Merge tag 'drm-fixes-2025-04-11-1' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Dave Airlie:
 "Weekly fixes, as expected it has a bit more in it than probably usual
  for rc2. amdgpu/xe/i915 lead the way with fixes all over for a bunch
  of other drivers. Nothing major stands out from what I can see.

  tests:
   - Clean up struct drm_display_mode in various places

  i915:
   - Fix scanline offset for LNL+ and BMG+
   - Fix GVT unterminated-string-initialization build warning
   - Fix DP rate limit when sink doesn't support TPS4
   - Handle GDDR + ECC memory type detection
   - Fix VRR parameter change check
   - Fix fence not released on early probe errors
   - Disable render power gating during live selftests

  xe:
   - Add another BMG PCI ID
   - Fix UAFs on migration paths
   - Fix shift-out-of-bounds access on TLB invalidation
   - Ensure ccs_mode is correctly set on gt reset
   - Extend some HW workarounds to Xe3
   - Fix PM runtime get/put on sysfs files
   - Fix u64 division on 32b
   - Fix flickering due to missing L3 invalidations
   - Fix missing error code return

  amdgpu:
   - MES FW version caching fixes
   - Only use GTT as a fallback if we already have a backing store
   - dma_buf fix
   - IP discovery fix
   - Replay and PSR with VRR fix
   - DC FP fixes
   - eDP fixes
   - KIQ TLB invalidate fix
   - Enable dmem groups support
   - Allow pinning VRAM dma bufs if imports can do P2P
   - Workload profile fixes
   - Prevent possible division by 0 in fan handling

  amdkfd:
   - Queue reset fixes

  imagination:
   - Fix overflow
   - Fix use-after-free

  ivpu:
   - Fix suspend/resume

  nouveau:
   - Do not deref dangling pointer

  rockchip:
   - Set DP/HDMI registers correctly

  udmabuf:
   - Fix overflow

  virtgpu:
   - Set reservation lock on dma-buf import
   - Fix error handling in prepare_fb"

* tag 'drm-fixes-2025-04-11-1' of https://gitlab.freedesktop.org/drm/kernel: (58 commits)
  drm/rockchip: dw_hdmi_qp: Fix io init for dw_hdmi_qp_rockchip_resume
  drm/rockchip: vop2: Fix interface enable/mux setting of DP1 on rk3588
  drm/amdgpu/mes12: optimize MES pipe FW version fetching
  drm/amd/pm/smu11: Prevent division by zero
  drm/amdgpu: cancel gfx idle work in device suspend for s0ix
  drm/amd/display: pause the workload setting in dm
  drm/amdgpu/pm/swsmu: implement pause workload profile
  drm/amdgpu/pm: add workload profile pause helper
  drm/i915/huc: Fix fence not released on early probe errors
  drm/i915/vrr: Add vrr.vsync_{start, end} in vrr_params_changed
  drm/tests: probe-helper: Fix drm_display_mode memory leak
  drm/tests: modes: Fix drm_display_mode memory leak
  drm/tests: modes: Fix drm_display_mode memory leak
  drm/tests: cmdline: Fix drm_display_mode memory leak
  drm/tests: modeset: Fix drm_display_mode memory leak
  drm/tests: modeset: Fix drm_display_mode memory leak
  drm/tests: helpers: Create kunit helper to destroy a drm_display_mode
  drm/xe: Restore EIO errno return when GuC PC start fails
  drm/xe: Invalidate L3 read-only cachelines for geometry streams too
  drm/xe: avoid plain 64-bit division
  ...

Author: torvalds

drivers/accel/ivpu/ivpu_debugfs.c

@@ -332,7 +332,7 @@ ivpu_force_recovery_fn(struct file *file, const char __user *user_buf, size_t si
 		return -EINVAL;
 
 	ret = ivpu_rpm_get(vdev);
-	if (ret)
+	if (ret < 0)
 		return ret;
 
 	ivpu_pm_trigger_recovery(vdev, "debugfs");
@@ -383,7 +383,7 @@ static int dct_active_set(void *data, u64 active_percent)
 		return -EINVAL;
 
 	ret = ivpu_rpm_get(vdev);
-	if (ret)
+	if (ret < 0)
 		return ret;
 
 	if (active_percent)

drivers/accel/ivpu/ivpu_ipc.c

@@ -302,7 +302,8 @@ ivpu_ipc_send_receive_internal(struct ivpu_device *vdev, struct vpu_jsm_msg *req
 	struct ivpu_ipc_consumer cons;
 	int ret;
 
