Merge tag 'drm-next-5.5-2019-11-08' of git://people.freedesktop.org/~agd5f/linux into drm-next

drm-next-5.5-2019-11-08:

amdgpu:
- Enable VCN dynamic powergating on RV/RV2
- Fixes for Navi14
- Misc Navi fixes
- Fix MSI-X tear down
- Misc Arturus fixes
- Fix xgmi powerstate handling
- Documenation fixes

scheduler:
- Fix static code checker warning
- Fix possible thread reactivation while thread is stopped
- Avoid cleanup if thread is parked

radeon:
- SI dpm fix ported from amdgpu
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191108212713.5078-1-alexander.deucher@amd.com

Documentation/gpu/amdgpu.rst

@@ -82,12 +82,21 @@ AMDGPU XGMI Support
 AMDGPU RAS Support
 ==================
 
+The AMDGPU RAS interfaces are exposed via sysfs (for informational queries) and
+debugfs (for error injection).
+
 RAS debugfs/sysfs Control and Error Injection Interfaces
 --------------------------------------------------------
 
 .. kernel-doc:: drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
    :doc: AMDGPU RAS debugfs control interface
 
+RAS Reboot Behavior for Unrecoverable Errors
+--------------------------------------------------------
+
+.. kernel-doc:: drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+   :doc: AMDGPU RAS Reboot Behavior for Unrecoverable Errors
+
 RAS Error Count sysfs Interface
 -------------------------------
 
@@ -109,6 +118,32 @@ RAS VRAM Bad Pages sysfs Interface
 .. kernel-doc:: drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
    :internal:
 
+Sample Code
+-----------
+Sample code for testing error injection can be found here:
+https://cgit.freedesktop.org/mesa/drm/tree/tests/amdgpu/ras_tests.c
+
+This is part of the libdrm amdgpu unit tests which cover several areas of the GPU.
+There are four sets of tests:
+
+RAS Basic Test
+
+The test verifies the RAS feature enabled status and makes sure the necessary sysfs and debugfs files
+are present.
+
+RAS Query Test
+
+This test checks the RAS availability and enablement status for each supported IP block as well as
+the error counts.
+
+RAS Inject Test
+
+This test injects errors for each IP.
+
+RAS Disable Test
+
+This test tests disabling of RAS features for each IP block.
+
 
 GPU Power/Thermal Controls and Monitoring
 =========================================

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

@@ -977,6 +977,9 @@ struct amdgpu_device {
 
 	uint64_t			unique_id;
 	uint64_t	df_perfmon_config_assign_mask[AMDGPU_MAX_DF_PERFMONS];
+
+	/* device pstate */
+	int				pstate;
 };
 
 static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev)

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

@@ -33,7 +33,7 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size,
 {
 	unsigned long start_jiffies;
 	unsigned long end_jiffies;
-	struct dma_fence *fence = NULL;
+	struct dma_fence *fence;
 	int i, r;
 
 	start_jiffies = jiffies;
@@ -44,16 +44,14 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size,
 		if (r)
 			goto exit_do_move;
 		r = dma_fence_wait(fence, false);
+		dma_fence_put(fence);
 		if (r)
 			goto exit_do_move;
-		dma_fence_put(fence);
 	}
 	end_jiffies = jiffies;
 	r = jiffies_to_msecs(end_jiffies - start_jiffies);
 
 exit_do_move:
-	if (fence)
-		dma_fence_put(fence);
 	return r;
 }
 

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

@@ -859,6 +859,9 @@ static int amdgpu_debugfs_test_ib(struct seq_file *m, void *data)
 	struct amdgpu_device *adev = dev->dev_private;
 	int r = 0, i;
 
+	/* Avoid accidently unparking the sched thread during GPU reset */
+	mutex_lock(&adev->lock_reset);
+
 	/* hold on the scheduler */
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
@@ -884,6 +887,8 @@ static int amdgpu_debugfs_test_ib(struct seq_file *m, void *data)
 		kthread_unpark(ring->sched.thread);
 	}
 
+	mutex_unlock(&adev->lock_reset);
+
 	return 0;
 }
 
@@ -1036,6 +1041,9 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
 	if (!fences)
 		return -ENOMEM;
 
+	/* Avoid accidently unparking the sched thread during GPU reset */
+	mutex_lock(&adev->lock_reset);
+
 	/* stop the scheduler */
 	kthread_park(ring->sched.thread);
 
@@ -1075,6 +1083,8 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
 	/* restart the scheduler */
 	kthread_unpark(ring->sched.thread);
 
+	mutex_unlock(&adev->lock_reset);
+
 	ttm_bo_unlock_delayed_workqueue(&adev->mman.bdev, resched);
 
 	kfree(fences);

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

@@ -2057,6 +2057,7 @@ static int amdgpu_device_enable_mgpu_fan_boost(void)
  */
 static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 {
+	struct amdgpu_gpu_instance *gpu_instance;
 	int i = 0, r;
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
@@ -2082,8 +2083,39 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 	if (r)
 		DRM_ERROR("enable mgpu fan boost failed (%d).\n", r);
 
