Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v2] drm/xe/vf: Start re-emission from first unsignaled job during VF migration
@ 2025-10-31 20:13 Matthew Brost
  2025-10-31 22:25 ` ✓ CI.KUnit: success for " Patchwork
                   ` (3 more replies)
  0 siblings, 4 replies; 9+ messages in thread
From: Matthew Brost @ 2025-10-31 20:13 UTC (permalink / raw)
  To: intel-xe; +Cc: michal.wajdeczko, tomasz.lis

The LRC software ring tail is reset to the first unsignaled pending
job's head.

Fix the re-emission logic to begin submitting from the first unsignaled
job detected, rather than scanning all pending jobs, which can cause
imbalance.

v2:
 - Include missing local changes

Fixes: c25c1010df88 ("drm/xe/vf: Replay GuC submission state on pause / unpause")
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
 drivers/gpu/drm/xe/xe_gpu_scheduler.h |  5 +++--
 drivers/gpu/drm/xe/xe_guc_submit.c    | 19 +++++++++++--------
 2 files changed, 14 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_gpu_scheduler.h b/drivers/gpu/drm/xe/xe_gpu_scheduler.h
index 9955397aaaa9..357afaec68d7 100644
--- a/drivers/gpu/drm/xe/xe_gpu_scheduler.h
+++ b/drivers/gpu/drm/xe/xe_gpu_scheduler.h
@@ -54,13 +54,14 @@ static inline void xe_sched_tdr_queue_imm(struct xe_gpu_scheduler *sched)
 static inline void xe_sched_resubmit_jobs(struct xe_gpu_scheduler *sched)
 {
 	struct drm_sched_job *s_job;
+	bool skip_emit = false;
 
 	list_for_each_entry(s_job, &sched->base.pending_list, list) {
 		struct drm_sched_fence *s_fence = s_job->s_fence;
 		struct dma_fence *hw_fence = s_fence->parent;
 
-		if (to_xe_sched_job(s_job)->skip_emit ||
-		    (hw_fence && !dma_fence_is_signaled(hw_fence)))
+		skip_emit |= to_xe_sched_job(s_job)->skip_emit;
+		if (skip_emit || (hw_fence && !dma_fence_is_signaled(hw_fence)))
 			sched->base.ops->run_job(s_job);
 	}
 }
diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
index d4ffdb71ef3d..f25b71aca498 100644
--- a/drivers/gpu/drm/xe/xe_guc_submit.c
+++ b/drivers/gpu/drm/xe/xe_guc_submit.c
@@ -2152,6 +2152,8 @@ static void guc_exec_queue_pause(struct xe_guc *guc, struct xe_exec_queue *q)
 
 	job = xe_sched_first_pending_job(sched);
 	if (job) {
+		job->skip_emit = true;
+
 		/*
 		 * Adjust software tail so jobs submitted overwrite previous
 		 * position in ring buffer with new GGTT addresses.
@@ -2241,17 +2243,18 @@ 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 drm_sched_job *s_job;
 	struct xe_sched_job *job = NULL;
+	bool skip_emit = false;
 
-	list_for_each_entry(s_job, &sched->base.pending_list, list) {
-		job = to_xe_sched_job(s_job);
-
-		xe_gt_dbg(guc_to_gt(guc), "Replay JOB - guc_id=%d, seqno=%d",
-			  q->guc->id, xe_sched_job_seqno(job));
+	list_for_each_entry(job, &sched->base.pending_list, drm.list) {
+		skip_emit |= job->skip_emit;
+		if (skip_emit) {
+			xe_gt_dbg(guc_to_gt(guc), "Replay JOB - guc_id=%d, seqno=%d",
+				  q->guc->id, xe_sched_job_seqno(job));
 
-		q->ring_ops->emit_job(job);
-		job->skip_emit = true;
+			q->ring_ops->emit_job(job);
+			job->skip_emit = true;
+		}
 	}
 
 	if (job)
-- 
2.34.1


^ permalink raw reply related	[flat|nested] 9+ messages in thread
end of thread, other threads:[~2025-11-20  0:32 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-10-31 20:13 [PATCH v2] drm/xe/vf: Start re-emission from first unsignaled job during VF migration Matthew Brost
2025-10-31 22:25 ` ✓ CI.KUnit: success for " Patchwork
2025-10-31 23:43 ` ✓ Xe.CI.BAT: " Patchwork
2025-11-01  1:33 ` [PATCH v2] " Lis, Tomasz
2025-11-07  7:48   ` Matthew Brost
2025-11-19  2:48     ` Lis, Tomasz
2025-11-19 19:55       ` Matthew Brost
2025-11-20  0:31         ` Lis, Tomasz
2025-11-01 13:52 ` ✗ Xe.CI.Full: failure for " Patchwork

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox