drm/xe/vf: Wakeup in GuC backend on VF post migration recovery

If VF post-migration recovery is in progress, the recovery flow will
rebuild all GuC submission state. In this case, exit all waiters to
ensure that submission queue scheduling can also be paused. Avoid taking
any adverse actions after aborting the wait.

As part of waking up the GuC backend, suspend_wait can now return
-EAGAIN indicating the waiter should be retried. If the caller is
running on work item, that work item need to be requeued to avoid a
deadlock for the work item blocking the VF migration recovery work item.

v3:
- Don't block in preempt fence work queue as this can interfere with VF
post-migration work queue scheduling leading to deadlock (Testing)
- Use xe_gt_recovery_inprogress (Michal)
v5:
- Use static function for vf_recovery (Michal)
- Add helper to wake CT waiters (Michal)
- Move some code to following patch (Michal)
- Adjust commit message to explain suspend_wait returning -EAGAIN (Michal)
- Add kernel doc to suspend_wait around returning -EAGAIN
v7:
- Add comment on why a shared wait queue is need on VFs (Michal)
- Guard again suspend_wait signaling early on resfix donw (Tomasz)
v8:
- Fix kernel doc (CI)

Signed-off-by: Matthew Brost <[e-mail]>
Reviewed-by: Michal Wajdeczko <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]

drivers/gpu/drm/xe/xe_exec_queue_types.h

···
 	 * call after suspend. In dma-fencing path thus must return within a
 	 * reasonable amount of time. -ETIME return shall indicate an error
 	 * waiting for suspend resulting in associated VM getting killed.
+	 * -EAGAIN return indicates the wait should be tried again, if the wait
+	 * is within a work item, the work item should be requeued as deadlock
+	 * avoidance mechanism.
 	 */
 	int (*suspend_wait)(struct xe_exec_queue *q);
 	/**

drivers/gpu/drm/xe/xe_gt_sriov_vf.c

···
 #include "xe_gt_sriov_vf.h"
 #include "xe_gt_sriov_vf_types.h"
 #include "xe_guc.h"
+#include "xe_guc_ct.h"
 #include "xe_guc_hxg_helpers.h"
 #include "xe_guc_relay.h"
 #include "xe_guc_submit.h"
···
 	    !gt->sriov.vf.migration.recovery_teardown) {
 		gt->sriov.vf.migration.recovery_queued = true;
 		WRITE_ONCE(gt->sriov.vf.migration.recovery_inprogress, true);
+		smp_wmb();	/* Ensure above write visable before wake */
+
+		xe_guc_ct_wake_waiters(&gt->uc.guc.ct);
 
 		started = queue_work(gt->ordered_wq, &gt->sriov.vf.migration.worker);
 		xe_gt_sriov_info(gt, "VF migration recovery %s\n", started ?

drivers/gpu/drm/xe/xe_guc_ct.h

···
 
 long xe_guc_ct_queue_proc_time_jiffies(struct xe_guc_ct *ct);
 
+/**
+ * xe_guc_ct_wake_waiters() - GuC CT wake up waiters
+ * @ct: GuC CT object
+ */
+static inline void xe_guc_ct_wake_waiters(struct xe_guc_ct *ct)
+{
+	wake_up_all(&ct->wq);
+}
+
 #endif

drivers/gpu/drm/xe/xe_guc_submit.c

···
 #include "xe_gt.h"
 #include "xe_gt_clock.h"
 #include "xe_gt_printk.h"
-#include "xe_gt_sriov_vf.h"
 #include "xe_guc.h"
 #include "xe_guc_capture.h"
 #include "xe_guc_ct.h"
···
 	return (WQ_SIZE - q->guc->wqi_tail);
 }
 
+static bool vf_recovery(struct xe_guc *guc)
+{
+	return xe_gt_recovery_pending(guc_to_gt(guc));
+}
+
 static int wq_wait_for_space(struct xe_exec_queue *q, u32 wqi_size)
 {
 	struct xe_guc *guc = exec_queue_to_guc(q);
···
 
