Merge tag 'amd-drm-fixes-6.15-2025-04-23' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-6.15-2025-04-23:

amdgpu:
- P2P DMA fixes
- Display reset fixes
- DCN 3.5 fixes
- ACPI EDID fix
- LTTPR fix
- mode_valid() fix

Signed-off-by: Dave Airlie <[email protected]>

From: Alex Deucher <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

Author: airlied

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

@@ -43,6 +43,29 @@
 #include <linux/dma-fence-array.h>
 #include <linux/pci-p2pdma.h>
 
+static const struct dma_buf_attach_ops amdgpu_dma_buf_attach_ops;
+
+/**
+ * dma_buf_attach_adev - Helper to get adev of an attachment
+ *
+ * @attach: attachment
+ *
+ * Returns:
+ * A struct amdgpu_device * if the attaching device is an amdgpu device or
+ * partition, NULL otherwise.
+ */
+static struct amdgpu_device *dma_buf_attach_adev(struct dma_buf_attachment *attach)
+{
+	if (attach->importer_ops == &amdgpu_dma_buf_attach_ops) {
+		struct drm_gem_object *obj = attach->importer_priv;
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
+		return amdgpu_ttm_adev(bo->tbo.bdev);
+	}
+
+	return NULL;
+}
+
 /**
  * amdgpu_dma_buf_attach - &dma_buf_ops.attach implementation
  *
@@ -54,11 +77,13 @@
 static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf,
 				 struct dma_buf_attachment *attach)
 {
+	struct amdgpu_device *attach_adev = dma_buf_attach_adev(attach);
 	struct drm_gem_object *obj = dmabuf->priv;
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 
-	if (pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
+	if (!amdgpu_dmabuf_is_xgmi_accessible(attach_adev, bo) &&
+	    pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
 		attach->peer2peer = false;
 
 	amdgpu_vm_bo_update_shared(bo);
@@ -77,22 +102,32 @@ static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach)
 {
 	struct dma_buf *dmabuf = attach->dmabuf;
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(dmabuf->priv);
-	u32 domains = bo->preferred_domains;
+	u32 domains = bo->allowed_domains;
 
 	dma_resv_assert_held(dmabuf->resv);
 
-	/*
-	 * Try pinning into VRAM to allow P2P with RDMA NICs without ODP
+	/* Try pinning into VRAM to allow P2P with RDMA NICs without ODP
 	 * support if all attachments can do P2P. If any attachment can't do
 	 * P2P just pin into GTT instead.
+	 *
+	 * To avoid with conflicting pinnings between GPUs and RDMA when move
+	 * notifiers are disabled, only allow pinning in VRAM when move
+	 * notiers are enabled.
 	 */
-	list_for_each_entry(attach, &dmabuf->attachments, node)
-		if (!attach->peer2peer)
-			domains &= ~AMDGPU_GEM_DOMAIN_VRAM;
+	if (!IS_ENABLED(CONFIG_DMABUF_MOVE_NOTIFY)) {
+		domains &= ~AMDGPU_GEM_DOMAIN_VRAM;
+	} else {
+		list_for_each_entry(attach, &dmabuf->attachments, node)
+			if (!attach->peer2peer)
+				domains &= ~AMDGPU_GEM_DOMAIN_VRAM;
+	}
 
 	if (domains & AMDGPU_GEM_DOMAIN_VRAM)
 		bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 
+	if (WARN_ON(!domains))
+		return -EINVAL;
+
 	return amdgpu_bo_pin(bo, domains);
 }
 
@@ -470,6 +505,9 @@ bool amdgpu_dmabuf_is_xgmi_accessible(struct amdgpu_device *adev,
 	struct drm_gem_object *obj = &bo->tbo.base;
 	struct drm_gem_object *gobj;
 
+	if (!adev)
+		return false;
+
 	if (obj->import_attach) {
 		struct dma_buf *dma_buf = obj->import_attach->dmabuf;
 

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

@@ -1920,26 +1920,6 @@ static enum dmub_ips_disable_type dm_get_default_ips_mode(
 	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 	case IP_VERSION(3, 5, 0):
 	case IP_VERSION(3, 6, 0):
-		/*
-		 * On DCN35 systems with Z8 enabled, it's possible for IPS2 + Z8 to
-		 * cause a hard hang. A fix exists for newer PMFW.
-		 *
-		 * As a workaround, for non-fixed PMFW, force IPS1+RCG as the deepest
-		 * IPS state in all cases, except for s0ix and all displays off (DPMS),
-		 * where IPS2 is allowed.
-		 *
-		 * When checking pmfw version, use the major and minor only.
-		 */
-		if ((adev->pm.fw_version & 0x00FFFF00) < 0x005D6300)
-			ret = DMUB_IPS_RCG_IN_ACTIVE_IPS2_IN_OFF;
-		else if (amdgpu_ip_version(adev, GC_HWIP, 0) > IP_VERSION(11, 5, 0))
-			/*
-			 * Other ASICs with DCN35 that have residency issues with
-			 * IPS2 in idle.
-			 * We want them to use IPS2 only in display off cases.
-			 */
-			ret =  DMUB_IPS_RCG_IN_ACTIVE_IPS2_IN_OFF;
-		break;
 	case IP_VERSION(3, 5, 1):
 		ret =  DMUB_IPS_RCG_IN_ACTIVE_IPS2_IN_OFF;
 		break;
@@ -3355,16 +3335,16 @@ static void dm_gpureset_commit_state(struct dc_state *dc_state,
 	for (k = 0; k < dc_state->stream_count; k++) {
 		bundle->stream_update.stream = dc_state->streams[k];
 
-		for (m = 0; m < dc_state->stream_status->plane_count; m++) {
+		for (m = 0; m < dc_state->stream_status[k].plane_count; m++) {
 			bundle->surface_updates[m].surface =
-				dc_state->stream_status->plane_states[m];
+				dc_state->stream_status[k].plane_states[m];
 			bundle->surface_updates[m].surface->force_full_update =
 				true;
 		}
 
 		update_planes_and_stream_adapter(dm->dc,
 					 UPDATE_TYPE_FULL,
-					 dc_state->stream_status->plane_count,
+					 dc_state->stream_status[k].plane_count,
 					 dc_state->streams[k],
 					 &bundle->stream_update,
 					 bundle->surface_updates);
@@ -6521,12 +6501,12 @@ decide_crtc_timing_for_drm_display_mode(struct drm_display_mode *drm_mode,
 					const struct drm_display_mode *native_mode,
 					bool scale_enabled)
 {
-	if (scale_enabled) {
-		copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
-	} else if (native_mode->clock == drm_mode->clock &&
-			native_mode->htotal == drm_mode->htotal &&
-			native_mode->vtotal == drm_mode->vtotal) {
-		copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
+	if (scale_enabled || (
+	    native_mode->clock == drm_mode->clock &&
+	    native_mode->htotal == drm_mode->htotal &&
+	    native_mode->vtotal == drm_mode->vtotal)) {
+		if (native_mode->crtc_clock)
+			copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
 	} else {
 		/* no scaling nor amdgpu inserted, no need to patch */
 	}
@@ -11043,6 +11023,9 @@ static bool should_reset_plane(struct drm_atomic_state *state,
 	    state->allow_modeset)
 		return true;
 
+	if (amdgpu_in_reset(adev) && state->allow_modeset)
+		return true;
+
 	/* Exit early if we know that we're adding or removing the plane. */
 	if (old_plane_state->crtc != new_plane_state->crtc)
 		return true;

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

@@ -918,7 +918,7 @@ dm_helpers_probe_acpi_edid(void *data, u8 *buf, unsigned int block, size_t len)
 {
 	struct drm_connector *connector = data;
 	struct acpi_device *acpidev = ACPI_COMPANION(connector->dev->dev);
-	unsigned char start = block * EDID_LENGTH;
+	unsigned short start = block * EDID_LENGTH;
 	struct edid *edid;
 	int r;
 

drivers/gpu/drm/amd/display/dc/dml/dcn35/dcn35_fpu.c

@@ -195,9 +195,9 @@ struct _vcs_dpi_soc_bounding_box_st dcn3_5_soc = {
 	.dcn_downspread_percent = 0.5,
 	.gpuvm_min_page_size_bytes = 4096,
 	.hostvm_min_page_size_bytes = 4096,
-	.do_urgent_latency_adjustment = 0,
+	.do_urgent_latency_adjustment = 1,
 	.urgent_latency_adjustment_fabric_clock_component_us = 0,
-	.urgent_latency_adjustment_fabric_clock_reference_mhz = 0,
+	.urgent_latency_adjustment_fabric_clock_reference_mhz = 3000,
 };
 
 void dcn35_build_wm_range_table_fpu(struct clk_mgr *clk_mgr)

drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training_8b_10b.c

@@ -35,6 +35,17 @@
 #define DC_LOGGER \
 	link->ctx->logger
 
+static void get_default_8b_10b_lttpr_aux_rd_interval(
+		union training_aux_rd_interval *training_rd_interval)
+{
+	/* LTTPR are required to program DPCD 0000Eh to 0x4 (16ms) upon AUX
+	 * read reply to this register. Since old sinks with DPCD rev 1.1
+	 * and earlier may not support this register, assume the mandatory
+	 * value is programmed by the LTTPR to avoid AUX timeout issues.
+	 */
+	training_rd_interval->raw = 0x4;
+}
+
 static int32_t get_cr_training_aux_rd_interval(struct dc_link *link,
 		const struct dc_link_settings *link_settings,
 		enum lttpr_mode lttpr_mode)
@@ -43,17 +54,22 @@ static int32_t get_cr_training_aux_rd_interval(struct dc_link *link,
 	uint32_t wait_in_micro_secs = 100;
 
 	memset(&training_rd_interval, 0, sizeof(training_rd_interval));
-	if (link_dp_get_encoding_format(link_settings) == DP_8b_10b_ENCODING &&
-			link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_12) {
-		core_link_read_dpcd(
-				link,
-				DP_TRAINING_AUX_RD_INTERVAL,
-				(uint8_t *)&training_rd_interval,
-				sizeof(training_rd_interval));
-		if (lttpr_mode != LTTPR_MODE_NON_TRANSPARENT)
-			wait_in_micro_secs = 400;
-		if (training_rd_interval.bits.TRAINIG_AUX_RD_INTERVAL)
-			wait_in_micro_secs = training_rd_interval.bits.TRAINIG_AUX_RD_INTERVAL * 4000;
+	if (link_dp_get_encoding_format(link_settings) == DP_8b_10b_ENCODING) {
+		if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_12)
+			core_link_read_dpcd(
+					link,
+					DP_TRAINING_AUX_RD_INTERVAL,
+					(uint8_t *)&training_rd_interval,
+					sizeof(training_rd_interval));
+		else if (dp_is_lttpr_present(link))
+			get_default_8b_10b_lttpr_aux_rd_interval(&training_rd_interval);
+
+		if (training_rd_interval.raw != 0) {
+			if (lttpr_mode != LTTPR_MODE_NON_TRANSPARENT)
+				wait_in_micro_secs = 400;
+			if (training_rd_interval.bits.TRAINIG_AUX_RD_INTERVAL)
+				wait_in_micro_secs = training_rd_interval.bits.TRAINIG_AUX_RD_INTERVAL * 4000;
+		}
 	}
 	return wait_in_micro_secs;
 }
@@ -71,13 +87,15 @@ static uint32_t get_eq_training_aux_rd_interval(
 				DP_128B132B_TRAINING_AUX_RD_INTERVAL,
 				(uint8_t *)&training_rd_interval,
 				sizeof(training_rd_interval));
-	} else if (link_dp_get_encoding_format(link_settings) == DP_8b_10b_ENCODING &&
-			link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_12) {
-		core_link_read_dpcd(
-				link,
-				DP_TRAINING_AUX_RD_INTERVAL,
-				(uint8_t *)&training_rd_interval,
-				sizeof(training_rd_interval));
+	} else if (link_dp_get_encoding_format(link_settings) == DP_8b_10b_ENCODING) {
+		if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_12)
+			core_link_read_dpcd(
+					link,
+					DP_TRAINING_AUX_RD_INTERVAL,
+					(uint8_t *)&training_rd_interval,
+					sizeof(training_rd_interval));
+		else if (dp_is_lttpr_present(link))
+			get_default_8b_10b_lttpr_aux_rd_interval(&training_rd_interval);
 	}
 
 	switch (training_rd_interval.bits.TRAINIG_AUX_RD_INTERVAL) {