* [PATCH 2/4] habanalabs: remove some f/w descriptor validations
2022-09-20 12:52 [PATCH 1/4] habanalabs: build ASICs from new to old Oded Gabbay
@ 2022-09-20 12:52 ` Oded Gabbay
2022-09-20 12:52 ` [PATCH 3/4] habanalabs/gaudi: use 8KB aligned address for TPC kernels Oded Gabbay
2022-09-20 12:52 ` [PATCH 4/4] habanalabs: eliminate aggregate use warning Oded Gabbay
2 siblings, 0 replies; 4+ messages in thread
From: Oded Gabbay @ 2022-09-20 12:52 UTC (permalink / raw)
To: linux-kernel; +Cc: farah kassabri
From: farah kassabri <fkassabri@habana.ai>
To be forward-backward compatible with the firmware in the initial
communication during preboot, we need to remove the validation of the
header size. This will allow us to add more fields to the
lkd_fw_comms_desc structure.
Instead of the validation of the header size, we just print warning
when some mismatch in descriptor has been revealed, and we calculate
the CRC base on descriptor size reported by the firmware instead of
calculating it ourselves.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
drivers/misc/habanalabs/common/firmware_if.c | 43 +++++++-------------
1 file changed, 14 insertions(+), 29 deletions(-)
diff --git a/drivers/misc/habanalabs/common/firmware_if.c b/drivers/misc/habanalabs/common/firmware_if.c
index 26a7529083e1..2de6a9bd564d 100644
--- a/drivers/misc/habanalabs/common/firmware_if.c
+++ b/drivers/misc/habanalabs/common/firmware_if.c
@@ -1900,50 +1900,36 @@ static int hl_fw_dynamic_validate_descriptor(struct hl_device *hdev,
u64 addr;
int rc;
- if (le32_to_cpu(fw_desc->header.magic) != HL_COMMS_DESC_MAGIC) {
- dev_err(hdev->dev, "Invalid magic for dynamic FW descriptor (%x)\n",
+ if (le32_to_cpu(fw_desc->header.magic) != HL_COMMS_DESC_MAGIC)
+ dev_warn(hdev->dev, "Invalid magic for dynamic FW descriptor (%x)\n",
fw_desc->header.magic);
- return -EIO;
- }
- if (fw_desc->header.version != HL_COMMS_DESC_VER) {
- dev_err(hdev->dev, "Invalid version for dynamic FW descriptor (%x)\n",
+ if (fw_desc->header.version != HL_COMMS_DESC_VER)
+ dev_warn(hdev->dev, "Invalid version for dynamic FW descriptor (%x)\n",
fw_desc->header.version);
- return -EIO;
- }
/*
- * calc CRC32 of data without header.
+ * Calc CRC32 of data without header. use the size of the descriptor
+ * reported by firmware, without calculating it ourself, to allow adding
+ * more fields to the lkd_fw_comms_desc structure.
* note that no alignment/stride address issues here as all structures
- * are 64 bit padded
+ * are 64 bit padded.
*/
- data_size = sizeof(struct lkd_fw_comms_desc) -
- sizeof(struct comms_desc_header);
data_ptr = (u8 *)fw_desc + sizeof(struct comms_desc_header);
-
- if (le16_to_cpu(fw_desc->header.size) != data_size) {
- dev_err(hdev->dev,
- "Invalid descriptor size 0x%x, expected size 0x%zx\n",
- le16_to_cpu(fw_desc->header.size), data_size);
- return -EIO;
- }
+ data_size = le16_to_cpu(fw_desc->header.size);
data_crc32 = hl_fw_compat_crc32(data_ptr, data_size);
-
if (data_crc32 != le32_to_cpu(fw_desc->header.crc32)) {
- dev_err(hdev->dev,
- "CRC32 mismatch for dynamic FW descriptor (%x:%x)\n",
- data_crc32, fw_desc->header.crc32);
+ dev_err(hdev->dev, "CRC32 mismatch for dynamic FW descriptor (%x:%x)\n",
+ data_crc32, fw_desc->header.crc32);
return -EIO;
}
/* find memory region to which to copy the image */
addr = le64_to_cpu(fw_desc->img_addr);
region_id = hl_get_pci_memory_region(hdev, addr);
- if ((region_id != PCI_REGION_SRAM) &&
- ((region_id != PCI_REGION_DRAM))) {
- dev_err(hdev->dev,
- "Invalid region to copy FW image address=%llx\n", addr);
+ if ((region_id != PCI_REGION_SRAM) && ((region_id != PCI_REGION_DRAM))) {
+ dev_err(hdev->dev, "Invalid region to copy FW image address=%llx\n", addr);
return -EIO;
}
@@ -1960,8 +1946,7 @@ static int hl_fw_dynamic_validate_descriptor(struct hl_device *hdev,
fw_loader->dynamic_loader.fw_image_size,
region);
if (rc) {
- dev_err(hdev->dev,
- "invalid mem transfer request for FW image\n");
+ dev_err(hdev->dev, "invalid mem transfer request for FW image\n");
return rc;
}
--
2.25.1
^ permalink raw reply related [flat|nested] 4+ messages in thread* [PATCH 3/4] habanalabs/gaudi: use 8KB aligned address for TPC kernels
2022-09-20 12:52 [PATCH 1/4] habanalabs: build ASICs from new to old Oded Gabbay
2022-09-20 12:52 ` [PATCH 2/4] habanalabs: remove some f/w descriptor validations Oded Gabbay
@ 2022-09-20 12:52 ` Oded Gabbay
2022-09-20 12:52 ` [PATCH 4/4] habanalabs: eliminate aggregate use warning Oded Gabbay
2 siblings, 0 replies; 4+ messages in thread
From: Oded Gabbay @ 2022-09-20 12:52 UTC (permalink / raw)
To: linux-kernel; +Cc: Tomer Tayar
From: Tomer Tayar <ttayar@habana.ai>
I$ prefetch is enabled when sending a TPC kernel to initialize the TPC
memory, and it has a restriction that the base address will be aligned
to 8KB.
Currently the base address is 128 bytes from the start address of the
device SRAM, so prefetching will start 128 bytes before the actual
kernel memory.
Modify the kernel address to be 8KB aligned.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
drivers/misc/habanalabs/gaudi/gaudi.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index e80ebace49c8..92560414e843 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -984,9 +984,10 @@ static int _gaudi_init_tpc_mem(struct hl_device *hdev,
init_tpc_mem_pkt->ctl = cpu_to_le32(ctl);
init_tpc_mem_pkt->src_addr = cpu_to_le64(tpc_kernel_src_addr);
- dst_addr = (prop->sram_user_base_address &
- GAUDI_PKT_LIN_DMA_DST_ADDR_MASK) >>
- GAUDI_PKT_LIN_DMA_DST_ADDR_SHIFT;
+
+ /* TPC_CMD is configured with I$ prefetch enabled, so address should be aligned to 8KB */
+ dst_addr = FIELD_PREP(GAUDI_PKT_LIN_DMA_DST_ADDR_MASK,
+ round_up(prop->sram_user_base_address, SZ_8K));
init_tpc_mem_pkt->dst_addr |= cpu_to_le64(dst_addr);
job = hl_cs_allocate_job(hdev, QUEUE_TYPE_EXT, true);
--
2.25.1
^ permalink raw reply related [flat|nested] 4+ messages in thread
* [PATCH 4/4] habanalabs: eliminate aggregate use warning
2022-09-20 12:52 [PATCH 1/4] habanalabs: build ASICs from new to old Oded Gabbay
2022-09-20 12:52 ` [PATCH 2/4] habanalabs: remove some f/w descriptor validations Oded Gabbay
2022-09-20 12:52 ` [PATCH 3/4] habanalabs/gaudi: use 8KB aligned address for TPC kernels Oded Gabbay
@ 2022-09-20 12:52 ` Oded Gabbay
2 siblings, 0 replies; 4+ messages in thread
From: Oded Gabbay @ 2022-09-20 12:52 UTC (permalink / raw)
To: linux-kernel
When doing sizeof() and giving as argument a dereference of
a pointer-to-a-pointer object, clang will issue a warning.
Eliminate the warning by passing struct <name>*
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
drivers/misc/habanalabs/common/command_submission.c | 4 ++--
drivers/misc/habanalabs/common/device.c | 4 ++--
drivers/misc/habanalabs/common/hw_queue.c | 4 +---
drivers/misc/habanalabs/common/hwmon.c | 3 ++-
drivers/misc/habanalabs/common/memory.c | 3 +--
5 files changed, 8 insertions(+), 10 deletions(-)
diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index fbe5003191bf..fa05770865c6 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -2571,7 +2571,7 @@ static int hl_cs_poll_fences(struct multi_cs_data *mcs_data, struct multi_cs_com
ktime_t max_ktime, first_cs_time;
enum hl_cs_wait_status status;
- memset(fence_ptr, 0, arr_len * sizeof(*fence_ptr));
+ memset(fence_ptr, 0, arr_len * sizeof(struct hl_fence *));
/* get all fences under the same lock */
rc = hl_ctx_get_fences(mcs_data->ctx, seq_arr, fence_ptr, arr_len);
@@ -2873,7 +2873,7 @@ static int hl_multi_cs_wait_ioctl(struct hl_fpriv *hpriv, void *data)
}
/* allocate array for the fences */
- fence_arr = kmalloc_array(seq_arr_len, sizeof(*fence_arr), GFP_KERNEL);
+ fence_arr = kmalloc_array(seq_arr_len, sizeof(struct hl_fence *), GFP_KERNEL);
if (!fence_arr) {
rc = -ENOMEM;
goto free_seq_arr;
diff --git a/drivers/misc/habanalabs/common/device.c b/drivers/misc/habanalabs/common/device.c
index c6a00bb259fb..233d8b46c831 100644
--- a/drivers/misc/habanalabs/common/device.c
+++ b/drivers/misc/habanalabs/common/device.c
@@ -719,7 +719,7 @@ static int device_early_init(struct hl_device *hdev)
if (hdev->asic_prop.completion_queues_count) {
hdev->cq_wq = kcalloc(hdev->asic_prop.completion_queues_count,
- sizeof(*hdev->cq_wq),
+ sizeof(struct workqueue_struct *),
GFP_KERNEL);
if (!hdev->cq_wq) {
rc = -ENOMEM;
@@ -1863,7 +1863,7 @@ int hl_device_init(struct hl_device *hdev, struct class *hclass)
}
hdev->shadow_cs_queue = kcalloc(hdev->asic_prop.max_pending_cs,
- sizeof(*hdev->shadow_cs_queue), GFP_KERNEL);
+ sizeof(struct hl_cs *), GFP_KERNEL);
if (!hdev->shadow_cs_queue) {
rc = -ENOMEM;
goto cq_fini;
diff --git a/drivers/misc/habanalabs/common/hw_queue.c b/drivers/misc/habanalabs/common/hw_queue.c
index 3f15ab9d827f..d0087c0ec48c 100644
--- a/drivers/misc/habanalabs/common/hw_queue.c
+++ b/drivers/misc/habanalabs/common/hw_queue.c
@@ -826,9 +826,7 @@ static int ext_and_cpu_queue_init(struct hl_device *hdev, struct hl_hw_queue *q,
q->kernel_address = p;
- q->shadow_queue = kmalloc_array(HL_QUEUE_LENGTH,
- sizeof(*q->shadow_queue),
- GFP_KERNEL);
+ q->shadow_queue = kmalloc_array(HL_QUEUE_LENGTH, sizeof(struct hl_cs_job *), GFP_KERNEL);
if (!q->shadow_queue) {
dev_err(hdev->dev,
"Failed to allocate shadow queue for H/W queue %d\n",
diff --git a/drivers/misc/habanalabs/common/hwmon.c b/drivers/misc/habanalabs/common/hwmon.c
index 8c262aeb425e..55eb0203817f 100644
--- a/drivers/misc/habanalabs/common/hwmon.c
+++ b/drivers/misc/habanalabs/common/hwmon.c
@@ -194,7 +194,8 @@ int hl_build_hwmon_channel_info(struct hl_device *hdev, struct cpucp_sensor *sen
curr_arr[sensors_by_type_next_index[type]++] = flags;
}
- channels_info = kcalloc(num_active_sensor_types + 1, sizeof(*channels_info), GFP_KERNEL);
+ channels_info = kcalloc(num_active_sensor_types + 1, sizeof(struct hwmon_channel_info *),
+ GFP_KERNEL);
if (!channels_info) {
rc = -ENOMEM;
goto channels_info_array_err;
diff --git a/drivers/misc/habanalabs/common/memory.c b/drivers/misc/habanalabs/common/memory.c
index 096fa3c1ae95..ef28f3b37b93 100644
--- a/drivers/misc/habanalabs/common/memory.c
+++ b/drivers/misc/habanalabs/common/memory.c
@@ -2308,8 +2308,7 @@ static int get_user_memory(struct hl_device *hdev, u64 addr, u64 size,
return -EFAULT;
}
- userptr->pages = kvmalloc_array(npages, sizeof(*userptr->pages),
- GFP_KERNEL);
+ userptr->pages = kvmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
if (!userptr->pages)
return -ENOMEM;
--
2.25.1
^ permalink raw reply related [flat|nested] 4+ messages in thread