-	/* set to low pstate by default */
-	amdgpu_xgmi_set_pstate(adev, 0);
+
+	if (adev->gmc.xgmi.num_physical_nodes > 1) {
+		mutex_lock(&mgpu_info.mutex);
+
+		/*
+		 * Reset device p-state to low as this was booted with high.
+		 *
+		 * This should be performed only after all devices from the same
+		 * hive get initialized.
+		 *
+		 * However, it's unknown how many device in the hive in advance.
+		 * As this is counted one by one during devices initializations.
+		 *
+		 * So, we wait for all XGMI interlinked devices initialized.
+		 * This may bring some delays as those devices may come from
+		 * different hives. But that should be OK.
+		 */
+		if (mgpu_info.num_dgpu == adev->gmc.xgmi.num_physical_nodes) {
+			for (i = 0; i < mgpu_info.num_gpu; i++) {
+				gpu_instance = &(mgpu_info.gpu_ins[i]);
+				if (gpu_instance->adev->flags & AMD_IS_APU)
+					continue;
+
+				r = amdgpu_xgmi_set_pstate(gpu_instance->adev, 0);
+				if (r) {
+					DRM_ERROR("pstate setting failed (%d).\n", r);
+					break;
+				}
+			}
+		}
+
+		mutex_unlock(&mgpu_info.mutex);
+	}
 
 	return 0;
 }

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

@@ -127,18 +127,48 @@ struct amdgpu_xgmi {
 };
 
 struct amdgpu_gmc {
+	/* FB's physical address in MMIO space (for CPU to
+	 * map FB). This is different compared to the agp/
+	 * gart/vram_start/end field as the later is from
+	 * GPU's view and aper_base is from CPU's view.
+	 */
 	resource_size_t		aper_size;
 	resource_size_t		aper_base;
 	/* for some chips with <= 32MB we need to lie
 	 * about vram size near mc fb location */
 	u64			mc_vram_size;
 	u64			visible_vram_size;
+	/* AGP aperture start and end in MC address space
+	 * Driver find a hole in the MC address space
+	 * to place AGP by setting MC_VM_AGP_BOT/TOP registers
+	 * Under VMID0, logical address == MC address. AGP
+	 * aperture maps to physical bus or IOVA addressed.
+	 * AGP aperture is used to simulate FB in ZFB case.
+	 * AGP aperture is also used for page table in system
+	 * memory (mainly for APU).
+	 *
+	 */
 	u64			agp_size;
 	u64			agp_start;
 	u64			agp_end;
+	/* GART aperture start and end in MC address space
+	 * Driver find a hole in the MC address space
+	 * to place GART by setting VM_CONTEXT0_PAGE_TABLE_START/END_ADDR
+	 * registers
+	 * Under VMID0, logical address inside GART aperture will
+	 * be translated through gpuvm gart page table to access
+	 * paged system memory
+	 */
 	u64			gart_size;
 	u64			gart_start;
 	u64			gart_end;
+	/* Frame buffer aperture of this GPU device. Different from
+	 * fb_start (see below), this only covers the local GPU device.
+	 * Driver get fb_start from MC_VM_FB_LOCATION_BASE (set by vbios)
+	 * and calculate vram_start of this local device by adding an
+	 * offset inside the XGMI hive.
+	 * Under VMID0, logical address == MC address
+	 */
 	u64			vram_start;
 	u64			vram_end;
 	/* FB region , it's same as local vram region in single GPU, in XGMI

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

@@ -311,7 +311,7 @@ void amdgpu_irq_fini(struct amdgpu_device *adev)
 		drm_irq_uninstall(adev->ddev);
 		adev->irq.installed = false;
 		if (adev->irq.msi_enabled)
-			pci_disable_msi(adev->pdev);
+			pci_free_irq_vectors(adev->pdev);
 		if (!amdgpu_device_has_dc_support(adev))
 			flush_work(&adev->hotplug_work);
 	}

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

@@ -220,7 +220,7 @@ static struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev,
  * As their names indicate, inject operation will write the
  * value to the address.
  *
- * Second member: struct ras_debug_if::op.
+ * The second member: struct ras_debug_if::op.
  * It has three kinds of operations.
  *
  * - 0: disable RAS on the block. Take ::head as its data.
@@ -228,14 +228,20 @@ static struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev,
  * - 2: inject errors on the block. Take ::inject as its data.
  *
  * How to use the interface?
- * programs:
- * copy the struct ras_debug_if in your codes and initialize it.
- * write the struct to the control node.
+ *
+ * Programs
+ *
+ * Copy the struct ras_debug_if in your codes and initialize it.
+ * Write the struct to the control node.
+ *
+ * Shells
  *
  * .. code-block:: bash
  *
  *	echo op block [error [sub_block address value]] > .../ras/ras_ctrl
  *
+ * Parameters:
+ *
  * op: disable, enable, inject
  *	disable: only block is needed
  *	enable: block and error are needed
@@ -265,8 +271,10 @@ static struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev,
  * /sys/class/drm/card[0/1/2...]/device/ras/[gfx/sdma/...]_err_count
  *
  * .. note::
- *	Operation is only allowed on blocks which are supported.
+ *	Operations are only allowed on blocks which are supported.
  *	Please check ras mask at /sys/module/amdgpu/parameters/ras_mask
+ *	to see which blocks support RAS on a particular asic.
+ *
  */
 static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, const char __user *buf,
 		size_t size, loff_t *pos)
@@ -322,7 +330,7 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, const char __user *
  * DOC: AMDGPU RAS debugfs EEPROM table reset interface
  *
  * Some boards contain an EEPROM which is used to persistently store a list of
- * bad pages containing ECC errors detected in vram.  This interface provides
+ * bad pages which experiences ECC errors in vram.  This interface provides
  * a way to reset the EEPROM, e.g., after testing error injection.
  *
  * Usage:
@@ -362,7 +370,7 @@ static const struct file_operations amdgpu_ras_debugfs_eeprom_ops = {
 /**
  * DOC: AMDGPU RAS sysfs Error Count Interface
  *
- * It allows user to read the error count for each IP block on the gpu through
+ * It allows the user to read the error count for each IP block on the gpu through
  * /sys/class/drm/card[0/1/2...]/device/ras/[gfx/sdma/...]_err_count
  *
  * It outputs the multiple lines which report the uncorrected (ue) and corrected
@@ -1027,6 +1035,24 @@ static int amdgpu_ras_sysfs_remove_all(struct amdgpu_device *adev)
 }
 /* sysfs end */
 
+/**
+ * DOC: AMDGPU RAS Reboot Behavior for Unrecoverable Errors
+ *
+ * Normally when there is an uncorrectable error, the driver will reset
+ * the GPU to recover.  However, in the event of an unrecoverable error,
+ * the driver provides an interface to reboot the system automatically
+ * in that event.
+ *
+ * The following file in debugfs provides that interface:
+ * /sys/kernel/debug/dri/[0/1/2...]/ras/auto_reboot
+ *
+ * Usage:
+ *
+ * .. code-block:: bash
+ *
+ *	echo true > .../ras/auto_reboot
+ *
+ */
 /* debugfs begin */
 static void amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev)
 {

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

@@ -138,6 +138,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
 		}
 
 		dma_fence_put(fence);
+		fence = NULL;
 
 		r = amdgpu_bo_kmap(vram_obj, &vram_map);
 		if (r) {
@@ -183,6 +184,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
 		}
 
 		dma_fence_put(fence);
+		fence = NULL;
 
 		r = amdgpu_bo_kmap(gtt_obj[i], &gtt_map);
 		if (r) {

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

@@ -1906,9 +1906,6 @@ void amdgpu_ttm_late_init(struct amdgpu_device *adev)
 	void *stolen_vga_buf;
 	/* return the VGA stolen memory (if any) back to VRAM */
 	amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, &stolen_vga_buf);
-
-	/* return the IP Discovery TMR memory back to VRAM */
-	amdgpu_bo_free_kernel(&adev->discovery_memory, NULL, NULL);
 }
 
 /**
@@ -1921,7 +1918,10 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev)
 
 	amdgpu_ttm_debugfs_fini(adev);
 	amdgpu_ttm_training_reserve_vram_fini(adev);
+	/* return the IP Discovery TMR memory back to VRAM */
+	amdgpu_bo_free_kernel(&adev->discovery_memory, NULL, NULL);
 	amdgpu_ttm_fw_reserve_vram_fini(adev);
+
 	if (adev->mman.aper_base_kaddr)
 		iounmap(adev->mman.aper_base_kaddr);
 	adev->mman.aper_base_kaddr = NULL;

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

@@ -1418,6 +1418,9 @@ static int amdgpu_vm_update_ptes(struct amdgpu_vm_update_params *params,
 		uint64_t incr, entry_end, pe_start;
 		struct amdgpu_bo *pt;
 
+		/* make sure that the page tables covering the address range are
+		 * actually allocated
+		 */
 		r = amdgpu_vm_alloc_pts(params->adev, params->vm, &cursor,
 					params->direct);
 		if (r)
@@ -1491,7 +1494,12 @@ static int amdgpu_vm_update_ptes(struct amdgpu_vm_update_params *params,
 		} while (frag_start < entry_end);
 
 		if (amdgpu_vm_pt_descendant(adev, &cursor)) {
-			/* Free all child entries */
+			/* Free all child entries.
+			 * Update the tables with the flags and addresses and free up subsequent
+			 * tables in the case of huge pages or freed up areas.
+			 * This is the maximum you can free, because all other page tables are not
+			 * completely covered by the range and so potentially still in use.
+			 */
 			while (cursor.pfn < frag_start) {
 				amdgpu_vm_free_pts(adev, params->vm, &cursor);
 				amdgpu_vm_pt_next(adev, &cursor);

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

@@ -274,22 +274,55 @@ int amdgpu_xgmi_set_pstate(struct amdgpu_device *adev, int pstate)
 {
 	int ret = 0;
 	struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev, 0);
+	struct amdgpu_device *tmp_adev;
+	bool update_hive_pstate = true;
+	bool is_high_pstate = pstate && adev->asic_type == CHIP_VEGA20;
 
 	if (!hive)
 		return 0;
 
-	if (hive->pstate == pstate)
-		return 0;
+	mutex_lock(&hive->hive_lock);
+
+	if (hive->pstate == pstate) {
+		adev->pstate = is_high_pstate ? pstate : adev->pstate;
+		goto out;
+	}
 
 	dev_dbg(adev->dev, "Set xgmi pstate %d.\n", pstate);
 
 	if (is_support_sw_smu_xgmi(adev))
 		ret = smu_set_xgmi_pstate(&adev->smu, pstate);
-	if (ret)
+	else if (adev->powerplay.pp_funcs &&
+		 adev->powerplay.pp_funcs->set_xgmi_pstate)
+		ret = adev->powerplay.pp_funcs->set_xgmi_pstate(adev->powerplay.pp_handle,
+								pstate);
+
+	if (ret) {
 		dev_err(adev->dev,
 			"XGMI: Set pstate failure on device %llx, hive %llx, ret %d",
 			adev->gmc.xgmi.node_id,
 			adev->gmc.xgmi.hive_id, ret);
+		goto out;
+	}
+
+	/* Update device pstate */
+	adev->pstate = pstate;
+
+	/*
+	 * Update the hive pstate only all devices of the hive
+	 * are in the same pstate
+	 */
+	list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+		if (tmp_adev->pstate != adev->pstate) {
+			update_hive_pstate = false;
+			break;
+		}
+	}
+	if (update_hive_pstate || is_high_pstate)
+		hive->pstate = pstate;
+
+out:
+	mutex_unlock(&hive->hive_lock);
 
 	return ret;
 }
@@ -364,6 +397,9 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 		goto exit;
 	}
 
+	/* Set default device pstate */
+	adev->pstate = -1;
+
 	top_info = &adev->psp.xgmi_context.top_info;
 
 	list_add_tail(&adev->gmc.xgmi.head, &hive->device_list);

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

@@ -2738,7 +2738,10 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev)
 	 * And it's needed by gfxoff feature.
 	 */
 	if (adev->gfx.rlc.is_rlc_v2_1) {
-		gfx_v9_1_init_rlc_save_restore_list(adev);
+		if (adev->asic_type == CHIP_VEGA12 ||
+		    (adev->asic_type == CHIP_RAVEN &&
+		     adev->rev_id >= 8))
+			gfx_v9_1_init_rlc_save_restore_list(adev);
 		gfx_v9_0_enable_save_restore_machine(adev);
 	}
 
@@ -3889,9 +3892,22 @@ static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 	uint64_t clock;
 
 	mutex_lock(&adev->gfx.gpu_clock_mutex);
-	WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
-	clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) |
-		((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
+	if (adev->asic_type == CHIP_VEGA10 && amdgpu_sriov_runtime(adev)) {
+		uint32_t tmp, lsb, msb, i = 0;
+		do {
+			if (i != 0)
+				udelay(1);
+			tmp = RREG32_SOC15(GC, 0, mmRLC_REFCLOCK_TIMESTAMP_MSB);
+			lsb = RREG32_SOC15(GC, 0, mmRLC_REFCLOCK_TIMESTAMP_LSB);
+			msb = RREG32_SOC15(GC, 0, mmRLC_REFCLOCK_TIMESTAMP_MSB);
+			i++;
+		} while (unlikely(tmp != msb) && (i < adev->usec_timeout));
+		clock = (uint64_t)lsb | ((uint64_t)msb << 32ULL);
+	} else {
+		WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
+		clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) |
+			((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
+	}
 	mutex_unlock(&adev->gfx.gpu_clock_mutex);
 	return clock;
 }

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

@@ -539,6 +539,16 @@ static bool nv_need_reset_on_init(struct amdgpu_device *adev)
 	return false;
 }
 
+static uint64_t nv_get_pcie_replay_count(struct amdgpu_device *adev)
+{
+
+	/* TODO
+	 * dummy implement for pcie_replay_count sysfs interface
+	 * */
+
+	return 0;
+}
+
 static void nv_init_doorbell_index(struct amdgpu_device *adev)
 {
 	adev->doorbell_index.kiq = AMDGPU_NAVI10_DOORBELL_KIQ;
@@ -586,6 +596,7 @@ static const struct amdgpu_asic_funcs nv_asic_funcs =
 	.need_full_reset = &nv_need_full_reset,
 	.get_pcie_usage = &nv_get_pcie_usage,
 	.need_reset_on_init = &nv_need_reset_on_init,
+	.get_pcie_replay_count = &nv_get_pcie_replay_count,
 };
 
 static int nv_common_early_init(void *handle)

drivers/gpu/drm/amd/amdgpu/soc15.c

@@ -1145,7 +1145,9 @@ static int soc15_common_early_init(void *handle)
 				AMD_CG_SUPPORT_SDMA_LS |
 				AMD_CG_SUPPORT_VCN_MGCG;
 
-			adev->pg_flags = AMD_PG_SUPPORT_SDMA | AMD_PG_SUPPORT_VCN;
+			adev->pg_flags = AMD_PG_SUPPORT_SDMA |
+				AMD_PG_SUPPORT_VCN |
+				AMD_PG_SUPPORT_VCN_DPG;
 		} else if (adev->pdev->device == 0x15d8) {
 			adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG |
 				AMD_CG_SUPPORT_GFX_MGLS |
@@ -1188,7 +1190,9 @@ static int soc15_common_early_init(void *handle)
 				AMD_CG_SUPPORT_SDMA_LS |
 				AMD_CG_SUPPORT_VCN_MGCG;
 
-			adev->pg_flags = AMD_PG_SUPPORT_SDMA | AMD_PG_SUPPORT_VCN;
+			adev->pg_flags = AMD_PG_SUPPORT_SDMA |
+				AMD_PG_SUPPORT_VCN |
+				AMD_PG_SUPPORT_VCN_DPG;
 		}
 		break;
 	case CHIP_ARCTURUS:

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c

@@ -4239,8 +4239,8 @@ enum drm_mode_status amdgpu_dm_connector_mode_valid(struct drm_connector *connec
 		result = MODE_OK;
 	else
 		DRM_DEBUG_KMS("Mode %dx%d (clk %d) failed DC validation with error %d\n",
-			      mode->vdisplay,
 			      mode->hdisplay,
+			      mode->vdisplay,
 			      mode->clock,
 			      dc_result);
 

drivers/gpu/drm/amd/display/dc/core/dc_link.c

@@ -3027,15 +3027,6 @@ void core_link_enable_stream(
 					CONTROLLER_DP_TEST_PATTERN_VIDEOMODE,
 					COLOR_DEPTH_UNDEFINED);
 
