Skip to contentSkip to conversation

MISC fixes

PATCH In review Oak Zeng 4 participants 11 messages Jul 16, 2021

Subsystems (3)
lore
Share & tools
lore

Oak Zeng

author
#0 lore raw

cover MISC fixes

Oak Zeng (5):
drm/amdgpu: Fix a printing message
drm/amdgpu: Change a few function names drm/amdkfd: Renaming dqm->packets to dqm->dpm drm/amdkfd: Set priv_queue to NULL after it is freed drm/amdkfd: Fix a concurrency issue during kfd recovery

 drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c            | 16 +++++-----
 drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c              |  2 +-
 drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c              |  2 +-
 drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c              |  2 +-
 drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c              |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_device.c            |  8 +----
 .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c  | 36 +++++++++++++---------
 .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.h  |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c    |  4 +++
 drivers/gpu/drm/amd/amdkfd/kfd_topology.c          |  2 +-
 10 files changed, 40 insertions(+), 36 deletions(-)
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
author
#1 lore raw

PATCH 1/5 drm/amdgpu: Fix a printing message

The printing message "PSP loading VCN firmware" is mis-leading because people might think driver is loading VCN firmware. Actually when this message is printed, driver is just preparing some VCN ucode, not loading VCN firmware yet. The actual VCN firmware loading will be in the PSP block hw_init. Fix the printing message

---
 drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 2 +-
 drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 2 +-
 drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 2 +-
 drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 2 +-
 4 files changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +1 -1
index 284bb42..121ee9f 100644
@@ -119,7 +119,7 @@ static int vcn_v1_0_sw_init(void *handle)
 		adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].fw = adev->vcn.fw;
 		adev->firmware.fw_size +=
 			ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
-		DRM_INFO("PSP loading VCN firmware\n");
+		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
 	}
 
 	r = amdgpu_vcn_resume(adev);
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c +1 -1
index 8af567c..f4686e9 100644
@@ -122,7 +122,7 @@ static int vcn_v2_0_sw_init(void *handle)
 		adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].fw = adev->vcn.fw;
 		adev->firmware.fw_size +=
 			ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
-		DRM_INFO("PSP loading VCN firmware\n");
+		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
 	}
 
 	r = amdgpu_vcn_resume(adev);
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -1
index 888b17d..e0c0c37 100644
@@ -152,7 +152,7 @@ static int vcn_v2_5_sw_init(void *handle)
 			adev->firmware.fw_size +=
 				ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
 		}
-		DRM_INFO("PSP loading VCN firmware\n");
+		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
 	}
 
 	r = amdgpu_vcn_resume(adev);
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +1 -1
index c3580de..a1bbe33 100644
@@ -158,7 +158,7 @@ static int vcn_v3_0_sw_init(void *handle)
 			adev->firmware.fw_size +=
 				ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
 		}
-		DRM_INFO("PSP loading VCN firmware\n");
+		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
 	}
 
 	r = amdgpu_vcn_resume(adev);
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Reviewed-by Acked-by on v1 1/5
#2 lore raw

Re: [PATCH 1/5] drm/amdgpu: Fix a printing message

Am 16.07.21 um 03:34 schrieb Oak Zeng:

Quoted 7 lines

The printing message "PSP loading VCN firmware" is mis-leading because people might think driver is loading VCN firmware. Actually when this message is printed, driver is just preparing some VCN ucode, not loading VCN firmware yet. The actual VCN firmware loading will be in the PSP block hw_init. Fix the printing message

Regards,
Christian.

Quoted 51 lines
---
  drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 2 +-
  drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 2 +-
  drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 2 +-
  drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 2 +-
  4 files changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +1 -1
index 284bb42..121ee9f 100644
@@ -119,7 +119,7 @@ static int vcn_v1_0_sw_init(void *handle)
  		adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].fw = adev->vcn.fw;
  		adev->firmware.fw_size +=
  			ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
-		DRM_INFO("PSP loading VCN firmware\n");
+		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
  	}
 
  	r = amdgpu_vcn_resume(adev);
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c +1 -1
index 8af567c..f4686e9 100644
@@ -122,7 +122,7 @@ static int vcn_v2_0_sw_init(void *handle)
  		adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].fw = adev->vcn.fw;
  		adev->firmware.fw_size +=
  			ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
-		DRM_INFO("PSP loading VCN firmware\n");
+		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
  	}
 
  	r = amdgpu_vcn_resume(adev);
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +1 -1
index 888b17d..e0c0c37 100644
@@ -152,7 +152,7 @@ static int vcn_v2_5_sw_init(void *handle)
  			adev->firmware.fw_size +=
  				ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
  		}
-		DRM_INFO("PSP loading VCN firmware\n");
+		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
  	}
 
  	r = amdgpu_vcn_resume(adev);
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +0 -0
index c3580de..a1bbe33 100644

drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c static int vcn_v3_0_sw_init(void *handle) · line 164

Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +1 -1
@@ -158,7 +158,7 @@ static int vcn_v3_0_sw_init(void *handle)
  			adev->firmware.fw_size +=
  				ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
  		}
-		DRM_INFO("PSP loading VCN firmware\n");
+		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
  	}
 
  	r = amdgpu_vcn_resume(adev);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

author
#3 lore raw

PATCH 2/5 drm/amdgpu: Change a few function names

In reply to Oak Zeng

Function name "psp_np_fw_load" is not proper as people don't know _np_fw_ means "non psp firmware". Change the function name to psp_load_non_psp_fw for better understanding. Same thing for function psp_execute_np_fw_load.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +8 -8
index d9ddb2c..8d1e2b2 100644
@@ -2351,7 +2351,7 @@ static int psp_prep_load_ip_fw_cmd_buf(struct amdgpu_firmware_info *ucode,
 	return ret;
 }
 
-static int psp_execute_np_fw_load(struct psp_context *psp,
+static int psp_execute_non_psp_fw_load(struct psp_context *psp,
 			          struct amdgpu_firmware_info *ucode)
 {
 	int ret = 0;
@@ -2387,7 +2387,7 @@ static int psp_load_smu_fw(struct psp_context *psp)
 		}
 	}
 
-	ret = psp_execute_np_fw_load(psp, ucode);
+	ret = psp_execute_non_psp_fw_load(psp, ucode);
 
 	if (ret)
 		DRM_ERROR("PSP load smu failed!\n");
@@ -2442,14 +2442,14 @@ int psp_load_fw_list(struct psp_context *psp,
 	for (i = 0; i < ucode_count; ++i) {
 		ucode = ucode_list[i];
 		psp_print_fw_hdr(psp, ucode);
-		ret = psp_execute_np_fw_load(psp, ucode);
+		ret = psp_execute_non_psp_fw_load(psp, ucode);
 		if (ret)
 			return ret;
 	}
 	return ret;
 }
 
-static int psp_np_fw_load(struct psp_context *psp)
+static int psp_load_non_psp_fw(struct psp_context *psp)
 {
 	int i, ret;
 	struct amdgpu_firmware_info *ucode;
@@ -2488,7 +2488,7 @@ static int psp_np_fw_load(struct psp_context *psp)
 
 		psp_print_fw_hdr(psp, ucode);
 
-		ret = psp_execute_np_fw_load(psp, ucode);
+		ret = psp_execute_non_psp_fw_load(psp, ucode);
 		if (ret)
 			return ret;
 
@@ -2565,7 +2565,7 @@ static int psp_load_fw(struct amdgpu_device *adev)
 	if (ret)
 		goto failed;
 
-	ret = psp_np_fw_load(psp);
+	ret = psp_load_non_psp_fw(psp);
 	if (ret)
 		goto failed;
 
@@ -2765,7 +2765,7 @@ static int psp_resume(void *handle)
 	if (ret)
 		goto failed;
 
-	ret = psp_np_fw_load(psp);
+	ret = psp_load_non_psp_fw(psp);
 	if (ret)
 		goto failed;
 
@@ -2863,7 +2863,7 @@ int psp_update_vcn_sram(struct amdgpu_device *adev, int inst_idx,
 	ucode.mc_addr = cmd_gpu_addr;
 	ucode.ucode_size = cmd_size;
 
-	return psp_execute_np_fw_load(&adev->psp, &ucode);
+	return psp_execute_non_psp_fw_load(&adev->psp, &ucode);
 }
 
 int psp_ring_cmd_submit(struct psp_context *psp,
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
author
#4 lore raw

PATCH 3/5 drm/amdkfd: Renaming dqm->packets to dqm->dpm

In reply to Oak Zeng

Renaming packets to dpm (device packet manager) to reflect the real meaning of this variable.

