Merge tag 'amd-drm-fixes-7.0-2026-03-25' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-7.0-2026-03-25:

amdgpu:
- DSC fix
- Module parameter parsing fix
- PASID reuse fix
- drm_edid leak fix
- SMU 13.x fixes
- SMU 14.x fix
- Fence fix in amdgpu_amdkfd_submit_ib()
- LVDS fixes
- GPU page fault fix for non-4K pages

amdkfd:
- Ordering fix in kfd_ioctl_create_process()

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

From: Alex Deucher <[email protected]>
Link: https://patch.msgid.link/[email protected]

Author: airlied

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

@@ -692,9 +692,9 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
 		goto err_ib_sched;
 	}
 
-	/* Drop the initial kref_init count (see drm_sched_main as example) */
-	dma_fence_put(f);
 	ret = dma_fence_wait(f, false);
+	/* Drop the returned fence reference after the wait completes */
+	dma_fence_put(f);
 
 err_ib_sched:
 	amdgpu_job_free(job);

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

@@ -4207,7 +4207,8 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
 
 static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev)
 {
-	char *input = amdgpu_lockup_timeout;
+	char buf[AMDGPU_MAX_TIMEOUT_PARAM_LENGTH];
+	char *input = buf;
 	char *timeout_setting = NULL;
 	int index = 0;
 	long timeout;
@@ -4217,9 +4218,17 @@ static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev)
 	adev->gfx_timeout = adev->compute_timeout = adev->sdma_timeout =
 		adev->video_timeout = msecs_to_jiffies(2000);
 
-	if (!strnlen(input, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH))
+	if (!strnlen(amdgpu_lockup_timeout, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH))
 		return 0;
 
+	/*
+	 * strsep() destructively modifies its input by replacing delimiters
+	 * with '\0'. Use a stack copy so the global module parameter buffer
+	 * remains intact for multi-GPU systems where this function is called
+	 * once per device.
+	 */
+	strscpy(buf, amdgpu_lockup_timeout, sizeof(buf));
+
 	while ((timeout_setting = strsep(&input, ",")) &&
 	       strnlen(timeout_setting, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) {
 		ret = kstrtol(timeout_setting, 0, &timeout);

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

@@ -35,10 +35,13 @@
  * PASIDs are global address space identifiers that can be shared
  * between the GPU, an IOMMU and the driver. VMs on different devices
  * may use the same PASID if they share the same address
- * space. Therefore PASIDs are allocated using a global IDA. VMs are
- * looked up from the PASID per amdgpu_device.
+ * space. Therefore PASIDs are allocated using IDR cyclic allocator
+ * (similar to kernel PID allocation) which naturally delays reuse.
+ * VMs are looked up from the PASID per amdgpu_device.
  */
-static DEFINE_IDA(amdgpu_pasid_ida);
+
+static DEFINE_IDR(amdgpu_pasid_idr);
+static DEFINE_SPINLOCK(amdgpu_pasid_idr_lock);
 
 /* Helper to free pasid from a fence callback */
 struct amdgpu_pasid_cb {
@@ -50,23 +53,24 @@ struct amdgpu_pasid_cb {
  * amdgpu_pasid_alloc - Allocate a PASID
  * @bits: Maximum width of the PASID in bits, must be at least 1
  *
- * Allocates a PASID of the given width while keeping smaller PASIDs
- * available if possible.
+ * Uses kernel's IDR cyclic allocator (same as PID allocation).
+ * Allocates sequentially with automatic wrap-around.
  *
  * Returns a positive integer on success. Returns %-EINVAL if bits==0.
  * Returns %-ENOSPC if no PASID was available. Returns %-ENOMEM on
  * memory allocation failure.
  */
 int amdgpu_pasid_alloc(unsigned int bits)
 {
-	int pasid = -EINVAL;
+	int pasid;
 
-	for (bits = min(bits, 31U); bits > 0; bits--) {
-		pasid = ida_alloc_range(&amdgpu_pasid_ida, 1U << (bits - 1),
-					(1U << bits) - 1, GFP_KERNEL);
-		if (pasid != -ENOSPC)
-			break;
-	}
+	if (bits == 0)
+		return -EINVAL;
+
+	spin_lock(&amdgpu_pasid_idr_lock);
+	pasid = idr_alloc_cyclic(&amdgpu_pasid_idr, NULL, 1,
+				 1U << bits, GFP_KERNEL);
+	spin_unlock(&amdgpu_pasid_idr_lock);
 
 	if (pasid >= 0)
 		trace_amdgpu_pasid_allocated(pasid);
@@ -81,7 +85,10 @@ int amdgpu_pasid_alloc(unsigned int bits)
 void amdgpu_pasid_free(u32 pasid)
 {
 	trace_amdgpu_pasid_freed(pasid);
-	ida_free(&amdgpu_pasid_ida, pasid);
+
+	spin_lock(&amdgpu_pasid_idr_lock);
+	idr_remove(&amdgpu_pasid_idr, pasid);
+	spin_unlock(&amdgpu_pasid_idr_lock);
 }
 
 static void amdgpu_pasid_free_cb(struct dma_fence *fence,
@@ -616,3 +623,15 @@ void amdgpu_vmid_mgr_fini(struct amdgpu_device *adev)
 		}
 	}
 }
+
+/**
+ * amdgpu_pasid_mgr_cleanup - cleanup PASID manager
+ *
+ * Cleanup the IDR allocator.
+ */
+void amdgpu_pasid_mgr_cleanup(void)
+{
+	spin_lock(&amdgpu_pasid_idr_lock);
+	idr_destroy(&amdgpu_pasid_idr);
+	spin_unlock(&amdgpu_pasid_idr_lock);
+}

