sema-svc: fake_rel applies releases immediately (skip multi-second CE rescue scans)

1 files changed, 54 insertions, 41 deletions

diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c
index 2127f38..7475ea7 100644
--- a/kernel/gsp_sema_service.c
+++ b/kernel/gsp_sema_service.c
@@ -3671,6 +3671,39 @@ static void svc_counter_publish(struct nv_vgpu_device *vgpu,
 			    ch->h_object, ch->counter_pa, val);
 }
 
+/*
+ * Acknowledge a CE tracking-semaphore release without reading the push that
+ * carried it: write the expected payload into the semaphore the channel last
+ * released to (learned from a readable push; the pool index is allocated by
+ * UVM and does not match the channel rank). Returns false when there is
+ * nothing to fall back on. Caller owns expected_sem_payload bookkeeping.
+ */
+static bool svc_fake_release(struct nv_vgpu_device *vgpu, struct svc_channel *ch)
+{
+	u64 pool_va;
+	u64 spa = 0;
+	int sap = SVC_APERTURE_SYS;
+	u32 scur = 0;
+
+	if (!ch->engine_type || !ch->expected_sem_payload)
+		return false;
+
+	pool_va = ch->last_rel_va ? ch->last_rel_va :
+		  0x121000000ULL + 4ULL * ch->ce_index;
+
+	if (!svc_lookup_trusted(vgpu, pool_va, &spa, &sap) &&
+	    !svc_read_ap(vgpu, spa, sap, &scur, sizeof(scur)) &&
+	    scur < ch->expected_sem_payload) {
+		u32 nv = ch->expected_sem_payload;
+
+		if (!svc_write_ap(vgpu, spa, sap, &nv, sizeof(nv)))
+			svc.releases_applied++;
+	}
+	svc.releases_faked++;
+	ch->expected_sem_payload++;
+	return true;
+}
+
 static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 {
 	/* Only ever called from the service thread. */
@@ -3837,6 +3870,24 @@ static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 						 * its whole slot.
 						 */
 						svc_log_stalled_va(va);
+						/*
+						 * With the fallback enabled the rescue
+						 * is pointless for CE: it never
+						 * executes the work anyway (only the
+						 * release), and its multi-second RAM
+						 * scan is what made the driver's
+						 * progress waits expire. Acknowledge
+						 * the release and move on.
+						 */
+						if (gsp_sema_fake_rel &&
+						    svc_fake_release(vgpu, ch)) {
+							ch->last_get = (idx + 1) %
+								       ch->gpfifo_entries;
+							svc.gpfifo_skipped++;
+							stalled = true;
+							kfree(pb);
+							continue;
+						}
 						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);
@@ -3915,47 +3966,9 @@ static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 								ch->h_object, va, len,
 								ch->rescue_fail);
 							ch->rescue_fail = 0;
-							/*
-							 * The entry carried a tracking-
-							 * semaphore release we could not
-							 * read. Acknowledge it the way the
-							 * GPU would so the driver's wait
-							 * clears in one pass.
-							 */
-							if (gsp_sema_fake_rel &&
-							    ch->engine_type &&
-							    ch->expected_sem_payload) {
-								u64 pool_va =
-									ch->last_rel_va ?
-									ch->last_rel_va :
-									0x121000000ULL +
-									4ULL * ch->ce_index;
-								u64 spa = 0;
-								int sap =
-									SVC_APERTURE_SYS;
-								u32 scur = 0;
-
-								if (!svc_lookup_trusted(
-									    vgpu,
-									    pool_va,
-									    &spa,
-									    &sap) &&
-								    !svc_read_ap(vgpu,
-										 spa, sap,
-										 &scur,
-										 sizeof(scur)) &&
-								    scur < ch->expected_sem_payload) {
-									u32 nv =
-										ch->expected_sem_payload;
-
-									if (!svc_write_ap(vgpu,
-											  spa, sap,
-											  &nv,
-											  sizeof(nv)))
-										svc.releases_applied++;
-								}
-								svc.releases_faked++;
-							}
+							if (gsp_sema_fake_rel)
+								svc_fake_release(vgpu,
+										 ch);
 							ch->last_get = (idx + 1) %
 								       ch->gpfifo_entries;
 							svc.gpfifo_skipped++;