---
 drivers/gpu/drm/amd/amdkfd/kfd_device.c            |  2 +-
 .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c  | 26 +++++++++++-----------
 .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.h  |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_topology.c          |  2 +-
 4 files changed, 16 insertions(+), 16 deletions(-)
drivers/gpu/drm/amd/amdkfd/kfd_device.c +1 -1
index 9e4a05e..c51402b 100644
@@ -1390,7 +1390,7 @@ int kfd_debugfs_hang_hws(struct kfd_dev *dev)
 		return -EINVAL;
 	}
 
-	r = pm_debugfs_hang_hws(&dev->dqm->packets);
+	r = pm_debugfs_hang_hws(&dev->dqm->dpm);
 	if (!r)
 		r = dqm_debugfs_execute_queues(dev->dqm);
 
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +13 -13
index 16a1713..f2984d3 100644
@@ -260,7 +260,7 @@ static int allocate_vmid(struct device_queue_manager *dqm,
 static int flush_texture_cache_nocpsch(struct kfd_dev *kdev,
 				struct qcm_process_device *qpd)
 {
-	const struct packet_manager_funcs *pmf = qpd->dqm->packets.pmf;
+	const struct packet_manager_funcs *pmf = qpd->dqm->dpm.pmf;
 	int ret;
 
 	if (!qpd->ib_kaddr)
@@ -1000,7 +1000,7 @@ static int start_nocpsch(struct device_queue_manager *dqm)
 	init_interrupts(dqm);
 	
 	if (dqm->dev->device_info->asic_family == CHIP_HAWAII)
-		return pm_init(&dqm->packets, dqm);
+		return pm_init(&dqm->dpm, dqm);
 	dqm->sched_running = true;
 
 	return 0;
@@ -1009,7 +1009,7 @@ static int start_nocpsch(struct device_queue_manager *dqm)
 static int stop_nocpsch(struct device_queue_manager *dqm)
 {
 	if (dqm->dev->device_info->asic_family == CHIP_HAWAII)
-		pm_uninit(&dqm->packets, false);
+		pm_uninit(&dqm->dpm, false);
 	dqm->sched_running = false;
 
 	return 0;
@@ -1124,7 +1124,7 @@ static int set_sched_resources(struct device_queue_manager *dqm)
 			"queue mask: 0x%8llX\n",
 			res.vmid_mask, res.queue_mask);
 
-	return pm_send_set_resources(&dqm->packets, &res);
+	return pm_send_set_resources(&dqm->dpm, &res);
 }
 
 static int initialize_cpsch(struct device_queue_manager *dqm)
@@ -1164,7 +1164,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	retval = 0;
 
-	retval = pm_init(&dqm->packets, dqm);
+	retval = pm_init(&dqm->dpm, dqm);
 	if (retval)
 		goto fail_packet_manager_init;
 
@@ -1197,7 +1197,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 	return 0;
 fail_allocate_vidmem:
 fail_set_sched_resources:
-	pm_uninit(&dqm->packets, false);
+	pm_uninit(&dqm->dpm, false);
 fail_packet_manager_init:
 	return retval;
 }
@@ -1213,10 +1213,10 @@ static int stop_cpsch(struct device_queue_manager *dqm)
 	dqm->sched_running = false;
 	dqm_unlock(dqm);
 
-	pm_release_ib(&dqm->packets);
+	pm_release_ib(&dqm->dpm);
 
 	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
-	pm_uninit(&dqm->packets, hanging);
+	pm_uninit(&dqm->dpm, hanging);
 
 	return 0;
 }
@@ -1390,7 +1390,7 @@ static int map_queues_cpsch(struct device_queue_manager *dqm)
 	if (dqm->active_runlist)
 		return 0;
 
-	retval = pm_send_runlist(&dqm->packets, &dqm->queues);
+	retval = pm_send_runlist(&dqm->dpm, &dqm->queues);
 	pr_debug("%s sent runlist\n", __func__);
 	if (retval) {
 		pr_err("failed to execute runlist\n");
@@ -1416,13 +1416,13 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 	if (!dqm->active_runlist)
 		return retval;
 
-	retval = pm_send_unmap_queue(&dqm->packets, KFD_QUEUE_TYPE_COMPUTE,
+	retval = pm_send_unmap_queue(&dqm->dpm, KFD_QUEUE_TYPE_COMPUTE,
 			filter, filter_param, false, 0);
 	if (retval)
 		return retval;
 
 	*dqm->fence_addr = KFD_FENCE_INIT;
-	pm_send_query_status(&dqm->packets, dqm->fence_gpu_addr,
+	pm_send_query_status(&dqm->dpm, dqm->fence_gpu_addr,
 				KFD_FENCE_COMPLETED);
 	/* should be timed out */
 	retval = amdkfd_fence_wait_timeout(dqm->fence_addr, KFD_FENCE_COMPLETED,
@@ -1448,14 +1448,14 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 	 * check those fields
 	 */
 	mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ];
-	if (mqd_mgr->read_doorbell_id(dqm->packets.priv_queue->queue->mqd)) {
+	if (mqd_mgr->read_doorbell_id(dqm->dpm.priv_queue->queue->mqd)) {
 		pr_err("HIQ MQD's queue_doorbell_id0 is not 0, Queue preemption time out\n");
 		while (halt_if_hws_hang)
 			schedule();
 		return -ETIME;
 	}
 
