Merge tag 'drm-xe-fixes-2025-12-19' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes

airlied · airlied · commit f66ac60dee28 · 2025-12-19T10:56:13.000+10:00
UAPI Changes: - Limit num_syncs to prevent oversized kernel allocations (Shuicheng) - Disallow 0 OA property values (Ashutosh) - Disallow 0 EU stall property values (Ashutosh) Driver Changes: - Fix kobject leak (Shuicheng) - Workaround (Vinay) - Loop variable reference fix (Matt Brost) - Fix a CONFIG corner-case incorrect number of arguments (Arnd Bergmann) - Skip reason prefix while emitting array (Raag) - VF migration fix (Tomasz) - Fix context in mei interrupt top half (Junxiao) - Don't include the CCS metadata in the dma-buf sg-table (Thomas) - VF queueing recovery work fix (Satyanarayana) - Increase TDF timeout (Jagmeet) - GT reset registers vs scheduler ordering fix (Jan) - Adjust long-running workload timeslices (Matt Brost) - Always set OA_OAGLBCTXCTRL_COUNTER_RESUME (Ashutosh) - Fix a return value (Dan Carpenter) - Drop preempt-fences when destroying imported dma-bufs (Thomas) - Use usleep_range for accurate long-running workload timeslicing (Matthew) Signed-off-by: Dave Airlie <airlied@redhat.com> From: Thomas Hellstrom <thomas.hellstrom@linux.intel.com> Link: https://patch.msgid.link/aUSMlQ4iruzm0NQR@fedora
diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
@@ -1527,7 +1527,7 @@ static bool xe_ttm_bo_lock_in_destructor(struct ttm_buffer_object *ttm_bo)
 	 * always succeed here, as long as we hold the lru lock.
 	 */
 	spin_lock(&ttm_bo->bdev->lru_lock);
-	locked = dma_resv_trylock(ttm_bo->base.resv);
+	locked = dma_resv_trylock(&ttm_bo->base._resv);
 	spin_unlock(&ttm_bo->bdev->lru_lock);
 	xe_assert(xe, locked);
 
@@ -1547,13 +1547,6 @@ static void xe_ttm_bo_release_notify(struct ttm_buffer_object *ttm_bo)
 	bo = ttm_to_xe_bo(ttm_bo);
 	xe_assert(xe_bo_device(bo), !(bo->created && kref_read(&ttm_bo->base.refcount)));
 
-	/*
-	 * Corner case where TTM fails to allocate memory and this BOs resv
-	 * still points the VMs resv
-	 */
-	if (ttm_bo->base.resv != &ttm_bo->base._resv)
-		return;
-
 	if (!xe_ttm_bo_lock_in_destructor(ttm_bo))
 		return;
 
@@ -1563,22 +1556,22 @@ static void xe_ttm_bo_release_notify(struct ttm_buffer_object *ttm_bo)
 	 * TODO: Don't do this for external bos once we scrub them after
 	 * unbind.
 	 */
-	dma_resv_for_each_fence(&cursor, ttm_bo->base.resv,
+	dma_resv_for_each_fence(&cursor, &ttm_bo->base._resv,
 				DMA_RESV_USAGE_BOOKKEEP, fence) {
 		if (xe_fence_is_xe_preempt(fence) &&
 		    !dma_fence_is_signaled(fence)) {
 			if (!replacement)
 				replacement = dma_fence_get_stub();
 
-			dma_resv_replace_fences(ttm_bo->base.resv,
+			dma_resv_replace_fences(&ttm_bo->base._resv,
 						fence->context,
 						replacement,
 						DMA_RESV_USAGE_BOOKKEEP);
 		}
 	}
 	dma_fence_put(replacement);
 
-	dma_resv_unlock(ttm_bo->base.resv);
+	dma_resv_unlock(&ttm_bo->base._resv);
 }
 
 static void xe_ttm_bo_delete_mem_notify(struct ttm_buffer_object *ttm_bo)
diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
@@ -1056,7 +1056,7 @@ static void tdf_request_sync(struct xe_device *xe)
 		 * transient and need to be flushed..
 		 */
 		if (xe_mmio_wait32(&gt->mmio, XE2_TDF_CTRL, TRANSIENT_FLUSH_REQUEST, 0,
-				   150, NULL, false))
+				   300, NULL, false))
 			xe_gt_err_once(gt, "TD flush timeout\n");
 