 #define AVAILABLE_SPACE \
 	CIRC_SPACE(q->guc->wqi_tail, q->guc->wqi_head, WQ_SIZE)
-	if (wqi_size > AVAILABLE_SPACE) {
+	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) {
···
 	ret = wait_event_timeout(guc->ct.wq,
 				 (!exec_queue_pending_enable(q) &&
 				  !exec_queue_pending_disable(q)) ||
-					 xe_guc_read_stopped(guc),
+					 xe_guc_read_stopped(guc) ||
+					 vf_recovery(guc),
 				 HZ * 5);
-	if (!ret) {
+	if (!ret && !vf_recovery(guc)) {
 		struct xe_gpu_scheduler *sched = &q->guc->sched;
 
 		xe_gt_warn(q->gt, "Pending enable/disable failed to respond\n");
···
 	bool wedged = false;
 
 	xe_gt_assert(guc_to_gt(guc), xe_exec_queue_is_lr(q));
+
+	if (vf_recovery(guc))
+		return;
+
 	trace_xe_exec_queue_lr_cleanup(q);
 
 	if (!exec_queue_killed(q))
···
 		 */
 		ret = wait_event_timeout(guc->ct.wq,
 					 !exec_queue_pending_disable(q) ||
-					 xe_guc_read_stopped(guc), HZ * 5);
+					 xe_guc_read_stopped(guc) ||
+					 vf_recovery(guc), HZ * 5);
+		if (vf_recovery(guc))
+			return;
+
 		if (!ret) {
 			xe_gt_warn(q->gt, "Schedule disable failed to respond, guc_id=%d\n",
 				   q->guc->id);
···
 
 	ret = wait_event_timeout(guc->ct.wq,
 				 !exec_queue_pending_enable(q) ||
-				 xe_guc_read_stopped(guc), HZ * 5);
-	if (!ret || xe_guc_read_stopped(guc)) {
+				 xe_guc_read_stopped(guc) ||
+				 vf_recovery(guc), HZ * 5);
+	if ((!ret && !vf_recovery(guc)) || xe_guc_read_stopped(guc)) {
 		xe_gt_warn(guc_to_gt(guc), "Schedule enable failed to respond");
 		set_exec_queue_banned(q);
 		xe_gt_reset_async(q->gt);
···
 	 * list so job can be freed and kick scheduler ensuring free job is not
 	 * lost.
 	 */
-	if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &job->fence->flags))
+	if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &job->fence->flags) ||
+	    vf_recovery(guc))
 		return DRM_GPU_SCHED_STAT_NO_HANG;
 
 	/* Kill the run_job entry point */
···
 			ret = wait_event_timeout(guc->ct.wq,
 						 (!exec_queue_pending_enable(q) &&
 						  !exec_queue_pending_disable(q)) ||
-						 xe_guc_read_stopped(guc), HZ * 5);
+						 xe_guc_read_stopped(guc) ||
+						 vf_recovery(guc), HZ * 5);
+			if (vf_recovery(guc))
+				goto handle_vf_resume;
 			if (!ret || xe_guc_read_stopped(guc))
 				goto trigger_reset;
 
···
 		smp_rmb();
 		ret = wait_event_timeout(guc->ct.wq,
 					 !exec_queue_pending_disable(q) ||
-					 xe_guc_read_stopped(guc), HZ * 5);
+					 xe_guc_read_stopped(guc) ||
+					 vf_recovery(guc), HZ * 5);
+		if (vf_recovery(guc))
+			goto handle_vf_resume;
 		if (!ret || xe_guc_read_stopped(guc)) {
 trigger_reset:
 			if (!ret)
···
 	 * some thought, do this in a follow up.
 	 */
 	xe_sched_submission_start(sched);
+handle_vf_resume:
 	return DRM_GPU_SCHED_STAT_NO_HANG;
 }
 
···
 
 static void __suspend_fence_signal(struct xe_exec_queue *q)
 {
+	struct xe_guc *guc = exec_queue_to_guc(q);
+	struct xe_device *xe = guc_to_xe(guc);
+
 	if (!q->guc->suspend_pending)
 		return;
 