-	pm_release_ib(&dqm->packets);
+	pm_release_ib(&dqm->dpm);
 	dqm->active_runlist = false;
 
 	return retval;
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h +1 -1
index 71e2fde..14479e8 100644
@@ -169,7 +169,7 @@ struct device_queue_manager {
 	struct device_queue_manager_asic_ops asic_ops;
 
 	struct mqd_manager	*mqd_mgrs[KFD_MQD_TYPE_MAX];
-	struct packet_manager	packets;
+	struct packet_manager	dpm;
 	struct kfd_dev		*dev;
 	struct mutex		lock_hidden; /* use dqm_lock/unlock(dqm) */
 	struct list_head	queues;
drivers/gpu/drm/amd/amdkfd/kfd_topology.c +1 -1
index b1ce072..748e82b 100644
@@ -1630,7 +1630,7 @@ int kfd_debugfs_rls_by_device(struct seq_file *m, void *data)
 		}
 
 		seq_printf(m, "Node %u, gpu_id %x:\n", i++, dev->gpu->id);
-		r = pm_debugfs_runlist(m, &dev->gpu->dqm->packets);
+		r = pm_debugfs_runlist(m, &dev->gpu->dqm->dpm);
 		if (r)
 			break;
 	}
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
maintainer
#5 lore raw

Re: [PATCH 3/5] drm/amdkfd: Renaming dqm->packets to dqm->dpm

Am 2021-07-15 um 9:34 p.m. schrieb Oak Zeng:

Quoted 2 lines

Renaming packets to dpm (device packet manager) to reflect the real meaning of this variable.

I don't think introducing another new acronym is helpful. Also "dpm" and "dqm" are visually too similar. Other places use "pm" for packet manager. If you feel "pm" is too short or too ambiguous, how about "pmgr" or "packet_mgr"?

Regards,
  Felix

Quoted 153 lines
---
 drivers/gpu/drm/amd/amdkfd/kfd_device.c            |  2 +-
 .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c  | 26 +++++++++++-----------
 .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.h  |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_topology.c          |  2 +-
 4 files changed, 16 insertions(+), 16 deletions(-)
drivers/gpu/drm/amd/amdkfd/kfd_device.c +1 -1
index 9e4a05e..c51402b 100644
@@ -1390,7 +1390,7 @@ int kfd_debugfs_hang_hws(struct kfd_dev *dev)
 		return -EINVAL;
 	}
 
-	r = pm_debugfs_hang_hws(&dev->dqm->packets);
+	r = pm_debugfs_hang_hws(&dev->dqm->dpm);
 	if (!r)
 		r = dqm_debugfs_execute_queues(dev->dqm);
 
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +13 -13
index 16a1713..f2984d3 100644
@@ -260,7 +260,7 @@ static int allocate_vmid(struct device_queue_manager *dqm,
 static int flush_texture_cache_nocpsch(struct kfd_dev *kdev,
 				struct qcm_process_device *qpd)
 {
-	const struct packet_manager_funcs *pmf = qpd->dqm->packets.pmf;
+	const struct packet_manager_funcs *pmf = qpd->dqm->dpm.pmf;
 	int ret;
 
 	if (!qpd->ib_kaddr)
@@ -1000,7 +1000,7 @@ static int start_nocpsch(struct device_queue_manager *dqm)
 	init_interrupts(dqm);
 
 	if (dqm->dev->device_info->asic_family == CHIP_HAWAII)
-		return pm_init(&dqm->packets, dqm);
+		return pm_init(&dqm->dpm, dqm);
 	dqm->sched_running = true;
 
 	return 0;
@@ -1009,7 +1009,7 @@ static int start_nocpsch(struct device_queue_manager *dqm)
 static int stop_nocpsch(struct device_queue_manager *dqm)
 {
 	if (dqm->dev->device_info->asic_family == CHIP_HAWAII)
-		pm_uninit(&dqm->packets, false);
+		pm_uninit(&dqm->dpm, false);
 	dqm->sched_running = false;
 
 	return 0;
@@ -1124,7 +1124,7 @@ static int set_sched_resources(struct device_queue_manager *dqm)
 			"queue mask: 0x%8llX\n",
 			res.vmid_mask, res.queue_mask);
 
-	return pm_send_set_resources(&dqm->packets, &res);
+	return pm_send_set_resources(&dqm->dpm, &res);
 }
 
 static int initialize_cpsch(struct device_queue_manager *dqm)
@@ -1164,7 +1164,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	retval = 0;
 
-	retval = pm_init(&dqm->packets, dqm);
+	retval = pm_init(&dqm->dpm, dqm);
 	if (retval)
 		goto fail_packet_manager_init;
 