 		xe_force_wake_put(gt_to_fw(gt), fw_ref);
diff --git a/drivers/gpu/drm/xe/xe_dma_buf.c b/drivers/gpu/drm/xe/xe_dma_buf.c
@@ -124,7 +124,7 @@ static struct sg_table *xe_dma_buf_map(struct dma_buf_attachment *attach,
 	case XE_PL_TT:
 		sgt = drm_prime_pages_to_sg(obj->dev,
 					    bo->ttm.ttm->pages,
-					    bo->ttm.ttm->num_pages);
+					    obj->size >> PAGE_SHIFT);
 		if (IS_ERR(sgt))
 			return sgt;
 
diff --git a/drivers/gpu/drm/xe/xe_eu_stall.c b/drivers/gpu/drm/xe/xe_eu_stall.c
@@ -315,7 +315,7 @@ static int xe_eu_stall_user_ext_set_property(struct xe_device *xe, u64 extension
 		return -EFAULT;
 
 	if (XE_IOCTL_DBG(xe, ext.property >= ARRAY_SIZE(xe_set_eu_stall_property_funcs)) ||
-	    XE_IOCTL_DBG(xe, ext.pad))
+	    XE_IOCTL_DBG(xe, !ext.property) || XE_IOCTL_DBG(xe, ext.pad))
 		return -EINVAL;
 
 	idx = array_index_nospec(ext.property, ARRAY_SIZE(xe_set_eu_stall_property_funcs));
diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c
@@ -132,7 +132,8 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 
 	if (XE_IOCTL_DBG(xe, args->extensions) ||
 	    XE_IOCTL_DBG(xe, args->pad[0] || args->pad[1] || args->pad[2]) ||
-	    XE_IOCTL_DBG(xe, args->reserved[0] || args->reserved[1]))
+	    XE_IOCTL_DBG(xe, args->reserved[0] || args->reserved[1]) ||
+	    XE_IOCTL_DBG(xe, args->num_syncs > DRM_XE_MAX_SYNCS))
 		return -EINVAL;
 
 	q = xe_exec_queue_lookup(xef, args->exec_queue_id);
diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c
@@ -797,16 +797,17 @@ static int do_gt_restart(struct xe_gt *gt)
 		xe_gt_sriov_pf_init_hw(gt);
 
 	xe_mocs_init(gt);
-	err = xe_uc_start(&gt->uc);
-	if (err)
-		return err;
 
 	for_each_hw_engine(hwe, gt, id)
 		xe_reg_sr_apply_mmio(&hwe->reg_sr, gt);
 
 	/* Get CCS mode in sync between sw/hw */
 	xe_gt_apply_ccs_mode(gt);
 
+	err = xe_uc_start(&gt->uc);
+	if (err)
+		return err;
+
 	/* Restore GT freq to expected values */
 	xe_gt_sanitize_freq(gt);
 
diff --git a/drivers/gpu/drm/xe/xe_gt_freq.c b/drivers/gpu/drm/xe/xe_gt_freq.c
@@ -293,8 +293,10 @@ int xe_gt_freq_init(struct xe_gt *gt)
 		return -ENOMEM;
 
 	err = sysfs_create_files(gt->freq, freq_attrs);
-	if (err)
+	if (err) {
+		kobject_put(gt->freq);
 		return err;
+	}
 
 	err = devm_add_action_or_reset(xe->drm.dev, freq_fini, gt->freq);
 	if (err)
diff --git a/drivers/gpu/drm/xe/xe_gt_idle.c b/drivers/gpu/drm/xe/xe_gt_idle.c
@@ -5,6 +5,7 @@
 
 #include <drm/drm_managed.h>
 
+#include <generated/xe_wa_oob.h>
 #include "xe_force_wake.h"
 #include "xe_device.h"
 #include "xe_gt.h"
@@ -16,6 +17,7 @@
 #include "xe_mmio.h"
 #include "xe_pm.h"
 #include "xe_sriov.h"
+#include "xe_wa.h"
 
 /**
  * DOC: Xe GT Idle
@@ -145,6 +147,12 @@ void xe_gt_idle_enable_pg(struct xe_gt *gt)
 		xe_mmio_write32(mmio, RENDER_POWERGATE_IDLE_HYSTERESIS, 25);
 	}
 
+	if (XE_GT_WA(gt, 14020316580))
+		gtidle->powergate_enable &= ~(VDN_HCP_POWERGATE_ENABLE(0) |
+					      VDN_MFXVDENC_POWERGATE_ENABLE(0) |
+					      VDN_HCP_POWERGATE_ENABLE(2) |
+					      VDN_MFXVDENC_POWERGATE_ENABLE(2));
+
 	xe_mmio_write32(mmio, POWERGATE_ENABLE, gtidle->powergate_enable);
 	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_vf.c b/drivers/gpu/drm/xe/xe_gt_sriov_vf.c
@@ -733,7 +733,7 @@ static void vf_start_migration_recovery(struct xe_gt *gt)
 
 	spin_lock(&gt->sriov.vf.migration.lock);
 
