vgsp: eng59b - do not let the arming launch consume the release carrier; re-arm in the pump

The saved-window release carrier was armed inside the pack-completing
launch's own build, and the client-first cancel treated that same launch
(and its retries) as the next client launch, silently dropping the claim.
c762: MME-CARRY-SAVE (win=240 held=68) -> MME-CARRY-ARM mode=release ->
the arming launch ran mac=0 (LOAD-ONLY, window dropped, held CALLs still
owed) -> no FIRE, no synthetic push, no mac=0x22.

Track the arming launch identity (h + push) and keep the claim armed across
it; only the MME-PACK-REARM edge (calls actually released), a real consumer,
context/handle changes, or a launch error cancels/consumes. Add a pump-side
re-arm once pack_done_gen is set, and bounded MME-CARRY-HOLD receipts for
every skip. One fire per pack completion, no retries.

Build rc 0; validation pending (c762-class boot with
sema_gl_mme_pack_release=1).

1 files changed, 22 insertions, 7 deletions

diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c
index 6d72a62..26d3b1a 100644
--- a/kernel/gsp_sema_service.c
+++ b/kernel/gsp_sema_service.c
@@ -50031,15 +50031,19 @@ static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 
 	/*
 	 * Saved-window release carrier (sema_gl_mme_pack_release). After the
-	 * channel loop (a client window decoded in this pass wins and cancels
-	 * the pending release), fire the saved held-call window once the pack
-	 * completed and the settle grace expired. Same gates as the
-	 * continuation carrier, except the FE-busy defer (this launch carries
-	 * a real window, not a LOAD-ONLY carrier).
+	 * channel loop (so a client window decoded in this pass is offered
+	 * first), re-arm the saved held-call window if a race left the record
+	 * valid but disarmed, then fire it once the pack completed and the
+	 * settle grace expired. Same gates as the continuation carrier, except
+	 * the FE-busy defer (this launch carries a real window, not a
+	 * LOAD-ONLY carrier). A real client window never cancels the release:
+	 * only the MME-PACK-REARM edge (calls actually released) does.
 	 */
 	{
 		u32 rel_h = 0;
+		bool found = false;
 
+		gsp_phys_gr_mme_release_rearm();
 		if (gsp_phys_gr_mme_release_peek(&rel_h)) {
 			int ci;
 
@@ -50048,14 +50052,25 @@ static void svc_pump_channels(struct nv_vgpu_device *vgpu)
 
 				if (c->h_object != rel_h || c->engine_type)
 					continue;
-				if (svc_gr_cand_unoffered_for(rel_h))
+				found = true;
+				if (svc_gr_cand_unoffered_for(rel_h)) {
+					pr_info_ratelimited("vgsp-gr: MME-CARRY-HOLD reason=client-candidate h=0x%x\n",
+							    rel_h);
 					break;
+				}
 				if (svc_fwd_hold.active ||
-				    test_bit(SVC_FWD_INFLIGHT, &svc_fwd_state))
+				    test_bit(SVC_FWD_INFLIGHT,
+					     &svc_fwd_state)) {
+					pr_info_ratelimited("vgsp-gr: MME-CARRY-HOLD reason=request-inflight h=0x%x\n",
+							    rel_h);
 					break;
+				}
 				(void)svc_gr_carrier_request(vgpu, c, true);
 				break;
 			}
+			if (!found)
+				pr_info_ratelimited("vgsp-gr: MME-CARRY-HOLD reason=no-channel h=0x%x\n",
+						    rel_h);
 		}
 	}