@@ -1197,7 +1197,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 	return 0;
 fail_allocate_vidmem:
 fail_set_sched_resources:
-	pm_uninit(&dqm->packets, false);
+	pm_uninit(&dqm->dpm, false);
 fail_packet_manager_init:
 	return retval;
 }
@@ -1213,10 +1213,10 @@ static int stop_cpsch(struct device_queue_manager *dqm)
 	dqm->sched_running = false;
 	dqm_unlock(dqm);
 
-	pm_release_ib(&dqm->packets);
+	pm_release_ib(&dqm->dpm);
 
 	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
-	pm_uninit(&dqm->packets, hanging);
+	pm_uninit(&dqm->dpm, hanging);
 
 	return 0;
 }
@@ -1390,7 +1390,7 @@ static int map_queues_cpsch(struct device_queue_manager *dqm)
 	if (dqm->active_runlist)
 		return 0;
 
-	retval = pm_send_runlist(&dqm->packets, &dqm->queues);
+	retval = pm_send_runlist(&dqm->dpm, &dqm->queues);
 	pr_debug("%s sent runlist\n", __func__);
 	if (retval) {
 		pr_err("failed to execute runlist\n");
@@ -1416,13 +1416,13 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 	if (!dqm->active_runlist)
 		return retval;
 
-	retval = pm_send_unmap_queue(&dqm->packets, KFD_QUEUE_TYPE_COMPUTE,
+	retval = pm_send_unmap_queue(&dqm->dpm, KFD_QUEUE_TYPE_COMPUTE,
 			filter, filter_param, false, 0);
 	if (retval)
 		return retval;
 
 	*dqm->fence_addr = KFD_FENCE_INIT;
-	pm_send_query_status(&dqm->packets, dqm->fence_gpu_addr,
+	pm_send_query_status(&dqm->dpm, dqm->fence_gpu_addr,
 				KFD_FENCE_COMPLETED);
 	/* should be timed out */
 	retval = amdkfd_fence_wait_timeout(dqm->fence_addr, KFD_FENCE_COMPLETED,
@@ -1448,14 +1448,14 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 	 * check those fields
 	 */
 	mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ];
-	if (mqd_mgr->read_doorbell_id(dqm->packets.priv_queue->queue->mqd)) {
+	if (mqd_mgr->read_doorbell_id(dqm->dpm.priv_queue->queue->mqd)) {
 		pr_err("HIQ MQD's queue_doorbell_id0 is not 0, Queue preemption time out\n");
 		while (halt_if_hws_hang)
 			schedule();
 		return -ETIME;
 	}
 
-	pm_release_ib(&dqm->packets);
+	pm_release_ib(&dqm->dpm);
 	dqm->active_runlist = false;
 
 	return retval;
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h +1 -1
index 71e2fde..14479e8 100644
@@ -169,7 +169,7 @@ struct device_queue_manager {
 	struct device_queue_manager_asic_ops asic_ops;
 
 	struct mqd_manager	*mqd_mgrs[KFD_MQD_TYPE_MAX];
-	struct packet_manager	packets;
+	struct packet_manager	dpm;
 	struct kfd_dev		*dev;
 	struct mutex		lock_hidden; /* use dqm_lock/unlock(dqm) */
 	struct list_head	queues;
drivers/gpu/drm/amd/amdkfd/kfd_topology.c +0 -0
index b1ce072..748e82b 100644

drivers/gpu/drm/amd/amdkfd/kfd_topology.c int kfd_debugfs_rls_by_device(struct seq_file *m, void *data) · line 1636

Quoted 8 lines
drivers/gpu/drm/amd/amdkfd/kfd_topology.c +1 -1
@@ -1630,7 +1630,7 @@ int kfd_debugfs_rls_by_device(struct seq_file *m, void *data)
 		}
 
 		seq_printf(m, "Node %u, gpu_id %x:\n", i++, dev->gpu->id);
-		r = pm_debugfs_runlist(m, &dev->gpu->dqm->packets);
+		r = pm_debugfs_runlist(m, &dev->gpu->dqm->dpm);
 		if (r)
 			break;
 	}

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

author
#6 lore raw

PATCH 4/5 drm/amdkfd: Set priv_queue to NULL after it is freed

In reply to Oak Zeng

This variable will be used to determine whether packet manager is initialized or not, in a future patch.

