sema-svc: rewind replays may only be skipped when their release was delivered

The per-put-epoch consumed-slot skip must not suppress a slot whose entry
was consumed *without* delivering its release (the CE grace expiry in real
mode): the driver-API smoke test's CE pushes are consumed unreadable before
their ring pin lands and depend on the pin-batch rewind replaying them once
the translation appears. The skip now applies only to slots whose
consumption delivered a release:

- tail consumption records a slot when the iteration decoded a release
  (svc_rel_count delta); the sweep-delivered entry_want consume and the
  fake-release consume always record
- the grace-expiry consume records only when a recovery (fake or
  sema_rel_recover) delivered the release
- svc_fake_release counts as a delivered release

So real-mode replays still re-process undelivered entries, while the
runtime-client case (readable pushes delivered immediately) keeps the
fast path that removed the 8 s stalls.

1 files changed, 10 insertions, 2 deletions

diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c
index 30c5561..e97c2ff 100644
--- a/kernel/gsp_sema_service.c
+++ b/kernel/gsp_sema_service.c
@@ -4995,6 +4995,7 @@ static bool svc_fake_release(struct nv_vgpu_device *vgpu, struct svc_channel *ch
 		svc.releases_applied++;
 	}
 	svc.releases_faked++;
+	svc_rel_count++;
 	ch->expected_sem_payload++;
 	return true;
 }
@@ -5149,6 +5150,9 @@ static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 				}
 			}
 
+			u32 rel_n_iter = svc_rel_count;
+
+			(void)rel_n_iter;
 			if (rrc) {
 				stalled = true;	/* ring unreadable: sweep RAM */
 				break;
@@ -5614,7 +5618,10 @@ if (gsp_sema_fake_rel &&
 							}
 							ch->rescue_fail = 0;
 							ch->defer_since = 0;
-							svc_replay_note(ch, idx, put);
+							if (gsp_sema_fake_rel ||
+							    gsp_sema_rel_recover)
+								svc_replay_note(ch, idx,
+										put);
 							ch->last_get = (idx + 1) %
 								       ch->gpfifo_entries;
 							svc.gpfifo_skipped++;
@@ -5650,7 +5657,8 @@ if (gsp_sema_fake_rel &&
 				}
 			}
 
-			svc_replay_note(ch, idx, put);
+			if (ch->engine_type && svc_rel_count != rel_n_iter)
+				svc_replay_note(ch, idx, put);
 			ch->last_get = (idx + 1) % ch->gpfifo_entries;
 			ch->rescue_fail = 0;
 			ch->defer_since = 0;