-		/* This second call is needed to reconfigure the DIG
-		 * as a workaround for the incorrect value being applied
-		 * from transmitter control.
-		 */
-		if (!dc_is_virtual_signal(pipe_ctx->stream->signal))
-			stream->link->link_enc->funcs->setup(
-				stream->link->link_enc,
-				pipe_ctx->stream->signal);
-
 #ifdef CONFIG_DRM_AMD_DC_DSC_SUPPORT
 		if (pipe_ctx->stream->timing.flags.DSC) {
 			if (dc_is_dp_signal(pipe_ctx->stream->signal) ||

drivers/gpu/drm/amd/include/kgd_pp_interface.h

@@ -220,6 +220,9 @@ enum pp_df_cstate {
 		((group) << PP_GROUP_SHIFT | (block) << PP_BLOCK_SHIFT | \
 		(support) << PP_STATE_SUPPORT_SHIFT | (state) << PP_STATE_SHIFT)
 
+#define XGMI_MODE_PSTATE_D3 0
+#define XGMI_MODE_PSTATE_D0 1
+
 struct seq_file;
 enum amd_pp_clock_type;
 struct amd_pp_simple_clock_info;
@@ -318,6 +321,7 @@ struct amd_pm_funcs {
 	int (*set_ppfeature_status)(void *handle, uint64_t ppfeature_masks);
 	int (*asic_reset_mode_2)(void *handle);
 	int (*set_df_cstate)(void *handle, enum pp_df_cstate state);
+	int (*set_xgmi_pstate)(void *handle, uint32_t pstate);
 };
 
 #endif

drivers/gpu/drm/amd/powerplay/amd_powerplay.c

@@ -969,6 +969,14 @@ static int pp_dpm_switch_power_profile(void *handle,
 		workload = hwmgr->workload_setting[index];
 	}
 
+	if (type == PP_SMC_POWER_PROFILE_COMPUTE &&
+		hwmgr->hwmgr_func->disable_power_features_for_compute_performance) {
+			if (hwmgr->hwmgr_func->disable_power_features_for_compute_performance(hwmgr, en)) {
+				mutex_unlock(&hwmgr->smu_lock);
+				return -EINVAL;
+			}
+	}
+
 	if (hwmgr->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL)
 		hwmgr->hwmgr_func->set_power_profile_mode(hwmgr, &workload, 0);
 	mutex_unlock(&hwmgr->smu_lock);
@@ -1566,6 +1574,23 @@ static int pp_set_df_cstate(void *handle, enum pp_df_cstate state)
 	return 0;
 }
 
+static int pp_set_xgmi_pstate(void *handle, uint32_t pstate)
+{
+	struct pp_hwmgr *hwmgr = handle;
+
+	if (!hwmgr)
+		return -EINVAL;
+
+	if (!hwmgr->pm_en || !hwmgr->hwmgr_func->set_xgmi_pstate)
+		return 0;
+
+	mutex_lock(&hwmgr->smu_lock);
+	hwmgr->hwmgr_func->set_xgmi_pstate(hwmgr, pstate);
+	mutex_unlock(&hwmgr->smu_lock);
+
+	return 0;
+}
+
 static const struct amd_pm_funcs pp_dpm_funcs = {
 	.load_firmware = pp_dpm_load_fw,
 	.wait_for_fw_loading_complete = pp_dpm_fw_loading_complete,
@@ -1625,4 +1650,5 @@ static const struct amd_pm_funcs pp_dpm_funcs = {
 	.asic_reset_mode_2 = pp_asic_reset_mode_2,
 	.smu_i2c_bus_access = pp_smu_i2c_bus_access,
 	.set_df_cstate = pp_set_df_cstate,
+	.set_xgmi_pstate = pp_set_xgmi_pstate,
 };

drivers/gpu/drm/amd/powerplay/amdgpu_smu.c

@@ -383,14 +383,25 @@ bool smu_clk_dpm_is_enabled(struct smu_context *smu, enum smu_clk_type clk_type)
 	return true;
 }
 
-
+/**
+ * smu_dpm_set_power_gate - power gate/ungate the specific IP block
+ *
+ * @smu:        smu_context pointer
+ * @block_type: the IP block to power gate/ungate
+ * @gate:       to power gate if true, ungate otherwise
+ *
+ * This API uses no smu->mutex lock protection due to:
+ * 1. It is either called by other IP block(gfx/sdma/vcn/uvd/vce).
+ *    This is guarded to be race condition free by the caller.
+ * 2. Or get called on user setting request of power_dpm_force_performance_level.
+ *    Under this case, the smu->mutex lock protection is already enforced on
+ *    the parent API smu_force_performance_level of the call path.
+ */
 int smu_dpm_set_power_gate(struct smu_context *smu, uint32_t block_type,
 			   bool gate)
 {
 	int ret = 0;
 
-	mutex_lock(&smu->mutex);
-
 	switch (block_type) {
 	case AMD_IP_BLOCK_TYPE_UVD:
 		ret = smu_dpm_set_uvd_enable(smu, gate);
@@ -408,8 +419,6 @@ int smu_dpm_set_power_gate(struct smu_context *smu, uint32_t block_type,
 		break;
 	}
 
-	mutex_unlock(&smu->mutex);
-
 	return ret;
 }
 
@@ -526,7 +535,7 @@ bool is_support_sw_smu(struct amdgpu_device *adev)
 
 bool is_support_sw_smu_xgmi(struct amdgpu_device *adev)
 {
-	if (amdgpu_dpm != 1)
+	if (!is_support_sw_smu(adev))
 		return false;
 
 	if (adev->asic_type == CHIP_VEGA20)
@@ -705,6 +714,9 @@ static int smu_set_funcs(struct amdgpu_device *adev)
 {
 	struct smu_context *smu = &adev->smu;
 
