Merge tag 'amd-drm-next-6.9-2024-03-01' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.9-2024-03-01:

amdgpu:
- GC 11.5.1 updates
- Misc display cleanups
- NBIO 7.9 updates
- Backlight fixes
- DMUB fixes
- MPO fixes
- atomfirmware table updates
- SR-IOV fixes
- VCN 4.x updates
- use RMW accessors for pci config registers
- PSR fixes
- Suspend/resume fixes
- RAS fixes
- ABM fixes
- Misc code cleanups
- SI DPM fix
- Revert freesync video

amdkfd:
- Misc cleanups
- Error handling fixes

radeon:
- use RMW accessors for pci config registers

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240301204857.13960-1-alexander.deucher@amd.com
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 files changed, 30 insertions, 10 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 80e90fdef291..f030cafc5a0a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -371,7 +371,7 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
 			goto err_wptr_map_gart;
 		}
 
-		err = amdgpu_amdkfd_map_gtt_bo_to_gart(dev->adev, wptr_bo);
+		err = amdgpu_amdkfd_map_gtt_bo_to_gart(wptr_bo);
 		if (err) {
 			pr_err("Failed to map wptr bo to GART\n");
 			goto err_wptr_map_gart;
@@ -2935,6 +2935,7 @@ static int kfd_ioctl_set_debug_trap(struct file *filep, struct kfd_process *p, v
 	if (IS_ERR_OR_NULL(target)) {
 		pr_debug("Cannot find process PID %i to debug\n", args->pid);
 		r = target ? PTR_ERR(target) : -ESRCH;
+		target = NULL;
 		goto out;
 	}
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index 002b08fa632f..7f2ae0d15d4a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -1676,6 +1676,7 @@ int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pc
 		case IP_VERSION(11, 0, 3):
 		case IP_VERSION(11, 0, 4):
 		case IP_VERSION(11, 5, 0):
+		case IP_VERSION(11, 5, 1):
 			num_of_cache_types =
 				kfd_fill_gpu_cache_info_from_gfx_config(kdev->kfd, *pcache_info);
 			break;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 0a9cf9dfc224..4d399c0c8a57 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -96,6 +96,7 @@ static void kfd_device_info_set_sdma_info(struct kfd_dev *kfd)
 	case IP_VERSION(6, 0, 2):
 	case IP_VERSION(6, 0, 3):
 	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
 		kfd->device_info.num_sdma_queues_per_engine = 8;
 		break;
 	default:
@@ -113,6 +114,7 @@ static void kfd_device_info_set_sdma_info(struct kfd_dev *kfd)
 	case IP_VERSION(6, 0, 2):
 	case IP_VERSION(6, 0, 3):
 	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
 		/* Reserve 1 for paging and 1 for gfx */
 		kfd->device_info.num_reserved_sdma_queues_per_engine = 2;
 		/* BIT(0)=engine-0 queue-0; BIT(1)=engine-1 queue-0; BIT(2)=engine-0 queue-1; ... */
@@ -165,6 +167,7 @@ static void kfd_device_info_set_event_interrupt_class(struct kfd_dev *kfd)
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
 		kfd->device_info.event_interrupt_class = &event_interrupt_class_v11;
 		break;
 	default:
@@ -420,6 +423,10 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 			gfx_target_version = 110500;
 			f2g = &gfx_v11_kfd2kgd;
 			break;
+		case IP_VERSION(11, 5, 1):
+			gfx_target_version = 110501;
+			f2g = &gfx_v11_kfd2kgd;
+			break;
 		default:
 			break;
 		}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index c0e71543389a..f4d395e38683 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -1903,6 +1903,10 @@ int amdkfd_fence_wait_timeout(struct device_queue_manager *dqm,
 	uint64_t *fence_addr =  dqm->fence_addr;
 
 	while (*fence_addr != fence_value) {
+		/* Fatal err detected, this response won't come */
+		if (amdgpu_amdkfd_is_fed(dqm->dev->adev))
+			return -EIO;
+
 		if (time_after(jiffies, end_jiffies)) {
 			dev_err(dev, "qcm fence wait loop timeout expired\n");
 			/* In HWS case, this is used to halt the driver thread
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
index 1bea629c49ca..32c926986dbb 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
@@ -286,7 +286,7 @@ err_no_space:
 	return -ENOMEM;
 }
 