-	drm_WARN_ON(&vdev->drm, pm_runtime_status_suspended(vdev->drm.dev));
+	drm_WARN_ON(&vdev->drm, pm_runtime_status_suspended(vdev->drm.dev) &&
+		    pm_runtime_enabled(vdev->drm.dev));
 
 	ivpu_ipc_consumer_add(vdev, &cons, channel, NULL);
 

drivers/accel/ivpu/ivpu_ms.c

@@ -4,6 +4,7 @@
  */
 
 #include <drm/drm_file.h>
+#include <linux/pm_runtime.h>
 
 #include "ivpu_drv.h"
 #include "ivpu_gem.h"
@@ -44,6 +45,10 @@ int ivpu_ms_start_ioctl(struct drm_device *dev, void *data, struct drm_file *fil
 	    args->sampling_period_ns < MS_MIN_SAMPLE_PERIOD_NS)
 		return -EINVAL;
 
+	ret = ivpu_rpm_get(vdev);
+	if (ret < 0)
+		return ret;
+
 	mutex_lock(&file_priv->ms_lock);
 
 	if (get_instance_by_mask(file_priv, args->metric_group_mask)) {
@@ -96,6 +101,8 @@ int ivpu_ms_start_ioctl(struct drm_device *dev, void *data, struct drm_file *fil
 	kfree(ms);
 unlock:
 	mutex_unlock(&file_priv->ms_lock);
+
+	ivpu_rpm_put(vdev);
 	return ret;
 }
 
@@ -160,6 +167,10 @@ int ivpu_ms_get_data_ioctl(struct drm_device *dev, void *data, struct drm_file *
 	if (!args->metric_group_mask)
 		return -EINVAL;
 
+	ret = ivpu_rpm_get(vdev);
+	if (ret < 0)
+		return ret;
+
 	mutex_lock(&file_priv->ms_lock);
 
 	ms = get_instance_by_mask(file_priv, args->metric_group_mask);
@@ -187,6 +198,7 @@ int ivpu_ms_get_data_ioctl(struct drm_device *dev, void *data, struct drm_file *
 unlock:
 	mutex_unlock(&file_priv->ms_lock);
 
+	ivpu_rpm_put(vdev);
 	return ret;
 }
 
@@ -204,11 +216,17 @@ int ivpu_ms_stop_ioctl(struct drm_device *dev, void *data, struct drm_file *file
 {
 	struct ivpu_file_priv *file_priv = file->driver_priv;
 	struct drm_ivpu_metric_streamer_stop *args = data;
+	struct ivpu_device *vdev = file_priv->vdev;
 	struct ivpu_ms_instance *ms;
+	int ret;
 
 	if (!args->metric_group_mask)
 		return -EINVAL;
 
+	ret = ivpu_rpm_get(vdev);
+	if (ret < 0)
+		return ret;
+
 	mutex_lock(&file_priv->ms_lock);
 
 	ms = get_instance_by_mask(file_priv, args->metric_group_mask);
@@ -217,6 +235,7 @@ int ivpu_ms_stop_ioctl(struct drm_device *dev, void *data, struct drm_file *file
 
 	mutex_unlock(&file_priv->ms_lock);
 
+	ivpu_rpm_put(vdev);
 	return ms ? 0 : -EINVAL;
 }
 
@@ -281,6 +300,9 @@ int ivpu_ms_get_info_ioctl(struct drm_device *dev, void *data, struct drm_file *
 void ivpu_ms_cleanup(struct ivpu_file_priv *file_priv)
 {
 	struct ivpu_ms_instance *ms, *tmp;
+	struct ivpu_device *vdev = file_priv->vdev;
+
+	pm_runtime_get_sync(vdev->drm.dev);
 
 	mutex_lock(&file_priv->ms_lock);
 
@@ -293,6 +315,8 @@ void ivpu_ms_cleanup(struct ivpu_file_priv *file_priv)
 		free_instance(file_priv, ms);
 
 	mutex_unlock(&file_priv->ms_lock);
+
+	pm_runtime_put_autosuspend(vdev->drm.dev);
 }
 
 void ivpu_ms_cleanup_all(struct ivpu_device *vdev)

drivers/dma-buf/udmabuf.c

@@ -393,7 +393,7 @@ static long udmabuf_create(struct miscdevice *device,
 	if (!ubuf)
 		return -ENOMEM;
 
-	pglimit = (size_limit_mb * 1024 * 1024) >> PAGE_SHIFT;
+	pglimit = ((u64)size_limit_mb * 1024 * 1024) >> PAGE_SHIFT;
 	for (i = 0; i < head->count; i++) {
 		pgoff_t subpgcnt;
 

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

@@ -353,7 +353,6 @@ enum amdgpu_kiq_irq {
 	AMDGPU_CP_KIQ_IRQ_DRIVER0 = 0,
 	AMDGPU_CP_KIQ_IRQ_LAST
 };
-#define SRIOV_USEC_TIMEOUT  1200000 /* wait 12 * 100ms for SRIOV */
 #define MAX_KIQ_REG_WAIT       5000 /* in usecs, 5ms */
 #define MAX_KIQ_REG_BAILOUT_INTERVAL   5 /* in msecs, 5ms */
 #define MAX_KIQ_REG_TRY 1000

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

@@ -3643,6 +3643,13 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 			    adev, adev->ip_blocks[i].version->type))
 			continue;
 
+		/* Since we skip suspend for S0i3, we need to cancel the delayed
+		 * idle work here as the suspend callback never gets called.
+		 */
+		if (adev->in_s0ix &&
+		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX &&
+		    amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 0, 0))
+			cancel_delayed_work_sync(&adev->gfx.idle_work);
 		/* skip suspend of gfx/mes and psp for S0ix
 		 * gfx is in gfxoff state, so on resume it will exit gfxoff just
 		 * like at runtime. PSP is also part of the always on hardware

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

@@ -120,6 +120,8 @@ MODULE_FIRMWARE("amdgpu/vega20_ip_discovery.bin");
 MODULE_FIRMWARE("amdgpu/raven_ip_discovery.bin");
 MODULE_FIRMWARE("amdgpu/raven2_ip_discovery.bin");
 MODULE_FIRMWARE("amdgpu/picasso_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/arcturus_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/aldebaran_ip_discovery.bin");
 
 #define mmIP_DISCOVERY_VERSION  0x16A00
 #define mmRCC_CONFIG_MEMSIZE	0xde3

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

@@ -75,11 +75,25 @@ static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf,
  */
 static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach)
 {
-	struct drm_gem_object *obj = attach->dmabuf->priv;
-	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+	struct dma_buf *dmabuf = attach->dmabuf;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(dmabuf->priv);
+	u32 domains = bo->preferred_domains;
 
-	/* pin buffer into GTT */
-	return amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT);
+	dma_resv_assert_held(dmabuf->resv);
+
+	/*
+	 * 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.
+	 */
+	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;
+
+	return amdgpu_bo_pin(bo, domains);
 }
 
 /**
@@ -134,9 +148,6 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach,
 		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 		if (r)
 			return ERR_PTR(r);
-
-	} else if (bo->tbo.resource->mem_type != TTM_PL_TT) {
-		return ERR_PTR(-EBUSY);
 	}
 
 	switch (bo->tbo.resource->mem_type) {
@@ -184,7 +195,7 @@ static void amdgpu_dma_buf_unmap(struct dma_buf_attachment *attach,
 				 struct sg_table *sgt,
 				 enum dma_data_direction dir)
 {
-	if (sgt->sgl->page_link) {
+	if (sg_page(sgt->sgl)) {
 		dma_unmap_sgtable(attach->dev, sgt, dir, 0);
 		sg_free_table(sgt);
 		kfree(sgt);

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

@@ -699,12 +699,10 @@ int amdgpu_gmc_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint16_t pasid,
 				   uint32_t flush_type, bool all_hub,
 				   uint32_t inst)
 {
-	u32 usec_timeout = amdgpu_sriov_vf(adev) ? SRIOV_USEC_TIMEOUT :
-		adev->usec_timeout;
 	struct amdgpu_ring *ring = &adev->gfx.kiq[inst].ring;
 	struct amdgpu_kiq *kiq = &adev->gfx.kiq[inst];
 	unsigned int ndw;
-	int r;
+	int r, cnt = 0;
 	uint32_t seq;
 
 	/*
@@ -761,10 +759,21 @@ int amdgpu_gmc_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint16_t pasid,
 
 		amdgpu_ring_commit(ring);
 		spin_unlock(&adev->gfx.kiq[inst].ring_lock);
-		if (amdgpu_fence_wait_polling(ring, seq, usec_timeout) < 1) {
+
+		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+
+		might_sleep();
+		while (r < 1 && cnt++ < MAX_KIQ_REG_TRY &&
+		       !amdgpu_reset_pending(adev->reset_domain)) {
+			msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
+			r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+		}
+
+		if (cnt > MAX_KIQ_REG_TRY) {
 			dev_err(adev->dev, "timeout waiting for kiq fence\n");
 			r = -ETIME;
-		}
+		} else
+			r = 0;
 	}
 
 error_unlock_reset:

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

@@ -163,8 +163,8 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 		 * When GTT is just an alternative to VRAM make sure that we
 		 * only use it as fallback and still try to fill up VRAM first.
 		 */
-		if (domain & abo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM &&
-		    !(adev->flags & AMD_IS_APU))
+		if (abo->tbo.resource && !(adev->flags & AMD_IS_APU) &&
+		    domain & abo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM)
 			places[c].flags |= TTM_PL_FLAG_FALLBACK;
 		c++;
 	}