-	if (!gt->sriov.vf.migration.recovery_queued ||
+	if (!gt->sriov.vf.migration.recovery_queued &&
 	    !gt->sriov.vf.migration.recovery_teardown) {
 		gt->sriov.vf.migration.recovery_queued = true;
 		WRITE_ONCE(gt->sriov.vf.migration.recovery_inprogress, true);
diff --git a/drivers/gpu/drm/xe/xe_gt_throttle.c b/drivers/gpu/drm/xe/xe_gt_throttle.c
@@ -140,7 +140,7 @@ static ssize_t reasons_show(struct kobject *kobj,
 		struct throttle_attribute *other_ta = kobj_attribute_to_throttle(kattr);
 
 		if (other_ta->mask != U32_MAX && reasons & other_ta->mask)
-			ret += sysfs_emit_at(buff, ret, "%s ", (*pother)->name);
+			ret += sysfs_emit_at(buff, ret, "%s ", (*pother)->name + strlen("reason_"));
 	}
 
 	if (drm_WARN_ONCE(&xe->drm, !ret, "Unknown reason: %#x\n", reasons))
diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
@@ -717,26 +717,46 @@ static bool vf_recovery(struct xe_guc *guc)
 	return xe_gt_recovery_pending(guc_to_gt(guc));
 }
 
+static inline void relaxed_ms_sleep(unsigned int delay_ms)
+{
+	unsigned long min_us, max_us;
+
+	if (!delay_ms)
+		return;
+
+	if (delay_ms > 20) {
+		msleep(delay_ms);
+		return;
+	}
+
+	min_us = mul_u32_u32(delay_ms, 1000);
+	max_us = min_us + 500;
+
+	usleep_range(min_us, max_us);
+}
+
 static int wq_wait_for_space(struct xe_exec_queue *q, u32 wqi_size)
 {
 	struct xe_guc *guc = exec_queue_to_guc(q);
 	struct xe_device *xe = guc_to_xe(guc);
 	struct iosys_map map = xe_lrc_parallel_map(q->lrc[0]);
-	unsigned int sleep_period_ms = 1;
+	unsigned int sleep_period_ms = 1, sleep_total_ms = 0;
 
 #define AVAILABLE_SPACE \
 	CIRC_SPACE(q->guc->wqi_tail, q->guc->wqi_head, WQ_SIZE)
 	if (wqi_size > AVAILABLE_SPACE && !vf_recovery(guc)) {
 try_again:
 		q->guc->wqi_head = parallel_read(xe, map, wq_desc.head);
-		if (wqi_size > AVAILABLE_SPACE) {
-			if (sleep_period_ms == 1024) {
+		if (wqi_size > AVAILABLE_SPACE && !vf_recovery(guc)) {
+			if (sleep_total_ms > 2000) {
 				xe_gt_reset_async(q->gt);
 				return -ENODEV;
 			}
 
 			msleep(sleep_period_ms);
-			sleep_period_ms <<= 1;
+			sleep_total_ms += sleep_period_ms;
+			if (sleep_period_ms < 64)
+				sleep_period_ms <<= 1;
 			goto try_again;
 		}
 	}
@@ -1585,7 +1605,7 @@ static void __guc_exec_queue_process_msg_suspend(struct xe_sched_msg *msg)
 				since_resume_ms;
 
 			if (wait_ms > 0 && q->guc->resume_time)
-				msleep(wait_ms);
+				relaxed_ms_sleep(wait_ms);
 
 			set_exec_queue_suspended(q);
 			disable_scheduling(q, false);
@@ -2253,10 +2273,11 @@ static void guc_exec_queue_unpause_prepare(struct xe_guc *guc,
 					   struct xe_exec_queue *q)
 {
 	struct xe_gpu_scheduler *sched = &q->guc->sched;
-	struct xe_sched_job *job = NULL;
+	struct xe_sched_job *job = NULL, *__job;
 	bool restore_replay = false;
 
-	list_for_each_entry(job, &sched->base.pending_list, drm.list) {
+	list_for_each_entry(__job, &sched->base.pending_list, drm.list) {
+		job = __job;
 		restore_replay |= job->restore_replay;
 		if (restore_replay) {
 			xe_gt_dbg(guc_to_gt(guc), "Replay JOB - guc_id=%d, seqno=%d",
diff --git a/drivers/gpu/drm/xe/xe_heci_gsc.c b/drivers/gpu/drm/xe/xe_heci_gsc.c
@@ -223,7 +223,7 @@ void xe_heci_gsc_irq_handler(struct xe_device *xe, u32 iir)
 	if (xe->heci_gsc.irq < 0)
 		return;
 