+	if (adev->pm.pp_feature & PP_OVERDRIVE_MASK)
+		smu->od_enabled = true;
+
 	switch (adev->asic_type) {
 	case CHIP_VEGA20:
 		vega20_set_ppt_funcs(smu);
@@ -716,6 +728,8 @@ static int smu_set_funcs(struct amdgpu_device *adev)
 		break;
 	case CHIP_ARCTURUS:
 		arcturus_set_ppt_funcs(smu);
+		/* OD is not supported on Arcturus */
+		smu->od_enabled =false;
 		break;
 	case CHIP_RENOIR:
 		renoir_set_ppt_funcs(smu);
@@ -724,9 +738,6 @@ static int smu_set_funcs(struct amdgpu_device *adev)
 		return -EINVAL;
 	}
 
-	if (adev->pm.pp_feature & PP_OVERDRIVE_MASK)
-		smu->od_enabled = true;
-
 	return 0;
 }
 

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c

@@ -3689,6 +3689,13 @@ static int vega10_set_power_state_tasks(struct pp_hwmgr *hwmgr,
 	PP_ASSERT_WITH_CODE(!result,
 			"Failed to upload PPtable!", return result);
 
+	/*
+	 * If a custom pp table is loaded, set DPMTABLE_OD_UPDATE_VDDC flag.
+	 * That effectively disables AVFS feature.
+	 */
+	if(hwmgr->hardcode_pp_table != NULL)
+		data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_VDDC;
+
 	vega10_update_avfs(hwmgr);
 
 	/*
@@ -5263,6 +5270,59 @@ static int vega10_get_performance_level(struct pp_hwmgr *hwmgr, const struct pp_
 	return 0;
 }
 
+static int vega10_disable_power_features_for_compute_performance(struct pp_hwmgr *hwmgr, bool disable)
+{
+	struct vega10_hwmgr *data = hwmgr->backend;
+	uint32_t feature_mask = 0;
+
+	if (disable) {
+		feature_mask |= data->smu_features[GNLD_ULV].enabled ?
+			data->smu_features[GNLD_ULV].smu_feature_bitmap : 0;
+		feature_mask |= data->smu_features[GNLD_DS_GFXCLK].enabled ?
+			data->smu_features[GNLD_DS_GFXCLK].smu_feature_bitmap : 0;
+		feature_mask |= data->smu_features[GNLD_DS_SOCCLK].enabled ?
+			data->smu_features[GNLD_DS_SOCCLK].smu_feature_bitmap : 0;
+		feature_mask |= data->smu_features[GNLD_DS_LCLK].enabled ?
+			data->smu_features[GNLD_DS_LCLK].smu_feature_bitmap : 0;
+		feature_mask |= data->smu_features[GNLD_DS_DCEFCLK].enabled ?
+			data->smu_features[GNLD_DS_DCEFCLK].smu_feature_bitmap : 0;
+	} else {
+		feature_mask |= (!data->smu_features[GNLD_ULV].enabled) ?
+			data->smu_features[GNLD_ULV].smu_feature_bitmap : 0;
+		feature_mask |= (!data->smu_features[GNLD_DS_GFXCLK].enabled) ?
+			data->smu_features[GNLD_DS_GFXCLK].smu_feature_bitmap : 0;
+		feature_mask |= (!data->smu_features[GNLD_DS_SOCCLK].enabled) ?
+			data->smu_features[GNLD_DS_SOCCLK].smu_feature_bitmap : 0;
+		feature_mask |= (!data->smu_features[GNLD_DS_LCLK].enabled) ?
+			data->smu_features[GNLD_DS_LCLK].smu_feature_bitmap : 0;
+		feature_mask |= (!data->smu_features[GNLD_DS_DCEFCLK].enabled) ?
+			data->smu_features[GNLD_DS_DCEFCLK].smu_feature_bitmap : 0;
+	}
+
+	if (feature_mask)
+		PP_ASSERT_WITH_CODE(!vega10_enable_smc_features(hwmgr,
+				!disable, feature_mask),
+				"enable/disable power features for compute performance Failed!",
+				return -EINVAL);
+
+	if (disable) {
+		data->smu_features[GNLD_ULV].enabled = false;
+		data->smu_features[GNLD_DS_GFXCLK].enabled = false;
+		data->smu_features[GNLD_DS_SOCCLK].enabled = false;
+		data->smu_features[GNLD_DS_LCLK].enabled = false;
+		data->smu_features[GNLD_DS_DCEFCLK].enabled = false;
+	} else {
+		data->smu_features[GNLD_ULV].enabled = true;
+		data->smu_features[GNLD_DS_GFXCLK].enabled = true;
+		data->smu_features[GNLD_DS_SOCCLK].enabled = true;
+		data->smu_features[GNLD_DS_LCLK].enabled = true;
+		data->smu_features[GNLD_DS_DCEFCLK].enabled = true;
+	}
+
+	return 0;
+
+}
+
 static const struct pp_hwmgr_func vega10_hwmgr_funcs = {
 	.backend_init = vega10_hwmgr_backend_init,
 	.backend_fini = vega10_hwmgr_backend_fini,
@@ -5330,6 +5390,8 @@ static const struct pp_hwmgr_func vega10_hwmgr_funcs = {
 	.get_ppfeature_status = vega10_get_ppfeature_status,
 	.set_ppfeature_status = vega10_set_ppfeature_status,
 	.set_mp1_state = vega10_set_mp1_state,
+	.disable_power_features_for_compute_performance =
+			vega10_disable_power_features_for_compute_performance,
 };
 
 int vega10_hwmgr_init(struct pp_hwmgr *hwmgr)

drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.c

@@ -4176,6 +4176,20 @@ static int vega20_set_df_cstate(struct pp_hwmgr *hwmgr,
 	return ret;
 }
 
+static int vega20_set_xgmi_pstate(struct pp_hwmgr *hwmgr,
+				  uint32_t pstate)
+{
+	int ret;
+
+	ret = smum_send_msg_to_smc_with_parameter(hwmgr,
+						  PPSMC_MSG_SetXgmiMode,
+						  pstate ? XGMI_MODE_PSTATE_D0 : XGMI_MODE_PSTATE_D3);
+	if (ret)
+		pr_err("SetXgmiPstate failed!\n");
+
+	return ret;
+}
+
 static const struct pp_hwmgr_func vega20_hwmgr_funcs = {
 	/* init/fini related */
 	.backend_init = vega20_hwmgr_backend_init,
@@ -4245,6 +4259,7 @@ static const struct pp_hwmgr_func vega20_hwmgr_funcs = {
 	.set_mp1_state = vega20_set_mp1_state,
 	.smu_i2c_bus_access = vega20_smu_i2c_bus_access,
 	.set_df_cstate = vega20_set_df_cstate,
+	.set_xgmi_pstate = vega20_set_xgmi_pstate,
 };
 
 int vega20_hwmgr_init(struct pp_hwmgr *hwmgr)

drivers/gpu/drm/amd/powerplay/inc/hwmgr.h

@@ -356,6 +356,9 @@ struct pp_hwmgr_func {
 	int (*asic_reset)(struct pp_hwmgr *hwmgr, enum SMU_ASIC_RESET_MODE mode);
 	int (*smu_i2c_bus_access)(struct pp_hwmgr *hwmgr, bool aquire);
 	int (*set_df_cstate)(struct pp_hwmgr *hwmgr, enum pp_df_cstate state);
+	int (*set_xgmi_pstate)(struct pp_hwmgr *hwmgr, uint32_t pstate);
+	int (*disable_power_features_for_compute_performance)(struct pp_hwmgr *hwmgr,
+					bool disable);
 };
 
 struct pp_table_func {

drivers/gpu/drm/amd/powerplay/inc/smu11_driver_if_arcturus.h

@@ -159,7 +159,7 @@
 //FIXME need updating
 // Debug Overrides Bitmask
 #define DPM_OVERRIDE_DISABLE_UCLK_PID               0x00000001
-#define DPM_OVERRIDE_ENABLE_VOLT_LINK_VCN_FCLK      0x00000002
+#define DPM_OVERRIDE_DISABLE_VOLT_LINK_VCN_FCLK     0x00000002
 
 // I2C Config Bit Defines
 #define I2C_CONTROLLER_ENABLED           1

drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h

@@ -27,7 +27,7 @@
 
 #define SMU11_DRIVER_IF_VERSION_INV 0xFFFFFFFF
 #define SMU11_DRIVER_IF_VERSION_VG20 0x13
-#define SMU11_DRIVER_IF_VERSION_ARCT 0x0F
+#define SMU11_DRIVER_IF_VERSION_ARCT 0x10
 #define SMU11_DRIVER_IF_VERSION_NV10 0x33
 #define SMU11_DRIVER_IF_VERSION_NV14 0x34
 

drivers/gpu/drm/amd/powerplay/renoir_ppt.c

@@ -180,11 +180,13 @@ static int renoir_print_clk_levels(struct smu_context *smu,
 	int i, size = 0, ret = 0;
 	uint32_t cur_value = 0, value = 0, count = 0, min = 0, max = 0;
 	DpmClocks_t *clk_table = smu->smu_table.clocks_table;
-	SmuMetrics_t metrics = {0};
+	SmuMetrics_t metrics;
 
 	if (!clk_table || clk_type >= SMU_CLK_COUNT)
 		return -EINVAL;
 
+	memset(&metrics, 0, sizeof(metrics));
+
 	ret = smu_update_table(smu, SMU_TABLE_SMU_METRICS, 0,
 			       (void *)&metrics, false);
 	if (ret)

drivers/gpu/drm/amd/powerplay/smu_v11_0.c

@@ -368,6 +368,7 @@ int smu_v11_0_setup_pptable(struct smu_context *smu)
 	version_major = le16_to_cpu(hdr->header.header_version_major);
 	version_minor = le16_to_cpu(hdr->header.header_version_minor);
 	if (version_major == 2 && smu->smu_table.boot_values.pp_table_id > 0) {
+		pr_info("use driver provided pptable %d\n", smu->smu_table.boot_values.pp_table_id);
 		switch (version_minor) {
 		case 0:
 			ret = smu_v11_0_set_pptable_v2_0(smu, &table, &size);
@@ -384,6 +385,7 @@ int smu_v11_0_setup_pptable(struct smu_context *smu)
 			return ret;
 
 	} else {
+		pr_info("use vbios provided pptable\n");
 		index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
 						    powerplayinfo);
 
@@ -1463,16 +1465,13 @@ int smu_v11_0_set_fan_speed_rpm(struct smu_context *smu,
 	return ret;
 }
 