---
 drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 1 +
 1 file changed, 1 insertion(+)
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +1 -0
index d8e940f..b130cc0 100644
@@ -278,6 +278,7 @@ void pm_uninit(struct packet_manager *pm, bool hanging)
 {
 	mutex_destroy(&pm->lock);
 	kernel_queue_uninit(pm->priv_queue, hanging);
+	pm->priv_queue = NULL;
 }
 
 int pm_send_set_resources(struct packet_manager *pm,
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
maintainer Reviewed-by on v1 4/5
#7 lore raw

Re: [PATCH 4/5] drm/amdkfd: Set priv_queue to NULL after it is freed

Am 2021-07-15 um 9:34 p.m. schrieb Oak Zeng:

Quoted 4 lines

This variable will be used to determine whether packet manager is initialized or not, in a future patch.

Quoted 9 lines
---
 drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 1 +
 1 file changed, 1 insertion(+)
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +0 -0
index d8e940f..b130cc0 100644

drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c void pm_uninit(struct packet_manager *pm, bool hanging) · line 284

Quoted 7 lines
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +1 -0
@@ -278,6 +278,7 @@ void pm_uninit(struct packet_manager *pm, bool hanging)
 {
 	mutex_destroy(&pm->lock);
 	kernel_queue_uninit(pm->priv_queue, hanging);
+	pm->priv_queue = NULL;
 }
 
 int pm_send_set_resources(struct packet_manager *pm,

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

author
#8 lore raw

PATCH 5/5 drm/amdkfd: Fix a concurrency issue during kfd recovery

In reply to Oak Zeng

start_cpsch and stop_cpsch can be called during kfd device initialization or during gpu reset/recovery. So they can run concurrently. Currently in start_cpsch and stop_cpsch, pm_init and pm_uninit is not protected by the dpm lock. Imagine such a case that user use packet manager's function to submit a pm4 packet to hang hws (ie through command cat /sys/class/kfd/kfd/topology/nodes/1/gpu_id | sudo tee /sys/kernel/debug/kfd/hang_hws), while kfd device is under device reset/recovery so packet manager can be not initialized. There will be unpredictable protection fault in such case.

This patch moves pm_init/uninit inside the dpm lock and check packet manager is initialized before using packet manager function.

---
 drivers/gpu/drm/amd/amdkfd/kfd_device.c               |  8 +-------
 drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 10 ++++++++--
 drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c       |  3 +++
 3 files changed, 12 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdkfd/kfd_device.c +1 -7
index c51402b..adc2342 100644
@@ -1383,18 +1383,12 @@ void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint32_t throttle_bitmask)
  */
 int kfd_debugfs_hang_hws(struct kfd_dev *dev)
 {
-	int r = 0;
-
 	if (dev->dqm->sched_policy != KFD_SCHED_POLICY_HWS) {
 		pr_err("HWS is not enabled");
 		return -EINVAL;
 	}
 
-	r = pm_debugfs_hang_hws(&dev->dqm->dpm);
-	if (!r)
-		r = dqm_debugfs_execute_queues(dev->dqm);
-
-	return r;
+	return dqm_debugfs_execute_queues(dev->dqm);
 }
 
 #endif
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +8 -2
index f2984d3..44136d2 100644
@@ -1164,6 +1164,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	retval = 0;
 
+	dqm_lock(dqm);
 	retval = pm_init(&dqm->dpm, dqm);
 	if (retval)
 		goto fail_packet_manager_init;
@@ -1186,7 +1187,6 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	init_interrupts(dqm);
 
-	dqm_lock(dqm);
 	/* clear hang status when driver try to start the hw scheduler */
 	dqm->is_hws_hang = false;
 	dqm->is_resetting = false;
@@ -1199,6 +1199,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 fail_set_sched_resources:
 	pm_uninit(&dqm->dpm, false);
 fail_packet_manager_init:
+	dqm_unlock(dqm);
 	return retval;
 }
 
@@ -1211,12 +1212,12 @@ static int stop_cpsch(struct device_queue_manager *dqm)
 		unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	hanging = dqm->is_hws_hang || dqm->is_resetting;
 	dqm->sched_running = false;
-	dqm_unlock(dqm);
 
 	pm_release_ib(&dqm->dpm);
 
 	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
 	pm_uninit(&dqm->dpm, hanging);
+	dqm_unlock(dqm);
 
 	return 0;
 }
@@ -2104,6 +2105,11 @@ int dqm_debugfs_execute_queues(struct device_queue_manager *dqm)
 	int r = 0;
 
 	dqm_lock(dqm);
+	r = pm_debugfs_hang_hws(&dqm->dpm);
+	if (r) {
+		dqm_unlock(dqm);
+		return r;
+	}
 	dqm->active_runlist = true;
 	r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	dqm_unlock(dqm);
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +3 -0
index b130cc0..0123e64 100644
@@ -448,6 +448,9 @@ int pm_debugfs_hang_hws(struct packet_manager *pm)
 	uint32_t *buffer, size;
 	int r = 0;
 
