Merge tag 'drm-fixes-2024-08-10' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Dave Airlie:
 "Weekly regular fixes, mostly amdgpu with i915/xe having a few each,
  and then some misc bits across the board, seems about right for rc3
  time.

  client:
   - fix null ptr deref

  bridge:
   - connector: fix double free

  atomic:
   - fix async flip update

  panel:
   - document panel

  omap:
   - add config dependency

  tests:
   - fix gem shmem test

  drm buddy:
   - Add start address to trim function

  amdgpu:
   - DMCUB fix
   - Fix DET programming on some DCNs
   - DCC fixes
   - DCN 4.0.1 fixes
   - SMU 14.0.x update
   - MMHUB fix
   - DCN 3.1.4 fix
   - GC 12.0 fixes
   - Fix soft recovery error propogation
   - SDMA 7.0 fixes
   - DSC fix

  xe:
   - Fix off-by-one when processing RTP rules
   - Use dma_fence_chain_free in chain fence unused as a sync
   - Fix PL1 disable flow in xe_hwmon_power_max_write
   - Take ref to VM in delayed dump snapshot

  i915:
   - correct dual pps handling for MTL_PCH+ [display]
   - Adjust vma offset for framebuffer mmap offset [gem]
   - Fix Virtual Memory mapping boundaries calculation [gem]
   - Allow evicting to use the requested placement
   - Attempt to get pages without eviction first"

* tag 'drm-fixes-2024-08-10' of https://gitlab.freedesktop.org/drm/kernel: (31 commits)
  drm/xe: Take ref to VM in delayed snapshot
  drm/xe/hwmon: Fix PL1 disable flow in xe_hwmon_power_max_write
  drm/xe: Use dma_fence_chain_free in chain fence unused as a sync
  drm/xe/rtp: Fix off-by-one when processing rules
  drm/amdgpu: Add DCC GFX12 flag to enable address alignment
  drm/amdgpu: correct sdma7 max dw
  drm/amdgpu: Add address alignment support to DCC buffers
  drm/amd/display: Skip Recompute DSC Params if no Stream on Link
  drm/amdgpu: change non-dcc buffer copy configuration
  drm/amdgpu: Forward soft recovery errors to userspace
  drm/amdgpu: add golden setting for gc v12
  drm/buddy: Add start address support to trim function
  drm/amd/display: Add missing program DET segment call to pipe init
  drm/amd/display: Add missing DCN314 to the DML Makefile
  drm/amdgpu: force to use legacy inv in mmhub
  drm/amd/pm: update powerplay structure on smu v14.0.2/3
  drm/amd/display: Add missing mcache registers
  drm/amd/display: Add dcc propagation value
  drm/amd/display: Add missing DET segments programming
  drm/amd/display: Replace dm_execute_dmub_cmd with dc_wake_and_execute_dmub_cmd
  ...

Author: torvalds

Documentation/devicetree/bindings/display/panel/samsung,atna33xc20.yaml

@@ -17,10 +17,13 @@ properties:
     oneOf:
       # Samsung 13.3" FHD (1920x1080 pixels) eDP AMOLED panel
       - const: samsung,atna33xc20
-      # Samsung 14.5" WQXGA+ (2880x1800 pixels) eDP AMOLED panel
       - items:
-          - const: samsung,atna45af01
-          - const: samsung,atna33xc20
+        - enum:
+          # Samsung 14.5" WQXGA+ (2880x1800 pixels) eDP AMOLED panel
+          - samsung,atna45af01
+          # Samsung 14.5" 3K (2944x1840 pixels) eDP AMOLED panel
+          - samsung,atna45dc02
+        - const: samsung,atna33xc20
 
   enable-gpios: true
   port: true

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

@@ -156,6 +156,8 @@ struct amdgpu_gmc_funcs {
 				      uint64_t addr, uint64_t *flags);
 	/* get the amount of memory used by the vbios for pre-OS console */
 	unsigned int (*get_vbios_fb_size)(struct amdgpu_device *adev);
+	/* get the DCC buffer alignment */
+	unsigned int (*get_dcc_alignment)(struct amdgpu_device *adev);
 
 	enum amdgpu_memory_partition (*query_mem_partition_mode)(
 		struct amdgpu_device *adev);
@@ -363,6 +365,10 @@ struct amdgpu_gmc {
 	(adev)->gmc.gmc_funcs->override_vm_pte_flags			\
 		((adev), (vm), (addr), (pte_flags))
 #define amdgpu_gmc_get_vbios_fb_size(adev) (adev)->gmc.gmc_funcs->get_vbios_fb_size((adev))
+#define amdgpu_gmc_get_dcc_alignment(adev) ({			\
+	typeof(adev) _adev = (adev);				\
+	_adev->gmc.gmc_funcs->get_dcc_alignment(_adev);		\
+})
 
