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

34 files changed, 1265 insertions, 562 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile
index fa26a4e3a99d..4536c8ad0e11 100644
--- a/drivers/gpu/drm/amd/amdgpu/Makefile
+++ b/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 += \
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index 1afbb2e932c6..f5d0fa207a88 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/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;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index e68bd6f8a6a4..1e9454e6e4cb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/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)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
index 502333725b49..a07e4b87d4ca 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
+++ b/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;
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
index 6acffedf648c..15b188aaf681 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
+++ b/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;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index 22aeee8adb71..67c234bcf89f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/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);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index 6aa3b1d845ab..8b512dc28df8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -131,7 +131,6 @@ 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;
@@ -139,10 +138,11 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 	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)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index 71a5cf37b472..4b3000c21ef2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/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);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
index 4baa300121d8..147100c27c2d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
+++ b/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;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
index 19899f6b9b2b..60522963aaca 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
+++ b/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
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
index fe1a61eb6e4c..582053f1cd56 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
@@ -209,8 +209,7 @@ struct amdgpu_ring_funcs {
 	void (*insert_end)(struct amdgpu_ring *ring);
 	/* pad the indirect buffer to the necessary number of dw */
 	void (*pad_ib)(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
-	unsigned (*init_cond_exec)(struct amdgpu_ring *ring);
-	void (*patch_cond_exec)(struct amdgpu_ring *ring, unsigned offset);
+	unsigned (*init_cond_exec)(struct amdgpu_ring *ring, uint64_t addr);
 	/* note usage for clock and power gating */
 	void (*begin_use)(struct amdgpu_ring *ring);
 	void (*end_use)(struct amdgpu_ring *ring);
@@ -286,6 +285,9 @@ struct amdgpu_ring {
 	unsigned		cond_exe_offs;
 	u64			cond_exe_gpu_addr;
 	volatile u32		*cond_exe_cpu_addr;
+	unsigned int		set_q_mode_offs;
+	volatile u32		*set_q_mode_ptr;
+	u64			set_q_mode_token;
 	unsigned		vm_hub;
 	unsigned		vm_inv_eng;
 	struct dma_fence	*vmid_wait;
@@ -327,8 +329,7 @@ struct amdgpu_ring {
 #define amdgpu_ring_emit_reg_write_reg_wait(r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m))
 #define amdgpu_ring_emit_frame_cntl(r, b, s) (r)->funcs->emit_frame_cntl((r), (b), (s))
 #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib)))
-#define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
-#define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
+#define amdgpu_ring_init_cond_exec(r, a) (r)->funcs->init_cond_exec((r), (a))
 #define amdgpu_ring_preempt_ib(r) (r)->funcs->preempt_ib(r)
 #define amdgpu_ring_patch_cntl(r, o) ((r)->funcs->patch_cntl((r), (o)))
 #define amdgpu_ring_patch_ce(r, o) ((r)->funcs->patch_ce((r), (o)))
@@ -411,6 +412,30 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
 	ring->count_dw -= count_dw;
 }
 
+/**
+ * amdgpu_ring_patch_cond_exec - patch dw count of conditional execute
+ * @ring: amdgpu_ring structure
+ * @offset: offset returned by amdgpu_ring_init_cond_exec
+ *
+ * Calculate the dw count and patch it into a cond_exec command.
+ */
+static inline void amdgpu_ring_patch_cond_exec(struct amdgpu_ring *ring,
+					       unsigned int offset)
+{
+	unsigned cur;
+
+	if (!ring->funcs->init_cond_exec)
+		return;
+
+	WARN_ON(offset > ring->buf_mask);
+	WARN_ON(ring->ring[offset] != 0);
+
+	cur = (ring->wptr - 1) & ring->buf_mask;
+	if (cur < offset)
+		cur += ring->ring_size >> 2;
+	ring->ring[offset] = cur - offset;
+}
+
 #define amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset)			\
 	(ring->is_mes_queue && ring->mes_ctx ?				\
 	 (ring->mes_ctx->meta_data_gpu_addr + offset) : 0)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index ed4a8c5d26d7..4299ce386322 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -513,8 +513,14 @@ int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		bo = bo_base->bo;
 
 		if (dma_resv_locking_ctx(bo->tbo.base.resv) != ticket) {
-			pr_warn_ratelimited("Evicted user BO is not reserved in pid %d\n",
-					    vm->task_info.pid);
+			struct amdgpu_task_info *ti = amdgpu_vm_get_task_info_vm(vm);
+
+			pr_warn_ratelimited("Evicted user BO is not reserved\n");
+			if (ti) {
+				pr_warn_ratelimited("pid %d\n", ti->pid);
+				amdgpu_vm_put_task_info(ti);
+			}
+
 			return -EINVAL;
 		}
 
@@ -652,7 +658,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 	bool vm_flush_needed = job->vm_needs_flush;
 	struct dma_fence *fence = NULL;
 	bool pasid_mapping_needed = false;
-	unsigned patch_offset = 0;
+	unsigned int patch;
 	int r;
 
 	if (amdgpu_vmid_had_gpu_reset(adev, id)) {
@@ -679,7 +685,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 
 	amdgpu_ring_ib_begin(ring);
 	if (ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+		patch = amdgpu_ring_init_cond_exec(ring,
+						   ring->cond_exe_gpu_addr);
 
 	if (need_pipe_sync)
 		amdgpu_ring_emit_pipeline_sync(ring);
@@ -727,8 +734,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 	}
 	dma_fence_put(fence);
 
-	if (ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, patch);
 
 	/* the double SWITCH_BUFFER here *cannot* be skipped by COND_EXEC */
 	if (ring->funcs->emit_switch_buffer) {
@@ -1385,10 +1391,6 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 			struct amdgpu_bo_va_mapping, list);
 		list_del(&mapping->list);
 
-		if (vm->pte_support_ats &&
-		    mapping->start < AMDGPU_GMC_HOLE_START)
-			init_pte_value = AMDGPU_PTE_DEFAULT_ATC;
-
 		r = amdgpu_vm_update_range(adev, vm, false, false, true, false,
 					   resv, mapping->start, mapping->last,
 					   init_pte_value, 0, 0, NULL, NULL,
@@ -2225,6 +2227,108 @@ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout)
 	return dma_fence_wait_timeout(vm->last_unlocked, true, timeout);
 }
 
+static void amdgpu_vm_destroy_task_info(struct kref *kref)
+{
+	struct amdgpu_task_info *ti = container_of(kref, struct amdgpu_task_info, refcount);
+
+	kfree(ti);
+}
+
+static inline struct amdgpu_vm *
+amdgpu_vm_get_vm_from_pasid(struct amdgpu_device *adev, u32 pasid)
+{
+	struct amdgpu_vm *vm;
+	unsigned long flags;
+
+	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
+	vm = xa_load(&adev->vm_manager.pasids, pasid);
+	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
+
+	return vm;
+}
+
+/**
+ * amdgpu_vm_put_task_info - reference down the vm task_info ptr
+ *
+ * @task_info: task_info struct under discussion.
+ *
+ * frees the vm task_info ptr at the last put
+ */
+void amdgpu_vm_put_task_info(struct amdgpu_task_info *task_info)
+{
+	kref_put(&task_info->refcount, amdgpu_vm_destroy_task_info);
+}
+
+/**
+ * amdgpu_vm_get_task_info_vm - Extracts task info for a vm.
+ *
+ * @vm: VM to get info from
+ *
+ * Returns the reference counted task_info structure, which must be
+ * referenced down with amdgpu_vm_put_task_info.
+ */
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_vm(struct amdgpu_vm *vm)
+{
+	struct amdgpu_task_info *ti = NULL;
+
+	if (vm) {
+		ti = vm->task_info;
+		kref_get(&vm->task_info->refcount);
+	}
+
+	return ti;
+}
+
+/**
+ * amdgpu_vm_get_task_info_pasid - Extracts task info for a PASID.
+ *
+ * @adev: drm device pointer
+ * @pasid: PASID identifier for VM
+ *
+ * Returns the reference counted task_info structure, which must be
+ * referenced down with amdgpu_vm_put_task_info.
+ */
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid)
+{
+	return amdgpu_vm_get_task_info_vm(
+			amdgpu_vm_get_vm_from_pasid(adev, pasid));
+}
+
+static int amdgpu_vm_create_task_info(struct amdgpu_vm *vm)
+{
+	vm->task_info = kzalloc(sizeof(struct amdgpu_task_info), GFP_KERNEL);
+	if (!vm->task_info)
+		return -ENOMEM;
+
+	kref_init(&vm->task_info->refcount);
+	return 0;
+}
+
+/**
+ * amdgpu_vm_set_task_info - Sets VMs task info.
+ *
+ * @vm: vm for which to set the info
+ */
+void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
+{
+	if (!vm->task_info)
+		return;
+
+	if (vm->task_info->pid == current->pid)
+		return;
+
+	vm->task_info->pid = current->pid;
+	get_task_comm(vm->task_info->task_name, current);
+
+	if (current->group_leader->mm != current->mm)
+		return;
+
+	vm->task_info->tgid = current->group_leader->pid;
+	get_task_comm(vm->task_info->process_name, current->group_leader);
+}
+
 /**
  * amdgpu_vm_init - initialize a vm instance
  *
@@ -2264,7 +2368,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (r)
 		return r;
 
-	vm->pte_support_ats = false;
 	vm->is_compute_context = false;
 
 	vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
@@ -2311,6 +2414,10 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (r)
 		goto error_free_root;
 
+	r = amdgpu_vm_create_task_info(vm);
+	if (r)
+		DRM_DEBUG("Failed to create task info for VM\n");
+
 	amdgpu_bo_unreserve(vm->root.bo);
 	amdgpu_bo_unref(&root_bo);
 
@@ -2350,30 +2457,12 @@ error_free_delayed:
  */
 int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
-	bool pte_support_ats = (adev->asic_type == CHIP_RAVEN);
 	int r;
 
 	r = amdgpu_bo_reserve(vm->root.bo, true);
 	if (r)
 		return r;
 
-	/* Check if PD needs to be reinitialized and do it before
-	 * changing any other state, in case it fails.
-	 */
-	if (pte_support_ats != vm->pte_support_ats) {
-		/* Sanity checks */
-		if (!amdgpu_vm_pt_is_root_clean(adev, vm)) {
-			r = -EINVAL;
-			goto unreserve_bo;
-		}
-
-		vm->pte_support_ats = pte_support_ats;
-		r = amdgpu_vm_pt_clear(adev, vm, to_amdgpu_bo_vm(vm->root.bo),
-				       false);
-		if (r)
-			goto unreserve_bo;
-	}
-
 	/* Update VM state */
 	vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
 				    AMDGPU_VM_USE_CPU_FOR_COMPUTE);
@@ -2450,6 +2539,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 
 	root = amdgpu_bo_ref(vm->root.bo);
 	amdgpu_bo_reserve(root, true);
+	amdgpu_vm_put_task_info(vm->task_info);
 	amdgpu_vm_set_pasid(adev, vm, 0);
 	dma_fence_wait(vm->last_unlocked, false);
 	dma_fence_put(vm->last_unlocked);
@@ -2607,48 +2697,6 @@ int amdgpu_vm_ioctl(str