Merge tag 'amd-drm-next-6.9-2024-03-08-1' of https://gitlab.freedeskt…

…op.org/agd5f/linux into drm-next

amd-drm-next-6.9-2024-03-08-1:

amdgpu:
- DCN 3.5.1 support
- Fixes for IOMMUv2 removal
- UAF fix
- Misc small fixes and cleanups
- SR-IOV fixes
- MCBP cleanup
- devcoredump update
- NBIF 6.3.1 support
- VPE 6.1.1 support

amdkfd:
- Misc fixes and cleanups
- GFX10.1 trap fixes

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240308170741.3691166-1-alexander.deucher@amd.com

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -98,7 +98,7 @@ amdgpu-y += \
 	vega20_reg_init.o nbio_v7_4.o nbio_v2_3.o nv.o arct_reg_init.o mxgpu_nv.o \
 	nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o soc21.o \
 	sienna_cichlid.o smu_v13_0_10.o nbio_v4_3.o hdp_v6_0.o nbio_v7_7.o hdp_v5_2.o lsdma_v6_0.o \
-	nbio_v7_9.o aqua_vanjaram.o nbio_v7_11.o lsdma_v7_0.o hdp_v7_0.o
+	nbio_v7_9.o aqua_vanjaram.o nbio_v7_11.o lsdma_v7_0.o hdp_v7_0.o nbif_v6_3_1.o
 
 # add DF block
 amdgpu-y += \

drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c

@@ -1782,9 +1782,14 @@ static int amdgpu_debugfs_vm_info_show(struct seq_file *m, void *unused)
 	list_for_each_entry(file, &dev->filelist, lhead) {
 		struct amdgpu_fpriv *fpriv = file->driver_priv;
 		struct amdgpu_vm *vm = &fpriv->vm;
+		struct amdgpu_task_info *ti;
+
+		ti = amdgpu_vm_get_task_info_vm(vm);
+		if (ti) {
+			seq_printf(m, "pid:%d\tProcess:%s ----------\n", ti->pid, ti->process_name);
+			amdgpu_vm_put_task_info(ti);
+		}
 
-		seq_printf(m, "pid:%d\tProcess:%s ----------\n",
-				vm->task_info.pid, vm->task_info.process_name);
 		r = amdgpu_bo_reserve(vm->root.bo, true);
 		if (r)
 			break;

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -4056,13 +4056,13 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		goto unmap_memory;
 	}
 
+	amdgpu_device_set_mcbp(adev);
+
 	/* early init functions */
 	r = amdgpu_device_ip_early_init(adev);
 	if (r)
 		goto unmap_memory;
 
-	amdgpu_device_set_mcbp(adev);
-
 	/* Get rid of things like offb */
 	r = drm_aperture_remove_conflicting_pci_framebuffers(adev->pdev, &amdgpu_kms_driver);
 	if (r)

drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c

@@ -61,6 +61,7 @@
 #include "nbio_v4_3.h"
 #include "nbio_v7_2.h"
 #include "nbio_v7_7.h"
+#include "nbif_v6_3_1.h"
 #include "hdp_v5_0.h"
 #include "hdp_v5_2.h"
 #include "hdp_v6_0.h"
@@ -1319,6 +1320,15 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 				}
 			}
 
+			if (le16_to_cpu(ip->hw_id) == VPE_HWID) {
+				if (adev->vpe.num_instances < AMDGPU_MAX_VPE_INSTANCES)
+					adev->vpe.num_instances++;
+				else
+					dev_err(adev->dev, "Too many VPE instances: %d vs %d\n",
+						adev->vpe.num_instances + 1,
+						AMDGPU_MAX_VPE_INSTANCES);
+			}
+
 			if (le16_to_cpu(ip->hw_id) == UMC_HWID) {
 				adev->gmc.num_umc++;
 				adev->umc.node_inst_num++;
@@ -1936,6 +1946,7 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		case IP_VERSION(3, 2, 0):
 		case IP_VERSION(3, 2, 1):
 		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
 			if (amdgpu_sriov_vf(adev))
 				amdgpu_discovery_set_sriov_display(adev);
 			else
@@ -2212,6 +2223,7 @@ static int amdgpu_discovery_set_vpe_ip_blocks(struct amdgpu_device *adev)
 {
 	switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) {
 	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
 		amdgpu_device_ip_block_add(adev, &vpe_v6_1_ip_block);
 		break;
 	default:
@@ -2558,6 +2570,10 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->nbio.funcs = &nbio_v7_7_funcs;
 		adev->nbio.hdp_flush_reg = &nbio_v7_7_hdp_flush_reg;
 		break;
+	case IP_VERSION(6, 3, 1):
+		adev->nbio.funcs = &nbif_v6_3_1_funcs;
+		adev->nbio.hdp_flush_reg = &nbif_v6_3_1_hdp_flush_reg;
+		break;
 	default:
 		break;
 	}

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -2693,7 +2693,7 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	}
 
 	adev->in_runpm = true;
-	if (amdgpu_device_supports_px(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
 
 	/*
@@ -2703,7 +2703,7 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	 * platforms.
 	 * TODO: this may be also needed for PX capable platform.
 	 */
-	if (amdgpu_device_supports_boco(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 		adev->mp1_state = PP_MP1_STATE_UNLOAD;
 
 	ret = amdgpu_device_prepare(drm_dev);
@@ -2712,15 +2712,15 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	ret = amdgpu_device_suspend(drm_dev, false);
 	if (ret) {
 		adev->in_runpm = false;
-		if (amdgpu_device_supports_boco(drm_dev))
+		if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 			adev->mp1_state = PP_MP1_STATE_NONE;
 		return ret;
 	}
 
-	if (amdgpu_device_supports_boco(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 		adev->mp1_state = PP_MP1_STATE_NONE;
 
-	if (amdgpu_device_supports_px(drm_dev)) {
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX) {
 		/* Only need to handle PCI state in the driver for ATPX
 		 * PCI core handles it for _PR3.
 		 */
@@ -2729,9 +2729,9 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 		pci_ignore_hotplug(pdev);
 		pci_set_power_state(pdev, PCI_D3cold);
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_DYNAMIC_OFF;
-	} else if (amdgpu_device_supports_boco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO) {
 		/* nothing to do */
-	} else if (amdgpu_device_supports_baco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) {
 		amdgpu_device_baco_enter(drm_dev);
 	}
 
@@ -2754,7 +2754,7 @@ static int amdgpu_pmops_runtime_resume(struct device *dev)
 	if (!pci_device_is_present(adev->pdev))
 		adev->no_hw_access = true;
 
-	if (amdgpu_device_supports_px(drm_dev)) {
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX) {
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
 
 		/* Only need to handle PCI state in the driver for ATPX
@@ -2766,22 +2766,22 @@ static int amdgpu_pmops_runtime_resume(struct device *dev)
 		if (ret)
 			return ret;
 		pci_set_master(pdev);
-	} else if (amdgpu_device_supports_boco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO) {
 		/* Only need to handle PCI state in the driver for ATPX
 		 * PCI core handles it for _PR3.
 		 */
 		pci_set_master(pdev);
-	} else if (amdgpu_device_supports_baco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) {
 		amdgpu_device_baco_exit(drm_dev);
 	}
 	ret = amdgpu_device_resume(drm_dev, false);
 	if (ret) {
-		if (amdgpu_device_supports_px(drm_dev))
+		if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 			pci_disable_device(pdev);
 		return ret;
 	}
 
-	if (amdgpu_device_supports_px(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_ON;
 	adev->in_runpm = false;
 	return 0;

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c

@@ -208,9 +208,15 @@ static int amdgpu_gem_object_open(struct drm_gem_object *obj,
 	if (!WARN_ON(!vm->process_info->eviction_fence)) {
 		r = amdgpu_amdkfd_bo_validate_and_fence(abo, AMDGPU_GEM_DOMAIN_GTT,
 							&vm->process_info->eviction_fence->base);
-		if (r)
-			dev_warn(adev->dev, "%d: validate_and_fence failed: %d\n",
-				 vm->task_info.pid, r);
+		if (r) {
+			struct amdgpu_task_info *ti = amdgpu_vm_get_task_info_vm(vm);
+
+			dev_warn(adev->dev, "validate_and_fence failed: %d\n", r);
+			if (ti) {
+				dev_warn(adev->dev, "pid %d\n", ti->pid);
+				amdgpu_vm_put_task_info(ti);
+			}
+		}
 	}
 	mutex_unlock(&vm->process_info->lock);
 

drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c

@@ -131,18 +131,18 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 	struct amdgpu_ib *ib = &ibs[0];
 	struct dma_fence *tmp = NULL;
 	bool need_ctx_switch;
-	unsigned int patch_offset = ~0;
 	struct amdgpu_vm *vm;
 	uint64_t fence_ctx;
 	uint32_t status = 0, alloc_size;
 	unsigned int fence_flags = 0;
 	bool secure, init_shadow;
 	u64 shadow_va, csa_va, gds_va;
 	int vmid = AMDGPU_JOB_GET_VMID(job);
+	bool need_pipe_sync = false;
+	unsigned int cond_exec;
 
 	unsigned int i;
 	int r = 0;
-	bool need_pipe_sync = false;
 
 	if (num_ibs == 0)
 		return -EINVAL;
@@ -228,7 +228,8 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 					    init_shadow, vmid);
 
 	if (ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+		cond_exec = amdgpu_ring_init_cond_exec(ring,
+						       ring->cond_exe_gpu_addr);
 
 	amdgpu_device_flush_hdp(adev, ring);
 