 /**
  * amdgpu_gmc_vram_full_visible - Check if full VRAM is visible through the BAR

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

@@ -264,9 +264,8 @@ amdgpu_job_prepare_job(struct drm_sched_job *sched_job,
 	struct dma_fence *fence = NULL;
 	int r;
 
-	/* Ignore soft recovered fences here */
 	r = drm_sched_entity_error(s_entity);
-	if (r && r != -ENODATA)
+	if (r)
 		goto error;
 
 	if (!fence && job->gang_submit)

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

@@ -456,6 +456,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	u64 vis_usage = 0, max_bytes, min_block_size;
 	struct amdgpu_vram_mgr_resource *vres;
 	u64 size, remaining_size, lpfn, fpfn;
+	unsigned int adjust_dcc_size = 0;
 	struct drm_buddy *mm = &mgr->mm;
 	struct drm_buddy_block *block;
 	unsigned long pages_per_block;
@@ -511,7 +512,19 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		/* Allocate blocks in desired range */
 		vres->flags |= DRM_BUDDY_RANGE_ALLOCATION;
 
+	if (bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC &&
+	    adev->gmc.gmc_funcs->get_dcc_alignment)
+		adjust_dcc_size = amdgpu_gmc_get_dcc_alignment(adev);
+
 	remaining_size = (u64)vres->base.size;
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) {
+		unsigned int dcc_size;
+
+		dcc_size = roundup_pow_of_two(vres->base.size + adjust_dcc_size);
+		remaining_size = (u64)dcc_size;
+
+		vres->flags |= DRM_BUDDY_TRIM_DISABLE;
+	}
 
 	mutex_lock(&mgr->lock);
 	while (remaining_size) {
@@ -521,8 +534,11 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 			min_block_size = mgr->default_page_size;
 
 		size = remaining_size;
-		if ((size >= (u64)pages_per_block << PAGE_SHIFT) &&
-		    !(size & (((u64)pages_per_block << PAGE_SHIFT) - 1)))
+
+		if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size)
+			min_block_size = size;
+		else if ((size >= (u64)pages_per_block << PAGE_SHIFT) &&
+			 !(size & (((u64)pages_per_block << PAGE_SHIFT) - 1)))
 			min_block_size = (u64)pages_per_block << PAGE_SHIFT;
 
 		BUG_ON(min_block_size < mm->chunk_size);
@@ -553,6 +569,22 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	}
 	mutex_unlock(&mgr->lock);
 
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) {
+		struct drm_buddy_block *dcc_block;
+		unsigned long dcc_start;
+		u64 trim_start;
+
+		dcc_block = amdgpu_vram_mgr_first_block(&vres->blocks);
+		/* Adjust the start address for DCC buffers only */
+		dcc_start =
+			roundup((unsigned long)amdgpu_vram_mgr_block_start(dcc_block),
+				adjust_dcc_size);
+		trim_start = (u64)dcc_start;
+		drm_buddy_block_trim(mm, &trim_start,
+				     (u64)vres->base.size,
+				     &vres->blocks);
+	}
+
 	vres->base.start = 0;
 	size = max_t(u64, amdgpu_vram_mgr_blocks_size(&vres->blocks),
 		     vres->base.size);

drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c

@@ -202,6 +202,12 @@ static const struct amdgpu_hwip_reg_entry gc_gfx_queue_reg_list_12[] = {
 	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ)
 };
 
+static const struct soc15_reg_golden golden_settings_gc_12_0[] = {
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regDB_MEM_CONFIG, 0x0000000f, 0x0000000f),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regCB_HW_CONTROL_1, 0x03000000, 0x03000000),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_CTRL5, 0x00000070, 0x00000020)
+};
+
 #define DEFAULT_SH_MEM_CONFIG \
 	((SH_MEM_ADDRESS_MODE_64 << SH_MEM_CONFIG__ADDRESS_MODE__SHIFT) | \
 	 (SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) | \
@@ -3432,6 +3438,24 @@ static void gfx_v12_0_disable_gpa_mode(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, regCPG_PSP_DEBUG, data);
 }
 
+static void gfx_v12_0_init_golden_registers(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if (adev->rev_id == 0)
+			soc15_program_register_sequence(adev,
+					golden_settings_gc_12_0,
+					(const u32)ARRAY_SIZE(golden_settings_gc_12_0));
+		break;
+	default:
+		break;
+	}
+}
+
 static int gfx_v12_0_hw_init(void *handle)
 {
 	int r;
@@ -3472,6 +3496,9 @@ static int gfx_v12_0_hw_init(void *handle)
 		}
 	}
 
+	if (!amdgpu_emu_mode)
+		gfx_v12_0_init_golden_registers(adev);
+
 	adev->gfx.is_poweron = true;
 
 	if (get_gb_addr_config(adev))

drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c

@@ -542,6 +542,23 @@ static unsigned gmc_v12_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	return 0;
 }
 
+static unsigned int gmc_v12_0_get_dcc_alignment(struct amdgpu_device *adev)
+{
+	unsigned int max_tex_channel_caches, alignment;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(12, 0, 0) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(12, 0, 1))
+		return 0;
+
+	max_tex_channel_caches = adev->gfx.config.max_texture_channel_caches;
+	if (is_power_of_2(max_tex_channel_caches))
+		alignment = (unsigned int)(max_tex_channel_caches / SZ_4);
+	else
+		alignment = roundup_pow_of_two(max_tex_channel_caches);
+
+	return (unsigned int)(alignment * max_tex_channel_caches * SZ_1K);
+}
+
 static const struct amdgpu_gmc_funcs gmc_v12_0_gmc_funcs = {
 	.flush_gpu_tlb = gmc_v12_0_flush_gpu_tlb,
 	.flush_gpu_tlb_pasid = gmc_v12_0_flush_gpu_tlb_pasid,
@@ -551,6 +568,7 @@ static const struct amdgpu_gmc_funcs gmc_v12_0_gmc_funcs = {
 	.get_vm_pde = gmc_v12_0_get_vm_pde,
 	.get_vm_pte = gmc_v12_0_get_vm_pte,
 	.get_vbios_fb_size = gmc_v12_0_get_vbios_fb_size,
+	.get_dcc_alignment = gmc_v12_0_get_dcc_alignment,
 };
 
 static void gmc_v12_0_set_gmc_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.c

@@ -80,7 +80,8 @@ static uint32_t mmhub_v4_1_0_get_invalidate_req(unsigned int vmid,
 	/* invalidate using legacy mode on vmid*/
 	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ,
 			    PER_VMID_INVALIDATE_REQ, 1 << vmid);
-	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type);
+	/* Only use legacy inv on mmhub side */
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, 0);
 	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1);
 	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1);
 	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1);

drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c

@@ -1575,8 +1575,7 @@ static void sdma_v7_0_emit_copy_buffer(struct amdgpu_ib *ib,
 	ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
 		SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0) |
-		SDMA_PKT_COPY_LINEAR_HEADER_CPV((copy_flags &
-			(AMDGPU_COPY_FLAGS_READ_DECOMPRESSED | AMDGPU_COPY_FLAGS_WRITE_COMPRESSED)) ? 1 : 0);
+		SDMA_PKT_COPY_LINEAR_HEADER_CPV(1);
 
 	ib->ptr[ib->length_dw++] = byte_count - 1;
 	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
@@ -1590,6 +1589,8 @@ static void sdma_v7_0_emit_copy_buffer(struct amdgpu_ib *ib,
 			((copy_flags & AMDGPU_COPY_FLAGS_READ_DECOMPRESSED) ? SDMA_DCC_READ_CM(2) : 0) |
 			((copy_flags & AMDGPU_COPY_FLAGS_WRITE_COMPRESSED) ? SDMA_DCC_WRITE_CM(1) : 0) |
 			SDMA_DCC_MAX_COM(max_com) | SDMA_DCC_MAX_UCOM(1);
+	else
+		ib->ptr[ib->length_dw++] = 0;
 }
 
 /**
@@ -1616,7 +1617,7 @@ static void sdma_v7_0_emit_fill_buffer(struct amdgpu_ib *ib,
 
 static const struct amdgpu_buffer_funcs sdma_v7_0_buffer_funcs = {
 	.copy_max_bytes = 0x400000,
-	.copy_num_dw = 7,
+	.copy_num_dw = 8,
 	.emit_copy_buffer = sdma_v7_0_emit_copy_buffer,
 	.fill_max_bytes = 0x400000,
 	.fill_num_dw = 5,

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

@@ -1270,6 +1270,9 @@ static bool is_dsc_need_re_compute(
 		}
 	}
 
+	if (new_stream_on_link_num == 0)
+		return false;
+
 	/* check current_state if there stream on link but it is not in
 	 * new request state
 	 */

drivers/gpu/drm/amd/display/dc/dce/dmub_replay.c

@@ -185,8 +185,7 @@ static bool dmub_replay_copy_settings(struct dmub_replay *dmub,
 	else
 		copy_settings_data->flags.bitfields.force_wakeup_by_tps3 = 0;
 
-
-	dm_execute_dmub_cmd(dc, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
+	dc_wake_and_execute_dmub_cmd(dc, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 
 	return true;
 }