 	WRITE_ONCE(q->guc->suspend_pending, false);
-	wake_up(&q->guc->suspend_wait);
+
+	/*
+	 * We use a GuC shared wait queue for VFs because the VF resfix start
+	 * interrupt must be able to wake all instances of suspend_wait. This
+	 * prevents the VF migration worker from being starved during
+	 * scheduling.
+	 */
+	if (IS_SRIOV_VF(xe))
+		wake_up_all(&guc->ct.wq);
+	else
+		wake_up(&q->guc->suspend_wait);
 }
 
 static void suspend_fence_signal(struct xe_exec_queue *q)
···
 
 	if (guc_exec_queue_allowed_to_change_state(q) && !exec_queue_suspended(q) &&
 	    exec_queue_enabled(q)) {
-		wait_event(guc->ct.wq, (q->guc->resume_time != RESUME_PENDING ||
-			   xe_guc_read_stopped(guc)) && !exec_queue_pending_disable(q));
+		wait_event(guc->ct.wq, vf_recovery(guc) ||
+			   ((q->guc->resume_time != RESUME_PENDING ||
+			   xe_guc_read_stopped(guc)) && !exec_queue_pending_disable(q)));
 
 		if (!xe_guc_read_stopped(guc)) {
 			s64 since_resume_ms =
···
 
 	q->entity = &ge->entity;
 
-	if (xe_guc_read_stopped(guc))
+	if (xe_guc_read_stopped(guc) || vf_recovery(guc))
 		xe_sched_stop(sched);
 
 	mutex_unlock(&guc->submission_state.lock);
···
 static int guc_exec_queue_suspend_wait(struct xe_exec_queue *q)
 {
 	struct xe_guc *guc = exec_queue_to_guc(q);
+	struct xe_device *xe = guc_to_xe(guc);
 	int ret;
 
 	/*
···
 	 * suspend_pending upon kill but to be paranoid but races in which
 	 * suspend_pending is set after kill also check kill here.
 	 */
-	ret = wait_event_interruptible_timeout(q->guc->suspend_wait,
-					       !READ_ONCE(q->guc->suspend_pending) ||
-					       exec_queue_killed(q) ||
-					       xe_guc_read_stopped(guc),
-					       HZ * 5);
+#define WAIT_COND \
+	(!READ_ONCE(q->guc->suspend_pending) ||	exec_queue_killed(q) || \
+	 xe_guc_read_stopped(guc))
+
+retry:
+	if (IS_SRIOV_VF(xe))
+		ret = wait_event_interruptible_timeout(guc->ct.wq, WAIT_COND ||
+						       vf_recovery(guc),
+						       HZ * 5);
+	else
+		ret = wait_event_interruptible_timeout(q->guc->suspend_wait,
+						       WAIT_COND, HZ * 5);
+
+	if (vf_recovery(guc) && !xe_device_wedged((guc_to_xe(guc))))
+		return -EAGAIN;
 
 	if (!ret) {
 		xe_gt_warn(guc_to_gt(guc),
···
 			   q->guc->id);
 		/* XXX: Trigger GT reset? */
 		return -ETIME;
+	} else if (IS_SRIOV_VF(xe) && !WAIT_COND) {
+		/* Corner case on RESFIX DONE where vf_recovery() changes */
+		goto retry;
 	}
+
+#undef WAIT_COND
 
 	return ret < 0 ? ret : 0;
 }
···
 {
 	int ret;
 
-	if (xe_gt_WARN_ON(guc_to_gt(guc),
-			  xe_gt_sriov_vf_recovery_pending(guc_to_gt(guc))))
+	if (xe_gt_WARN_ON(guc_to_gt(guc), vf_recovery(guc)))
 		return 0;
 
 	if (!guc->submission_state.initialized)

drivers/gpu/drm/xe/xe_preempt_fence.c

···
 #include <linux/slab.h>
 
 #include "xe_exec_queue.h"
+#include "xe_gt_printk.h"
+#include "xe_guc_exec_queue_types.h"
 #include "xe_vm.h"
 
 static void preempt_fence_work_func(struct work_struct *w)
···
 		dma_fence_set_error(&pfence->base, pfence->error);
 	} else if (!q->ops->reset_status(q)) {
 		int err = q->ops->suspend_wait(q);
+
+		if (err == -EAGAIN) {
+			xe_gt_dbg(q->gt, "PREEMPT FENCE RETRY guc_id=%d",
+				  q->guc->id);
+			queue_work(q->vm->xe->preempt_fence_wq,
+				   &pfence->preempt_work);
+			dma_fence_end_signalling(cookie);
+			return;
+		}
 
 		if (err)
 			dma_fence_set_error(&pfence->base, err);