-	ret = generic_handle_irq(xe->heci_gsc.irq);
+	ret = generic_handle_irq_safe(xe->heci_gsc.irq);
 	if (ret)
 		drm_err_ratelimited(&xe->drm, "error handling GSC irq: %d\n", ret);
 }
@@ -243,7 +243,7 @@ void xe_heci_csc_irq_handler(struct xe_device *xe, u32 iir)
 	if (xe->heci_gsc.irq < 0)
 		return;
 
-	ret = generic_handle_irq(xe->heci_gsc.irq);
+	ret = generic_handle_irq_safe(xe->heci_gsc.irq);
 	if (ret)
 		drm_err_ratelimited(&xe->drm, "error handling GSC irq: %d\n", ret);
 }
diff --git a/drivers/gpu/drm/xe/xe_oa.c b/drivers/gpu/drm/xe/xe_oa.c
@@ -1105,11 +1105,12 @@ static int xe_oa_enable_metric_set(struct xe_oa_stream *stream)
 			oag_buf_size_select(stream) |
 			oag_configure_mmio_trigger(stream, true));
 
-	xe_mmio_write32(mmio, __oa_regs(stream)->oa_ctx_ctrl, stream->periodic ?
-			(OAG_OAGLBCTXCTRL_COUNTER_RESUME |
+	xe_mmio_write32(mmio, __oa_regs(stream)->oa_ctx_ctrl,
+			OAG_OAGLBCTXCTRL_COUNTER_RESUME |
+			(stream->periodic ?
 			 OAG_OAGLBCTXCTRL_TIMER_ENABLE |
 			 REG_FIELD_PREP(OAG_OAGLBCTXCTRL_TIMER_PERIOD_MASK,
-					stream->period_exponent)) : 0);
+					 stream->period_exponent) : 0));
 
 	/*
 	 * Initialize Super Queue Internal Cnt Register
@@ -1254,6 +1255,9 @@ static int xe_oa_set_no_preempt(struct xe_oa *oa, u64 value,
 static int xe_oa_set_prop_num_syncs(struct xe_oa *oa, u64 value,
 				    struct xe_oa_open_param *param)
 {
+	if (XE_IOCTL_DBG(oa->xe, value > DRM_XE_MAX_SYNCS))
+		return -EINVAL;
+
 	param->num_syncs = value;
 	return 0;
 }
@@ -1343,7 +1347,7 @@ static int xe_oa_user_ext_set_property(struct xe_oa *oa, enum xe_oa_user_extn_fr
 		     ARRAY_SIZE(xe_oa_set_property_funcs_config));
 
 	if (XE_IOCTL_DBG(oa->xe, ext.property >= ARRAY_SIZE(xe_oa_set_property_funcs_open)) ||
-	    XE_IOCTL_DBG(oa->xe, ext.pad))
+	    XE_IOCTL_DBG(oa->xe, !ext.property) || XE_IOCTL_DBG(oa->xe, ext.pad))
 		return -EINVAL;
 
 	idx = array_index_nospec(ext.property, ARRAY_SIZE(xe_oa_set_property_funcs_open));
diff --git a/drivers/gpu/drm/xe/xe_sriov_vfio.c b/drivers/gpu/drm/xe/xe_sriov_vfio.c
@@ -21,7 +21,7 @@ EXPORT_SYMBOL_FOR_MODULES(xe_sriov_vfio_get_pf, "xe-vfio-pci");
 bool xe_sriov_vfio_migration_supported(struct xe_device *xe)
 {
 	if (!IS_SRIOV_PF(xe))
-		return -EPERM;
+		return false;
 
 	return xe_sriov_pf_migration_supported(xe);
 }
diff --git a/drivers/gpu/drm/xe/xe_svm.h b/drivers/gpu/drm/xe/xe_svm.h
@@ -214,7 +214,7 @@ int xe_svm_init(struct xe_vm *vm)
 {
 #if IS_ENABLED(CONFIG_DRM_GPUSVM)
 	return drm_gpusvm_init(&vm->svm.gpusvm, "Xe SVM (simple)", &vm->xe->drm,
-			       NULL, NULL, 0, 0, 0, NULL, NULL, 0);
+			       NULL, 0, 0, 0, NULL, NULL, 0);
 #else
 	return 0;
 #endif
diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
@@ -1508,7 +1508,10 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags, struct xe_file *xef)
 	INIT_WORK(&vm->destroy_work, vm_destroy_work_func);
 
 	INIT_LIST_HEAD(&vm->preempt.exec_queues);
-	vm->preempt.min_run_period_ms = 10;	/* FIXME: Wire up to uAPI */
+	if (flags & XE_VM_FLAG_FAULT_MODE)
+		vm->preempt.min_run_period_ms = 0;
+	else
+		vm->preempt.min_run_period_ms = 5;
 
 	for_each_tile(tile, xe, id)
 		xe_range_fence_tree_init(&vm->rftree[id]);
@@ -3324,6 +3327,9 @@ static int vm_bind_ioctl_check_args(struct xe_device *xe, struct xe_vm *vm,
 	if (XE_IOCTL_DBG(xe, args->extensions))
 		return -EINVAL;
 
+	if (XE_IOCTL_DBG(xe, args->num_syncs > DRM_XE_MAX_SYNCS))
+		return -EINVAL;
+
 	if (args->num_binds > 1) {
 		u64 __user *bind_user =
 			u64_to_user_ptr(args->vector_of_binds);
diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h
@@ -263,7 +263,7 @@ struct xe_vm {
 		 * @min_run_period_ms: The minimum run period before preempting
 		 * an engine again
 		 */
-		s64 min_run_period_ms;
+		unsigned int min_run_period_ms;
 		/** @exec_queues: list of exec queues attached to this VM */
 		struct list_head exec_queues;
 		/** @num_exec_queues: number exec queues attached to this VM */
diff --git a/drivers/gpu/drm/xe/xe_wa.c b/drivers/gpu/drm/xe/xe_wa.c
@@ -270,14 +270,6 @@ static const struct xe_rtp_entry_sr gt_was[] = {
 	  XE_RTP_ACTIONS(SET(VDBOX_CGCTL3F1C(0), MFXPIPE_CLKGATE_DIS)),
 	  XE_RTP_ENTRY_FLAG(FOREACH_ENGINE),
 	},
-	{ XE_RTP_NAME("14020316580"),
-	  XE_RTP_RULES(MEDIA_VERSION(1301)),
-	  XE_RTP_ACTIONS(CLR(POWERGATE_ENABLE,
-			     VDN_HCP_POWERGATE_ENABLE(0) |
-			     VDN_MFXVDENC_POWERGATE_ENABLE(0) |
-			     VDN_HCP_POWERGATE_ENABLE(2) |
-			     VDN_MFXVDENC_POWERGATE_ENABLE(2))),
-	},
 	{ XE_RTP_NAME("14019449301"),
 	  XE_RTP_RULES(MEDIA_VERSION(1301), ENGINE_CLASS(VIDEO_DECODE)),
 	  XE_RTP_ACTIONS(SET(VDBOX_CGCTL3F08(0), CG3DDISHRS_CLKGATE_DIS)),
diff --git a/drivers/gpu/drm/xe/xe_wa_oob.rules b/drivers/gpu/drm/xe/xe_wa_oob.rules
@@ -76,3 +76,4 @@
 
 15015404425_disable	PLATFORM(PANTHERLAKE), MEDIA_STEP(B0, FOREVER)
 16026007364    MEDIA_VERSION(3000)
+14020316580    MEDIA_VERSION(1301)
diff --git a/include/uapi/drm/xe_drm.h b/include/uapi/drm/xe_drm.h
@@ -1463,6 +1463,7 @@ struct drm_xe_exec {
 	/** @exec_queue_id: Exec queue ID for the batch buffer */
 	__u32 exec_queue_id;
 
+#define DRM_XE_MAX_SYNCS 1024
 	/** @num_syncs: Amount of struct drm_xe_sync in array. */
 	__u32 num_syncs;
 

Original file line number	Diff line number	Diff line change
`@@ -140,7 +140,7 @@ static ssize_t reasons_show(struct kobject *kobj,`
`140`	`140`	`struct throttle_attribute *other_ta = kobj_attribute_to_throttle(kattr);`
`141`	`141`
`142`	`142`	`if (other_ta->mask != U32_MAX && reasons & other_ta->mask)`
`143`		`- ret += sysfs_emit_at(buff, ret, "%s ", (*pother)->name);`
	`143`	`+ ret += sysfs_emit_at(buff, ret, "%s ", (*pother)->name + strlen("reason_"));`
`144`	`144`	`}`
`145`	`145`
`146`	`146`	`if (drm_WARN_ONCE(&xe->drm, !ret, "Unknown reason: %#x\n", reasons))`