@@ -278,16 +279,9 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 				       fence_flags | AMDGPU_FENCE_FLAG_64BIT);
 	}
 
-	if (ring->funcs->emit_gfx_shadow) {
+	if (ring->funcs->emit_gfx_shadow && ring->funcs->init_cond_exec) {
 		amdgpu_ring_emit_gfx_shadow(ring, 0, 0, 0, false, 0);
-
-		if (ring->funcs->init_cond_exec) {
-			unsigned int ce_offset = ~0;
-
-			ce_offset = amdgpu_ring_init_cond_exec(ring);
-			if (ce_offset != ~0 && ring->funcs->patch_cond_exec)
-				amdgpu_ring_patch_cond_exec(ring, ce_offset);
-		}
+		amdgpu_ring_init_cond_exec(ring, ring->cond_exe_gpu_addr);
 	}
 
 	r = amdgpu_fence_emit(ring, f, job, fence_flags);
@@ -302,8 +296,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 	if (ring->funcs->insert_end)
 		ring->funcs->insert_end(ring);
 
-	if (patch_offset != ~0 && ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, cond_exec);
 
 	ring->current_ctx = fence_ctx;
 	if (vm && ring->funcs->emit_switch_buffer)

drivers/gpu/drm/amd/amdgpu/amdgpu_job.c

@@ -35,7 +35,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 {
 	struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
 	struct amdgpu_job *job = to_amdgpu_job(s_job);
-	struct amdgpu_task_info ti;
+	struct amdgpu_task_info *ti;
 	struct amdgpu_device *adev = ring->adev;
 	int idx;
 	int r;
@@ -48,7 +48,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 		return DRM_GPU_SCHED_STAT_ENODEV;
 	}
 
-	memset(&ti, 0, sizeof(struct amdgpu_task_info));
+
 	adev->job_hang = true;
 
 	if (amdgpu_gpu_recovery &&
@@ -58,12 +58,16 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 		goto exit;
 	}
 
-	amdgpu_vm_get_task_info(ring->adev, job->pasid, &ti);
 	DRM_ERROR("ring %s timeout, signaled seq=%u, emitted seq=%u\n",
-		  job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
-		  ring->fence_drv.sync_seq);
-	DRM_ERROR("Process information: process %s pid %d thread %s pid %d\n",
-		  ti.process_name, ti.tgid, ti.task_name, ti.pid);
+		   job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
+		   ring->fence_drv.sync_seq);
+
+	ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
+	if (ti) {
+		DRM_ERROR("Process information: process %s pid %d thread %s pid %d\n",
+			  ti->process_name, ti->tgid, ti->task_name, ti->pid);
+		amdgpu_vm_put_task_info(ti);
+	}
 
 	dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
 

drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c

@@ -196,6 +196,13 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,
 			   coredump->reset_task_info.process_name,
 			   coredump->reset_task_info.pid);
 
+	if (coredump->ring) {
+		drm_printf(&p, "\nRing timed out details\n");
+		drm_printf(&p, "IP Type: %d Ring Name: %s\n",
+			   coredump->ring->funcs->type,
+			   coredump->ring->name);
+	}
+
 	if (coredump->reset_vram_lost)
 		drm_printf(&p, "VRAM is lost due to GPU reset!\n");
 	if (coredump->adev->reset_info.num_regs) {
@@ -220,6 +227,8 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,
 {
 	struct amdgpu_coredump_info *coredump;
 	struct drm_device *dev = adev_to_drm(adev);
+	struct amdgpu_job *job = reset_context->job;
+	struct drm_sched_job *s_job;
 
 	coredump = kzalloc(sizeof(*coredump), GFP_NOWAIT);
 
@@ -230,8 +239,21 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,
 
 	coredump->reset_vram_lost = vram_lost;
 
-	if (reset_context->job && reset_context->job->vm)
-		coredump->reset_task_info = reset_context->job->vm->task_info;
+	if (reset_context->job && reset_context->job->vm) {
+		struct amdgpu_task_info *ti;
+		struct amdgpu_vm *vm = reset_context->job->vm;
+
+		ti = amdgpu_vm_get_task_info_vm(vm);
+		if (ti) {
+			coredump->reset_task_info = *ti;
+			amdgpu_vm_put_task_info(ti);
+		}
+	}
+
+	if (job) {
+		s_job = &job->base;
+		coredump->ring = to_amdgpu_ring(s_job->sched);
+	}
 
 	coredump->adev = adev;
 

drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h

@@ -97,6 +97,7 @@ struct amdgpu_coredump_info {
 	struct amdgpu_task_info         reset_task_info;
 	struct timespec64               reset_time;
 	bool                            reset_vram_lost;
+	struct amdgpu_ring			*ring;
 };
 #endif