+	if (!pm->priv_queue)
+		return -EBUSY;
+
 	size = pm->pmf->query_status_size;
 	mutex_lock(&pm->lock);
 	kq_acquire_packet_buffer(pm->priv_queue,
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#9 lore raw

Re: [PATCH 5/5] drm/amdkfd: Fix a concurrency issue during kfd recovery

Quoted 89 lines · Oak Zeng

start_cpsch and stop_cpsch can be called during kfd device initialization or during gpu reset/recovery. So they can run concurrently. Currently in start_cpsch and stop_cpsch, pm_init and pm_uninit is not protected by the dpm lock. Imagine such a case that user use packet manager's function to submit a pm4 packet to hang hws (ie through command cat /sys/class/kfd/kfd/topology/nodes/1/gpu_id | sudo tee /sys/kernel/debug/kfd/hang_hws), while kfd device is under device reset/recovery so packet manager can be not initialized. There will be unpredictable protection fault in such case.

This patch moves pm_init/uninit inside the dpm lock and check packet manager is initialized before using packet manager function.

---
  drivers/gpu/drm/amd/amdkfd/kfd_device.c               |  8 +-------
  drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 10 ++++++++--
  drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c       |  3 +++
  3 files changed, 12 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdkfd/kfd_device.c +1 -7
index c51402b..adc2342 100644
@@ -1383,18 +1383,12 @@ void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint32_t throttle_bitmask)
   */
  int kfd_debugfs_hang_hws(struct kfd_dev *dev)
  {
-	int r = 0;
-
  	if (dev->dqm->sched_policy != KFD_SCHED_POLICY_HWS) {
  		pr_err("HWS is not enabled");
  		return -EINVAL;
  	}
 
-	r = pm_debugfs_hang_hws(&dev->dqm->dpm);
-	if (!r)
-		r = dqm_debugfs_execute_queues(dev->dqm);
-
-	return r;
+	return dqm_debugfs_execute_queues(dev->dqm);
  }
 
  #endif
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +3 -2
index f2984d3..44136d2 100644
@@ -1164,6 +1164,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
  	retval = 0;
 
+	dqm_lock(dqm);
  	retval = pm_init(&dqm->dpm, dqm);
  	if (retval)
  		goto fail_packet_manager_init;
@@ -1186,7 +1187,6 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
  	init_interrupts(dqm);
 
-	dqm_lock(dqm);
  	/* clear hang status when driver try to start the hw scheduler */
  	dqm->is_hws_hang = false;
  	dqm->is_resetting = false;
@@ -1199,6 +1199,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
  fail_set_sched_resources:
  	pm_uninit(&dqm->dpm, false);
  fail_packet_manager_init:
+	dqm_unlock(dqm);
  	return retval;
  }
 
@@ -1211,12 +1212,12 @@ static int stop_cpsch(struct device_queue_manager *dqm)
  		unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
  	hanging = dqm->is_hws_hang || dqm->is_resetting;
  	dqm->sched_running = false;
-	dqm_unlock(dqm);
 
  	pm_release_ib(&dqm->dpm);
 
  	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
  	pm_uninit(&dqm->dpm, hanging);
+	dqm_unlock(dqm);
 
  	return 0;
  }

drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c int dqm_debugfs_execute_queues(struct device_queue_manager *dqm) · line 2108

Quoted 4 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +1 -0
@@ -2104,6 +2105,11 @@ int dqm_debugfs_execute_queues(struct device_queue_manager *dqm)
  	int r = 0;
 
  	dqm_lock(dqm);
+	r = pm_debugfs_hang_hws(&dqm->dpm);

execute_queues look like a misnomer now as it combines the logic to hang HWS.

Quoted 12 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +4 -0
+	if (r) {
+		dqm_unlock(dqm);
+		return r;
+	}
  	dqm->active_runlist = true;
  	r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
  	dqm_unlock(dqm);
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +0 -0
index b130cc0..0123e64 100644

drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c int pm_debugfs_hang_hws(struct packet_manager *pm) · line 453

Quoted 6 lines
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +3 -0
@@ -448,6 +448,9 @@ int pm_debugfs_hang_hws(struct packet_manager *pm)
  	uint32_t *buffer, size;
  	int r = 0;
 
+	if (!pm->priv_queue)
+		return -EBUSY;
+

EAGAIN may be more appropriate

Thanks,
Lijo

drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c int pm_debugfs_hang_hws(struct packet_manager *pm) · line 456

Quoted 4 lines
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +0 -0
  	size = pm->pmf->query_status_size;
  	mutex_lock(&pm->lock);
  	kq_acquire_packet_buffer(pm->priv_queue,