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

@@ -74,6 +74,7 @@ int amdgpu_pasid_alloc(unsigned int bits);
 void amdgpu_pasid_free(u32 pasid);
 void amdgpu_pasid_free_delayed(struct dma_resv *resv,
 			       u32 pasid);
+void amdgpu_pasid_mgr_cleanup(void);
 
 bool amdgpu_vmid_had_gpu_reset(struct amdgpu_device *adev,
 			       struct amdgpu_vmid *id);

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

@@ -2898,6 +2898,7 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev)
 	xa_destroy(&adev->vm_manager.pasids);
 
 	amdgpu_vmid_mgr_fini(adev);
+	amdgpu_pasid_mgr_cleanup();
 }
 
 /**
@@ -2973,14 +2974,14 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 	if (!root)
 		return false;
 
-	addr /= AMDGPU_GPU_PAGE_SIZE;
-
 	if (is_compute_context && !svm_range_restore_pages(adev, pasid, vmid,
-	    node_id, addr, ts, write_fault)) {
+	    node_id, addr >> PAGE_SHIFT, ts, write_fault)) {
 		amdgpu_bo_unref(&root);
 		return true;
 	}
 
+	addr /= AMDGPU_GPU_PAGE_SIZE;
+
 	r = amdgpu_bo_reserve(root, true);
 	if (r)
 		goto error_unref;

drivers/gpu/drm/amd/amdkfd/kfd_chardev.c

@@ -3170,11 +3170,11 @@ static int kfd_ioctl_create_process(struct file *filep, struct kfd_process *p, v
 	struct kfd_process *process;
 	int ret;
 
-	/* Each FD owns only one kfd_process */
-	if (p->context_id != KFD_CONTEXT_ID_PRIMARY)
+	if (!filep->private_data || !p)
 		return -EINVAL;
 
-	if (!filep->private_data || !p)
+	/* Each FD owns only one kfd_process */
+	if (p->context_id != KFD_CONTEXT_ID_PRIMARY)
 		return -EINVAL;
 
 	mutex_lock(&kfd_processes_mutex);

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

@@ -3909,8 +3909,9 @@ void amdgpu_dm_update_connector_after_detect(
 
 		aconnector->dc_sink = sink;
 		dc_sink_retain(aconnector->dc_sink);
+		drm_edid_free(aconnector->drm_edid);
+		aconnector->drm_edid = NULL;
 		if (sink->dc_edid.length == 0) {
-			aconnector->drm_edid = NULL;
 			hdmi_cec_unset_edid(aconnector);
 			if (aconnector->dc_link->aux_mode) {
 				drm_dp_cec_unset_edid(&aconnector->dm_dp_aux.aux);
@@ -5422,7 +5423,7 @@ static void setup_backlight_device(struct amdgpu_display_manager *dm,
 	caps = &dm->backlight_caps[aconnector->bl_idx];
 
 	/* Only offer ABM property when non-OLED and user didn't turn off by module parameter */
-	if (!caps->ext_caps->bits.oled && amdgpu_dm_abm_level < 0)
+	if (caps->ext_caps && !caps->ext_caps->bits.oled && amdgpu_dm_abm_level < 0)
 		drm_object_attach_property(&aconnector->base.base,
 					   dm->adev->mode_info.abm_level_property,
 					   ABM_SYSFS_CONTROL);
@@ -12523,6 +12524,11 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 	}
 
 	if (dc_resource_is_dsc_encoding_supported(dc)) {
+		for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) {
+			dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+			dm_new_crtc_state->mode_changed_independent_from_dsc = new_crtc_state->mode_changed;
+		}
+
 		for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) {
 			if (drm_atomic_crtc_needs_modeset(new_crtc_state)) {
 				ret = add_affected_mst_dsc_crtcs(state, crtc);

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

@@ -984,6 +984,7 @@ struct dm_crtc_state {
 
 	bool freesync_vrr_info_changed;
 
+	bool mode_changed_independent_from_dsc;
 	bool dsc_force_changed;
 	bool vrr_supported;
 	struct mod_freesync_config freesync_config;

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

@@ -1744,9 +1744,11 @@ int pre_validate_dsc(struct drm_atomic_state *state,
 			int ind = find_crtc_index_in_state_by_stream(state, stream);
 
 			if (ind >= 0) {
+				struct dm_crtc_state *dm_new_crtc_state = to_dm_crtc_state(state->crtcs[ind].new_state);
+
 				DRM_INFO_ONCE("%s:%d MST_DSC no mode changed for stream 0x%p\n",
 						__func__, __LINE__, stream);
-				state->crtcs[ind].new_state->mode_changed = 0;
+				dm_new_crtc_state->base.mode_changed = dm_new_crtc_state->mode_changed_independent_from_dsc;
 			}
 		}
 	}

drivers/gpu/drm/amd/display/dc/resource/dce100/dce100_resource.c

@@ -650,9 +650,6 @@ static struct link_encoder *dce100_link_encoder_create(
 		return &enc110->base;
 	}
 
-	if (enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs))
-		return NULL;
-
 	link_regs_id =
 		map_transmitter_id_to_phy_instance(enc_init_data->transmitter);
 
@@ -661,7 +658,8 @@ static struct link_encoder *dce100_link_encoder_create(
 				      &link_enc_feature,
 				      &link_enc_regs[link_regs_id],
 				      &link_enc_aux_regs[enc_init_data->channel - 1],
-				      &link_enc_hpd_regs[enc_init_data->hpd_source]);
+				      enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs) ?
+				      NULL : &link_enc_hpd_regs[enc_init_data->hpd_source]);
 	return &enc110->base;
 }