-#define XGMI_STATE_D0 1
-#define XGMI_STATE_D3 0
-
 int smu_v11_0_set_xgmi_pstate(struct smu_context *smu,
 				     uint32_t pstate)
 {
 	int ret = 0;
 	ret = smu_send_smc_msg_with_param(smu,
 					  SMU_MSG_SetXgmiMode,
-					  pstate ? XGMI_STATE_D0 : XGMI_STATE_D3);
+					  pstate ? XGMI_MODE_PSTATE_D0 : XGMI_MODE_PSTATE_D3);
 	return ret;
 }
 

drivers/gpu/drm/scheduler/sched_entity.c

@@ -23,6 +23,7 @@
 
 #include <linux/kthread.h>
 #include <linux/slab.h>
+#include <linux/completion.h>
 
 #include <drm/drm_print.h>
 #include <drm/gpu_scheduler.h>
@@ -68,6 +69,8 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
 	if (!entity->rq_list)
 		return -ENOMEM;
 
+	init_completion(&entity->entity_idle);
+
 	for (i = 0; i < num_rq_list; ++i)
 		entity->rq_list[i] = rq_list[i];
 
@@ -286,11 +289,12 @@ void drm_sched_entity_fini(struct drm_sched_entity *entity)
 	 */
 	if (spsc_queue_count(&entity->job_queue)) {
 		if (sched) {
-			/* Park the kernel for a moment to make sure it isn't processing
-			 * our enity.
+			/*
+			 * Wait for thread to idle to make sure it isn't processing
+			 * this entity.
 			 */
-			kthread_park(sched->thread);
-			kthread_unpark(sched->thread);
+			wait_for_completion(&entity->entity_idle);
+
 		}
 		if (entity->dependency) {
 			dma_fence_remove_callback(entity->dependency,

drivers/gpu/drm/scheduler/sched_main.c

@@ -47,6 +47,7 @@
 #include <linux/kthread.h>
 #include <linux/wait.h>
 #include <linux/sched.h>
+#include <linux/completion.h>
 #include <uapi/linux/sched/types.h>
 
 #include <drm/drm_print.h>
@@ -134,6 +135,7 @@ drm_sched_rq_select_entity(struct drm_sched_rq *rq)
 		list_for_each_entry_continue(entity, &rq->entities, list) {
 			if (drm_sched_entity_is_ready(entity)) {
 				rq->current_entity = entity;
+				reinit_completion(&entity->entity_idle);
 				spin_unlock(&rq->lock);
 				return entity;
 			}
@@ -144,6 +146,7 @@ drm_sched_rq_select_entity(struct drm_sched_rq *rq)
 
 		if (drm_sched_entity_is_ready(entity)) {
 			rq->current_entity = entity;
+			reinit_completion(&entity->entity_idle);
 			spin_unlock(&rq->lock);
 			return entity;
 		}
@@ -496,8 +499,10 @@ void drm_sched_resubmit_jobs(struct drm_gpu_scheduler *sched)
 		fence = sched->ops->run_job(s_job);
 
 		if (IS_ERR_OR_NULL(fence)) {
+			if (IS_ERR(fence))
+				dma_fence_set_error(&s_fence->finished, PTR_ERR(fence));
+
 			s_job->s_fence->parent = NULL;
-			dma_fence_set_error(&s_fence->finished, PTR_ERR(fence));
 		} else {
 			s_job->s_fence->parent = fence;
 		}
@@ -645,9 +650,13 @@ drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched)
 	struct drm_sched_job *job;
 	unsigned long flags;
 
-	/* Don't destroy jobs while the timeout worker is running */
-	if (sched->timeout != MAX_SCHEDULE_TIMEOUT &&
-	    !cancel_delayed_work(&sched->work_tdr))
+	/*
+	 * Don't destroy jobs while the timeout worker is running  OR thread
+	 * is being parked and hence assumed to not touch ring_mirror_list
+	 */
+	if ((sched->timeout != MAX_SCHEDULE_TIMEOUT &&
+	    !cancel_delayed_work(&sched->work_tdr)) ||
+	    __kthread_should_park(sched->thread))
 		return NULL;
 
 	spin_lock_irqsave(&sched->job_list_lock, flags);
@@ -724,6 +733,9 @@ static int drm_sched_main(void *param)
 			continue;
 
 		sched_job = drm_sched_entity_pop_job(entity);
+
+		complete(&entity->entity_idle);
+
 		if (!sched_job)
 			continue;
 
@@ -746,8 +758,9 @@ static int drm_sched_main(void *param)
 					  r);
 			dma_fence_put(fence);
 		} else {
+			if (IS_ERR(fence))
+				dma_fence_set_error(&s_fence->finished, PTR_ERR(fence));
 
-			dma_fence_set_error(&s_fence->finished, PTR_ERR(fence));
 			drm_sched_process_job(NULL, &sched_job->cb);
 		}
 

include/drm/gpu_scheduler.h

@@ -26,6 +26,7 @@
 
 #include <drm/spsc_queue.h>
 #include <linux/dma-fence.h>
+#include <linux/completion.h>
 
 #define MAX_WAIT_SCHED_ENTITY_Q_EMPTY msecs_to_jiffies(1000)
 
@@ -71,6 +72,7 @@ enum drm_sched_priority {
  * @last_scheduled: points to the finished fence of the last scheduled job.
  * @last_user: last group leader pushing a job into the entity.
  * @stopped: Marks the enity as removed from rq and destined for termination.
+ * @entity_idle: Signals when enityt is not in use
  *
  * Entities will emit jobs in order to their corresponding hardware
  * ring, and the scheduler will alternate between entities based on
@@ -94,6 +96,7 @@ struct drm_sched_entity {
 	struct dma_fence                *last_scheduled;
 	struct task_struct		*last_user;
 	bool 				stopped;
+	struct completion		entity_idle;
 };
 
 /**