64 offset, u64 size)
> > +{
> > + struct panthor_device *ptdev = vm->ptdev;
> > + unsigned int count;
> > + struct scatterlist *sgl;
> > + struct io_pgtable_ops *ops = vm->pgtbl_ops;
> > + u64 start_iova = iova;
> > + int ret;
> > +
> > + if (!size)
> > + return 0;
> > +
> > + for_each_sgtable_dma_sg(sgt, sgl, count) {
> > + dma_addr_t paddr = sg_dma_address(sgl);
> > + size_t len = sg_dma_len(sgl);
> > +
> > + if (len <= offset) {
> > + offset -= len;
> > + continue;
> > + }
> > +
> > + paddr -= offset;
>
> Shouldn't that be "+="?
It should definitely be +=. Will fix that in v5.
>
> > + len -= offset;
> > +
> > + if (size >= 0) {
>
> size is unsigned... so this is always true!
I'll drop the condition.
>
> > + len = min_t(size_t, len, size);
> > + size -= len;
> > + }
> > +
> > + drm_dbg(&ptdev->base, "map: as=%d, iova=%llx, paddr=%pad, len=%zx",
> > + vm->as.id, iova, &paddr, len);
> > +
> > + while (len) {
> > + size_t pgcount, mapped = 0;
> > + size_t pgsize = get_pgsize(iova | paddr, len, &pgcount);
> > +
> > + ret = ops->map_pages(ops, iova, paddr, pgsize, pgcount, prot,
> > + GFP_KERNEL, &mapped);
> > + iova += mapped;
> > + paddr += mapped;
> > + len -= mapped;
> > +
> > + if (drm_WARN_ON(&ptdev->base, !ret && !mapped))
> > + ret = -ENOMEM;
> > +
> > + if (ret) {
> > + /* If something failed, unmap what we've already mapped before
> > + * returning. The unmap call is not supposed to fail.
> > + */
> > + drm_WARN_ON(&ptdev->base,
> > + panthor_vm_unmap_pages(vm, start_iova,
> > + iova - start_iova));
> > + return ret;
> > + }
> > + }
> > +
> > + if (!size)
> > + break;
> > + }
> > +
> > + return panthor_vm_flush_range(vm, start_iova, iova - start_iova);
> > +}
> > +
>
> <snip>
>
> > +static void panthor_vm_destroy(struct panthor_vm *vm)
> > +{
> > + if (!vm)
> > + return;
> > +
> > + vm->destroyed = true;
> > +
> > + mutex_lock(&vm->heaps.lock);
> > + panthor_heap_pool_destroy(vm->heaps.pool);
> > + vm->heaps.pool = NULL;
> > + mutex_unlock(&vm->heaps.lock);
> > +
> > + drm_WARN_ON(&vm->ptdev->base,
> > + panthor_vm_unmap_range(vm, vm->base.mm_start, vm->base.mm_range));
> > + panthor_vm_put(vm);
> > +}
> > +
> > +/**
> > + * panthor_vm_destroy() - Destroy a VM.
> ^^^^^^^^^^^^^^^^^^ needs updating to the new function name.
>
And I'll fix the kernel doc.
Thanks,
Boris
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 08/14] drm/panthor: Add the FW logical block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (6 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 07/14] drm/panthor: Add the MMU/VM " Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-01-22 16:30 ` [PATCH v4 09/14] drm/panthor: Add the heap " Boris Brezillon
` (9 subsequent siblings)
17 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
Contains everything that's FW related, that includes the code dealing
with the microcontroller unit (MCU) that's running the FW, and anything
related to allocating memory shared between the FW and the CPU.
A few global FW events are processed in the IRQ handler, the rest is
forwarded to the scheduler, since scheduling is the primary reason for
the FW existence, and also the main source of FW <-> kernel
interactions.
v4:
- Add a MODULE_FIRMWARE() entry for gen 10.8
- Fix a wrong return ERR_PTR() in panthor_fw_load_section_entry()
- Fix typos
- Add Steve's R-b
v3:
- Make the FW path more future-proof (Liviu)
- Use one waitqueue for all FW events
- Simplify propagation of FW events to the scheduler logic
- Drop the panthor_fw_mem abstraction and use panthor_kernel_bo instead
- Account for the panthor_vm changes
- Replace magic number with 0x7fffffff with ~0 to better signify that
it's the maximum permitted value.
- More accurate rounding when computing the firmware timeout.
- Add a 'sub iterator' helper function. This also adds a check that a
firmware entry doesn't overflow the firmware image.
- Drop __packed from FW structures, natural alignment is good enough.
- Other minor code improvements.
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
---
drivers/gpu/drm/panthor/panthor_fw.c | 1334 ++++++++++++++++++++++++++
drivers/gpu/drm/panthor/panthor_fw.h | 504 ++++++++++
2 files changed, 1838 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_fw.c
create mode 100644 drivers/gpu/drm/panthor/panthor_fw.h
diff --git a/drivers/gpu/drm/panthor/panthor_fw.c b/drivers/gpu/drm/panthor/panthor_fw.c
new file mode 100644
index 000000000000..dc40f350a0de
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_fw.c
@@ -0,0 +1,1334 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2023 Collabora ltd. */
+
+#ifdef CONFIG_ARM_ARCH_TIMER
+#include <asm/arch_timer.h>
+#endif
+
+#include <linux/clk.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/iopoll.h>
+#include <linux/iosys-map.h>
+#include <linux/mutex.h>
+#include <linux/platform_device.h>
+
+#include <drm/drm_drv.h>
+#include <drm/drm_managed.h>
+
+#include "panthor_device.h"
+#include "panthor_gem.h"
+#include "panthor_gpu.h"
+#include "panthor_regs.h"
+#include "panthor_fw.h"
+#include "panthor_mmu.h"
+#include "panthor_sched.h"
+
+#define CSF_FW_NAME "mali_csffw.bin"
+
+#define PING_INTERVAL_MS 12000
+#define PROGRESS_TIMEOUT_CYCLES (5ull * 500 * 1024 * 1024)
+#define PROGRESS_TIMEOUT_SCALE_SHIFT 10
+#define IDLE_HYSTERESIS_US 800
+#define PWROFF_HYSTERESIS_US 10000
+
+/**
+ * struct panthor_fw_binary_hdr - Firmware binary header.
+ */
+struct panthor_fw_binary_hdr {
+ /** @magic: Magic value to check binary validity. */
+ u32 magic;
+#define CSF_FW_BINARY_HEADER_MAGIC 0xc3f13a6e
+
+ /** @minor: Minor FW version. */
+ u8 minor;
+
+ /** @major: Major FW version. */
+ u8 major;
+#define CSF_FW_BINARY_HEADER_MAJOR_MAX 0
+
+ /** @padding1: MBZ. */
+ u16 padding1;
+
+ /** @version_hash: FW version hash. */
+ u32 version_hash;
+
+ /** @padding2: MBZ. */
+ u32 padding2;
+
+ /** @size: FW binary size. */
+ u32 size;
+};
+
+/**
+ * enum panthor_fw_binary_entry_type - Firmware binary entry type
+ */
+enum panthor_fw_binary_entry_type {
+ /** @CSF_FW_BINARY_ENTRY_TYPE_IFACE: Host <-> FW interface. */
+ CSF_FW_BINARY_ENTRY_TYPE_IFACE = 0,
+
+ /** @CSF_FW_BINARY_ENTRY_TYPE_CONFIG: FW config. */
+ CSF_FW_BINARY_ENTRY_TYPE_CONFIG = 1,
+
+ /** @CSF_FW_BINARY_ENTRY_TYPE_FUTF_TEST: Unit-tests. */
+ CSF_FW_BINARY_ENTRY_TYPE_FUTF_TEST = 2,
+
+ /** @CSF_FW_BINARY_ENTRY_TYPE_TRACE_BUFFER: Trace buffer interface. */
+ CSF_FW_BINARY_ENTRY_TYPE_TRACE_BUFFER = 3,
+
+ /** @CSF_FW_BINARY_ENTRY_TYPE_TIMELINE_METADATA: Timeline metadata interface. */
+ CSF_FW_BINARY_ENTRY_TYPE_TIMELINE_METADATA = 4,
+};
+
+#define CSF_FW_BINARY_ENTRY_TYPE(ehdr) ((ehdr) & 0xff)
+#define CSF_FW_BINARY_ENTRY_SIZE(ehdr) (((ehdr) >> 8) & 0xff)
+#define CSF_FW_BINARY_ENTRY_UPDATE BIT(30)
+#define CSF_FW_BINARY_ENTRY_OPTIONAL BIT(31)
+
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_RD BIT(0)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_WR BIT(1)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_EX BIT(2)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_NONE (0 << 3)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_CACHED (1 << 3)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_UNCACHED_COHERENT (2 << 3)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_CACHED_COHERENT (3 << 3)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_MASK GENMASK(4, 3)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_PROT BIT(5)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_SHARED BIT(30)
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_ZERO BIT(31)
+
+#define CSF_FW_BINARY_IFACE_ENTRY_RD_SUPPORTED_FLAGS \
+ (CSF_FW_BINARY_IFACE_ENTRY_RD_RD | \
+ CSF_FW_BINARY_IFACE_ENTRY_RD_WR | \
+ CSF_FW_BINARY_IFACE_ENTRY_RD_EX | \
+ CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_MASK | \
+ CSF_FW_BINARY_IFACE_ENTRY_RD_PROT | \
+ CSF_FW_BINARY_IFACE_ENTRY_RD_SHARED | \
+ CSF_FW_BINARY_IFACE_ENTRY_RD_ZERO)
+
+/**
+ * struct panthor_fw_binary_section_entry_hdr - Describes a section of FW binary
+ */
+struct panthor_fw_binary_section_entry_hdr {
+ /** @flags: Section flags. */
+ u32 flags;
+
+ /** @va: MCU virtual range to map this binary section to. */
+ struct {
+ /** @start: Start address. */
+ u32 start;
+
+ /** @end: End address. */
+ u32 end;
+ } va;
+
+ /** @data: Data to initialize the FW section with. */
+ struct {
+ /** @start: Start offset in the FW binary. */
+ u32 start;
+
+ /** @end: End offset in the FW binary. */
+ u32 end;
+ } data;
+};
+
+/**
+ * struct panthor_fw_binary_iter - Firmware binary iterator
+ *
+ * Used to parse a firmware binary.
+ */
+struct panthor_fw_binary_iter {
+ /** @data: FW binary data. */
+ const void *data;
+
+ /** @size: FW binary size. */
+ size_t size;
+
+ /** @offset: Iterator offset. */
+ size_t offset;
+};
+
+/**
+ * struct panthor_fw_section - FW section
+ */
+struct panthor_fw_section {
+ /** @node: Used to keep track of FW sections. */
+ struct list_head node;
+
+ /** @flags: Section flags, as encoded in the FW binary. */
+ u32 flags;
+
+ /** @mem: Section memory. */
+ struct panthor_kernel_bo *mem;
+
+ /**
+ * @name: Name of the section, as specified in the binary.
+ *
+ * Can be NULL.
+ */
+ const char *name;
+
+ /**
+ * @data: Initial data copied to the FW memory.
+ *
+ * We keep data around so we can reload sections after a reset.
+ */
+ struct {
+ /** @buf: Buffed used to store init data. */
+ const void *buf;
+
+ /** @size: Size of @buf in bytes. */
+ size_t size;
+ } data;
+};
+
+#define CSF_MCU_SHARED_REGION_START 0x04000000ULL
+#define CSF_MCU_SHARED_REGION_SIZE 0x04000000ULL
+
+#define MIN_CS_PER_CSG 8
+#define MIN_CSGS 3
+#define MAX_CSG_PRIO 0xf
+
+#define CSF_IFACE_VERSION(major, minor, patch) \
+ (((major) << 24) | ((minor) << 16) | (patch))
+#define CSF_IFACE_VERSION_MAJOR(v) ((v) >> 24)
+#define CSF_IFACE_VERSION_MINOR(v) (((v) >> 16) & 0xff)
+#define CSF_IFACE_VERSION_PATCH(v) ((v) & 0xffff)
+
+#define CSF_GROUP_CONTROL_OFFSET 0x1000
+#define CSF_STREAM_CONTROL_OFFSET 0x40
+#define CSF_UNPRESERVED_REG_COUNT 4
+
+/**
+ * struct panthor_fw_iface - FW interfaces
+ */
+struct panthor_fw_iface {
+ /** @global: Global interface. */
+ struct panthor_fw_global_iface global;
+
+ /** @groups: Group slot interfaces. */
+ struct panthor_fw_csg_iface groups[MAX_CSGS];
+
+ /** @streams: Command stream slot interfaces. */
+ struct panthor_fw_cs_iface streams[MAX_CSGS][MAX_CS_PER_CSG];
+};
+
+/**
+ * struct panthor_fw - Firmware management
+ */
+struct panthor_fw {
+ /** @vm: MCU VM. */
+ struct panthor_vm *vm;
+
+ /** @sections: List of FW sections. */
+ struct list_head sections;
+
+ /** @shared_section: The section containing the FW interfaces. */
+ struct panthor_fw_section *shared_section;
+
+ /** @iface: FW interfaces. */
+ struct panthor_fw_iface iface;
+
+ /** @watchdog: Collection of fields relating to the FW watchdog. */
+ struct {
+ /** @ping_work: Delayed work used to ping the FW. */
+ struct delayed_work ping_work;
+ } watchdog;
+
+ /**
+ * @req_waitqueue: FW request waitqueue.
+ *
+ * Everytime a request is sent to a command stream group or the global
+ * interface, the caller will first busy wait for the request to be
+ * acknowledged, and then fallback to a sleeping wait.
+ *
+ * This wait queue is here to support the sleeping wait flavor.
+ */
+ wait_queue_head_t req_waitqueue;
+
+ /** @booted: True is the FW is booted */
+ bool booted;
+
+ /**
+ * @fast_reset: True if the post_reset logic can proceed with a fast reset.
+ *
+ * A fast reset is just a reset where the driver doesn't reload the FW sections.
+ *
+ * Any time the firmware is properly suspended, a fast reset can take place.
+ * On the other hand, if the halt operation failed, the driver will reload
+ * all sections to make sure we start from a fresh state.
+ */
+ bool fast_reset;
+
+ /** @irq: Job irq data. */
+ struct panthor_irq irq;
+};
+
+struct panthor_vm *panthor_fw_vm(struct panthor_device *ptdev)
+{
+ return ptdev->fw->vm;
+}
+
+/**
+ * panthor_fw_get_glb_iface() - Get the global interface
+ * @ptdev: Device.
+ *
+ * Return: The global interface.
+ */
+struct panthor_fw_global_iface *
+panthor_fw_get_glb_iface(struct panthor_device *ptdev)
+{
+ return &ptdev->fw->iface.global;
+}
+
+/**
+ * panthor_fw_get_csg_iface() - Get a command stream group slot interface
+ * @ptdev: Device.
+ * @csg_slot: Index of the command stream group slot.
+ *
+ * Return: The command stream group slot interface.
+ */
+struct panthor_fw_csg_iface *
+panthor_fw_get_csg_iface(struct panthor_device *ptdev, u32 csg_slot)
+{
+ if (drm_WARN_ON(&ptdev->base, csg_slot >= MAX_CSGS))
+ return NULL;
+
+ return &ptdev->fw->iface.groups[csg_slot];
+}
+
+/**
+ * panthor_fw_get_cs_iface() - Get a command stream slot interface
+ * @ptdev: Device.
+ * @csg_slot: Index of the command stream group slot.
+ * @cs_slot: Index of the command stream slot.
+ *
+ * Return: The command stream slot interface.
+ */
+struct panthor_fw_cs_iface *
+panthor_fw_get_cs_iface(struct panthor_device *ptdev, u32 csg_slot, u32 cs_slot)
+{
+ if (drm_WARN_ON(&ptdev->base, csg_slot >= MAX_CSGS || cs_slot > MAX_CS_PER_CSG))
+ return NULL;
+
+ return &ptdev->fw->iface.streams[csg_slot][cs_slot];
+}
+
+/**
+ * panthor_fw_conv_timeout() - Convert a timeout into a cycle-count
+ * @ptdev: Device.
+ * @timeout_us: Timeout expressed in micro-seconds.
+ *
+ * The FW has two timer sources: the GPU counter or arch-timer. We need
+ * to express timeouts in term of number of cycles and specify which
+ * timer source should be used.
+ *
+ * Return: A value suitable for timeout fields in the global interface.
+ */
+static u32 panthor_fw_conv_timeout(struct panthor_device *ptdev, u32 timeout_us)
+{
+ bool use_cycle_counter = false;
+ u32 timer_rate = 0;
+ u64 mod_cycles;
+
+#ifdef CONFIG_ARM_ARCH_TIMER
+ timer_rate = arch_timer_get_cntfrq();
+#endif
+
+ if (!timer_rate) {
+ use_cycle_counter = true;
+ timer_rate = clk_get_rate(ptdev->clks.core);
+ }
+
+ if (drm_WARN_ON(&ptdev->base, !timer_rate)) {
+ /* We couldn't get a valid clock rate, let's just pick the
+ * maximum value so the FW still handles the core
+ * power on/off requests.
+ */
+ return GLB_TIMER_VAL(~0) |
+ GLB_TIMER_SOURCE_GPU_COUNTER;
+ }
+
+ mod_cycles = DIV_ROUND_UP_ULL((u64)timeout_us * timer_rate,
+ 1000000ull << 10);
+ if (drm_WARN_ON(&ptdev->base, mod_cycles > GLB_TIMER_VAL(~0)))
+ mod_cycles = GLB_TIMER_VAL(~0);
+
+ return GLB_TIMER_VAL(mod_cycles) |
+ (use_cycle_counter ? GLB_TIMER_SOURCE_GPU_COUNTER : 0);
+}
+
+static int panthor_fw_binary_iter_read(struct panthor_device *ptdev,
+ struct panthor_fw_binary_iter *iter,
+ void *out, size_t size)
+{
+ size_t new_offset = iter->offset + size;
+
+ if (new_offset > iter->size || new_offset < iter->offset) {
+ drm_err(&ptdev->base, "Firmware too small\n");
+ return -EINVAL;
+ }
+
+ memcpy(out, iter->data + iter->offset, size);
+ iter->offset = new_offset;
+ return 0;
+}
+
+static int panthor_fw_binary_sub_iter_init(struct panthor_device *ptdev,
+ struct panthor_fw_binary_iter *iter,
+ struct panthor_fw_binary_iter *sub_iter,
+ size_t size)
+{
+ size_t new_offset = iter->offset + size;
+
+ if (new_offset > iter->size || new_offset < iter->offset) {
+ drm_err(&ptdev->base, "Firmware entry too long\n");
+ return -EINVAL;
+ }
+
+ sub_iter->offset = 0;
+ sub_iter->data = iter->data + iter->offset;
+ sub_iter->size = size;
+ iter->offset = new_offset;
+ return 0;
+}
+
+static void panthor_fw_init_section_mem(struct panthor_device *ptdev,
+ struct panthor_fw_section *section)
+{
+ bool was_mapped = !!section->mem->kmap;
+ int ret;
+
+ if (!section->data.size &&
+ !(section->flags & CSF_FW_BINARY_IFACE_ENTRY_RD_ZERO))
+ return;
+
+ ret = panthor_kernel_bo_vmap(section->mem);
+ if (drm_WARN_ON(&ptdev->base, ret))
+ return;
+
+ memcpy(section->mem->kmap, section->data.buf, section->data.size);
+ if (section->flags & CSF_FW_BINARY_IFACE_ENTRY_RD_ZERO) {
+ memset(section->mem->kmap + section->data.size, 0,
+ panthor_kernel_bo_size(section->mem) - section->data.size);
+ }
+
+ if (!was_mapped)
+ panthor_kernel_bo_vunmap(section->mem);
+}
+
+/**
+ * panthor_fw_alloc_queue_iface_mem() - Allocate a ring-buffer interfaces.
+ * @ptdev: Device.
+ * @input: Pointer holding the input interface on success.
+ * Should be ignored on failure.
+ * @output: Pointer holding the output interface on success.
+ * Should be ignored on failure.
+ * @input_fw_va: Pointer holding the input interface FW VA on success.
+ * Should be ignored on failure.
+ * @output_fw_va: Pointer holding the output interface FW VA on success.
+ * Should be ignored on failure.
+ *
+ * Allocates panthor_fw_ringbuf_{input,out}_iface interfaces. The input
+ * interface is at offset 0, and the output interface at offset 4096.
+ *
+ * Return: A valid pointer in case of success, an ERR_PTR() otherwise.
+ */
+struct panthor_kernel_bo *
+panthor_fw_alloc_queue_iface_mem(struct panthor_device *ptdev,
+ struct panthor_fw_ringbuf_input_iface **input,
+ const struct panthor_fw_ringbuf_output_iface **output,
+ u32 *input_fw_va, u32 *output_fw_va)
+{
+ struct panthor_kernel_bo *mem;
+ int ret;
+
+ mem = panthor_kernel_bo_create(ptdev, ptdev->fw->vm, SZ_8K,
+ DRM_PANTHOR_BO_NO_MMAP,
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC |
+ DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED,
+ PANTHOR_VM_KERNEL_AUTO_VA);
+ if (IS_ERR(mem))
+ return mem;
+
+ ret = panthor_kernel_bo_vmap(mem);
+ if (ret) {
+ panthor_kernel_bo_destroy(panthor_fw_vm(ptdev), mem);
+ return ERR_PTR(ret);
+ }
+
+ memset(mem->kmap, 0, panthor_kernel_bo_size(mem));
+ *input = mem->kmap;
+ *output = mem->kmap + SZ_4K;
+ *input_fw_va = panthor_kernel_bo_gpuva(mem);
+ *output_fw_va = *input_fw_va + SZ_4K;
+
+ return mem;
+}
+
+/**
+ * panthor_fw_alloc_suspend_buf_mem() - Allocate a suspend buffer for a command stream group.
+ * @ptdev: Device.
+ * @size: Size of the suspend buffer.
+ *
+ * Return: A valid pointer in case of success, an ERR_PTR() otherwise.
+ */
+struct panthor_kernel_bo *
+panthor_fw_alloc_suspend_buf_mem(struct panthor_device *ptdev, size_t size)
+{
+ return panthor_kernel_bo_create(ptdev, panthor_fw_vm(ptdev), size,
+ DRM_PANTHOR_BO_NO_MMAP,
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC,
+ PANTHOR_VM_KERNEL_AUTO_VA);
+}
+
+static int panthor_fw_load_section_entry(struct panthor_device *ptdev,
+ const struct firmware *fw,
+ struct panthor_fw_binary_iter *iter,
+ u32 ehdr)
+{
+ struct panthor_fw_binary_section_entry_hdr hdr;
+ struct panthor_fw_section *section;
+ u32 section_size;
+ u32 name_len;
+ int ret;
+
+ ret = panthor_fw_binary_iter_read(ptdev, iter, &hdr, sizeof(hdr));
+ if (ret)
+ return ret;
+
+ if (hdr.data.end < hdr.data.start) {
+ drm_err(&ptdev->base, "Firmware corrupted, data.end < data.start (0x%x < 0x%x)\n",
+ hdr.data.end, hdr.data.start);
+ return -EINVAL;
+ }
+
+ if (hdr.va.end < hdr.va.start) {
+ drm_err(&ptdev->base, "Firmware corrupted, hdr.va.end < hdr.va.start (0x%x < 0x%x)\n",
+ hdr.va.end, hdr.va.start);
+ return -EINVAL;
+ }
+
+ if (hdr.data.end > fw->size) {
+ drm_err(&ptdev->base, "Firmware corrupted, file truncated? data_end=0x%x > fw size=0x%zx\n",
+ hdr.data.end, fw->size);
+ return -EINVAL;
+ }
+
+ if ((hdr.va.start & ~PAGE_MASK) != 0 ||
+ (hdr.va.end & ~PAGE_MASK) != 0) {
+ drm_err(&ptdev->base, "Firmware corrupted, virtual addresses not page aligned: 0x%x-0x%x\n",
+ hdr.va.start, hdr.va.end);
+ return -EINVAL;
+ }
+
+ if (hdr.flags & ~CSF_FW_BINARY_IFACE_ENTRY_RD_SUPPORTED_FLAGS) {
+ drm_err(&ptdev->base, "Firmware contains interface with unsupported flags (0x%x)\n",
+ hdr.flags);
+ return -EINVAL;
+ }
+
+ if (hdr.flags & CSF_FW_BINARY_IFACE_ENTRY_RD_PROT) {
+ drm_warn(&ptdev->base,
+ "Firmware protected mode entry not be supported, ignoring");
+ return 0;
+ }
+
+ if (hdr.va.start == CSF_MCU_SHARED_REGION_START &&
+ !(hdr.flags & CSF_FW_BINARY_IFACE_ENTRY_RD_SHARED)) {
+ drm_err(&ptdev->base,
+ "Interface at 0x%llx must be shared", CSF_MCU_SHARED_REGION_START);
+ return -EINVAL;
+ }
+
+ name_len = iter->size - iter->offset;
+
+ section = drmm_kzalloc(&ptdev->base, sizeof(*section), GFP_KERNEL);
+ if (!section)
+ return -ENOMEM;
+
+ list_add_tail(§ion->node, &ptdev->fw->sections);
+ section->flags = hdr.flags;
+ section->data.size = hdr.data.end - hdr.data.start;
+
+ if (section->data.size > 0) {
+ void *data = drmm_kmalloc(&ptdev->base, section->data.size, GFP_KERNEL);
+
+ if (!data)
+ return -ENOMEM;
+
+ memcpy(data, fw->data + hdr.data.start, section->data.size);
+ section->data.buf = data;
+ }
+
+ if (name_len > 0) {
+ char *name = drmm_kmalloc(&ptdev->base, name_len + 1, GFP_KERNEL);
+
+ if (!name)
+ return -ENOMEM;
+
+ memcpy(name, iter->data + iter->offset, name_len);
+ name[name_len] = '\0';
+ section->name = name;
+ }
+
+ section_size = hdr.va.end - hdr.va.start;
+ if (section_size) {
+ u32 cache_mode = hdr.flags & CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_MASK;
+ struct panthor_gem_object *bo;
+ u32 vm_map_flags = 0;
+ struct sg_table *sgt;
+ u64 va = hdr.va.start;
+
+ if (!(hdr.flags & CSF_FW_BINARY_IFACE_ENTRY_RD_WR))
+ vm_map_flags |= DRM_PANTHOR_VM_BIND_OP_MAP_READONLY;
+
+ if (!(hdr.flags & CSF_FW_BINARY_IFACE_ENTRY_RD_EX))
+ vm_map_flags |= DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC;
+
+ /* TODO: CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_*_COHERENT are mapped to
+ * non-cacheable for now. We might want to introduce a new
+ * IOMMU_xxx flag (or abuse IOMMU_MMIO, which maps to device
+ * memory and is currently not used by our driver) for
+ * AS_MEMATTR_AARCH64_SHARED memory, so we can take benefit
+ * of IO-coherent systems.
+ */
+ if (cache_mode != CSF_FW_BINARY_IFACE_ENTRY_RD_CACHE_MODE_CACHED)
+ vm_map_flags |= DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED;
+
+ section->mem = panthor_kernel_bo_create(ptdev, panthor_fw_vm(ptdev),
+ section_size,
+ DRM_PANTHOR_BO_NO_MMAP,
+ vm_map_flags, va);
+ if (IS_ERR(section->mem))
+ return PTR_ERR(section->mem);
+
+ if (drm_WARN_ON(&ptdev->base, section->mem->va_node.start != hdr.va.start))
+ return -EINVAL;
+
+ if (section->flags & CSF_FW_BINARY_IFACE_ENTRY_RD_SHARED) {
+ ret = panthor_kernel_bo_vmap(section->mem);
+ if (ret)
+ return ret;
+ }
+
+ panthor_fw_init_section_mem(ptdev, section);
+
+ bo = to_panthor_bo(section->mem->obj);
+ sgt = drm_gem_shmem_get_pages_sgt(&bo->base);
+ if (IS_ERR(sgt))
+ return PTR_ERR(sgt);
+
+ dma_sync_sgtable_for_device(ptdev->base.dev, sgt, DMA_TO_DEVICE);
+ }
+
+ if (hdr.va.start == CSF_MCU_SHARED_REGION_START)
+ ptdev->fw->shared_section = section;
+
+ return 0;
+}
+
+static void
+panthor_reload_fw_sections(struct panthor_device *ptdev, bool full_reload)
+{
+ struct panthor_fw_section *section;
+
+ list_for_each_entry(section, &ptdev->fw->sections, node) {
+ struct sg_table *sgt;
+
+ if (!full_reload && !(section->flags & CSF_FW_BINARY_IFACE_ENTRY_RD_WR))
+ continue;
+
+ panthor_fw_init_section_mem(ptdev, section);
+ sgt = drm_gem_shmem_get_pages_sgt(&to_panthor_bo(section->mem->obj)->base);
+ if (!drm_WARN_ON(&ptdev->base, IS_ERR_OR_NULL(sgt)))
+ dma_sync_sgtable_for_device(ptdev->base.dev, sgt, DMA_TO_DEVICE);
+ }
+}
+
+static int panthor_fw_load_entry(struct panthor_device *ptdev,
+ const struct firmware *fw,
+ struct panthor_fw_binary_iter *iter)
+{
+ struct panthor_fw_binary_iter eiter;
+ u32 ehdr;
+ int ret;
+
+ ret = panthor_fw_binary_iter_read(ptdev, iter, &ehdr, sizeof(ehdr));
+ if (ret)
+ return ret;
+
+ if ((iter->offset % sizeof(u32)) ||
+ (CSF_FW_BINARY_ENTRY_SIZE(ehdr) % sizeof(u32))) {
+ drm_err(&ptdev->base, "Firmware entry isn't 32 bit aligned, offset=0x%x size=0x%x\n",
+ (u32)(iter->offset - sizeof(u32)), CSF_FW_BINARY_ENTRY_SIZE(ehdr));
+ return -EINVAL;
+ }
+
+ if (panthor_fw_binary_sub_iter_init(ptdev, iter, &eiter,
+ CSF_FW_BINARY_ENTRY_SIZE(ehdr) - sizeof(ehdr)))
+ return -EINVAL;
+
+ switch (CSF_FW_BINARY_ENTRY_TYPE(ehdr)) {
+ case CSF_FW_BINARY_ENTRY_TYPE_IFACE:
+ return panthor_fw_load_section_entry(ptdev, fw, &eiter, ehdr);
+
+ /* FIXME: handle those entry types? */
+ case CSF_FW_BINARY_ENTRY_TYPE_CONFIG:
+ case CSF_FW_BINARY_ENTRY_TYPE_FUTF_TEST:
+ case CSF_FW_BINARY_ENTRY_TYPE_TRACE_BUFFER:
+ case CSF_FW_BINARY_ENTRY_TYPE_TIMELINE_METADATA:
+ return 0;
+ default:
+ break;
+ }
+
+ if (ehdr & CSF_FW_BINARY_ENTRY_OPTIONAL)
+ return 0;
+
+ drm_err(&ptdev->base,
+ "Unsupported non-optional entry type %u in firmware\n",
+ CSF_FW_BINARY_ENTRY_TYPE(ehdr));
+ return -EINVAL;
+}
+
+static int panthor_fw_load(struct panthor_device *ptdev)
+{
+ const struct firmware *fw = NULL;
+ struct panthor_fw_binary_iter iter = {};
+ struct panthor_fw_binary_hdr hdr;
+ char fw_path[128];
+ int ret;
+
+ snprintf(fw_path, sizeof(fw_path), "arm/mali/arch%d.%d/%s",
+ (u32)GPU_ARCH_MAJOR(ptdev->gpu_info.gpu_id),
+ (u32)GPU_ARCH_MINOR(ptdev->gpu_info.gpu_id),
+ CSF_FW_NAME);
+
+ ret = request_firmware(&fw, fw_path, ptdev->base.dev);
+ if (ret) {
+ drm_err(&ptdev->base, "Failed to load firmware image '%s'\n",
+ CSF_FW_NAME);
+ return ret;
+ }
+
+ iter.data = fw->data;
+ iter.size = fw->size;
+ ret = panthor_fw_binary_iter_read(ptdev, &iter, &hdr, sizeof(hdr));
+ if (ret)
+ goto out;
+
+ if (hdr.magic != CSF_FW_BINARY_HEADER_MAGIC) {
+ ret = -EINVAL;
+ drm_err(&ptdev->base, "Invalid firmware magic\n");
+ goto out;
+ }
+
+ if (hdr.major != CSF_FW_BINARY_HEADER_MAJOR_MAX) {
+ ret = -EINVAL;
+ drm_err(&ptdev->base, "Unsupported firmware binary header version %d.%d (expected %d.x)\n",
+ hdr.major, hdr.minor, CSF_FW_BINARY_HEADER_MAJOR_MAX);
+ goto out;
+ }
+
+ if (hdr.size > iter.size) {
+ drm_err(&ptdev->base, "Firmware image is truncated\n");
+ goto out;
+ }
+
+ iter.size = hdr.size;
+
+ while (iter.offset < hdr.size) {
+ ret = panthor_fw_load_entry(ptdev, fw, &iter);
+ if (ret)
+ goto out;
+ }
+
+ if (!ptdev->fw->shared_section) {
+ drm_err(&ptdev->base, "Shared interface region not found\n");
+ ret = -EINVAL;
+ goto out;
+ }
+
+out:
+ release_firmware(fw);
+ return ret;
+}
+
+/**
+ * iface_fw_to_cpu_addr() - Turn an MCU address into a CPU address
+ * @ptdev: Device.
+ * @mcu_va: MCU address.
+ *
+ * Return: NULL if the address is not part of the shared section, non-NULL otherwise.
+ */
+static void *iface_fw_to_cpu_addr(struct panthor_device *ptdev, u32 mcu_va)
+{
+ u64 shared_mem_start = panthor_kernel_bo_gpuva(ptdev->fw->shared_section->mem);
+ u64 shared_mem_end = shared_mem_start +
+ panthor_kernel_bo_size(ptdev->fw->shared_section->mem);
+ if (mcu_va < shared_mem_start || mcu_va >= shared_mem_end)
+ return NULL;
+
+ return ptdev->fw->shared_section->mem->kmap + (mcu_va - shared_mem_start);
+}
+
+static int panthor_init_cs_iface(struct panthor_device *ptdev,
+ unsigned int csg_idx, unsigned int cs_idx)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+ struct panthor_fw_csg_iface *csg_iface = panthor_fw_get_csg_iface(ptdev, csg_idx);
+ struct panthor_fw_cs_iface *cs_iface = &ptdev->fw->iface.streams[csg_idx][cs_idx];
+ u64 shared_section_sz = panthor_kernel_bo_size(ptdev->fw->shared_section->mem);
+ u32 iface_offset = CSF_GROUP_CONTROL_OFFSET +
+ (csg_idx * glb_iface->control->group_stride) +
+ CSF_STREAM_CONTROL_OFFSET +
+ (cs_idx * csg_iface->control->stream_stride);
+ struct panthor_fw_cs_iface *first_cs_iface =
+ panthor_fw_get_cs_iface(ptdev, 0, 0);
+
+ if (iface_offset + sizeof(*cs_iface) >= shared_section_sz)
+ return -EINVAL;
+
+ spin_lock_init(&cs_iface->lock);
+ cs_iface->control = ptdev->fw->shared_section->mem->kmap + iface_offset;
+ cs_iface->input = iface_fw_to_cpu_addr(ptdev, cs_iface->control->input_va);
+ cs_iface->output = iface_fw_to_cpu_addr(ptdev, cs_iface->control->output_va);
+
+ if (!cs_iface->input || !cs_iface->output) {
+ drm_err(&ptdev->base, "Invalid stream control interface input/output VA");
+ return -EINVAL;
+ }
+
+ if (cs_iface != first_cs_iface) {
+ if (cs_iface->control->features != first_cs_iface->control->features) {
+ drm_err(&ptdev->base, "Expecting identical CS slots");
+ return -EINVAL;
+ }
+ } else {
+ u32 reg_count = CS_FEATURES_WORK_REGS(cs_iface->control->features);
+
+ ptdev->csif_info.cs_reg_count = reg_count;
+ ptdev->csif_info.unpreserved_cs_reg_count = CSF_UNPRESERVED_REG_COUNT;
+ }
+
+ return 0;
+}
+
+static bool compare_csg(const struct panthor_fw_csg_control_iface *a,
+ const struct panthor_fw_csg_control_iface *b)
+{
+ if (a->features != b->features)
+ return false;
+ if (a->suspend_size != b->suspend_size)
+ return false;
+ if (a->protm_suspend_size != b->protm_suspend_size)
+ return false;
+ if (a->stream_num != b->stream_num)
+ return false;
+ return true;
+}
+
+static int panthor_init_csg_iface(struct panthor_device *ptdev,
+ unsigned int csg_idx)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+ struct panthor_fw_csg_iface *csg_iface = &ptdev->fw->iface.groups[csg_idx];
+ u64 shared_section_sz = panthor_kernel_bo_size(ptdev->fw->shared_section->mem);
+ u32 iface_offset = CSF_GROUP_CONTROL_OFFSET + (csg_idx * glb_iface->control->group_stride);
+ unsigned int i;
+
+ if (iface_offset + sizeof(*csg_iface) >= shared_section_sz)
+ return -EINVAL;
+
+ spin_lock_init(&csg_iface->lock);
+ csg_iface->control = ptdev->fw->shared_section->mem->kmap + iface_offset;
+ csg_iface->input = iface_fw_to_cpu_addr(ptdev, csg_iface->control->input_va);
+ csg_iface->output = iface_fw_to_cpu_addr(ptdev, csg_iface->control->output_va);
+
+ if (csg_iface->control->stream_num < MIN_CS_PER_CSG ||
+ csg_iface->control->stream_num > MAX_CS_PER_CSG)
+ return -EINVAL;
+
+ if (!csg_iface->input || !csg_iface->output) {
+ drm_err(&ptdev->base, "Invalid group control interface input/output VA");
+ return -EINVAL;
+ }
+
+ if (csg_idx > 0) {
+ struct panthor_fw_csg_iface *first_csg_iface =
+ panthor_fw_get_csg_iface(ptdev, 0);
+
+ if (!compare_csg(first_csg_iface->control, csg_iface->control)) {
+ drm_err(&ptdev->base, "Expecting identical CSG slots");
+ return -EINVAL;
+ }
+ }
+
+ for (i = 0; i < csg_iface->control->stream_num; i++) {
+ int ret = panthor_init_cs_iface(ptdev, csg_idx, i);
+
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+static u32 panthor_get_instr_features(struct panthor_device *ptdev)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+
+ if (glb_iface->control->version < CSF_IFACE_VERSION(1, 1, 0))
+ return 0;
+
+ return glb_iface->control->instr_features;
+}
+
+static int panthor_fw_init_ifaces(struct panthor_device *ptdev)
+{
+ struct panthor_fw_global_iface *glb_iface = &ptdev->fw->iface.global;
+ unsigned int i;
+
+ if (!ptdev->fw->shared_section->mem->kmap)
+ return -EINVAL;
+
+ spin_lock_init(&glb_iface->lock);
+ glb_iface->control = ptdev->fw->shared_section->mem->kmap;
+
+ if (!glb_iface->control->version) {
+ drm_err(&ptdev->base, "Firmware version is 0. Firmware may have failed to boot");
+ return -EINVAL;
+ }
+
+ glb_iface->input = iface_fw_to_cpu_addr(ptdev, glb_iface->control->input_va);
+ glb_iface->output = iface_fw_to_cpu_addr(ptdev, glb_iface->control->output_va);
+ if (!glb_iface->input || !glb_iface->output) {
+ drm_err(&ptdev->base, "Invalid global control interface input/output VA");
+ return -EINVAL;
+ }
+
+ if (glb_iface->control->group_num > MAX_CSGS ||
+ glb_iface->control->group_num < MIN_CSGS) {
+ drm_err(&ptdev->base, "Invalid number of control groups");
+ return -EINVAL;
+ }
+
+ for (i = 0; i < glb_iface->control->group_num; i++) {
+ int ret = panthor_init_csg_iface(ptdev, i);
+
+ if (ret)
+ return ret;
+ }
+
+ drm_info(&ptdev->base, "CSF FW v%d.%d.%d, Features %#x Instrumentation features %#x",
+ CSF_IFACE_VERSION_MAJOR(glb_iface->control->version),
+ CSF_IFACE_VERSION_MINOR(glb_iface->control->version),
+ CSF_IFACE_VERSION_PATCH(glb_iface->control->version),
+ glb_iface->control->features,
+ panthor_get_instr_features(ptdev));
+ return 0;
+}
+
+static void panthor_fw_init_global_iface(struct panthor_device *ptdev)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+
+ /* Enable all cores. */
+ glb_iface->input->core_en_mask = ptdev->gpu_info.shader_present;
+
+ /* Setup timers. */
+ glb_iface->input->poweroff_timer = panthor_fw_conv_timeout(ptdev, PWROFF_HYSTERESIS_US);
+ glb_iface->input->progress_timer = PROGRESS_TIMEOUT_CYCLES >> PROGRESS_TIMEOUT_SCALE_SHIFT;
+ glb_iface->input->idle_timer = panthor_fw_conv_timeout(ptdev, IDLE_HYSTERESIS_US);
+
+ /* Enable interrupts we care about. */
+ glb_iface->input->ack_irq_mask = GLB_CFG_ALLOC_EN |
+ GLB_PING |
+ GLB_CFG_PROGRESS_TIMER |
+ GLB_CFG_POWEROFF_TIMER |
+ GLB_IDLE_EN |
+ GLB_IDLE;
+
+ panthor_fw_update_reqs(glb_iface, req, GLB_IDLE_EN, GLB_IDLE_EN);
+ panthor_fw_toggle_reqs(glb_iface, req, ack,
+ GLB_CFG_ALLOC_EN |
+ GLB_CFG_POWEROFF_TIMER |
+ GLB_CFG_PROGRESS_TIMER);
+
+ gpu_write(ptdev, CSF_DOORBELL(CSF_GLB_DOORBELL_ID), 1);
+
+ /* Kick the watchdog. */
+ mod_delayed_work(ptdev->reset.wq, &ptdev->fw->watchdog.ping_work,
+ msecs_to_jiffies(PING_INTERVAL_MS));
+}
+
+static void panthor_job_irq_handler(struct panthor_device *ptdev, u32 status)
+{
+ if (!ptdev->fw->booted && (status & JOB_INT_GLOBAL_IF))
+ ptdev->fw->booted = true;
+
+ wake_up_all(&ptdev->fw->req_waitqueue);
+
+ /* If the FW is not booted, don't process IRQs, just flag the FW as booted. */
+ if (!ptdev->fw->booted)
+ return;
+
+ panthor_sched_report_fw_events(ptdev, status);
+}
+PANTHOR_IRQ_HANDLER(job, JOB, panthor_job_irq_handler);
+
+static int panthor_fw_start(struct panthor_device *ptdev)
+{
+ bool timedout = false;
+
+ ptdev->fw->booted = false;
+ panthor_job_irq_resume(&ptdev->fw->irq, ~0);
+ gpu_write(ptdev, MCU_CONTROL, MCU_CONTROL_AUTO);
+
+ if (!wait_event_timeout(ptdev->fw->req_waitqueue,
+ ptdev->fw->booted,
+ msecs_to_jiffies(1000))) {
+ if (!ptdev->fw->booted &&
+ !(gpu_read(ptdev, JOB_INT_STAT) & JOB_INT_GLOBAL_IF))
+ timedout = true;
+ }
+
+ if (timedout) {
+ drm_err(&ptdev->base, "Failed to boot MCU");
+ return -ETIMEDOUT;
+ }
+
+ return 0;
+}
+
+static void panthor_fw_stop(struct panthor_device *ptdev)
+{
+ u32 status;
+
+ gpu_write(ptdev, MCU_CONTROL, MCU_CONTROL_DISABLE);
+ if (readl_poll_timeout(ptdev->iomem + MCU_STATUS, status,
+ status == MCU_STATUS_DISABLED, 10, 100000))
+ drm_err(&ptdev->base, "Failed to stop MCU");
+}
+
+/**
+ * panthor_fw_pre_reset() - Call before a reset.
+ * @ptdev: Device.
+ * @on_hang: true if the reset was triggered on a GPU hang.
+ *
+ * If the reset is not triggered on a hang, we try to gracefully halt the
+ * MCU, so we can do a fast-reset when panthor_fw_post_reset() is called.
+ */
+void panthor_fw_pre_reset(struct panthor_device *ptdev, bool on_hang)
+{
+ /* Make sure we won't be woken up by a ping. */
+ cancel_delayed_work_sync(&ptdev->fw->watchdog.ping_work);
+
+ ptdev->fw->fast_reset = false;
+
+ if (!on_hang) {
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+ u32 status;
+
+ panthor_fw_update_reqs(glb_iface, req, GLB_HALT, GLB_HALT);
+ gpu_write(ptdev, CSF_DOORBELL(CSF_GLB_DOORBELL_ID), 1);
+ if (!readl_poll_timeout(ptdev->iomem + MCU_STATUS, status,
+ status == MCU_STATUS_HALT, 10, 100000) &&
+ glb_iface->output->halt_status == PANTHOR_FW_HALT_OK) {
+ ptdev->fw->fast_reset = true;
+ } else {
+ drm_warn(&ptdev->base, "Failed to cleanly suspend MCU");
+ }
+
+ /* The FW detects 0 -> 1 transitions. Make sure we reset
+ * the HALT bit before the FW is rebooted.
+ */
+ panthor_fw_update_reqs(glb_iface, req, 0, GLB_HALT);
+ }
+
+ panthor_job_irq_suspend(&ptdev->fw->irq);
+}
+
+/**
+ * panthor_fw_post_reset() - Call after a reset.
+ * @ptdev: Device.
+ *
+ * Start the FW. If this is not a fast reset, all FW sections are reloaded to
+ * make sure we can recover from a memory corruption.
+ */
+int panthor_fw_post_reset(struct panthor_device *ptdev)
+{
+ int ret;
+
+ /* Make the MCU VM active. */
+ ret = panthor_vm_active(ptdev->fw->vm);
+ if (ret)
+ return ret;
+
+ /* Reload all sections, including RO ones. We're not supposed
+ * to end up here anyway, let's just assume the overhead of
+ * reloading everything is acceptable.
+ */
+ if (!ptdev->fw->fast_reset)
+ panthor_reload_fw_sections(ptdev, true);
+
+ ret = panthor_fw_start(ptdev);
+ if (ret)
+ return ret;
+
+ /* We must re-initialize the global interface even on fast-reset. */
+ panthor_fw_init_global_iface(ptdev);
+ return 0;
+}
+
+/**
+ * panthor_fw_unplug() - Called when the device is unplugged.
+ * @ptdev: Device.
+ *
+ * This function must make sure all pending operations are flushed before
+ * will release device resources, thus preventing any interaction with
+ * the HW.
+ *
+ * If there is still FW-related work running after this function returns,
+ * they must use drm_dev_{enter,exit}() and skip any HW access when
+ * drm_dev_enter() returns false.
+ */
+void panthor_fw_unplug(struct panthor_device *ptdev)
+{
+ struct panthor_fw_section *section;
+
+ cancel_delayed_work_sync(&ptdev->fw->watchdog.ping_work);
+
+ /* Make sure the IRQ handler can be called after that point. */
+ if (ptdev->fw->irq.irq)
+ panthor_job_irq_suspend(&ptdev->fw->irq);
+
+ panthor_fw_stop(ptdev);
+
+ if (ptdev->fw->vm)
+ panthor_vm_idle(ptdev->fw->vm);
+
+ list_for_each_entry(section, &ptdev->fw->sections, node)
+ panthor_kernel_bo_destroy(panthor_fw_vm(ptdev), section->mem);
+
+ panthor_vm_put(ptdev->fw->vm);
+
+ panthor_gpu_power_off(ptdev, L2, ptdev->gpu_info.l2_present, 20000);
+}
+
+/**
+ * panthor_fw_wait_acks() - Wait for requests to be acknowledged by the FW.
+ * @req_ptr: Pointer to the req register.
+ * @ack_ptr: Pointer to the ack register.
+ * @wq: Wait queue to use for the sleeping wait.
+ * @req_mask: Mask of requests to wait for.
+ * @acked: Pointer to field that's updated with the acked requests.
+ * If the function returns 0, *acked == req_mask.
+ * @timeout_ms: Timeout expressed in milliseconds.
+ *
+ * Return: 0 on success, -ETIMEDOUT otherwise.
+ */
+static int panthor_fw_wait_acks(const u32 *req_ptr, const u32 *ack_ptr,
+ wait_queue_head_t *wq,
+ u32 req_mask, u32 *acked,
+ u32 timeout_ms)
+{
+ u32 ack, req = READ_ONCE(*req_ptr) & req_mask;
+ int ret;
+
+ /* Busy wait for a few µsecs before falling back to a sleeping wait. */
+ *acked = req_mask;
+ ret = read_poll_timeout_atomic(READ_ONCE, ack,
+ (ack & req_mask) == req,
+ 0, 10, 0,
+ *ack_ptr);
+ if (!ret)
+ return 0;
+
+ if (wait_event_timeout(*wq, (READ_ONCE(*ack_ptr) & req_mask) == req,
+ msecs_to_jiffies(timeout_ms)))
+ return 0;
+
+ /* Check one last time, in case we were not woken up for some reason. */
+ ack = READ_ONCE(*ack_ptr);
+ if ((ack & req_mask) == req)
+ return 0;
+
+ *acked = ~(req ^ ack) & req_mask;
+ return -ETIMEDOUT;
+}
+
+/**
+ * panthor_fw_glb_wait_acks() - Wait for global requests to be acknowledged.
+ * @ptdev: Device.
+ * @req_mask: Mask of requests to wait for.
+ * @acked: Pointer to field that's updated with the acked requests.
+ * If the function returns 0, *acked == req_mask.
+ * @timeout_ms: Timeout expressed in milliseconds.
+ *
+ * Return: 0 on success, -ETIMEDOUT otherwise.
+ */
+int panthor_fw_glb_wait_acks(struct panthor_device *ptdev,
+ u32 req_mask, u32 *acked,
+ u32 timeout_ms)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+
+ /* GLB_HALT doesn't get acked through the FW interface. */
+ if (drm_WARN_ON(&ptdev->base, req_mask & (~GLB_REQ_MASK | GLB_HALT)))
+ return -EINVAL;
+
+ return panthor_fw_wait_acks(&glb_iface->input->req,
+ &glb_iface->output->ack,
+ &ptdev->fw->req_waitqueue,
+ req_mask, acked, timeout_ms);
+}
+
+/**
+ * panthor_fw_csg_wait_acks() - Wait for command stream group requests to be acknowledged.
+ * @ptdev: Device.
+ * @csg_slot: CSG slot ID.
+ * @req_mask: Mask of requests to wait for.
+ * @acked: Pointer to field that's updated with the acked requests.
+ * If the function returns 0, *acked == req_mask.
+ * @timeout_ms: Timeout expressed in milliseconds.
+ *
+ * Return: 0 on success, -ETIMEDOUT otherwise.
+ */
+int panthor_fw_csg_wait_acks(struct panthor_device *ptdev, u32 csg_slot,
+ u32 req_mask, u32 *acked, u32 timeout_ms)
+{
+ struct panthor_fw_csg_iface *csg_iface = panthor_fw_get_csg_iface(ptdev, csg_slot);
+ int ret;
+
+ if (drm_WARN_ON(&ptdev->base, req_mask & ~CSG_REQ_MASK))
+ return -EINVAL;
+
+ ret = panthor_fw_wait_acks(&csg_iface->input->req,
+ &csg_iface->output->ack,
+ &ptdev->fw->req_waitqueue,
+ req_mask, acked, timeout_ms);
+
+ /*
+ * Check that all bits in the state field were updated, if any mismatch
+ * then clear all bits in the state field. This allows code to do
+ * (acked & CSG_STATE_MASK) and get the right value.
+ */
+
+ if ((*acked & CSG_STATE_MASK) != CSG_STATE_MASK)
+ *acked &= ~CSG_STATE_MASK;
+
+ return ret;
+}
+
+/**
+ * panthor_fw_ring_csg_doorbells() - Ring command stream group doorbells.
+ * @ptdev: Device.
+ * @csg_mask: Bitmask encoding the command stream group doorbells to ring.
+ *
+ * This function is toggling bits in the doorbell_req and ringing the
+ * global doorbell. It doesn't require a user doorbell to be attached to
+ * the group.
+ */
+void panthor_fw_ring_csg_doorbells(struct panthor_device *ptdev, u32 csg_mask)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+
+ panthor_fw_toggle_reqs(glb_iface, doorbell_req, doorbell_ack, csg_mask);
+ gpu_write(ptdev, CSF_DOORBELL(CSF_GLB_DOORBELL_ID), 1);
+}
+
+static void panthor_fw_ping_work(struct work_struct *work)
+{
+ struct panthor_fw *fw = container_of(work, struct panthor_fw, watchdog.ping_work.work);
+ struct panthor_device *ptdev = fw->irq.ptdev;
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+ u32 acked;
+ int ret;
+
+ if (panthor_device_reset_is_pending(ptdev))
+ return;
+
+ panthor_fw_toggle_reqs(glb_iface, req, ack, GLB_PING);
+ gpu_write(ptdev, CSF_DOORBELL(CSF_GLB_DOORBELL_ID), 1);
+
+ ret = panthor_fw_glb_wait_acks(ptdev, GLB_PING, &acked, 100);
+ if (ret) {
+ panthor_device_schedule_reset(ptdev);
+ drm_err(&ptdev->base, "FW ping timeout, scheduling a reset");
+ } else {
+ mod_delayed_work(ptdev->reset.wq, &fw->watchdog.ping_work,
+ msecs_to_jiffies(PING_INTERVAL_MS));
+ }
+}
+
+/**
+ * panthor_fw_init() - Initialize FW related data.
+ * @ptdev: Device.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+int panthor_fw_init(struct panthor_device *ptdev)
+{
+ struct panthor_fw *fw;
+ int ret, irq;
+
+ fw = drmm_kzalloc(&ptdev->base, sizeof(*fw), GFP_KERNEL);
+ if (!fw)
+ return -ENOMEM;
+
+ ptdev->fw = fw;
+ init_waitqueue_head(&fw->req_waitqueue);
+ INIT_LIST_HEAD(&fw->sections);
+ INIT_DELAYED_WORK(&fw->watchdog.ping_work, panthor_fw_ping_work);
+
+ irq = platform_get_irq_byname(to_platform_device(ptdev->base.dev), "job");
+ if (irq <= 0)
+ return -ENODEV;
+
+ ret = panthor_request_job_irq(ptdev, &fw->irq, irq, 0);
+ if (ret) {
+ drm_err(&ptdev->base, "failed to request job irq");
+ return ret;
+ }
+
+ ret = panthor_gpu_l2_power_on(ptdev);
+ if (ret)
+ return ret;
+
+ fw->vm = panthor_vm_create(ptdev, true,
+ 0, SZ_4G,
+ CSF_MCU_SHARED_REGION_START,
+ CSF_MCU_SHARED_REGION_SIZE);
+ if (IS_ERR(fw->vm)) {
+ ret = PTR_ERR(fw->vm);
+ fw->vm = NULL;
+ goto err_unplug_fw;
+ }
+
+ ret = panthor_fw_load(ptdev);
+ if (ret)
+ goto err_unplug_fw;
+
+ ret = panthor_vm_active(fw->vm);
+ if (ret)
+ goto err_unplug_fw;
+
+ ret = panthor_fw_start(ptdev);
+ if (ret)
+ goto err_unplug_fw;
+
+ ret = panthor_fw_init_ifaces(ptdev);
+ if (ret)
+ goto err_unplug_fw;
+
+ panthor_fw_init_global_iface(ptdev);
+ return 0;
+
+err_unplug_fw:
+ panthor_fw_unplug(ptdev);
+ return ret;
+}
+
+MODULE_FIRMWARE("arm/mali/arch10.8/mali_csffw.bin");
diff --git a/drivers/gpu/drm/panthor/panthor_fw.h b/drivers/gpu/drm/panthor/panthor_fw.h
new file mode 100644
index 000000000000..1126b1ea199a
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_fw.h
@@ -0,0 +1,504 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2023 Collabora ltd. */
+
+#ifndef __PANTHOR_MCU_H__
+#define __PANTHOR_MCU_H__
+
+#include <linux/types.h>
+
+#include "panthor_device.h"
+
+struct panthor_kernel_bo;
+
+#define MAX_CSGS 31
+#define MAX_CS_PER_CSG 32
+
+struct panthor_fw_ringbuf_input_iface {
+ u64 insert;
+ u64 extract;
+};
+
+struct panthor_fw_ringbuf_output_iface {
+ u64 extract;
+ u32 active;
+};
+
+struct panthor_fw_cs_control_iface {
+#define CS_FEATURES_WORK_REGS(x) (((x) & GENMASK(7, 0)) + 1)
+#define CS_FEATURES_SCOREBOARDS(x) (((x) & GENMASK(15, 8)) >> 8)
+#define CS_FEATURES_COMPUTE BIT(16)
+#define CS_FEATURES_FRAGMENT BIT(17)
+#define CS_FEATURES_TILER BIT(18)
+ u32 features;
+ u32 input_va;
+ u32 output_va;
+};
+
+struct panthor_fw_cs_input_iface {
+#define CS_STATE_MASK GENMASK(2, 0)
+#define CS_STATE_STOP 0
+#define CS_STATE_START 1
+#define CS_EXTRACT_EVENT BIT(4)
+#define CS_IDLE_SYNC_WAIT BIT(8)
+#define CS_IDLE_PROTM_PENDING BIT(9)
+#define CS_IDLE_EMPTY BIT(10)
+#define CS_IDLE_RESOURCE_REQ BIT(11)
+#define CS_TILER_OOM BIT(26)
+#define CS_PROTM_PENDING BIT(27)
+#define CS_FATAL BIT(30)
+#define CS_FAULT BIT(31)
+#define CS_REQ_MASK (CS_STATE_MASK | \
+ CS_EXTRACT_EVENT | \
+ CS_IDLE_SYNC_WAIT | \
+ CS_IDLE_PROTM_PENDING | \
+ CS_IDLE_EMPTY | \
+ CS_IDLE_RESOURCE_REQ)
+#define CS_EVT_MASK (CS_TILER_OOM | \
+ CS_PROTM_PENDING | \
+ CS_FATAL | \
+ CS_FAULT)
+ u32 req;
+
+#define CS_CONFIG_PRIORITY(x) ((x) & GENMASK(3, 0))
+#define CS_CONFIG_DOORBELL(x) (((x) << 8) & GENMASK(15, 8))
+ u32 config;
+ u32 reserved1;
+ u32 ack_irq_mask;
+ u64 ringbuf_base;
+ u32 ringbuf_size;
+ u32 reserved2;
+ u64 heap_start;
+ u64 heap_end;
+ u64 ringbuf_input;
+ u64 ringbuf_output;
+ u32 instr_config;
+ u32 instrbuf_size;
+ u64 instrbuf_base;
+ u64 instrbuf_offset_ptr;
+};
+
+struct panthor_fw_cs_output_iface {
+ u32 ack;
+ u32 reserved1[15];
+ u64 status_cmd_ptr;
+
+#define CS_STATUS_WAIT_SB_MASK GENMASK(15, 0)
+#define CS_STATUS_WAIT_SB_SRC_MASK GENMASK(19, 16)
+#define CS_STATUS_WAIT_SB_SRC_NONE (0 << 16)
+#define CS_STATUS_WAIT_SB_SRC_WAIT (8 << 16)
+#define CS_STATUS_WAIT_SYNC_COND_LE (0 << 24)
+#define CS_STATUS_WAIT_SYNC_COND_GT (1 << 24)
+#define CS_STATUS_WAIT_SYNC_COND_MASK GENMASK(27, 24)
+#define CS_STATUS_WAIT_PROGRESS BIT(28)
+#define CS_STATUS_WAIT_PROTM BIT(29)
+#define CS_STATUS_WAIT_SYNC_64B BIT(30)
+#define CS_STATUS_WAIT_SYNC BIT(31)
+ u32 status_wait;
+ u32 status_req_resource;
+ u64 status_wait_sync_ptr;
+ u32 status_wait_sync_value;
+ u32 status_scoreboards;
+
+#define CS_STATUS_BLOCKED_REASON_UNBLOCKED 0
+#define CS_STATUS_BLOCKED_REASON_SB_WAIT 1
+#define CS_STATUS_BLOCKED_REASON_PROGRESS_WAIT 2
+#define CS_STATUS_BLOCKED_REASON_SYNC_WAIT 3
+#define CS_STATUS_BLOCKED_REASON_DEFERRED 5
+#define CS_STATUS_BLOCKED_REASON_RES 6
+#define CS_STATUS_BLOCKED_REASON_FLUSH 7
+#define CS_STATUS_BLOCKED_REASON_MASK GENMASK(3, 0)
+ u32 status_blocked_reason;
+ u32 status_wait_sync_value_hi;
+ u32 reserved2[6];
+
+#define CS_EXCEPTION_TYPE(x) ((x) & GENMASK(7, 0))
+#define CS_EXCEPTION_DATA(x) (((x) >> 8) & GENMASK(23, 0))
+ u32 fault;
+ u32 fatal;
+ u64 fault_info;
+ u64 fatal_info;
+ u32 reserved3[10];
+ u32 heap_vt_start;
+ u32 heap_vt_end;
+ u32 reserved4;
+ u32 heap_frag_end;
+ u64 heap_address;
+};
+
+struct panthor_fw_csg_control_iface {
+ u32 features;
+ u32 input_va;
+ u32 output_va;
+ u32 suspend_size;
+ u32 protm_suspend_size;
+ u32 stream_num;
+ u32 stream_stride;
+};
+
+struct panthor_fw_csg_input_iface {
+#define CSG_STATE_MASK GENMASK(2, 0)
+#define CSG_STATE_TERMINATE 0
+#define CSG_STATE_START 1
+#define CSG_STATE_SUSPEND 2
+#define CSG_STATE_RESUME 3
+#define CSG_ENDPOINT_CONFIG BIT(4)
+#define CSG_STATUS_UPDATE BIT(5)
+#define CSG_SYNC_UPDATE BIT(28)
+#define CSG_IDLE BIT(29)
+#define CSG_DOORBELL BIT(30)
+#define CSG_PROGRESS_TIMER_EVENT BIT(31)
+#define CSG_REQ_MASK (CSG_STATE_MASK | \
+ CSG_ENDPOINT_CONFIG | \
+ CSG_STATUS_UPDATE)
+#define CSG_EVT_MASK (CSG_SYNC_UPDATE | \
+ CSG_IDLE | \
+ CSG_PROGRESS_TIMER_EVENT)
+ u32 req;
+ u32 ack_irq_mask;
+
+ u32 doorbell_req;
+ u32 cs_irq_ack;
+ u32 reserved1[4];
+ u64 allow_compute;
+ u64 allow_fragment;
+ u32 allow_other;
+
+#define CSG_EP_REQ_COMPUTE(x) ((x) & GENMASK(7, 0))
+#define CSG_EP_REQ_FRAGMENT(x) (((x) << 8) & GENMASK(15, 8))
+#define CSG_EP_REQ_TILER(x) (((x) << 16) & GENMASK(19, 16))
+#define CSG_EP_REQ_EXCL_COMPUTE BIT(20)
+#define CSG_EP_REQ_EXCL_FRAGMENT BIT(21)
+#define CSG_EP_REQ_PRIORITY(x) (((x) << 28) & GENMASK(31, 28))
+#define CSG_EP_REQ_PRIORITY_MASK GENMASK(31, 28)
+ u32 endpoint_req;
+ u32 reserved2[2];
+ u64 suspend_buf;
+ u64 protm_suspend_buf;
+ u32 config;
+ u32 iter_trace_config;
+};
+
+struct panthor_fw_csg_output_iface {
+ u32 ack;
+ u32 reserved1;
+ u32 doorbell_ack;
+ u32 cs_irq_req;
+ u32 status_endpoint_current;
+ u32 status_endpoint_req;
+
+#define CSG_STATUS_STATE_IS_IDLE BIT(0)
+ u32 status_state;
+ u32 resource_dep;
+};
+
+struct panthor_fw_global_control_iface {
+ u32 version;
+ u32 features;
+ u32 input_va;
+ u32 output_va;
+ u32 group_num;
+ u32 group_stride;
+ u32 perfcnt_size;
+ u32 instr_features;
+};
+
+struct panthor_fw_global_input_iface {
+#define GLB_HALT BIT(0)
+#define GLB_CFG_PROGRESS_TIMER BIT(1)
+#define GLB_CFG_ALLOC_EN BIT(2)
+#define GLB_CFG_POWEROFF_TIMER BIT(3)
+#define GLB_PROTM_ENTER BIT(4)
+#define GLB_PERFCNT_EN BIT(5)
+#define GLB_PERFCNT_SAMPLER BIT(6)
+#define GLB_COUNTER_EN BIT(7)
+#define GLB_PING BIT(8)
+#define GLB_FWCFG_UPDATE BIT(9)
+#define GLB_IDLE_EN BIT(10)
+#define GLB_SLEEP BIT(12)
+#define GLB_INACTIVE_COMPUTE BIT(20)
+#define GLB_INACTIVE_FRAGMENT BIT(21)
+#define GLB_INACTIVE_TILER BIT(22)
+#define GLB_PROTM_EXIT BIT(23)
+#define GLB_PERFCNT_THRESHOLD BIT(24)
+#define GLB_PERFCNT_OVERFLOW BIT(25)
+#define GLB_IDLE BIT(26)
+#define GLB_DBG_CSF BIT(30)
+#define GLB_DBG_HOST BIT(31)
+#define GLB_REQ_MASK GENMASK(10, 0)
+#define GLB_EVT_MASK GENMASK(26, 20)
+ u32 req;
+ u32 ack_irq_mask;
+ u32 doorbell_req;
+ u32 reserved1;
+ u32 progress_timer;
+
+#define GLB_TIMER_VAL(x) ((x) & GENMASK(30, 0))
+#define GLB_TIMER_SOURCE_GPU_COUNTER BIT(31)
+ u32 poweroff_timer;
+ u64 core_en_mask;
+ u32 reserved2;
+ u32 perfcnt_as;
+ u64 perfcnt_base;
+ u32 perfcnt_extract;
+ u32 reserved3[3];
+ u32 perfcnt_config;
+ u32 perfcnt_csg_select;
+ u32 perfcnt_fw_enable;
+ u32 perfcnt_csg_enable;
+ u32 perfcnt_csf_enable;
+ u32 perfcnt_shader_enable;
+ u32 perfcnt_tiler_enable;
+ u32 perfcnt_mmu_l2_enable;
+ u32 reserved4[8];
+ u32 idle_timer;
+};
+
+enum panthor_fw_halt_status {
+ PANTHOR_FW_HALT_OK = 0,
+ PANTHOR_FW_HALT_ON_PANIC = 0x4e,
+ PANTHOR_FW_HALT_ON_WATCHDOG_EXPIRATION = 0x4f,
+};
+
+struct panthor_fw_global_output_iface {
+ u32 ack;
+ u32 reserved1;
+ u32 doorbell_ack;
+ u32 reserved2;
+ u32 halt_status;
+ u32 perfcnt_status;
+ u32 perfcnt_insert;
+};
+
+/**
+ * struct panthor_fw_cs_iface - Firmware command stream slot interface
+ */
+struct panthor_fw_cs_iface {
+ /**
+ * @lock: Lock protecting access to the panthor_fw_cs_input_iface::req
+ * field.
+ *
+ * Needed so we can update the req field concurrently from the interrupt
+ * handler and the scheduler logic.
+ *
+ * TODO: Ideally we'd want to use a cmpxchg() to update the req, but FW
+ * interface sections are mapped uncached/write-combined right now, and
+ * using cmpxchg() on such mappings leads to SError faults. Revisit when
+ * we have 'SHARED' GPU mappings hooked up.
+ */
+ spinlock_t lock;
+
+ /**
+ * @control: Command stream slot control interface.
+ *
+ * Used to expose command stream slot properties.
+ *
+ * This interface is read-only.
+ */
+ struct panthor_fw_cs_control_iface *control;
+
+ /**
+ * @input: Command stream slot input interface.
+ *
+ * Used for host updates/events.
+ */
+ struct panthor_fw_cs_input_iface *input;
+
+ /**
+ * @output: Command stream slot output interface.
+ *
+ * Used for FW updates/events.
+ *
+ * This interface is read-only.
+ */
+ const struct panthor_fw_cs_output_iface *output;
+};
+
+/**
+ * struct panthor_fw_csg_iface - Firmware command stream group slot interface
+ */
+struct panthor_fw_csg_iface {
+ /**
+ * @lock: Lock protecting access to the panthor_fw_csg_input_iface::req
+ * field.
+ *
+ * Needed so we can update the req field concurrently from the interrupt
+ * handler and the scheduler logic.
+ *
+ * TODO: Ideally we'd want to use a cmpxchg() to update the req, but FW
+ * interface sections are mapped uncached/write-combined right now, and
+ * using cmpxchg() on such mappings leads to SError faults. Revisit when
+ * we have 'SHARED' GPU mappings hooked up.
+ */
+ spinlock_t lock;
+
+ /**
+ * @control: Command stream group slot control interface.
+ *
+ * Used to expose command stream group slot properties.
+ *
+ * This interface is read-only.
+ */
+ const struct panthor_fw_csg_control_iface *control;
+
+ /**
+ * @input: Command stream slot input interface.
+ *
+ * Used for host updates/events.
+ */
+ struct panthor_fw_csg_input_iface *input;
+
+ /**
+ * @output: Command stream group slot output interface.
+ *
+ * Used for FW updates/events.
+ *
+ * This interface is read-only.
+ */
+ const struct panthor_fw_csg_output_iface *output;
+};
+
+/**
+ * struct panthor_fw_global_iface - Firmware global interface
+ */
+struct panthor_fw_global_iface {
+ /**
+ * @lock: Lock protecting access to the panthor_fw_global_input_iface::req
+ * field.
+ *
+ * Needed so we can update the req field concurrently from the interrupt
+ * handler and the scheduler/FW management logic.
+ *
+ * TODO: Ideally we'd want to use a cmpxchg() to update the req, but FW
+ * interface sections are mapped uncached/write-combined right now, and
+ * using cmpxchg() on such mappings leads to SError faults. Revisit when
+ * we have 'SHARED' GPU mappings hooked up.
+ */
+ spinlock_t lock;
+
+ /**
+ * @control: Command stream group slot control interface.
+ *
+ * Used to expose global FW properties.
+ *
+ * This interface is read-only.
+ */
+ const struct panthor_fw_global_control_iface *control;
+
+ /**
+ * @input: Global input interface.
+ *
+ * Used for host updates/events.
+ */
+ struct panthor_fw_global_input_iface *input;
+
+ /**
+ * @output: Global output interface.
+ *
+ * Used for FW updates/events.
+ *
+ * This interface is read-only.
+ */
+ const struct panthor_fw_global_output_iface *output;
+};
+
+/**
+ * panthor_fw_toggle_reqs() - Toggle acknowledge bits to send an event to the FW
+ * @__iface: The interface to operate on.
+ * @__in_reg: Name of the register to update in the input section of the interface.
+ * @__out_reg: Name of the register to take as a reference in the output section of the
+ * interface.
+ * @__mask: Mask to apply to the update.
+ *
+ * The Host -> FW event/message passing was designed to be lockless, with each side of
+ * the channel having its writeable section. Events are signaled as a difference between
+ * the host and FW side in the req/ack registers (when a bit differs, there's an event
+ * pending, when they are the same, nothing needs attention).
+ *
+ * This helper allows one to update the req register based on the current value of the
+ * ack register managed by the FW. Toggling a specific bit will flag an event. In order
+ * for events to be re-evaluated, the interface doorbell needs to be rung.
+ *
+ * Concurrent accesses to the same req register is covered.
+ *
+ * Anything requiring atomic updates to multiple registers requires a dedicated lock.
+ */
+#define panthor_fw_toggle_reqs(__iface, __in_reg, __out_reg, __mask) \
+ do { \
+ u32 __cur_val, __new_val, __out_val; \
+ spin_lock(&(__iface)->lock); \
+ __cur_val = READ_ONCE((__iface)->input->__in_reg); \
+ __out_val = READ_ONCE((__iface)->output->__out_reg); \
+ __new_val = ((__out_val ^ (__mask)) & (__mask)) | (__cur_val & ~(__mask)); \
+ WRITE_ONCE((__iface)->input->__in_reg, __new_val); \
+ spin_unlock(&(__iface)->lock); \
+ } while (0)
+
+/**
+ * panthor_fw_update_reqs() - Update bits to reflect a configuration change
+ * @__iface: The interface to operate on.
+ * @__in_reg: Name of the register to update in the input section of the interface.
+ * @__val: Value to set.
+ * @__mask: Mask to apply to the update.
+ *
+ * Some configuration get passed through req registers that are also used to
+ * send events to the FW. Those req registers being updated from the interrupt
+ * handler, they require special helpers to update the configuration part as well.
+ *
+ * Concurrent accesses to the same req register is covered.
+ *
+ * Anything requiring atomic updates to multiple registers requires a dedicated lock.
+ */
+#define panthor_fw_update_reqs(__iface, __in_reg, __val, __mask) \
+ do { \
+ u32 __cur_val, __new_val; \
+ spin_lock(&(__iface)->lock); \
+ __cur_val = READ_ONCE((__iface)->input->__in_reg); \
+ __new_val = (__cur_val & ~(__mask)) | ((__val) & (__mask)); \
+ WRITE_ONCE((__iface)->input->__in_reg, __new_val); \
+ spin_unlock(&(__iface)->lock); \
+ } while (0)
+
+struct panthor_fw_global_iface *
+panthor_fw_get_glb_iface(struct panthor_device *ptdev);
+
+struct panthor_fw_csg_iface *
+panthor_fw_get_csg_iface(struct panthor_device *ptdev, u32 csg_slot);
+
+struct panthor_fw_cs_iface *
+panthor_fw_get_cs_iface(struct panthor_device *ptdev, u32 csg_slot, u32 cs_slot);
+
+int panthor_fw_csg_wait_acks(struct panthor_device *ptdev, u32 csg_id, u32 req_mask,
+ u32 *acked, u32 timeout_ms);
+
+int panthor_fw_glb_wait_acks(struct panthor_device *ptdev, u32 req_mask, u32 *acked,
+ u32 timeout_ms);
+
+void panthor_fw_ring_csg_doorbells(struct panthor_device *ptdev, u32 csg_slot);
+
+struct panthor_kernel_bo *
+panthor_fw_alloc_queue_iface_mem(struct panthor_device *ptdev,
+ struct panthor_fw_ringbuf_input_iface **input,
+ const struct panthor_fw_ringbuf_output_iface **output,
+ u32 *input_fw_va, u32 *output_fw_va);
+struct panthor_kernel_bo *
+panthor_fw_alloc_suspend_buf_mem(struct panthor_device *ptdev, size_t size);
+
+struct panthor_vm *panthor_fw_vm(struct panthor_device *ptdev);
+
+void panthor_fw_pre_reset(struct panthor_device *ptdev, bool on_hang);
+int panthor_fw_post_reset(struct panthor_device *ptdev);
+
+static inline void panthor_fw_suspend(struct panthor_device *ptdev)
+{
+ panthor_fw_pre_reset(ptdev, false);
+}
+
+static inline int panthor_fw_resume(struct panthor_device *ptdev)
+{
+ return panthor_fw_post_reset(ptdev);
+}
+
+int panthor_fw_init(struct panthor_device *ptdev);
+void panthor_fw_unplug(struct panthor_device *ptdev);
+
+#endif
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* [PATCH v4 09/14] drm/panthor: Add the heap logical block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (7 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 08/14] drm/panthor: Add the FW " Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-02-12 11:40 ` Steven Price
2024-01-22 16:30 ` [PATCH v4 10/14] drm/panthor: Add the scheduler " Boris Brezillon
` (8 subsequent siblings)
17 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
Tiler heap growing requires some kernel driver involvement: when the
tiler runs out of heap memory, it will raise an exception which is
either directly handled by the firmware if some free heap chunks are
available in the heap context, or passed back to the kernel otherwise.
The heap helpers will be used by the scheduler logic to allocate more
heap chunks to a heap context, when such a situation happens.
Heap context creation is explicitly requested by userspace (using
the TILER_HEAP_CREATE ioctl), and the returned context is attached to a
queue through some command stream instruction.
All the kernel does is keep the list of heap chunks allocated to a
context, so they can be freed when TILER_HEAP_DESTROY is called, or
extended when the FW requests a new chunk.
v4:
- Rework locking to allow concurrent calls to panthor_heap_grow()
- Add a helper to return a heap chunk if we couldn't pass it to the
FW because the group was scheduled out
v3:
- Add a FIXME for the heap OOM deadlock
- Use the panthor_kernel_bo abstraction for the heap context and heap
chunks
- Drop the panthor_heap_gpu_ctx struct as it is opaque to the driver
- Ensure that the heap context is aligned to the GPU cache line size
- Minor code tidy ups
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
---
drivers/gpu/drm/panthor/panthor_heap.c | 596 +++++++++++++++++++++++++
drivers/gpu/drm/panthor/panthor_heap.h | 39 ++
2 files changed, 635 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_heap.c
create mode 100644 drivers/gpu/drm/panthor/panthor_heap.h
diff --git a/drivers/gpu/drm/panthor/panthor_heap.c b/drivers/gpu/drm/panthor/panthor_heap.c
new file mode 100644
index 000000000000..fac51c24b3e9
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_heap.c
@@ -0,0 +1,596 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2023 Collabora ltd. */
+
+#include <linux/iosys-map.h>
+#include <linux/rwsem.h>
+
+#include <drm/panthor_drm.h>
+
+#include "panthor_device.h"
+#include "panthor_gem.h"
+#include "panthor_heap.h"
+#include "panthor_mmu.h"
+#include "panthor_regs.h"
+
+/*
+ * The GPU heap context is an opaque structure used by the GPU to track the
+ * heap allocations. The driver should only touch it to initialize it (zero all
+ * fields). Because the CPU and GPU can both access this structure it is
+ * required to be GPU cache line aligned.
+ */
+#define HEAP_CONTEXT_SIZE 32
+
+/**
+ * struct panthor_heap_chunk_header - Heap chunk header
+ */
+struct panthor_heap_chunk_header {
+ /**
+ * @next: Next heap chunk in the list.
+ *
+ * This is a GPU VA.
+ */
+ u64 next;
+
+ /** @unknown: MBZ. */
+ u32 unknown[14];
+};
+
+/**
+ * struct panthor_heap_chunk - Structure used to keep track of allocated heap chunks.
+ */
+struct panthor_heap_chunk {
+ /** @node: Used to insert the heap chunk in panthor_heap::chunks. */
+ struct list_head node;
+
+ /** @bo: Buffer object backing the heap chunk. */
+ struct panthor_kernel_bo *bo;
+};
+
+/**
+ * struct panthor_heap - Structure used to manage tiler heap contexts.
+ */
+struct panthor_heap {
+ /** @chunks: List containing all heap chunks allocated so far. */
+ struct list_head chunks;
+
+ /** @lock: Lock protecting insertion in the chunks list. */
+ struct mutex lock;
+
+ /** @chunk_size: Size of each chunk. */
+ u32 chunk_size;
+
+ /** @max_chunks: Maximum number of chunks. */
+ u32 max_chunks;
+
+ /**
+ * @target_in_flight: Number of in-flight render passes after which
+ * we'd let the FW wait for fragment job to finish instead of allocating new chunks.
+ */
+ u32 target_in_flight;
+
+ /** @chunk_count: Number of heap chunks currently allocated. */
+ u32 chunk_count;
+};
+
+#define MAX_HEAPS_PER_POOL 128
+
+/**
+ * struct panthor_heap_pool - Pool of heap contexts
+ *
+ * The pool is attached to a panthor_file and can't be shared across processes.
+ */
+struct panthor_heap_pool {
+ /** @refcount: Reference count. */
+ struct kref refcount;
+
+ /** @ptdev: Device. */
+ struct panthor_device *ptdev;
+
+ /** @vm: VM this pool is bound to. */
+ struct panthor_vm *vm;
+
+ /** @lock: Lock protecting access to @xa. */
+ struct rw_semaphore lock;
+
+ /** @xa: Array storing panthor_heap objects. */
+ struct xarray xa;
+
+ /** @gpu_contexts: Buffer object containing the GPU heap contexts. */
+ struct panthor_kernel_bo *gpu_contexts;
+};
+
+static int panthor_heap_ctx_stride(struct panthor_device *ptdev)
+{
+ u32 l2_features = ptdev->gpu_info.l2_features;
+ u32 gpu_cache_line_size = GPU_L2_FEATURES_LINE_SIZE(l2_features);
+
+ return ALIGN(HEAP_CONTEXT_SIZE, gpu_cache_line_size);
+}
+
+static int panthor_get_heap_ctx_offset(struct panthor_heap_pool *pool, int id)
+{
+ return panthor_heap_ctx_stride(pool->ptdev) * id;
+}
+
+static void *panthor_get_heap_ctx(struct panthor_heap_pool *pool, int id)
+{
+ return pool->gpu_contexts->kmap +
+ panthor_get_heap_ctx_offset(pool, id);
+}
+
+static void panthor_free_heap_chunk(struct panthor_vm *vm,
+ struct panthor_heap *heap,
+ struct panthor_heap_chunk *chunk)
+{
+ mutex_lock(&heap->lock);
+ list_del(&chunk->node);
+ heap->chunk_count--;
+ mutex_unlock(&heap->lock);
+
+ panthor_kernel_bo_destroy(vm, chunk->bo);
+ kfree(chunk);
+}
+
+static int panthor_alloc_heap_chunk(struct panthor_device *ptdev,
+ struct panthor_vm *vm,
+ struct panthor_heap *heap,
+ bool initial_chunk)
+{
+ struct panthor_heap_chunk *chunk;
+ struct panthor_heap_chunk_header *hdr;
+ int ret;
+
+ chunk = kmalloc(sizeof(*chunk), GFP_KERNEL);
+ if (!chunk)
+ return -ENOMEM;
+
+ chunk->bo = panthor_kernel_bo_create(ptdev, vm, heap->chunk_size,
+ DRM_PANTHOR_BO_NO_MMAP,
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC,
+ PANTHOR_VM_KERNEL_AUTO_VA);
+ if (IS_ERR(chunk->bo)) {
+ ret = PTR_ERR(chunk->bo);
+ goto err_free_chunk;
+ }
+
+ ret = panthor_kernel_bo_vmap(chunk->bo);
+ if (ret)
+ goto err_destroy_bo;
+
+ hdr = chunk->bo->kmap;
+ memset(hdr, 0, sizeof(*hdr));
+
+ if (initial_chunk && !list_empty(&heap->chunks)) {
+ struct panthor_heap_chunk *prev_chunk;
+ u64 prev_gpuva;
+
+ prev_chunk = list_first_entry(&heap->chunks,
+ struct panthor_heap_chunk,
+ node);
+
+ prev_gpuva = panthor_kernel_bo_gpuva(prev_chunk->bo);
+ hdr->next = (prev_gpuva & GENMASK_ULL(63, 12)) |
+ (heap->chunk_size >> 12);
+ }
+
+ panthor_kernel_bo_vunmap(chunk->bo);
+
+ mutex_lock(&heap->lock);
+ list_add(&chunk->node, &heap->chunks);
+ heap->chunk_count++;
+ mutex_unlock(&heap->lock);
+
+ return 0;
+
+err_destroy_bo:
+ panthor_kernel_bo_destroy(vm, chunk->bo);
+
+err_free_chunk:
+ kfree(chunk);
+
+ return ret;
+}
+
+static void panthor_free_heap_chunks(struct panthor_vm *vm,
+ struct panthor_heap *heap)
+{
+ struct panthor_heap_chunk *chunk, *tmp;
+
+ list_for_each_entry_safe(chunk, tmp, &heap->chunks, node)
+ panthor_free_heap_chunk(vm, heap, chunk);
+}
+
+static int panthor_alloc_heap_chunks(struct panthor_device *ptdev,
+ struct panthor_vm *vm,
+ struct panthor_heap *heap,
+ u32 chunk_count)
+{
+ int ret;
+ u32 i;
+
+ for (i = 0; i < chunk_count; i++) {
+ ret = panthor_alloc_heap_chunk(ptdev, vm, heap, true);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+static int
+panthor_heap_destroy_locked(struct panthor_heap_pool *pool, u32 handle)
+{
+ struct panthor_heap *heap;
+
+ heap = xa_erase(&pool->xa, handle);
+ if (!heap)
+ return -EINVAL;
+
+ panthor_free_heap_chunks(pool->vm, heap);
+ mutex_destroy(&heap->lock);
+ kfree(heap);
+ return 0;
+}
+
+/**
+ * panthor_heap_destroy() - Destroy a heap context
+ * @pool: Pool this context belongs to.
+ * @handle: Handle returned by panthor_heap_create().
+ */
+int panthor_heap_destroy(struct panthor_heap_pool *pool, u32 handle)
+{
+ int ret;
+
+ down_write(&pool->lock);
+ ret = panthor_heap_destroy_locked(pool, handle);
+ up_write(&pool->lock);
+
+ return ret;
+}
+
+/**
+ * panthor_heap_create() - Create a heap context
+ * @pool: Pool to instantiate the heap context from.
+ * @initial_chunk_count: Number of chunk allocated at initialization time.
+ * Must be at least 1.
+ * @chunk_size: The size of each chunk. Must be a power of two between 256k
+ * and 2M.
+ * @max_chunks: Maximum number of chunks that can be allocated.
+ * @target_in_flight: Maximum number of in-flight render passes.
+ * @heap_ctx_gpu_va: Pointer holding the GPU address of the allocated heap
+ * context.
+ * @first_chunk_gpu_va: Pointer holding the GPU address of the first chunk
+ * assigned to the heap context.
+ *
+ * Return: a positive handle on success, a negative error otherwise.
+ */
+int panthor_heap_create(struct panthor_heap_pool *pool,
+ u32 initial_chunk_count,
+ u32 chunk_size,
+ u32 max_chunks,
+ u32 target_in_flight,
+ u64 *heap_ctx_gpu_va,
+ u64 *first_chunk_gpu_va)
+{
+ struct panthor_heap *heap;
+ struct panthor_heap_chunk *first_chunk;
+ struct panthor_vm *vm;
+ int ret = 0;
+ u32 id;
+
+ if (initial_chunk_count == 0)
+ return -EINVAL;
+
+ if (hweight32(chunk_size) != 1 ||
+ chunk_size < SZ_256K || chunk_size > SZ_2M)
+ return -EINVAL;
+
+ down_read(&pool->lock);
+ vm = panthor_vm_get(pool->vm);
+ up_read(&pool->lock);
+
+ /* The pool has been destroyed, we can't create a new heap. */
+ if (!vm)
+ return -EINVAL;
+
+ heap = kzalloc(sizeof(*heap), GFP_KERNEL);
+ if (!heap) {
+ ret = -ENOMEM;
+ goto err_put_vm;
+ }
+
+ mutex_init(&heap->lock);
+ INIT_LIST_HEAD(&heap->chunks);
+ heap->chunk_size = chunk_size;
+ heap->max_chunks = max_chunks;
+ heap->target_in_flight = target_in_flight;
+
+ ret = panthor_alloc_heap_chunks(pool->ptdev, vm, heap,
+ initial_chunk_count);
+ if (ret)
+ goto err_free_heap;
+
+ first_chunk = list_first_entry(&heap->chunks,
+ struct panthor_heap_chunk,
+ node);
+ *first_chunk_gpu_va = panthor_kernel_bo_gpuva(first_chunk->bo);
+
+ down_write(&pool->lock);
+ /* The pool has been destroyed, we can't create a new heap. */
+ if (!pool->vm) {
+ ret = -EINVAL;
+ } else {
+ ret = xa_alloc(&pool->xa, &id, heap, XA_LIMIT(1, MAX_HEAPS_PER_POOL), GFP_KERNEL);
+ if (!ret) {
+ void *gpu_ctx = panthor_get_heap_ctx(pool, id);
+
+ memset(gpu_ctx, 0, panthor_heap_ctx_stride(pool->ptdev));
+ *heap_ctx_gpu_va = panthor_kernel_bo_gpuva(pool->gpu_contexts) +
+ panthor_get_heap_ctx_offset(pool, id);
+ }
+ }
+ up_write(&pool->lock);
+
+ if (ret)
+ goto err_free_heap;
+
+ panthor_vm_put(vm);
+ return id;
+
+err_free_heap:
+ panthor_free_heap_chunks(pool->vm, heap);
+ mutex_destroy(&heap->lock);
+ kfree(heap);
+
+err_put_vm:
+ panthor_vm_put(vm);
+ return ret;
+}
+
+/**
+ * panthor_heap_return_chunk() - Return an unused heap chunk
+ * @pool: The pool this heap belongs to.
+ * @heap_gpu_va: The GPU address of the heap context.
+ * @chunk_gpu_va: The chunk VA to return.
+ *
+ * This function is used when a chunk allocated with panthor_heap_grow()
+ * couldn't be linked to the heap context through the FW interface because
+ * the group requesting the allocation was scheduled out in the meantime.
+ */
+int panthor_heap_return_chunk(struct panthor_heap_pool *pool,
+ u64 heap_gpu_va,
+ u64 chunk_gpu_va)
+{
+ u64 offset = heap_gpu_va - panthor_kernel_bo_gpuva(pool->gpu_contexts);
+ u32 heap_id = (u32)offset / panthor_heap_ctx_stride(pool->ptdev);
+ struct panthor_heap_chunk *chunk, *tmp, *removed = NULL;
+ struct panthor_heap *heap;
+ int ret;
+
+ if (offset > U32_MAX || heap_id >= MAX_HEAPS_PER_POOL)
+ return -EINVAL;
+
+ down_read(&pool->lock);
+ heap = xa_load(&pool->xa, heap_id);
+ if (!heap) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
+ chunk_gpu_va &= GENMASK_ULL(63, 12);
+
+ mutex_lock(&heap->lock);
+ list_for_each_entry_safe(chunk, tmp, &heap->chunks, node) {
+ if (panthor_kernel_bo_gpuva(chunk->bo) == chunk_gpu_va) {
+ removed = chunk;
+ list_del(&chunk->node);
+ heap->chunk_count--;
+ break;
+ }
+ }
+ mutex_unlock(&heap->lock);
+
+ if (removed) {
+ panthor_kernel_bo_destroy(pool->vm, chunk->bo);
+ kfree(chunk);
+ ret = 0;
+ } else {
+ ret = -EINVAL;
+ }
+
+out_unlock:
+ up_read(&pool->lock);
+ return ret;
+}
+
+/**
+ * panthor_heap_grow() - Make a heap context grow.
+ * @pool: The pool this heap belongs to.
+ * @heap_gpu_va: The GPU address of the heap context.
+ * @renderpasses_in_flight: Number of render passes currently in-flight.
+ * @pending_frag_count: Number of fragment jobs waiting for execution/completion.
+ * @new_chunk_gpu_va: Pointer used to return the chunk VA.
+ */
+int panthor_heap_grow(struct panthor_heap_pool *pool,
+ u64 heap_gpu_va,
+ u32 renderpasses_in_flight,
+ u32 pending_frag_count,
+ u64 *new_chunk_gpu_va)
+{
+ u64 offset = heap_gpu_va - panthor_kernel_bo_gpuva(pool->gpu_contexts);
+ u32 heap_id = (u32)offset / panthor_heap_ctx_stride(pool->ptdev);
+ struct panthor_heap_chunk *chunk;
+ struct panthor_heap *heap;
+ int ret;
+
+ if (offset > U32_MAX || heap_id >= MAX_HEAPS_PER_POOL)
+ return -EINVAL;
+
+ down_read(&pool->lock);
+ heap = xa_load(&pool->xa, heap_id);
+ if (!heap) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
+ /* If we reached the target in-flight render passes, or if we
+ * reached the maximum number of chunks, let the FW figure another way to
+ * find some memory (wait for render passes to finish, or call the exception
+ * handler provided by the userspace driver, if any).
+ */
+ if (renderpasses_in_flight > heap->target_in_flight ||
+ (pending_frag_count > 0 && heap->chunk_count >= heap->max_chunks)) {
+ ret = -EBUSY;
+ goto out_unlock;
+ } else if (heap->chunk_count >= heap->max_chunks) {
+ ret = -ENOMEM;
+ goto out_unlock;
+ }
+
+ /* FIXME: panthor_alloc_heap_chunk() triggers a kernel BO creation, which
+ * relies on blocking allocations (both for the BO itself, and backing
+ * memory), which might cause a deadlock because we're called from a context
+ * where we hold the panthor scheduler lock, thus preventing job cleanups
+ * that could free up some memory. The jobs themselves will timeout, but
+ * we'll still be blocked there. The only solution here is to implement
+ * something similar to shmem_sg_alloc_table() in i915, so we can do
+ * non-blocking allocations, and just kill the job when we run out-of-memory
+ * for the tiler context.
+ */
+ ret = panthor_alloc_heap_chunk(pool->ptdev, pool->vm, heap, false);
+ if (ret)
+ goto out_unlock;
+
+ chunk = list_first_entry(&heap->chunks,
+ struct panthor_heap_chunk,
+ node);
+ *new_chunk_gpu_va = (panthor_kernel_bo_gpuva(chunk->bo) & GENMASK_ULL(63, 12)) |
+ (heap->chunk_size >> 12);
+ ret = 0;
+
+out_unlock:
+ up_read(&pool->lock);
+ return ret;
+}
+
+static void panthor_heap_pool_release(struct kref *refcount)
+{
+ struct panthor_heap_pool *pool =
+ container_of(refcount, struct panthor_heap_pool, refcount);
+
+ xa_destroy(&pool->xa);
+ kfree(pool);
+}
+
+/**
+ * panthor_heap_pool_put() - Release a heap pool reference
+ * @pool: Pool to release the reference on. Can be NULL.
+ */
+void panthor_heap_pool_put(struct panthor_heap_pool *pool)
+{
+ if (pool)
+ kref_put(&pool->refcount, panthor_heap_pool_release);
+}
+
+/**
+ * panthor_heap_pool_get() - Get a heap pool reference
+ * @pool: Pool to get the reference on. Can be NULL.
+ *
+ * Return: @pool.
+ */
+struct panthor_heap_pool *
+panthor_heap_pool_get(struct panthor_heap_pool *pool)
+{
+ if (pool)
+ kref_get(&pool->refcount);
+
+ return pool;
+}
+
+/**
+ * panthor_heap_pool_create() - Create a heap pool
+ * @ptdev: Device.
+ * @vm: The VM this heap pool will be attached to.
+ *
+ * Heap pools might contain up to 128 heap contexts, and are per-VM.
+ *
+ * Return: A valid pointer on success, a negative error code otherwise.
+ */
+struct panthor_heap_pool *
+panthor_heap_pool_create(struct panthor_device *ptdev, struct panthor_vm *vm)
+{
+ size_t bosize = ALIGN(MAX_HEAPS_PER_POOL *
+ panthor_heap_ctx_stride(ptdev),
+ 4096);
+ struct panthor_heap_pool *pool;
+ int ret = 0;
+
+ pool = kzalloc(sizeof(*pool), GFP_KERNEL);
+ if (!pool)
+ return ERR_PTR(-ENOMEM);
+
+ /* We want a weak ref here: the heap pool belongs to the VM, so we're
+ * sure that, as long as the heap pool exists, the VM exists too.
+ */
+ pool->vm = vm;
+ pool->ptdev = ptdev;
+ init_rwsem(&pool->lock);
+ xa_init_flags(&pool->xa, XA_FLAGS_ALLOC1);
+ kref_init(&pool->refcount);
+
+ pool->gpu_contexts = panthor_kernel_bo_create(ptdev, vm, bosize,
+ DRM_PANTHOR_BO_NO_MMAP,
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC,
+ PANTHOR_VM_KERNEL_AUTO_VA);
+ if (IS_ERR(pool->gpu_contexts)) {
+ ret = PTR_ERR(pool->gpu_contexts);
+ goto err_destroy_pool;
+ }
+
+ ret = panthor_kernel_bo_vmap(pool->gpu_contexts);
+ if (ret)
+ goto err_destroy_pool;
+
+ return pool;
+
+err_destroy_pool:
+ panthor_heap_pool_destroy(pool);
+ return ERR_PTR(ret);
+}
+
+/**
+ * panthor_heap_pool_destroy() - Destroy a heap pool.
+ * @pool: Pool to destroy.
+ *
+ * This function destroys all heap contexts and their resources. Thus
+ * preventing any use of the heap context or the chunk attached to them
+ * after that point.
+ *
+ * If the GPU still has access to some heap contexts, a fault should be
+ * triggered, which should flag the command stream groups using these
+ * context as faulty.
+ *
+ * The heap pool object is only released when all references to this pool
+ * are released.
+ */
+void panthor_heap_pool_destroy(struct panthor_heap_pool *pool)
+{
+ struct panthor_heap *heap;
+ unsigned long i;
+
+ if (!pool)
+ return;
+
+ down_write(&pool->lock);
+ xa_for_each(&pool->xa, i, heap)
+ drm_WARN_ON(&pool->ptdev->base, panthor_heap_destroy_locked(pool, i));
+
+ if (!IS_ERR_OR_NULL(pool->gpu_contexts))
+ panthor_kernel_bo_destroy(pool->vm, pool->gpu_contexts);
+
+ /* Reflects the fact the pool has been destroyed. */
+ pool->vm = NULL;
+ up_write(&pool->lock);
+
+ panthor_heap_pool_put(pool);
+}
diff --git a/drivers/gpu/drm/panthor/panthor_heap.h b/drivers/gpu/drm/panthor/panthor_heap.h
new file mode 100644
index 000000000000..25a5f2bba445
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_heap.h
@@ -0,0 +1,39 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2023 Collabora ltd. */
+
+#ifndef __PANTHOR_HEAP_H__
+#define __PANTHOR_HEAP_H__
+
+#include <linux/types.h>
+
+struct panthor_device;
+struct panthor_heap_pool;
+struct panthor_vm;
+
+int panthor_heap_create(struct panthor_heap_pool *pool,
+ u32 initial_chunk_count,
+ u32 chunk_size,
+ u32 max_chunks,
+ u32 target_in_flight,
+ u64 *heap_ctx_gpu_va,
+ u64 *first_chunk_gpu_va);
+int panthor_heap_destroy(struct panthor_heap_pool *pool, u32 handle);
+
+struct panthor_heap_pool *
+panthor_heap_pool_create(struct panthor_device *ptdev, struct panthor_vm *vm);
+void panthor_heap_pool_destroy(struct panthor_heap_pool *pool);
+
+struct panthor_heap_pool *
+panthor_heap_pool_get(struct panthor_heap_pool *pool);
+void panthor_heap_pool_put(struct panthor_heap_pool *pool);
+
+int panthor_heap_grow(struct panthor_heap_pool *pool,
+ u64 heap_gpu_va,
+ u32 renderpasses_in_flight,
+ u32 pending_frag_count,
+ u64 *new_chunk_gpu_va);
+int panthor_heap_return_chunk(struct panthor_heap_pool *pool,
+ u64 heap_gpu_va,
+ u64 chunk_gpu_va);
+
+#endif
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 09/14] drm/panthor: Add the heap logical block
2024-01-22 16:30 ` [PATCH v4 09/14] drm/panthor: Add the heap " Boris Brezillon
@ 2024-02-12 11:40 ` Steven Price
2024-02-14 17:33 ` Boris Brezillon
0 siblings, 1 reply; 51+ messages in thread
From: Steven Price @ 2024-02-12 11:40 UTC (permalink / raw)
To: Boris Brezillon, dri-devel
Cc: Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen
On 22/01/2024 16:30, Boris Brezillon wrote:
> Tiler heap growing requires some kernel driver involvement: when the
> tiler runs out of heap memory, it will raise an exception which is
> either directly handled by the firmware if some free heap chunks are
> available in the heap context, or passed back to the kernel otherwise.
> The heap helpers will be used by the scheduler logic to allocate more
> heap chunks to a heap context, when such a situation happens.
>
> Heap context creation is explicitly requested by userspace (using
> the TILER_HEAP_CREATE ioctl), and the returned context is attached to a
> queue through some command stream instruction.
>
> All the kernel does is keep the list of heap chunks allocated to a
> context, so they can be freed when TILER_HEAP_DESTROY is called, or
> extended when the FW requests a new chunk.
>
> v4:
> - Rework locking to allow concurrent calls to panthor_heap_grow()
> - Add a helper to return a heap chunk if we couldn't pass it to the
> FW because the group was scheduled out
>
> v3:
> - Add a FIXME for the heap OOM deadlock
> - Use the panthor_kernel_bo abstraction for the heap context and heap
> chunks
> - Drop the panthor_heap_gpu_ctx struct as it is opaque to the driver
> - Ensure that the heap context is aligned to the GPU cache line size
> - Minor code tidy ups
>
> Co-developed-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
It looks fine, but there's a confusing FIXME comment:
> + /* FIXME: panthor_alloc_heap_chunk() triggers a kernel BO creation, which
> + * relies on blocking allocations (both for the BO itself, and backing
> + * memory), which might cause a deadlock because we're called from a context
> + * where we hold the panthor scheduler lock, thus preventing job cleanups
> + * that could free up some memory. The jobs themselves will timeout, but
> + * we'll still be blocked there. The only solution here is to implement
> + * something similar to shmem_sg_alloc_table() in i915, so we can do
> + * non-blocking allocations, and just kill the job when we run out-of-memory
> + * for the tiler context.
> + */
Whereas at the call site (group_process_tiler_oom()) there's the comment:
> /* We do the allocation without holding the scheduler lock to avoid
> * blocking the scheduling.
> */
AFAICT that FIXME comment can just be deleted now. Assuming you agree
then with that change:
Reviewed-by: Steven Price <steven.price@arm.com>
Thanks,
Steve
> ---
> drivers/gpu/drm/panthor/panthor_heap.c | 596 +++++++++++++++++++++++++
> drivers/gpu/drm/panthor/panthor_heap.h | 39 ++
> 2 files changed, 635 insertions(+)
> create mode 100644 drivers/gpu/drm/panthor/panthor_heap.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_heap.h
>
> diff --git a/drivers/gpu/drm/panthor/panthor_heap.c b/drivers/gpu/drm/panthor/panthor_heap.c
> new file mode 100644
> index 000000000000..fac51c24b3e9
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_heap.c
> @@ -0,0 +1,596 @@
> +// SPDX-License-Identifier: GPL-2.0 or MIT
> +/* Copyright 2023 Collabora ltd. */
> +
> +#include <linux/iosys-map.h>
> +#include <linux/rwsem.h>
> +
> +#include <drm/panthor_drm.h>
> +
> +#include "panthor_device.h"
> +#include "panthor_gem.h"
> +#include "panthor_heap.h"
> +#include "panthor_mmu.h"
> +#include "panthor_regs.h"
> +
> +/*
> + * The GPU heap context is an opaque structure used by the GPU to track the
> + * heap allocations. The driver should only touch it to initialize it (zero all
> + * fields). Because the CPU and GPU can both access this structure it is
> + * required to be GPU cache line aligned.
> + */
> +#define HEAP_CONTEXT_SIZE 32
> +
> +/**
> + * struct panthor_heap_chunk_header - Heap chunk header
> + */
> +struct panthor_heap_chunk_header {
> + /**
> + * @next: Next heap chunk in the list.
> + *
> + * This is a GPU VA.
> + */
> + u64 next;
> +
> + /** @unknown: MBZ. */
> + u32 unknown[14];
> +};
> +
> +/**
> + * struct panthor_heap_chunk - Structure used to keep track of allocated heap chunks.
> + */
> +struct panthor_heap_chunk {
> + /** @node: Used to insert the heap chunk in panthor_heap::chunks. */
> + struct list_head node;
> +
> + /** @bo: Buffer object backing the heap chunk. */
> + struct panthor_kernel_bo *bo;
> +};
> +
> +/**
> + * struct panthor_heap - Structure used to manage tiler heap contexts.
> + */
> +struct panthor_heap {
> + /** @chunks: List containing all heap chunks allocated so far. */
> + struct list_head chunks;
> +
> + /** @lock: Lock protecting insertion in the chunks list. */
> + struct mutex lock;
> +
> + /** @chunk_size: Size of each chunk. */
> + u32 chunk_size;
> +
> + /** @max_chunks: Maximum number of chunks. */
> + u32 max_chunks;
> +
> + /**
> + * @target_in_flight: Number of in-flight render passes after which
> + * we'd let the FW wait for fragment job to finish instead of allocating new chunks.
> + */
> + u32 target_in_flight;
> +
> + /** @chunk_count: Number of heap chunks currently allocated. */
> + u32 chunk_count;
> +};
> +
> +#define MAX_HEAPS_PER_POOL 128
> +
> +/**
> + * struct panthor_heap_pool - Pool of heap contexts
> + *
> + * The pool is attached to a panthor_file and can't be shared across processes.
> + */
> +struct panthor_heap_pool {
> + /** @refcount: Reference count. */
> + struct kref refcount;
> +
> + /** @ptdev: Device. */
> + struct panthor_device *ptdev;
> +
> + /** @vm: VM this pool is bound to. */
> + struct panthor_vm *vm;
> +
> + /** @lock: Lock protecting access to @xa. */
> + struct rw_semaphore lock;
> +
> + /** @xa: Array storing panthor_heap objects. */
> + struct xarray xa;
> +
> + /** @gpu_contexts: Buffer object containing the GPU heap contexts. */
> + struct panthor_kernel_bo *gpu_contexts;
> +};
> +
> +static int panthor_heap_ctx_stride(struct panthor_device *ptdev)
> +{
> + u32 l2_features = ptdev->gpu_info.l2_features;
> + u32 gpu_cache_line_size = GPU_L2_FEATURES_LINE_SIZE(l2_features);
> +
> + return ALIGN(HEAP_CONTEXT_SIZE, gpu_cache_line_size);
> +}
> +
> +static int panthor_get_heap_ctx_offset(struct panthor_heap_pool *pool, int id)
> +{
> + return panthor_heap_ctx_stride(pool->ptdev) * id;
> +}
> +
> +static void *panthor_get_heap_ctx(struct panthor_heap_pool *pool, int id)
> +{
> + return pool->gpu_contexts->kmap +
> + panthor_get_heap_ctx_offset(pool, id);
> +}
> +
> +static void panthor_free_heap_chunk(struct panthor_vm *vm,
> + struct panthor_heap *heap,
> + struct panthor_heap_chunk *chunk)
> +{
> + mutex_lock(&heap->lock);
> + list_del(&chunk->node);
> + heap->chunk_count--;
> + mutex_unlock(&heap->lock);
> +
> + panthor_kernel_bo_destroy(vm, chunk->bo);
> + kfree(chunk);
> +}
> +
> +static int panthor_alloc_heap_chunk(struct panthor_device *ptdev,
> + struct panthor_vm *vm,
> + struct panthor_heap *heap,
> + bool initial_chunk)
> +{
> + struct panthor_heap_chunk *chunk;
> + struct panthor_heap_chunk_header *hdr;
> + int ret;
> +
> + chunk = kmalloc(sizeof(*chunk), GFP_KERNEL);
> + if (!chunk)
> + return -ENOMEM;
> +
> + chunk->bo = panthor_kernel_bo_create(ptdev, vm, heap->chunk_size,
> + DRM_PANTHOR_BO_NO_MMAP,
> + DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC,
> + PANTHOR_VM_KERNEL_AUTO_VA);
> + if (IS_ERR(chunk->bo)) {
> + ret = PTR_ERR(chunk->bo);
> + goto err_free_chunk;
> + }
> +
> + ret = panthor_kernel_bo_vmap(chunk->bo);
> + if (ret)
> + goto err_destroy_bo;
> +
> + hdr = chunk->bo->kmap;
> + memset(hdr, 0, sizeof(*hdr));
> +
> + if (initial_chunk && !list_empty(&heap->chunks)) {
> + struct panthor_heap_chunk *prev_chunk;
> + u64 prev_gpuva;
> +
> + prev_chunk = list_first_entry(&heap->chunks,
> + struct panthor_heap_chunk,
> + node);
> +
> + prev_gpuva = panthor_kernel_bo_gpuva(prev_chunk->bo);
> + hdr->next = (prev_gpuva & GENMASK_ULL(63, 12)) |
> + (heap->chunk_size >> 12);
> + }
> +
> + panthor_kernel_bo_vunmap(chunk->bo);
> +
> + mutex_lock(&heap->lock);
> + list_add(&chunk->node, &heap->chunks);
> + heap->chunk_count++;
> + mutex_unlock(&heap->lock);
> +
> + return 0;
> +
> +err_destroy_bo:
> + panthor_kernel_bo_destroy(vm, chunk->bo);
> +
> +err_free_chunk:
> + kfree(chunk);
> +
> + return ret;
> +}
> +
> +static void panthor_free_heap_chunks(struct panthor_vm *vm,
> + struct panthor_heap *heap)
> +{
> + struct panthor_heap_chunk *chunk, *tmp;
> +
> + list_for_each_entry_safe(chunk, tmp, &heap->chunks, node)
> + panthor_free_heap_chunk(vm, heap, chunk);
> +}
> +
> +static int panthor_alloc_heap_chunks(struct panthor_device *ptdev,
> + struct panthor_vm *vm,
> + struct panthor_heap *heap,
> + u32 chunk_count)
> +{
> + int ret;
> + u32 i;
> +
> + for (i = 0; i < chunk_count; i++) {
> + ret = panthor_alloc_heap_chunk(ptdev, vm, heap, true);
> + if (ret)
> + return ret;
> + }
> +
> + return 0;
> +}
> +
> +static int
> +panthor_heap_destroy_locked(struct panthor_heap_pool *pool, u32 handle)
> +{
> + struct panthor_heap *heap;
> +
> + heap = xa_erase(&pool->xa, handle);
> + if (!heap)
> + return -EINVAL;
> +
> + panthor_free_heap_chunks(pool->vm, heap);
> + mutex_destroy(&heap->lock);
> + kfree(heap);
> + return 0;
> +}
> +
> +/**
> + * panthor_heap_destroy() - Destroy a heap context
> + * @pool: Pool this context belongs to.
> + * @handle: Handle returned by panthor_heap_create().
> + */
> +int panthor_heap_destroy(struct panthor_heap_pool *pool, u32 handle)
> +{
> + int ret;
> +
> + down_write(&pool->lock);
> + ret = panthor_heap_destroy_locked(pool, handle);
> + up_write(&pool->lock);
> +
> + return ret;
> +}
> +
> +/**
> + * panthor_heap_create() - Create a heap context
> + * @pool: Pool to instantiate the heap context from.
> + * @initial_chunk_count: Number of chunk allocated at initialization time.
> + * Must be at least 1.
> + * @chunk_size: The size of each chunk. Must be a power of two between 256k
> + * and 2M.
> + * @max_chunks: Maximum number of chunks that can be allocated.
> + * @target_in_flight: Maximum number of in-flight render passes.
> + * @heap_ctx_gpu_va: Pointer holding the GPU address of the allocated heap
> + * context.
> + * @first_chunk_gpu_va: Pointer holding the GPU address of the first chunk
> + * assigned to the heap context.
> + *
> + * Return: a positive handle on success, a negative error otherwise.
> + */
> +int panthor_heap_create(struct panthor_heap_pool *pool,
> + u32 initial_chunk_count,
> + u32 chunk_size,
> + u32 max_chunks,
> + u32 target_in_flight,
> + u64 *heap_ctx_gpu_va,
> + u64 *first_chunk_gpu_va)
> +{
> + struct panthor_heap *heap;
> + struct panthor_heap_chunk *first_chunk;
> + struct panthor_vm *vm;
> + int ret = 0;
> + u32 id;
> +
> + if (initial_chunk_count == 0)
> + return -EINVAL;
> +
> + if (hweight32(chunk_size) != 1 ||
> + chunk_size < SZ_256K || chunk_size > SZ_2M)
> + return -EINVAL;
> +
> + down_read(&pool->lock);
> + vm = panthor_vm_get(pool->vm);
> + up_read(&pool->lock);
> +
> + /* The pool has been destroyed, we can't create a new heap. */
> + if (!vm)
> + return -EINVAL;
> +
> + heap = kzalloc(sizeof(*heap), GFP_KERNEL);
> + if (!heap) {
> + ret = -ENOMEM;
> + goto err_put_vm;
> + }
> +
> + mutex_init(&heap->lock);
> + INIT_LIST_HEAD(&heap->chunks);
> + heap->chunk_size = chunk_size;
> + heap->max_chunks = max_chunks;
> + heap->target_in_flight = target_in_flight;
> +
> + ret = panthor_alloc_heap_chunks(pool->ptdev, vm, heap,
> + initial_chunk_count);
> + if (ret)
> + goto err_free_heap;
> +
> + first_chunk = list_first_entry(&heap->chunks,
> + struct panthor_heap_chunk,
> + node);
> + *first_chunk_gpu_va = panthor_kernel_bo_gpuva(first_chunk->bo);
> +
> + down_write(&pool->lock);
> + /* The pool has been destroyed, we can't create a new heap. */
> + if (!pool->vm) {
> + ret = -EINVAL;
> + } else {
> + ret = xa_alloc(&pool->xa, &id, heap, XA_LIMIT(1, MAX_HEAPS_PER_POOL), GFP_KERNEL);
> + if (!ret) {
> + void *gpu_ctx = panthor_get_heap_ctx(pool, id);
> +
> + memset(gpu_ctx, 0, panthor_heap_ctx_stride(pool->ptdev));
> + *heap_ctx_gpu_va = panthor_kernel_bo_gpuva(pool->gpu_contexts) +
> + panthor_get_heap_ctx_offset(pool, id);
> + }
> + }
> + up_write(&pool->lock);
> +
> + if (ret)
> + goto err_free_heap;
> +
> + panthor_vm_put(vm);
> + return id;
> +
> +err_free_heap:
> + panthor_free_heap_chunks(pool->vm, heap);
> + mutex_destroy(&heap->lock);
> + kfree(heap);
> +
> +err_put_vm:
> + panthor_vm_put(vm);
> + return ret;
> +}
> +
> +/**
> + * panthor_heap_return_chunk() - Return an unused heap chunk
> + * @pool: The pool this heap belongs to.
> + * @heap_gpu_va: The GPU address of the heap context.
> + * @chunk_gpu_va: The chunk VA to return.
> + *
> + * This function is used when a chunk allocated with panthor_heap_grow()
> + * couldn't be linked to the heap context through the FW interface because
> + * the group requesting the allocation was scheduled out in the meantime.
> + */
> +int panthor_heap_return_chunk(struct panthor_heap_pool *pool,
> + u64 heap_gpu_va,
> + u64 chunk_gpu_va)
> +{
> + u64 offset = heap_gpu_va - panthor_kernel_bo_gpuva(pool->gpu_contexts);
> + u32 heap_id = (u32)offset / panthor_heap_ctx_stride(pool->ptdev);
> + struct panthor_heap_chunk *chunk, *tmp, *removed = NULL;
> + struct panthor_heap *heap;
> + int ret;
> +
> + if (offset > U32_MAX || heap_id >= MAX_HEAPS_PER_POOL)
> + return -EINVAL;
> +
> + down_read(&pool->lock);
> + heap = xa_load(&pool->xa, heap_id);
> + if (!heap) {
> + ret = -EINVAL;
> + goto out_unlock;
> + }
> +
> + chunk_gpu_va &= GENMASK_ULL(63, 12);
> +
> + mutex_lock(&heap->lock);
> + list_for_each_entry_safe(chunk, tmp, &heap->chunks, node) {
> + if (panthor_kernel_bo_gpuva(chunk->bo) == chunk_gpu_va) {
> + removed = chunk;
> + list_del(&chunk->node);
> + heap->chunk_count--;
> + break;
> + }
> + }
> + mutex_unlock(&heap->lock);
> +
> + if (removed) {
> + panthor_kernel_bo_destroy(pool->vm, chunk->bo);
> + kfree(chunk);
> + ret = 0;
> + } else {
> + ret = -EINVAL;
> + }
> +
> +out_unlock:
> + up_read(&pool->lock);
> + return ret;
> +}
> +
> +/**
> + * panthor_heap_grow() - Make a heap context grow.
> + * @pool: The pool this heap belongs to.
> + * @heap_gpu_va: The GPU address of the heap context.
> + * @renderpasses_in_flight: Number of render passes currently in-flight.
> + * @pending_frag_count: Number of fragment jobs waiting for execution/completion.
> + * @new_chunk_gpu_va: Pointer used to return the chunk VA.
> + */
> +int panthor_heap_grow(struct panthor_heap_pool *pool,
> + u64 heap_gpu_va,
> + u32 renderpasses_in_flight,
> + u32 pending_frag_count,
> + u64 *new_chunk_gpu_va)
> +{
> + u64 offset = heap_gpu_va - panthor_kernel_bo_gpuva(pool->gpu_contexts);
> + u32 heap_id = (u32)offset / panthor_heap_ctx_stride(pool->ptdev);
> + struct panthor_heap_chunk *chunk;
> + struct panthor_heap *heap;
> + int ret;
> +
> + if (offset > U32_MAX || heap_id >= MAX_HEAPS_PER_POOL)
> + return -EINVAL;
> +
> + down_read(&pool->lock);
> + heap = xa_load(&pool->xa, heap_id);
> + if (!heap) {
> + ret = -EINVAL;
> + goto out_unlock;
> + }
> +
> + /* If we reached the target in-flight render passes, or if we
> + * reached the maximum number of chunks, let the FW figure another way to
> + * find some memory (wait for render passes to finish, or call the exception
> + * handler provided by the userspace driver, if any).
> + */
> + if (renderpasses_in_flight > heap->target_in_flight ||
> + (pending_frag_count > 0 && heap->chunk_count >= heap->max_chunks)) {
> + ret = -EBUSY;
> + goto out_unlock;
> + } else if (heap->chunk_count >= heap->max_chunks) {
> + ret = -ENOMEM;
> + goto out_unlock;
> + }
> +
> + /* FIXME: panthor_alloc_heap_chunk() triggers a kernel BO creation, which
> + * relies on blocking allocations (both for the BO itself, and backing
> + * memory), which might cause a deadlock because we're called from a context
> + * where we hold the panthor scheduler lock, thus preventing job cleanups
> + * that could free up some memory. The jobs themselves will timeout, but
> + * we'll still be blocked there. The only solution here is to implement
> + * something similar to shmem_sg_alloc_table() in i915, so we can do
> + * non-blocking allocations, and just kill the job when we run out-of-memory
> + * for the tiler context.
> + */
> + ret = panthor_alloc_heap_chunk(pool->ptdev, pool->vm, heap, false);
> + if (ret)
> + goto out_unlock;
> +
> + chunk = list_first_entry(&heap->chunks,
> + struct panthor_heap_chunk,
> + node);
> + *new_chunk_gpu_va = (panthor_kernel_bo_gpuva(chunk->bo) & GENMASK_ULL(63, 12)) |
> + (heap->chunk_size >> 12);
> + ret = 0;
> +
> +out_unlock:
> + up_read(&pool->lock);
> + return ret;
> +}
> +
> +static void panthor_heap_pool_release(struct kref *refcount)
> +{
> + struct panthor_heap_pool *pool =
> + container_of(refcount, struct panthor_heap_pool, refcount);
> +
> + xa_destroy(&pool->xa);
> + kfree(pool);
> +}
> +
> +/**
> + * panthor_heap_pool_put() - Release a heap pool reference
> + * @pool: Pool to release the reference on. Can be NULL.
> + */
> +void panthor_heap_pool_put(struct panthor_heap_pool *pool)
> +{
> + if (pool)
> + kref_put(&pool->refcount, panthor_heap_pool_release);
> +}
> +
> +/**
> + * panthor_heap_pool_get() - Get a heap pool reference
> + * @pool: Pool to get the reference on. Can be NULL.
> + *
> + * Return: @pool.
> + */
> +struct panthor_heap_pool *
> +panthor_heap_pool_get(struct panthor_heap_pool *pool)
> +{
> + if (pool)
> + kref_get(&pool->refcount);
> +
> + return pool;
> +}
> +
> +/**
> + * panthor_heap_pool_create() - Create a heap pool
> + * @ptdev: Device.
> + * @vm: The VM this heap pool will be attached to.
> + *
> + * Heap pools might contain up to 128 heap contexts, and are per-VM.
> + *
> + * Return: A valid pointer on success, a negative error code otherwise.
> + */
> +struct panthor_heap_pool *
> +panthor_heap_pool_create(struct panthor_device *ptdev, struct panthor_vm *vm)
> +{
> + size_t bosize = ALIGN(MAX_HEAPS_PER_POOL *
> + panthor_heap_ctx_stride(ptdev),
> + 4096);
> + struct panthor_heap_pool *pool;
> + int ret = 0;
> +
> + pool = kzalloc(sizeof(*pool), GFP_KERNEL);
> + if (!pool)
> + return ERR_PTR(-ENOMEM);
> +
> + /* We want a weak ref here: the heap pool belongs to the VM, so we're
> + * sure that, as long as the heap pool exists, the VM exists too.
> + */
> + pool->vm = vm;
> + pool->ptdev = ptdev;
> + init_rwsem(&pool->lock);
> + xa_init_flags(&pool->xa, XA_FLAGS_ALLOC1);
> + kref_init(&pool->refcount);
> +
> + pool->gpu_contexts = panthor_kernel_bo_create(ptdev, vm, bosize,
> + DRM_PANTHOR_BO_NO_MMAP,
> + DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC,
> + PANTHOR_VM_KERNEL_AUTO_VA);
> + if (IS_ERR(pool->gpu_contexts)) {
> + ret = PTR_ERR(pool->gpu_contexts);
> + goto err_destroy_pool;
> + }
> +
> + ret = panthor_kernel_bo_vmap(pool->gpu_contexts);
> + if (ret)
> + goto err_destroy_pool;
> +
> + return pool;
> +
> +err_destroy_pool:
> + panthor_heap_pool_destroy(pool);
> + return ERR_PTR(ret);
> +}
> +
> +/**
> + * panthor_heap_pool_destroy() - Destroy a heap pool.
> + * @pool: Pool to destroy.
> + *
> + * This function destroys all heap contexts and their resources. Thus
> + * preventing any use of the heap context or the chunk attached to them
> + * after that point.
> + *
> + * If the GPU still has access to some heap contexts, a fault should be
> + * triggered, which should flag the command stream groups using these
> + * context as faulty.
> + *
> + * The heap pool object is only released when all references to this pool
> + * are released.
> + */
> +void panthor_heap_pool_destroy(struct panthor_heap_pool *pool)
> +{
> + struct panthor_heap *heap;
> + unsigned long i;
> +
> + if (!pool)
> + return;
> +
> + down_write(&pool->lock);
> + xa_for_each(&pool->xa, i, heap)
> + drm_WARN_ON(&pool->ptdev->base, panthor_heap_destroy_locked(pool, i));
> +
> + if (!IS_ERR_OR_NULL(pool->gpu_contexts))
> + panthor_kernel_bo_destroy(pool->vm, pool->gpu_contexts);
> +
> + /* Reflects the fact the pool has been destroyed. */
> + pool->vm = NULL;
> + up_write(&pool->lock);
> +
> + panthor_heap_pool_put(pool);
> +}
> diff --git a/drivers/gpu/drm/panthor/panthor_heap.h b/drivers/gpu/drm/panthor/panthor_heap.h
> new file mode 100644
> index 000000000000..25a5f2bba445
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_heap.h
> @@ -0,0 +1,39 @@
> +/* SPDX-License-Identifier: GPL-2.0 or MIT */
> +/* Copyright 2023 Collabora ltd. */
> +
> +#ifndef __PANTHOR_HEAP_H__
> +#define __PANTHOR_HEAP_H__
> +
> +#include <linux/types.h>
> +
> +struct panthor_device;
> +struct panthor_heap_pool;
> +struct panthor_vm;
> +
> +int panthor_heap_create(struct panthor_heap_pool *pool,
> + u32 initial_chunk_count,
> + u32 chunk_size,
> + u32 max_chunks,
> + u32 target_in_flight,
> + u64 *heap_ctx_gpu_va,
> + u64 *first_chunk_gpu_va);
> +int panthor_heap_destroy(struct panthor_heap_pool *pool, u32 handle);
> +
> +struct panthor_heap_pool *
> +panthor_heap_pool_create(struct panthor_device *ptdev, struct panthor_vm *vm);
> +void panthor_heap_pool_destroy(struct panthor_heap_pool *pool);
> +
> +struct panthor_heap_pool *
> +panthor_heap_pool_get(struct panthor_heap_pool *pool);
> +void panthor_heap_pool_put(struct panthor_heap_pool *pool);
> +
> +int panthor_heap_grow(struct panthor_heap_pool *pool,
> + u64 heap_gpu_va,
> + u32 renderpasses_in_flight,
> + u32 pending_frag_count,
> + u64 *new_chunk_gpu_va);
> +int panthor_heap_return_chunk(struct panthor_heap_pool *pool,
> + u64 heap_gpu_va,
> + u64 chunk_gpu_va);
> +
> +#endif
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 09/14] drm/panthor: Add the heap logical block
2024-02-12 11:40 ` Steven Price
@ 2024-02-14 17:33 ` Boris Brezillon
2024-02-15 9:34 ` Steven Price
0 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-02-14 17:33 UTC (permalink / raw)
To: Steven Price
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen
On Mon, 12 Feb 2024 11:40:55 +0000
Steven Price <steven.price@arm.com> wrote:
> On 22/01/2024 16:30, Boris Brezillon wrote:
> > Tiler heap growing requires some kernel driver involvement: when the
> > tiler runs out of heap memory, it will raise an exception which is
> > either directly handled by the firmware if some free heap chunks are
> > available in the heap context, or passed back to the kernel otherwise.
> > The heap helpers will be used by the scheduler logic to allocate more
> > heap chunks to a heap context, when such a situation happens.
> >
> > Heap context creation is explicitly requested by userspace (using
> > the TILER_HEAP_CREATE ioctl), and the returned context is attached to a
> > queue through some command stream instruction.
> >
> > All the kernel does is keep the list of heap chunks allocated to a
> > context, so they can be freed when TILER_HEAP_DESTROY is called, or
> > extended when the FW requests a new chunk.
> >
> > v4:
> > - Rework locking to allow concurrent calls to panthor_heap_grow()
> > - Add a helper to return a heap chunk if we couldn't pass it to the
> > FW because the group was scheduled out
> >
> > v3:
> > - Add a FIXME for the heap OOM deadlock
> > - Use the panthor_kernel_bo abstraction for the heap context and heap
> > chunks
> > - Drop the panthor_heap_gpu_ctx struct as it is opaque to the driver
> > - Ensure that the heap context is aligned to the GPU cache line size
> > - Minor code tidy ups
> >
> > Co-developed-by: Steven Price <steven.price@arm.com>
> > Signed-off-by: Steven Price <steven.price@arm.com>
> > Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
>
> It looks fine, but there's a confusing FIXME comment:
>
> > + /* FIXME: panthor_alloc_heap_chunk() triggers a kernel BO creation, which
> > + * relies on blocking allocations (both for the BO itself, and backing
> > + * memory), which might cause a deadlock because we're called from a context
> > + * where we hold the panthor scheduler lock, thus preventing job cleanups
> > + * that could free up some memory. The jobs themselves will timeout, but
> > + * we'll still be blocked there. The only solution here is to implement
> > + * something similar to shmem_sg_alloc_table() in i915, so we can do
> > + * non-blocking allocations, and just kill the job when we run out-of-memory
> > + * for the tiler context.
> > + */
>
> Whereas at the call site (group_process_tiler_oom()) there's the comment:
>
> > /* We do the allocation without holding the scheduler lock to avoid
> > * blocking the scheduling.
> > */
>
> AFAICT that FIXME comment can just be deleted now. Assuming you agree
> then with that change:
The FIXME is no longer accurate indeed, but I'd like to keep a FIXME
here to reflect the fact the solution we have is not the one we want, as
it prevents the GPU from immediately falling back to the user provided
OOM exception handler, or killing the job if there's no way it can
reclaim tiler memory.
How about:
FIXME: panthor_alloc_heap_chunk() triggers a kernel BO creation, which
goes through the blocking allocation path. Ultimately, we want
a non-blocking allocation, so we can immediately report to the FW when
the system is running out of memory. In that case, the FW can call a
user-provided exception handler, which might try to free some tiler
memory by issuing an intermediate fragment job. If the exception handler
can't do anything, it will flag the queue as faulty so the job that
triggered this tiler chunk allocation and all further jobs in this
queue fail immediately instead of having to wait for the job
timeout.
>
> Reviewed-by: Steven Price <steven.price@arm.com>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 09/14] drm/panthor: Add the heap logical block
2024-02-14 17:33 ` Boris Brezillon
@ 2024-02-15 9:34 ` Steven Price
0 siblings, 0 replies; 51+ messages in thread
From: Steven Price @ 2024-02-15 9:34 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen
On 14/02/2024 17:33, Boris Brezillon wrote:
> On Mon, 12 Feb 2024 11:40:55 +0000
> Steven Price <steven.price@arm.com> wrote:
>
>> On 22/01/2024 16:30, Boris Brezillon wrote:
>>> Tiler heap growing requires some kernel driver involvement: when the
>>> tiler runs out of heap memory, it will raise an exception which is
>>> either directly handled by the firmware if some free heap chunks are
>>> available in the heap context, or passed back to the kernel otherwise.
>>> The heap helpers will be used by the scheduler logic to allocate more
>>> heap chunks to a heap context, when such a situation happens.
>>>
>>> Heap context creation is explicitly requested by userspace (using
>>> the TILER_HEAP_CREATE ioctl), and the returned context is attached to a
>>> queue through some command stream instruction.
>>>
>>> All the kernel does is keep the list of heap chunks allocated to a
>>> context, so they can be freed when TILER_HEAP_DESTROY is called, or
>>> extended when the FW requests a new chunk.
>>>
>>> v4:
>>> - Rework locking to allow concurrent calls to panthor_heap_grow()
>>> - Add a helper to return a heap chunk if we couldn't pass it to the
>>> FW because the group was scheduled out
>>>
>>> v3:
>>> - Add a FIXME for the heap OOM deadlock
>>> - Use the panthor_kernel_bo abstraction for the heap context and heap
>>> chunks
>>> - Drop the panthor_heap_gpu_ctx struct as it is opaque to the driver
>>> - Ensure that the heap context is aligned to the GPU cache line size
>>> - Minor code tidy ups
>>>
>>> Co-developed-by: Steven Price <steven.price@arm.com>
>>> Signed-off-by: Steven Price <steven.price@arm.com>
>>> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
>>
>> It looks fine, but there's a confusing FIXME comment:
>>
>>> + /* FIXME: panthor_alloc_heap_chunk() triggers a kernel BO creation, which
>>> + * relies on blocking allocations (both for the BO itself, and backing
>>> + * memory), which might cause a deadlock because we're called from a context
>>> + * where we hold the panthor scheduler lock, thus preventing job cleanups
>>> + * that could free up some memory. The jobs themselves will timeout, but
>>> + * we'll still be blocked there. The only solution here is to implement
>>> + * something similar to shmem_sg_alloc_table() in i915, so we can do
>>> + * non-blocking allocations, and just kill the job when we run out-of-memory
>>> + * for the tiler context.
>>> + */
>>
>> Whereas at the call site (group_process_tiler_oom()) there's the comment:
>>
>>> /* We do the allocation without holding the scheduler lock to avoid
>>> * blocking the scheduling.
>>> */
>>
>> AFAICT that FIXME comment can just be deleted now. Assuming you agree
>> then with that change:
>
> The FIXME is no longer accurate indeed, but I'd like to keep a FIXME
> here to reflect the fact the solution we have is not the one we want, as
> it prevents the GPU from immediately falling back to the user provided
> OOM exception handler, or killing the job if there's no way it can
> reclaim tiler memory.
>
> How about:
>
> FIXME: panthor_alloc_heap_chunk() triggers a kernel BO creation, which
> goes through the blocking allocation path. Ultimately, we want
> a non-blocking allocation, so we can immediately report to the FW when
> the system is running out of memory. In that case, the FW can call a
> user-provided exception handler, which might try to free some tiler
> memory by issuing an intermediate fragment job. If the exception handler
> can't do anything, it will flag the queue as faulty so the job that
> triggered this tiler chunk allocation and all further jobs in this
> queue fail immediately instead of having to wait for the job
> timeout.
Sounds good - it was mostly the talk of deadlock that worried me. The
code as it stands is good enough for merging (as it's not going to
deadlock the kernel) but we can obviously look to improve the low-memory
behaviour in future.
Steve
>>
>> Reviewed-by: Steven Price <steven.price@arm.com>
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 10/14] drm/panthor: Add the scheduler logical block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (8 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 09/14] drm/panthor: Add the heap " Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-02-14 16:48 ` Steven Price
2024-01-22 16:30 ` [PATCH v4 11/14] drm/panthor: Add the driver frontend block Boris Brezillon
` (7 subsequent siblings)
17 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
This is the piece of software interacting with the FW scheduler, and
taking care of some scheduling aspects when the FW comes short of slots
scheduling slots. Indeed, the FW only expose a few slots, and the kernel
has to give all submission contexts, a chance to execute their jobs.
The kernel-side scheduler is timeslice-based, with a round-robin queue
per priority level.
Job submission is handled with a 1:1 drm_sched_entity:drm_gpu_scheduler,
allowing us to delegate the dependency tracking to the core.
All the gory details should be documented inline.
v4:
- Check drmm_mutex_init() return code
- s/drm_gem_vmap_unlocked/drm_gem_vunmap_unlocked/ in
panthor_queue_put_syncwait_obj()
- Drop unneeded WARN_ON() in cs_slot_sync_queue_state_locked()
- Use atomic_xchg() instead of atomic_fetch_and(0)
- Fix typos
- Let panthor_kernel_bo_destroy() check for IS_ERR_OR_NULL() BOs
- Defer TILER_OOM event handling to a separate workqueue to prevent
deadlocks when the heap chunk allocation is blocked on mem-reclaim.
This is just a temporary solution, until we add support for
non-blocking/failable allocations
- Pass the scheduler workqueue to drm_sched instead of instantiating
a separate one (no longer needed now that heap chunk allocation
happens on a dedicated wq)
- Set WQ_MEM_RECLAIM on the scheduler workqueue, so we can handle
job timeouts when the system is under mem pressure, and hopefully
free up some memory retained by these jobs
v3:
- Rework the FW event handling logic to avoid races
- Make sure MMU faults kill the group immediately
- Use the panthor_kernel_bo abstraction for group/queue buffers
- Make in_progress an atomic_t, so we can check it without the reset lock
held
- Don't limit the number of groups per context to the FW scheduler
capacity. Fix the limit to 128 for now.
- Add a panthor_job_vm() helper
- Account for panthor_vm changes
- Add our job fence as DMA_RESV_USAGE_WRITE to all external objects
(was previously DMA_RESV_USAGE_BOOKKEEP). I don't get why, given
we're supposed to be fully-explicit, but other drivers do that, so
there must be a good reason
- Account for drm_sched changes
- Provide a panthor_queue_put_syncwait_obj()
- Unconditionally return groups to their idle list in
panthor_sched_suspend()
- Condition of sched_queue_{,delayed_}work fixed to be only when a reset
isn't pending or in progress.
- Several typos in comments fixed.
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
---
drivers/gpu/drm/panthor/panthor_sched.c | 3500 +++++++++++++++++++++++
drivers/gpu/drm/panthor/panthor_sched.h | 48 +
2 files changed, 3548 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_sched.c
create mode 100644 drivers/gpu/drm/panthor/panthor_sched.h
diff --git a/drivers/gpu/drm/panthor/panthor_sched.c b/drivers/gpu/drm/panthor/panthor_sched.c
new file mode 100644
index 000000000000..b57f4a5f5176
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_sched.c
@@ -0,0 +1,3500 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2023 Collabora ltd. */
+
+#include <drm/panthor_drm.h>
+#include <drm/drm_drv.h>
+#include <drm/drm_exec.h>
+#include <drm/drm_gem_shmem_helper.h>
+#include <drm/drm_managed.h>
+#include <drm/gpu_scheduler.h>
+
+#include <linux/build_bug.h>
+#include <linux/clk.h>
+#include <linux/delay.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/iopoll.h>
+#include <linux/iosys-map.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+#include <linux/dma-resv.h>
+
+#include "panthor_sched.h"
+#include "panthor_devfreq.h"
+#include "panthor_device.h"
+#include "panthor_gem.h"
+#include "panthor_heap.h"
+#include "panthor_regs.h"
+#include "panthor_gpu.h"
+#include "panthor_fw.h"
+#include "panthor_mmu.h"
+
+/**
+ * DOC: Scheduler
+ *
+ * Mali CSF hardware adopts a firmware-assisted scheduling model, where
+ * the firmware takes care of scheduling aspects, to some extend.
+ *
+ * The scheduling happens at the scheduling group level, each group
+ * contains 1 to N queues (N is FW/hardware dependent, and exposed
+ * through the firmware interface). Each queue is assigned a command
+ * stream ring buffer, which serves as a way to get jobs submitted to
+ * the GPU, among other things.
+ *
+ * The firmware can schedule a maximum of M groups (M is FW/hardware
+ * dependent, and exposed through the firmware interface). Passed
+ * this maximum number of groups, the kernel must take care of
+ * rotating the groups passed to the firmware so every group gets
+ * a chance to have his queues scheduled for execution.
+ *
+ * The current implementation only supports with kernel-mode queues.
+ * In other terms, userspace doesn't have access to the ring-buffer.
+ * Instead, userspace passes indirect command stream buffers that are
+ * called from the queue ring-buffer by the kernel using a pre-defined
+ * sequence of command stream instructions to ensure the userspace driver
+ * always gets consistent results (cache maintenance,
+ * synchronization, ...).
+ *
+ * We rely on the drm_gpu_scheduler framework to deal with job
+ * dependencies and submission. As any other driver dealing with a
+ * FW-scheduler, we use the 1:1 entity:scheduler mode, such that each
+ * entity has its own job scheduler. When a job is ready to be executed
+ * (all its dependencies are met), it is pushed to the appropriate
+ * queue ring-buffer, and the group is scheduled for execution if it
+ * wasn't already active.
+ *
+ * Kernel-side group scheduling is timeslice-based. When we have less
+ * groups than there are slots, the periodic tick is disabled and we
+ * just let the FW schedule the active groups. When there are more
+ * groups than slots, we let each group a chance to execute stuff for
+ * a given amount of time, and then re-evaluate and pick new groups
+ * to schedule. The group selection algorithm is based on
+ * priority+round-robin.
+ *
+ * Even though user-mode queues is out of the scope right now, the
+ * current design takes them into account by avoiding any guess on the
+ * group/queue state that would be based on information we wouldn't have
+ * if userspace was in charge of the ring-buffer. That's also one of the
+ * reason we don't do 'cooperative' scheduling (encoding FW group slot
+ * reservation as dma_fence that would be returned from the
+ * drm_gpu_scheduler::prepare_job() hook, and treating group rotation as
+ * a queue of waiters, ordered by job submission order). This approach
+ * would work for kernel-mode queues, but would make user-mode queues a
+ * lot more complicated to retrofit.
+ */
+
+#define JOB_TIMEOUT_MS 5000
+
+#define MIN_CS_PER_CSG 8
+
+#define MIN_CSGS 3
+#define MAX_CSG_PRIO 0xf
+
+struct panthor_group;
+
+/**
+ * struct panthor_csg_slot - Command stream group slot
+ *
+ * This represents a FW slot for a scheduling group.
+ */
+struct panthor_csg_slot {
+ /** @group: Scheduling group bound to this slot. */
+ struct panthor_group *group;
+
+ /** @priority: Group priority. */
+ u8 priority;
+
+ /**
+ * @idle: True if the group bound to this slot is idle.
+ *
+ * A group is idle when it has nothing waiting for execution on
+ * all its queues, or when queues are blocked waiting for something
+ * to happen (synchronization object).
+ */
+ bool idle;
+};
+
+/**
+ * enum panthor_csg_priority - Group priority
+ */
+enum panthor_csg_priority {
+ /** @PANTHOR_CSG_PRIORITY_LOW: Low priority group. */
+ PANTHOR_CSG_PRIORITY_LOW = 0,
+
+ /** @PANTHOR_CSG_PRIORITY_MEDIUM: Medium priority group. */
+ PANTHOR_CSG_PRIORITY_MEDIUM,
+
+ /** @PANTHOR_CSG_PRIORITY_HIGH: High priority group. */
+ PANTHOR_CSG_PRIORITY_HIGH,
+
+ /**
+ * @PANTHOR_CSG_PRIORITY_RT: Real-time priority group.
+ *
+ * Real-time priority allows one to preempt scheduling of other
+ * non-real-time groups. When such a group becomes executable,
+ * it will evict the group with the lowest non-rt priority if
+ * there's no free group slot available.
+ *
+ * Currently not exposed to userspace.
+ */
+ PANTHOR_CSG_PRIORITY_RT,
+
+ /** @PANTHOR_CSG_PRIORITY_COUNT: Number of priority levels. */
+ PANTHOR_CSG_PRIORITY_COUNT,
+};
+
+/**
+ * struct panthor_scheduler - Object used to manage the scheduler
+ */
+struct panthor_scheduler {
+ /** @ptdev: Device. */
+ struct panthor_device *ptdev;
+
+ /**
+ * @wq: Workqueue used by our internal scheduler logic and
+ * drm_gpu_scheduler.
+ *
+ * Used for the scheduler tick, group update or other kind of FW
+ * event processing that can't be handled in the threaded interrupt
+ * path. Also passed to the drm_gpu_scheduler instances embedded
+ * in panthor_queue.
+ */
+ struct workqueue_struct *wq;
+
+ /**
+ * @heap_alloc_wq: Workqueue used to schedule tiler_oom works.
+ *
+ * We have a queue dedicated to heap chunk allocation works to avoid
+ * blocking the rest of the scheduler if the allocation tries to
+ * reclaim memory.
+ */
+ struct workqueue_struct *heap_alloc_wq;
+
+ /** @tick_work: Work executed on a scheduling tick. */
+ struct delayed_work tick_work;
+
+ /**
+ * @sync_upd_work: Work used to process synchronization object updates.
+ *
+ * We use this work to unblock queues/groups that were waiting on a
+ * synchronization object.
+ */
+ struct work_struct sync_upd_work;
+
+ /**
+ * @fw_events_work: Work used to process FW events outside the interrupt path.
+ *
+ * Even if the interrupt is threaded, we need any event processing
+ * that require taking the panthor_scheduler::lock to be processed
+ * outside the interrupt path so we don't block the tick logic when
+ * it calls panthor_fw_{csg,wait}_wait_acks(). Since most of the
+ * event processing requires taking this lock, we just delegate all
+ * FW event processing to the scheduler workqueue.
+ */
+ struct work_struct fw_events_work;
+
+ /**
+ * @fw_events: Bitmask encoding pending FW events.
+ */
+ atomic_t fw_events;
+
+ /**
+ * @resched_target: When the next tick should occur.
+ *
+ * Expressed in jiffies.
+ */
+ u64 resched_target;
+
+ /**
+ * @last_tick: When the last tick occurred.
+ *
+ * Expressed in jiffies.
+ */
+ u64 last_tick;
+
+ /** @tick_period: Tick period in jiffies. */
+ u64 tick_period;
+
+ /**
+ * @lock: Lock protecting access to all the scheduler fields.
+ *
+ * Should be taken in the tick work, the irq handler, and anywhere the @groups
+ * fields are touched.
+ */
+ struct mutex lock;
+
+ /** @groups: Various lists used to classify groups. */
+ struct {
+ /**
+ * @runnable: Runnable group lists.
+ *
+ * When a group has queues that want to execute something,
+ * its panthor_group::run_node should be inserted here.
+ *
+ * One list per-priority.
+ */
+ struct list_head runnable[PANTHOR_CSG_PRIORITY_COUNT];
+
+ /**
+ * @idle: Idle group lists.
+ *
+ * When all queues of a group are idle (either because they
+ * have nothing to execute, or because they are blocked), the
+ * panthor_group::run_node field should be inserted here.
+ *
+ * One list per-priority.
+ */
+ struct list_head idle[PANTHOR_CSG_PRIORITY_COUNT];
+
+ /**
+ * @waiting: List of groups whose queues are blocked on a
+ * synchronization object.
+ *
+ * Insert panthor_group::wait_node here when a group is waiting
+ * for synchronization objects to be signaled.
+ *
+ * This list is evaluated in the @sync_upd_work work.
+ */
+ struct list_head waiting;
+ } groups;
+
+ /**
+ * @csg_slots: FW command stream group slots.
+ */
+ struct panthor_csg_slot csg_slots[MAX_CSGS];
+
+ /** @csg_slot_count: Number of command stream group slots exposed by the FW. */
+ u32 csg_slot_count;
+
+ /** @cs_slot_count: Number of command stream slot per group slot exposed by the FW. */
+ u32 cs_slot_count;
+
+ /** @as_slot_count: Number of address space slots supported by the MMU. */
+ u32 as_slot_count;
+
+ /** @used_csg_slot_count: Number of command stream group slot currently used. */
+ u32 used_csg_slot_count;
+
+ /** @sb_slot_count: Number of scoreboard slots. */
+ u32 sb_slot_count;
+
+ /**
+ * @might_have_idle_groups: True if an active group might have become idle.
+ *
+ * This will force a tick, so other runnable groups can be scheduled if one
+ * or more active groups became idle.
+ */
+ bool might_have_idle_groups;
+
+ /** @pm: Power management related fields. */
+ struct {
+ /** @has_ref: True if the scheduler owns a runtime PM reference. */
+ bool has_ref;
+ } pm;
+
+ /** @reset: Reset related fields. */
+ struct {
+ /** @lock: Lock protecting the other reset fields. */
+ struct mutex lock;
+
+ /**
+ * @in_progress: True if a reset is in progress.
+ *
+ * Set to true in panthor_sched_pre_reset() and back to false in
+ * panthor_sched_post_reset().
+ */
+ atomic_t in_progress;
+
+ /**
+ * @stopped_groups: List containing all groups that were stopped
+ * before a reset.
+ *
+ * Insert panthor_group::run_node in the pre_reset path.
+ */
+ struct list_head stopped_groups;
+ } reset;
+};
+
+/**
+ * struct panthor_syncobj_32b - 32-bit FW synchronization object
+ */
+struct panthor_syncobj_32b {
+ /** @seqno: Sequence number. */
+ u32 seqno;
+
+ /**
+ * @status: Status.
+ *
+ * Not zero on failure.
+ */
+ u32 status;
+};
+
+/**
+ * struct panthor_syncobj_64b - 64-bit FW synchronization object
+ */
+struct panthor_syncobj_64b {
+ /** @seqno: Sequence number. */
+ u64 seqno;
+
+ /**
+ * @status: Status.
+ *
+ * Not zero on failure.
+ */
+ u32 status;
+
+ /** @pad: MBZ. */
+ u32 pad;
+};
+
+/**
+ * struct panthor_queue - Execution queue
+ */
+struct panthor_queue {
+ /** @scheduler: DRM scheduler used for this queue. */
+ struct drm_gpu_scheduler scheduler;
+
+ /** @entity: DRM scheduling entity used for this queue. */
+ struct drm_sched_entity entity;
+
+ /**
+ * @remaining_time: Time remaining before the job timeout expires.
+ *
+ * The job timeout is suspended when the queue is not scheduled by the
+ * FW. Every time we suspend the timer, we need to save the remaining
+ * time so we can restore it later on.
+ */
+ unsigned long remaining_time;
+
+ /** @timeout_suspended: True if the job timeout was suspended. */
+ bool timeout_suspended;
+
+ /**
+ * @doorbell_id: Doorbell assigned to this queue.
+ *
+ * Right now, all groups share the same doorbell, and the doorbell ID
+ * is assigned to group_slot + 1 when the group is assigned a slot. But
+ * we might decide to provide fine grained doorbell assignment at some
+ * point, so don't have to wake up all queues in a group every time one
+ * of them is updated.
+ */
+ u8 doorbell_id;
+
+ /**
+ * @priority: Priority of the queue inside the group.
+ *
+ * Must be less than 16 (Only 4 bits available).
+ */
+ u8 priority;
+#define CSF_MAX_QUEUE_PRIO GENMASK(3, 0)
+
+ /** @ringbuf: Command stream ring-buffer. */
+ struct panthor_kernel_bo *ringbuf;
+
+ /** @iface: Firmware interface. */
+ struct {
+ /** @mem: FW memory allocated for this interface. */
+ struct panthor_kernel_bo *mem;
+
+ /** @input: Input interface. */
+ struct panthor_fw_ringbuf_input_iface *input;
+
+ /** @output: Output interface. */
+ const struct panthor_fw_ringbuf_output_iface *output;
+
+ /** @input_fw_va: FW virtual address of the input interface buffer. */
+ u32 input_fw_va;
+
+ /** @output_fw_va: FW virtual address of the output interface buffer. */
+ u32 output_fw_va;
+ } iface;
+
+ /**
+ * @syncwait: Stores information about the synchronization object this
+ * queue is waiting on.
+ */
+ struct {
+ /** @gpu_va: GPU address of the synchronization object. */
+ u64 gpu_va;
+
+ /** @ref: Reference value to compare against. */
+ u64 ref;
+
+ /** @gt: True if this is a greater-than test. */
+ bool gt;
+
+ /** @sync64: True if this is a 64-bit sync object. */
+ bool sync64;
+
+ /** @bo: Buffer object holding the synchronization object. */
+ struct drm_gem_object *obj;
+
+ /** @offset: Offset of the synchronization object inside @bo. */
+ u64 offset;
+
+ /**
+ * @kmap: Kernel mapping of the buffer object holding the
+ * synchronization object.
+ */
+ void *kmap;
+ } syncwait;
+
+ /** @fence_ctx: Fence context fields. */
+ struct {
+ /** @lock: Used to protect access to all fences allocated by this context. */
+ spinlock_t lock;
+
+ /**
+ * @id: Fence context ID.
+ *
+ * Allocated with dma_fence_context_alloc().
+ */
+ u64 id;
+
+ /** @seqno: Sequence number of the last initialized fence. */
+ atomic64_t seqno;
+
+ /**
+ * @in_flight_jobs: List containing all in-flight jobs.
+ *
+ * Used to keep track and signal panthor_job::done_fence when the
+ * synchronization object attached to the queue is signaled.
+ */
+ struct list_head in_flight_jobs;
+ } fence_ctx;
+};
+
+/**
+ * enum panthor_group_state - Scheduling group state.
+ */
+enum panthor_group_state {
+ /** @PANTHOR_CS_GROUP_CREATED: Group was created, but not scheduled yet. */
+ PANTHOR_CS_GROUP_CREATED,
+
+ /** @PANTHOR_CS_GROUP_ACTIVE: Group is currently scheduled. */
+ PANTHOR_CS_GROUP_ACTIVE,
+
+ /**
+ * @PANTHOR_CS_GROUP_SUSPENDED: Group was scheduled at least once, but is
+ * inactive/suspended right now.
+ */
+ PANTHOR_CS_GROUP_SUSPENDED,
+
+ /**
+ * @PANTHOR_CS_GROUP_TERMINATED: Group was terminated.
+ *
+ * Can no longer be scheduled. The only allowed action is a destruction.
+ */
+ PANTHOR_CS_GROUP_TERMINATED,
+};
+
+/**
+ * struct panthor_group - Scheduling group object
+ */
+struct panthor_group {
+ /** @refcount: Reference count */
+ struct kref refcount;
+
+ /** @ptdev: Device. */
+ struct panthor_device *ptdev;
+
+ /** @vm: VM bound to the group. */
+ struct panthor_vm *vm;
+
+ /** @compute_core_mask: Mask of shader cores that can be used for compute jobs. */
+ u64 compute_core_mask;
+
+ /** @fragment_core_mask: Mask of shader cores that can be used for fragment jobs. */
+ u64 fragment_core_mask;
+
+ /** @tiler_core_mask: Mask of tiler cores that can be used for tiler jobs. */
+ u64 tiler_core_mask;
+
+ /** @max_compute_cores: Maximum number of shader cores used for compute jobs. */
+ u8 max_compute_cores;
+
+ /** @max_compute_cores: Maximum number of shader cores used for fragment jobs. */
+ u8 max_fragment_cores;
+
+ /** @max_tiler_cores: Maximum number of tiler cores used for tiler jobs. */
+ u8 max_tiler_cores;
+
+ /** @priority: Group priority (check panthor_csg_priority). */
+ u8 priority;
+
+ /** @blocked_queues: Bitmask reflecting the blocked queues. */
+ u32 blocked_queues;
+
+ /** @idle_queues: Bitmask reflecting the blocked queues. */
+ u32 idle_queues;
+
+ /** @fatal_lock: Lock used to protect access to fatal fields. */
+ spinlock_t fatal_lock;
+
+ /** @fatal_queues: Bitmask reflecting the queues that hit a fatal exception. */
+ u32 fatal_queues;
+
+ /** @tiler_oom: Mask of queues that have a tiler OOM event to process. */
+ atomic_t tiler_oom;
+
+ /** @queue_count: Number of queues in this group. */
+ u32 queue_count;
+
+ /** @queues: Queues owned by this group. */
+ struct panthor_queue *queues[MAX_CS_PER_CSG];
+
+ /**
+ * @csg_id: ID of the FW group slot.
+ *
+ * -1 when the group is not scheduled/active.
+ */
+ int csg_id;
+
+ /**
+ * @destroyed: True when the group has been destroyed.
+ *
+ * If a group is destroyed it becomes useless: no further jobs can be submitted
+ * to its queues. We simply wait for all references to be dropped so we can
+ * release the group object.
+ */
+ bool destroyed;
+
+ /**
+ * @timedout: True when a timeout occurred on any of the queues owned by
+ * this group.
+ *
+ * Timeouts can be reported by drm_sched or by the FW. In any case, any
+ * timeout situation is unrecoverable, and the group becomes useless.
+ * We simply wait for all references to be dropped so we can release the
+ * group object.
+ */
+ bool timedout;
+
+ /**
+ * @syncobjs: Pool of per-queue synchronization objects.
+ *
+ * One sync object per queue. The position of the sync object is
+ * determined by the queue index.
+ */
+ struct panthor_kernel_bo *syncobjs;
+
+ /** @state: Group state. */
+ enum panthor_group_state state;
+
+ /**
+ * @suspend_buf: Suspend buffer.
+ *
+ * Stores the state of the group and its queues when a group is suspended.
+ * Used at resume time to restore the group in its previous state.
+ *
+ * The size of the suspend buffer is exposed through the FW interface.
+ */
+ struct panthor_kernel_bo *suspend_buf;
+
+ /**
+ * @protm_suspend_buf: Protection mode suspend buffer.
+ *
+ * Stores the state of the group and its queues when a group that's in
+ * protection mode is suspended.
+ *
+ * Used at resume time to restore the group in its previous state.
+ *
+ * The size of the protection mode suspend buffer is exposed through the
+ * FW interface.
+ */
+ struct panthor_kernel_bo *protm_suspend_buf;
+
+ /** @sync_upd_work: Work used to check/signal job fences. */
+ struct work_struct sync_upd_work;
+
+ /** @tiler_oom_work: Work used to process tiler OOM events happening on this group. */
+ struct work_struct tiler_oom_work;
+
+ /** @term_work: Work used to finish the group termination procedure. */
+ struct work_struct term_work;
+
+ /**
+ * @release_work: Work used to release group resources.
+ *
+ * We need to postpone the group release to avoid a deadlock when
+ * the last ref is released in the tick work.
+ */
+ struct work_struct release_work;
+
+ /**
+ * @run_node: Node used to insert the group in the
+ * panthor_group::groups::{runnable,idle} and
+ * panthor_group::reset.stopped_groups lists.
+ */
+ struct list_head run_node;
+
+ /**
+ * @wait_node: Node used to insert the group in the
+ * panthor_group::groups::waiting list.
+ */
+ struct list_head wait_node;
+};
+
+/**
+ * group_queue_work() - Queue a group work
+ * @group: Group to queue the work for.
+ * @wname: Work name.
+ *
+ * Grabs a ref and queue a work item to the scheduler workqueue. If
+ * the work was already queued, we release the reference we grabbed.
+ *
+ * Work callbacks must release the reference we grabbed here.
+ */
+#define group_queue_work(group, wname) \
+ do { \
+ group_get(group); \
+ if (!queue_work((group)->ptdev->scheduler->wq, &(group)->wname ## _work)) \
+ group_put(group); \
+ } while (0)
+
+/**
+ * sched_queue_work() - Queue a scheduler work.
+ * @sched: Scheduler object.
+ * @wname: Work name.
+ *
+ * Conditionally queues a scheduler work if no reset is pending/in-progress.
+ */
+#define sched_queue_work(sched, wname) \
+ do { \
+ if (!atomic_read(&sched->reset.in_progress) && \
+ !panthor_device_reset_is_pending((sched)->ptdev)) \
+ queue_work((sched)->wq, &(sched)->wname ## _work); \
+ } while (0)
+
+/**
+ * sched_queue_delayed_work() - Queue a scheduler delayed work.
+ * @sched: Scheduler object.
+ * @wname: Work name.
+ * @delay: Work delay in jiffies.
+ *
+ * Conditionally queues a scheduler delayed work if no reset is
+ * pending/in-progress.
+ */
+#define sched_queue_delayed_work(sched, wname, delay) \
+ do { \
+ if (!atomic_read(&sched->reset.in_progress) && \
+ !panthor_device_reset_is_pending((sched)->ptdev)) \
+ mod_delayed_work((sched)->wq, &(sched)->wname ## _work, delay); \
+ } while (0)
+
+/*
+ * We currently set the maximum of groups per file to an arbitrary low value.
+ * But this can be updated if we need more.
+ */
+#define MAX_GROUPS_PER_POOL 128
+
+/**
+ * struct panthor_group_pool - Group pool
+ *
+ * Each file get assigned a group pool.
+ */
+struct panthor_group_pool {
+ /** @xa: Xarray used to manage group handles. */
+ struct xarray xa;
+};
+
+/**
+ * struct panthor_job - Used to manage GPU job
+ */
+struct panthor_job {
+ /** @base: Inherit from drm_sched_job. */
+ struct drm_sched_job base;
+
+ /** @refcount: Reference count. */
+ struct kref refcount;
+
+ /** @group: Group of the queue this job will be pushed to. */
+ struct panthor_group *group;
+
+ /** @queue_idx: Index of the queue inside @group. */
+ u32 queue_idx;
+
+ /** @call_info: Information about the userspace command stream call. */
+ struct {
+ /** @start: GPU address of the userspace command stream. */
+ u64 start;
+
+ /** @size: Size of the userspace command stream. */
+ u32 size;
+
+ /**
+ * @latest_flush: Flush ID at the time the userspace command
+ * stream was built.
+ *
+ * Needed for the flush reduction mechanism.
+ */
+ u32 latest_flush;
+ } call_info;
+
+ /** @ringbuf: Position of this job is in the ring buffer. */
+ struct {
+ /** @start: Start offset. */
+ u64 start;
+
+ /** @end: End offset. */
+ u64 end;
+ } ringbuf;
+
+ /**
+ * @node: Used to insert the job in the panthor_queue::fence_ctx::in_flight_jobs
+ * list.
+ */
+ struct list_head node;
+
+ /** @done_fence: Fence signaled when the job is finished or cancelled. */
+ struct dma_fence *done_fence;
+};
+
+static void
+panthor_queue_put_syncwait_obj(struct panthor_queue *queue)
+{
+ if (queue->syncwait.kmap) {
+ struct iosys_map map = IOSYS_MAP_INIT_VADDR(queue->syncwait.kmap);
+
+ drm_gem_vunmap_unlocked(queue->syncwait.obj, &map);
+ queue->syncwait.kmap = NULL;
+ }
+
+ drm_gem_object_put(queue->syncwait.obj);
+ queue->syncwait.obj = NULL;
+}
+
+static void *
+panthor_queue_get_syncwait_obj(struct panthor_group *group, struct panthor_queue *queue)
+{
+ struct panthor_device *ptdev = group->ptdev;
+ struct panthor_gem_object *bo;
+ struct iosys_map map;
+ int ret;
+
+ if (queue->syncwait.kmap)
+ return queue->syncwait.kmap + queue->syncwait.offset;
+
+ bo = panthor_vm_get_bo_for_va(group->vm,
+ queue->syncwait.gpu_va,
+ &queue->syncwait.offset);
+ if (drm_WARN_ON(&ptdev->base, IS_ERR_OR_NULL(bo)))
+ goto err_put_syncwait_obj;
+
+ queue->syncwait.obj = &bo->base.base;
+ ret = drm_gem_vmap_unlocked(queue->syncwait.obj, &map);
+ if (drm_WARN_ON(&ptdev->base, ret))
+ goto err_put_syncwait_obj;
+
+ queue->syncwait.kmap = map.vaddr;
+ if (drm_WARN_ON(&ptdev->base, !queue->syncwait.kmap))
+ goto err_put_syncwait_obj;
+
+ return queue->syncwait.kmap + queue->syncwait.offset;
+
+err_put_syncwait_obj:
+ panthor_queue_put_syncwait_obj(queue);
+ return NULL;
+}
+
+static void group_free_queue(struct panthor_group *group, struct panthor_queue *queue)
+{
+ if (IS_ERR_OR_NULL(queue))
+ return;
+
+ if (queue->entity.fence_context)
+ drm_sched_entity_destroy(&queue->entity);
+
+ if (queue->scheduler.ops)
+ drm_sched_fini(&queue->scheduler);
+
+ panthor_queue_put_syncwait_obj(queue);
+
+ panthor_kernel_bo_destroy(group->vm, queue->ringbuf);
+ panthor_kernel_bo_destroy(panthor_fw_vm(group->ptdev), queue->iface.mem);
+
+ kfree(queue);
+}
+
+static void group_release_work(struct work_struct *work)
+{
+ struct panthor_group *group = container_of(work,
+ struct panthor_group,
+ release_work);
+ struct panthor_device *ptdev = group->ptdev;
+ u32 i;
+
+ for (i = 0; i < group->queue_count; i++)
+ group_free_queue(group, group->queues[i]);
+
+ panthor_kernel_bo_destroy(panthor_fw_vm(ptdev), group->suspend_buf);
+ panthor_kernel_bo_destroy(panthor_fw_vm(ptdev), group->protm_suspend_buf);
+
+ if (!IS_ERR_OR_NULL(group->syncobjs))
+ panthor_kernel_bo_destroy(group->vm, group->syncobjs);
+
+ panthor_vm_put(group->vm);
+ kfree(group);
+}
+
+static void group_release(struct kref *kref)
+{
+ struct panthor_group *group = container_of(kref,
+ struct panthor_group,
+ refcount);
+ struct panthor_device *ptdev = group->ptdev;
+
+ drm_WARN_ON(&ptdev->base, group->csg_id >= 0);
+ drm_WARN_ON(&ptdev->base, !list_empty(&group->run_node));
+ drm_WARN_ON(&ptdev->base, !list_empty(&group->wait_node));
+
+ queue_work(panthor_cleanup_wq, &group->release_work);
+}
+
+static void group_put(struct panthor_group *group)
+{
+ if (group)
+ kref_put(&group->refcount, group_release);
+}
+
+static struct panthor_group *
+group_get(struct panthor_group *group)
+{
+ if (group)
+ kref_get(&group->refcount);
+
+ return group;
+}
+
+/**
+ * group_bind_locked() - Bind a group to a group slot
+ * @group: Group.
+ * @csg_id: Slot.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+group_bind_locked(struct panthor_group *group, u32 csg_id)
+{
+ struct panthor_device *ptdev = group->ptdev;
+ struct panthor_csg_slot *csg_slot;
+ int ret;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ if (drm_WARN_ON(&ptdev->base, group->csg_id != -1 || csg_id >= MAX_CSGS ||
+ ptdev->scheduler->csg_slots[csg_id].group))
+ return -EINVAL;
+
+ ret = panthor_vm_active(group->vm);
+ if (ret)
+ return ret;
+
+ csg_slot = &ptdev->scheduler->csg_slots[csg_id];
+ group_get(group);
+ group->csg_id = csg_id;
+
+ /* Dummy doorbell allocation: doorbell is assigned to the group and
+ * all queues use the same doorbell.
+ *
+ * TODO: Implement LRU-based doorbell assignment, so the most often
+ * updated queues get their own doorbell, thus avoiding useless checks
+ * on queues belonging to the same group that are rarely updated.
+ */
+ for (u32 i = 0; i < group->queue_count; i++)
+ group->queues[i]->doorbell_id = csg_id + 1;
+
+ csg_slot->group = group;
+
+ return 0;
+}
+
+/**
+ * group_unbind_locked() - Unbind a group from a slot.
+ * @group: Group to unbind.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+group_unbind_locked(struct panthor_group *group)
+{
+ struct panthor_device *ptdev = group->ptdev;
+ struct panthor_csg_slot *slot;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ if (drm_WARN_ON(&ptdev->base, group->csg_id < 0 || group->csg_id >= MAX_CSGS))
+ return -EINVAL;
+
+ if (drm_WARN_ON(&ptdev->base, group->state == PANTHOR_CS_GROUP_ACTIVE))
+ return -EINVAL;
+
+ slot = &ptdev->scheduler->csg_slots[group->csg_id];
+ panthor_vm_idle(group->vm);
+ group->csg_id = -1;
+
+ /* Tiler OOM events will be re-issued next time the group is scheduled. */
+ atomic_set(&group->tiler_oom, 0);
+ cancel_work(&group->tiler_oom_work);
+
+ for (u32 i = 0; i < group->queue_count; i++)
+ group->queues[i]->doorbell_id = -1;
+
+ slot->group = NULL;
+
+ group_put(group);
+ return 0;
+}
+
+/**
+ * cs_slot_prog_locked() - Program a queue slot
+ * @ptdev: Device.
+ * @csg_id: Group slot ID.
+ * @cs_id: Queue slot ID.
+ *
+ * Program a queue slot with the queue information so things can start being
+ * executed on this queue.
+ *
+ * The group slot must have a group bound to it already (group_bind_locked()).
+ */
+static void
+cs_slot_prog_locked(struct panthor_device *ptdev, u32 csg_id, u32 cs_id)
+{
+ struct panthor_queue *queue = ptdev->scheduler->csg_slots[csg_id].group->queues[cs_id];
+ struct panthor_fw_cs_iface *cs_iface = panthor_fw_get_cs_iface(ptdev, csg_id, cs_id);
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ queue->iface.input->extract = queue->iface.output->extract;
+ drm_WARN_ON(&ptdev->base, queue->iface.input->insert < queue->iface.input->extract);
+
+ cs_iface->input->ringbuf_base = panthor_kernel_bo_gpuva(queue->ringbuf);
+ cs_iface->input->ringbuf_size = panthor_kernel_bo_size(queue->ringbuf);
+ cs_iface->input->ringbuf_input = queue->iface.input_fw_va;
+ cs_iface->input->ringbuf_output = queue->iface.output_fw_va;
+ cs_iface->input->config = CS_CONFIG_PRIORITY(queue->priority) |
+ CS_CONFIG_DOORBELL(queue->doorbell_id);
+ cs_iface->input->ack_irq_mask = ~0;
+ panthor_fw_update_reqs(cs_iface, req,
+ CS_IDLE_SYNC_WAIT |
+ CS_IDLE_EMPTY |
+ CS_STATE_START |
+ CS_EXTRACT_EVENT,
+ CS_IDLE_SYNC_WAIT |
+ CS_IDLE_EMPTY |
+ CS_STATE_MASK |
+ CS_EXTRACT_EVENT);
+ if (queue->iface.input->insert != queue->iface.input->extract && queue->timeout_suspended) {
+ drm_sched_resume_timeout(&queue->scheduler, queue->remaining_time);
+ queue->timeout_suspended = false;
+ }
+}
+
+/**
+ * @cs_slot_reset_locked() - Reset a queue slot
+ * @ptdev: Device.
+ * @csg_id: Group slot.
+ * @cs_id: Queue slot.
+ *
+ * Change the queue slot state to STOP and suspend the queue timeout if
+ * the queue is not blocked.
+ *
+ * The group slot must have a group bound to it (group_bind_locked()).
+ */
+static int
+cs_slot_reset_locked(struct panthor_device *ptdev, u32 csg_id, u32 cs_id)
+{
+ struct panthor_fw_cs_iface *cs_iface = panthor_fw_get_cs_iface(ptdev, csg_id, cs_id);
+ struct panthor_group *group = ptdev->scheduler->csg_slots[csg_id].group;
+ struct panthor_queue *queue = group->queues[cs_id];
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ panthor_fw_update_reqs(cs_iface, req,
+ CS_STATE_STOP,
+ CS_STATE_MASK);
+
+ /* If the queue is blocked, we want to keep the timeout running, so
+ * we can detect unbounded waits and kill the group when that happens.
+ */
+ if (!(group->blocked_queues & BIT(cs_id)) && !queue->timeout_suspended) {
+ queue->remaining_time = drm_sched_suspend_timeout(&queue->scheduler);
+ queue->timeout_suspended = true;
+ WARN_ON(queue->remaining_time > msecs_to_jiffies(JOB_TIMEOUT_MS));
+ }
+
+ return 0;
+}
+
+/**
+ * csg_slot_sync_priority_locked() - Synchronize the group slot priority
+ * @ptdev: Device.
+ * @csg_id: Group slot ID.
+ *
+ * Group slot priority update happens asynchronously. When we receive a
+ * %CSG_ENDPOINT_CONFIG, we know the update is effective, and can
+ * reflect it to our panthor_csg_slot object.
+ */
+static void
+csg_slot_sync_priority_locked(struct panthor_device *ptdev, u32 csg_id)
+{
+ struct panthor_csg_slot *csg_slot = &ptdev->scheduler->csg_slots[csg_id];
+ struct panthor_fw_csg_iface *csg_iface;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+ csg_slot->priority = (csg_iface->input->endpoint_req & CSG_EP_REQ_PRIORITY_MASK) >> 28;
+}
+
+/**
+ * cs_slot_sync_queue_state_locked() - Synchronize the queue slot priority
+ * @ptdev: Device.
+ * @csg_id: Group slot.
+ * @cs_id: Queue slot.
+ *
+ * Queue state is updated on group suspend or STATUS_UPDATE event.
+ */
+static void
+cs_slot_sync_queue_state_locked(struct panthor_device *ptdev, u32 csg_id, u32 cs_id)
+{
+ struct panthor_group *group = ptdev->scheduler->csg_slots[csg_id].group;
+ struct panthor_queue *queue = group->queues[cs_id];
+ struct panthor_fw_cs_iface *cs_iface =
+ panthor_fw_get_cs_iface(group->ptdev, csg_id, cs_id);
+
+ u32 status_wait_cond;
+
+ switch (cs_iface->output->status_blocked_reason) {
+ case CS_STATUS_BLOCKED_REASON_UNBLOCKED:
+ if (queue->iface.input->insert == queue->iface.output->extract &&
+ cs_iface->output->status_scoreboards == 0)
+ group->idle_queues |= BIT(cs_id);
+ break;
+
+ case CS_STATUS_BLOCKED_REASON_SYNC_WAIT:
+ list_move_tail(&group->wait_node, &group->ptdev->scheduler->groups.waiting);
+ group->blocked_queues |= BIT(cs_id);
+ queue->syncwait.gpu_va = cs_iface->output->status_wait_sync_ptr;
+ queue->syncwait.ref = cs_iface->output->status_wait_sync_value;
+ status_wait_cond = cs_iface->output->status_wait & CS_STATUS_WAIT_SYNC_COND_MASK;
+ queue->syncwait.gt = status_wait_cond == CS_STATUS_WAIT_SYNC_COND_GT;
+ if (cs_iface->output->status_wait & CS_STATUS_WAIT_SYNC_64B) {
+ u64 sync_val_hi = cs_iface->output->status_wait_sync_value_hi;
+
+ queue->syncwait.sync64 = true;
+ queue->syncwait.ref |= sync_val_hi << 32;
+ } else {
+ queue->syncwait.sync64 = false;
+ }
+ break;
+
+ default:
+ /* Other reasons are not blocking. Consider the queue as runnable
+ * in those cases.
+ */
+ break;
+ }
+}
+
+static void
+csg_slot_sync_queues_state_locked(struct panthor_device *ptdev, u32 csg_id)
+{
+ struct panthor_csg_slot *csg_slot = &ptdev->scheduler->csg_slots[csg_id];
+ struct panthor_group *group = csg_slot->group;
+ u32 i;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ group->idle_queues = 0;
+ group->blocked_queues = 0;
+
+ for (i = 0; i < group->queue_count; i++) {
+ if (group->queues[i])
+ cs_slot_sync_queue_state_locked(ptdev, csg_id, i);
+ }
+}
+
+static void
+csg_slot_sync_state_locked(struct panthor_device *ptdev, u32 csg_id)
+{
+ struct panthor_csg_slot *csg_slot = &ptdev->scheduler->csg_slots[csg_id];
+ struct panthor_fw_csg_iface *csg_iface;
+ struct panthor_group *group;
+ enum panthor_group_state new_state, old_state;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+ group = csg_slot->group;
+
+ if (!group)
+ return;
+
+ old_state = group->state;
+ switch (csg_iface->output->ack & CSG_STATE_MASK) {
+ case CSG_STATE_START:
+ case CSG_STATE_RESUME:
+ new_state = PANTHOR_CS_GROUP_ACTIVE;
+ break;
+ case CSG_STATE_TERMINATE:
+ new_state = PANTHOR_CS_GROUP_TERMINATED;
+ break;
+ case CSG_STATE_SUSPEND:
+ new_state = PANTHOR_CS_GROUP_SUSPENDED;
+ break;
+ }
+
+ if (old_state == new_state)
+ return;
+
+ if (new_state == PANTHOR_CS_GROUP_SUSPENDED)
+ csg_slot_sync_queues_state_locked(ptdev, csg_id);
+
+ if (old_state == PANTHOR_CS_GROUP_ACTIVE) {
+ u32 i;
+
+ /* Reset the queue slots so we start from a clean
+ * state when starting/resuming a new group on this
+ * CSG slot. No wait needed here, and no ringbell
+ * either, since the CS slot will only be re-used
+ * on the next CSG start operation.
+ */
+ for (i = 0; i < group->queue_count; i++) {
+ if (group->queues[i])
+ cs_slot_reset_locked(ptdev, csg_id, i);
+ }
+ }
+
+ group->state = new_state;
+}
+
+static int
+csg_slot_prog_locked(struct panthor_device *ptdev, u32 csg_id, u32 priority)
+{
+ struct panthor_fw_csg_iface *csg_iface;
+ struct panthor_csg_slot *csg_slot;
+ struct panthor_group *group;
+ u32 queue_mask = 0, i;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ if (priority > MAX_CSG_PRIO)
+ return -EINVAL;
+
+ if (drm_WARN_ON(&ptdev->base, csg_id >= MAX_CSGS))
+ return -EINVAL;
+
+ csg_slot = &ptdev->scheduler->csg_slots[csg_id];
+ group = csg_slot->group;
+ if (!group || group->state == PANTHOR_CS_GROUP_ACTIVE)
+ return 0;
+
+ csg_iface = panthor_fw_get_csg_iface(group->ptdev, csg_id);
+
+ for (i = 0; i < group->queue_count; i++) {
+ if (group->queues[i]) {
+ cs_slot_prog_locked(ptdev, csg_id, i);
+ queue_mask |= BIT(i);
+ }
+ }
+
+ csg_iface->input->allow_compute = group->compute_core_mask;
+ csg_iface->input->allow_fragment = group->fragment_core_mask;
+ csg_iface->input->allow_other = group->tiler_core_mask;
+ csg_iface->input->endpoint_req = CSG_EP_REQ_COMPUTE(group->max_compute_cores) |
+ CSG_EP_REQ_FRAGMENT(group->max_fragment_cores) |
+ CSG_EP_REQ_TILER(group->max_tiler_cores) |
+ CSG_EP_REQ_PRIORITY(priority);
+ csg_iface->input->config = panthor_vm_as(group->vm);
+
+ if (group->suspend_buf)
+ csg_iface->input->suspend_buf = panthor_kernel_bo_gpuva(group->suspend_buf);
+ else
+ csg_iface->input->suspend_buf = 0;
+
+ if (group->protm_suspend_buf) {
+ csg_iface->input->protm_suspend_buf =
+ panthor_kernel_bo_gpuva(group->protm_suspend_buf);
+ } else {
+ csg_iface->input->protm_suspend_buf = 0;
+ }
+
+ csg_iface->input->ack_irq_mask = ~0;
+ panthor_fw_toggle_reqs(csg_iface, doorbell_req, doorbell_ack, queue_mask);
+ return 0;
+}
+
+static void
+cs_slot_process_fatal_event_locked(struct panthor_device *ptdev,
+ u32 csg_id, u32 cs_id)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[csg_id];
+ struct panthor_group *group = csg_slot->group;
+ struct panthor_fw_cs_iface *cs_iface;
+ u32 fatal;
+ u64 info;
+
+ lockdep_assert_held(&sched->lock);
+
+ cs_iface = panthor_fw_get_cs_iface(ptdev, csg_id, cs_id);
+ fatal = cs_iface->output->fatal;
+ info = cs_iface->output->fatal_info;
+
+ if (group)
+ group->fatal_queues |= BIT(cs_id);
+
+ sched_queue_delayed_work(sched, tick, 0);
+ drm_warn(&ptdev->base,
+ "CSG slot %d CS slot: %d\n"
+ "CS_FATAL.EXCEPTION_TYPE: 0x%x (%s)\n"
+ "CS_FATAL.EXCEPTION_DATA: 0x%x\n"
+ "CS_FATAL_INFO.EXCEPTION_DATA: 0x%llx\n",
+ csg_id, cs_id,
+ (unsigned int)CS_EXCEPTION_TYPE(fatal),
+ panthor_exception_name(ptdev, CS_EXCEPTION_TYPE(fatal)),
+ (unsigned int)CS_EXCEPTION_DATA(fatal),
+ info);
+}
+
+static void
+cs_slot_process_fault_event_locked(struct panthor_device *ptdev,
+ u32 csg_id, u32 cs_id)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[csg_id];
+ struct panthor_group *group = csg_slot->group;
+ struct panthor_queue *queue = group && cs_id < group->queue_count ?
+ group->queues[cs_id] : NULL;
+ struct panthor_fw_cs_iface *cs_iface;
+ u32 fault;
+ u64 info;
+
+ lockdep_assert_held(&sched->lock);
+
+ cs_iface = panthor_fw_get_cs_iface(ptdev, csg_id, cs_id);
+ fault = cs_iface->output->fault;
+ info = cs_iface->output->fault_info;
+
+ if (queue && CS_EXCEPTION_TYPE(fault) == DRM_PANTHOR_EXCEPTION_CS_INHERIT_FAULT) {
+ u64 cs_extract = queue->iface.output->extract;
+ struct panthor_job *job;
+
+ spin_lock(&queue->fence_ctx.lock);
+ list_for_each_entry(job, &queue->fence_ctx.in_flight_jobs, node) {
+ if (cs_extract >= job->ringbuf.end)
+ continue;
+
+ if (cs_extract < job->ringbuf.start)
+ break;
+
+ dma_fence_set_error(job->done_fence, -EINVAL);
+ }
+ spin_unlock(&queue->fence_ctx.lock);
+ }
+
+ drm_warn(&ptdev->base,
+ "CSG slot %d CS slot: %d\n"
+ "CS_FAULT.EXCEPTION_TYPE: 0x%x (%s)\n"
+ "CS_FAULT.EXCEPTION_DATA: 0x%x\n"
+ "CS_FAULT_INFO.EXCEPTION_DATA: 0x%llx\n",
+ csg_id, cs_id,
+ (unsigned int)CS_EXCEPTION_TYPE(fault),
+ panthor_exception_name(ptdev, CS_EXCEPTION_TYPE(fault)),
+ (unsigned int)CS_EXCEPTION_DATA(fault),
+ info);
+}
+
+static int group_process_tiler_oom(struct panthor_group *group, u32 cs_id)
+{
+ struct panthor_device *ptdev = group->ptdev;
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ u32 renderpasses_in_flight, pending_frag_count;
+ struct panthor_heap_pool *heaps = NULL;
+ u64 heap_address, new_chunk_va = 0;
+ u32 vt_start, vt_end, frag_end;
+ int ret, csg_id;
+
+ mutex_lock(&sched->lock);
+ csg_id = group->csg_id;
+ if (csg_id >= 0) {
+ struct panthor_fw_cs_iface *cs_iface;
+
+ cs_iface = panthor_fw_get_cs_iface(ptdev, csg_id, cs_id);
+ heaps = panthor_vm_get_heap_pool(group->vm, false);
+ heap_address = cs_iface->output->heap_address;
+ vt_start = cs_iface->output->heap_vt_start;
+ vt_end = cs_iface->output->heap_vt_end;
+ frag_end = cs_iface->output->heap_frag_end;
+ renderpasses_in_flight = vt_start - frag_end;
+ pending_frag_count = vt_end - frag_end;
+ }
+ mutex_unlock(&sched->lock);
+
+ /* The group got scheduled out, we stop here. We will get a new tiler OOM event
+ * when it's scheduled again.
+ */
+ if (unlikely(csg_id < 0))
+ return 0;
+
+ if (!heaps || frag_end > vt_end || vt_end >= vt_start) {
+ ret = -EINVAL;
+ } else {
+ /* We do the allocation without holding the scheduler lock to avoid
+ * blocking the scheduling.
+ */
+ ret = panthor_heap_grow(heaps, heap_address,
+ renderpasses_in_flight,
+ pending_frag_count, &new_chunk_va);
+ }
+
+ if (ret && ret != -EBUSY) {
+ group->fatal_queues |= BIT(csg_id);
+ sched_queue_delayed_work(sched, tick, 0);
+ goto out_put_heap_pool;
+ }
+
+ mutex_lock(&sched->lock);
+ csg_id = group->csg_id;
+ if (csg_id >= 0) {
+ struct panthor_fw_csg_iface *csg_iface;
+ struct panthor_fw_cs_iface *cs_iface;
+
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+ cs_iface = panthor_fw_get_cs_iface(ptdev, csg_id, cs_id);
+
+ cs_iface->input->heap_start = new_chunk_va;
+ cs_iface->input->heap_end = new_chunk_va;
+ panthor_fw_update_reqs(cs_iface, req, cs_iface->output->ack, CS_TILER_OOM);
+ panthor_fw_toggle_reqs(csg_iface, doorbell_req, doorbell_ack, BIT(cs_id));
+ panthor_fw_ring_csg_doorbells(ptdev, BIT(csg_id));
+ }
+ mutex_unlock(&sched->lock);
+
+ /* We allocated a chunck, but couldn't link it to the heap
+ * context because the group was scheduled out while we were
+ * allocating memory. We need to return this chunk to the heap.
+ */
+ if (unlikely(csg_id < 0 && new_chunk_va))
+ panthor_heap_return_chunk(heaps, heap_address, new_chunk_va);
+
+ ret = 0;
+
+out_put_heap_pool:
+ panthor_heap_pool_put(heaps);
+ return ret;
+}
+
+static void group_tiler_oom_work(struct work_struct *work)
+{
+ struct panthor_group *group =
+ container_of(work, struct panthor_group, tiler_oom_work);
+ u32 tiler_oom = atomic_xchg(&group->tiler_oom, 0);
+
+ while (tiler_oom) {
+ u32 cs_id = ffs(tiler_oom) - 1;
+
+ group_process_tiler_oom(group, cs_id);
+ tiler_oom &= ~BIT(cs_id);
+ }
+
+ group_put(group);
+}
+
+static void
+cs_slot_process_tiler_oom_event_locked(struct panthor_device *ptdev,
+ u32 csg_id, u32 cs_id)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[csg_id];
+ struct panthor_group *group = csg_slot->group;
+
+ lockdep_assert_held(&sched->lock);
+
+ if (drm_WARN_ON(&ptdev->base, !group))
+ return;
+
+ atomic_or(BIT(cs_id), &group->tiler_oom);
+
+ /* We don't use group_queue_work() here because we want to queue the
+ * work item to the heap_alloc_wq.
+ */
+ group_get(group);
+ if (!queue_work(sched->heap_alloc_wq, &group->tiler_oom_work))
+ group_put(group);
+}
+
+static bool cs_slot_process_irq_locked(struct panthor_device *ptdev,
+ u32 csg_id, u32 cs_id)
+{
+ struct panthor_fw_cs_iface *cs_iface;
+ u32 req, ack, events;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ cs_iface = panthor_fw_get_cs_iface(ptdev, csg_id, cs_id);
+ req = cs_iface->input->req;
+ ack = cs_iface->output->ack;
+ events = (req ^ ack) & CS_EVT_MASK;
+
+ if (events & CS_FATAL)
+ cs_slot_process_fatal_event_locked(ptdev, csg_id, cs_id);
+
+ if (events & CS_FAULT)
+ cs_slot_process_fault_event_locked(ptdev, csg_id, cs_id);
+
+ if (events & CS_TILER_OOM)
+ cs_slot_process_tiler_oom_event_locked(ptdev, csg_id, cs_id);
+
+ /* We don't acknowledge the TILER_OOM event since its handling is
+ * deferred to a separate work.
+ */
+ panthor_fw_update_reqs(cs_iface, req, ack, CS_FATAL | CS_FAULT);
+
+ return (events & (CS_FAULT | CS_TILER_OOM)) != 0;
+}
+
+static void csg_slot_sync_idle_state_locked(struct panthor_device *ptdev, u32 csg_id)
+{
+ struct panthor_csg_slot *csg_slot = &ptdev->scheduler->csg_slots[csg_id];
+ struct panthor_fw_csg_iface *csg_iface;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+ csg_slot->idle = csg_iface->output->status_state & CSG_STATUS_STATE_IS_IDLE;
+}
+
+static void csg_slot_process_idle_event_locked(struct panthor_device *ptdev, u32 csg_id)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+
+ lockdep_assert_held(&sched->lock);
+
+ sched->might_have_idle_groups = true;
+
+ /* Schedule a tick so we can evict idle groups and schedule non-idle
+ * ones. This will also update runtime PM and devfreq busy/idle states,
+ * so the device can lower its frequency or get suspended.
+ */
+ sched_queue_delayed_work(sched, tick, 0);
+}
+
+static void csg_slot_sync_update_locked(struct panthor_device *ptdev,
+ u32 csg_id)
+{
+ struct panthor_csg_slot *csg_slot = &ptdev->scheduler->csg_slots[csg_id];
+ struct panthor_group *group = csg_slot->group;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ if (group)
+ group_queue_work(group, sync_upd);
+
+ sched_queue_work(ptdev->scheduler, sync_upd);
+}
+
+static void
+csg_slot_process_progress_timer_event_locked(struct panthor_device *ptdev, u32 csg_id)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[csg_id];
+ struct panthor_group *group = csg_slot->group;
+
+ lockdep_assert_held(&sched->lock);
+
+ drm_warn(&ptdev->base, "CSG slot %d progress timeout\n", csg_id);
+
+ group = csg_slot->group;
+ if (!drm_WARN_ON(&ptdev->base, !group))
+ group->timedout = true;
+
+ sched_queue_delayed_work(sched, tick, 0);
+}
+
+static void sched_process_csg_irq_locked(struct panthor_device *ptdev, u32 csg_id)
+{
+ u32 req, ack, cs_irq_req, cs_irq_ack, cs_irqs, csg_events;
+ struct panthor_fw_csg_iface *csg_iface;
+ u32 ring_cs_db_mask = 0;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ if (drm_WARN_ON(&ptdev->base, csg_id >= ptdev->scheduler->csg_slot_count))
+ return;
+
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+ req = READ_ONCE(csg_iface->input->req);
+ ack = READ_ONCE(csg_iface->output->ack);
+ cs_irq_req = READ_ONCE(csg_iface->output->cs_irq_req);
+ cs_irq_ack = READ_ONCE(csg_iface->input->cs_irq_ack);
+ csg_events = (req ^ ack) & CSG_EVT_MASK;
+
+ /* There may not be any pending CSG/CS interrupts to process */
+ if (req == ack && cs_irq_req == cs_irq_ack)
+ return;
+
+ /* Immediately set IRQ_ACK bits to be same as the IRQ_REQ bits before
+ * examining the CS_ACK & CS_REQ bits. This would ensure that Host
+ * doesn't miss an interrupt for the CS in the race scenario where
+ * whilst Host is servicing an interrupt for the CS, firmware sends
+ * another interrupt for that CS.
+ */
+ csg_iface->input->cs_irq_ack = cs_irq_req;
+
+ panthor_fw_update_reqs(csg_iface, req, ack,
+ CSG_SYNC_UPDATE |
+ CSG_IDLE |
+ CSG_PROGRESS_TIMER_EVENT);
+
+ if (csg_events & CSG_IDLE)
+ csg_slot_process_idle_event_locked(ptdev, csg_id);
+
+ if (csg_events & CSG_PROGRESS_TIMER_EVENT)
+ csg_slot_process_progress_timer_event_locked(ptdev, csg_id);
+
+ cs_irqs = cs_irq_req ^ cs_irq_ack;
+ while (cs_irqs) {
+ u32 cs_id = ffs(cs_irqs) - 1;
+
+ if (cs_slot_process_irq_locked(ptdev, csg_id, cs_id))
+ ring_cs_db_mask |= BIT(cs_id);
+
+ cs_irqs &= ~BIT(cs_id);
+ }
+
+ if (csg_events & CSG_SYNC_UPDATE)
+ csg_slot_sync_update_locked(ptdev, csg_id);
+
+ if (ring_cs_db_mask)
+ panthor_fw_toggle_reqs(csg_iface, doorbell_req, doorbell_ack, ring_cs_db_mask);
+
+ panthor_fw_ring_csg_doorbells(ptdev, BIT(csg_id));
+}
+
+static void sched_process_idle_event_locked(struct panthor_device *ptdev)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ /* Acknowledge the idle event and schedule a tick. */
+ panthor_fw_update_reqs(glb_iface, req, glb_iface->output->ack, GLB_IDLE);
+ sched_queue_delayed_work(ptdev->scheduler, tick, 0);
+}
+
+/**
+ * panthor_sched_process_global_irq() - Process the scheduling part of a global IRQ
+ * @ptdev: Device.
+ */
+static void sched_process_global_irq_locked(struct panthor_device *ptdev)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+ u32 req, ack, evts;
+
+ lockdep_assert_held(&ptdev->scheduler->lock);
+
+ req = READ_ONCE(glb_iface->input->req);
+ ack = READ_ONCE(glb_iface->output->ack);
+ evts = (req ^ ack) & GLB_EVT_MASK;
+
+ if (evts & GLB_IDLE)
+ sched_process_idle_event_locked(ptdev);
+}
+
+static void process_fw_events_work(struct work_struct *work)
+{
+ struct panthor_scheduler *sched = container_of(work, struct panthor_scheduler,
+ fw_events_work);
+ u32 events = atomic_xchg(&sched->fw_events, 0);
+ struct panthor_device *ptdev = sched->ptdev;
+
+ mutex_lock(&sched->lock);
+
+ if (events & JOB_INT_GLOBAL_IF) {
+ sched_process_global_irq_locked(ptdev);
+ events &= ~JOB_INT_GLOBAL_IF;
+ }
+
+ while (events) {
+ u32 csg_id = ffs(events) - 1;
+
+ sched_process_csg_irq_locked(ptdev, csg_id);
+ events &= ~BIT(csg_id);
+ }
+
+ mutex_unlock(&sched->lock);
+}
+
+/**
+ * panthor_sched_report_fw_events() - Report FW events to the scheduler.
+ */
+void panthor_sched_report_fw_events(struct panthor_device *ptdev, u32 events)
+{
+ if (!ptdev->scheduler)
+ return;
+
+ atomic_or(events, &ptdev->scheduler->fw_events);
+ sched_queue_work(ptdev->scheduler, fw_events);
+}
+
+static const char *fence_get_driver_name(struct dma_fence *fence)
+{
+ return "panthor";
+}
+
+static const char *queue_fence_get_timeline_name(struct dma_fence *fence)
+{
+ return "queue-fence";
+}
+
+static const struct dma_fence_ops panthor_queue_fence_ops = {
+ .get_driver_name = fence_get_driver_name,
+ .get_timeline_name = queue_fence_get_timeline_name,
+};
+
+/**
+ */
+struct panthor_csg_slots_upd_ctx {
+ u32 update_mask;
+ u32 timedout_mask;
+ struct {
+ u32 value;
+ u32 mask;
+ } requests[MAX_CSGS];
+};
+
+static void csgs_upd_ctx_init(struct panthor_csg_slots_upd_ctx *ctx)
+{
+ memset(ctx, 0, sizeof(*ctx));
+}
+
+static void csgs_upd_ctx_queue_reqs(struct panthor_device *ptdev,
+ struct panthor_csg_slots_upd_ctx *ctx,
+ u32 csg_id, u32 value, u32 mask)
+{
+ if (drm_WARN_ON(&ptdev->base, !mask) ||
+ drm_WARN_ON(&ptdev->base, csg_id >= ptdev->scheduler->csg_slot_count))
+ return;
+
+ ctx->requests[csg_id].value = (ctx->requests[csg_id].value & ~mask) | (value & mask);
+ ctx->requests[csg_id].mask |= mask;
+ ctx->update_mask |= BIT(csg_id);
+}
+
+static int csgs_upd_ctx_apply_locked(struct panthor_device *ptdev,
+ struct panthor_csg_slots_upd_ctx *ctx)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ u32 update_slots = ctx->update_mask;
+
+ lockdep_assert_held(&sched->lock);
+
+ if (!ctx->update_mask)
+ return 0;
+
+ while (update_slots) {
+ struct panthor_fw_csg_iface *csg_iface;
+ u32 csg_id = ffs(update_slots) - 1;
+
+ update_slots &= ~BIT(csg_id);
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+ panthor_fw_update_reqs(csg_iface, req,
+ ctx->requests[csg_id].value,
+ ctx->requests[csg_id].mask);
+ }
+
+ panthor_fw_ring_csg_doorbells(ptdev, ctx->update_mask);
+
+ update_slots = ctx->update_mask;
+ while (update_slots) {
+ struct panthor_fw_csg_iface *csg_iface;
+ u32 csg_id = ffs(update_slots) - 1;
+ u32 req_mask = ctx->requests[csg_id].mask, acked;
+ int ret;
+
+ update_slots &= ~BIT(csg_id);
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+
+ ret = panthor_fw_csg_wait_acks(ptdev, csg_id, req_mask, &acked, 100);
+
+ if (acked & CSG_ENDPOINT_CONFIG)
+ csg_slot_sync_priority_locked(ptdev, csg_id);
+
+ if (acked & CSG_STATE_MASK)
+ csg_slot_sync_state_locked(ptdev, csg_id);
+
+ if (acked & CSG_STATUS_UPDATE) {
+ csg_slot_sync_queues_state_locked(ptdev, csg_id);
+ csg_slot_sync_idle_state_locked(ptdev, csg_id);
+ }
+
+ if (ret && acked != req_mask &&
+ ((csg_iface->input->req ^ csg_iface->output->ack) & req_mask) != 0) {
+ drm_err(&ptdev->base, "CSG %d update request timedout", csg_id);
+ ctx->timedout_mask |= BIT(csg_id);
+ }
+ }
+
+ if (ctx->timedout_mask)
+ return -ETIMEDOUT;
+
+ return 0;
+}
+
+struct panthor_sched_tick_ctx {
+ struct list_head old_groups[PANTHOR_CSG_PRIORITY_COUNT];
+ struct list_head groups[PANTHOR_CSG_PRIORITY_COUNT];
+ u32 idle_group_count;
+ u32 group_count;
+ enum panthor_csg_priority min_priority;
+ struct panthor_vm *vms[MAX_CS_PER_CSG];
+ u32 as_count;
+ bool immediate_tick;
+ u32 csg_upd_failed_mask;
+};
+
+static bool
+tick_ctx_is_full(const struct panthor_scheduler *sched,
+ const struct panthor_sched_tick_ctx *ctx)
+{
+ return ctx->group_count == sched->csg_slot_count;
+}
+
+static bool
+group_is_idle(struct panthor_group *group)
+{
+ struct panthor_device *ptdev = group->ptdev;
+ u32 inactive_queues;
+
+ if (group->csg_id >= 0)
+ return ptdev->scheduler->csg_slots[group->csg_id].idle;
+
+ inactive_queues = group->idle_queues | group->blocked_queues;
+ return hweight32(inactive_queues) == group->queue_count;
+}
+
+static bool
+group_can_run(struct panthor_group *group)
+{
+ return group->state != PANTHOR_CS_GROUP_TERMINATED &&
+ !group->destroyed && group->fatal_queues == 0 &&
+ !group->timedout;
+}
+
+static void
+tick_ctx_pick_groups_from_list(const struct panthor_scheduler *sched,
+ struct panthor_sched_tick_ctx *ctx,
+ struct list_head *queue,
+ bool skip_idle_groups,
+ bool owned_by_tick_ctx)
+{
+ struct panthor_group *group, *tmp;
+
+ if (tick_ctx_is_full(sched, ctx))
+ return;
+
+ list_for_each_entry_safe(group, tmp, queue, run_node) {
+ u32 i;
+
+ if (!group_can_run(group))
+ continue;
+
+ if (skip_idle_groups && group_is_idle(group))
+ continue;
+
+ for (i = 0; i < ctx->as_count; i++) {
+ if (ctx->vms[i] == group->vm)
+ break;
+ }
+
+ if (i == ctx->as_count && ctx->as_count == sched->as_slot_count)
+ continue;
+
+ if (!owned_by_tick_ctx)
+ group_get(group);
+
+ list_move_tail(&group->run_node, &ctx->groups[group->priority]);
+ ctx->group_count++;
+ if (group_is_idle(group))
+ ctx->idle_group_count++;
+
+ if (i == ctx->as_count)
+ ctx->vms[ctx->as_count++] = group->vm;
+
+ if (ctx->min_priority > group->priority)
+ ctx->min_priority = group->priority;
+
+ if (tick_ctx_is_full(sched, ctx))
+ return;
+ }
+}
+
+static void
+tick_ctx_insert_old_group(struct panthor_scheduler *sched,
+ struct panthor_sched_tick_ctx *ctx,
+ struct panthor_group *group,
+ bool full_tick)
+{
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[group->csg_id];
+ struct panthor_group *other_group;
+
+ if (!full_tick) {
+ list_add_tail(&group->run_node, &ctx->old_groups[group->priority]);
+ return;
+ }
+
+ /* Rotate to make sure groups with lower CSG slot
+ * priorities have a chance to get a higher CSG slot
+ * priority next time they get picked. This priority
+ * has an impact on resource request ordering, so it's
+ * important to make sure we don't let one group starve
+ * all other groups with the same group priority.
+ */
+ list_for_each_entry(other_group,
+ &ctx->old_groups[csg_slot->group->priority],
+ run_node) {
+ struct panthor_csg_slot *other_csg_slot = &sched->csg_slots[other_group->csg_id];
+
+ if (other_csg_slot->priority > csg_slot->priority) {
+ list_add_tail(&csg_slot->group->run_node, &other_group->run_node);
+ return;
+ }
+ }
+
+ list_add_tail(&group->run_node, &ctx->old_groups[group->priority]);
+}
+
+static void
+tick_ctx_init(struct panthor_scheduler *sched,
+ struct panthor_sched_tick_ctx *ctx,
+ bool full_tick)
+{
+ struct panthor_device *ptdev = sched->ptdev;
+ struct panthor_csg_slots_upd_ctx upd_ctx;
+ int ret;
+ u32 i;
+
+ memset(ctx, 0, sizeof(*ctx));
+ csgs_upd_ctx_init(&upd_ctx);
+
+ ctx->min_priority = PANTHOR_CSG_PRIORITY_COUNT;
+ for (i = 0; i < ARRAY_SIZE(ctx->groups); i++) {
+ INIT_LIST_HEAD(&ctx->groups[i]);
+ INIT_LIST_HEAD(&ctx->old_groups[i]);
+ }
+
+ for (i = 0; i < sched->csg_slot_count; i++) {
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[i];
+ struct panthor_group *group = csg_slot->group;
+ struct panthor_fw_csg_iface *csg_iface;
+
+ if (!group)
+ continue;
+
+ csg_iface = panthor_fw_get_csg_iface(ptdev, i);
+ group_get(group);
+
+ /* If there was unhandled faults on the VM, force processing of
+ * CSG IRQs, so we can flag the faulty queue.
+ */
+ if (panthor_vm_has_unhandled_faults(group->vm)) {
+ sched_process_csg_irq_locked(ptdev, i);
+
+ /* No fatal fault reported, flag all queues as faulty. */
+ if (!group->fatal_queues)
+ group->fatal_queues |= GENMASK(group->queue_count - 1, 0);
+ }
+
+ tick_ctx_insert_old_group(sched, ctx, group, full_tick);
+ csgs_upd_ctx_queue_reqs(ptdev, &upd_ctx, i,
+ csg_iface->output->ack ^ CSG_STATUS_UPDATE,
+ CSG_STATUS_UPDATE);
+ }
+
+ ret = csgs_upd_ctx_apply_locked(ptdev, &upd_ctx);
+ if (ret) {
+ panthor_device_schedule_reset(ptdev);
+ ctx->csg_upd_failed_mask |= upd_ctx.timedout_mask;
+ }
+}
+
+#define NUM_INSTRS_PER_SLOT 16
+
+static void
+group_term_post_processing(struct panthor_group *group)
+{
+ struct panthor_job *job, *tmp;
+ LIST_HEAD(faulty_jobs);
+ bool cookie;
+ u32 i = 0;
+
+ if (drm_WARN_ON(&group->ptdev->base, group_can_run(group)))
+ return;
+
+ cookie = dma_fence_begin_signalling();
+ for (i = 0; i < group->queue_count; i++) {
+ struct panthor_queue *queue = group->queues[i];
+ struct panthor_syncobj_64b *syncobj;
+ int err;
+
+ if (group->fatal_queues & BIT(i))
+ err = -EINVAL;
+ else if (group->timedout)
+ err = -ETIMEDOUT;
+ else
+ err = -ECANCELED;
+
+ if (!queue)
+ continue;
+
+ spin_lock(&queue->fence_ctx.lock);
+ list_for_each_entry_safe(job, tmp, &queue->fence_ctx.in_flight_jobs, node) {
+ list_move_tail(&job->node, &faulty_jobs);
+ dma_fence_set_error(job->done_fence, err);
+ dma_fence_signal_locked(job->done_fence);
+ }
+ spin_unlock(&queue->fence_ctx.lock);
+
+ /* Manually update the syncobj seqno to unblock waiters. */
+ syncobj = group->syncobjs->kmap + (i * sizeof(*syncobj));
+ syncobj->status = ~0;
+ syncobj->seqno = atomic64_read(&queue->fence_ctx.seqno);
+ sched_queue_work(group->ptdev->scheduler, sync_upd);
+ }
+ dma_fence_end_signalling(cookie);
+
+ list_for_each_entry_safe(job, tmp, &faulty_jobs, node) {
+ list_del_init(&job->node);
+ panthor_job_put(&job->base);
+ }
+}
+
+static void group_term_work(struct work_struct *work)
+{
+ struct panthor_group *group =
+ container_of(work, struct panthor_group, term_work);
+
+ group_term_post_processing(group);
+ group_put(group);
+}
+
+static void
+tick_ctx_cleanup(struct panthor_scheduler *sched,
+ struct panthor_sched_tick_ctx *ctx)
+{
+ struct panthor_group *group, *tmp;
+ u32 i;
+
+ for (i = 0; i < ARRAY_SIZE(ctx->old_groups); i++) {
+ list_for_each_entry_safe(group, tmp, &ctx->old_groups[i], run_node) {
+ /* If everything went fine, we should only have groups
+ * to be terminated in the old_groups lists.
+ */
+ drm_WARN_ON(&group->ptdev->base, !ctx->csg_upd_failed_mask &&
+ group_can_run(group));
+
+ if (!group_can_run(group)) {
+ list_del_init(&group->run_node);
+ list_del_init(&group->wait_node);
+ group_queue_work(group, term);
+ } else if (group->csg_id >= 0) {
+ list_del_init(&group->run_node);
+ } else {
+ list_move(&group->run_node,
+ group_is_idle(group) ?
+ &sched->groups.idle[group->priority] :
+ &sched->groups.runnable[group->priority]);
+ }
+ group_put(group);
+ }
+ }
+
+ for (i = 0; i < ARRAY_SIZE(ctx->groups); i++) {
+ /* If everything went fine, the groups to schedule lists should
+ * be empty.
+ */
+ drm_WARN_ON(&group->ptdev->base,
+ !ctx->csg_upd_failed_mask && !list_empty(&ctx->groups[i]));
+
+ list_for_each_entry_safe(group, tmp, &ctx->groups[i], run_node) {
+ if (group->csg_id >= 0) {
+ list_del_init(&group->run_node);
+ } else {
+ list_move(&group->run_node,
+ group_is_idle(group) ?
+ &sched->groups.idle[group->priority] :
+ &sched->groups.runnable[group->priority]);
+ }
+ group_put(group);
+ }
+ }
+}
+
+static void
+tick_ctx_apply(struct panthor_scheduler *sched, struct panthor_sched_tick_ctx *ctx)
+{
+ struct panthor_group *group, *tmp;
+ struct panthor_device *ptdev = sched->ptdev;
+ struct panthor_csg_slot *csg_slot;
+ int prio, new_csg_prio = MAX_CSG_PRIO, i;
+ u32 csg_mod_mask = 0, free_csg_slots = 0;
+ struct panthor_csg_slots_upd_ctx upd_ctx;
+ int ret;
+
+ csgs_upd_ctx_init(&upd_ctx);
+
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1; prio >= 0; prio--) {
+ /* Suspend or terminate evicted groups. */
+ list_for_each_entry(group, &ctx->old_groups[prio], run_node) {
+ bool term = !group_can_run(group);
+ int csg_id = group->csg_id;
+
+ if (drm_WARN_ON(&ptdev->base, csg_id < 0))
+ continue;
+
+ csg_slot = &sched->csg_slots[csg_id];
+ csgs_upd_ctx_queue_reqs(ptdev, &upd_ctx, csg_id,
+ term ? CSG_STATE_TERMINATE : CSG_STATE_SUSPEND,
+ CSG_STATE_MASK);
+ }
+
+ /* Update priorities on already running groups. */
+ list_for_each_entry(group, &ctx->groups[prio], run_node) {
+ struct panthor_fw_csg_iface *csg_iface;
+ int csg_id = group->csg_id;
+
+ if (csg_id < 0) {
+ new_csg_prio--;
+ continue;
+ }
+
+ csg_slot = &sched->csg_slots[csg_id];
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+ if (csg_slot->priority == new_csg_prio) {
+ new_csg_prio--;
+ continue;
+ }
+
+ panthor_fw_update_reqs(csg_iface, endpoint_req,
+ CSG_EP_REQ_PRIORITY(new_csg_prio),
+ CSG_EP_REQ_PRIORITY_MASK);
+ csgs_upd_ctx_queue_reqs(ptdev, &upd_ctx, csg_id,
+ csg_iface->output->ack ^ CSG_ENDPOINT_CONFIG,
+ CSG_ENDPOINT_CONFIG);
+ new_csg_prio--;
+ }
+ }
+
+ ret = csgs_upd_ctx_apply_locked(ptdev, &upd_ctx);
+ if (ret) {
+ panthor_device_schedule_reset(ptdev);
+ ctx->csg_upd_failed_mask |= upd_ctx.timedout_mask;
+ return;
+ }
+
+ /* Unbind evicted groups. */
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1; prio >= 0; prio--) {
+ list_for_each_entry(group, &ctx->old_groups[prio], run_node) {
+ /* This group is gone. Process interrupts to clear
+ * any pending interrupts before we start the new
+ * group.
+ */
+ if (group->csg_id >= 0)
+ sched_process_csg_irq_locked(ptdev, group->csg_id);
+
+ group_unbind_locked(group);
+ }
+ }
+
+ for (i = 0; i < sched->csg_slot_count; i++) {
+ if (!sched->csg_slots[i].group)
+ free_csg_slots |= BIT(i);
+ }
+
+ csgs_upd_ctx_init(&upd_ctx);
+ new_csg_prio = MAX_CSG_PRIO;
+
+ /* Start new groups. */
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1; prio >= 0; prio--) {
+ list_for_each_entry(group, &ctx->groups[prio], run_node) {
+ int csg_id = group->csg_id;
+ struct panthor_fw_csg_iface *csg_iface;
+
+ if (csg_id >= 0) {
+ new_csg_prio--;
+ continue;
+ }
+
+ csg_id = ffs(free_csg_slots) - 1;
+ if (drm_WARN_ON(&ptdev->base, csg_id < 0))
+ break;
+
+ csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
+ csg_slot = &sched->csg_slots[csg_id];
+ csg_mod_mask |= BIT(csg_id);
+ group_bind_locked(group, csg_id);
+ csg_slot_prog_locked(ptdev, csg_id, new_csg_prio--);
+ csgs_upd_ctx_queue_reqs(ptdev, &upd_ctx, csg_id,
+ group->state == PANTHOR_CS_GROUP_SUSPENDED ?
+ CSG_STATE_RESUME : CSG_STATE_START,
+ CSG_STATE_MASK);
+ csgs_upd_ctx_queue_reqs(ptdev, &upd_ctx, csg_id,
+ csg_iface->output->ack ^ CSG_ENDPOINT_CONFIG,
+ CSG_ENDPOINT_CONFIG);
+ free_csg_slots &= ~BIT(csg_id);
+ }
+ }
+
+ ret = csgs_upd_ctx_apply_locked(ptdev, &upd_ctx);
+ if (ret) {
+ panthor_device_schedule_reset(ptdev);
+ ctx->csg_upd_failed_mask |= upd_ctx.timedout_mask;
+ return;
+ }
+
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1; prio >= 0; prio--) {
+ list_for_each_entry_safe(group, tmp, &ctx->groups[prio], run_node) {
+ list_del_init(&group->run_node);
+
+ /* If the group has been destroyed while we were
+ * scheduling, ask for an immediate tick to
+ * re-evaluate as soon as possible and get rid of
+ * this dangling group.
+ */
+ if (group->destroyed)
+ ctx->immediate_tick = true;
+ group_put(group);
+ }
+
+ /* Return evicted groups to the idle or run queues. Groups
+ * that can no longer be run (because they've been destroyed
+ * or experienced an unrecoverable error) will be scheduled
+ * for destruction in tick_ctx_cleanup().
+ */
+ list_for_each_entry_safe(group, tmp, &ctx->old_groups[prio], run_node) {
+ if (!group_can_run(group))
+ continue;
+
+ if (group_is_idle(group))
+ list_move_tail(&group->run_node, &sched->groups.idle[prio]);
+ else
+ list_move_tail(&group->run_node, &sched->groups.runnable[prio]);
+ group_put(group);
+ }
+ }
+
+ sched->used_csg_slot_count = ctx->group_count;
+ sched->might_have_idle_groups = ctx->idle_group_count > 0;
+}
+
+static u64
+tick_ctx_update_resched_target(struct panthor_scheduler *sched,
+ const struct panthor_sched_tick_ctx *ctx)
+{
+ /* We had space left, no need to reschedule until some external event happens. */
+ if (!tick_ctx_is_full(sched, ctx))
+ goto no_tick;
+
+ /* If idle groups were scheduled, no need to wake up until some external
+ * event happens (group unblocked, new job submitted, ...).
+ */
+ if (ctx->idle_group_count)
+ goto no_tick;
+
+ if (drm_WARN_ON(&sched->ptdev->base, ctx->min_priority >= PANTHOR_CSG_PRIORITY_COUNT))
+ goto no_tick;
+
+ /* If there are groups of the same priority waiting, we need to
+ * keep the scheduler ticking, otherwise, we'll just wait for
+ * new groups with higher priority to be queued.
+ */
+ if (!list_empty(&sched->groups.runnable[ctx->min_priority])) {
+ u64 resched_target = sched->last_tick + sched->tick_period;
+
+ if (time_before64(sched->resched_target, sched->last_tick) ||
+ time_before64(resched_target, sched->resched_target))
+ sched->resched_target = resched_target;
+
+ return sched->resched_target - sched->last_tick;
+ }
+
+no_tick:
+ sched->resched_target = U64_MAX;
+ return U64_MAX;
+}
+
+static void tick_work(struct work_struct *work)
+{
+ struct panthor_scheduler *sched = container_of(work, struct panthor_scheduler,
+ tick_work.work);
+ struct panthor_device *ptdev = sched->ptdev;
+ struct panthor_sched_tick_ctx ctx;
+ u64 remaining_jiffies = 0, resched_delay;
+ u64 now = get_jiffies_64();
+ int prio, ret, cookie;
+
+ if (!drm_dev_enter(&ptdev->base, &cookie))
+ return;
+
+ ret = pm_runtime_resume_and_get(ptdev->base.dev);
+ if (drm_WARN_ON(&ptdev->base, ret))
+ goto out_dev_exit;
+
+ if (time_before64(now, sched->resched_target))
+ remaining_jiffies = sched->resched_target - now;
+
+ mutex_lock(&sched->lock);
+ if (panthor_device_reset_is_pending(sched->ptdev))
+ goto out_unlock;
+
+ tick_ctx_init(sched, &ctx, remaining_jiffies != 0);
+ if (ctx.csg_upd_failed_mask)
+ goto out_cleanup_ctx;
+
+ if (remaining_jiffies) {
+ /* Scheduling forced in the middle of a tick. Only RT groups
+ * can preempt non-RT ones. Currently running RT groups can't be
+ * preempted.
+ */
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1;
+ prio >= 0 && !tick_ctx_is_full(sched, &ctx);
+ prio--) {
+ tick_ctx_pick_groups_from_list(sched, &ctx, &ctx.old_groups[prio],
+ true, true);
+ if (prio == PANTHOR_CSG_PRIORITY_RT) {
+ tick_ctx_pick_groups_from_list(sched, &ctx,
+ &sched->groups.runnable[prio],
+ true, false);
+ }
+ }
+ }
+
+ /* First pick non-idle groups */
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1;
+ prio >= 0 && !tick_ctx_is_full(sched, &ctx);
+ prio--) {
+ tick_ctx_pick_groups_from_list(sched, &ctx, &sched->groups.runnable[prio],
+ true, false);
+ tick_ctx_pick_groups_from_list(sched, &ctx, &ctx.old_groups[prio], true, true);
+ }
+
+ /* If we have free CSG slots left, pick idle groups */
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1;
+ prio >= 0 && !tick_ctx_is_full(sched, &ctx);
+ prio--) {
+ /* Check the old_group queue first to avoid reprogramming the slots */
+ tick_ctx_pick_groups_from_list(sched, &ctx, &ctx.old_groups[prio], false, true);
+ tick_ctx_pick_groups_from_list(sched, &ctx, &sched->groups.idle[prio],
+ false, false);
+ }
+
+ tick_ctx_apply(sched, &ctx);
+ if (ctx.csg_upd_failed_mask)
+ goto out_cleanup_ctx;
+
+ if (ctx.idle_group_count == ctx.group_count) {
+ panthor_devfreq_record_idle(sched->ptdev);
+ if (sched->pm.has_ref) {
+ pm_runtime_put_autosuspend(ptdev->base.dev);
+ sched->pm.has_ref = false;
+ }
+ } else {
+ panthor_devfreq_record_busy(sched->ptdev);
+ if (!sched->pm.has_ref) {
+ pm_runtime_get(ptdev->base.dev);
+ sched->pm.has_ref = true;
+ }
+ }
+
+ sched->last_tick = now;
+ resched_delay = tick_ctx_update_resched_target(sched, &ctx);
+ if (ctx.immediate_tick)
+ resched_delay = 0;
+
+ if (resched_delay != U64_MAX)
+ sched_queue_delayed_work(sched, tick, resched_delay);
+
+out_cleanup_ctx:
+ tick_ctx_cleanup(sched, &ctx);
+
+out_unlock:
+ mutex_unlock(&sched->lock);
+ pm_runtime_mark_last_busy(ptdev->base.dev);
+ pm_runtime_put_autosuspend(ptdev->base.dev);
+
+out_dev_exit:
+ drm_dev_exit(cookie);
+}
+
+static int panthor_queue_eval_syncwait(struct panthor_group *group, u8 queue_idx)
+{
+ struct panthor_queue *queue = group->queues[queue_idx];
+ union {
+ struct panthor_syncobj_64b sync64;
+ struct panthor_syncobj_32b sync32;
+ } *syncobj;
+ bool result;
+ u64 value;
+
+ syncobj = panthor_queue_get_syncwait_obj(group, queue);
+ if (!syncobj)
+ return -EINVAL;
+
+ value = queue->syncwait.sync64 ?
+ syncobj->sync64.seqno :
+ syncobj->sync32.seqno;
+
+ if (queue->syncwait.gt)
+ result = value > queue->syncwait.ref;
+ else
+ result = value <= queue->syncwait.ref;
+
+ if (result)
+ panthor_queue_put_syncwait_obj(queue);
+
+ return result;
+}
+
+static void sync_upd_work(struct work_struct *work)
+{
+ struct panthor_scheduler *sched = container_of(work,
+ struct panthor_scheduler,
+ sync_upd_work);
+ struct panthor_group *group, *tmp;
+ bool immediate_tick = false;
+
+ mutex_lock(&sched->lock);
+ list_for_each_entry_safe(group, tmp, &sched->groups.waiting, wait_node) {
+ u32 tested_queues = group->blocked_queues;
+ u32 unblocked_queues = 0;
+
+ while (tested_queues) {
+ u32 cs_id = ffs(tested_queues) - 1;
+ int ret;
+
+ ret = panthor_queue_eval_syncwait(group, cs_id);
+ drm_WARN_ON(&group->ptdev->base, ret < 0);
+ if (ret)
+ unblocked_queues |= BIT(cs_id);
+
+ tested_queues &= ~BIT(cs_id);
+ }
+
+ if (unblocked_queues) {
+ group->blocked_queues &= ~unblocked_queues;
+
+ if (group->csg_id < 0) {
+ list_move(&group->run_node,
+ &sched->groups.runnable[group->priority]);
+ if (group->priority == PANTHOR_CSG_PRIORITY_RT)
+ immediate_tick = true;
+ }
+ }
+
+ if (!group->blocked_queues)
+ list_del_init(&group->wait_node);
+ }
+ mutex_unlock(&sched->lock);
+
+ if (immediate_tick)
+ sched_queue_delayed_work(sched, tick, 0);
+}
+
+static void group_schedule_locked(struct panthor_group *group, u32 queue_mask)
+{
+ struct panthor_device *ptdev = group->ptdev;
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct list_head *queue = &sched->groups.runnable[group->priority];
+ u64 delay_jiffies = 0;
+ bool was_idle;
+ u64 now;
+
+ if (!group_can_run(group))
+ return;
+
+ /* All updated queues are blocked, no need to wake up the scheduler. */
+ if ((queue_mask & group->blocked_queues) == queue_mask)
+ return;
+
+ was_idle = group_is_idle(group);
+ group->idle_queues &= ~queue_mask;
+
+ /* Don't mess up with the lists if we're in a middle of a reset. */
+ if (atomic_read(&sched->reset.in_progress))
+ return;
+
+ if (was_idle && !group_is_idle(group))
+ list_move_tail(&group->run_node, queue);
+
+ /* RT groups are preemptive. */
+ if (group->priority == PANTHOR_CSG_PRIORITY_RT) {
+ sched_queue_delayed_work(sched, tick, 0);
+ return;
+ }
+
+ /* Some groups might be idle, force an immediate tick to
+ * re-evaluate.
+ */
+ if (sched->might_have_idle_groups) {
+ sched_queue_delayed_work(sched, tick, 0);
+ return;
+ }
+
+ /* Scheduler is ticking, nothing to do. */
+ if (sched->resched_target != U64_MAX) {
+ /* If there are free slots, force immediating ticking. */
+ if (sched->used_csg_slot_count < sched->csg_slot_count)
+ sched_queue_delayed_work(sched, tick, 0);
+
+ return;
+ }
+
+ /* Scheduler tick was off, recalculate the resched_target based on the
+ * last tick event, and queue the scheduler work.
+ */
+ now = get_jiffies_64();
+ sched->resched_target = sched->last_tick + sched->tick_period;
+ if (sched->used_csg_slot_count == sched->csg_slot_count &&
+ time_before64(now, sched->resched_target))
+ delay_jiffies = min_t(unsigned long, sched->resched_target - now, ULONG_MAX);
+
+ sched_queue_delayed_work(sched, tick, delay_jiffies);
+}
+
+static void queue_stop(struct panthor_queue *queue,
+ struct panthor_job *bad_job)
+{
+ drm_sched_stop(&queue->scheduler, bad_job ? &bad_job->base : NULL);
+}
+
+static void queue_start(struct panthor_queue *queue)
+{
+ struct panthor_job *job;
+
+ /* Re-assign the parent fences. */
+ list_for_each_entry(job, &queue->scheduler.pending_list, base.list)
+ job->base.s_fence->parent = dma_fence_get(job->done_fence);
+
+ drm_sched_start(&queue->scheduler, true);
+}
+
+static void panthor_group_stop(struct panthor_group *group)
+{
+ struct panthor_scheduler *sched = group->ptdev->scheduler;
+
+ lockdep_assert_held(&sched->reset.lock);
+
+ for (u32 i = 0; i < group->queue_count; i++)
+ queue_stop(group->queues[i], NULL);
+
+ group_get(group);
+ list_move_tail(&group->run_node, &sched->reset.stopped_groups);
+}
+
+static void panthor_group_start(struct panthor_group *group)
+{
+ struct panthor_scheduler *sched = group->ptdev->scheduler;
+
+ lockdep_assert_held(&group->ptdev->scheduler->reset.lock);
+
+ for (u32 i = 0; i < group->queue_count; i++)
+ queue_start(group->queues[i]);
+
+ if (group_can_run(group)) {
+ list_move_tail(&group->run_node,
+ group_is_idle(group) ?
+ &sched->groups.idle[group->priority] :
+ &sched->groups.runnable[group->priority]);
+ } else {
+ list_del_init(&group->run_node);
+ list_del_init(&group->wait_node);
+ group_queue_work(group, term);
+ }
+
+ group_put(group);
+}
+
+static void panthor_sched_immediate_tick(struct panthor_device *ptdev)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+
+ sched_queue_delayed_work(sched, tick, 0);
+}
+
+/**
+ * panthor_sched_report_mmu_fault() - Report MMU faults to the scheduler.
+ */
+void panthor_sched_report_mmu_fault(struct panthor_device *ptdev)
+{
+ /* Force a tick to immediately kill faulty groups. */
+ panthor_sched_immediate_tick(ptdev);
+}
+
+void panthor_sched_resume(struct panthor_device *ptdev)
+{
+ /* Force a tick to re-evaluate after a resume. */
+ panthor_sched_immediate_tick(ptdev);
+}
+
+void panthor_sched_suspend(struct panthor_device *ptdev)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_csg_slots_upd_ctx upd_ctx;
+ u64 suspended_slots, faulty_slots;
+ struct panthor_group *group;
+ u32 i;
+
+ mutex_lock(&sched->lock);
+ csgs_upd_ctx_init(&upd_ctx);
+ for (i = 0; i < sched->csg_slot_count; i++) {
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[i];
+
+ if (csg_slot->group) {
+ csgs_upd_ctx_queue_reqs(ptdev, &upd_ctx, i,
+ CSG_STATE_SUSPEND,
+ CSG_STATE_MASK);
+ }
+ }
+
+ suspended_slots = upd_ctx.update_mask;
+
+ csgs_upd_ctx_apply_locked(ptdev, &upd_ctx);
+ suspended_slots &= ~upd_ctx.timedout_mask;
+ faulty_slots = upd_ctx.timedout_mask;
+
+ if (faulty_slots) {
+ u32 slot_mask = faulty_slots;
+
+ drm_err(&ptdev->base, "CSG suspend failed, escalating to termination");
+ csgs_upd_ctx_init(&upd_ctx);
+ while (slot_mask) {
+ u32 csg_id = ffs(slot_mask) - 1;
+
+ csgs_upd_ctx_queue_reqs(ptdev, &upd_ctx, csg_id,
+ CSG_STATE_TERMINATE,
+ CSG_STATE_MASK);
+ slot_mask &= ~BIT(csg_id);
+ }
+
+ csgs_upd_ctx_apply_locked(ptdev, &upd_ctx);
+
+ slot_mask = upd_ctx.timedout_mask;
+ while (slot_mask) {
+ u32 csg_id = ffs(slot_mask) - 1;
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[csg_id];
+
+ /* Terminate command timedout, but the soft-reset will
+ * automatically terminate all active groups, so let's
+ * force the state to halted here.
+ */
+ if (csg_slot->group->state != PANTHOR_CS_GROUP_TERMINATED)
+ csg_slot->group->state = PANTHOR_CS_GROUP_TERMINATED;
+ slot_mask &= ~BIT(csg_id);
+ }
+ }
+
+ /* Flush L2 and LSC caches to make sure suspend state is up-to-date.
+ * If the flush fails, flag all queues for termination.
+ */
+ if (suspended_slots) {
+ bool flush_caches_failed = false;
+ u32 slot_mask = suspended_slots;
+
+ if (panthor_gpu_flush_caches(ptdev, CACHE_CLEAN, CACHE_CLEAN, 0))
+ flush_caches_failed = true;
+
+ while (slot_mask) {
+ u32 csg_id = ffs(slot_mask) - 1;
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[csg_id];
+
+ if (flush_caches_failed)
+ csg_slot->group->state = PANTHOR_CS_GROUP_TERMINATED;
+ else
+ csg_slot_sync_update_locked(ptdev, csg_id);
+
+ slot_mask &= ~BIT(csg_id);
+ }
+
+ if (flush_caches_failed)
+ faulty_slots |= suspended_slots;
+ }
+
+ for (i = 0; i < sched->csg_slot_count; i++) {
+ struct panthor_csg_slot *csg_slot = &sched->csg_slots[i];
+
+ group = csg_slot->group;
+ if (!group)
+ continue;
+
+ group_get(group);
+
+ if (group->csg_id >= 0)
+ sched_process_csg_irq_locked(ptdev, group->csg_id);
+
+ group_unbind_locked(group);
+
+ drm_WARN_ON(&group->ptdev->base, !list_empty(&group->run_node));
+
+ if (group_can_run(group)) {
+ list_add(&group->run_node,
+ &sched->groups.idle[group->priority]);
+ } else {
+ /* We don't bother stopping the scheduler if the group is
+ * faulty, the group termination work will finish the job.
+ */
+ list_del_init(&group->wait_node);
+ group_queue_work(group, term);
+ }
+ group_put(group);
+ }
+ mutex_unlock(&sched->lock);
+}
+
+void panthor_sched_pre_reset(struct panthor_device *ptdev)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_group *group, *group_tmp;
+ u32 i;
+
+ mutex_lock(&sched->reset.lock);
+ atomic_set(&sched->reset.in_progress, true);
+
+ /* Cancel all scheduler works. Once this is done, these works can't be
+ * scheduled again until the reset operation is complete.
+ */
+ cancel_work_sync(&sched->sync_upd_work);
+ cancel_delayed_work_sync(&sched->tick_work);
+
+ panthor_sched_suspend(ptdev);
+
+ /* Stop all groups that might still accept jobs, so we don't get passed
+ * new jobs while we're resetting.
+ */
+ for (i = 0; i < ARRAY_SIZE(sched->groups.runnable); i++) {
+ /* All groups should be in the idle lists. */
+ drm_WARN_ON(&ptdev->base, !list_empty(&sched->groups.runnable[i]));
+ list_for_each_entry_safe(group, group_tmp, &sched->groups.runnable[i], run_node)
+ panthor_group_stop(group);
+ }
+
+ for (i = 0; i < ARRAY_SIZE(sched->groups.idle); i++) {
+ list_for_each_entry_safe(group, group_tmp, &sched->groups.idle[i], run_node)
+ panthor_group_stop(group);
+ }
+
+ mutex_unlock(&sched->reset.lock);
+}
+
+void panthor_sched_post_reset(struct panthor_device *ptdev)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_group *group, *group_tmp;
+
+ mutex_lock(&sched->reset.lock);
+
+ list_for_each_entry_safe(group, group_tmp, &sched->reset.stopped_groups, run_node)
+ panthor_group_start(group);
+
+ /* We're done resetting the GPU, clear the reset.in_progress bit so we can
+ * kick the scheduler.
+ */
+ atomic_set(&sched->reset.in_progress, false);
+ mutex_unlock(&sched->reset.lock);
+
+ sched_queue_delayed_work(sched, tick, 0);
+
+ sched_queue_work(sched, sync_upd);
+}
+
+static void group_sync_upd_work(struct work_struct *work)
+{
+ struct panthor_group *group =
+ container_of(work, struct panthor_group, sync_upd_work);
+ struct panthor_job *job, *job_tmp;
+ LIST_HEAD(done_jobs);
+ u32 queue_idx;
+ bool cookie;
+
+ cookie = dma_fence_begin_signalling();
+ for (queue_idx = 0; queue_idx < group->queue_count; queue_idx++) {
+ struct panthor_queue *queue = group->queues[queue_idx];
+ struct panthor_syncobj_64b *syncobj;
+
+ if (!queue)
+ continue;
+
+ syncobj = group->syncobjs->kmap + (queue_idx * sizeof(*syncobj));
+
+ spin_lock(&queue->fence_ctx.lock);
+ list_for_each_entry_safe(job, job_tmp, &queue->fence_ctx.in_flight_jobs, node) {
+ if (!job->call_info.size)
+ continue;
+
+ if (syncobj->seqno < job->done_fence->seqno)
+ break;
+
+ list_move_tail(&job->node, &done_jobs);
+ dma_fence_signal_locked(job->done_fence);
+ }
+ spin_unlock(&queue->fence_ctx.lock);
+ }
+ dma_fence_end_signalling(cookie);
+
+ list_for_each_entry_safe(job, job_tmp, &done_jobs, node) {
+ list_del_init(&job->node);
+ panthor_job_put(&job->base);
+ }
+
+ group_put(group);
+}
+
+static struct dma_fence *
+queue_run_job(struct drm_sched_job *sched_job)
+{
+ struct panthor_job *job = container_of(sched_job, struct panthor_job, base);
+ struct panthor_group *group = job->group;
+ struct panthor_queue *queue = group->queues[job->queue_idx];
+ struct panthor_device *ptdev = group->ptdev;
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ u32 ringbuf_size = panthor_kernel_bo_size(queue->ringbuf);
+ u32 ringbuf_insert = queue->iface.input->insert & (ringbuf_size - 1);
+ u64 addr_reg = ptdev->csif_info.cs_reg_count -
+ ptdev->csif_info.unpreserved_cs_reg_count;
+ u64 val_reg = addr_reg + 2;
+ u64 sync_addr = panthor_kernel_bo_gpuva(group->syncobjs) +
+ job->queue_idx * sizeof(struct panthor_syncobj_64b);
+ u32 waitall_mask = GENMASK(sched->sb_slot_count - 1, 0);
+ struct dma_fence *done_fence;
+ int ret;
+
+ u64 call_instrs[NUM_INSTRS_PER_SLOT] = {
+ /* MOV32 rX+2, cs.latest_flush */
+ (2ull << 56) | (val_reg << 48) | job->call_info.latest_flush,
+
+ /* FLUSH_CACHE2.clean_inv_all.no_wait.signal(0) rX+2 */
+ (36ull << 56) | (0ull << 48) | (val_reg << 40) | (0 << 16) | 0x233,
+
+ /* MOV48 rX:rX+1, cs.start */
+ (1ull << 56) | (addr_reg << 48) | job->call_info.start,
+
+ /* MOV32 rX+2, cs.size */
+ (2ull << 56) | (val_reg << 48) | job->call_info.size,
+
+ /* WAIT(0) => waits for FLUSH_CACHE2 instruction */
+ (3ull << 56) | (1 << 16),
+
+ /* CALL rX:rX+1, rX+2 */
+ (32ull << 56) | (addr_reg << 40) | (val_reg << 32),
+
+ /* MOV48 rX:rX+1, sync_addr */
+ (1ull << 56) | (addr_reg << 48) | sync_addr,
+
+ /* MOV48 rX+2, #1 */
+ (1ull << 56) | (val_reg << 48) | 1,
+
+ /* WAIT(all) */
+ (3ull << 56) | (waitall_mask << 16),
+
+ /* SYNC_ADD64.system_scope.propage_err.nowait rX:rX+1, rX+2*/
+ (51ull << 56) | (0ull << 48) | (addr_reg << 40) | (val_reg << 32) | (0 << 16) | 1,
+
+ /* ERROR_BARRIER, so we can recover from faults at job
+ * boundaries.
+ */
+ (47ull << 56),
+ };
+
+ /* Need to be cacheline aligned to please the prefetcher. */
+ static_assert(sizeof(call_instrs) % 64 == 0,
+ "call_instrs is not aligned on a cacheline");
+
+ /* Stream size is zero, nothing to do => return a NULL fence and let
+ * drm_sched signal the parent.
+ */
+ if (!job->call_info.size)
+ return NULL;
+
+ ret = pm_runtime_resume_and_get(ptdev->base.dev);
+ if (drm_WARN_ON(&ptdev->base, ret))
+ return ERR_PTR(ret);
+
+ mutex_lock(&sched->lock);
+ if (!group_can_run(group)) {
+ done_fence = ERR_PTR(-ECANCELED);
+ goto out_unlock;
+ }
+
+ dma_fence_init(job->done_fence,
+ &panthor_queue_fence_ops,
+ &queue->fence_ctx.lock,
+ queue->fence_ctx.id,
+ atomic64_inc_return(&queue->fence_ctx.seqno));
+
+ memcpy(queue->ringbuf->kmap + ringbuf_insert,
+ call_instrs, sizeof(call_instrs));
+
+ panthor_job_get(&job->base);
+ spin_lock(&queue->fence_ctx.lock);
+ list_add_tail(&job->node, &queue->fence_ctx.in_flight_jobs);
+ spin_unlock(&queue->fence_ctx.lock);
+
+ job->ringbuf.start = queue->iface.input->insert;
+ job->ringbuf.end = job->ringbuf.start + sizeof(call_instrs);
+
+ /* Make sure the ring buffer is updated before the INSERT
+ * register.
+ */
+ wmb();
+
+ queue->iface.input->extract = queue->iface.output->extract;
+ queue->iface.input->insert = job->ringbuf.end;
+
+ if (group->csg_id < 0) {
+ /* If the queue is blocked, we want to keep the timeout running, so we
+ * can detect unbounded waits and kill the group when that happens.
+ * Otherwise, we suspend the timeout so the time we spend waiting for
+ * a CSG slot is not counted.
+ */
+ if (!(group->blocked_queues & BIT(job->queue_idx)) &&
+ !queue->timeout_suspended) {
+ queue->remaining_time = drm_sched_suspend_timeout(&queue->scheduler);
+ queue->timeout_suspended = true;
+ }
+
+ group_schedule_locked(group, BIT(job->queue_idx));
+ } else {
+ gpu_write(ptdev, CSF_DOORBELL(queue->doorbell_id), 1);
+ if (!sched->pm.has_ref &&
+ !(group->blocked_queues & BIT(job->queue_idx))) {
+ pm_runtime_get(ptdev->base.dev);
+ sched->pm.has_ref = true;
+ }
+ }
+
+ done_fence = dma_fence_get(job->done_fence);
+
+out_unlock:
+ mutex_unlock(&sched->lock);
+ pm_runtime_mark_last_busy(ptdev->base.dev);
+ pm_runtime_put_autosuspend(ptdev->base.dev);
+
+ return done_fence;
+}
+
+static enum drm_gpu_sched_stat
+queue_timedout_job(struct drm_sched_job *sched_job)
+{
+ struct panthor_job *job = container_of(sched_job, struct panthor_job, base);
+ struct panthor_group *group = job->group;
+ struct panthor_device *ptdev = group->ptdev;
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_queue *queue = group->queues[job->queue_idx];
+
+ drm_warn(&ptdev->base, "job timeout\n");
+
+ drm_WARN_ON(&ptdev->base, atomic_read(&sched->reset.in_progress));
+
+ queue_stop(queue, job);
+
+ mutex_lock(&sched->lock);
+ group->timedout = true;
+ if (group->csg_id >= 0) {
+ sched_queue_delayed_work(ptdev->scheduler, tick, 0);
+ } else {
+ /* Remove from the run queues, so the scheduler can't
+ * pick the group on the next tick.
+ */
+ drm_WARN_ON(&ptdev->base, list_empty(&group->run_node));
+ list_del_init(&group->run_node);
+ list_del_init(&group->wait_node);
+
+ group_queue_work(group, term);
+ }
+ mutex_unlock(&sched->lock);
+
+ queue_start(queue);
+
+ return DRM_GPU_SCHED_STAT_NOMINAL;
+}
+
+static void queue_free_job(struct drm_sched_job *sched_job)
+{
+ drm_sched_job_cleanup(sched_job);
+ panthor_job_put(sched_job);
+}
+
+static const struct drm_sched_backend_ops panthor_queue_sched_ops = {
+ .run_job = queue_run_job,
+ .timedout_job = queue_timedout_job,
+ .free_job = queue_free_job,
+};
+
+static struct panthor_queue *
+group_create_queue(struct panthor_group *group,
+ const struct drm_panthor_queue_create *args)
+{
+ struct drm_gpu_scheduler *drm_sched;
+ struct panthor_queue *queue;
+ int ret;
+
+ if (args->pad[0] || args->pad[1] || args->pad[2])
+ return ERR_PTR(-EINVAL);
+
+ if (args->ringbuf_size < SZ_4K || args->ringbuf_size > SZ_64K ||
+ !is_power_of_2(args->ringbuf_size))
+ return ERR_PTR(-EINVAL);
+
+ if (args->priority > CSF_MAX_QUEUE_PRIO)
+ return ERR_PTR(-EINVAL);
+
+ queue = kzalloc(sizeof(*queue), GFP_KERNEL);
+ if (!queue)
+ return ERR_PTR(-ENOMEM);
+
+ queue->fence_ctx.id = dma_fence_context_alloc(1);
+ spin_lock_init(&queue->fence_ctx.lock);
+ INIT_LIST_HEAD(&queue->fence_ctx.in_flight_jobs);
+
+ queue->priority = args->priority;
+
+ queue->ringbuf = panthor_kernel_bo_create(group->ptdev, group->vm,
+ args->ringbuf_size,
+ DRM_PANTHOR_BO_NO_MMAP,
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC |
+ DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED,
+ PANTHOR_VM_KERNEL_AUTO_VA);
+ if (IS_ERR(queue->ringbuf)) {
+ ret = PTR_ERR(queue->ringbuf);
+ goto err_free_queue;
+ }
+
+ ret = panthor_kernel_bo_vmap(queue->ringbuf);
+ if (ret)
+ goto err_free_queue;
+
+ queue->iface.mem = panthor_fw_alloc_queue_iface_mem(group->ptdev,
+ &queue->iface.input,
+ &queue->iface.output,
+ &queue->iface.input_fw_va,
+ &queue->iface.output_fw_va);
+ if (IS_ERR(queue->iface.mem)) {
+ ret = PTR_ERR(queue->iface.mem);
+ goto err_free_queue;
+ }
+
+ ret = drm_sched_init(&queue->scheduler, &panthor_queue_sched_ops,
+ group->ptdev->scheduler->wq, 1,
+ args->ringbuf_size / (NUM_INSTRS_PER_SLOT * sizeof(u64)),
+ 0, msecs_to_jiffies(JOB_TIMEOUT_MS),
+ group->ptdev->reset.wq,
+ NULL, "panthor-queue", group->ptdev->base.dev);
+ if (ret)
+ goto err_free_queue;
+
+ drm_sched = &queue->scheduler;
+ ret = drm_sched_entity_init(&queue->entity, 0, &drm_sched, 1, NULL);
+
+ return queue;
+
+err_free_queue:
+ group_free_queue(group, queue);
+ return ERR_PTR(ret);
+}
+
+#define MAX_GROUPS_PER_POOL 128
+
+int panthor_group_create(struct panthor_file *pfile,
+ const struct drm_panthor_group_create *group_args,
+ const struct drm_panthor_queue_create *queue_args)
+{
+ struct panthor_device *ptdev = pfile->ptdev;
+ struct panthor_group_pool *gpool = pfile->groups;
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_fw_csg_iface *csg_iface = panthor_fw_get_csg_iface(ptdev, 0);
+ struct panthor_group *group = NULL;
+ u32 gid, i, suspend_size;
+ int ret;
+
+ if (group_args->pad)
+ return -EINVAL;
+
+ if (group_args->priority > PANTHOR_CSG_PRIORITY_HIGH)
+ return -EINVAL;
+
+ if ((group_args->compute_core_mask & ~ptdev->gpu_info.shader_present) ||
+ (group_args->fragment_core_mask & ~ptdev->gpu_info.shader_present) ||
+ (group_args->tiler_core_mask & ~ptdev->gpu_info.tiler_present))
+ return -EINVAL;
+
+ if (hweight64(group_args->compute_core_mask) < group_args->max_compute_cores ||
+ hweight64(group_args->fragment_core_mask) < group_args->max_fragment_cores ||
+ hweight64(group_args->tiler_core_mask) < group_args->max_tiler_cores)
+ return -EINVAL;
+
+ group = kzalloc(sizeof(*group), GFP_KERNEL);
+ if (!group)
+ return -ENOMEM;
+
+ spin_lock_init(&group->fatal_lock);
+ kref_init(&group->refcount);
+ group->state = PANTHOR_CS_GROUP_CREATED;
+ group->csg_id = -1;
+
+ group->ptdev = ptdev;
+ group->max_compute_cores = group_args->max_compute_cores;
+ group->compute_core_mask = group_args->compute_core_mask;
+ group->max_fragment_cores = group_args->max_fragment_cores;
+ group->fragment_core_mask = group_args->fragment_core_mask;
+ group->max_tiler_cores = group_args->max_tiler_cores;
+ group->tiler_core_mask = group_args->tiler_core_mask;
+ group->priority = group_args->priority;
+
+ INIT_LIST_HEAD(&group->wait_node);
+ INIT_LIST_HEAD(&group->run_node);
+ INIT_WORK(&group->term_work, group_term_work);
+ INIT_WORK(&group->sync_upd_work, group_sync_upd_work);
+ INIT_WORK(&group->tiler_oom_work, group_tiler_oom_work);
+ INIT_WORK(&group->release_work, group_release_work);
+
+ group->vm = panthor_vm_pool_get_vm(pfile->vms, group_args->vm_id);
+ if (!group->vm) {
+ ret = -EINVAL;
+ goto err_put_group;
+ }
+
+ suspend_size = csg_iface->control->suspend_size;
+ group->suspend_buf = panthor_fw_alloc_suspend_buf_mem(ptdev, suspend_size);
+ if (IS_ERR(group->suspend_buf)) {
+ ret = PTR_ERR(group->suspend_buf);
+ group->suspend_buf = NULL;
+ goto err_put_group;
+ }
+
+ suspend_size = csg_iface->control->protm_suspend_size;
+ group->protm_suspend_buf = panthor_fw_alloc_suspend_buf_mem(ptdev, suspend_size);
+ if (IS_ERR(group->protm_suspend_buf)) {
+ ret = PTR_ERR(group->protm_suspend_buf);
+ group->protm_suspend_buf = NULL;
+ goto err_put_group;
+ }
+
+ group->syncobjs = panthor_kernel_bo_create(ptdev, group->vm,
+ group_args->queues.count *
+ sizeof(struct panthor_syncobj_64b),
+ DRM_PANTHOR_BO_NO_MMAP,
+ DRM_PANTHOR_VM_BIND_OP_MAP_NOEXEC |
+ DRM_PANTHOR_VM_BIND_OP_MAP_UNCACHED,
+ PANTHOR_VM_KERNEL_AUTO_VA);
+ if (IS_ERR(group->syncobjs)) {
+ ret = PTR_ERR(group->syncobjs);
+ goto err_put_group;
+ }
+
+ ret = panthor_kernel_bo_vmap(group->syncobjs);
+ if (ret)
+ goto err_put_group;
+
+ memset(group->syncobjs->kmap, 0,
+ group_args->queues.count * sizeof(struct panthor_syncobj_64b));
+
+ for (i = 0; i < group_args->queues.count; i++) {
+ group->queues[i] = group_create_queue(group, &queue_args[i]);
+ if (IS_ERR(group->queues[i])) {
+ ret = PTR_ERR(group->queues[i]);
+ group->queues[i] = NULL;
+ goto err_put_group;
+ }
+
+ group->queue_count++;
+ }
+
+ group->idle_queues = GENMASK(group->queue_count - 1, 0);
+
+ ret = xa_alloc(&gpool->xa, &gid, group, XA_LIMIT(1, MAX_GROUPS_PER_POOL), GFP_KERNEL);
+ if (ret)
+ goto err_put_group;
+
+ mutex_lock(&sched->reset.lock);
+ if (atomic_read(&sched->reset.in_progress)) {
+ panthor_group_stop(group);
+ } else {
+ mutex_lock(&sched->lock);
+ list_add_tail(&group->run_node,
+ &sched->groups.idle[group->priority]);
+ mutex_unlock(&sched->lock);
+ }
+ mutex_unlock(&sched->reset.lock);
+
+ return gid;
+
+err_put_group:
+ group_put(group);
+ return ret;
+}
+
+int panthor_group_destroy(struct panthor_file *pfile, u32 group_handle)
+{
+ struct panthor_group_pool *gpool = pfile->groups;
+ struct panthor_device *ptdev = pfile->ptdev;
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_group *group;
+
+ group = xa_erase(&gpool->xa, group_handle);
+ if (!group)
+ return -EINVAL;
+
+ for (u32 i = 0; i < group->queue_count; i++) {
+ if (group->queues[i])
+ drm_sched_entity_destroy(&group->queues[i]->entity);
+ }
+
+ mutex_lock(&sched->reset.lock);
+ mutex_lock(&sched->lock);
+ group->destroyed = true;
+ if (group->csg_id >= 0) {
+ sched_queue_delayed_work(sched, tick, 0);
+ } else if (!atomic_read(&sched->reset.in_progress)) {
+ /* Remove from the run queues, so the scheduler can't
+ * pick the group on the next tick.
+ */
+ list_del_init(&group->run_node);
+ list_del_init(&group->wait_node);
+ group_queue_work(group, term);
+ }
+ mutex_unlock(&sched->lock);
+ mutex_unlock(&sched->reset.lock);
+
+ group_put(group);
+ return 0;
+}
+
+int panthor_group_get_state(struct panthor_file *pfile,
+ struct drm_panthor_group_get_state *get_state)
+{
+ struct panthor_group_pool *gpool = pfile->groups;
+ struct panthor_device *ptdev = pfile->ptdev;
+ struct panthor_scheduler *sched = ptdev->scheduler;
+ struct panthor_group *group;
+
+ if (get_state->pad)
+ return -EINVAL;
+
+ group = group_get(xa_load(&gpool->xa, get_state->group_handle));
+ if (!group)
+ return -EINVAL;
+
+ memset(get_state, 0, sizeof(*get_state));
+
+ mutex_lock(&sched->lock);
+ if (group->timedout)
+ get_state->state |= DRM_PANTHOR_GROUP_STATE_TIMEDOUT;
+ if (group->fatal_queues) {
+ get_state->state |= DRM_PANTHOR_GROUP_STATE_FATAL_FAULT;
+ get_state->fatal_queues = group->fatal_queues;
+ }
+ mutex_unlock(&sched->lock);
+
+ group_put(group);
+ return 0;
+}
+
+int panthor_group_pool_create(struct panthor_file *pfile)
+{
+ struct panthor_group_pool *gpool;
+
+ gpool = kzalloc(sizeof(*gpool), GFP_KERNEL);
+ if (!gpool)
+ return -ENOMEM;
+
+ xa_init_flags(&gpool->xa, XA_FLAGS_ALLOC1);
+ pfile->groups = gpool;
+ return 0;
+}
+
+void panthor_group_pool_destroy(struct panthor_file *pfile)
+{
+ struct panthor_group_pool *gpool = pfile->groups;
+ struct panthor_group *group;
+ unsigned long i;
+
+ if (IS_ERR_OR_NULL(gpool))
+ return;
+
+ xa_for_each(&gpool->xa, i, group)
+ panthor_group_destroy(pfile, i);
+
+ xa_destroy(&gpool->xa);
+ kfree(gpool);
+ pfile->groups = NULL;
+}
+
+static void job_release(struct kref *ref)
+{
+ struct panthor_job *job = container_of(ref, struct panthor_job, refcount);
+
+ drm_WARN_ON(&job->group->ptdev->base, !list_empty(&job->node));
+
+ if (job->base.s_fence)
+ drm_sched_job_cleanup(&job->base);
+
+ if (job->done_fence && job->done_fence->ops)
+ dma_fence_put(job->done_fence);
+ else
+ dma_fence_free(job->done_fence);
+
+ group_put(job->group);
+
+ kfree(job);
+}
+
+struct drm_sched_job *panthor_job_get(struct drm_sched_job *sched_job)
+{
+ if (sched_job) {
+ struct panthor_job *job = container_of(sched_job, struct panthor_job, base);
+
+ kref_get(&job->refcount);
+ }
+
+ return sched_job;
+}
+
+void panthor_job_put(struct drm_sched_job *sched_job)
+{
+ struct panthor_job *job = container_of(sched_job, struct panthor_job, base);
+
+ if (sched_job)
+ kref_put(&job->refcount, job_release);
+}
+
+struct panthor_vm *panthor_job_vm(struct drm_sched_job *sched_job)
+{
+ struct panthor_job *job = container_of(sched_job, struct panthor_job, base);
+
+ return job->group->vm;
+}
+
+struct drm_sched_job *
+panthor_job_create(struct panthor_file *pfile,
+ u16 group_handle,
+ const struct drm_panthor_queue_submit *qsubmit)
+{
+ struct panthor_group_pool *gpool = pfile->groups;
+ struct panthor_job *job;
+ int ret;
+
+ if (qsubmit->pad)
+ return ERR_PTR(-EINVAL);
+
+ /* If stream_addr is zero, so stream_size should be. */
+ if ((qsubmit->stream_size == 0) != (qsubmit->stream_addr == 0))
+ return ERR_PTR(-EINVAL);
+
+ /* Make sure the address is aligned on 64-byte (cacheline) and the size is
+ * aligned on 8-byte (instruction size).
+ */
+ if ((qsubmit->stream_addr & 63) || (qsubmit->stream_size & 7))
+ return ERR_PTR(-EINVAL);
+
+ /* bits 24:30 must be zero. */
+ if (qsubmit->latest_flush & GENMASK(30, 24))
+ return ERR_PTR(-EINVAL);
+
+ job = kzalloc(sizeof(*job), GFP_KERNEL);
+ if (!job)
+ return ERR_PTR(-ENOMEM);
+
+ kref_init(&job->refcount);
+ job->queue_idx = qsubmit->queue_index;
+ job->call_info.size = qsubmit->stream_size;
+ job->call_info.start = qsubmit->stream_addr;
+ job->call_info.latest_flush = qsubmit->latest_flush;
+ INIT_LIST_HEAD(&job->node);
+
+ job->group = group_get(xa_load(&gpool->xa, group_handle));
+ if (!job->group) {
+ ret = -EINVAL;
+ goto err_put_job;
+ }
+
+ if (job->queue_idx >= job->group->queue_count ||
+ !job->group->queues[job->queue_idx]) {
+ ret = -EINVAL;
+ goto err_put_job;
+ }
+
+ job->done_fence = kzalloc(sizeof(*job->done_fence), GFP_KERNEL);
+ if (!job->done_fence) {
+ ret = -ENOMEM;
+ goto err_put_job;
+ }
+
+ ret = drm_sched_job_init(&job->base,
+ &job->group->queues[job->queue_idx]->entity,
+ 1, job->group);
+ if (ret)
+ goto err_put_job;
+
+ return &job->base;
+
+err_put_job:
+ panthor_job_put(&job->base);
+ return ERR_PTR(ret);
+}
+
+void panthor_job_update_resvs(struct drm_exec *exec, struct drm_sched_job *sched_job)
+{
+ struct panthor_job *job = container_of(sched_job, struct panthor_job, base);
+
+ /* Still not sure why we want USAGE_WRITE for external objects, since I
+ * was assuming this would be handled through explicit syncs being imported
+ * to external BOs with DMA_BUF_IOCTL_IMPORT_SYNC_FILE, but other drivers
+ * seem to pass DMA_RESV_USAGE_WRITE, so there must be a good reason.
+ */
+ panthor_vm_update_resvs(job->group->vm, exec, &sched_job->s_fence->finished,
+ DMA_RESV_USAGE_BOOKKEEP, DMA_RESV_USAGE_WRITE);
+}
+
+void panthor_sched_unplug(struct panthor_device *ptdev)
+{
+ struct panthor_scheduler *sched = ptdev->scheduler;
+
+ cancel_delayed_work_sync(&sched->tick_work);
+
+ mutex_lock(&sched->lock);
+ if (sched->pm.has_ref) {
+ pm_runtime_put(ptdev->base.dev);
+ sched->pm.has_ref = false;
+ }
+ mutex_unlock(&sched->lock);
+}
+
+static void panthor_sched_fini(struct drm_device *ddev, void *res)
+{
+ struct panthor_scheduler *sched = res;
+ int prio;
+
+ if (!sched || !sched->csg_slot_count)
+ return;
+
+ cancel_delayed_work_sync(&sched->tick_work);
+
+ if (sched->wq)
+ destroy_workqueue(sched->wq);
+
+ if (sched->heap_alloc_wq)
+ destroy_workqueue(sched->heap_alloc_wq);
+
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1; prio >= 0; prio--) {
+ drm_WARN_ON(ddev, !list_empty(&sched->groups.runnable[prio]));
+ drm_WARN_ON(ddev, !list_empty(&sched->groups.idle[prio]));
+ }
+
+ drm_WARN_ON(ddev, !list_empty(&sched->groups.waiting));
+}
+
+int panthor_sched_init(struct panthor_device *ptdev)
+{
+ struct panthor_fw_global_iface *glb_iface = panthor_fw_get_glb_iface(ptdev);
+ struct panthor_fw_csg_iface *csg_iface = panthor_fw_get_csg_iface(ptdev, 0);
+ struct panthor_fw_cs_iface *cs_iface = panthor_fw_get_cs_iface(ptdev, 0, 0);
+ struct panthor_scheduler *sched;
+ u32 gpu_as_count, num_groups;
+ int prio, ret;
+
+ sched = drmm_kzalloc(&ptdev->base, sizeof(*sched), GFP_KERNEL);
+ if (!sched)
+ return -ENOMEM;
+
+ /* The highest bit in JOB_INT_* is reserved for globabl IRQs. That
+ * leaves 31 bits for CSG IRQs, hence the MAX_CSGS clamp here.
+ */
+ num_groups = min_t(u32, MAX_CSGS, glb_iface->control->group_num);
+
+ /* The FW-side scheduler might deadlock if two groups with the same
+ * priority try to access a set of resources that overlaps, with part
+ * of the resources being allocated to one group and the other part to
+ * the other group, both groups waiting for the remaining resources to
+ * be allocated. To avoid that, it is recommended to assign each CSG a
+ * different priority. In theory we could allow several groups to have
+ * the same CSG priority if they don't request the same resources, but
+ * that makes the scheduling logic more complicated, so let's clamp
+ * the number of CSG slots to MAX_CSG_PRIO + 1 for now.
+ */
+ num_groups = min_t(u32, MAX_CSG_PRIO + 1, num_groups);
+
+ /* We need at least one AS for the MCU and one for the GPU contexts. */
+ gpu_as_count = hweight32(ptdev->gpu_info.as_present & GENMASK(31, 1));
+ if (!gpu_as_count) {
+ drm_err(&ptdev->base, "Not enough AS (%d, expected at least 2)",
+ gpu_as_count + 1);
+ return -EINVAL;
+ }
+
+ sched->ptdev = ptdev;
+ sched->sb_slot_count = CS_FEATURES_SCOREBOARDS(cs_iface->control->features);
+ sched->csg_slot_count = num_groups;
+ sched->cs_slot_count = csg_iface->control->stream_num;
+ sched->as_slot_count = gpu_as_count;
+ ptdev->csif_info.csg_slot_count = sched->csg_slot_count;
+ ptdev->csif_info.cs_slot_count = sched->cs_slot_count;
+ ptdev->csif_info.scoreboard_slot_count = sched->sb_slot_count;
+
+ sched->last_tick = 0;
+ sched->resched_target = U64_MAX;
+ sched->tick_period = msecs_to_jiffies(10);
+ INIT_DELAYED_WORK(&sched->tick_work, tick_work);
+ INIT_WORK(&sched->sync_upd_work, sync_upd_work);
+ INIT_WORK(&sched->fw_events_work, process_fw_events_work);
+
+ ret = drmm_mutex_init(&ptdev->base, &sched->lock);
+ if (ret)
+ return ret;
+
+ for (prio = PANTHOR_CSG_PRIORITY_COUNT - 1; prio >= 0; prio--) {
+ INIT_LIST_HEAD(&sched->groups.runnable[prio]);
+ INIT_LIST_HEAD(&sched->groups.idle[prio]);
+ }
+ INIT_LIST_HEAD(&sched->groups.waiting);
+
+ ret = drmm_mutex_init(&ptdev->base, &sched->reset.lock);
+ if (ret)
+ return ret;
+
+ INIT_LIST_HEAD(&sched->reset.stopped_groups);
+
+ /* sched->heap_alloc_wq will be used for heap chunk allocation on
+ * tiler OOM events, which means we can't use the same workqueue for
+ * the scheduler because works queued by the scheduler are in
+ * the dma-signalling path. Allocate a dedicated heap_alloc_wq to
+ * work around this limitation.
+ *
+ * FIXME: Ultimately, what we need is a failable/non-blocking GEM
+ * allocation path that we can call when a heap OOM is reported. The
+ * FW is smart enough to fall back on other methods if the kernel can't
+ * allocate memory, and fail the tiling job if none of these
+ * countermeasures worked.
+ *
+ * Set WQ_MEM_RECLAIM on sched->wq to unblock the situation when the
+ * system is running out of memory.
+ */
+ sched->heap_alloc_wq = alloc_workqueue("panthor-heap-alloc", WQ_UNBOUND, 0);
+ sched->wq = alloc_workqueue("panthor-csf-sched", WQ_MEM_RECLAIM | WQ_UNBOUND, 0);
+ if (!sched->wq || !sched->heap_alloc_wq) {
+ panthor_sched_fini(&ptdev->base, sched);
+ drm_err(&ptdev->base, "Failed to allocate the workqueues");
+ return -ENOMEM;
+ }
+
+ ret = drmm_add_action_or_reset(&ptdev->base, panthor_sched_fini, sched);
+ if (ret)
+ return ret;
+
+ ptdev->scheduler = sched;
+ return 0;
+}
diff --git a/drivers/gpu/drm/panthor/panthor_sched.h b/drivers/gpu/drm/panthor/panthor_sched.h
new file mode 100644
index 000000000000..bd6912942bc1
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_sched.h
@@ -0,0 +1,48 @@
+/* SPDX-License-Identifier: GPL-2.0 or MIT */
+/* Copyright 2023 Collabora ltd. */
+
+#ifndef __PANTHOR_SCHED_H__
+#define __PANTHOR_SCHED_H__
+
+#include <drm/panthor_drm.h>
+
+struct drm_exec;
+struct dma_fence;
+struct drm_file;
+struct drm_gem_object;
+struct drm_sched_job;
+struct panthor_device;
+struct panthor_file;
+struct panthor_group_pool;
+struct panthor_job;
+
+int panthor_group_create(struct panthor_file *pfile,
+ const struct drm_panthor_group_create *group_args,
+ const struct drm_panthor_queue_create *queue_args);
+int panthor_group_destroy(struct panthor_file *pfile, u32 group_handle);
+int panthor_group_get_state(struct panthor_file *pfile,
+ struct drm_panthor_group_get_state *get_state);
+
+struct drm_sched_job *
+panthor_job_create(struct panthor_file *pfile,
+ u16 group_handle,
+ const struct drm_panthor_queue_submit *qsubmit);
+struct drm_sched_job *panthor_job_get(struct drm_sched_job *job);
+struct panthor_vm *panthor_job_vm(struct drm_sched_job *sched_job);
+void panthor_job_put(struct drm_sched_job *job);
+void panthor_job_update_resvs(struct drm_exec *exec, struct drm_sched_job *job);
+
+int panthor_group_pool_create(struct panthor_file *pfile);
+void panthor_group_pool_destroy(struct panthor_file *pfile);
+
+int panthor_sched_init(struct panthor_device *ptdev);
+void panthor_sched_unplug(struct panthor_device *ptdev);
+void panthor_sched_pre_reset(struct panthor_device *ptdev);
+void panthor_sched_post_reset(struct panthor_device *ptdev);
+void panthor_sched_suspend(struct panthor_device *ptdev);
+void panthor_sched_resume(struct panthor_device *ptdev);
+
+void panthor_sched_report_mmu_fault(struct panthor_device *ptdev);
+void panthor_sched_report_fw_events(struct panthor_device *ptdev, u32 events);
+
+#endif
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 10/14] drm/panthor: Add the scheduler logical block
2024-01-22 16:30 ` [PATCH v4 10/14] drm/panthor: Add the scheduler " Boris Brezillon
@ 2024-02-14 16:48 ` Steven Price
0 siblings, 0 replies; 51+ messages in thread
From: Steven Price @ 2024-02-14 16:48 UTC (permalink / raw)
To: Boris Brezillon, dri-devel
Cc: Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen
On 22/01/2024 16:30, Boris Brezillon wrote:
> This is the piece of software interacting with the FW scheduler, and
> taking care of some scheduling aspects when the FW comes short of slots
> scheduling slots. Indeed, the FW only expose a few slots, and the kernel
> has to give all submission contexts, a chance to execute their jobs.
>
> The kernel-side scheduler is timeslice-based, with a round-robin queue
> per priority level.
>
> Job submission is handled with a 1:1 drm_sched_entity:drm_gpu_scheduler,
> allowing us to delegate the dependency tracking to the core.
>
> All the gory details should be documented inline.
>
> v4:
> - Check drmm_mutex_init() return code
> - s/drm_gem_vmap_unlocked/drm_gem_vunmap_unlocked/ in
> panthor_queue_put_syncwait_obj()
> - Drop unneeded WARN_ON() in cs_slot_sync_queue_state_locked()
> - Use atomic_xchg() instead of atomic_fetch_and(0)
> - Fix typos
> - Let panthor_kernel_bo_destroy() check for IS_ERR_OR_NULL() BOs
> - Defer TILER_OOM event handling to a separate workqueue to prevent
> deadlocks when the heap chunk allocation is blocked on mem-reclaim.
> This is just a temporary solution, until we add support for
> non-blocking/failable allocations
> - Pass the scheduler workqueue to drm_sched instead of instantiating
> a separate one (no longer needed now that heap chunk allocation
> happens on a dedicated wq)
> - Set WQ_MEM_RECLAIM on the scheduler workqueue, so we can handle
> job timeouts when the system is under mem pressure, and hopefully
> free up some memory retained by these jobs
>
> v3:
> - Rework the FW event handling logic to avoid races
> - Make sure MMU faults kill the group immediately
> - Use the panthor_kernel_bo abstraction for group/queue buffers
> - Make in_progress an atomic_t, so we can check it without the reset lock
> held
> - Don't limit the number of groups per context to the FW scheduler
> capacity. Fix the limit to 128 for now.
> - Add a panthor_job_vm() helper
> - Account for panthor_vm changes
> - Add our job fence as DMA_RESV_USAGE_WRITE to all external objects
> (was previously DMA_RESV_USAGE_BOOKKEEP). I don't get why, given
> we're supposed to be fully-explicit, but other drivers do that, so
> there must be a good reason
> - Account for drm_sched changes
> - Provide a panthor_queue_put_syncwait_obj()
> - Unconditionally return groups to their idle list in
> panthor_sched_suspend()
> - Condition of sched_queue_{,delayed_}work fixed to be only when a reset
> isn't pending or in progress.
> - Several typos in comments fixed.
>
> Co-developed-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Looks good, a few minor NITs below, but otherwise.
Reviewed-by: Steven Price <steven.price@arm.com>
> ---
> drivers/gpu/drm/panthor/panthor_sched.c | 3500 +++++++++++++++++++++++
> drivers/gpu/drm/panthor/panthor_sched.h | 48 +
> 2 files changed, 3548 insertions(+)
> create mode 100644 drivers/gpu/drm/panthor/panthor_sched.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_sched.h
>
> diff --git a/drivers/gpu/drm/panthor/panthor_sched.c b/drivers/gpu/drm/panthor/panthor_sched.c
> new file mode 100644
> index 000000000000..b57f4a5f5176
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_sched.c
> @@ -0,0 +1,3500 @@
> +// SPDX-License-Identifier: GPL-2.0 or MIT
> +/* Copyright 2023 Collabora ltd. */
> +
> +#include <drm/panthor_drm.h>
> +#include <drm/drm_drv.h>
> +#include <drm/drm_exec.h>
> +#include <drm/drm_gem_shmem_helper.h>
> +#include <drm/drm_managed.h>
> +#include <drm/gpu_scheduler.h>
> +
> +#include <linux/build_bug.h>
> +#include <linux/clk.h>
> +#include <linux/delay.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/firmware.h>
> +#include <linux/interrupt.h>
> +#include <linux/io.h>
> +#include <linux/iopoll.h>
> +#include <linux/iosys-map.h>
> +#include <linux/module.h>
> +#include <linux/platform_device.h>
> +#include <linux/pm_runtime.h>
> +#include <linux/dma-resv.h>
> +
> +#include "panthor_sched.h"
> +#include "panthor_devfreq.h"
> +#include "panthor_device.h"
> +#include "panthor_gem.h"
> +#include "panthor_heap.h"
> +#include "panthor_regs.h"
> +#include "panthor_gpu.h"
> +#include "panthor_fw.h"
> +#include "panthor_mmu.h"
> +
> +/**
> + * DOC: Scheduler
> + *
> + * Mali CSF hardware adopts a firmware-assisted scheduling model, where
> + * the firmware takes care of scheduling aspects, to some extend.
NIT: s/extend/extent/
> + *
> + * The scheduling happens at the scheduling group level, each group
> + * contains 1 to N queues (N is FW/hardware dependent, and exposed
> + * through the firmware interface). Each queue is assigned a command
> + * stream ring buffer, which serves as a way to get jobs submitted to
> + * the GPU, among other things.
> + *
> + * The firmware can schedule a maximum of M groups (M is FW/hardware
> + * dependent, and exposed through the firmware interface). Passed
> + * this maximum number of groups, the kernel must take care of
> + * rotating the groups passed to the firmware so every group gets
> + * a chance to have his queues scheduled for execution.
> + *
> + * The current implementation only supports with kernel-mode queues.
> + * In other terms, userspace doesn't have access to the ring-buffer.
> + * Instead, userspace passes indirect command stream buffers that are
> + * called from the queue ring-buffer by the kernel using a pre-defined
> + * sequence of command stream instructions to ensure the userspace driver
> + * always gets consistent results (cache maintenance,
> + * synchronization, ...).
> + *
> + * We rely on the drm_gpu_scheduler framework to deal with job
> + * dependencies and submission. As any other driver dealing with a
> + * FW-scheduler, we use the 1:1 entity:scheduler mode, such that each
> + * entity has its own job scheduler. When a job is ready to be executed
> + * (all its dependencies are met), it is pushed to the appropriate
> + * queue ring-buffer, and the group is scheduled for execution if it
> + * wasn't already active.
> + *
> + * Kernel-side group scheduling is timeslice-based. When we have less
> + * groups than there are slots, the periodic tick is disabled and we
> + * just let the FW schedule the active groups. When there are more
> + * groups than slots, we let each group a chance to execute stuff for
> + * a given amount of time, and then re-evaluate and pick new groups
> + * to schedule. The group selection algorithm is based on
> + * priority+round-robin.
> + *
> + * Even though user-mode queues is out of the scope right now, the
> + * current design takes them into account by avoiding any guess on the
> + * group/queue state that would be based on information we wouldn't have
> + * if userspace was in charge of the ring-buffer. That's also one of the
> + * reason we don't do 'cooperative' scheduling (encoding FW group slot
> + * reservation as dma_fence that would be returned from the
> + * drm_gpu_scheduler::prepare_job() hook, and treating group rotation as
> + * a queue of waiters, ordered by job submission order). This approach
> + * would work for kernel-mode queues, but would make user-mode queues a
> + * lot more complicated to retrofit.
> + */
> +
<snip>
> +
> +/**
> + * struct panthor_group - Scheduling group object
> + */
> +struct panthor_group {
> + /** @refcount: Reference count */
> + struct kref refcount;
> +
> + /** @ptdev: Device. */
> + struct panthor_device *ptdev;
> +
> + /** @vm: VM bound to the group. */
> + struct panthor_vm *vm;
> +
> + /** @compute_core_mask: Mask of shader cores that can be used for compute jobs. */
> + u64 compute_core_mask;
> +
> + /** @fragment_core_mask: Mask of shader cores that can be used for fragment jobs. */
> + u64 fragment_core_mask;
> +
> + /** @tiler_core_mask: Mask of tiler cores that can be used for tiler jobs. */
> + u64 tiler_core_mask;
> +
> + /** @max_compute_cores: Maximum number of shader cores used for compute jobs. */
> + u8 max_compute_cores;
> +
> + /** @max_compute_cores: Maximum number of shader cores used for fragment jobs. */
> + u8 max_fragment_cores;
> +
> + /** @max_tiler_cores: Maximum number of tiler cores used for tiler jobs. */
> + u8 max_tiler_cores;
> +
> + /** @priority: Group priority (check panthor_csg_priority). */
> + u8 priority;
> +
> + /** @blocked_queues: Bitmask reflecting the blocked queues. */
> + u32 blocked_queues;
> +
> + /** @idle_queues: Bitmask reflecting the blocked queues. */
s/blocked/idle/
> + u32 idle_queues;
> +
> + /** @fatal_lock: Lock used to protect access to fatal fields. */
> + spinlock_t fatal_lock;
> +
> + /** @fatal_queues: Bitmask reflecting the queues that hit a fatal exception. */
> + u32 fatal_queues;
> +
> + /** @tiler_oom: Mask of queues that have a tiler OOM event to process. */
> + atomic_t tiler_oom;
> +
> + /** @queue_count: Number of queues in this group. */
> + u32 queue_count;
> +
> + /** @queues: Queues owned by this group. */
> + struct panthor_queue *queues[MAX_CS_PER_CSG];
> +
> + /**
> + * @csg_id: ID of the FW group slot.
> + *
> + * -1 when the group is not scheduled/active.
> + */
> + int csg_id;
> +
> + /**
> + * @destroyed: True when the group has been destroyed.
> + *
> + * If a group is destroyed it becomes useless: no further jobs can be submitted
> + * to its queues. We simply wait for all references to be dropped so we can
> + * release the group object.
> + */
> + bool destroyed;
> +
> + /**
> + * @timedout: True when a timeout occurred on any of the queues owned by
> + * this group.
> + *
> + * Timeouts can be reported by drm_sched or by the FW. In any case, any
> + * timeout situation is unrecoverable, and the group becomes useless.
> + * We simply wait for all references to be dropped so we can release the
> + * group object.
> + */
> + bool timedout;
> +
> + /**
> + * @syncobjs: Pool of per-queue synchronization objects.
> + *
> + * One sync object per queue. The position of the sync object is
> + * determined by the queue index.
> + */
> + struct panthor_kernel_bo *syncobjs;
> +
> + /** @state: Group state. */
> + enum panthor_group_state state;
> +
> + /**
> + * @suspend_buf: Suspend buffer.
> + *
> + * Stores the state of the group and its queues when a group is suspended.
> + * Used at resume time to restore the group in its previous state.
> + *
> + * The size of the suspend buffer is exposed through the FW interface.
> + */
> + struct panthor_kernel_bo *suspend_buf;
> +
> + /**
> + * @protm_suspend_buf: Protection mode suspend buffer.
> + *
> + * Stores the state of the group and its queues when a group that's in
> + * protection mode is suspended.
> + *
> + * Used at resume time to restore the group in its previous state.
> + *
> + * The size of the protection mode suspend buffer is exposed through the
> + * FW interface.
> + */
> + struct panthor_kernel_bo *protm_suspend_buf;
> +
> + /** @sync_upd_work: Work used to check/signal job fences. */
> + struct work_struct sync_upd_work;
> +
> + /** @tiler_oom_work: Work used to process tiler OOM events happening on this group. */
> + struct work_struct tiler_oom_work;
> +
> + /** @term_work: Work used to finish the group termination procedure. */
> + struct work_struct term_work;
> +
> + /**
> + * @release_work: Work used to release group resources.
> + *
> + * We need to postpone the group release to avoid a deadlock when
> + * the last ref is released in the tick work.
> + */
> + struct work_struct release_work;
> +
> + /**
> + * @run_node: Node used to insert the group in the
> + * panthor_group::groups::{runnable,idle} and
> + * panthor_group::reset.stopped_groups lists.
> + */
> + struct list_head run_node;
> +
> + /**
> + * @wait_node: Node used to insert the group in the
> + * panthor_group::groups::waiting list.
> + */
> + struct list_head wait_node;
> +};
> +
> +/**
> + * group_queue_work() - Queue a group work
> + * @group: Group to queue the work for.
> + * @wname: Work name.
> + *
> + * Grabs a ref and queue a work item to the scheduler workqueue. If
> + * the work was already queued, we release the reference we grabbed.
> + *
> + * Work callbacks must release the reference we grabbed here.
> + */
> +#define group_queue_work(group, wname) \
> + do { \
> + group_get(group); \
> + if (!queue_work((group)->ptdev->scheduler->wq, &(group)->wname ## _work)) \
> + group_put(group); \
> + } while (0)
> +
> +/**
> + * sched_queue_work() - Queue a scheduler work.
> + * @sched: Scheduler object.
> + * @wname: Work name.
> + *
> + * Conditionally queues a scheduler work if no reset is pending/in-progress.
> + */
> +#define sched_queue_work(sched, wname) \
> + do { \
> + if (!atomic_read(&sched->reset.in_progress) && \
> + !panthor_device_reset_is_pending((sched)->ptdev)) \
> + queue_work((sched)->wq, &(sched)->wname ## _work); \
> + } while (0)
> +
> +/**
> + * sched_queue_delayed_work() - Queue a scheduler delayed work.
> + * @sched: Scheduler object.
> + * @wname: Work name.
> + * @delay: Work delay in jiffies.
> + *
> + * Conditionally queues a scheduler delayed work if no reset is
> + * pending/in-progress.
> + */
> +#define sched_queue_delayed_work(sched, wname, delay) \
> + do { \
> + if (!atomic_read(&sched->reset.in_progress) && \
> + !panthor_device_reset_is_pending((sched)->ptdev)) \
> + mod_delayed_work((sched)->wq, &(sched)->wname ## _work, delay); \
> + } while (0)
> +
> +/*
> + * We currently set the maximum of groups per file to an arbitrary low value.
> + * But this can be updated if we need more.
> + */
> +#define MAX_GROUPS_PER_POOL 128
> +
> +/**
> + * struct panthor_group_pool - Group pool
> + *
> + * Each file get assigned a group pool.
> + */
> +struct panthor_group_pool {
> + /** @xa: Xarray used to manage group handles. */
> + struct xarray xa;
> +};
> +
> +/**
> + * struct panthor_job - Used to manage GPU job
> + */
> +struct panthor_job {
> + /** @base: Inherit from drm_sched_job. */
> + struct drm_sched_job base;
> +
> + /** @refcount: Reference count. */
> + struct kref refcount;
> +
> + /** @group: Group of the queue this job will be pushed to. */
> + struct panthor_group *group;
> +
> + /** @queue_idx: Index of the queue inside @group. */
> + u32 queue_idx;
> +
> + /** @call_info: Information about the userspace command stream call. */
> + struct {
> + /** @start: GPU address of the userspace command stream. */
> + u64 start;
> +
> + /** @size: Size of the userspace command stream. */
> + u32 size;
> +
> + /**
> + * @latest_flush: Flush ID at the time the userspace command
> + * stream was built.
> + *
> + * Needed for the flush reduction mechanism.
> + */
> + u32 latest_flush;
> + } call_info;
> +
> + /** @ringbuf: Position of this job is in the ring buffer. */
> + struct {
> + /** @start: Start offset. */
> + u64 start;
> +
> + /** @end: End offset. */
> + u64 end;
> + } ringbuf;
> +
> + /**
> + * @node: Used to insert the job in the panthor_queue::fence_ctx::in_flight_jobs
> + * list.
> + */
> + struct list_head node;
> +
> + /** @done_fence: Fence signaled when the job is finished or cancelled. */
> + struct dma_fence *done_fence;
> +};
> +
> +static void
> +panthor_queue_put_syncwait_obj(struct panthor_queue *queue)
> +{
> + if (queue->syncwait.kmap) {
> + struct iosys_map map = IOSYS_MAP_INIT_VADDR(queue->syncwait.kmap);
> +
> + drm_gem_vunmap_unlocked(queue->syncwait.obj, &map);
> + queue->syncwait.kmap = NULL;
> + }
> +
> + drm_gem_object_put(queue->syncwait.obj);
> + queue->syncwait.obj = NULL;
> +}
> +
> +static void *
> +panthor_queue_get_syncwait_obj(struct panthor_group *group, struct panthor_queue *queue)
> +{
> + struct panthor_device *ptdev = group->ptdev;
> + struct panthor_gem_object *bo;
> + struct iosys_map map;
> + int ret;
> +
> + if (queue->syncwait.kmap)
> + return queue->syncwait.kmap + queue->syncwait.offset;
> +
> + bo = panthor_vm_get_bo_for_va(group->vm,
> + queue->syncwait.gpu_va,
> + &queue->syncwait.offset);
> + if (drm_WARN_ON(&ptdev->base, IS_ERR_OR_NULL(bo)))
> + goto err_put_syncwait_obj;
> +
> + queue->syncwait.obj = &bo->base.base;
> + ret = drm_gem_vmap_unlocked(queue->syncwait.obj, &map);
> + if (drm_WARN_ON(&ptdev->base, ret))
> + goto err_put_syncwait_obj;
> +
> + queue->syncwait.kmap = map.vaddr;
> + if (drm_WARN_ON(&ptdev->base, !queue->syncwait.kmap))
> + goto err_put_syncwait_obj;
> +
> + return queue->syncwait.kmap + queue->syncwait.offset;
> +
> +err_put_syncwait_obj:
> + panthor_queue_put_syncwait_obj(queue);
> + return NULL;
> +}
> +
> +static void group_free_queue(struct panthor_group *group, struct panthor_queue *queue)
> +{
> + if (IS_ERR_OR_NULL(queue))
> + return;
> +
> + if (queue->entity.fence_context)
> + drm_sched_entity_destroy(&queue->entity);
> +
> + if (queue->scheduler.ops)
> + drm_sched_fini(&queue->scheduler);
> +
> + panthor_queue_put_syncwait_obj(queue);
> +
> + panthor_kernel_bo_destroy(group->vm, queue->ringbuf);
> + panthor_kernel_bo_destroy(panthor_fw_vm(group->ptdev), queue->iface.mem);
> +
> + kfree(queue);
> +}
> +
> +static void group_release_work(struct work_struct *work)
> +{
> + struct panthor_group *group = container_of(work,
> + struct panthor_group,
> + release_work);
> + struct panthor_device *ptdev = group->ptdev;
> + u32 i;
> +
> + for (i = 0; i < group->queue_count; i++)
> + group_free_queue(group, group->queues[i]);
> +
> + panthor_kernel_bo_destroy(panthor_fw_vm(ptdev), group->suspend_buf);
> + panthor_kernel_bo_destroy(panthor_fw_vm(ptdev), group->protm_suspend_buf);
> +
> + if (!IS_ERR_OR_NULL(group->syncobjs))
No need for this check - panthor_kernel_bo_destroy() does it.
> + panthor_kernel_bo_destroy(group->vm, group->syncobjs);
> +
> + panthor_vm_put(group->vm);
> + kfree(group);
> +}
Thanks,
Steve
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 11/14] drm/panthor: Add the driver frontend block
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (9 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 10/14] drm/panthor: Add the scheduler " Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-01-23 16:29 ` Heiko Stübner
2024-01-22 16:30 ` [PATCH v4 12/14] drm/panthor: Allow driver compilation Boris Brezillon
` (6 subsequent siblings)
17 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
This is the last piece missing to expose the driver to the outside
world.
This is basically a wrapper between the ioctls and the other logical
blocks.
v4:
- Add an ioctl to let the UMD query the VM state
- Fix kernel doc
- Let panthor_device_init() call panthor_device_init()
- Fix cleanup ordering in the panthor_init() error path
- Add Steve's and Liviu's R-b
v3:
- Add acks for the MIT/GPL2 relicensing
- Fix 32-bit support
- Account for panthor_vm and panthor_sched changes
- Simplify the resv preparation/update logic
- Use a linked list rather than xarray for list of signals.
- Simplify panthor_get_uobj_array by returning the newly allocated
array.
- Drop the "DOC" for job submission helpers and move the relevant
comments to panthor_ioctl_group_submit().
- Add helpers sync_op_is_signal()/sync_op_is_wait().
- Simplify return type of panthor_submit_ctx_add_sync_signal() and
panthor_submit_ctx_get_sync_signal().
- Drop WARN_ON from panthor_submit_ctx_add_job().
- Fix typos in comments.
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
---
drivers/gpu/drm/panthor/panthor_drv.c | 1470 +++++++++++++++++++++++++
1 file changed, 1470 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/panthor_drv.c
diff --git a/drivers/gpu/drm/panthor/panthor_drv.c b/drivers/gpu/drm/panthor/panthor_drv.c
new file mode 100644
index 000000000000..207aacaccd39
--- /dev/null
+++ b/drivers/gpu/drm/panthor/panthor_drv.c
@@ -0,0 +1,1470 @@
+// SPDX-License-Identifier: GPL-2.0 or MIT
+/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
+/* Copyright 2019 Linaro, Ltd., Rob Herring <robh@kernel.org> */
+/* Copyright 2019 Collabora ltd. */
+
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/of_platform.h>
+#include <linux/pagemap.h>
+#include <linux/pm_runtime.h>
+
+#include <drm/drm_drv.h>
+#include <drm/drm_exec.h>
+#include <drm/drm_ioctl.h>
+#include <drm/drm_syncobj.h>
+#include <drm/drm_utils.h>
+#include <drm/drm_debugfs.h>
+#include <drm/gpu_scheduler.h>
+#include <drm/panthor_drm.h>
+
+#include "panthor_sched.h"
+#include "panthor_device.h"
+#include "panthor_gem.h"
+#include "panthor_heap.h"
+#include "panthor_fw.h"
+#include "panthor_mmu.h"
+#include "panthor_gpu.h"
+#include "panthor_regs.h"
+
+/**
+ * DOC: user <-> kernel object copy helpers.
+ */
+
+/**
+ * panthor_set_uobj() - Copy kernel object to user object.
+ * @usr_ptr: Users pointer.
+ * @usr_size: Size of the user object.
+ * @min_size: Minimum size for this object.
+ * @kern_size: Size of the kernel object.
+ * @in: Address of the kernel object to copy.
+ *
+ * Helper automating kernel -> user object copies.
+ *
+ * Don't use this function directly, use PANTHOR_UOBJ_SET() instead.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+panthor_set_uobj(u64 usr_ptr, u32 usr_size, u32 min_size, u32 kern_size, const void *in)
+{
+ /* User size shouldn't be smaller than the minimal object size. */
+ if (usr_size < min_size)
+ return -EINVAL;
+
+ if (copy_to_user(u64_to_user_ptr(usr_ptr), in, min_t(u32, usr_size, kern_size)))
+ return -EFAULT;
+
+ /* When the kernel object is smaller than the user object, we fill the gap with
+ * zeros.
+ */
+ if (usr_size > kern_size &&
+ clear_user(u64_to_user_ptr(usr_ptr + kern_size), usr_size - kern_size)) {
+ return -EFAULT;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_get_uobj_array() - Copy a user object array into a kernel accessible object array.
+ * @in: The object array to copy.
+ * @min_stride: Minimum array stride.
+ * @obj_size: Kernel object size.
+ *
+ * Helper automating user -> kernel object copies.
+ *
+ * Don't use this function directly, use PANTHOR_UOBJ_GET_ARRAY() instead.
+ *
+ * Return: newly allocated object array or an ERR_PTR on error.
+ */
+static void *
+panthor_get_uobj_array(const struct drm_panthor_obj_array *in, u32 min_stride,
+ u32 obj_size)
+{
+ int ret = 0;
+ void *out_alloc;
+
+ /* User stride must be at least the minimum object size, otherwise it might
+ * lack useful information.
+ */
+ if (in->stride < min_stride)
+ return ERR_PTR(-EINVAL);
+
+ if (!in->count)
+ return NULL;
+
+ out_alloc = kvmalloc_array(in->count, obj_size, GFP_KERNEL);
+ if (!out_alloc)
+ return ERR_PTR(-ENOMEM);
+
+ if (obj_size == in->stride) {
+ /* Fast path when user/kernel have the same uAPI header version. */
+ if (copy_from_user(out_alloc, u64_to_user_ptr(in->array),
+ (unsigned long)obj_size * in->count))
+ ret = -EFAULT;
+ } else {
+ void __user *in_ptr = u64_to_user_ptr(in->array);
+ void *out_ptr = out_alloc;
+
+ /* If the sizes differ, we need to copy elements one by one. */
+ for (u32 i = 0; i < in->count; i++) {
+ ret = copy_struct_from_user(out_ptr, obj_size, in_ptr, in->stride);
+ if (ret)
+ break;
+
+ out_ptr += obj_size;
+ in_ptr += in->stride;
+ }
+ }
+
+ if (ret) {
+ kvfree(out_alloc);
+ return ERR_PTR(ret);
+ }
+
+ return out_alloc;
+}
+
+/**
+ * PANTHOR_UOBJ_MIN_SIZE_INTERNAL() - Get the minimum user object size
+ * @_typename: Object type.
+ * @_last_mandatory_field: Last mandatory field.
+ *
+ * Get the minimum user object size based on the last mandatory field name,
+ * A.K.A, the name of the last field of the structure at the time this
+ * structure was added to the uAPI.
+ *
+ * Don't use directly, use PANTHOR_UOBJ_DECL() instead.
+ */
+#define PANTHOR_UOBJ_MIN_SIZE_INTERNAL(_typename, _last_mandatory_field) \
+ (offsetof(_typename, _last_mandatory_field) + \
+ sizeof(((_typename *)NULL)->_last_mandatory_field))
+
+/**
+ * PANTHOR_UOBJ_DECL() - Declare a new uAPI object whose subject to
+ * evolutions.
+ * @_typename: Object type.
+ * @_last_mandatory_field: Last mandatory field.
+ *
+ * Should be used to extend the PANTHOR_UOBJ_MIN_SIZE() list.
+ */
+#define PANTHOR_UOBJ_DECL(_typename, _last_mandatory_field) \
+ _typename : PANTHOR_UOBJ_MIN_SIZE_INTERNAL(_typename, _last_mandatory_field)
+
+/**
+ * PANTHOR_UOBJ_MIN_SIZE() - Get the minimum size of a given uAPI object
+ * @_obj_name: Object to get the minimum size of.
+ *
+ * Don't use this macro directly, it's automatically called by
+ * PANTHOR_UOBJ_{SET,GET_ARRAY}().
+ */
+#define PANTHOR_UOBJ_MIN_SIZE(_obj_name) \
+ _Generic(_obj_name, \
+ PANTHOR_UOBJ_DECL(struct drm_panthor_gpu_info, tiler_present), \
+ PANTHOR_UOBJ_DECL(struct drm_panthor_csif_info, pad), \
+ PANTHOR_UOBJ_DECL(struct drm_panthor_sync_op, timeline_value), \
+ PANTHOR_UOBJ_DECL(struct drm_panthor_queue_submit, syncs), \
+ PANTHOR_UOBJ_DECL(struct drm_panthor_queue_create, ringbuf_size), \
+ PANTHOR_UOBJ_DECL(struct drm_panthor_vm_bind_op, syncs))
+
+/**
+ * PANTHOR_UOBJ_SET() - Copy a kernel object to a user object.
+ * @_dest_usr_ptr: User pointer to copy to.
+ * @_usr_size: Size of the user object.
+ * @_src_obj: Kernel object to copy (not a pointer).
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+#define PANTHOR_UOBJ_SET(_dest_usr_ptr, _usr_size, _src_obj) \
+ panthor_set_uobj(_dest_usr_ptr, _usr_size, \
+ PANTHOR_UOBJ_MIN_SIZE(_src_obj), \
+ sizeof(_src_obj), &(_src_obj))
+
+/**
+ * PANTHOR_UOBJ_GET_ARRAY() - Copy a user object array to a kernel accessible
+ * object array.
+ * @_dest_array: Local variable that will hold the newly allocated kernel
+ * object array.
+ * @_uobj_array: The drm_panthor_obj_array object describing the user object
+ * array.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+#define PANTHOR_UOBJ_GET_ARRAY(_dest_array, _uobj_array) \
+ ({ \
+ typeof(_dest_array) _tmp; \
+ _tmp = panthor_get_uobj_array(_uobj_array, \
+ PANTHOR_UOBJ_MIN_SIZE((_dest_array)[0]), \
+ sizeof((_dest_array)[0])); \
+ if (!IS_ERR(_tmp)) \
+ _dest_array = _tmp; \
+ PTR_ERR_OR_ZERO(_tmp); \
+ })
+
+/**
+ * struct panthor_sync_signal - Represent a synchronization object point to attach
+ * our job fence to.
+ *
+ * This structure is here to keep track of fences that are currently bound to
+ * a specific syncobj point.
+ *
+ * At the beginning of a job submission, the fence
+ * is retrieved from the syncobj itself, and can be NULL if no fence was attached
+ * to this point.
+ *
+ * At the end, it points to the fence of the last job that had a
+ * %DRM_PANTHOR_SYNC_OP_SIGNAL on this syncobj.
+ *
+ * With jobs being submitted in batches, the fence might change several times during
+ * the process, allowing one job to wait on a job that's part of the same submission
+ * but appears earlier in the drm_panthor_group_submit::queue_submits array.
+ */
+struct panthor_sync_signal {
+ /** @node: list_head to track signal ops within a submit operation */
+ struct list_head node;
+
+ /** @handle: The syncobj handle. */
+ u32 handle;
+
+ /**
+ * @point: The syncobj point.
+ *
+ * Zero for regular syncobjs, and non-zero for timeline syncobjs.
+ */
+ u64 point;
+
+ /**
+ * @syncobj: The sync object pointed by @handle.
+ */
+ struct drm_syncobj *syncobj;
+
+ /**
+ * @chain: Chain object used to link the new fence to an existing
+ * timeline syncobj.
+ *
+ * NULL for regular syncobj, non-NULL for timeline syncobjs.
+ */
+ struct dma_fence_chain *chain;
+
+ /**
+ * @fence: The fence to assign to the syncobj or syncobj-point.
+ */
+ struct dma_fence *fence;
+};
+
+/**
+ * struct panthor_job_ctx - Job context
+ */
+struct panthor_job_ctx {
+ /** @job: The job that is about to be submitted to drm_sched. */
+ struct drm_sched_job *job;
+
+ /** @syncops: Array of sync operations. */
+ struct drm_panthor_sync_op *syncops;
+
+ /** @syncop_count: Number of sync operations. */
+ u32 syncop_count;
+};
+
+/**
+ * struct panthor_submit_ctx - Submission context
+ *
+ * Anything that's related to a submission (%DRM_IOCTL_PANTHOR_VM_BIND or
+ * %DRM_IOCTL_PANTHOR_GROUP_SUBMIT) is kept here, so we can automate the
+ * initialization and cleanup steps.
+ */
+struct panthor_submit_ctx {
+ /** @file: DRM file this submission happens on. */
+ struct drm_file *file;
+
+ /**
+ * @signals: List of struct panthor_sync_signal.
+ *
+ * %DRM_PANTHOR_SYNC_OP_SIGNAL operations will be recorded here,
+ * and %DRM_PANTHOR_SYNC_OP_WAIT will first check if an entry
+ * matching the syncobj+point exists before calling
+ * drm_syncobj_find_fence(). This allows us to describe dependencies
+ * existing between jobs that are part of the same batch.
+ */
+ struct list_head signals;
+
+ /** @jobs: Array of jobs. */
+ struct panthor_job_ctx *jobs;
+
+ /** @job_count: Number of entries in the @jobs array. */
+ u32 job_count;
+
+ /** @exec: drm_exec context used to acquire and prepare resv objects. */
+ struct drm_exec exec;
+};
+
+#define PANTHOR_SYNC_OP_FLAGS_MASK \
+ (DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_MASK | DRM_PANTHOR_SYNC_OP_SIGNAL)
+
+static bool sync_op_is_signal(const struct drm_panthor_sync_op *sync_op)
+{
+ return !!(sync_op->flags & DRM_PANTHOR_SYNC_OP_SIGNAL);
+}
+
+static bool sync_op_is_wait(const struct drm_panthor_sync_op *sync_op)
+{
+ /* Note that DRM_PANTHOR_SYNC_OP_WAIT == 0 */
+ return !(sync_op->flags & DRM_PANTHOR_SYNC_OP_SIGNAL);
+}
+
+/**
+ * panthor_check_sync_op() - Check drm_panthor_sync_op fields
+ * @sync_op: The sync operation to check.
+ *
+ * Return: 0 on success, -EINVAL otherwise.
+ */
+static int
+panthor_check_sync_op(const struct drm_panthor_sync_op *sync_op)
+{
+ u8 handle_type;
+
+ if (sync_op->flags & ~PANTHOR_SYNC_OP_FLAGS_MASK)
+ return -EINVAL;
+
+ handle_type = sync_op->flags & DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_MASK;
+ if (handle_type != DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_SYNCOBJ &&
+ handle_type != DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_TIMELINE_SYNCOBJ)
+ return -EINVAL;
+
+ if (handle_type == DRM_PANTHOR_SYNC_OP_HANDLE_TYPE_SYNCOBJ &&
+ sync_op->timeline_value != 0)
+ return -EINVAL;
+
+ return 0;
+}
+
+/**
+ * panthor_sync_signal_free() - Release resources and free a panthor_sync_signal object
+ * @sig_sync: Signal object to free.
+ */
+static void
+panthor_sync_signal_free(struct panthor_sync_signal *sig_sync)
+{
+ if (!sig_sync)
+ return;
+
+ drm_syncobj_put(sig_sync->syncobj);
+ dma_fence_chain_free(sig_sync->chain);
+ dma_fence_put(sig_sync->fence);
+ kfree(sig_sync);
+}
+
+/**
+ * panthor_submit_ctx_add_sync_signal() - Add a signal operation to a submit context
+ * @ctx: Context to add the signal operation to.
+ * @handle: Syncobj handle.
+ * @point: Syncobj point.
+ *
+ * Return: 0 on success, otherwise negative error value.
+ */
+static int
+panthor_submit_ctx_add_sync_signal(struct panthor_submit_ctx *ctx, u32 handle, u64 point)
+{
+ struct panthor_sync_signal *sig_sync;
+ struct dma_fence *cur_fence;
+ int ret;
+
+ sig_sync = kzalloc(sizeof(*sig_sync), GFP_KERNEL);
+ if (!sig_sync)
+ return -ENOMEM;
+
+ sig_sync->handle = handle;
+ sig_sync->point = point;
+
+ if (point > 0) {
+ sig_sync->chain = dma_fence_chain_alloc();
+ if (!sig_sync->chain) {
+ ret = -ENOMEM;
+ goto err_free_sig_sync;
+ }
+ }
+
+ sig_sync->syncobj = drm_syncobj_find(ctx->file, handle);
+ if (!sig_sync->syncobj) {
+ ret = -EINVAL;
+ goto err_free_sig_sync;
+ }
+
+ /* Retrieve the current fence attached to that point. It's
+ * perfectly fine to get a NULL fence here, it just means there's
+ * no fence attached to that point yet.
+ */
+ if (!drm_syncobj_find_fence(ctx->file, handle, point, 0, &cur_fence))
+ sig_sync->fence = cur_fence;
+
+ list_add_tail(&sig_sync->node, &ctx->signals);
+
+ return 0;
+
+err_free_sig_sync:
+ panthor_sync_signal_free(sig_sync);
+ return ret;
+}
+
+/**
+ * panthor_submit_ctx_search_sync_signal() - Search an existing signal operation in a
+ * submit context.
+ * @ctx: Context to search the signal operation in.
+ * @handle: Syncobj handle.
+ * @point: Syncobj point.
+ *
+ * Return: A valid panthor_sync_signal object if found, NULL otherwise.
+ */
+static struct panthor_sync_signal *
+panthor_submit_ctx_search_sync_signal(struct panthor_submit_ctx *ctx, u32 handle, u64 point)
+{
+ struct panthor_sync_signal *sig_sync;
+
+ list_for_each_entry(sig_sync, &ctx->signals, node) {
+ if (handle == sig_sync->handle && point == sig_sync->point)
+ return sig_sync;
+ }
+
+ return NULL;
+}
+
+/**
+ * panthor_submit_ctx_add_job() - Add a job to a submit context
+ * @ctx: Context to search the signal operation in.
+ * @idx: Index of the job in the context.
+ * @job: Job to add.
+ * @syncs: Sync operations provided by userspace.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+panthor_submit_ctx_add_job(struct panthor_submit_ctx *ctx, u32 idx,
+ struct drm_sched_job *job,
+ const struct drm_panthor_obj_array *syncs)
+{
+ int ret;
+
+ ctx->jobs[idx].job = job;
+
+ ret = PANTHOR_UOBJ_GET_ARRAY(ctx->jobs[idx].syncops, syncs);
+ if (ret)
+ return ret;
+
+ ctx->jobs[idx].syncop_count = syncs->count;
+ return 0;
+}
+
+/**
+ * panthor_submit_ctx_get_sync_signal() - Search signal operation and add one if none was found.
+ * @ctx: Context to search the signal operation in.
+ * @handle: Syncobj handle.
+ * @point: Syncobj point.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+panthor_submit_ctx_get_sync_signal(struct panthor_submit_ctx *ctx, u32 handle, u64 point)
+{
+ struct panthor_sync_signal *sig_sync;
+
+ sig_sync = panthor_submit_ctx_search_sync_signal(ctx, handle, point);
+ if (sig_sync)
+ return 0;
+
+ return panthor_submit_ctx_add_sync_signal(ctx, handle, point);
+}
+
+/**
+ * panthor_submit_ctx_update_job_sync_signal_fences() - Update fences
+ * on the signal operations specified by a job.
+ * @ctx: Context to search the signal operation in.
+ * @job_idx: Index of the job to operate on.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+panthor_submit_ctx_update_job_sync_signal_fences(struct panthor_submit_ctx *ctx,
+ u32 job_idx)
+{
+ struct panthor_device *ptdev = container_of(ctx->file->minor->dev,
+ struct panthor_device,
+ base);
+ struct dma_fence *done_fence = &ctx->jobs[job_idx].job->s_fence->finished;
+ const struct drm_panthor_sync_op *sync_ops = ctx->jobs[job_idx].syncops;
+ u32 sync_op_count = ctx->jobs[job_idx].syncop_count;
+
+ for (u32 i = 0; i < sync_op_count; i++) {
+ struct dma_fence *old_fence;
+ struct panthor_sync_signal *sig_sync;
+
+ if (!sync_op_is_signal(&sync_ops[i]))
+ continue;
+
+ sig_sync = panthor_submit_ctx_search_sync_signal(ctx, sync_ops[i].handle,
+ sync_ops[i].timeline_value);
+ if (drm_WARN_ON(&ptdev->base, !sig_sync))
+ return -EINVAL;
+
+ old_fence = sig_sync->fence;
+ sig_sync->fence = dma_fence_get(done_fence);
+ dma_fence_put(old_fence);
+
+ if (drm_WARN_ON(&ptdev->base, !sig_sync->fence))
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_submit_ctx_collect_job_signal_ops() - Iterate over all job signal operations
+ * and add them to the context.
+ * @ctx: Context to search the signal operation in.
+ * @job_idx: Index of the job to operate on.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+panthor_submit_ctx_collect_job_signal_ops(struct panthor_submit_ctx *ctx,
+ u32 job_idx)
+{
+ const struct drm_panthor_sync_op *sync_ops = ctx->jobs[job_idx].syncops;
+ u32 sync_op_count = ctx->jobs[job_idx].syncop_count;
+
+ for (u32 i = 0; i < sync_op_count; i++) {
+ int ret;
+
+ if (!sync_op_is_signal(&sync_ops[i]))
+ continue;
+
+ ret = panthor_check_sync_op(&sync_ops[i]);
+ if (ret)
+ return ret;
+
+ ret = panthor_submit_ctx_get_sync_signal(ctx,
+ sync_ops[i].handle,
+ sync_ops[i].timeline_value);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_submit_ctx_push_fences() - Iterate over the signal array, and for each entry, push
+ * the currently assigned fence to the associated syncobj.
+ * @ctx: Context to push fences on.
+ *
+ * This is the last step of a submission procedure, and is done once we know the submission
+ * is effective and job fences are guaranteed to be signaled in finite time.
+ */
+static void
+panthor_submit_ctx_push_fences(struct panthor_submit_ctx *ctx)
+{
+ struct panthor_sync_signal *sig_sync;
+
+ list_for_each_entry(sig_sync, &ctx->signals, node) {
+ if (sig_sync->chain) {
+ drm_syncobj_add_point(sig_sync->syncobj, sig_sync->chain,
+ sig_sync->fence, sig_sync->point);
+ sig_sync->chain = NULL;
+ } else {
+ drm_syncobj_replace_fence(sig_sync->syncobj, sig_sync->fence);
+ }
+ }
+}
+
+/**
+ * panthor_submit_ctx_add_sync_deps_to_job() - Add sync wait operations as
+ * job dependencies.
+ * @ctx: Submit context.
+ * @job_idx: Index of the job to operate on.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+panthor_submit_ctx_add_sync_deps_to_job(struct panthor_submit_ctx *ctx,
+ u32 job_idx)
+{
+ struct panthor_device *ptdev = container_of(ctx->file->minor->dev,
+ struct panthor_device,
+ base);
+ const struct drm_panthor_sync_op *sync_ops = ctx->jobs[job_idx].syncops;
+ struct drm_sched_job *job = ctx->jobs[job_idx].job;
+ u32 sync_op_count = ctx->jobs[job_idx].syncop_count;
+ int ret = 0;
+
+ for (u32 i = 0; i < sync_op_count; i++) {
+ struct panthor_sync_signal *sig_sync;
+ struct dma_fence *fence;
+
+ if (!sync_op_is_wait(&sync_ops[i]))
+ continue;
+
+ ret = panthor_check_sync_op(&sync_ops[i]);
+ if (ret)
+ return ret;
+
+ sig_sync = panthor_submit_ctx_search_sync_signal(ctx, sync_ops[i].handle,
+ sync_ops[i].timeline_value);
+ if (sig_sync) {
+ if (drm_WARN_ON(&ptdev->base, !sig_sync->fence))
+ return -EINVAL;
+
+ fence = dma_fence_get(sig_sync->fence);
+ } else {
+ ret = drm_syncobj_find_fence(ctx->file, sync_ops[i].handle,
+ sync_ops[i].timeline_value,
+ 0, &fence);
+ if (ret)
+ return ret;
+ }
+
+ ret = drm_sched_job_add_dependency(job, fence);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_submit_ctx_collect_jobs_signal_ops() - Collect all signal operations
+ * and add them to the submit context.
+ * @ctx: Submit context.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+panthor_submit_ctx_collect_jobs_signal_ops(struct panthor_submit_ctx *ctx)
+{
+ for (u32 i = 0; i < ctx->job_count; i++) {
+ int ret;
+
+ ret = panthor_submit_ctx_collect_job_signal_ops(ctx, i);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_submit_ctx_add_deps_and_arm_jobs() - Add jobs dependencies and arm jobs
+ * @ctx: Submit context.
+ *
+ * Must be called after the resv preparation has been taken care of.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int
+panthor_submit_ctx_add_deps_and_arm_jobs(struct panthor_submit_ctx *ctx)
+{
+ for (u32 i = 0; i < ctx->job_count; i++) {
+ int ret;
+
+ ret = panthor_submit_ctx_add_sync_deps_to_job(ctx, i);
+ if (ret)
+ return ret;
+
+ drm_sched_job_arm(ctx->jobs[i].job);
+
+ ret = panthor_submit_ctx_update_job_sync_signal_fences(ctx, i);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+/**
+ * panthor_submit_ctx_push_jobs() - Push jobs to their scheduling entities.
+ * @ctx: Submit context.
+ * @upd_resvs: Callback used to update reservation objects that were previously
+ * preapred.
+ */
+static void
+panthor_submit_ctx_push_jobs(struct panthor_submit_ctx *ctx,
+ void (*upd_resvs)(struct drm_exec *, struct drm_sched_job *))
+{
+ for (u32 i = 0; i < ctx->job_count; i++) {
+ upd_resvs(&ctx->exec, ctx->jobs[i].job);
+ drm_sched_entity_push_job(ctx->jobs[i].job);
+
+ /* Job is owned by the scheduler now. */
+ ctx->jobs[i].job = NULL;
+ }
+
+ panthor_submit_ctx_push_fences(ctx);
+}
+
+/**
+ * panthor_submit_ctx_init() - Initializes a submission context
+ * @ctx: Submit context to initialize.
+ * @file: drm_file this submission happens on.
+ * @job_count: Number of jobs that will be submitted.
+ *
+ * Return: 0 on success, a negative error code otherwise.
+ */
+static int panthor_submit_ctx_init(struct panthor_submit_ctx *ctx,
+ struct drm_file *file, u32 job_count)
+{
+ ctx->jobs = kvmalloc_array(job_count, sizeof(*ctx->jobs),
+ GFP_KERNEL | __GFP_ZERO);
+ if (!ctx->jobs)
+ return -ENOMEM;
+
+ ctx->file = file;
+ ctx->job_count = job_count;
+ INIT_LIST_HEAD(&ctx->signals);
+ drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT | DRM_EXEC_IGNORE_DUPLICATES);
+ return 0;
+}
+
+/**
+ * panthor_submit_ctx_cleanup() - Cleanup a submission context
+ * @ctx: Submit context to cleanup.
+ * @job_put: Job put callback.
+ */
+static void panthor_submit_ctx_cleanup(struct panthor_submit_ctx *ctx,
+ void (*job_put)(struct drm_sched_job *))
+{
+ struct panthor_sync_signal *sig_sync, *tmp;
+ unsigned long i;
+
+ drm_exec_fini(&ctx->exec);
+
+ list_for_each_entry_safe(sig_sync, tmp, &ctx->signals, node)
+ panthor_sync_signal_free(sig_sync);
+
+ for (i = 0; i < ctx->job_count; i++) {
+ job_put(ctx->jobs[i].job);
+ kvfree(ctx->jobs[i].syncops);
+ }
+
+ kvfree(ctx->jobs);
+}
+
+static int panthor_ioctl_dev_query(struct drm_device *ddev, void *data, struct drm_file *file)
+{
+ struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
+ struct drm_panthor_dev_query *args = data;
+
+ if (!args->pointer) {
+ switch (args->type) {
+ case DRM_PANTHOR_DEV_QUERY_GPU_INFO:
+ args->size = sizeof(ptdev->gpu_info);
+ return 0;
+
+ case DRM_PANTHOR_DEV_QUERY_CSIF_INFO:
+ args->size = sizeof(ptdev->csif_info);
+ return 0;
+
+ default:
+ return -EINVAL;
+ }
+ }
+
+ switch (args->type) {
+ case DRM_PANTHOR_DEV_QUERY_GPU_INFO:
+ return PANTHOR_UOBJ_SET(args->pointer, args->size, ptdev->gpu_info);
+
+ case DRM_PANTHOR_DEV_QUERY_CSIF_INFO:
+ return PANTHOR_UOBJ_SET(args->pointer, args->size, ptdev->csif_info);
+
+ default:
+ return -EINVAL;
+ }
+}
+
+#define PANTHOR_VM_CREATE_FLAGS 0
+
+static int panthor_ioctl_vm_create(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_vm_create *args = data;
+ int cookie, ret;
+
+ if (!drm_dev_enter(ddev, &cookie))
+ return -ENODEV;
+
+ ret = panthor_vm_pool_create_vm(ptdev, pfile->vms, args);
+ if (ret >= 0) {
+ args->id = ret;
+ ret = 0;
+ }
+
+ drm_dev_exit(cookie);
+ return ret;
+}
+
+static int panthor_ioctl_vm_destroy(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_vm_destroy *args = data;
+
+ if (args->pad)
+ return -EINVAL;
+
+ return panthor_vm_pool_destroy_vm(pfile->vms, args->id);
+}
+
+#define PANTHOR_BO_FLAGS DRM_PANTHOR_BO_NO_MMAP
+
+static int panthor_ioctl_bo_create(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_bo_create *args = data;
+ struct panthor_vm *vm = NULL;
+ int cookie, ret;
+
+ if (!drm_dev_enter(ddev, &cookie))
+ return -ENODEV;
+
+ if (!args->size || args->pad ||
+ (args->flags & ~PANTHOR_BO_FLAGS)) {
+ ret = -EINVAL;
+ goto out_dev_exit;
+ }
+
+ if (args->exclusive_vm_id) {
+ vm = panthor_vm_pool_get_vm(pfile->vms, args->exclusive_vm_id);
+ if (!vm) {
+ ret = -EINVAL;
+ goto out_dev_exit;
+ }
+ }
+
+ ret = panthor_gem_create_with_handle(file, ddev, vm, args->size,
+ args->flags, &args->handle);
+
+ panthor_vm_put(vm);
+
+out_dev_exit:
+ drm_dev_exit(cookie);
+ return ret;
+}
+
+static int panthor_ioctl_bo_mmap_offset(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct drm_panthor_bo_mmap_offset *args = data;
+ struct drm_gem_object *obj;
+ int ret;
+
+ if (args->pad)
+ return -EINVAL;
+
+ obj = drm_gem_object_lookup(file, args->handle);
+ if (!obj)
+ return -ENOENT;
+
+ ret = drm_gem_create_mmap_offset(obj);
+ if (ret)
+ goto out;
+
+ args->offset = drm_vma_node_offset_addr(&obj->vma_node);
+
+out:
+ drm_gem_object_put(obj);
+ return ret;
+}
+
+static int panthor_ioctl_group_submit(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_group_submit *args = data;
+ struct drm_panthor_queue_submit *jobs_args;
+ struct panthor_submit_ctx ctx;
+ int ret = 0, cookie;
+
+ if (args->pad)
+ return -EINVAL;
+
+ if (!drm_dev_enter(ddev, &cookie))
+ return -ENODEV;
+
+ ret = PANTHOR_UOBJ_GET_ARRAY(jobs_args, &args->queue_submits);
+ if (ret)
+ goto out_dev_exit;
+
+ ret = panthor_submit_ctx_init(&ctx, file, args->queue_submits.count);
+ if (ret)
+ goto out_free_jobs_args;
+
+ /* Create jobs and attach sync operations */
+ for (u32 i = 0; i < args->queue_submits.count; i++) {
+ const struct drm_panthor_queue_submit *qsubmit = &jobs_args[i];
+ struct drm_sched_job *job;
+
+ job = panthor_job_create(pfile, args->group_handle, qsubmit);
+ if (IS_ERR(job)) {
+ ret = PTR_ERR(job);
+ goto out_cleanup_submit_ctx;
+ }
+
+ ret = panthor_submit_ctx_add_job(&ctx, i, job, &qsubmit->syncs);
+ if (ret)
+ goto out_cleanup_submit_ctx;
+ }
+
+ /*
+ * Collect signal operations on all jobs, such that each job can pick
+ * from it for its dependencies and update the fence to signal when the
+ * job is submitted.
+ */
+ ret = panthor_submit_ctx_collect_jobs_signal_ops(&ctx);
+ if (ret)
+ goto out_cleanup_submit_ctx;
+
+ /*
+ * We acquire/prepare revs on all jobs before proceeding with the
+ * dependency registration.
+ *
+ * This is solving two problems:
+ * 1. drm_sched_job_arm() and drm_sched_entity_push_job() must be
+ * protected by a lock to make sure no concurrent access to the same
+ * entity get interleaved, which would mess up with the fence seqno
+ * ordering. Luckily, one of the resv being acquired is the VM resv,
+ * and a scheduling entity is only bound to a single VM. As soon as
+ * we acquire the VM resv, we should be safe.
+ * 2. Jobs might depend on fences that were issued by previous jobs in
+ * the same batch, so we can't add dependencies on all jobs before
+ * arming previous jobs and registering the fence to the signal
+ * array, otherwise we might miss dependencies, or point to an
+ * outdated fence.
+ */
+ if (args->queue_submits.count > 0) {
+ /* All jobs target the same group, so they also point to the same VM. */
+ struct panthor_vm *vm = panthor_job_vm(ctx.jobs[0].job);
+
+ drm_exec_until_all_locked(&ctx.exec) {
+ ret = panthor_vm_prepare_mapped_bos_resvs(&ctx.exec, vm,
+ args->queue_submits.count);
+ }
+
+ if (ret)
+ goto out_cleanup_submit_ctx;
+ }
+
+ /*
+ * Now that resvs are locked/prepared, we can iterate over each job to
+ * add the dependencies, arm the job fence, register the job fence to
+ * the signal array.
+ */
+ ret = panthor_submit_ctx_add_deps_and_arm_jobs(&ctx);
+ if (ret)
+ goto out_cleanup_submit_ctx;
+
+ /* Nothing can fail after that point, so we can make our job fences
+ * visible to the outside world. Push jobs and set the job fences to
+ * the resv slots we reserved. This also pushes the fences to the
+ * syncobjs that are part of the signal array.
+ */
+ panthor_submit_ctx_push_jobs(&ctx, panthor_job_update_resvs);
+
+out_cleanup_submit_ctx:
+ panthor_submit_ctx_cleanup(&ctx, panthor_job_put);
+
+out_free_jobs_args:
+ kvfree(jobs_args);
+
+out_dev_exit:
+ drm_dev_exit(cookie);
+ return ret;
+}
+
+static int panthor_ioctl_group_destroy(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_group_destroy *args = data;
+
+ if (args->pad)
+ return -EINVAL;
+
+ return panthor_group_destroy(pfile, args->group_handle);
+}
+
+static int panthor_ioctl_group_create(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_group_create *args = data;
+ struct drm_panthor_queue_create *queue_args;
+ int ret;
+
+ if (!args->queues.count)
+ return -EINVAL;
+
+ ret = PANTHOR_UOBJ_GET_ARRAY(queue_args, &args->queues);
+ if (ret)
+ return ret;
+
+ ret = panthor_group_create(pfile, args, queue_args);
+ if (ret >= 0) {
+ args->group_handle = ret;
+ ret = 0;
+ }
+
+ kvfree(queue_args);
+ return ret;
+}
+
+static int panthor_ioctl_group_get_state(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_group_get_state *args = data;
+
+ return panthor_group_get_state(pfile, args);
+}
+
+static int panthor_ioctl_tiler_heap_create(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_tiler_heap_create *args = data;
+ struct panthor_heap_pool *pool;
+ struct panthor_vm *vm;
+ int ret;
+
+ vm = panthor_vm_pool_get_vm(pfile->vms, args->vm_id);
+ if (!vm)
+ return -EINVAL;
+
+ pool = panthor_vm_get_heap_pool(vm, true);
+ if (IS_ERR(pool)) {
+ ret = PTR_ERR(pool);
+ goto out_put_vm;
+ }
+
+ ret = panthor_heap_create(pool,
+ args->initial_chunk_count,
+ args->chunk_size,
+ args->max_chunks,
+ args->target_in_flight,
+ &args->tiler_heap_ctx_gpu_va,
+ &args->first_heap_chunk_gpu_va);
+ if (ret < 0)
+ goto out_put_heap_pool;
+
+ /* Heap pools are per-VM. We combine the VM and HEAP id to make
+ * a unique heap handle.
+ */
+ args->handle = (args->vm_id << 16) | ret;
+ ret = 0;
+
+out_put_heap_pool:
+ panthor_heap_pool_put(pool);
+
+out_put_vm:
+ panthor_vm_put(vm);
+ return ret;
+}
+
+static int panthor_ioctl_tiler_heap_destroy(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_tiler_heap_destroy *args = data;
+ struct panthor_heap_pool *pool;
+ struct panthor_vm *vm;
+ int ret;
+
+ if (args->pad)
+ return -EINVAL;
+
+ vm = panthor_vm_pool_get_vm(pfile->vms, args->handle >> 16);
+ if (!vm)
+ return -EINVAL;
+
+ pool = panthor_vm_get_heap_pool(vm, false);
+ if (!pool) {
+ ret = -EINVAL;
+ goto out_put_vm;
+ }
+
+ ret = panthor_heap_destroy(pool, args->handle & GENMASK(15, 0));
+ panthor_heap_pool_put(pool);
+
+out_put_vm:
+ panthor_vm_put(vm);
+ return ret;
+}
+
+static int panthor_ioctl_vm_bind_async(struct drm_device *ddev,
+ struct drm_panthor_vm_bind *args,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_vm_bind_op *jobs_args;
+ struct panthor_submit_ctx ctx;
+ struct panthor_vm *vm;
+ int ret = 0;
+
+ vm = panthor_vm_pool_get_vm(pfile->vms, args->vm_id);
+ if (!vm)
+ return -EINVAL;
+
+ ret = PANTHOR_UOBJ_GET_ARRAY(jobs_args, &args->ops);
+ if (ret)
+ goto out_put_vm;
+
+ ret = panthor_submit_ctx_init(&ctx, file, args->ops.count);
+ if (ret)
+ goto out_free_jobs_args;
+
+ for (u32 i = 0; i < args->ops.count; i++) {
+ struct drm_panthor_vm_bind_op *op = &jobs_args[i];
+ struct drm_sched_job *job;
+
+ job = panthor_vm_bind_job_create(file, vm, op);
+ if (IS_ERR(job)) {
+ ret = PTR_ERR(job);
+ goto out_cleanup_submit_ctx;
+ }
+
+ ret = panthor_submit_ctx_add_job(&ctx, i, job, &op->syncs);
+ if (ret)
+ goto out_cleanup_submit_ctx;
+ }
+
+ ret = panthor_submit_ctx_collect_jobs_signal_ops(&ctx);
+ if (ret)
+ goto out_cleanup_submit_ctx;
+
+ /* Prepare reservation objects for each VM_BIND job. */
+ drm_exec_until_all_locked(&ctx.exec) {
+ for (u32 i = 0; i < ctx.job_count; i++) {
+ ret = panthor_vm_bind_job_prepare_resvs(&ctx.exec, ctx.jobs[i].job);
+ drm_exec_retry_on_contention(&ctx.exec);
+ if (ret)
+ goto out_cleanup_submit_ctx;
+ }
+ }
+
+ ret = panthor_submit_ctx_add_deps_and_arm_jobs(&ctx);
+ if (ret)
+ goto out_cleanup_submit_ctx;
+
+ /* Nothing can fail after that point. */
+ panthor_submit_ctx_push_jobs(&ctx, panthor_vm_bind_job_update_resvs);
+
+out_cleanup_submit_ctx:
+ panthor_submit_ctx_cleanup(&ctx, panthor_vm_bind_job_put);
+
+out_free_jobs_args:
+ kvfree(jobs_args);
+
+out_put_vm:
+ panthor_vm_put(vm);
+ return ret;
+}
+
+static int panthor_ioctl_vm_bind_sync(struct drm_device *ddev,
+ struct drm_panthor_vm_bind *args,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_vm_bind_op *jobs_args;
+ struct panthor_vm *vm;
+ int ret;
+
+ vm = panthor_vm_pool_get_vm(pfile->vms, args->vm_id);
+ if (!vm)
+ return -EINVAL;
+
+ ret = PANTHOR_UOBJ_GET_ARRAY(jobs_args, &args->ops);
+ if (ret)
+ goto out_put_vm;
+
+ for (u32 i = 0; i < args->ops.count; i++) {
+ ret = panthor_vm_bind_exec_sync_op(file, vm, &jobs_args[i]);
+ if (ret) {
+ /* Update ops.count so the user knows where things failed. */
+ args->ops.count = i;
+ break;
+ }
+ }
+
+ kvfree(jobs_args);
+
+out_put_vm:
+ panthor_vm_put(vm);
+ return ret;
+}
+
+#define PANTHOR_VM_BIND_FLAGS DRM_PANTHOR_VM_BIND_ASYNC
+
+static int panthor_ioctl_vm_bind(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct drm_panthor_vm_bind *args = data;
+ int cookie, ret;
+
+ if (!drm_dev_enter(ddev, &cookie))
+ return -ENODEV;
+
+ if (args->flags & DRM_PANTHOR_VM_BIND_ASYNC)
+ ret = panthor_ioctl_vm_bind_async(ddev, args, file);
+ else
+ ret = panthor_ioctl_vm_bind_sync(ddev, args, file);
+
+ drm_dev_exit(cookie);
+ return ret;
+}
+
+static int panthor_ioctl_vm_get_state(struct drm_device *ddev, void *data,
+ struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+ struct drm_panthor_vm_get_state *args = data;
+ struct panthor_vm *vm;
+
+ vm = panthor_vm_pool_get_vm(pfile->vms, args->vm_id);
+ if (!vm)
+ return -EINVAL;
+
+ if (panthor_vm_is_unusable(vm))
+ args->state = DRM_PANTHOR_VM_STATE_UNUSABLE;
+ else
+ args->state = DRM_PANTHOR_VM_STATE_USABLE;
+
+ panthor_vm_put(vm);
+ return 0;
+}
+
+static int
+panthor_open(struct drm_device *ddev, struct drm_file *file)
+{
+ struct panthor_device *ptdev = container_of(ddev, struct panthor_device, base);
+ struct panthor_file *pfile;
+ int ret;
+
+ if (!try_module_get(THIS_MODULE))
+ return -EINVAL;
+
+ pfile = kzalloc(sizeof(*pfile), GFP_KERNEL);
+ if (!pfile) {
+ ret = -ENOMEM;
+ goto err_put_mod;
+ }
+
+ pfile->ptdev = ptdev;
+
+ ret = panthor_vm_pool_create(pfile);
+ if (ret)
+ goto err_free_file;
+
+ ret = panthor_group_pool_create(pfile);
+ if (ret)
+ goto err_destroy_vm_pool;
+
+ file->driver_priv = pfile;
+ return 0;
+
+err_destroy_vm_pool:
+ panthor_vm_pool_destroy(pfile);
+
+err_free_file:
+ kfree(pfile);
+
+err_put_mod:
+ module_put(THIS_MODULE);
+ return ret;
+}
+
+static void
+panthor_postclose(struct drm_device *ddev, struct drm_file *file)
+{
+ struct panthor_file *pfile = file->driver_priv;
+
+ panthor_group_pool_destroy(pfile);
+ panthor_vm_pool_destroy(pfile);
+
+ kfree(pfile);
+ module_put(THIS_MODULE);
+}
+
+static const struct drm_ioctl_desc panthor_drm_driver_ioctls[] = {
+#define PANTHOR_IOCTL(n, func, flags) \
+ DRM_IOCTL_DEF_DRV(PANTHOR_##n, panthor_ioctl_##func, flags)
+
+ PANTHOR_IOCTL(DEV_QUERY, dev_query, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(VM_CREATE, vm_create, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(VM_DESTROY, vm_destroy, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(VM_BIND, vm_bind, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(VM_GET_STATE, vm_get_state, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(BO_CREATE, bo_create, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(BO_MMAP_OFFSET, bo_mmap_offset, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(GROUP_CREATE, group_create, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(GROUP_DESTROY, group_destroy, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(GROUP_GET_STATE, group_get_state, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(TILER_HEAP_CREATE, tiler_heap_create, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(TILER_HEAP_DESTROY, tiler_heap_destroy, DRM_RENDER_ALLOW),
+ PANTHOR_IOCTL(GROUP_SUBMIT, group_submit, DRM_RENDER_ALLOW),
+};
+
+static int panthor_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+ struct drm_file *file = filp->private_data;
+ struct panthor_file *pfile = file->driver_priv;
+ struct panthor_device *ptdev = pfile->ptdev;
+ u64 offset = (u64)vma->vm_pgoff << PAGE_SHIFT;
+ int ret, cookie;
+
+ if (!drm_dev_enter(file->minor->dev, &cookie))
+ return -ENODEV;
+
+ if (panthor_device_mmio_offset(offset) >= DRM_PANTHOR_USER_MMIO_OFFSET)
+ ret = panthor_device_mmap_io(ptdev, vma);
+ else
+ ret = drm_gem_mmap(filp, vma);
+
+ drm_dev_exit(cookie);
+ return ret;
+}
+
+static const struct file_operations panthor_drm_driver_fops = {
+ .open = drm_open,
+ .release = drm_release,
+ .unlocked_ioctl = drm_ioctl,
+ .compat_ioctl = drm_compat_ioctl,
+ .poll = drm_poll,
+ .read = drm_read,
+ .llseek = noop_llseek,
+ .mmap = panthor_mmap,
+};
+
+#ifdef CONFIG_DEBUG_FS
+static void panthor_debugfs_init(struct drm_minor *minor)
+{
+ panthor_mmu_debugfs_init(minor);
+}
+#endif
+
+/*
+ * PanCSF driver version:
+ * - 1.0 - initial interface
+ */
+static const struct drm_driver panthor_drm_driver = {
+ .driver_features = DRIVER_RENDER | DRIVER_GEM | DRIVER_SYNCOBJ |
+ DRIVER_SYNCOBJ_TIMELINE | DRIVER_GEM_GPUVA,
+ .open = panthor_open,
+ .postclose = panthor_postclose,
+ .ioctls = panthor_drm_driver_ioctls,
+ .num_ioctls = ARRAY_SIZE(panthor_drm_driver_ioctls),
+ .fops = &panthor_drm_driver_fops,
+ .name = "panthor",
+ .desc = "Panthor DRM driver",
+ .date = "20230801",
+ .major = 1,
+ .minor = 0,
+
+ .gem_create_object = panthor_gem_create_object,
+ .gem_prime_import_sg_table = drm_gem_shmem_prime_import_sg_table,
+#ifdef CONFIG_DEBUG_FS
+ .debugfs_init = panthor_debugfs_init,
+#endif
+};
+
+static int panthor_probe(struct platform_device *pdev)
+{
+ struct panthor_device *ptdev;
+
+ ptdev = devm_drm_dev_alloc(&pdev->dev, &panthor_drm_driver,
+ struct panthor_device, base);
+ if (!ptdev)
+ return -ENOMEM;
+
+ platform_set_drvdata(pdev, ptdev);
+
+ return panthor_device_init(ptdev);
+}
+
+static void panthor_remove(struct platform_device *pdev)
+{
+ struct panthor_device *ptdev = platform_get_drvdata(pdev);
+
+ panthor_device_unplug(ptdev);
+}
+
+static const struct of_device_id dt_match[] = {
+ { .compatible = "rockchip,rk3588-mali" },
+ { .compatible = "arm,mali-valhall-csf" },
+ {}
+};
+MODULE_DEVICE_TABLE(of, dt_match);
+
+static DEFINE_RUNTIME_DEV_PM_OPS(panthor_pm_ops,
+ panthor_device_suspend,
+ panthor_device_resume,
+ NULL);
+
+static struct platform_driver panthor_driver = {
+ .probe = panthor_probe,
+ .remove_new = panthor_remove,
+ .driver = {
+ .name = "panthor",
+ .pm = &panthor_pm_ops,
+ .of_match_table = dt_match,
+ },
+};
+
+/*
+ * Workqueue used to cleanup stuff.
+ *
+ * We create a dedicated workqueue so we can drain on unplug and
+ * make sure all resources are freed before the module is unloaded.
+ */
+struct workqueue_struct *panthor_cleanup_wq;
+
+static int __init panthor_init(void)
+{
+ int ret;
+
+ ret = panthor_mmu_pt_cache_init();
+ if (ret)
+ return ret;
+
+ panthor_cleanup_wq = alloc_workqueue("panthor-cleanup", WQ_UNBOUND, 0);
+ if (!panthor_cleanup_wq) {
+ pr_err("panthor: Failed to allocate the workqueues");
+ ret = -ENOMEM;
+ goto err_mmu_pt_cache_fini;
+ }
+
+ ret = platform_driver_register(&panthor_driver);
+ if (ret)
+ goto err_destroy_cleanup_wq;
+
+ return 0;
+
+err_destroy_cleanup_wq:
+ destroy_workqueue(panthor_cleanup_wq);
+
+err_mmu_pt_cache_fini:
+ panthor_mmu_pt_cache_fini();
+ return ret;
+}
+module_init(panthor_init);
+
+static void __exit panthor_exit(void)
+{
+ platform_driver_unregister(&panthor_driver);
+ destroy_workqueue(panthor_cleanup_wq);
+ panthor_mmu_pt_cache_fini();
+}
+module_exit(panthor_exit);
+
+MODULE_AUTHOR("Panthor Project Developers");
+MODULE_DESCRIPTION("Panthor DRM Driver");
+MODULE_LICENSE("Dual MIT/GPL");
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 11/14] drm/panthor: Add the driver frontend block
2024-01-22 16:30 ` [PATCH v4 11/14] drm/panthor: Add the driver frontend block Boris Brezillon
@ 2024-01-23 16:29 ` Heiko Stübner
2024-01-23 17:06 ` Boris Brezillon
0 siblings, 1 reply; 51+ messages in thread
From: Heiko Stübner @ 2024-01-23 16:29 UTC (permalink / raw)
To: dri-devel, Boris Brezillon
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
Am Montag, 22. Januar 2024, 17:30:42 CET schrieb Boris Brezillon:
> This is the last piece missing to expose the driver to the outside
> world.
>
> This is basically a wrapper between the ioctls and the other logical
> blocks.
>
> v4:
> - Add an ioctl to let the UMD query the VM state
> - Fix kernel doc
> - Let panthor_device_init() call panthor_device_init()
> - Fix cleanup ordering in the panthor_init() error path
> - Add Steve's and Liviu's R-b
>
> v3:
> - Add acks for the MIT/GPL2 relicensing
> - Fix 32-bit support
> - Account for panthor_vm and panthor_sched changes
> - Simplify the resv preparation/update logic
> - Use a linked list rather than xarray for list of signals.
> - Simplify panthor_get_uobj_array by returning the newly allocated
> array.
> - Drop the "DOC" for job submission helpers and move the relevant
> comments to panthor_ioctl_group_submit().
> - Add helpers sync_op_is_signal()/sync_op_is_wait().
> - Simplify return type of panthor_submit_ctx_add_sync_signal() and
> panthor_submit_ctx_get_sync_signal().
> - Drop WARN_ON from panthor_submit_ctx_add_job().
> - Fix typos in comments.
>
> Co-developed-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
> Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
> Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
> Reviewed-by: Steven Price <steven.price@arm.com>
> Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
> ---
> drivers/gpu/drm/panthor/panthor_drv.c | 1470 +++++++++++++++++++++++++
> 1 file changed, 1470 insertions(+)
> create mode 100644 drivers/gpu/drm/panthor/panthor_drv.c
>
> diff --git a/drivers/gpu/drm/panthor/panthor_drv.c b/drivers/gpu/drm/panthor/panthor_drv.c
> new file mode 100644
> index 000000000000..207aacaccd39
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_drv.c
> @@ -0,0 +1,1470 @@
> +// SPDX-License-Identifier: GPL-2.0 or MIT
> +/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
> +/* Copyright 2019 Linaro, Ltd., Rob Herring <robh@kernel.org> */
> +/* Copyright 2019 Collabora ltd. */
> +
> +#include <linux/list.h>
> +#include <linux/module.h>
> +#include <linux/of_platform.h>
> +#include <linux/pagemap.h>
> +#include <linux/pm_runtime.h>
@@ -7,6 +7,7 @@
#include <linux/module.h>
#include <linux/of_platform.h>
#include <linux/pagemap.h>
+#include <linux/platform_device.h>
#include <linux/pm_runtime.h>
#include <drm/drm_drv.h>
with v6.8-rc1 this needs a linux/platform_device.h include to keep
finding struct platform_device and friends
[...]
> +static int panthor_submit_ctx_init(struct panthor_submit_ctx *ctx,
> + struct drm_file *file, u32 job_count)
> +{
> + ctx->jobs = kvmalloc_array(job_count, sizeof(*ctx->jobs),
> + GFP_KERNEL | __GFP_ZERO);
> + if (!ctx->jobs)
> + return -ENOMEM;
> +
> + ctx->file = file;
> + ctx->job_count = job_count;
> + INIT_LIST_HEAD(&ctx->signals);
> + drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT | DRM_EXEC_IGNORE_DUPLICATES);
../drivers/gpu/drm/panthor/panthor_drv.c: In function ‘panthor_submit_ctx_init’:
../drivers/gpu/drm/panthor/panthor_drv.c:722:9: error: too few arguments to function ‘drm_exec_init’
722 | drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT | DRM_EXEC_IGNORE_DUPLICATES);
| ^~~~~~~~~~~~~
In v6.8-rc1 (or drm-misc-next I guess) the calling convention of
drm_exec_init changed to include a number of initial objects, see
commit 05d249352f1a ("drm/exec: Pass in initial # of objects")
Heiko
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 11/14] drm/panthor: Add the driver frontend block
2024-01-23 16:29 ` Heiko Stübner
@ 2024-01-23 17:06 ` Boris Brezillon
0 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-23 17:06 UTC (permalink / raw)
To: Heiko Stübner
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, dri-devel,
Steven Price, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
On Tue, 23 Jan 2024 17:29:12 +0100
Heiko Stübner <heiko@sntech.de> wrote:
> Am Montag, 22. Januar 2024, 17:30:42 CET schrieb Boris Brezillon:
> > This is the last piece missing to expose the driver to the outside
> > world.
> >
> > This is basically a wrapper between the ioctls and the other logical
> > blocks.
> >
> > v4:
> > - Add an ioctl to let the UMD query the VM state
> > - Fix kernel doc
> > - Let panthor_device_init() call panthor_device_init()
> > - Fix cleanup ordering in the panthor_init() error path
> > - Add Steve's and Liviu's R-b
> >
> > v3:
> > - Add acks for the MIT/GPL2 relicensing
> > - Fix 32-bit support
> > - Account for panthor_vm and panthor_sched changes
> > - Simplify the resv preparation/update logic
> > - Use a linked list rather than xarray for list of signals.
> > - Simplify panthor_get_uobj_array by returning the newly allocated
> > array.
> > - Drop the "DOC" for job submission helpers and move the relevant
> > comments to panthor_ioctl_group_submit().
> > - Add helpers sync_op_is_signal()/sync_op_is_wait().
> > - Simplify return type of panthor_submit_ctx_add_sync_signal() and
> > panthor_submit_ctx_get_sync_signal().
> > - Drop WARN_ON from panthor_submit_ctx_add_job().
> > - Fix typos in comments.
> >
> > Co-developed-by: Steven Price <steven.price@arm.com>
> > Signed-off-by: Steven Price <steven.price@arm.com>
> > Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> > Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
> > Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
> > Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
> > Reviewed-by: Steven Price <steven.price@arm.com>
> > Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
> > ---
> > drivers/gpu/drm/panthor/panthor_drv.c | 1470 +++++++++++++++++++++++++
> > 1 file changed, 1470 insertions(+)
> > create mode 100644 drivers/gpu/drm/panthor/panthor_drv.c
> >
> > diff --git a/drivers/gpu/drm/panthor/panthor_drv.c b/drivers/gpu/drm/panthor/panthor_drv.c
> > new file mode 100644
> > index 000000000000..207aacaccd39
> > --- /dev/null
> > +++ b/drivers/gpu/drm/panthor/panthor_drv.c
> > @@ -0,0 +1,1470 @@
> > +// SPDX-License-Identifier: GPL-2.0 or MIT
> > +/* Copyright 2018 Marty E. Plummer <hanetzer@startmail.com> */
> > +/* Copyright 2019 Linaro, Ltd., Rob Herring <robh@kernel.org> */
> > +/* Copyright 2019 Collabora ltd. */
> > +
> > +#include <linux/list.h>
> > +#include <linux/module.h>
> > +#include <linux/of_platform.h>
> > +#include <linux/pagemap.h>
> > +#include <linux/pm_runtime.h>
>
> @@ -7,6 +7,7 @@
> #include <linux/module.h>
> #include <linux/of_platform.h>
> #include <linux/pagemap.h>
> +#include <linux/platform_device.h>
> #include <linux/pm_runtime.h>
>
> #include <drm/drm_drv.h>
>
> with v6.8-rc1 this needs a linux/platform_device.h include to keep
> finding struct platform_device and friends
Will add this include in v5.
>
> [...]
>
>
> > +static int panthor_submit_ctx_init(struct panthor_submit_ctx *ctx,
> > + struct drm_file *file, u32 job_count)
> > +{
> > + ctx->jobs = kvmalloc_array(job_count, sizeof(*ctx->jobs),
> > + GFP_KERNEL | __GFP_ZERO);
> > + if (!ctx->jobs)
> > + return -ENOMEM;
> > +
> > + ctx->file = file;
> > + ctx->job_count = job_count;
> > + INIT_LIST_HEAD(&ctx->signals);
> > + drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT | DRM_EXEC_IGNORE_DUPLICATES);
>
> ../drivers/gpu/drm/panthor/panthor_drv.c: In function ‘panthor_submit_ctx_init’:
> ../drivers/gpu/drm/panthor/panthor_drv.c:722:9: error: too few arguments to function ‘drm_exec_init’
> 722 | drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT | DRM_EXEC_IGNORE_DUPLICATES);
> | ^~~~~~~~~~~~~
>
> In v6.8-rc1 (or drm-misc-next I guess) the calling convention of
> drm_exec_init changed to include a number of initial objects, see
> commit 05d249352f1a ("drm/exec: Pass in initial # of objects")
>
Looks like this patch went through drm-misc-fixes, and v4.8-rc1 has not
been back merged to drm-misc-next yet.
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 12/14] drm/panthor: Allow driver compilation
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (10 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 11/14] drm/panthor: Add the driver frontend block Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-01-22 16:30 ` Boris Brezillon
` (5 subsequent siblings)
17 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Grant Likely, Marty E . Plummer, Robin Murphy,
Faith Ekstrand
Now that all blocks are available, we can add/update Kconfig/Makefile
files to allow compilation.
v4:
- Add Steve's R-b
v3:
- Add a dep on DRM_GPUVM
- Fix dependencies in Kconfig
- Expand help text to (hopefully) describe which GPUs are to be
supported by this driver and which are for panfrost.
Co-developed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Steven Price <steven.price@arm.com> # MIT+GPL2 relicensing,Arm
Acked-by: Grant Likely <grant.likely@linaro.org> # MIT+GPL2 relicensing,Linaro
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> # MIT+GPL2 relicensing,Collabora
Reviewed-by: Steven Price <steven.price@arm.com>
---
drivers/gpu/drm/Kconfig | 2 ++
drivers/gpu/drm/Makefile | 1 +
drivers/gpu/drm/panthor/Kconfig | 23 +++++++++++++++++++++++
drivers/gpu/drm/panthor/Makefile | 15 +++++++++++++++
4 files changed, 41 insertions(+)
create mode 100644 drivers/gpu/drm/panthor/Kconfig
create mode 100644 drivers/gpu/drm/panthor/Makefile
diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
index 628f90ed8a9b..88c1ca148507 100644
--- a/drivers/gpu/drm/Kconfig
+++ b/drivers/gpu/drm/Kconfig
@@ -369,6 +369,8 @@ source "drivers/gpu/drm/lima/Kconfig"
source "drivers/gpu/drm/panfrost/Kconfig"
+source "drivers/gpu/drm/panthor/Kconfig"
+
source "drivers/gpu/drm/aspeed/Kconfig"
source "drivers/gpu/drm/mcde/Kconfig"
diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile
index 8ac6f4b9546e..9023d5440cd7 100644
--- a/drivers/gpu/drm/Makefile
+++ b/drivers/gpu/drm/Makefile
@@ -178,6 +178,7 @@ obj-$(CONFIG_DRM_XEN) += xen/
obj-$(CONFIG_DRM_VBOXVIDEO) += vboxvideo/
obj-$(CONFIG_DRM_LIMA) += lima/
obj-$(CONFIG_DRM_PANFROST) += panfrost/
+obj-$(CONFIG_DRM_PANTHOR) += panthor/
obj-$(CONFIG_DRM_ASPEED_GFX) += aspeed/
obj-$(CONFIG_DRM_MCDE) += mcde/
obj-$(CONFIG_DRM_TIDSS) += tidss/
diff --git a/drivers/gpu/drm/panthor/Kconfig b/drivers/gpu/drm/panthor/Kconfig
new file mode 100644
index 000000000000..159845e51116
--- /dev/null
+++ b/drivers/gpu/drm/panthor/Kconfig
@@ -0,0 +1,23 @@
+# SPDX-License-Identifier: GPL-2.0 or MIT
+
+config DRM_PANTHOR
+ tristate "Panthor (DRM support for ARM Mali CSF-based GPUs)"
+ depends on DRM
+ depends on ARM || ARM64 || COMPILE_TEST
+ depends on !GENERIC_ATOMIC64 # for IOMMU_IO_PGTABLE_LPAE
+ depends on MMU
+ select DRM_GPUVM
+ select DRM_EXEC
+ select DRM_SCHED
+ select IOMMU_SUPPORT
+ select IOMMU_IO_PGTABLE_LPAE
+ select DRM_GEM_SHMEM_HELPER
+ select PM_DEVFREQ
+ select DEVFREQ_GOV_SIMPLE_ONDEMAND
+ help
+ DRM driver for ARM Mali CSF-based GPUs.
+
+ This driver is for Mali (or Immortalis) Valhall Gxxx GPUs.
+
+ Note that the Mali-G68 and Mali-G78, while Valhall architecture, will
+ be supported with the panfrost driver as they are not CSF GPUs.
diff --git a/drivers/gpu/drm/panthor/Makefile b/drivers/gpu/drm/panthor/Makefile
new file mode 100644
index 000000000000..64193a484879
--- /dev/null
+++ b/drivers/gpu/drm/panthor/Makefile
@@ -0,0 +1,15 @@
+# SPDX-License-Identifier: GPL-2.0 or MIT
+
+panthor-y := \
+ panthor_devfreq.o \
+ panthor_device.o \
+ panthor_drv.o \
+ panthor_gem.o \
+ panthor_gpu.o \
+ panthor_heap.o \
+ panthor_heap.o \
+ panthor_fw.o \
+ panthor_mmu.o \
+ panthor_sched.o
+
+obj-$(CONFIG_DRM_PANTHOR) += panthor.o
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* [PATCH v4 13/14] dt-bindings: gpu: mali-valhall-csf: Add support for Arm Mali CSF GPUs
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-01-22 16:30 ` [PATCH v4 02/14] drm/panthor: Add GPU register definitions Boris Brezillon
` (16 subsequent siblings)
17 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Steven Price,
Robin Murphy, kernel, Heiko Stuebner, Tatsuyuki Ishi,
Chris Diamand, Ketil Johnsen, Liviu Dudau, Krzysztof Kozlowski,
Rob Herring, Conor Dooley, devicetree, Boris Brezillon
From: Liviu Dudau <liviu.dudau@arm.com>
Arm has introduced a new v10 GPU architecture that replaces the Job Manager
interface with a new Command Stream Frontend. It adds firmware driven
command stream queues that can be used by kernel and user space to submit
jobs to the GPU.
Add the initial schema for the device tree that is based on support for
RK3588 SoC. The minimum number of clocks is one for the IP, but on Rockchip
platforms they will tend to expose the semi-independent clocks for better
power management.
v4:
- Fix formatting issue
v3:
- Cleanup commit message to remove redundant text
- Added opp-table property and re-ordered entries
- Clarified power-domains and power-domain-names requirements for RK3588.
- Cleaned up example
Note: power-domains and power-domain-names requirements for other platforms
are still work in progress, hence the bindings are left incomplete here.
v2:
- New commit
Signed-off-by: Liviu Dudau <liviu.dudau@arm.com>
Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Conor Dooley <conor+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
---
.../bindings/gpu/arm,mali-valhall-csf.yaml | 147 ++++++++++++++++++
1 file changed, 147 insertions(+)
create mode 100644 Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
new file mode 100644
index 000000000000..be1f6bacc3f3
--- /dev/null
+++ b/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
@@ -0,0 +1,147 @@
+# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/gpu/arm,mali-valhall-csf.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: ARM Mali Valhall GPU
+
+maintainers:
+ - Liviu Dudau <liviu.dudau@arm.com>
+ - Boris Brezillon <boris.brezillon@collabora.com>
+
+properties:
+ $nodename:
+ pattern: '^gpu@[a-f0-9]+$'
+
+ compatible:
+ oneOf:
+ - items:
+ - enum:
+ - rockchip,rk3588-mali
+ - const: arm,mali-valhall-csf # Mali Valhall GPU model/revision is fully discoverable
+
+ reg:
+ maxItems: 1
+
+ interrupts:
+ items:
+ - description: Job interrupt
+ - description: MMU interrupt
+ - description: GPU interrupt
+
+ interrupt-names:
+ items:
+ - const: job
+ - const: mmu
+ - const: gpu
+
+ clocks:
+ minItems: 1
+ maxItems: 3
+
+ clock-names:
+ minItems: 1
+ items:
+ - const: core
+ - const: coregroup
+ - const: stacks
+
+ mali-supply: true
+
+ operating-points-v2: true
+ opp-table:
+ type: object
+
+ power-domains:
+ minItems: 1
+ maxItems: 5
+
+ power-domain-names:
+ minItems: 1
+ maxItems: 5
+
+ sram-supply: true
+
+ "#cooling-cells":
+ const: 2
+
+ dynamic-power-coefficient:
+ $ref: /schemas/types.yaml#/definitions/uint32
+ description:
+ A u32 value that represents the running time dynamic
+ power coefficient in units of uW/MHz/V^2. The
+ coefficient can either be calculated from power
+ measurements or derived by analysis.
+
+ The dynamic power consumption of the GPU is
+ proportional to the square of the Voltage (V) and
+ the clock frequency (f). The coefficient is used to
+ calculate the dynamic power as below -
+
+ Pdyn = dynamic-power-coefficient * V^2 * f
+
+ where voltage is in V, frequency is in MHz.
+
+ dma-coherent: true
+
+required:
+ - compatible
+ - reg
+ - interrupts
+ - interrupt-names
+ - clocks
+ - mali-supply
+
+additionalProperties: false
+
+allOf:
+ - if:
+ properties:
+ compatible:
+ contains:
+ const: rockchip,rk3588-mali
+ then:
+ properties:
+ clocks:
+ minItems: 3
+ power-domains:
+ maxItems: 1
+ power-domain-names: false
+
+examples:
+ - |
+ #include <dt-bindings/clock/rockchip,rk3588-cru.h>
+ #include <dt-bindings/interrupt-controller/irq.h>
+ #include <dt-bindings/interrupt-controller/arm-gic.h>
+ #include <dt-bindings/power/rk3588-power.h>
+
+ gpu: gpu@fb000000 {
+ compatible = "rockchip,rk3588-mali", "arm,mali-valhall-csf";
+ reg = <0xfb000000 0x200000>;
+ interrupts = <GIC_SPI 92 IRQ_TYPE_LEVEL_HIGH 0>,
+ <GIC_SPI 93 IRQ_TYPE_LEVEL_HIGH 0>,
+ <GIC_SPI 94 IRQ_TYPE_LEVEL_HIGH 0>;
+ interrupt-names = "job", "mmu", "gpu";
+ clock-names = "core", "coregroup", "stacks";
+ clocks = <&cru CLK_GPU>, <&cru CLK_GPU_COREGROUP>,
+ <&cru CLK_GPU_STACKS>;
+ power-domains = <&power RK3588_PD_GPU>;
+ operating-points-v2 = <&gpu_opp_table>;
+ mali-supply = <&vdd_gpu_s0>;
+ sram-supply = <&vdd_gpu_mem_s0>;
+ };
+
+ gpu_opp_table: opp-table {
+ compatible = "operating-points-v2";
+ opp-300000000 {
+ opp-hz = /bits/ 64 <300000000>;
+ opp-microvolt = <675000 675000 850000>;
+ };
+ opp-400000000 {
+ opp-hz = /bits/ 64 <400000000>;
+ opp-microvolt = <675000 675000 850000>;
+ };
+ };
+
+...
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* [PATCH v4 13/14] dt-bindings: gpu: mali-valhall-csf: Add support for Arm Mali CSF GPUs
@ 2024-01-22 16:30 ` Boris Brezillon
0 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Krzysztof Kozlowski, Neil Armstrong, Ketil Johnsen, Liviu Dudau,
Conor Dooley, Boris Brezillon, Steven Price, devicetree,
Rob Herring, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
From: Liviu Dudau <liviu.dudau@arm.com>
Arm has introduced a new v10 GPU architecture that replaces the Job Manager
interface with a new Command Stream Frontend. It adds firmware driven
command stream queues that can be used by kernel and user space to submit
jobs to the GPU.
Add the initial schema for the device tree that is based on support for
RK3588 SoC. The minimum number of clocks is one for the IP, but on Rockchip
platforms they will tend to expose the semi-independent clocks for better
power management.
v4:
- Fix formatting issue
v3:
- Cleanup commit message to remove redundant text
- Added opp-table property and re-ordered entries
- Clarified power-domains and power-domain-names requirements for RK3588.
- Cleaned up example
Note: power-domains and power-domain-names requirements for other platforms
are still work in progress, hence the bindings are left incomplete here.
v2:
- New commit
Signed-off-by: Liviu Dudau <liviu.dudau@arm.com>
Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Conor Dooley <conor+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
---
.../bindings/gpu/arm,mali-valhall-csf.yaml | 147 ++++++++++++++++++
1 file changed, 147 insertions(+)
create mode 100644 Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
new file mode 100644
index 000000000000..be1f6bacc3f3
--- /dev/null
+++ b/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
@@ -0,0 +1,147 @@
+# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/gpu/arm,mali-valhall-csf.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: ARM Mali Valhall GPU
+
+maintainers:
+ - Liviu Dudau <liviu.dudau@arm.com>
+ - Boris Brezillon <boris.brezillon@collabora.com>
+
+properties:
+ $nodename:
+ pattern: '^gpu@[a-f0-9]+$'
+
+ compatible:
+ oneOf:
+ - items:
+ - enum:
+ - rockchip,rk3588-mali
+ - const: arm,mali-valhall-csf # Mali Valhall GPU model/revision is fully discoverable
+
+ reg:
+ maxItems: 1
+
+ interrupts:
+ items:
+ - description: Job interrupt
+ - description: MMU interrupt
+ - description: GPU interrupt
+
+ interrupt-names:
+ items:
+ - const: job
+ - const: mmu
+ - const: gpu
+
+ clocks:
+ minItems: 1
+ maxItems: 3
+
+ clock-names:
+ minItems: 1
+ items:
+ - const: core
+ - const: coregroup
+ - const: stacks
+
+ mali-supply: true
+
+ operating-points-v2: true
+ opp-table:
+ type: object
+
+ power-domains:
+ minItems: 1
+ maxItems: 5
+
+ power-domain-names:
+ minItems: 1
+ maxItems: 5
+
+ sram-supply: true
+
+ "#cooling-cells":
+ const: 2
+
+ dynamic-power-coefficient:
+ $ref: /schemas/types.yaml#/definitions/uint32
+ description:
+ A u32 value that represents the running time dynamic
+ power coefficient in units of uW/MHz/V^2. The
+ coefficient can either be calculated from power
+ measurements or derived by analysis.
+
+ The dynamic power consumption of the GPU is
+ proportional to the square of the Voltage (V) and
+ the clock frequency (f). The coefficient is used to
+ calculate the dynamic power as below -
+
+ Pdyn = dynamic-power-coefficient * V^2 * f
+
+ where voltage is in V, frequency is in MHz.
+
+ dma-coherent: true
+
+required:
+ - compatible
+ - reg
+ - interrupts
+ - interrupt-names
+ - clocks
+ - mali-supply
+
+additionalProperties: false
+
+allOf:
+ - if:
+ properties:
+ compatible:
+ contains:
+ const: rockchip,rk3588-mali
+ then:
+ properties:
+ clocks:
+ minItems: 3
+ power-domains:
+ maxItems: 1
+ power-domain-names: false
+
+examples:
+ - |
+ #include <dt-bindings/clock/rockchip,rk3588-cru.h>
+ #include <dt-bindings/interrupt-controller/irq.h>
+ #include <dt-bindings/interrupt-controller/arm-gic.h>
+ #include <dt-bindings/power/rk3588-power.h>
+
+ gpu: gpu@fb000000 {
+ compatible = "rockchip,rk3588-mali", "arm,mali-valhall-csf";
+ reg = <0xfb000000 0x200000>;
+ interrupts = <GIC_SPI 92 IRQ_TYPE_LEVEL_HIGH 0>,
+ <GIC_SPI 93 IRQ_TYPE_LEVEL_HIGH 0>,
+ <GIC_SPI 94 IRQ_TYPE_LEVEL_HIGH 0>;
+ interrupt-names = "job", "mmu", "gpu";
+ clock-names = "core", "coregroup", "stacks";
+ clocks = <&cru CLK_GPU>, <&cru CLK_GPU_COREGROUP>,
+ <&cru CLK_GPU_STACKS>;
+ power-domains = <&power RK3588_PD_GPU>;
+ operating-points-v2 = <&gpu_opp_table>;
+ mali-supply = <&vdd_gpu_s0>;
+ sram-supply = <&vdd_gpu_mem_s0>;
+ };
+
+ gpu_opp_table: opp-table {
+ compatible = "operating-points-v2";
+ opp-300000000 {
+ opp-hz = /bits/ 64 <300000000>;
+ opp-microvolt = <675000 675000 850000>;
+ };
+ opp-400000000 {
+ opp-hz = /bits/ 64 <400000000>;
+ opp-microvolt = <675000 675000 850000>;
+ };
+ };
+
+...
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 13/14] dt-bindings: gpu: mali-valhall-csf: Add support for Arm Mali CSF GPUs
2024-01-22 16:30 ` Boris Brezillon
@ 2024-01-30 19:54 ` Rob Herring
-1 siblings, 0 replies; 51+ messages in thread
From: Rob Herring @ 2024-01-30 19:54 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Daniel Vetter, Marty E . Plummer,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Liviu Dudau, Steven Price,
Robin Murphy, kernel, Heiko Stuebner, Tatsuyuki Ishi,
Chris Diamand, Ketil Johnsen, Krzysztof Kozlowski, Conor Dooley,
devicetree
On Mon, Jan 22, 2024 at 05:30:44PM +0100, Boris Brezillon wrote:
> From: Liviu Dudau <liviu.dudau@arm.com>
>
> Arm has introduced a new v10 GPU architecture that replaces the Job Manager
> interface with a new Command Stream Frontend. It adds firmware driven
> command stream queues that can be used by kernel and user space to submit
> jobs to the GPU.
>
> Add the initial schema for the device tree that is based on support for
> RK3588 SoC. The minimum number of clocks is one for the IP, but on Rockchip
> platforms they will tend to expose the semi-independent clocks for better
> power management.
>
> v4:
> - Fix formatting issue
>
> v3:
> - Cleanup commit message to remove redundant text
> - Added opp-table property and re-ordered entries
> - Clarified power-domains and power-domain-names requirements for RK3588.
> - Cleaned up example
>
> Note: power-domains and power-domain-names requirements for other platforms
> are still work in progress, hence the bindings are left incomplete here.
>
> v2:
> - New commit
>
> Signed-off-by: Liviu Dudau <liviu.dudau@arm.com>
> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
> Cc: Rob Herring <robh+dt@kernel.org>
> Cc: Conor Dooley <conor+dt@kernel.org>
> Cc: devicetree@vger.kernel.org
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> ---
> .../bindings/gpu/arm,mali-valhall-csf.yaml | 147 ++++++++++++++++++
> 1 file changed, 147 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
>
> diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
> new file mode 100644
> index 000000000000..be1f6bacc3f3
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
> @@ -0,0 +1,147 @@
> +# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/gpu/arm,mali-valhall-csf.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: ARM Mali Valhall GPU
> +
> +maintainers:
> + - Liviu Dudau <liviu.dudau@arm.com>
> + - Boris Brezillon <boris.brezillon@collabora.com>
> +
> +properties:
> + $nodename:
> + pattern: '^gpu@[a-f0-9]+$'
> +
> + compatible:
> + oneOf:
> + - items:
> + - enum:
> + - rockchip,rk3588-mali
> + - const: arm,mali-valhall-csf # Mali Valhall GPU model/revision is fully discoverable
> +
> + reg:
> + maxItems: 1
> +
> + interrupts:
> + items:
> + - description: Job interrupt
> + - description: MMU interrupt
> + - description: GPU interrupt
> +
> + interrupt-names:
> + items:
> + - const: job
> + - const: mmu
> + - const: gpu
> +
> + clocks:
> + minItems: 1
> + maxItems: 3
> +
> + clock-names:
> + minItems: 1
> + items:
> + - const: core
> + - const: coregroup
> + - const: stacks
> +
> + mali-supply: true
> +
> + operating-points-v2: true
> + opp-table:
> + type: object
> +
> + power-domains:
> + minItems: 1
> + maxItems: 5
> +
> + power-domain-names:
> + minItems: 1
> + maxItems: 5
> +
> + sram-supply: true
> +
> + "#cooling-cells":
> + const: 2
> +
> + dynamic-power-coefficient:
> + $ref: /schemas/types.yaml#/definitions/uint32
> + description:
> + A u32 value that represents the running time dynamic
> + power coefficient in units of uW/MHz/V^2. The
> + coefficient can either be calculated from power
> + measurements or derived by analysis.
> +
> + The dynamic power consumption of the GPU is
> + proportional to the square of the Voltage (V) and
> + the clock frequency (f). The coefficient is used to
> + calculate the dynamic power as below -
> +
> + Pdyn = dynamic-power-coefficient * V^2 * f
> +
> + where voltage is in V, frequency is in MHz.
> +
> + dma-coherent: true
> +
> +required:
> + - compatible
> + - reg
> + - interrupts
> + - interrupt-names
> + - clocks
> + - mali-supply
> +
> +additionalProperties: false
> +
> +allOf:
> + - if:
> + properties:
> + compatible:
> + contains:
> + const: rockchip,rk3588-mali
> + then:
> + properties:
> + clocks:
> + minItems: 3
> + power-domains:
> + maxItems: 1
> + power-domain-names: false
> +
> +examples:
> + - |
> + #include <dt-bindings/clock/rockchip,rk3588-cru.h>
> + #include <dt-bindings/interrupt-controller/irq.h>
> + #include <dt-bindings/interrupt-controller/arm-gic.h>
> + #include <dt-bindings/power/rk3588-power.h>
> +
> + gpu: gpu@fb000000 {
> + compatible = "rockchip,rk3588-mali", "arm,mali-valhall-csf";
> + reg = <0xfb000000 0x200000>;
> + interrupts = <GIC_SPI 92 IRQ_TYPE_LEVEL_HIGH 0>,
> + <GIC_SPI 93 IRQ_TYPE_LEVEL_HIGH 0>,
> + <GIC_SPI 94 IRQ_TYPE_LEVEL_HIGH 0>;
> + interrupt-names = "job", "mmu", "gpu";
> + clock-names = "core", "coregroup", "stacks";
> + clocks = <&cru CLK_GPU>, <&cru CLK_GPU_COREGROUP>,
> + <&cru CLK_GPU_STACKS>;
> + power-domains = <&power RK3588_PD_GPU>;
> + operating-points-v2 = <&gpu_opp_table>;
> + mali-supply = <&vdd_gpu_s0>;
> + sram-supply = <&vdd_gpu_mem_s0>;
> + };
> +
> + gpu_opp_table: opp-table {
Move this into the gpu node. With that,
Reviewed-by: Rob Herring <robh@kernel.org>
> + compatible = "operating-points-v2";
> + opp-300000000 {
> + opp-hz = /bits/ 64 <300000000>;
> + opp-microvolt = <675000 675000 850000>;
> + };
> + opp-400000000 {
> + opp-hz = /bits/ 64 <400000000>;
> + opp-microvolt = <675000 675000 850000>;
> + };
> + };
> +
> +...
> --
> 2.43.0
>
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 13/14] dt-bindings: gpu: mali-valhall-csf: Add support for Arm Mali CSF GPUs
@ 2024-01-30 19:54 ` Rob Herring
0 siblings, 0 replies; 51+ messages in thread
From: Rob Herring @ 2024-01-30 19:54 UTC (permalink / raw)
To: Boris Brezillon
Cc: Neil Armstrong, Tatsuyuki Ishi, Nicolas Boichat, kernel,
Daniel Stone, Krzysztof Kozlowski, devicetree, Ketil Johnsen,
Liviu Dudau, Conor Dooley, dri-devel, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand,
Marty E . Plummer, Robin Murphy, Faith Ekstrand
On Mon, Jan 22, 2024 at 05:30:44PM +0100, Boris Brezillon wrote:
> From: Liviu Dudau <liviu.dudau@arm.com>
>
> Arm has introduced a new v10 GPU architecture that replaces the Job Manager
> interface with a new Command Stream Frontend. It adds firmware driven
> command stream queues that can be used by kernel and user space to submit
> jobs to the GPU.
>
> Add the initial schema for the device tree that is based on support for
> RK3588 SoC. The minimum number of clocks is one for the IP, but on Rockchip
> platforms they will tend to expose the semi-independent clocks for better
> power management.
>
> v4:
> - Fix formatting issue
>
> v3:
> - Cleanup commit message to remove redundant text
> - Added opp-table property and re-ordered entries
> - Clarified power-domains and power-domain-names requirements for RK3588.
> - Cleaned up example
>
> Note: power-domains and power-domain-names requirements for other platforms
> are still work in progress, hence the bindings are left incomplete here.
>
> v2:
> - New commit
>
> Signed-off-by: Liviu Dudau <liviu.dudau@arm.com>
> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
> Cc: Rob Herring <robh+dt@kernel.org>
> Cc: Conor Dooley <conor+dt@kernel.org>
> Cc: devicetree@vger.kernel.org
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> ---
> .../bindings/gpu/arm,mali-valhall-csf.yaml | 147 ++++++++++++++++++
> 1 file changed, 147 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
>
> diff --git a/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml b/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
> new file mode 100644
> index 000000000000..be1f6bacc3f3
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
> @@ -0,0 +1,147 @@
> +# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/gpu/arm,mali-valhall-csf.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: ARM Mali Valhall GPU
> +
> +maintainers:
> + - Liviu Dudau <liviu.dudau@arm.com>
> + - Boris Brezillon <boris.brezillon@collabora.com>
> +
> +properties:
> + $nodename:
> + pattern: '^gpu@[a-f0-9]+$'
> +
> + compatible:
> + oneOf:
> + - items:
> + - enum:
> + - rockchip,rk3588-mali
> + - const: arm,mali-valhall-csf # Mali Valhall GPU model/revision is fully discoverable
> +
> + reg:
> + maxItems: 1
> +
> + interrupts:
> + items:
> + - description: Job interrupt
> + - description: MMU interrupt
> + - description: GPU interrupt
> +
> + interrupt-names:
> + items:
> + - const: job
> + - const: mmu
> + - const: gpu
> +
> + clocks:
> + minItems: 1
> + maxItems: 3
> +
> + clock-names:
> + minItems: 1
> + items:
> + - const: core
> + - const: coregroup
> + - const: stacks
> +
> + mali-supply: true
> +
> + operating-points-v2: true
> + opp-table:
> + type: object
> +
> + power-domains:
> + minItems: 1
> + maxItems: 5
> +
> + power-domain-names:
> + minItems: 1
> + maxItems: 5
> +
> + sram-supply: true
> +
> + "#cooling-cells":
> + const: 2
> +
> + dynamic-power-coefficient:
> + $ref: /schemas/types.yaml#/definitions/uint32
> + description:
> + A u32 value that represents the running time dynamic
> + power coefficient in units of uW/MHz/V^2. The
> + coefficient can either be calculated from power
> + measurements or derived by analysis.
> +
> + The dynamic power consumption of the GPU is
> + proportional to the square of the Voltage (V) and
> + the clock frequency (f). The coefficient is used to
> + calculate the dynamic power as below -
> +
> + Pdyn = dynamic-power-coefficient * V^2 * f
> +
> + where voltage is in V, frequency is in MHz.
> +
> + dma-coherent: true
> +
> +required:
> + - compatible
> + - reg
> + - interrupts
> + - interrupt-names
> + - clocks
> + - mali-supply
> +
> +additionalProperties: false
> +
> +allOf:
> + - if:
> + properties:
> + compatible:
> + contains:
> + const: rockchip,rk3588-mali
> + then:
> + properties:
> + clocks:
> + minItems: 3
> + power-domains:
> + maxItems: 1
> + power-domain-names: false
> +
> +examples:
> + - |
> + #include <dt-bindings/clock/rockchip,rk3588-cru.h>
> + #include <dt-bindings/interrupt-controller/irq.h>
> + #include <dt-bindings/interrupt-controller/arm-gic.h>
> + #include <dt-bindings/power/rk3588-power.h>
> +
> + gpu: gpu@fb000000 {
> + compatible = "rockchip,rk3588-mali", "arm,mali-valhall-csf";
> + reg = <0xfb000000 0x200000>;
> + interrupts = <GIC_SPI 92 IRQ_TYPE_LEVEL_HIGH 0>,
> + <GIC_SPI 93 IRQ_TYPE_LEVEL_HIGH 0>,
> + <GIC_SPI 94 IRQ_TYPE_LEVEL_HIGH 0>;
> + interrupt-names = "job", "mmu", "gpu";
> + clock-names = "core", "coregroup", "stacks";
> + clocks = <&cru CLK_GPU>, <&cru CLK_GPU_COREGROUP>,
> + <&cru CLK_GPU_STACKS>;
> + power-domains = <&power RK3588_PD_GPU>;
> + operating-points-v2 = <&gpu_opp_table>;
> + mali-supply = <&vdd_gpu_s0>;
> + sram-supply = <&vdd_gpu_mem_s0>;
> + };
> +
> + gpu_opp_table: opp-table {
Move this into the gpu node. With that,
Reviewed-by: Rob Herring <robh@kernel.org>
> + compatible = "operating-points-v2";
> + opp-300000000 {
> + opp-hz = /bits/ 64 <300000000>;
> + opp-microvolt = <675000 675000 850000>;
> + };
> + opp-400000000 {
> + opp-hz = /bits/ 64 <400000000>;
> + opp-microvolt = <675000 675000 850000>;
> + };
> + };
> +
> +...
> --
> 2.43.0
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 14/14] drm/panthor: Add an entry to MAINTAINERS
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (12 preceding siblings ...)
2024-01-22 16:30 ` Boris Brezillon
@ 2024-01-22 16:30 ` Boris Brezillon
2024-01-23 16:18 ` [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Heiko Stübner
` (3 subsequent siblings)
17 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-22 16:30 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
Add an entry for the Panthor driver to the MAINTAINERS file.
v4:
- Add Steve's R-b
v3:
- Add bindings document as an 'F:' line.
- Add Steven and Liviu as co-maintainers.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
---
MAINTAINERS | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index 9d959a6881f7..4d6e27f0a2e4 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -1648,6 +1648,17 @@ F: Documentation/gpu/panfrost.rst
F: drivers/gpu/drm/panfrost/
F: include/uapi/drm/panfrost_drm.h
+ARM MALI PANTHOR DRM DRIVER
+M: Boris Brezillon <boris.brezillon@collabora.com>
+M: Steven Price <steven.price@arm.com>
+M: Liviu Dudau <liviu.dudau@arm.com>
+L: dri-devel@lists.freedesktop.org
+S: Supported
+T: git git://anongit.freedesktop.org/drm/drm-misc
+F: Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
+F: drivers/gpu/drm/panthor/
+F: include/uapi/drm/panthor_drm.h
+
ARM MALI-DP DRM DRIVER
M: Liviu Dudau <liviu.dudau@arm.com>
S: Supported
--
2.43.0
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (13 preceding siblings ...)
2024-01-22 16:30 ` [PATCH v4 14/14] drm/panthor: Add an entry to MAINTAINERS Boris Brezillon
@ 2024-01-23 16:18 ` Heiko Stübner
2024-01-29 7:58 ` Boris Brezillon
` (2 subsequent siblings)
17 siblings, 0 replies; 51+ messages in thread
From: Heiko Stübner @ 2024-01-23 16:18 UTC (permalink / raw)
To: dri-devel, Boris Brezillon
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Boris Brezillon, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
Am Montag, 22. Januar 2024, 17:30:31 CET schrieb Boris Brezillon:
> Hello,
>
> This is the 4th version of the kernel driver for Mali CSF-based GPUs.
>
> A branch based on drm-misc-next and containing all the dependencies
> that are not yet available in drm-misc-next here[1], and another [2]
> containing extra patches to have things working on rk3588. The CSF
> firmware binary can be found here[3], and should be placed under
> /lib/firmware/arm/mali/arch10.8/mali_csffw.bin.
>
> The mesa MR adding v10 support on top of panthor is available here [4].
>
> Steve, I intentionally dropped your R-b on "drm/panthor: Add the heap
> logical block" and "drm/panthor: Add the scheduler logical block"
> because the tiler-OOM handling changed enough to require a new review
> IMHO.
>
> Regarding the GPL2+MIT relicensing, I collected Clément's R-b for the
> devfreq code, but am still lacking Alexey Sheplyakov for some bits in
> panthor_gpu.c. The rest of the code is either new, or covered by the
> Linaro, Arm and Collabora acks.
On a rk3588-jaguar with a corresponding mesa build, series:
Tested-by: Heiko Stuebner <heiko@sntech.de>
Thanks for all the work on this
Heiko
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (14 preceding siblings ...)
2024-01-23 16:18 ` [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Heiko Stübner
@ 2024-01-29 7:58 ` Boris Brezillon
2024-01-29 9:20 ` Andy Yan
2024-02-08 16:02 ` Liviu Dudau
17 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-01-29 7:58 UTC (permalink / raw)
To: dri-devel
Cc: Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand,
Marty E . Plummer, Robin Murphy, Faith Ekstrand
On Mon, 22 Jan 2024 17:30:31 +0100
Boris Brezillon <boris.brezillon@collabora.com> wrote:
> Hello,
>
> This is the 4th version of the kernel driver for Mali CSF-based GPUs.
>
> A branch based on drm-misc-next and containing all the dependencies
> that are not yet available in drm-misc-next here[1], and another [2]
> containing extra patches to have things working on rk3588. The CSF
> firmware binary can be found here[3], and should be placed under
> /lib/firmware/arm/mali/arch10.8/mali_csffw.bin.
>
> The mesa MR adding v10 support on top of panthor is available here [4].
>
> Steve, I intentionally dropped your R-b on "drm/panthor: Add the heap
> logical block" and "drm/panthor: Add the scheduler logical block"
> because the tiler-OOM handling changed enough to require a new review
> IMHO.
>
> Regarding the GPL2+MIT relicensing, I collected Clément's R-b for the
> devfreq code, but am still lacking Alexey Sheplyakov for some bits in
> panthor_gpu.c. The rest of the code is either new, or covered by the
> Linaro, Arm and Collabora acks.
>
> And here is a non-exhaustive changelog, check each commit for a detailed
> changelog.
>
> v4:
> - Fix various bugs in the VM logic
> - Address comments from Steven, Liviu, Ketil and Chris
> - Move tiler OOM handling out of the scheduler interrupt handling path
> so we can properly recover when the system runs out of memory, and
> panthor is blocked trying to allocate heap chunks
> - Rework the heap locking to support concurrent chunk allocation. Not
> sure if this is supposed to happen, but we need to be robust against
> userspace passing the same heap context to two scheduling groups.
> Wasn't needed before the tiler_oom rework, because heap allocation
> base serialized by the scheduler lock.
> - Make kernel BO destruction robust to NULL/ERR pointers
>
> v3;
> - Quite a few changes at the MMU/sched level to make the fix some
> race conditions and deadlocks
> - Addition of the a sync-only VM_BIND operation (to support
> vkQueueSparseBind with zero commands).
> - Addition of a VM_GET_STATE ioctl
> - Various cosmetic changes (see the commit changelogs for more details)
> - Various fixes (see the commit changelogs for more details)
>
> v2:
> - Rename the driver (pancsf -> panthor)
> - Split the commit adding the driver to ease review
> - Use drm_sched for dependency tracking/job submission
> - Add a VM_BIND ioctl
> - Add the concept of exclusive VM for BOs that are only ever mapped to a
> single VM
> - Document the code and uAPI
> - Add a DT binding doc
>
> Regards,
>
> Boris
>
> [1]https://gitlab.freedesktop.org/panfrost/linux/-/tree/panthor-v4
> [2]https://gitlab.freedesktop.org/panfrost/linux/-/tree/panthor-v4+rk3588
> [3]https://gitlab.com/firefly-linux/external/libmali/-/raw/firefly/firmware/g610/mali_csffw.bin
Here's a link to a more recent/maintained libmali tree:
[3]https://github.com/JeffyCN/mirrors/raw/libmali/firmware/g610/mali_csffw.bin
> [4]https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26358
>
> Boris Brezillon (13):
> drm/panthor: Add uAPI
> drm/panthor: Add GPU register definitions
> drm/panthor: Add the device logical block
> drm/panthor: Add the GPU logical block
> drm/panthor: Add GEM logical block
> drm/panthor: Add the devfreq logical block
> drm/panthor: Add the MMU/VM logical block
> drm/panthor: Add the FW logical block
> drm/panthor: Add the heap logical block
> drm/panthor: Add the scheduler logical block
> drm/panthor: Add the driver frontend block
> drm/panthor: Allow driver compilation
> drm/panthor: Add an entry to MAINTAINERS
>
> Liviu Dudau (1):
> dt-bindings: gpu: mali-valhall-csf: Add support for Arm Mali CSF GPUs
>
> .../bindings/gpu/arm,mali-valhall-csf.yaml | 147 +
> Documentation/gpu/driver-uapi.rst | 5 +
> MAINTAINERS | 11 +
> drivers/gpu/drm/Kconfig | 2 +
> drivers/gpu/drm/Makefile | 1 +
> drivers/gpu/drm/panthor/Kconfig | 23 +
> drivers/gpu/drm/panthor/Makefile | 15 +
> drivers/gpu/drm/panthor/panthor_devfreq.c | 283 ++
> drivers/gpu/drm/panthor/panthor_devfreq.h | 25 +
> drivers/gpu/drm/panthor/panthor_device.c | 544 +++
> drivers/gpu/drm/panthor/panthor_device.h | 393 ++
> drivers/gpu/drm/panthor/panthor_drv.c | 1470 +++++++
> drivers/gpu/drm/panthor/panthor_fw.c | 1334 +++++++
> drivers/gpu/drm/panthor/panthor_fw.h | 504 +++
> drivers/gpu/drm/panthor/panthor_gem.c | 228 ++
> drivers/gpu/drm/panthor/panthor_gem.h | 144 +
> drivers/gpu/drm/panthor/panthor_gpu.c | 482 +++
> drivers/gpu/drm/panthor/panthor_gpu.h | 52 +
> drivers/gpu/drm/panthor/panthor_heap.c | 596 +++
> drivers/gpu/drm/panthor/panthor_heap.h | 39 +
> drivers/gpu/drm/panthor/panthor_mmu.c | 2760 +++++++++++++
> drivers/gpu/drm/panthor/panthor_mmu.h | 102 +
> drivers/gpu/drm/panthor/panthor_regs.h | 239 ++
> drivers/gpu/drm/panthor/panthor_sched.c | 3500 +++++++++++++++++
> drivers/gpu/drm/panthor/panthor_sched.h | 48 +
> include/uapi/drm/panthor_drm.h | 945 +++++
> 26 files changed, 13892 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
> create mode 100644 drivers/gpu/drm/panthor/Kconfig
> create mode 100644 drivers/gpu/drm/panthor/Makefile
> create mode 100644 drivers/gpu/drm/panthor/panthor_devfreq.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_devfreq.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_device.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_device.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_drv.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_fw.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_fw.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_heap.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_heap.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_regs.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_sched.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_sched.h
> create mode 100644 include/uapi/drm/panthor_drm.h
>
^ permalink raw reply [flat|nested] 51+ messages in thread* Re:[PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (15 preceding siblings ...)
2024-01-29 7:58 ` Boris Brezillon
@ 2024-01-29 9:20 ` Andy Yan
2024-01-29 10:41 ` [PATCH " Boris Brezillon
2024-02-08 16:02 ` Liviu Dudau
17 siblings, 1 reply; 51+ messages in thread
From: Andy Yan @ 2024-01-29 9:20 UTC (permalink / raw)
To: Boris Brezillon
Cc: Neil Armstrong, Tatsuyuki Ishi, Nicolas Boichat, Daniel Stone,
Ketil Johnsen, Liviu Dudau, dri-devel, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand, kernel,
Robin Murphy, Faith Ekstrand, Marty E . Plummer
Hi Boris：
Thanks for you great work。
One thing please take note：
commit (arm64: dts: rockchip: rk3588: Add GPU nodes) in [1] seems remove the "disabled" status
of usb_host2_xhci, this may cause a boot issue on some boards that use combphy2_psu phy for other functions.
https://gitlab.freedesktop.org/panfrost/linux/-/commit/bf5b542294894219c2d366f6dd0d32a9dcf17252
At 2024-01-23 00:30:31, "Boris Brezillon" <boris.brezillon@collabora.com> wrote:
>Hello,
>
>This is the 4th version of the kernel driver for Mali CSF-based GPUs.
>
>A branch based on drm-misc-next and containing all the dependencies
>that are not yet available in drm-misc-next here[1], and another [2]
>containing extra patches to have things working on rk3588. The CSF
>firmware binary can be found here[3], and should be placed under
>/lib/firmware/arm/mali/arch10.8/mali_csffw.bin.
>
>The mesa MR adding v10 support on top of panthor is available here [4].
>
>Steve, I intentionally dropped your R-b on "drm/panthor: Add the heap
>logical block" and "drm/panthor: Add the scheduler logical block"
>because the tiler-OOM handling changed enough to require a new review
>IMHO.
>
>Regarding the GPL2+MIT relicensing, I collected Clément's R-b for the
>devfreq code, but am still lacking Alexey Sheplyakov for some bits in
>panthor_gpu.c. The rest of the code is either new, or covered by the
>Linaro, Arm and Collabora acks.
>
>And here is a non-exhaustive changelog, check each commit for a detailed
>changelog.
>
>v4:
>- Fix various bugs in the VM logic
>- Address comments from Steven, Liviu, Ketil and Chris
>- Move tiler OOM handling out of the scheduler interrupt handling path
> so we can properly recover when the system runs out of memory, and
> panthor is blocked trying to allocate heap chunks
>- Rework the heap locking to support concurrent chunk allocation. Not
> sure if this is supposed to happen, but we need to be robust against
> userspace passing the same heap context to two scheduling groups.
> Wasn't needed before the tiler_oom rework, because heap allocation
> base serialized by the scheduler lock.
>- Make kernel BO destruction robust to NULL/ERR pointers
>
>v3;
>- Quite a few changes at the MMU/sched level to make the fix some
> race conditions and deadlocks
>- Addition of the a sync-only VM_BIND operation (to support
> vkQueueSparseBind with zero commands).
>- Addition of a VM_GET_STATE ioctl
>- Various cosmetic changes (see the commit changelogs for more details)
>- Various fixes (see the commit changelogs for more details)
>
>v2:
>- Rename the driver (pancsf -> panthor)
>- Split the commit adding the driver to ease review
>- Use drm_sched for dependency tracking/job submission
>- Add a VM_BIND ioctl
>- Add the concept of exclusive VM for BOs that are only ever mapped to a
> single VM
>- Document the code and uAPI
>- Add a DT binding doc
>
>Regards,
>
>Boris
>
>[1]https://gitlab.freedesktop.org/panfrost/linux/-/tree/panthor-v4
>[2]https://gitlab.freedesktop.org/panfrost/linux/-/tree/panthor-v4+rk3588
>[3]https://gitlab.com/firefly-linux/external/libmali/-/raw/firefly/firmware/g610/mali_csffw.bin
>[4]https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26358
>
>Boris Brezillon (13):
> drm/panthor: Add uAPI
> drm/panthor: Add GPU register definitions
> drm/panthor: Add the device logical block
> drm/panthor: Add the GPU logical block
> drm/panthor: Add GEM logical block
> drm/panthor: Add the devfreq logical block
> drm/panthor: Add the MMU/VM logical block
> drm/panthor: Add the FW logical block
> drm/panthor: Add the heap logical block
> drm/panthor: Add the scheduler logical block
> drm/panthor: Add the driver frontend block
> drm/panthor: Allow driver compilation
> drm/panthor: Add an entry to MAINTAINERS
>
>Liviu Dudau (1):
> dt-bindings: gpu: mali-valhall-csf: Add support for Arm Mali CSF GPUs
>
> .../bindings/gpu/arm,mali-valhall-csf.yaml | 147 +
> Documentation/gpu/driver-uapi.rst | 5 +
> MAINTAINERS | 11 +
> drivers/gpu/drm/Kconfig | 2 +
> drivers/gpu/drm/Makefile | 1 +
> drivers/gpu/drm/panthor/Kconfig | 23 +
> drivers/gpu/drm/panthor/Makefile | 15 +
> drivers/gpu/drm/panthor/panthor_devfreq.c | 283 ++
> drivers/gpu/drm/panthor/panthor_devfreq.h | 25 +
> drivers/gpu/drm/panthor/panthor_device.c | 544 +++
> drivers/gpu/drm/panthor/panthor_device.h | 393 ++
> drivers/gpu/drm/panthor/panthor_drv.c | 1470 +++++++
> drivers/gpu/drm/panthor/panthor_fw.c | 1334 +++++++
> drivers/gpu/drm/panthor/panthor_fw.h | 504 +++
> drivers/gpu/drm/panthor/panthor_gem.c | 228 ++
> drivers/gpu/drm/panthor/panthor_gem.h | 144 +
> drivers/gpu/drm/panthor/panthor_gpu.c | 482 +++
> drivers/gpu/drm/panthor/panthor_gpu.h | 52 +
> drivers/gpu/drm/panthor/panthor_heap.c | 596 +++
> drivers/gpu/drm/panthor/panthor_heap.h | 39 +
> drivers/gpu/drm/panthor/panthor_mmu.c | 2760 +++++++++++++
> drivers/gpu/drm/panthor/panthor_mmu.h | 102 +
> drivers/gpu/drm/panthor/panthor_regs.h | 239 ++
> drivers/gpu/drm/panthor/panthor_sched.c | 3500 +++++++++++++++++
> drivers/gpu/drm/panthor/panthor_sched.h | 48 +
> include/uapi/drm/panthor_drm.h | 945 +++++
> 26 files changed, 13892 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
> create mode 100644 drivers/gpu/drm/panthor/Kconfig
> create mode 100644 drivers/gpu/drm/panthor/Makefile
> create mode 100644 drivers/gpu/drm/panthor/panthor_devfreq.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_devfreq.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_device.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_device.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_drv.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_fw.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_fw.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_heap.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_heap.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_regs.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_sched.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_sched.h
> create mode 100644 include/uapi/drm/panthor_drm.h
>
>--
>2.43.0
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-01-29 9:20 ` Andy Yan
@ 2024-01-29 10:41 ` Boris Brezillon
2024-02-04 1:14 ` Andy Yan
0 siblings, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-01-29 10:41 UTC (permalink / raw)
To: Andy Yan
Cc: Neil Armstrong, Tatsuyuki Ishi, Nicolas Boichat, Daniel Stone,
Ketil Johnsen, Liviu Dudau, dri-devel, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand, kernel,
Robin Murphy, Faith Ekstrand, Marty E . Plummer
On Mon, 29 Jan 2024 17:20:47 +0800 (CST)
"Andy Yan" <andyshrk@163.com> wrote:
> Hi Boris：
>
> Thanks for you great work。
>
> One thing please take note：
> commit (arm64: dts: rockchip: rk3588: Add GPU nodes) in [1] seems remove the "disabled" status
> of usb_host2_xhci, this may cause a boot issue on some boards that use combphy2_psu phy for other functions.
Oops, should be fixed in
https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next+rk3588
now.
Thanks,
Boris
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re:Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-01-29 10:41 ` [PATCH " Boris Brezillon
@ 2024-02-04 1:14 ` Andy Yan
2024-02-04 10:07 ` Boris Brezillon
2024-02-05 9:03 ` Boris Brezillon
0 siblings, 2 replies; 51+ messages in thread
From: Andy Yan @ 2024-02-04 1:14 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand,
Marty E . Plummer, Robin Murphy, Faith Ekstrand
Hi Boris：
I saw this warning sometimes（Run on a armbain based bookworm），not sure is a know issue or something else。
[15368.293031] systemd-journald[715]: Received client request to relinquish /var/log/journal/1bc4a340506142af9bd31a6a3d2170ba access.
[37743.040737] ------------[ cut here ]------------
[37743.040764] panthor fb000000.gpu: drm_WARN_ON(shmem->pages_use_count)
[37743.040890] WARNING: CPU: 2 PID: 5702 at drivers/gpu/drm/drm_gem_shmem_helper.c:158 drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
[37743.040929] Modules linked in: joydev rfkill sunrpc lz4hc lz4 zram binfmt_misc hantro_vpu crct10dif_ce v4l2_vp9 v4l2_h264 snd_soc_simple_amplifier v4l2_mem2mem videobuf2_dma_contig snd_soc_es8328_i2c videobuf2_memops rk_crypto2 snd_soc_es8328 videobuf2_v4l2 sm3_generic videodev crypto_engine sm3 rockchip_rng videobuf2_common nvmem_rockchip_otp snd_soc_rockchip_i2s_tdm snd_soc_hdmi_codec snd_soc_simple_card mc snd_soc_simple_card_utils snd_soc_core snd_compress ac97_bus snd_pcm_dmaengine snd_pcm snd_timer snd soundcore dm_mod ip_tables x_tables autofs4 dw_hdmi_qp_i2s_audio dw_hdmi_qp_cec rk808_regulator rockchipdrm dw_mipi_dsi dw_hdmi_qp dw_hdmi analogix_dp drm_dma_helper fusb302 display_connector rk8xx_spi drm_display_helper phy_rockchip_snps_pcie3 phy_rockchip_samsung_hdptx_hdmi panthor tcpm rk8xx_core cec drm_gpuvm gpu_sched drm_kms_helper drm_shmem_helper drm_exec r8169 drm pwm_bl adc_keys
[37743.041108] CPU: 2 PID: 5702 Comm: kworker/u16:8 Not tainted 6.8.0-rc1-edge-rockchip-rk3588 #2
[37743.041115] Hardware name: Rockchip RK3588 EVB1 V10 Board (DT)
[37743.041120] Workqueue: panthor-cleanup panthor_vm_bind_job_cleanup_op_ctx_work [panthor]
[37743.041151] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[37743.041157] pc : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
[37743.041169] lr : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
[37743.041181] sp : ffff80008d37bcc0
[37743.041184] x29: ffff80008d37bcc0 x28: ffff800081d379c0 x27: ffff800081d37000
[37743.041196] x26: ffff00019909a280 x25: ffff00019909a2c0 x24: ffff0001017a4c05
[37743.041206] x23: dead000000000100 x22: dead000000000122 x21: ffff0001627ac1a0
[37743.041217] x20: 0000000000000000 x19: ffff0001627ac000 x18: 0000000000000000
[37743.041227] x17: 000000040044ffff x16: 005000f2b5503510 x15: fffffffffff91b77
[37743.041238] x14: 0000000000000001 x13: 00000000000003c5 x12: 00000000ffffffea
[37743.041248] x11: 00000000ffffdfff x10: 00000000ffffdfff x9 : ffff800081e0e818
[37743.041259] x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
[37743.041269] x5 : 0000000000001fff x4 : 0000000000000000 x3 : ffff8000819a6008
[37743.041279] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00018465e900
[37743.041290] Call trace:
[37743.041293] drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
[37743.041306] panthor_gem_free_object+0x24/0xa0 [panthor]
[37743.041321] drm_gem_object_free+0x1c/0x30 [drm]
[37743.041452] panthor_vm_bo_put+0xc4/0x12c [panthor]
[37743.041475] panthor_vm_cleanup_op_ctx.constprop.0+0xb0/0x104 [panthor]
[37743.041491] panthor_vm_bind_job_cleanup_op_ctx_work+0x28/0xd0 [panthor]
[37743.041507] process_one_work+0x15c/0x3a4
[37743.041526] worker_thread+0x32c/0x438
[37743.041536] kthread+0x108/0x10c
[37743.041546] ret_from_fork+0x10/0x20
[37743.041557] ---[ end trace 0000000000000000 ]---
rk3588@rk3588-evb1:~$
rk3588@rk3588-evb1:~$ neofetch
------------------
█ █ █ █ █ █ █ █ █ █ █ OS: Armbian (24.2.0-trunk) aarch64
███████████████████████ Host: Rockchip RK3588 EVB1 V10 Board
▄▄██ ██▄▄ Kernel: 6.8.0-rc1-edge-rockchip-rk3588
▄▄██ ███████████ ██▄▄ Uptime: 13 hours, 28 mins
▄▄██ ██ ██ ██▄▄ Packages: 1455 (dpkg)
▄▄██ ██ ██ ██▄▄ Shell: bash 5.2.15
▄▄██ ██ ██ ██▄▄ Resolution: 3840x2160
▄▄██ █████████████ ██▄▄ Terminal: /dev/pts/1
▄▄██ ██ ██ ██▄▄ CPU: (8) @ 1.800GHz
▄▄██ ██ ██ ██▄▄ Memory: 2062MiB / 7687MiB
▄▄██ ██ ██ ██▄▄
▄▄██ ██▄▄
███████████████████████
█ █ █ █ █ █ █ █ █ █ █
在 2024-01-29 18:41:47，"Boris Brezillon" <boris.brezillon@collabora.com> 写道：
>On Mon, 29 Jan 2024 17:20:47 +0800 (CST)
>"Andy Yan" <andyshrk@163.com> wrote:
>
>> Hi Boris：
>>
>> Thanks for you great work。
>>
>> One thing please take note：
>> commit (arm64: dts: rockchip: rk3588: Add GPU nodes) in [1] seems remove the "disabled" status
>> of usb_host2_xhci, this may cause a boot issue on some boards that use combphy2_psu phy for other functions.
>
>Oops, should be fixed in
>https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next+rk3588
>now.
>
>Thanks,
>
>Boris
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-02-04 1:14 ` Andy Yan
@ 2024-02-04 10:07 ` Boris Brezillon
2024-02-04 10:36 ` Andy Yan
2024-02-05 9:03 ` Boris Brezillon
1 sibling, 1 reply; 51+ messages in thread
From: Boris Brezillon @ 2024-02-04 10:07 UTC (permalink / raw)
To: Andy Yan
Cc: dri-devel, Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand,
Marty E . Plummer, Robin Murphy, Faith Ekstrand
On Sun, 4 Feb 2024 09:14:44 +0800 (CST)
"Andy Yan" <andyshrk@163.com> wrote:
> Hi Boris：
> I saw this warning sometimes（Run on a armbain based bookworm），not sure is a know issue or something else。
No it's not, and I didn't manage to reproduce locally. Looks like
you're using a 6.8 kernel, but my panthor-v4/next branches are still
based on drm-misc-next from 2 weeks ago, which was based on a 6.7
kernel. Can you share the kernel branch you're using?
> [15368.293031] systemd-journald[715]: Received client request to relinquish /var/log/journal/1bc4a340506142af9bd31a6a3d2170ba access.
> [37743.040737] ------------[ cut here ]------------
> [37743.040764] panthor fb000000.gpu: drm_WARN_ON(shmem->pages_use_count)
> [37743.040890] WARNING: CPU: 2 PID: 5702 at drivers/gpu/drm/drm_gem_shmem_helper.c:158 drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> [37743.040929] Modules linked in: joydev rfkill sunrpc lz4hc lz4 zram binfmt_misc hantro_vpu crct10dif_ce v4l2_vp9 v4l2_h264 snd_soc_simple_amplifier v4l2_mem2mem videobuf2_dma_contig snd_soc_es8328_i2c videobuf2_memops rk_crypto2 snd_soc_es8328 videobuf2_v4l2 sm3_generic videodev crypto_engine sm3 rockchip_rng videobuf2_common nvmem_rockchip_otp snd_soc_rockchip_i2s_tdm snd_soc_hdmi_codec snd_soc_simple_card mc snd_soc_simple_card_utils snd_soc_core snd_compress ac97_bus snd_pcm_dmaengine snd_pcm snd_timer snd soundcore dm_mod ip_tables x_tables autofs4 dw_hdmi_qp_i2s_audio dw_hdmi_qp_cec rk808_regulator rockchipdrm dw_mipi_dsi dw_hdmi_qp dw_hdmi analogix_dp drm_dma_helper fusb302 display_connector rk8xx_spi drm_display_helper phy_rockchip_snps_pcie3 phy_rockchip_samsung_hdptx_hdmi panthor tcpm rk8xx_core cec drm_gpuvm gpu_sched drm_kms_helper drm_shmem_helper drm_exec r8169 drm pwm_bl adc_keys
> [37743.041108] CPU: 2 PID: 5702 Comm: kworker/u16:8 Not tainted 6.8.0-rc1-edge-rockchip-rk3588 #2
> [37743.041115] Hardware name: Rockchip RK3588 EVB1 V10 Board (DT)
> [37743.041120] Workqueue: panthor-cleanup panthor_vm_bind_job_cleanup_op_ctx_work [panthor]
> [37743.041151] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> [37743.041157] pc : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> [37743.041169] lr : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> [37743.041181] sp : ffff80008d37bcc0
> [37743.041184] x29: ffff80008d37bcc0 x28: ffff800081d379c0 x27: ffff800081d37000
> [37743.041196] x26: ffff00019909a280 x25: ffff00019909a2c0 x24: ffff0001017a4c05
> [37743.041206] x23: dead000000000100 x22: dead000000000122 x21: ffff0001627ac1a0
> [37743.041217] x20: 0000000000000000 x19: ffff0001627ac000 x18: 0000000000000000
> [37743.041227] x17: 000000040044ffff x16: 005000f2b5503510 x15: fffffffffff91b77
> [37743.041238] x14: 0000000000000001 x13: 00000000000003c5 x12: 00000000ffffffea
> [37743.041248] x11: 00000000ffffdfff x10: 00000000ffffdfff x9 : ffff800081e0e818
> [37743.041259] x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
> [37743.041269] x5 : 0000000000001fff x4 : 0000000000000000 x3 : ffff8000819a6008
> [37743.041279] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00018465e900
> [37743.041290] Call trace:
> [37743.041293] drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> [37743.041306] panthor_gem_free_object+0x24/0xa0 [panthor]
> [37743.041321] drm_gem_object_free+0x1c/0x30 [drm]
> [37743.041452] panthor_vm_bo_put+0xc4/0x12c [panthor]
I checked the _pin/_unpin calls in panthor, and they seem to be
balanced (we take a ref when we allocate a gpuvm_bo and release it
when the gpuvm_bo is gone). I wonder if something else is calling
_pin_pages() or _get_pages() without holding a GEM ref...
While investigating I found a double-cleanup in the code (see below)
which explains why those memset(0) were required in
panthor_vm_cleanup_op_ctx()), but I doubt it fixes your issue.
--->8---
diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
index d3ce29cd0662..5606ab4d6289 100644
--- a/drivers/gpu/drm/panthor/panthor_mmu.c
+++ b/drivers/gpu/drm/panthor/panthor_mmu.c
@@ -1085,17 +1085,12 @@ static void panthor_vm_cleanup_op_ctx(struct panthor_vm_op_ctx *op_ctx,
}
kfree(op_ctx->rsvd_page_tables.pages);
- memset(&op_ctx->rsvd_page_tables, 0, sizeof(op_ctx->rsvd_page_tables));
if (op_ctx->map.vm_bo)
panthor_vm_bo_put(op_ctx->map.vm_bo);
- memset(&op_ctx->map, 0, sizeof(op_ctx->map));
-
- for (u32 i = 0; i < ARRAY_SIZE(op_ctx->preallocated_vmas); i++) {
+ for (u32 i = 0; i < ARRAY_SIZE(op_ctx->preallocated_vmas); i++)
kfree(op_ctx->preallocated_vmas[i]);
- op_ctx->preallocated_vmas[i] = NULL;
- }
list_for_each_entry_safe(vma, tmp_vma, &op_ctx->returned_vmas, node) {
list_del(&vma->node);
@@ -2382,7 +2377,6 @@ static void panthor_vm_bind_job_cleanup_op_ctx_work(struct work_struct *work)
struct panthor_vm_bind_job *job =
container_of(work, struct panthor_vm_bind_job, cleanup_op_ctx_work);
- panthor_vm_cleanup_op_ctx(&job->ctx, job->vm);
panthor_vm_bind_job_put(&job->base);
}
^ permalink raw reply related [flat|nested] 51+ messages in thread* Re:Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-02-04 10:07 ` Boris Brezillon
@ 2024-02-04 10:36 ` Andy Yan
0 siblings, 0 replies; 51+ messages in thread
From: Andy Yan @ 2024-02-04 10:36 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand,
Marty E . Plummer, Robin Murphy, Faith Ekstrand
Hi Boris:
在 2024-02-04 18:07:56，"Boris Brezillon" <boris.brezillon@collabora.com> 写道：
>On Sun, 4 Feb 2024 09:14:44 +0800 (CST)
>"Andy Yan" <andyshrk@163.com> wrote:
>
>> Hi Boris：
>> I saw this warning sometimes（Run on a armbain based bookworm），not sure is a know issue or something else。
>
>No it's not, and I didn't manage to reproduce locally. Looks like
>you're using a 6.8 kernel, but my panthor-v4/next branches are still
>based on drm-misc-next from 2 weeks ago, which was based on a 6.7
>kernel. Can you share the kernel branch you're using?
>
Here is my kernel branch:
https://github.com/andyshrk/linux/commits/linux-6.8-rc1-rk3588_panthor_v4/
>> [15368.293031] systemd-journald[715]: Received client request to relinquish /var/log/journal/1bc4a340506142af9bd31a6a3d2170ba access.
>> [37743.040737] ------------[ cut here ]------------
>> [37743.040764] panthor fb000000.gpu: drm_WARN_ON(shmem->pages_use_count)
>> [37743.040890] WARNING: CPU: 2 PID: 5702 at drivers/gpu/drm/drm_gem_shmem_helper.c:158 drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.040929] Modules linked in: joydev rfkill sunrpc lz4hc lz4 zram binfmt_misc hantro_vpu crct10dif_ce v4l2_vp9 v4l2_h264 snd_soc_simple_amplifier v4l2_mem2mem videobuf2_dma_contig snd_soc_es8328_i2c videobuf2_memops rk_crypto2 snd_soc_es8328 videobuf2_v4l2 sm3_generic videodev crypto_engine sm3 rockchip_rng videobuf2_common nvmem_rockchip_otp snd_soc_rockchip_i2s_tdm snd_soc_hdmi_codec snd_soc_simple_card mc snd_soc_simple_card_utils snd_soc_core snd_compress ac97_bus snd_pcm_dmaengine snd_pcm snd_timer snd soundcore dm_mod ip_tables x_tables autofs4 dw_hdmi_qp_i2s_audio dw_hdmi_qp_cec rk808_regulator rockchipdrm dw_mipi_dsi dw_hdmi_qp dw_hdmi analogix_dp drm_dma_helper fusb302 display_connector rk8xx_spi drm_display_helper phy_rockchip_snps_pcie3 phy_rockchip_samsung_hdptx_hdmi panthor tcpm rk8xx_core cec drm_gpuvm gpu_sched drm_kms_helper drm_shmem_helper drm_exec r8169 drm pwm_bl adc_keys
>> [37743.041108] CPU: 2 PID: 5702 Comm: kworker/u16:8 Not tainted 6.8.0-rc1-edge-rockchip-rk3588 #2
>> [37743.041115] Hardware name: Rockchip RK3588 EVB1 V10 Board (DT)
>> [37743.041120] Workqueue: panthor-cleanup panthor_vm_bind_job_cleanup_op_ctx_work [panthor]
>> [37743.041151] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>> [37743.041157] pc : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041169] lr : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041181] sp : ffff80008d37bcc0
>> [37743.041184] x29: ffff80008d37bcc0 x28: ffff800081d379c0 x27: ffff800081d37000
>> [37743.041196] x26: ffff00019909a280 x25: ffff00019909a2c0 x24: ffff0001017a4c05
>> [37743.041206] x23: dead000000000100 x22: dead000000000122 x21: ffff0001627ac1a0
>> [37743.041217] x20: 0000000000000000 x19: ffff0001627ac000 x18: 0000000000000000
>> [37743.041227] x17: 000000040044ffff x16: 005000f2b5503510 x15: fffffffffff91b77
>> [37743.041238] x14: 0000000000000001 x13: 00000000000003c5 x12: 00000000ffffffea
>> [37743.041248] x11: 00000000ffffdfff x10: 00000000ffffdfff x9 : ffff800081e0e818
>> [37743.041259] x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
>> [37743.041269] x5 : 0000000000001fff x4 : 0000000000000000 x3 : ffff8000819a6008
>> [37743.041279] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00018465e900
>> [37743.041290] Call trace:
>> [37743.041293] drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041306] panthor_gem_free_object+0x24/0xa0 [panthor]
>> [37743.041321] drm_gem_object_free+0x1c/0x30 [drm]
>> [37743.041452] panthor_vm_bo_put+0xc4/0x12c [panthor]
>
>I checked the _pin/_unpin calls in panthor, and they seem to be
>balanced (we take a ref when we allocate a gpuvm_bo and release it
>when the gpuvm_bo is gone). I wonder if something else is calling
>_pin_pages() or _get_pages() without holding a GEM ref...
>
>While investigating I found a double-cleanup in the code (see below)
>which explains why those memset(0) were required in
>panthor_vm_cleanup_op_ctx()), but I doubt it fixes your issue.
>
>--->8---
>diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
>index d3ce29cd0662..5606ab4d6289 100644
>--- a/drivers/gpu/drm/panthor/panthor_mmu.c
>+++ b/drivers/gpu/drm/panthor/panthor_mmu.c
>@@ -1085,17 +1085,12 @@ static void panthor_vm_cleanup_op_ctx(struct panthor_vm_op_ctx *op_ctx,
> }
>
> kfree(op_ctx->rsvd_page_tables.pages);
>- memset(&op_ctx->rsvd_page_tables, 0, sizeof(op_ctx->rsvd_page_tables));
>
> if (op_ctx->map.vm_bo)
> panthor_vm_bo_put(op_ctx->map.vm_bo);
>
>- memset(&op_ctx->map, 0, sizeof(op_ctx->map));
>-
>- for (u32 i = 0; i < ARRAY_SIZE(op_ctx->preallocated_vmas); i++) {
>+ for (u32 i = 0; i < ARRAY_SIZE(op_ctx->preallocated_vmas); i++)
> kfree(op_ctx->preallocated_vmas[i]);
>- op_ctx->preallocated_vmas[i] = NULL;
>- }
>
> list_for_each_entry_safe(vma, tmp_vma, &op_ctx->returned_vmas, node) {
> list_del(&vma->node);
>@@ -2382,7 +2377,6 @@ static void panthor_vm_bind_job_cleanup_op_ctx_work(struct work_struct *work)
> struct panthor_vm_bind_job *job =
> container_of(work, struct panthor_vm_bind_job, cleanup_op_ctx_work);
>
>- panthor_vm_cleanup_op_ctx(&job->ctx, job->vm);
> panthor_vm_bind_job_put(&job->base);
> }
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-02-04 1:14 ` Andy Yan
2024-02-04 10:07 ` Boris Brezillon
@ 2024-02-05 9:03 ` Boris Brezillon
2024-02-05 9:41 ` Andy Yan
` (2 more replies)
1 sibling, 3 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-02-05 9:03 UTC (permalink / raw)
To: Andy Yan, Danilo Krummrich
Cc: dri-devel, Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand,
Marty E . Plummer, Robin Murphy, Faith Ekstrand
+Danilo for the panthor gpuvm-needs update.
On Sun, 4 Feb 2024 09:14:44 +0800 (CST)
"Andy Yan" <andyshrk@163.com> wrote:
> Hi Boris：
> I saw this warning sometimes（Run on a armbain based bookworm），not sure is a know issue or something else。
> [15368.293031] systemd-journald[715]: Received client request to relinquish /var/log/journal/1bc4a340506142af9bd31a6a3d2170ba access.
> [37743.040737] ------------[ cut here ]------------
> [37743.040764] panthor fb000000.gpu: drm_WARN_ON(shmem->pages_use_count)
> [37743.040890] WARNING: CPU: 2 PID: 5702 at drivers/gpu/drm/drm_gem_shmem_helper.c:158 drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> [37743.040929] Modules linked in: joydev rfkill sunrpc lz4hc lz4 zram binfmt_misc hantro_vpu crct10dif_ce v4l2_vp9 v4l2_h264 snd_soc_simple_amplifier v4l2_mem2mem videobuf2_dma_contig snd_soc_es8328_i2c videobuf2_memops rk_crypto2 snd_soc_es8328 videobuf2_v4l2 sm3_generic videodev crypto_engine sm3 rockchip_rng videobuf2_common nvmem_rockchip_otp snd_soc_rockchip_i2s_tdm snd_soc_hdmi_codec snd_soc_simple_card mc snd_soc_simple_card_utils snd_soc_core snd_compress ac97_bus snd_pcm_dmaengine snd_pcm snd_timer snd soundcore dm_mod ip_tables x_tables autofs4 dw_hdmi_qp_i2s_audio dw_hdmi_qp_cec rk808_regulator rockchipdrm dw_mipi_dsi dw_hdmi_qp dw_hdmi analogix_dp drm_dma_helper fusb302 display_connector rk8xx_spi drm_display_helper phy_rockchip_snps_pcie3 phy_rockchip_samsung_hdptx_hdmi panthor tcpm rk8xx_core cec drm_gpuvm gpu_sched drm_kms_helper drm_shmem_helper drm_exec r8169 drm pwm_bl adc_keys
> [37743.041108] CPU: 2 PID: 5702 Comm: kworker/u16:8 Not tainted 6.8.0-rc1-edge-rockchip-rk3588 #2
> [37743.041115] Hardware name: Rockchip RK3588 EVB1 V10 Board (DT)
> [37743.041120] Workqueue: panthor-cleanup panthor_vm_bind_job_cleanup_op_ctx_work [panthor]
> [37743.041151] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> [37743.041157] pc : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> [37743.041169] lr : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> [37743.041181] sp : ffff80008d37bcc0
> [37743.041184] x29: ffff80008d37bcc0 x28: ffff800081d379c0 x27: ffff800081d37000
> [37743.041196] x26: ffff00019909a280 x25: ffff00019909a2c0 x24: ffff0001017a4c05
> [37743.041206] x23: dead000000000100 x22: dead000000000122 x21: ffff0001627ac1a0
> [37743.041217] x20: 0000000000000000 x19: ffff0001627ac000 x18: 0000000000000000
> [37743.041227] x17: 000000040044ffff x16: 005000f2b5503510 x15: fffffffffff91b77
> [37743.041238] x14: 0000000000000001 x13: 00000000000003c5 x12: 00000000ffffffea
> [37743.041248] x11: 00000000ffffdfff x10: 00000000ffffdfff x9 : ffff800081e0e818
> [37743.041259] x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
> [37743.041269] x5 : 0000000000001fff x4 : 0000000000000000 x3 : ffff8000819a6008
> [37743.041279] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00018465e900
> [37743.041290] Call trace:
> [37743.041293] drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> [37743.041306] panthor_gem_free_object+0x24/0xa0 [panthor]
> [37743.041321] drm_gem_object_free+0x1c/0x30 [drm]
> [37743.041452] panthor_vm_bo_put+0xc4/0x12c [panthor]
> [37743.041475] panthor_vm_cleanup_op_ctx.constprop.0+0xb0/0x104 [panthor]
> [37743.041491] panthor_vm_bind_job_cleanup_op_ctx_work+0x28/0xd0 [panthor]
Ok, I think I found the culprit: there's a race between
the drm_gpuvm_bo_put() call in panthor_vm_bo_put() and the list
iteration done by drm_gpuvm_prepare_objects(). Because we're not
setting DRM_GPUVM_RESV_PROTECTED, the code goes through the 'lockless'
iteration loop, and takes/release a vm_bo ref at each iteration. This
means our 'were we the last vm_bo user?' test in panthor_vm_bo_put()
might return false even if we were actually the last user, and when
for_each_vm_bo_in_list() releases the ref it acquired, it not only leaks
the pin reference, thus leaving GEM pages pinned (which explains this
WARN_ON() splat), but it also calls drm_gpuvm_bo_destroy() in a path
where we don't hold the GPUVA list lock, which is bad.
Long story short, I'll have to use DRM_GPUVM_RESV_PROTECTED, which is
fine because I'm deferring vm_bo removal to a work where taking the VM
resv lock is allowed. Since I was the one asking for this lockless
iterator in the first place, I wonder if we should kill that and make
DRM_GPUVM_RESV_PROTECTED the default (this would greatly simplify
the code). AFAICT, The PowerVR driver shouldn't be impacted because it's
using drm_gpuvm in synchronous mode only, and Xe already uses the
resv-protected mode. That leaves Nouveau, but IIRC, it's also doing VM
updates in the ioctl path.
Danilo, any opinions?
Andy, I pushed a new version to the panthor-next [1] and
panthor-next+rk3588 [2] branches. The fix I'm talking about is [3], but
you probably want to consider taking all the fixups in your branch.
Regards,
Boris
[1]https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next
[2]https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next+rk3588
[3]https://gitlab.freedesktop.org/panfrost/linux/-/commit/df48c09662a403275e76e679ee004085badea7c1
^ permalink raw reply [flat|nested] 51+ messages in thread* Re:Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-02-05 9:03 ` Boris Brezillon
@ 2024-02-05 9:41 ` Andy Yan
2024-02-05 9:54 ` Danilo Krummrich
2024-02-06 10:35 ` Andy Yan
2 siblings, 0 replies; 51+ messages in thread
From: Andy Yan @ 2024-02-05 9:41 UTC (permalink / raw)
To: Boris Brezillon
Cc: Danilo Krummrich, dri-devel, Tatsuyuki Ishi, Nicolas Boichat,
kernel, Daniel Stone, Neil Armstrong, Ketil Johnsen, Liviu Dudau,
Steven Price, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
Hi Boris：
在 2024-02-05 17:03:21，"Boris Brezillon" <boris.brezillon@collabora.com> 写道：
>+Danilo for the panthor gpuvm-needs update.
>
>On Sun, 4 Feb 2024 09:14:44 +0800 (CST)
>"Andy Yan" <andyshrk@163.com> wrote:
>
>> Hi Boris：
>> I saw this warning sometimes（Run on a armbain based bookworm），not sure is a know issue or something else。
>> [15368.293031] systemd-journald[715]: Received client request to relinquish /var/log/journal/1bc4a340506142af9bd31a6a3d2170ba access.
>> [37743.040737] ------------[ cut here ]------------
>> [37743.040764] panthor fb000000.gpu: drm_WARN_ON(shmem->pages_use_count)
>> [37743.040890] WARNING: CPU: 2 PID: 5702 at drivers/gpu/drm/drm_gem_shmem_helper.c:158 drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.040929] Modules linked in: joydev rfkill sunrpc lz4hc lz4 zram binfmt_misc hantro_vpu crct10dif_ce v4l2_vp9 v4l2_h264 snd_soc_simple_amplifier v4l2_mem2mem videobuf2_dma_contig snd_soc_es8328_i2c videobuf2_memops rk_crypto2 snd_soc_es8328 videobuf2_v4l2 sm3_generic videodev crypto_engine sm3 rockchip_rng videobuf2_common nvmem_rockchip_otp snd_soc_rockchip_i2s_tdm snd_soc_hdmi_codec snd_soc_simple_card mc snd_soc_simple_card_utils snd_soc_core snd_compress ac97_bus snd_pcm_dmaengine snd_pcm snd_timer snd soundcore dm_mod ip_tables x_tables autofs4 dw_hdmi_qp_i2s_audio dw_hdmi_qp_cec rk808_regulator rockchipdrm dw_mipi_dsi dw_hdmi_qp dw_hdmi analogix_dp drm_dma_helper fusb302 display_connector rk8xx_spi drm_display_helper phy_rockchip_snps_pcie3 phy_rockchip_samsung_hdptx_hdmi panthor tcpm rk8xx_core cec drm_gpuvm gpu_sched drm_kms_helper drm_shmem_helper drm_exec r8169 drm pwm_bl adc_keys
>> [37743.041108] CPU: 2 PID: 5702 Comm: kworker/u16:8 Not tainted 6.8.0-rc1-edge-rockchip-rk3588 #2
>> [37743.041115] Hardware name: Rockchip RK3588 EVB1 V10 Board (DT)
>> [37743.041120] Workqueue: panthor-cleanup panthor_vm_bind_job_cleanup_op_ctx_work [panthor]
>> [37743.041151] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>> [37743.041157] pc : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041169] lr : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041181] sp : ffff80008d37bcc0
>> [37743.041184] x29: ffff80008d37bcc0 x28: ffff800081d379c0 x27: ffff800081d37000
>> [37743.041196] x26: ffff00019909a280 x25: ffff00019909a2c0 x24: ffff0001017a4c05
>> [37743.041206] x23: dead000000000100 x22: dead000000000122 x21: ffff0001627ac1a0
>> [37743.041217] x20: 0000000000000000 x19: ffff0001627ac000 x18: 0000000000000000
>> [37743.041227] x17: 000000040044ffff x16: 005000f2b5503510 x15: fffffffffff91b77
>> [37743.041238] x14: 0000000000000001 x13: 00000000000003c5 x12: 00000000ffffffea
>> [37743.041248] x11: 00000000ffffdfff x10: 00000000ffffdfff x9 : ffff800081e0e818
>> [37743.041259] x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
>> [37743.041269] x5 : 0000000000001fff x4 : 0000000000000000 x3 : ffff8000819a6008
>> [37743.041279] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00018465e900
>> [37743.041290] Call trace:
>> [37743.041293] drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041306] panthor_gem_free_object+0x24/0xa0 [panthor]
>> [37743.041321] drm_gem_object_free+0x1c/0x30 [drm]
>> [37743.041452] panthor_vm_bo_put+0xc4/0x12c [panthor]
>> [37743.041475] panthor_vm_cleanup_op_ctx.constprop.0+0xb0/0x104 [panthor]
>> [37743.041491] panthor_vm_bind_job_cleanup_op_ctx_work+0x28/0xd0 [panthor]
>
>Ok, I think I found the culprit: there's a race between
>the drm_gpuvm_bo_put() call in panthor_vm_bo_put() and the list
>iteration done by drm_gpuvm_prepare_objects(). Because we're not
>setting DRM_GPUVM_RESV_PROTECTED, the code goes through the 'lockless'
>iteration loop, and takes/release a vm_bo ref at each iteration. This
>means our 'were we the last vm_bo user?' test in panthor_vm_bo_put()
>might return false even if we were actually the last user, and when
>for_each_vm_bo_in_list() releases the ref it acquired, it not only leaks
>the pin reference, thus leaving GEM pages pinned (which explains this
>WARN_ON() splat), but it also calls drm_gpuvm_bo_destroy() in a path
>where we don't hold the GPUVA list lock, which is bad.
>
>Long story short, I'll have to use DRM_GPUVM_RESV_PROTECTED, which is
>fine because I'm deferring vm_bo removal to a work where taking the VM
>resv lock is allowed. Since I was the one asking for this lockless
>iterator in the first place, I wonder if we should kill that and make
>DRM_GPUVM_RESV_PROTECTED the default (this would greatly simplify
>the code). AFAICT, The PowerVR driver shouldn't be impacted because it's
>using drm_gpuvm in synchronous mode only, and Xe already uses the
>resv-protected mode. That leaves Nouveau, but IIRC, it's also doing VM
>updates in the ioctl path.
>
>Danilo, any opinions?
>
>Andy, I pushed a new version to the panthor-next [1] and
>panthor-next+rk3588 [2] branches. The fix I'm talking about is [3], but
>you probably want to consider taking all the fixups in your branch.
Okay， I will have a try。
Thanks
>
>Regards,
>
>Boris
>
>[1]https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next
>[2]https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next+rk3588
>[3]https://gitlab.freedesktop.org/panfrost/linux/-/commit/df48c09662a403275e76e679ee004085badea7c1
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-02-05 9:03 ` Boris Brezillon
2024-02-05 9:41 ` Andy Yan
@ 2024-02-05 9:54 ` Danilo Krummrich
2024-02-05 10:31 ` Boris Brezillon
2024-02-06 10:35 ` Andy Yan
2 siblings, 1 reply; 51+ messages in thread
From: Danilo Krummrich @ 2024-02-05 9:54 UTC (permalink / raw)
To: Boris Brezillon, Andy Yan
Cc: dri-devel, Tatsuyuki Ishi, Nicolas Boichat, kernel, Daniel Stone,
Neil Armstrong, Ketil Johnsen, Liviu Dudau, Steven Price,
Clément Péron, Daniel Vetter, Chris Diamand,
Marty E . Plummer, Robin Murphy, Faith Ekstrand
On 2/5/24 10:03, Boris Brezillon wrote:
> +Danilo for the panthor gpuvm-needs update.
>
> On Sun, 4 Feb 2024 09:14:44 +0800 (CST)
> "Andy Yan" <andyshrk@163.com> wrote:
>
>> Hi Boris：
>> I saw this warning sometimes（Run on a armbain based bookworm），not sure is a know issue or something else。
>> [15368.293031] systemd-journald[715]: Received client request to relinquish /var/log/journal/1bc4a340506142af9bd31a6a3d2170ba access.
>> [37743.040737] ------------[ cut here ]------------
>> [37743.040764] panthor fb000000.gpu: drm_WARN_ON(shmem->pages_use_count)
>> [37743.040890] WARNING: CPU: 2 PID: 5702 at drivers/gpu/drm/drm_gem_shmem_helper.c:158 drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.040929] Modules linked in: joydev rfkill sunrpc lz4hc lz4 zram binfmt_misc hantro_vpu crct10dif_ce v4l2_vp9 v4l2_h264 snd_soc_simple_amplifier v4l2_mem2mem videobuf2_dma_contig snd_soc_es8328_i2c videobuf2_memops rk_crypto2 snd_soc_es8328 videobuf2_v4l2 sm3_generic videodev crypto_engine sm3 rockchip_rng videobuf2_common nvmem_rockchip_otp snd_soc_rockchip_i2s_tdm snd_soc_hdmi_codec snd_soc_simple_card mc snd_soc_simple_card_utils snd_soc_core snd_compress ac97_bus snd_pcm_dmaengine snd_pcm snd_timer snd soundcore dm_mod ip_tables x_tables autofs4 dw_hdmi_qp_i2s_audio dw_hdmi_qp_cec rk808_regulator rockchipdrm dw_mipi_dsi dw_hdmi_qp dw_hdmi analogix_dp drm_dma_helper fusb302 display_connector rk8xx_spi drm_display_helper phy_rockchip_snps_pcie3 phy_rockchip_samsung_hdptx_hdmi panthor tcpm rk8xx_core cec drm_gpuvm gpu_sched drm_kms_helper drm_shmem_helper drm_exec r8169 drm pwm_bl adc_keys
>> [37743.041108] CPU: 2 PID: 5702 Comm: kworker/u16:8 Not tainted 6.8.0-rc1-edge-rockchip-rk3588 #2
>> [37743.041115] Hardware name: Rockchip RK3588 EVB1 V10 Board (DT)
>> [37743.041120] Workqueue: panthor-cleanup panthor_vm_bind_job_cleanup_op_ctx_work [panthor]
>> [37743.041151] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>> [37743.041157] pc : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041169] lr : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041181] sp : ffff80008d37bcc0
>> [37743.041184] x29: ffff80008d37bcc0 x28: ffff800081d379c0 x27: ffff800081d37000
>> [37743.041196] x26: ffff00019909a280 x25: ffff00019909a2c0 x24: ffff0001017a4c05
>> [37743.041206] x23: dead000000000100 x22: dead000000000122 x21: ffff0001627ac1a0
>> [37743.041217] x20: 0000000000000000 x19: ffff0001627ac000 x18: 0000000000000000
>> [37743.041227] x17: 000000040044ffff x16: 005000f2b5503510 x15: fffffffffff91b77
>> [37743.041238] x14: 0000000000000001 x13: 00000000000003c5 x12: 00000000ffffffea
>> [37743.041248] x11: 00000000ffffdfff x10: 00000000ffffdfff x9 : ffff800081e0e818
>> [37743.041259] x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
>> [37743.041269] x5 : 0000000000001fff x4 : 0000000000000000 x3 : ffff8000819a6008
>> [37743.041279] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00018465e900
>> [37743.041290] Call trace:
>> [37743.041293] drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041306] panthor_gem_free_object+0x24/0xa0 [panthor]
>> [37743.041321] drm_gem_object_free+0x1c/0x30 [drm]
>> [37743.041452] panthor_vm_bo_put+0xc4/0x12c [panthor]
>> [37743.041475] panthor_vm_cleanup_op_ctx.constprop.0+0xb0/0x104 [panthor]
>> [37743.041491] panthor_vm_bind_job_cleanup_op_ctx_work+0x28/0xd0 [panthor]
>
> Ok, I think I found the culprit: there's a race between
> the drm_gpuvm_bo_put() call in panthor_vm_bo_put() and the list
> iteration done by drm_gpuvm_prepare_objects(). Because we're not
> setting DRM_GPUVM_RESV_PROTECTED, the code goes through the 'lockless'
> iteration loop, and takes/release a vm_bo ref at each iteration. This
> means our 'were we the last vm_bo user?' test in panthor_vm_bo_put()
> might return false even if we were actually the last user, and when
> for_each_vm_bo_in_list() releases the ref it acquired, it not only leaks
> the pin reference, thus leaving GEM pages pinned (which explains this
> WARN_ON() splat), but it also calls drm_gpuvm_bo_destroy() in a path
> where we don't hold the GPUVA list lock, which is bad.
IIUC, GPUVM generally behaves correctly. It's just that the return value
of drm_gpuvm_bo_put() needs to be treated with care.
Maybe we should extend c50a291d621a ("drm/gpuvm: Let drm_gpuvm_bo_put()
report when the vm_bo object is destroyed") by a note explaining this
unexpected case, or, if not required anymore, simply revert the patch.
>
> Long story short, I'll have to use DRM_GPUVM_RESV_PROTECTED, which is
> fine because I'm deferring vm_bo removal to a work where taking the VM
> resv lock is allowed. Since I was the one asking for this lockless
> iterator in the first place, I wonder if we should kill that and make
> DRM_GPUVM_RESV_PROTECTED the default (this would greatly simplify
> the code). AFAICT, The PowerVR driver shouldn't be impacted because it's
> using drm_gpuvm in synchronous mode only, and Xe already uses the
> resv-protected mode. That leaves Nouveau, but IIRC, it's also doing VM
> updates in the ioctl path.
>
> Danilo, any opinions?
I agree that we should remove it once we got enough evidence that updating
the VA space in the asynchronous path isn't going to be a thing. I'm not
entirely sure, whether we'll attempt to re-work Nouveau, but I'd like to try
this approach with Nova. Hence, I'd still like to leave it in for a while.
OOC, knowing that you went a little back and forth with one or the other
approach, last I heard was that using DRM_GPUVM_RESV_PROTECTED didn't work
out well with shmem buffers. How did you fix it?
>
> Andy, I pushed a new version to the panthor-next [1] and
> panthor-next+rk3588 [2] branches. The fix I'm talking about is [3], but
> you probably want to consider taking all the fixups in your branch.
>
> Regards,
>
> Boris
>
> [1]https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next
> [2]https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next+rk3588
> [3]https://gitlab.freedesktop.org/panfrost/linux/-/commit/df48c09662a403275e76e679ee004085badea7c1
>
^ permalink raw reply [flat|nested] 51+ messages in thread* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-02-05 9:54 ` Danilo Krummrich
@ 2024-02-05 10:31 ` Boris Brezillon
0 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-02-05 10:31 UTC (permalink / raw)
To: Danilo Krummrich
Cc: Andy Yan, dri-devel, Tatsuyuki Ishi, Nicolas Boichat, kernel,
Daniel Stone, Neil Armstrong, Ketil Johnsen, Liviu Dudau,
Steven Price, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
On Mon, 5 Feb 2024 10:54:05 +0100
Danilo Krummrich <dakr@redhat.com> wrote:
> On 2/5/24 10:03, Boris Brezillon wrote:
> > +Danilo for the panthor gpuvm-needs update.
> >
> > On Sun, 4 Feb 2024 09:14:44 +0800 (CST)
> > "Andy Yan" <andyshrk@163.com> wrote:
> >
> >> Hi Boris：
> >> I saw this warning sometimes（Run on a armbain based bookworm），not sure is a know issue or something else。
> >> [15368.293031] systemd-journald[715]: Received client request to relinquish /var/log/journal/1bc4a340506142af9bd31a6a3d2170ba access.
> >> [37743.040737] ------------[ cut here ]------------
> >> [37743.040764] panthor fb000000.gpu: drm_WARN_ON(shmem->pages_use_count)
> >> [37743.040890] WARNING: CPU: 2 PID: 5702 at drivers/gpu/drm/drm_gem_shmem_helper.c:158 drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> >> [37743.040929] Modules linked in: joydev rfkill sunrpc lz4hc lz4 zram binfmt_misc hantro_vpu crct10dif_ce v4l2_vp9 v4l2_h264 snd_soc_simple_amplifier v4l2_mem2mem videobuf2_dma_contig snd_soc_es8328_i2c videobuf2_memops rk_crypto2 snd_soc_es8328 videobuf2_v4l2 sm3_generic videodev crypto_engine sm3 rockchip_rng videobuf2_common nvmem_rockchip_otp snd_soc_rockchip_i2s_tdm snd_soc_hdmi_codec snd_soc_simple_card mc snd_soc_simple_card_utils snd_soc_core snd_compress ac97_bus snd_pcm_dmaengine snd_pcm snd_timer snd soundcore dm_mod ip_tables x_tables autofs4 dw_hdmi_qp_i2s_audio dw_hdmi_qp_cec rk808_regulator rockchipdrm dw_mipi_dsi dw_hdmi_qp dw_hdmi analogix_dp drm_dma_helper fusb302 display_connector rk8xx_spi drm_display_helper phy_rockchip_snps_pcie3 phy_rockchip_samsung_hdptx_hdmi panthor tcpm rk8xx_core cec drm_gpuvm gpu_sched drm_kms_helper drm_shmem_helper drm_exec r8169 drm pwm_bl adc_keys
> >> [37743.041108] CPU: 2 PID: 5702 Comm: kworker/u16:8 Not tainted 6.8.0-rc1-edge-rockchip-rk3588 #2
> >> [37743.041115] Hardware name: Rockchip RK3588 EVB1 V10 Board (DT)
> >> [37743.041120] Workqueue: panthor-cleanup panthor_vm_bind_job_cleanup_op_ctx_work [panthor]
> >> [37743.041151] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> >> [37743.041157] pc : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> >> [37743.041169] lr : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> >> [37743.041181] sp : ffff80008d37bcc0
> >> [37743.041184] x29: ffff80008d37bcc0 x28: ffff800081d379c0 x27: ffff800081d37000
> >> [37743.041196] x26: ffff00019909a280 x25: ffff00019909a2c0 x24: ffff0001017a4c05
> >> [37743.041206] x23: dead000000000100 x22: dead000000000122 x21: ffff0001627ac1a0
> >> [37743.041217] x20: 0000000000000000 x19: ffff0001627ac000 x18: 0000000000000000
> >> [37743.041227] x17: 000000040044ffff x16: 005000f2b5503510 x15: fffffffffff91b77
> >> [37743.041238] x14: 0000000000000001 x13: 00000000000003c5 x12: 00000000ffffffea
> >> [37743.041248] x11: 00000000ffffdfff x10: 00000000ffffdfff x9 : ffff800081e0e818
> >> [37743.041259] x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
> >> [37743.041269] x5 : 0000000000001fff x4 : 0000000000000000 x3 : ffff8000819a6008
> >> [37743.041279] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00018465e900
> >> [37743.041290] Call trace:
> >> [37743.041293] drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
> >> [37743.041306] panthor_gem_free_object+0x24/0xa0 [panthor]
> >> [37743.041321] drm_gem_object_free+0x1c/0x30 [drm]
> >> [37743.041452] panthor_vm_bo_put+0xc4/0x12c [panthor]
> >> [37743.041475] panthor_vm_cleanup_op_ctx.constprop.0+0xb0/0x104 [panthor]
> >> [37743.041491] panthor_vm_bind_job_cleanup_op_ctx_work+0x28/0xd0 [panthor]
> >
> > Ok, I think I found the culprit: there's a race between
> > the drm_gpuvm_bo_put() call in panthor_vm_bo_put() and the list
> > iteration done by drm_gpuvm_prepare_objects(). Because we're not
> > setting DRM_GPUVM_RESV_PROTECTED, the code goes through the 'lockless'
> > iteration loop, and takes/release a vm_bo ref at each iteration. This
> > means our 'were we the last vm_bo user?' test in panthor_vm_bo_put()
> > might return false even if we were actually the last user, and when
> > for_each_vm_bo_in_list() releases the ref it acquired, it not only leaks
> > the pin reference, thus leaving GEM pages pinned (which explains this
> > WARN_ON() splat), but it also calls drm_gpuvm_bo_destroy() in a path
> > where we don't hold the GPUVA list lock, which is bad.
>
> IIUC, GPUVM generally behaves correctly. It's just that the return value
> of drm_gpuvm_bo_put() needs to be treated with care.
>
> Maybe we should extend c50a291d621a ("drm/gpuvm: Let drm_gpuvm_bo_put()
> report when the vm_bo object is destroyed") by a note explaining this
> unexpected case,
I can do that.
> or, if not required anymore, simply revert the patch.
Having drm_gpuvm_bo_put() reflect when it released the last ref is
definitely needed in panthor, it's just that it can't be trusted if
there's no lock protecting against concurrent gpuvm lockless list
iterations, because of the get/put dance that happens here.
>
> >
> > Long story short, I'll have to use DRM_GPUVM_RESV_PROTECTED, which is
> > fine because I'm deferring vm_bo removal to a work where taking the VM
> > resv lock is allowed. Since I was the one asking for this lockless
> > iterator in the first place, I wonder if we should kill that and make
> > DRM_GPUVM_RESV_PROTECTED the default (this would greatly simplify
> > the code). AFAICT, The PowerVR driver shouldn't be impacted because it's
> > using drm_gpuvm in synchronous mode only, and Xe already uses the
> > resv-protected mode. That leaves Nouveau, but IIRC, it's also doing VM
> > updates in the ioctl path.
> >
> > Danilo, any opinions?
>
> I agree that we should remove it once we got enough evidence that updating
> the VA space in the asynchronous path isn't going to be a thing. I'm not
> entirely sure, whether we'll attempt to re-work Nouveau, but I'd like to try
> this approach with Nova. Hence, I'd still like to leave it in for a while.
Oh, that' has nothing to do with updating the VM RB tree in the async
path. I keep doing it this way in panthor if that's what you're worried
about. The only thing I defer to a workqueue is the gpuvm_bo_put() call,
because this one needs to do all sort of locking that are not allowed
in the dma-signalling path.
>
> OOC, knowing that you went a little back and forth with one or the other
> approach, last I heard was that using DRM_GPUVM_RESV_PROTECTED didn't work
> out well with shmem buffers. How did you fix it?
The problem I had with the shmem GEM implementation was not related to
DRM_GPUVM_RESV_PROTECTED, it was related to the fact I wasn't able to
take an extra pages ref for the prev/next remap blocks in the async VM
operation path, because drm_gem_shmem_{get,pin}_pages take the BO resv
lock. The idea to solve that was to use atomic refcounts and avoid
taking the lock when refcnt > 1, which we did, but for a different
reason (not merged yet, but it's part of [1] which should land soon).
In the meantime, I realized the vm_bo concept could solve my problem.
If, instead of having each drm_gpuva own a pages ref, I make the
drm_gpuvm_bo owns this ref, I no longer need to call
drm_gem_shmem_{get,pin}_pages() in the async update path.
All I need to do is acquire the pages ref in the ioctl path, when
obtaining a vm_bo that didn't exist previously
(drm_gpuvm_bo_obtain_prealloc() allows us to figure out when this is a
new vm_bo vs when this is an pre-existing one). This is the very reason
I need to know when a vm_bo is destroyed, because when that happens, I
also need to release the pages ref owned by this vm_bo.
Regarding DRM_GPUVM_RESV_PROTECTED, I honestly don't remember where all
the churn around lockless list iteration came from, but I checked, and
DRM_GPUVM_RESV_PROTECTED only matters for:
- objects resv preparation, which we do with the VM resv lock held in
the ioctl path anyway
- VM BO destruction (drm_gpuvm_bo_put() calls leading to a
drm_gpuvm_bo_destroy()), which we now defer to a workqueue, where we
can take the VM resv lock and the GPUVA list lock
- BO validation, which we don't use yet because we don't have
mem-reclaim hooked up, but once we have it, it will happen in the
ioctl path too
So, from panthor's perspective, we should be good without the
vm-resv-lockless stuff.
[1]https://lore.kernel.org/lkml/20240105184624.508603-1-dmitry.osipenko@collabora.com/T/
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re:Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-02-05 9:03 ` Boris Brezillon
2024-02-05 9:41 ` Andy Yan
2024-02-05 9:54 ` Danilo Krummrich
@ 2024-02-06 10:35 ` Andy Yan
2024-02-06 11:11 ` Boris Brezillon
2 siblings, 1 reply; 51+ messages in thread
From: Andy Yan @ 2024-02-06 10:35 UTC (permalink / raw)
To: Boris Brezillon
Cc: Danilo Krummrich, dri-devel, Tatsuyuki Ishi, Nicolas Boichat,
kernel, Daniel Stone, Neil Armstrong, Ketil Johnsen, Liviu Dudau,
Steven Price, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
[-- Attachment #1: Type: text/plain, Size: 5464 bytes --]
HI Boris：
在 2024-02-05 17:03:21，"Boris Brezillon" <boris.brezillon@collabora.com> 写道：
>+Danilo for the panthor gpuvm-needs update.
>
>On Sun, 4 Feb 2024 09:14:44 +0800 (CST)
>"Andy Yan" <andyshrk@163.com> wrote:
>
>> Hi Boris：
>> I saw this warning sometimes（Run on a armbain based bookworm），not sure is a know issue or something else。
>> [15368.293031] systemd-journald[715]: Received client request to relinquish /var/log/journal/1bc4a340506142af9bd31a6a3d2170ba access.
>> [37743.040737] ------------[ cut here ]------------
>> [37743.040764] panthor fb000000.gpu: drm_WARN_ON(shmem->pages_use_count)
>> [37743.040890] WARNING: CPU: 2 PID: 5702 at drivers/gpu/drm/drm_gem_shmem_helper.c:158 drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.040929] Modules linked in: joydev rfkill sunrpc lz4hc lz4 zram binfmt_misc hantro_vpu crct10dif_ce v4l2_vp9 v4l2_h264 snd_soc_simple_amplifier v4l2_mem2mem videobuf2_dma_contig snd_soc_es8328_i2c videobuf2_memops rk_crypto2 snd_soc_es8328 videobuf2_v4l2 sm3_generic videodev crypto_engine sm3 rockchip_rng videobuf2_common nvmem_rockchip_otp snd_soc_rockchip_i2s_tdm snd_soc_hdmi_codec snd_soc_simple_card mc snd_soc_simple_card_utils snd_soc_core snd_compress ac97_bus snd_pcm_dmaengine snd_pcm snd_timer snd soundcore dm_mod ip_tables x_tables autofs4 dw_hdmi_qp_i2s_audio dw_hdmi_qp_cec rk808_regulator rockchipdrm dw_mipi_dsi dw_hdmi_qp dw_hdmi analogix_dp drm_dma_helper fusb302 display_connector rk8xx_spi drm_display_helper phy_rockchip_snps_pcie3 phy_rockchip_samsung_hdptx_hdmi panthor tcpm rk8xx_core cec drm_gpuvm gpu_sched drm_kms_helper drm_shmem_helper drm_exec r8169 drm pwm_bl adc_keys
>> [37743.041108] CPU: 2 PID: 5702 Comm: kworker/u16:8 Not tainted 6.8.0-rc1-edge-rockchip-rk3588 #2
>> [37743.041115] Hardware name: Rockchip RK3588 EVB1 V10 Board (DT)
>> [37743.041120] Workqueue: panthor-cleanup panthor_vm_bind_job_cleanup_op_ctx_work [panthor]
>> [37743.041151] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>> [37743.041157] pc : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041169] lr : drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041181] sp : ffff80008d37bcc0
>> [37743.041184] x29: ffff80008d37bcc0 x28: ffff800081d379c0 x27: ffff800081d37000
>> [37743.041196] x26: ffff00019909a280 x25: ffff00019909a2c0 x24: ffff0001017a4c05
>> [37743.041206] x23: dead000000000100 x22: dead000000000122 x21: ffff0001627ac1a0
>> [37743.041217] x20: 0000000000000000 x19: ffff0001627ac000 x18: 0000000000000000
>> [37743.041227] x17: 000000040044ffff x16: 005000f2b5503510 x15: fffffffffff91b77
>> [37743.041238] x14: 0000000000000001 x13: 00000000000003c5 x12: 00000000ffffffea
>> [37743.041248] x11: 00000000ffffdfff x10: 00000000ffffdfff x9 : ffff800081e0e818
>> [37743.041259] x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
>> [37743.041269] x5 : 0000000000001fff x4 : 0000000000000000 x3 : ffff8000819a6008
>> [37743.041279] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00018465e900
>> [37743.041290] Call trace:
>> [37743.041293] drm_gem_shmem_free+0x144/0x14c [drm_shmem_helper]
>> [37743.041306] panthor_gem_free_object+0x24/0xa0 [panthor]
>> [37743.041321] drm_gem_object_free+0x1c/0x30 [drm]
>> [37743.041452] panthor_vm_bo_put+0xc4/0x12c [panthor]
>> [37743.041475] panthor_vm_cleanup_op_ctx.constprop.0+0xb0/0x104 [panthor]
>> [37743.041491] panthor_vm_bind_job_cleanup_op_ctx_work+0x28/0xd0 [panthor]
>
>Ok, I think I found the culprit: there's a race between
>the drm_gpuvm_bo_put() call in panthor_vm_bo_put() and the list
>iteration done by drm_gpuvm_prepare_objects(). Because we're not
>setting DRM_GPUVM_RESV_PROTECTED, the code goes through the 'lockless'
>iteration loop, and takes/release a vm_bo ref at each iteration. This
>means our 'were we the last vm_bo user?' test in panthor_vm_bo_put()
>might return false even if we were actually the last user, and when
>for_each_vm_bo_in_list() releases the ref it acquired, it not only leaks
>the pin reference, thus leaving GEM pages pinned (which explains this
>WARN_ON() splat), but it also calls drm_gpuvm_bo_destroy() in a path
>where we don't hold the GPUVA list lock, which is bad.
>
>Long story short, I'll have to use DRM_GPUVM_RESV_PROTECTED, which is
>fine because I'm deferring vm_bo removal to a work where taking the VM
>resv lock is allowed. Since I was the one asking for this lockless
>iterator in the first place, I wonder if we should kill that and make
>DRM_GPUVM_RESV_PROTECTED the default (this would greatly simplify
>the code). AFAICT, The PowerVR driver shouldn't be impacted because it's
>using drm_gpuvm in synchronous mode only, and Xe already uses the
>resv-protected mode. That leaves Nouveau, but IIRC, it's also doing VM
>updates in the ioctl path.
>
>Danilo, any opinions?
>
>Andy, I pushed a new version to the panthor-next [1] and
>panthor-next+rk3588 [2] branches. The fix I'm talking about is [3], but
>you probably want to consider taking all the fixups in your branch.
>
Just for your info，
After taking all the fixups in my branch，It has been running for 24 hours, and so far, this warning hasn't shown up.
>Regards,
>
>Boris
>
>[1]https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next
>[2]https://gitlab.freedesktop.org/panfrost/linux/-/commits/panthor-next+rk3588
>[3]https://gitlab.freedesktop.org/panfrost/linux/-/commit/df48c09662a403275e76e679ee004085badea7c1
[-- Attachment #2: Type: text/html, Size: 5946 bytes --]
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-02-06 10:35 ` Andy Yan
@ 2024-02-06 11:11 ` Boris Brezillon
0 siblings, 0 replies; 51+ messages in thread
From: Boris Brezillon @ 2024-02-06 11:11 UTC (permalink / raw)
To: Andy Yan
Cc: Danilo Krummrich, dri-devel, Tatsuyuki Ishi, Nicolas Boichat,
kernel, Daniel Stone, Neil Armstrong, Ketil Johnsen, Liviu Dudau,
Steven Price, Clément Péron, Daniel Vetter,
Chris Diamand, Marty E . Plummer, Robin Murphy, Faith Ekstrand
Hi Andy,
On Tue, 6 Feb 2024 18:35:36 +0800 (CST)
"Andy Yan" <andyshrk@163.com> wrote:
> >Andy, I pushed a new version to the panthor-next [1] and
> >panthor-next+rk3588 [2] branches. The fix I'm talking about is [3],
> >but you probably want to consider taking all the fixups in your
> >branch.
>
> >
> Just for your info，
> After taking all the fixups in my branch，It has been running for 24
> hours, and so far, this warning hasn't shown up.
Thanks for the update, and thanks for reporting the problem in the
first place.
Regards,
Boris
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs
2024-01-22 16:30 [PATCH v4 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
` (16 preceding siblings ...)
2024-01-29 9:20 ` Andy Yan
@ 2024-02-08 16:02 ` Liviu Dudau
17 siblings, 0 replies; 51+ messages in thread
From: Liviu Dudau @ 2024-02-08 16:02 UTC (permalink / raw)
To: Boris Brezillon
Cc: dri-devel, Daniel Vetter, Marty E . Plummer, Rob Herring,
Clément Péron, Nicolas Boichat, Neil Armstrong,
Faith Ekstrand, Daniel Stone, Steven Price, Robin Murphy, kernel,
Heiko Stuebner, Tatsuyuki Ishi, Chris Diamand, Ketil Johnsen
On Mon, Jan 22, 2024 at 05:30:31PM +0100, Boris Brezillon wrote:
> Hello,
Hi,
>
> This is the 4th version of the kernel driver for Mali CSF-based GPUs.
>
> A branch based on drm-misc-next and containing all the dependencies
> that are not yet available in drm-misc-next here[1], and another [2]
> containing extra patches to have things working on rk3588. The CSF
> firmware binary can be found here[3], and should be placed under
> /lib/firmware/arm/mali/arch10.8/mali_csffw.bin.
Since today the linux-firmware 'main' branch also contains an official
copy of the CSF firmware for the 10.8 architecture. Thanks to Mario for
the quick inclusion!
Best regards,
Liviu
>
> The mesa MR adding v10 support on top of panthor is available here [4].
>
> Steve, I intentionally dropped your R-b on "drm/panthor: Add the heap
> logical block" and "drm/panthor: Add the scheduler logical block"
> because the tiler-OOM handling changed enough to require a new review
> IMHO.
>
> Regarding the GPL2+MIT relicensing, I collected Clément's R-b for the
> devfreq code, but am still lacking Alexey Sheplyakov for some bits in
> panthor_gpu.c. The rest of the code is either new, or covered by the
> Linaro, Arm and Collabora acks.
>
> And here is a non-exhaustive changelog, check each commit for a detailed
> changelog.
>
> v4:
> - Fix various bugs in the VM logic
> - Address comments from Steven, Liviu, Ketil and Chris
> - Move tiler OOM handling out of the scheduler interrupt handling path
> so we can properly recover when the system runs out of memory, and
> panthor is blocked trying to allocate heap chunks
> - Rework the heap locking to support concurrent chunk allocation. Not
> sure if this is supposed to happen, but we need to be robust against
> userspace passing the same heap context to two scheduling groups.
> Wasn't needed before the tiler_oom rework, because heap allocation
> base serialized by the scheduler lock.
> - Make kernel BO destruction robust to NULL/ERR pointers
>
> v3;
> - Quite a few changes at the MMU/sched level to make the fix some
> race conditions and deadlocks
> - Addition of the a sync-only VM_BIND operation (to support
> vkQueueSparseBind with zero commands).
> - Addition of a VM_GET_STATE ioctl
> - Various cosmetic changes (see the commit changelogs for more details)
> - Various fixes (see the commit changelogs for more details)
>
> v2:
> - Rename the driver (pancsf -> panthor)
> - Split the commit adding the driver to ease review
> - Use drm_sched for dependency tracking/job submission
> - Add a VM_BIND ioctl
> - Add the concept of exclusive VM for BOs that are only ever mapped to a
> single VM
> - Document the code and uAPI
> - Add a DT binding doc
>
> Regards,
>
> Boris
>
> [1]https://gitlab.freedesktop.org/panfrost/linux/-/tree/panthor-v4
> [2]https://gitlab.freedesktop.org/panfrost/linux/-/tree/panthor-v4+rk3588
> [3]https://gitlab.com/firefly-linux/external/libmali/-/raw/firefly/firmware/g610/mali_csffw.bin
> [4]https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26358
>
> Boris Brezillon (13):
> drm/panthor: Add uAPI
> drm/panthor: Add GPU register definitions
> drm/panthor: Add the device logical block
> drm/panthor: Add the GPU logical block
> drm/panthor: Add GEM logical block
> drm/panthor: Add the devfreq logical block
> drm/panthor: Add the MMU/VM logical block
> drm/panthor: Add the FW logical block
> drm/panthor: Add the heap logical block
> drm/panthor: Add the scheduler logical block
> drm/panthor: Add the driver frontend block
> drm/panthor: Allow driver compilation
> drm/panthor: Add an entry to MAINTAINERS
>
> Liviu Dudau (1):
> dt-bindings: gpu: mali-valhall-csf: Add support for Arm Mali CSF GPUs
>
> .../bindings/gpu/arm,mali-valhall-csf.yaml | 147 +
> Documentation/gpu/driver-uapi.rst | 5 +
> MAINTAINERS | 11 +
> drivers/gpu/drm/Kconfig | 2 +
> drivers/gpu/drm/Makefile | 1 +
> drivers/gpu/drm/panthor/Kconfig | 23 +
> drivers/gpu/drm/panthor/Makefile | 15 +
> drivers/gpu/drm/panthor/panthor_devfreq.c | 283 ++
> drivers/gpu/drm/panthor/panthor_devfreq.h | 25 +
> drivers/gpu/drm/panthor/panthor_device.c | 544 +++
> drivers/gpu/drm/panthor/panthor_device.h | 393 ++
> drivers/gpu/drm/panthor/panthor_drv.c | 1470 +++++++
> drivers/gpu/drm/panthor/panthor_fw.c | 1334 +++++++
> drivers/gpu/drm/panthor/panthor_fw.h | 504 +++
> drivers/gpu/drm/panthor/panthor_gem.c | 228 ++
> drivers/gpu/drm/panthor/panthor_gem.h | 144 +
> drivers/gpu/drm/panthor/panthor_gpu.c | 482 +++
> drivers/gpu/drm/panthor/panthor_gpu.h | 52 +
> drivers/gpu/drm/panthor/panthor_heap.c | 596 +++
> drivers/gpu/drm/panthor/panthor_heap.h | 39 +
> drivers/gpu/drm/panthor/panthor_mmu.c | 2760 +++++++++++++
> drivers/gpu/drm/panthor/panthor_mmu.h | 102 +
> drivers/gpu/drm/panthor/panthor_regs.h | 239 ++
> drivers/gpu/drm/panthor/panthor_sched.c | 3500 +++++++++++++++++
> drivers/gpu/drm/panthor/panthor_sched.h | 48 +
> include/uapi/drm/panthor_drm.h | 945 +++++
> 26 files changed, 13892 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/gpu/arm,mali-valhall-csf.yaml
> create mode 100644 drivers/gpu/drm/panthor/Kconfig
> create mode 100644 drivers/gpu/drm/panthor/Makefile
> create mode 100644 drivers/gpu/drm/panthor/panthor_devfreq.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_devfreq.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_device.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_device.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_drv.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_fw.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_fw.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gem.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_gpu.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_heap.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_heap.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_mmu.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_regs.h
> create mode 100644 drivers/gpu/drm/panthor/panthor_sched.c
> create mode 100644 drivers/gpu/drm/panthor/panthor_sched.h
> create mode 100644 include/uapi/drm/panthor_drm.h
>
> --
> 2.43.0
>
--
====================
| I would like to |
| fix the world, |
| but they're not |
| giving me the |
\ source code! /
---------------
¯\_(ツ)_/¯
^ permalink raw reply [flat|nested] 51+ messages in thread