-void kq_submit_packet(struct kernel_queue *kq)
+int kq_submit_packet(struct kernel_queue *kq)
 {
 #ifdef DEBUG
 	int i;
@@ -298,6 +298,10 @@ void kq_submit_packet(struct kernel_queue *kq)
 	}
 	pr_debug("\n");
 #endif
+	/* Fatal err detected, packet submission won't go through */
+	if (amdgpu_amdkfd_is_fed(kq->dev->adev))
+		return -EIO;
+
 	if (kq->dev->kfd->device_info.doorbell_size == 8) {
 		*kq->wptr64_kernel = kq->pending_wptr64;
 		write_kernel_doorbell64(kq->queue->properties.doorbell_ptr,
@@ -307,6 +311,8 @@ void kq_submit_packet(struct kernel_queue *kq)
 		write_kernel_doorbell(kq->queue->properties.doorbell_ptr,
 					kq->pending_wptr);
 	}
+
+	return 0;
 }
 
 void kq_rollback_packet(struct kernel_queue *kq)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
index 9a6244430845..e24ee50acdf0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
@@ -47,7 +47,7 @@
 int kq_acquire_packet_buffer(struct kernel_queue *kq,
 				size_t packet_size_in_dwords,
 				unsigned int **buffer_ptr);
-void kq_submit_packet(struct kernel_queue *kq);
+int kq_submit_packet(struct kernel_queue *kq);
 void kq_rollback_packet(struct kernel_queue *kq);
 
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
index 401096c103b2..d6f65f39072b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
@@ -288,7 +288,7 @@ int pm_send_set_resources(struct packet_manager *pm,
 
 	retval = pm->pmf->set_resources(pm, buffer, res);
 	if (!retval)
-		kq_submit_packet(pm->priv_queue);
+		retval = kq_submit_packet(pm->priv_queue);
 	else
 		kq_rollback_packet(pm->priv_queue);
 
@@ -325,7 +325,7 @@ int pm_send_runlist(struct packet_manager *pm, struct list_head *dqm_queues)
 	if (retval)
 		goto fail_create_runlist;
 
-	kq_submit_packet(pm->priv_queue);
+	retval = kq_submit_packet(pm->priv_queue);
 
 	mutex_unlock(&pm->lock);
 
@@ -361,7 +361,7 @@ int pm_send_query_status(struct packet_manager *pm, uint64_t fence_address,
 
 	retval = pm->pmf->query_status(pm, buffer, fence_address, fence_value);
 	if (!retval)
-		kq_submit_packet(pm->priv_queue);
+		retval = kq_submit_packet(pm->priv_queue);
 	else
 		kq_rollback_packet(pm->priv_queue);
 
@@ -392,7 +392,7 @@ int pm_update_grace_period(struct packet_manager *pm, uint32_t grace_period)
 
 		retval = pm->pmf->set_grace_period(pm, buffer, grace_period);
 		if (!retval)
-			kq_submit_packet(pm->priv_queue);
+			retval = kq_submit_packet(pm->priv_queue);
 		else
 			kq_rollback_packet(pm->priv_queue);
 	}
@@ -421,7 +421,7 @@ int pm_send_unmap_queue(struct packet_manager *pm,
 
 	retval = pm->pmf->unmap_queues(pm, buffer, filter, filter_param, reset);
 	if (!retval)
-		kq_submit_packet(pm->priv_queue);
+		retval = kq_submit_packet(pm->priv_queue);
 	else
 		kq_rollback_packet(pm->priv_queue);
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index bc9eb847ecfe..c51f131eaa2f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1997,8 +1997,9 @@ int kfd_topology_add_device(struct kfd_node *gpu)
 			HSA_CAP_ASIC_REVISION_MASK);
 
 	dev->node_props.location_id = pci_dev_id(gpu->adev->pdev);
-	if (KFD_GC_VERSION(dev->gpu->kfd) == IP_VERSION(9, 4, 3))
-		dev->node_props.location_id |= dev->gpu->node_id;
+	/* On multi-partition nodes, node id = location_id[31:28] */
+	if (gpu->kfd->num_nodes > 1)
+		dev->node_props.location_id |= (dev->gpu->node_id << 28);
 
 	dev->node_props.domain = pci_domain_nr(gpu->adev->pdev->bus);
 	dev->node_props.max_engine_clk_fcompute =