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

maintainer
#10 lore raw

Re: [PATCH 5/5] drm/amdkfd: Fix a concurrency issue during kfd recovery

In reply to Oak Zeng

Am 2021-07-15 um 9:34 p.m. schrieb Oak Zeng:

Quoted 33 lines

start_cpsch and stop_cpsch can be called during kfd device initialization or during gpu reset/recovery. So they can run concurrently. Currently in start_cpsch and stop_cpsch, pm_init and pm_uninit is not protected by the dpm lock. Imagine such a case that user use packet manager's function to submit a pm4 packet to hang hws (ie through command cat /sys/class/kfd/kfd/topology/nodes/1/gpu_id | sudo tee /sys/kernel/debug/kfd/hang_hws), while kfd device is under device reset/recovery so packet manager can be not initialized. There will be unpredictable protection fault in such case.

This patch moves pm_init/uninit inside the dpm lock and check packet manager is initialized before using packet manager function.

---
 drivers/gpu/drm/amd/amdkfd/kfd_device.c               |  8 +-------
 drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 10 ++++++++--
 drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c       |  3 +++
 3 files changed, 12 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdkfd/kfd_device.c +0 -2
index c51402b..adc2342 100644
@@ -1383,18 +1383,12 @@ void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint32_t throttle_bitmask)
  */
 int kfd_debugfs_hang_hws(struct kfd_dev *dev)
 {
-	int r = 0;
-
 	if (dev->dqm->sched_policy != KFD_SCHED_POLICY_HWS) {

drivers/gpu/drm/amd/amdkfd/kfd_device.c void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint32_t throttle_bitmask) · line 1391

Quoted 10 lines
drivers/gpu/drm/amd/amdkfd/kfd_device.c +1 -5
 		pr_err("HWS is not enabled");
 		return -EINVAL;
 	}
 
-	r = pm_debugfs_hang_hws(&dev->dqm->dpm);
-	if (!r)
-		r = dqm_debugfs_execute_queues(dev->dqm);
-
-	return r;
+	return dqm_debugfs_execute_queues(dev->dqm);

This function should now probably be renamed to something like dqm_debugfs_hang_hws. Other than that, this patch looks good to me.

Regards,
  Felix

Quoted 62 lines
drivers/gpu/drm/amd/amdkfd/kfd_device.c +0 -0
 }
 
 #endif
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +8 -2
index f2984d3..44136d2 100644
@@ -1164,6 +1164,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	retval = 0;
 
+	dqm_lock(dqm);
 	retval = pm_init(&dqm->dpm, dqm);
 	if (retval)
 		goto fail_packet_manager_init;
@@ -1186,7 +1187,6 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	init_interrupts(dqm);
 
-	dqm_lock(dqm);
 	/* clear hang status when driver try to start the hw scheduler */
 	dqm->is_hws_hang = false;
 	dqm->is_resetting = false;
@@ -1199,6 +1199,7 @@ static int start_cpsch(struct device_queue_manager *dqm)
 fail_set_sched_resources:
 	pm_uninit(&dqm->dpm, false);
 fail_packet_manager_init:
+	dqm_unlock(dqm);
 	return retval;
 }
 
@@ -1211,12 +1212,12 @@ static int stop_cpsch(struct device_queue_manager *dqm)
 		unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	hanging = dqm->is_hws_hang || dqm->is_resetting;
 	dqm->sched_running = false;
-	dqm_unlock(dqm);
 
 	pm_release_ib(&dqm->dpm);
 
 	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
 	pm_uninit(&dqm->dpm, hanging);
+	dqm_unlock(dqm);
 
 	return 0;
 }
@@ -2104,6 +2105,11 @@ int dqm_debugfs_execute_queues(struct device_queue_manager *dqm)
 	int r = 0;
 
 	dqm_lock(dqm);
+	r = pm_debugfs_hang_hws(&dqm->dpm);
+	if (r) {
+		dqm_unlock(dqm);
+		return r;
+	}
 	dqm->active_runlist = true;
 	r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	dqm_unlock(dqm);
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +0 -0
index b130cc0..0123e64 100644

drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c int pm_debugfs_hang_hws(struct packet_manager *pm) · line 456

Quoted 9 lines
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +3 -0
@@ -448,6 +448,9 @@ int pm_debugfs_hang_hws(struct packet_manager *pm)
 	uint32_t *buffer, size;
 	int r = 0;
 
+	if (!pm->priv_queue)
+		return -EBUSY;
+
 	size = pm->pmf->query_status_size;
 	mutex_lock(&pm->lock);
 	kq_acquire_packet_buffer(pm->priv_queue,

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx