vgsp-sema: bounded fallback registration for a parked VAS candidate (15 s)

The VAS candidate verification needs the client's ring to be found by the
RAM content scan; on a boot where that never anchors the ring the
candidate stays parked forever and every launch operand walk fails rc=-2
(observed: pd_found=0, vas[0] holding an internal pdb, walks stuck at
9555 successes with 114 K attempts while the client spun for minutes).
After 15 s parked, register the candidate unverified with a loud log: it
came from the client's own RM publish and the newest-first walk order
keeps an older stale directory from winning, so the bounded wait's
registration is the lesser evil.

1 files changed, 49 insertions, 0 deletions

diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c
index dcce4b5..cd7ff73 100644
--- a/kernel/gsp_sema_service.c
+++ b/kernel/gsp_sema_service.c
@@ -1674,6 +1674,7 @@ static struct {
 	u64 pdb;
 	u32 entries;
 	bool vidmem;
+	unsigned long stamp;	/* r537: park time for the bounded fallback */
 } svc_vas_cand[SVC_VAS_CAND_MAX];
 
 static int svc_gmmu_walk(struct nv_vgpu_device *vgpu, u64 pdb, u32 entries,
@@ -1697,12 +1698,53 @@ void gsp_sema_service_vaspace_candidate_add(u32 h_vaspace, u64 pdb,
 			svc_vas_cand[i].pdb = pdb;
 			svc_vas_cand[i].entries = entries;
 			svc_vas_cand[i].vidmem = vidmem;
+			svc_vas_cand[i].stamp = jiffies;
 			return;
 		}
 	}
 }
 
 /*
+ * r537 bounded fallback: register a parked candidate that has waited for
+ * its content-verified ring walk longer than SVC_VAS_CAND_AGE_MS. The
+ * verification (svc_vas_candidates_verify) needs the client's ring to be
+ * found by the RAM content scan; on a boot where that scan never anchors
+ * the ring (observed: pd_found=0 and every launch operand walk rc=-2 for
+ * minutes, the client spinning with its data VAs unresolvable) the
+ * candidate stays parked forever and the client can never proceed. The
+ * candidate itself came from the client's own RM publish (its VAS
+ * handle), and the r459 newest-first walk order keeps an older stale
+ * directory from winning, so after the bounded wait the registration is
+ * the lesser evil; it is logged loudly for the record.
+ */
+#define SVC_VAS_CAND_AGE_MS	15000u
+
+static void svc_vas_candidates_age_out(struct nv_vgpu_device *vgpu)
+{
+	int i;
+
+	for (i = 0; i < SVC_VAS_CAND_MAX; i++) {
+		struct sema_vaspace_desc d;
+
+		if (!svc_vas_cand[i].valid)
+			continue;
+		if (time_before(jiffies, svc_vas_cand[i].stamp +
+				msecs_to_jiffies(SVC_VAS_CAND_AGE_MS)))
+			continue;
+		memset(&d, 0, sizeof(d));
+		d.h_vaspace = svc_vas_cand[i].h_vaspace;
+		d.pdb = svc_vas_cand[i].pdb;
+		d.num_entries = svc_vas_cand[i].entries;
+		d.b_vidmem = svc_vas_cand[i].vidmem;
+		pr_warn("sema-svc: VAS 0x%x page directory registered UNVERIFIED after %u ms parked (pdb=0x%llx entries=%u; the content-verified ring walk never anchored this boot)\n",
+			d.h_vaspace, SVC_VAS_CAND_AGE_MS, d.pdb,
+			d.num_entries);
+		gsp_sema_service_vaspace_add(&d);
+		svc_vas_cand[i].valid = false;
+	}
+}
+
+/*
  * Verify each parked candidate against one content-validated translation
  * (the ring VA -> PA the RAM scan just proved): walk the ring page through
  * the candidate and register only the directory that reproduces it. Runs on
@@ -24296,6 +24338,13 @@ static int svc_thread(void *arg)
 			}
 
 			/*
+			 * r537: bounded fallback for a parked VAS candidate
+			 * whose content-verified ring walk never anchored
+			 * (see svc_vas_candidates_age_out).
+			 */
+			svc_vas_candidates_age_out(vgpu);
+
+			/*
 			 * Channel-driven execution: notice submissions as they
 			 * are made and run the pushed work immediately, so the
 			 * guest's UVM progress waits complete in milliseconds