sema: consume stale GPFIFO entries after repeated rescue failures

Entries whose push buffer was overwritten (the rescue cannot validate any
candidate page) were retried forever, rescanning RAM every 5 s and
leaving the ring stuck behind them. After three real rescue failures
(rate-limited attempts do not count) the entry is consumed with a
'skipping stale entry' log and the pump moves on. New skip= stat.

1 files changed, 32 insertions, 3 deletions

diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c
index 0a4e723..930118e 100644
--- a/kernel/gsp_sema_service.c
+++ b/kernel/gsp_sema_service.c
@@ -141,6 +141,7 @@ struct svc_channel {
 	u32	expected_sem_payload;	/* next semaphore payload we expect for this channel's tracking sem */
 	u32	logged_expected;	/* last pool value logged (rate limit) */
 	unsigned long rescue_last;	/* rate limit for the full-RAM rescue scan */
+	u32	rescue_fail;	/* consecutive failed rescues for the entry */
 };
 
 /*
@@ -317,6 +318,7 @@ struct sema_svc {
 	u64			entry_log;		/* per-push summary budget */
 	u64			ops_denied;		/* work refused: implausible stream */
 	u64			false_trans;		/* resolvable VA mapped to non-push */
+	u64			gpfifo_skipped;		/* stale entries consumed */
 	u64			releases_seen;
 	u64			releases_applied;
 	u64			releases_missed;	/* no VA->PA entry */
@@ -1463,11 +1465,11 @@ static void svc_rebuild(struct nv_vgpu_device *vgpu)
 		svc.releases_stale, svc.releases_jumped,
 		svc.walks_ok, svc.walks_ok + svc.walks_miss, svc.lat_hits,
 		svc_scan_cursor);
-	pr_info("sema-svc:   pages=%llu pushes=%llu pending=%llu nomap=%llu memset=%llu/%llu/%llu ringops=%llu denied=%llu false=%llu\n",
+	pr_info("sema-svc:   pages=%llu pushes=%llu pending=%llu nomap=%llu memset=%llu/%llu/%llu ringops=%llu denied=%llu false=%llu skip=%llu\n",
 		svc.pages_decoded, svc.pushes_seen, svc.pushes_pending,
 		svc.pushes_nomap, svc.memsets_done, svc.memsets_nomap,
 		svc.memsets_fail, svc.ring_ops, svc.ops_denied,
-		svc.false_trans);
+		svc.false_trans, svc.gpfifo_skipped);
 	pr_info("sema-svc:   gpfifo done=%llu nomap=%llu vidmem_writes=%llu ring_auto=%u pb=%u/%u\n",
 		svc.gpfifo_entries_done, svc.gpfifo_nomap,
 		svc.vidmem_writes, svc_ring_n, svc_pb_n, svc_pb_passes);
@@ -3357,6 +3359,7 @@ static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 				u8 *pb = kmalloc(len, GFP_KERNEL);
 
 				if (pb) {
+					int rescue_rc;
 					bool unreadable =
 						svc_read_va_trusted(vgpu, va,
 								    pb, len) != 0;
@@ -3408,7 +3411,10 @@ static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 						if (svc.false_trans && svc.false_trans < 8)
 							pr_info("sema-svc: false translation for push va=0x%llx (h=0x%x len=%u) - rescuing\n",
 								va, ch->h_object, len);
-						if (!svc_rescue_push_page(vgpu, ch, va, len)) {
+						rescue_rc = svc_rescue_push_page(vgpu,
+										 ch, va, len);
+						if (!rescue_rc) {
+							ch->rescue_fail = 0;
 							/* Rescue succeeded, retry the read. */
 							if (!svc_read_va_trusted(vgpu, va, pb,
 										 len)) {
@@ -3438,6 +3444,29 @@ static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 								kfree(pb);
 								break;
 							}
+						} else if (rescue_rc == -EAGAIN) {
+							/* Rate-limited: try again later. */
+							svc.gpfifo_nomap++;
+							kfree(pb);
+							break;
+						} else if (++ch->rescue_fail >= 3) {
+							/*
+							 * The slot no longer holds this
+							 * push (overwritten buffer): the
+							 * rescue cannot validate any
+							 * candidate. Consume the entry so
+							 * the pump moves on instead of
+							 * rescanning RAM every 5 s.
+							 */
+							pr_info("sema-chan: skipping stale entry h=0x%x va=0x%llx len=%u after %u failed rescues\n",
+								ch->h_object, va, len,
+								ch->rescue_fail);
+							ch->rescue_fail = 0;
+							ch->last_get = (idx + 1) %
+								       ch->gpfifo_entries;
+							svc.gpfifo_skipped++;
+							kfree(pb);
+							continue;
 						} else {
 							svc.gpfifo_nomap++;
 							kfree(pb);