[PATCH 0/2] drm/panthor: Expose scheduler groups info in debugfs

[PATCH 0/2] drm/panthor: Expose scheduler groups info in debugfs

[Mary Guillemard]

This patch series adds a new debugfs file named "sched_groups" allowing
a user to query information about all userspace clients scheduler groups.

This patch series assumes that [1] and [2] are applied.

[1]https://lore.kernel.org/all/20240903144955.144278-2-mary.guillemard@collabora.com/
[2]https://lore.kernel.org/all/20240909064820.34982-2-mary.guillemard@collabora.com/

Mary Guillemard (2):
  drm/panthor: Add csg_priority to panthor_group
  drm/panthor: Expose scheduler groups info in debugfs

 drivers/gpu/drm/panthor/panthor_drv.c   |   1 +
 drivers/gpu/drm/panthor/panthor_sched.c | 145 +++++++++++++++++++++++-
 drivers/gpu/drm/panthor/panthor_sched.h |   4 +
 3 files changed, 147 insertions(+), 3 deletions(-)

-- 
2.46.0

[PATCH 1/2] drm/panthor: Add csg_priority to panthor_group

[Mary Guillemard]

Keep track of the csg priority in panthor_group when the group is
scheduled/active.

This is useful to know the actual priority in use in the firmware
group slot.

Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
---
 drivers/gpu/drm/panthor/panthor_sched.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/panthor/panthor_sched.c b/drivers/gpu/drm/panthor/panthor_sched.c
index 86908ada7335..f15abeef4ece 100644
--- a/drivers/gpu/drm/panthor/panthor_sched.c
+++ b/drivers/gpu/drm/panthor/panthor_sched.c
@@ -574,6 +574,13 @@ struct panthor_group {
 	 */
 	int csg_id;
 
+	/**
+	 * @csg_id: Priority of the FW group slot.
+	 *
+	 * -1 when the group is not scheduled/active.
+	 */
+	int csg_priority;
+
 	/**
 	 * @destroyed: True when the group has been destroyed.
 	 *
@@ -894,11 +901,12 @@ group_get(struct panthor_group *group)
  * group_bind_locked() - Bind a group to a group slot
  * @group: Group.
  * @csg_id: Slot.
+ * @csg_priority: Priority of the slot.
  *
  * Return: 0 on success, a negative error code otherwise.
  */
 static int
-group_bind_locked(struct panthor_group *group, u32 csg_id)
+group_bind_locked(struct panthor_group *group, u32 csg_id, u32 csg_priority)
 {
 	struct panthor_device *ptdev = group->ptdev;
 	struct panthor_csg_slot *csg_slot;
@@ -917,6 +925,7 @@ group_bind_locked(struct panthor_group *group, u32 csg_id)
 	csg_slot = &ptdev->scheduler->csg_slots[csg_id];
 	group_get(group);
 	group->csg_id = csg_id;
+	group->csg_priority = csg_priority;
 
 	/* Dummy doorbell allocation: doorbell is assigned to the group and
 	 * all queues use the same doorbell.
@@ -956,6 +965,7 @@ group_unbind_locked(struct panthor_group *group)
 	slot = &ptdev->scheduler->csg_slots[group->csg_id];
 	panthor_vm_idle(group->vm);
 	group->csg_id = -1;
+	group->csg_priority = -1;
 
 	/* Tiler OOM events will be re-issued next time the group is scheduled. */
 	atomic_set(&group->tiler_oom, 0);
@@ -2193,8 +2203,9 @@ tick_ctx_apply(struct panthor_scheduler *sched, struct panthor_sched_tick_ctx *c
 
 			csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
 			csg_slot = &sched->csg_slots[csg_id];
-			group_bind_locked(group, csg_id);
-			csg_slot_prog_locked(ptdev, csg_id, new_csg_prio--);
+			group_bind_locked(group, csg_id, new_csg_prio);
+			csg_slot_prog_locked(ptdev, csg_id, new_csg_prio);
+			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,
@@ -3111,6 +3122,7 @@ int panthor_group_create(struct panthor_file *pfile,
 	kref_init(&group->refcount);
 	group->state = PANTHOR_CS_GROUP_CREATED;
 	group->csg_id = -1;
+	group->csg_priority = -1;
 
 	group->ptdev = ptdev;
 	group->max_compute_cores = group_args->max_compute_cores;
-- 
2.46.0

[PATCH 2/2] drm/panthor: Expose scheduler groups info in debugfs

[Mary Guillemard]

This adds a new debugfs file named "sched_groups" allowing a user to
query information about all userspace clients scheduler groups.

As we uses drm_device.filelist to achieve it, we also expose the
client_id with the task information to differentiate one client
from another inside the same task.

Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
---
 drivers/gpu/drm/panthor/panthor_drv.c   |   1 +
 drivers/gpu/drm/panthor/panthor_sched.c | 127 ++++++++++++++++++++++++
 drivers/gpu/drm/panthor/panthor_sched.h |   4 +
 3 files changed, 132 insertions(+)

diff --git a/drivers/gpu/drm/panthor/panthor_drv.c b/drivers/gpu/drm/panthor/panthor_drv.c
index 0d825d63d712..aa390597d355 100644
--- a/drivers/gpu/drm/panthor/panthor_drv.c
+++ b/drivers/gpu/drm/panthor/panthor_drv.c
@@ -1450,6 +1450,7 @@ static const struct file_operations panthor_drm_driver_fops = {
 static void panthor_debugfs_init(struct drm_minor *minor)
 {
 	panthor_mmu_debugfs_init(minor);
+	panthor_sched_debugfs_init(minor);
 }
 #endif
 
diff --git a/drivers/gpu/drm/panthor/panthor_sched.c b/drivers/gpu/drm/panthor/panthor_sched.c
index f15abeef4ece..0c1ddbfbe164 100644
--- a/drivers/gpu/drm/panthor/panthor_sched.c
+++ b/drivers/gpu/drm/panthor/panthor_sched.c
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0 or MIT
 /* Copyright 2023 Collabora ltd. */
 
+#include <drm/drm_debugfs.h>
 #include <drm/drm_drv.h>
 #include <drm/drm_exec.h>
 #include <drm/drm_gem_shmem_helper.h>
@@ -3581,3 +3582,129 @@ int panthor_sched_init(struct panthor_device *ptdev)
 	ptdev->scheduler = sched;
 	return 0;
 }
+
+#ifdef CONFIG_DEBUG_FS
+static const char *panthor_csg_priority_names[PANTHOR_CSG_PRIORITY_COUNT] = {
+	"LOW",
+	"MEDIUM",
+	"HIGH",
+	"REALTIME"
+};
+
+static const char *panthor_group_state_names[PANTHOR_CS_GROUP_UNKNOWN_STATE] = {
+	"CREATED",
+	"ACTIVE",
+	"SUSPENDED",
+	"TERMINATED"
+};
+
+static void show_panthor_queue(const struct panthor_queue *queue,
+			       u32 queue_index,
+			       struct seq_file *m)
+{
+	seq_printf(m, "queue %u:", queue_index);
+	seq_printf(m, " priority %u", queue->priority);
+	seq_printf(m, " doorbell_id %d", queue->doorbell_id);
+	seq_puts(m, "\n");
+}
+
+static void show_panthor_group(struct panthor_group *group,
+			       u32 group_handle,
+			       struct seq_file *m)
+{
+	u32 i;
+
+	group_get(group);
+
+	seq_printf(m, "group %u:", group_handle);
+	seq_printf(m, " priority %s", group->priority < PANTHOR_CSG_PRIORITY_COUNT ?
+		   panthor_csg_priority_names[group->priority] : "UNKNOWN");
+	seq_printf(m, " state %s", group->state < PANTHOR_CS_GROUP_UNKNOWN_STATE ?
+		   panthor_group_state_names[group->state] : "UNKNOWN");
+	seq_printf(m, " csg_id %d", group->csg_id);
+	seq_printf(m, " csg_priority %d", group->csg_priority);
+	seq_printf(m, " compute_core_mask 0x%016llx", group->compute_core_mask);
+	seq_printf(m, " fragment_core_mask 0x%016llx", group->fragment_core_mask);
+	seq_printf(m, " tiler_core_mask 0x%016llx", group->tiler_core_mask);
+	seq_printf(m, " max_compute_cores %d", group->max_compute_cores);
+	seq_printf(m, " max_fragment_cores %d", group->max_fragment_cores);
+	seq_printf(m, " max_tiler_cores %d", group->max_tiler_cores);
+	seq_puts(m, "\n");
+
+	for (i = 0; i < group->queue_count; i++)
+		show_panthor_queue(group->queues[i], i, m);
+
+	group_put(group);
+}
+
+static int show_file_group_pool(const struct panthor_file *pfile, struct seq_file *m)
+{
+	struct panthor_group_pool *gpool = pfile->groups;
+	struct panthor_group *group;
+	unsigned long i;
+
+	if (IS_ERR_OR_NULL(gpool))
+		return 0;
+
+	xa_for_each(&gpool->xa, i, group)
+		show_panthor_group(group, i, m);
+
+	return 0;
+}
+
+static int show_each_file(struct seq_file *m, void *arg)
+{
+	struct drm_info_node *node = (struct drm_info_node *)m->private;
+	struct drm_device *ddev = node->minor->dev;
+	int (*show)(const struct panthor_file *, struct seq_file *) = node->info_ent->data;
+	struct drm_file *file;
+	int ret;
+
+	ret = mutex_lock_interruptible(&ddev->filelist_mutex);
+	if (ret)
+		return ret;
+
+	list_for_each_entry(file, &ddev->filelist, lhead) {
+		struct task_struct *task;
+		struct panthor_file *pfile = file->driver_priv;
+		struct pid *pid;
+
+		/*
+		 * Although we have a valid reference on file->pid, that does
+		 * not guarantee that the task_struct who called get_pid() is
+		 * still alive (e.g. get_pid(current) => fork() => exit()).
+		 * Therefore, we need to protect this ->comm access using RCU.
+		 */
+		rcu_read_lock();
+		pid = rcu_dereference(file->pid);
+		task = pid_task(pid, PIDTYPE_TGID);
+		seq_printf(m, "client_id %8llu pid %8d command %s:\n", file->client_id,
+			   pid_nr(pid), task ? task->comm : "<unknown>");
+		rcu_read_unlock();
+
+		ret = show(pfile, m);
+		if (ret < 0)
+			break;
+
+		seq_puts(m, "\n");
+	}
+
+	mutex_unlock(&ddev->filelist_mutex);
+	return ret;
+}
+
+static struct drm_info_list panthor_sched_debugfs_list[] = {
+	{ "sched_groups", show_each_file, 0, show_file_group_pool },
+};
+
+/**
+ * panthor_sched_debugfs_init() - Initialize scheduler debugfs entries
+ * @minor: Minor.
+ */
+void panthor_sched_debugfs_init(struct drm_minor *minor)
+{
+	drm_debugfs_create_files(panthor_sched_debugfs_list,
+				 ARRAY_SIZE(panthor_sched_debugfs_list),
+				 minor->debugfs_root, minor);
+}
+#endif /* CONFIG_DEBUG_FS */
diff --git a/drivers/gpu/drm/panthor/panthor_sched.h b/drivers/gpu/drm/panthor/panthor_sched.h
index 3a30d2328b30..577239aa66bf 100644
--- a/drivers/gpu/drm/panthor/panthor_sched.h
+++ b/drivers/gpu/drm/panthor/panthor_sched.h
@@ -47,4 +47,8 @@ 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);
 
+#ifdef CONFIG_DEBUG_FS
+void panthor_sched_debugfs_init(struct drm_minor *minor);
+#endif
+
 #endif
-- 
2.46.0

Re: [PATCH 1/2] drm/panthor: Add csg_priority to panthor_group

[kernel test robot]

Hi Mary,

kernel test robot noticed the following build warnings:

[auto build test WARNING on drm/drm-next]
[also build test WARNING on drm-exynos/exynos-drm-next drm-intel/for-linux-next drm-intel/for-linux-next-fixes drm-misc/drm-misc-next drm-tip/drm-tip linus/master v6.11 next-20240920]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Mary-Guillemard/drm-panthor-Add-csg_priority-to-panthor_group/20240918-165602
base:   git://anongit.freedesktop.org/drm/drm drm-next
patch link:    https://lore.kernel.org/r/20240918085056.24422-3-mary.guillemard%40collabora.com
patch subject: [PATCH 1/2] drm/panthor: Add csg_priority to panthor_group
config: x86_64-allyesconfig (https://download.01.org/0day-ci/archive/20240922/202409220025.v4cCQjaI-lkp@intel.com/config)
compiler: clang version 18.1.8 (https://github.com/llvm/llvm-project 3b5b5c1ec4a3095ab096dd780e84d7ab81f3d7ff)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240922/202409220025.v4cCQjaI-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202409220025.v4cCQjaI-lkp@intel.com/

All warnings (new ones prefixed by >>):

   drivers/gpu/drm/panthor/panthor_sched.c:319: warning: Excess struct member 'runnable' description in 'panthor_scheduler'
   drivers/gpu/drm/panthor/panthor_sched.c:319: warning: Excess struct member 'idle' description in 'panthor_scheduler'
   drivers/gpu/drm/panthor/panthor_sched.c:319: warning: Excess struct member 'waiting' description in 'panthor_scheduler'
   drivers/gpu/drm/panthor/panthor_sched.c:319: warning: Excess struct member 'has_ref' description in 'panthor_scheduler'
   drivers/gpu/drm/panthor/panthor_sched.c:319: warning: Excess struct member 'in_progress' description in 'panthor_scheduler'
   drivers/gpu/drm/panthor/panthor_sched.c:319: warning: Excess struct member 'stopped_groups' description in 'panthor_scheduler'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'mem' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'input' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'output' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'input_fw_va' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'output_fw_va' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'gpu_va' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'ref' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'gt' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'sync64' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'bo' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'offset' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'kmap' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'lock' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'id' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'seqno' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'last_fence' description in 'panthor_queue'
   drivers/gpu/drm/panthor/panthor_sched.c:479: warning: Excess struct member 'in_flight_jobs' description in 'panthor_queue'
>> drivers/gpu/drm/panthor/panthor_sched.c:669: warning: Function parameter or struct member 'csg_priority' not described in 'panthor_group'
   drivers/gpu/drm/panthor/panthor_sched.c:784: warning: Excess struct member 'start' description in 'panthor_job'
   drivers/gpu/drm/panthor/panthor_sched.c:784: warning: Excess struct member 'size' description in 'panthor_job'
   drivers/gpu/drm/panthor/panthor_sched.c:784: warning: Excess struct member 'latest_flush' description in 'panthor_job'
   drivers/gpu/drm/panthor/panthor_sched.c:784: warning: Excess struct member 'start' description in 'panthor_job'
   drivers/gpu/drm/panthor/panthor_sched.c:784: warning: Excess struct member 'end' description in 'panthor_job'
   drivers/gpu/drm/panthor/panthor_sched.c:1707: warning: Function parameter or struct member 'ptdev' not described in 'panthor_sched_report_fw_events'
   drivers/gpu/drm/panthor/panthor_sched.c:1707: warning: Function parameter or struct member 'events' not described in 'panthor_sched_report_fw_events'
   drivers/gpu/drm/panthor/panthor_sched.c:2602: warning: Function parameter or struct member 'ptdev' not described in 'panthor_sched_report_mmu_fault'


vim +669 drivers/gpu/drm/panthor/panthor_sched.c

de85488138247d Boris Brezillon 2024-02-29  516  
de85488138247d Boris Brezillon 2024-02-29  517  /**
de85488138247d Boris Brezillon 2024-02-29  518   * struct panthor_group - Scheduling group object
de85488138247d Boris Brezillon 2024-02-29  519   */
de85488138247d Boris Brezillon 2024-02-29  520  struct panthor_group {
de85488138247d Boris Brezillon 2024-02-29  521  	/** @refcount: Reference count */
de85488138247d Boris Brezillon 2024-02-29  522  	struct kref refcount;
de85488138247d Boris Brezillon 2024-02-29  523  
de85488138247d Boris Brezillon 2024-02-29  524  	/** @ptdev: Device. */
de85488138247d Boris Brezillon 2024-02-29  525  	struct panthor_device *ptdev;
de85488138247d Boris Brezillon 2024-02-29  526  
de85488138247d Boris Brezillon 2024-02-29  527  	/** @vm: VM bound to the group. */
de85488138247d Boris Brezillon 2024-02-29  528  	struct panthor_vm *vm;
de85488138247d Boris Brezillon 2024-02-29  529  
de85488138247d Boris Brezillon 2024-02-29  530  	/** @compute_core_mask: Mask of shader cores that can be used for compute jobs. */
de85488138247d Boris Brezillon 2024-02-29  531  	u64 compute_core_mask;
de85488138247d Boris Brezillon 2024-02-29  532  
de85488138247d Boris Brezillon 2024-02-29  533  	/** @fragment_core_mask: Mask of shader cores that can be used for fragment jobs. */
de85488138247d Boris Brezillon 2024-02-29  534  	u64 fragment_core_mask;
de85488138247d Boris Brezillon 2024-02-29  535  
de85488138247d Boris Brezillon 2024-02-29  536  	/** @tiler_core_mask: Mask of tiler cores that can be used for tiler jobs. */
de85488138247d Boris Brezillon 2024-02-29  537  	u64 tiler_core_mask;
de85488138247d Boris Brezillon 2024-02-29  538  
de85488138247d Boris Brezillon 2024-02-29  539  	/** @max_compute_cores: Maximum number of shader cores used for compute jobs. */
de85488138247d Boris Brezillon 2024-02-29  540  	u8 max_compute_cores;
de85488138247d Boris Brezillon 2024-02-29  541  
be7ffc821f5fc2 Liviu Dudau     2024-04-02  542  	/** @max_fragment_cores: Maximum number of shader cores used for fragment jobs. */
de85488138247d Boris Brezillon 2024-02-29  543  	u8 max_fragment_cores;
de85488138247d Boris Brezillon 2024-02-29  544  
de85488138247d Boris Brezillon 2024-02-29  545  	/** @max_tiler_cores: Maximum number of tiler cores used for tiler jobs. */
de85488138247d Boris Brezillon 2024-02-29  546  	u8 max_tiler_cores;
de85488138247d Boris Brezillon 2024-02-29  547  
de85488138247d Boris Brezillon 2024-02-29  548  	/** @priority: Group priority (check panthor_csg_priority). */
de85488138247d Boris Brezillon 2024-02-29  549  	u8 priority;
de85488138247d Boris Brezillon 2024-02-29  550  
de85488138247d Boris Brezillon 2024-02-29  551  	/** @blocked_queues: Bitmask reflecting the blocked queues. */
de85488138247d Boris Brezillon 2024-02-29  552  	u32 blocked_queues;
de85488138247d Boris Brezillon 2024-02-29  553  
de85488138247d Boris Brezillon 2024-02-29  554  	/** @idle_queues: Bitmask reflecting the idle queues. */
de85488138247d Boris Brezillon 2024-02-29  555  	u32 idle_queues;
de85488138247d Boris Brezillon 2024-02-29  556  
de85488138247d Boris Brezillon 2024-02-29  557  	/** @fatal_lock: Lock used to protect access to fatal fields. */
de85488138247d Boris Brezillon 2024-02-29  558  	spinlock_t fatal_lock;
de85488138247d Boris Brezillon 2024-02-29  559  
de85488138247d Boris Brezillon 2024-02-29  560  	/** @fatal_queues: Bitmask reflecting the queues that hit a fatal exception. */
de85488138247d Boris Brezillon 2024-02-29  561  	u32 fatal_queues;
de85488138247d Boris Brezillon 2024-02-29  562  
de85488138247d Boris Brezillon 2024-02-29  563  	/** @tiler_oom: Mask of queues that have a tiler OOM event to process. */
de85488138247d Boris Brezillon 2024-02-29  564  	atomic_t tiler_oom;
de85488138247d Boris Brezillon 2024-02-29  565  
de85488138247d Boris Brezillon 2024-02-29  566  	/** @queue_count: Number of queues in this group. */
de85488138247d Boris Brezillon 2024-02-29  567  	u32 queue_count;
de85488138247d Boris Brezillon 2024-02-29  568  
de85488138247d Boris Brezillon 2024-02-29  569  	/** @queues: Queues owned by this group. */
de85488138247d Boris Brezillon 2024-02-29  570  	struct panthor_queue *queues[MAX_CS_PER_CSG];
de85488138247d Boris Brezillon 2024-02-29  571  
de85488138247d Boris Brezillon 2024-02-29  572  	/**
de85488138247d Boris Brezillon 2024-02-29  573  	 * @csg_id: ID of the FW group slot.
de85488138247d Boris Brezillon 2024-02-29  574  	 *
de85488138247d Boris Brezillon 2024-02-29  575  	 * -1 when the group is not scheduled/active.
de85488138247d Boris Brezillon 2024-02-29  576  	 */
de85488138247d Boris Brezillon 2024-02-29  577  	int csg_id;
de85488138247d Boris Brezillon 2024-02-29  578  
474e5b301372a6 Mary Guillemard 2024-09-18  579  	/**
474e5b301372a6 Mary Guillemard 2024-09-18  580  	 * @csg_id: Priority of the FW group slot.
474e5b301372a6 Mary Guillemard 2024-09-18  581  	 *
474e5b301372a6 Mary Guillemard 2024-09-18  582  	 * -1 when the group is not scheduled/active.
474e5b301372a6 Mary Guillemard 2024-09-18  583  	 */
474e5b301372a6 Mary Guillemard 2024-09-18  584  	int csg_priority;
474e5b301372a6 Mary Guillemard 2024-09-18  585  
de85488138247d Boris Brezillon 2024-02-29  586  	/**
de85488138247d Boris Brezillon 2024-02-29  587  	 * @destroyed: True when the group has been destroyed.
de85488138247d Boris Brezillon 2024-02-29  588  	 *
de85488138247d Boris Brezillon 2024-02-29  589  	 * If a group is destroyed it becomes useless: no further jobs can be submitted
de85488138247d Boris Brezillon 2024-02-29  590  	 * to its queues. We simply wait for all references to be dropped so we can
de85488138247d Boris Brezillon 2024-02-29  591  	 * release the group object.
de85488138247d Boris Brezillon 2024-02-29  592  	 */
de85488138247d Boris Brezillon 2024-02-29  593  	bool destroyed;
de85488138247d Boris Brezillon 2024-02-29  594  
de85488138247d Boris Brezillon 2024-02-29  595  	/**
de85488138247d Boris Brezillon 2024-02-29  596  	 * @timedout: True when a timeout occurred on any of the queues owned by
de85488138247d Boris Brezillon 2024-02-29  597  	 * this group.
de85488138247d Boris Brezillon 2024-02-29  598  	 *
de85488138247d Boris Brezillon 2024-02-29  599  	 * Timeouts can be reported by drm_sched or by the FW. In any case, any
de85488138247d Boris Brezillon 2024-02-29  600  	 * timeout situation is unrecoverable, and the group becomes useless.
de85488138247d Boris Brezillon 2024-02-29  601  	 * We simply wait for all references to be dropped so we can release the
de85488138247d Boris Brezillon 2024-02-29  602  	 * group object.
de85488138247d Boris Brezillon 2024-02-29  603  	 */
de85488138247d Boris Brezillon 2024-02-29  604  	bool timedout;
de85488138247d Boris Brezillon 2024-02-29  605  
de85488138247d Boris Brezillon 2024-02-29  606  	/**
de85488138247d Boris Brezillon 2024-02-29  607  	 * @syncobjs: Pool of per-queue synchronization objects.
de85488138247d Boris Brezillon 2024-02-29  608  	 *
de85488138247d Boris Brezillon 2024-02-29  609  	 * One sync object per queue. The position of the sync object is
de85488138247d Boris Brezillon 2024-02-29  610  	 * determined by the queue index.
de85488138247d Boris Brezillon 2024-02-29  611  	 */
de85488138247d Boris Brezillon 2024-02-29  612  	struct panthor_kernel_bo *syncobjs;
de85488138247d Boris Brezillon 2024-02-29  613  
de85488138247d Boris Brezillon 2024-02-29  614  	/** @state: Group state. */
de85488138247d Boris Brezillon 2024-02-29  615  	enum panthor_group_state state;
de85488138247d Boris Brezillon 2024-02-29  616  
de85488138247d Boris Brezillon 2024-02-29  617  	/**
de85488138247d Boris Brezillon 2024-02-29  618  	 * @suspend_buf: Suspend buffer.
de85488138247d Boris Brezillon 2024-02-29  619  	 *
de85488138247d Boris Brezillon 2024-02-29  620  	 * Stores the state of the group and its queues when a group is suspended.
de85488138247d Boris Brezillon 2024-02-29  621  	 * Used at resume time to restore the group in its previous state.
de85488138247d Boris Brezillon 2024-02-29  622  	 *
de85488138247d Boris Brezillon 2024-02-29  623  	 * The size of the suspend buffer is exposed through the FW interface.
de85488138247d Boris Brezillon 2024-02-29  624  	 */
de85488138247d Boris Brezillon 2024-02-29  625  	struct panthor_kernel_bo *suspend_buf;
de85488138247d Boris Brezillon 2024-02-29  626  
de85488138247d Boris Brezillon 2024-02-29  627  	/**
de85488138247d Boris Brezillon 2024-02-29  628  	 * @protm_suspend_buf: Protection mode suspend buffer.
de85488138247d Boris Brezillon 2024-02-29  629  	 *
de85488138247d Boris Brezillon 2024-02-29  630  	 * Stores the state of the group and its queues when a group that's in
de85488138247d Boris Brezillon 2024-02-29  631  	 * protection mode is suspended.
de85488138247d Boris Brezillon 2024-02-29  632  	 *
de85488138247d Boris Brezillon 2024-02-29  633  	 * Used at resume time to restore the group in its previous state.
de85488138247d Boris Brezillon 2024-02-29  634  	 *
de85488138247d Boris Brezillon 2024-02-29  635  	 * The size of the protection mode suspend buffer is exposed through the
de85488138247d Boris Brezillon 2024-02-29  636  	 * FW interface.
de85488138247d Boris Brezillon 2024-02-29  637  	 */
de85488138247d Boris Brezillon 2024-02-29  638  	struct panthor_kernel_bo *protm_suspend_buf;
de85488138247d Boris Brezillon 2024-02-29  639  
de85488138247d Boris Brezillon 2024-02-29  640  	/** @sync_upd_work: Work used to check/signal job fences. */
de85488138247d Boris Brezillon 2024-02-29  641  	struct work_struct sync_upd_work;
de85488138247d Boris Brezillon 2024-02-29  642  
de85488138247d Boris Brezillon 2024-02-29  643  	/** @tiler_oom_work: Work used to process tiler OOM events happening on this group. */
de85488138247d Boris Brezillon 2024-02-29  644  	struct work_struct tiler_oom_work;
de85488138247d Boris Brezillon 2024-02-29  645  
de85488138247d Boris Brezillon 2024-02-29  646  	/** @term_work: Work used to finish the group termination procedure. */
de85488138247d Boris Brezillon 2024-02-29  647  	struct work_struct term_work;
de85488138247d Boris Brezillon 2024-02-29  648  
de85488138247d Boris Brezillon 2024-02-29  649  	/**
de85488138247d Boris Brezillon 2024-02-29  650  	 * @release_work: Work used to release group resources.
de85488138247d Boris Brezillon 2024-02-29  651  	 *
de85488138247d Boris Brezillon 2024-02-29  652  	 * We need to postpone the group release to avoid a deadlock when
de85488138247d Boris Brezillon 2024-02-29  653  	 * the last ref is released in the tick work.
de85488138247d Boris Brezillon 2024-02-29  654  	 */
de85488138247d Boris Brezillon 2024-02-29  655  	struct work_struct release_work;
de85488138247d Boris Brezillon 2024-02-29  656  
de85488138247d Boris Brezillon 2024-02-29  657  	/**
de85488138247d Boris Brezillon 2024-02-29  658  	 * @run_node: Node used to insert the group in the
de85488138247d Boris Brezillon 2024-02-29  659  	 * panthor_group::groups::{runnable,idle} and
de85488138247d Boris Brezillon 2024-02-29  660  	 * panthor_group::reset.stopped_groups lists.
de85488138247d Boris Brezillon 2024-02-29  661  	 */
de85488138247d Boris Brezillon 2024-02-29  662  	struct list_head run_node;
de85488138247d Boris Brezillon 2024-02-29  663  
de85488138247d Boris Brezillon 2024-02-29  664  	/**
de85488138247d Boris Brezillon 2024-02-29  665  	 * @wait_node: Node used to insert the group in the
de85488138247d Boris Brezillon 2024-02-29  666  	 * panthor_group::groups::waiting list.
de85488138247d Boris Brezillon 2024-02-29  667  	 */
de85488138247d Boris Brezillon 2024-02-29  668  	struct list_head wait_node;
de85488138247d Boris Brezillon 2024-02-29 @669  };
de85488138247d Boris Brezillon 2024-02-29  670  

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [PATCH 2/2] drm/panthor: Expose scheduler groups info in debugfs

[Steven Price]

On 18/09/2024 09:50, Mary Guillemard wrote:
> This adds a new debugfs file named "sched_groups" allowing a user to
> query information about all userspace clients scheduler groups.
> 
> As we uses drm_device.filelist to achieve it, we also expose the
> client_id with the task information to differentiate one client
> from another inside the same task.
> 
> Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>

Can you explain a bit more why you want this? I'm not against the idea
but I'm wary of adding 'debugging features' because often they fall into
one of two camps:

 * Genuinely useful features that shouldn't be in debugfs because people
want a stable API.

 * Niche features that take maintainance effort despite not being
particularly useful.

> ---
>  drivers/gpu/drm/panthor/panthor_drv.c   |   1 +
>  drivers/gpu/drm/panthor/panthor_sched.c | 127 ++++++++++++++++++++++++
>  drivers/gpu/drm/panthor/panthor_sched.h |   4 +
>  3 files changed, 132 insertions(+)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_drv.c b/drivers/gpu/drm/panthor/panthor_drv.c
> index 0d825d63d712..aa390597d355 100644
> --- a/drivers/gpu/drm/panthor/panthor_drv.c
> +++ b/drivers/gpu/drm/panthor/panthor_drv.c
> @@ -1450,6 +1450,7 @@ static const struct file_operations panthor_drm_driver_fops = {
>  static void panthor_debugfs_init(struct drm_minor *minor)
>  {
>  	panthor_mmu_debugfs_init(minor);
> +	panthor_sched_debugfs_init(minor);
>  }
>  #endif
>  
> diff --git a/drivers/gpu/drm/panthor/panthor_sched.c b/drivers/gpu/drm/panthor/panthor_sched.c
> index f15abeef4ece..0c1ddbfbe164 100644
> --- a/drivers/gpu/drm/panthor/panthor_sched.c
> +++ b/drivers/gpu/drm/panthor/panthor_sched.c
> @@ -1,6 +1,7 @@
>  // SPDX-License-Identifier: GPL-2.0 or MIT
>  /* Copyright 2023 Collabora ltd. */
>  
> +#include <drm/drm_debugfs.h>
>  #include <drm/drm_drv.h>
>  #include <drm/drm_exec.h>
>  #include <drm/drm_gem_shmem_helper.h>
> @@ -3581,3 +3582,129 @@ int panthor_sched_init(struct panthor_device *ptdev)
>  	ptdev->scheduler = sched;
>  	return 0;
>  }
> +
> +#ifdef CONFIG_DEBUG_FS
> +static const char *panthor_csg_priority_names[PANTHOR_CSG_PRIORITY_COUNT] = {
> +	"LOW",
> +	"MEDIUM",
> +	"HIGH",
> +	"REALTIME"
> +};
> +
> +static const char *panthor_group_state_names[PANTHOR_CS_GROUP_UNKNOWN_STATE] = {
> +	"CREATED",
> +	"ACTIVE",
> +	"SUSPENDED",
> +	"TERMINATED"
> +};
> +
> +static void show_panthor_queue(const struct panthor_queue *queue,
> +			       u32 queue_index,
> +			       struct seq_file *m)
> +{
> +	seq_printf(m, "queue %u:", queue_index);
> +	seq_printf(m, " priority %u", queue->priority);
> +	seq_printf(m, " doorbell_id %d", queue->doorbell_id);
> +	seq_puts(m, "\n");
> +}
> +
> +static void show_panthor_group(struct panthor_group *group,
> +			       u32 group_handle,
> +			       struct seq_file *m)
> +{
> +	u32 i;
> +
> +	group_get(group);

This looks wrong - this function relies on group not becoming invalid
before the call, so we'd better be holding a sufficient lock/reference
on the group before show_panthor_group() is called otherwise we've got a
race before the group_get() call.

Although I can see this bug is present elsewhere in panthor_sched.c now
I look...

> +
> +	seq_printf(m, "group %u:", group_handle);
> +	seq_printf(m, " priority %s", group->priority < PANTHOR_CSG_PRIORITY_COUNT ?
> +		   panthor_csg_priority_names[group->priority] : "UNKNOWN");
> +	seq_printf(m, " state %s", group->state < PANTHOR_CS_GROUP_UNKNOWN_STATE ?
> +		   panthor_group_state_names[group->state] : "UNKNOWN");
> +	seq_printf(m, " csg_id %d", group->csg_id);
> +	seq_printf(m, " csg_priority %d", group->csg_priority);
> +	seq_printf(m, " compute_core_mask 0x%016llx", group->compute_core_mask);
> +	seq_printf(m, " fragment_core_mask 0x%016llx", group->fragment_core_mask);
> +	seq_printf(m, " tiler_core_mask 0x%016llx", group->tiler_core_mask);
> +	seq_printf(m, " max_compute_cores %d", group->max_compute_cores);
> +	seq_printf(m, " max_fragment_cores %d", group->max_fragment_cores);
> +	seq_printf(m, " max_tiler_cores %d", group->max_tiler_cores);
> +	seq_puts(m, "\n");
> +
> +	for (i = 0; i < group->queue_count; i++)
> +		show_panthor_queue(group->queues[i], i, m);
> +
> +	group_put(group);
> +}
> +
> +static int show_file_group_pool(const struct panthor_file *pfile, struct seq_file *m)
> +{
> +	struct panthor_group_pool *gpool = pfile->groups;
> +	struct panthor_group *group;
> +	unsigned long i;
> +
> +	if (IS_ERR_OR_NULL(gpool))
> +		return 0;
> +
> +	xa_for_each(&gpool->xa, i, group)
> +		show_panthor_group(group, i, m);

Here I see no protection against a racing call to PANTHOR_GROUP_DESTROY,
so group could become invalid between xa_for_each returning it and the
call to show_panthor_group().

Steve

> +
> +	return 0;
> +}
> +
> +static int show_each_file(struct seq_file *m, void *arg)
> +{
> +	struct drm_info_node *node = (struct drm_info_node *)m->private;
> +	struct drm_device *ddev = node->minor->dev;
> +	int (*show)(const struct panthor_file *, struct seq_file *) = node->info_ent->data;
> +	struct drm_file *file;
> +	int ret;
> +
> +	ret = mutex_lock_interruptible(&ddev->filelist_mutex);
> +	if (ret)
> +		return ret;
> +
> +	list_for_each_entry(file, &ddev->filelist, lhead) {
> +		struct task_struct *task;
> +		struct panthor_file *pfile = file->driver_priv;
> +		struct pid *pid;
> +
> +		/*
> +		 * Although we have a valid reference on file->pid, that does
> +		 * not guarantee that the task_struct who called get_pid() is
> +		 * still alive (e.g. get_pid(current) => fork() => exit()).
> +		 * Therefore, we need to protect this ->comm access using RCU.
> +		 */
> +		rcu_read_lock();
> +		pid = rcu_dereference(file->pid);
> +		task = pid_task(pid, PIDTYPE_TGID);
> +		seq_printf(m, "client_id %8llu pid %8d command %s:\n", file->client_id,
> +			   pid_nr(pid), task ? task->comm : "<unknown>");
> +		rcu_read_unlock();
> +
> +		ret = show(pfile, m);
> +		if (ret < 0)
> +			break;
> +
> +		seq_puts(m, "\n");
> +	}
> +
> +	mutex_unlock(&ddev->filelist_mutex);
> +	return ret;
> +}
> +
> +static struct drm_info_list panthor_sched_debugfs_list[] = {
> +	{ "sched_groups", show_each_file, 0, show_file_group_pool },
> +};
> +
> +/**
> + * panthor_sched_debugfs_init() - Initialize scheduler debugfs entries
> + * @minor: Minor.
> + */
> +void panthor_sched_debugfs_init(struct drm_minor *minor)
> +{
> +	drm_debugfs_create_files(panthor_sched_debugfs_list,
> +				 ARRAY_SIZE(panthor_sched_debugfs_list),
> +				 minor->debugfs_root, minor);
> +}
> +#endif /* CONFIG_DEBUG_FS */
> diff --git a/drivers/gpu/drm/panthor/panthor_sched.h b/drivers/gpu/drm/panthor/panthor_sched.h
> index 3a30d2328b30..577239aa66bf 100644
> --- a/drivers/gpu/drm/panthor/panthor_sched.h
> +++ b/drivers/gpu/drm/panthor/panthor_sched.h
> @@ -47,4 +47,8 @@ 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);
>  
> +#ifdef CONFIG_DEBUG_FS
> +void panthor_sched_debugfs_init(struct drm_minor *minor);
> +#endif
> +
>  #endif

Re: [PATCH 1/2] drm/panthor: Add csg_priority to panthor_group

[Steven Price]

On 18/09/2024 09:50, Mary Guillemard wrote:
> Keep track of the csg priority in panthor_group when the group is
> scheduled/active.
> 
> This is useful to know the actual priority in use in the firmware
> group slot.
> 
> Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
> ---
>  drivers/gpu/drm/panthor/panthor_sched.c | 18 +++++++++++++++---
>  1 file changed, 15 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_sched.c b/drivers/gpu/drm/panthor/panthor_sched.c
> index 86908ada7335..f15abeef4ece 100644
> --- a/drivers/gpu/drm/panthor/panthor_sched.c
> +++ b/drivers/gpu/drm/panthor/panthor_sched.c
> @@ -574,6 +574,13 @@ struct panthor_group {
>  	 */
>  	int csg_id;
>  
> +	/**
> +	 * @csg_id: Priority of the FW group slot.

kerneldoc name is wrong: s/csg_id/csg_priority/

Otherwise this looks reasonable, but see my reply to the second patch.

Thanks,
Steve

> +	 *
> +	 * -1 when the group is not scheduled/active.
> +	 */
> +	int csg_priority;
> +
>  	/**
>  	 * @destroyed: True when the group has been destroyed.
>  	 *
> @@ -894,11 +901,12 @@ group_get(struct panthor_group *group)
>   * group_bind_locked() - Bind a group to a group slot
>   * @group: Group.
>   * @csg_id: Slot.
> + * @csg_priority: Priority of the slot.
>   *
>   * Return: 0 on success, a negative error code otherwise.
>   */
>  static int
> -group_bind_locked(struct panthor_group *group, u32 csg_id)
> +group_bind_locked(struct panthor_group *group, u32 csg_id, u32 csg_priority)
>  {
>  	struct panthor_device *ptdev = group->ptdev;
>  	struct panthor_csg_slot *csg_slot;
> @@ -917,6 +925,7 @@ group_bind_locked(struct panthor_group *group, u32 csg_id)
>  	csg_slot = &ptdev->scheduler->csg_slots[csg_id];
>  	group_get(group);
>  	group->csg_id = csg_id;
> +	group->csg_priority = csg_priority;
>  
>  	/* Dummy doorbell allocation: doorbell is assigned to the group and
>  	 * all queues use the same doorbell.
> @@ -956,6 +965,7 @@ group_unbind_locked(struct panthor_group *group)
>  	slot = &ptdev->scheduler->csg_slots[group->csg_id];
>  	panthor_vm_idle(group->vm);
>  	group->csg_id = -1;
> +	group->csg_priority = -1;
>  
>  	/* Tiler OOM events will be re-issued next time the group is scheduled. */
>  	atomic_set(&group->tiler_oom, 0);
> @@ -2193,8 +2203,9 @@ tick_ctx_apply(struct panthor_scheduler *sched, struct panthor_sched_tick_ctx *c
>  
>  			csg_iface = panthor_fw_get_csg_iface(ptdev, csg_id);
>  			csg_slot = &sched->csg_slots[csg_id];
> -			group_bind_locked(group, csg_id);
> -			csg_slot_prog_locked(ptdev, csg_id, new_csg_prio--);
> +			group_bind_locked(group, csg_id, new_csg_prio);
> +			csg_slot_prog_locked(ptdev, csg_id, new_csg_prio);
> +			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,
> @@ -3111,6 +3122,7 @@ int panthor_group_create(struct panthor_file *pfile,
>  	kref_init(&group->refcount);
>  	group->state = PANTHOR_CS_GROUP_CREATED;
>  	group->csg_id = -1;
> +	group->csg_priority = -1;
>  
>  	group->ptdev = ptdev;
>  	group->max_compute_cores = group_args->max_compute_cores;