vgsp-sema: re-arm push-buffer and GR-ring discovery with the session (ISSUE-0002)

The push-buffer pairing bitmap (svc_pb_done), its pass budgets
(svc_pb_passes/svc_pb_idle/svc_pb_last*) and the GR ring bitmap
(svc_gr_ring_done) are boot-scoped, but svc_session_flush() only wiped
the pins they stood for. The next client reuses the same CE/GR ranks,
so discovery skipped every rank, its push pages could never be paired,
and its decoded releases were dropped ('no slot registered'): the
guest waited on a tracking semaphore that could never advance - the
second-client hang. Re-arm the discovery state in the flush.

1 files changed, 16 insertions, 0 deletions

diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c
index f46158e..39e0096 100644
--- a/kernel/gsp_sema_service.c
+++ b/kernel/gsp_sema_service.c
@@ -6401,6 +6401,22 @@ static void svc_session_flush(void)
 	svc_grce_dst_ring_next = 0;
 	svc_ring_done = 0;
 	/*
+	 * Push-buffer and GR-ring discovery are boot-scoped bitmaps (and pass
+	 * budgets): a rank marked done stays marked even though this flush
+	 * just wiped the pins the mark stood for. The next client reuses the
+	 * same ranks, so discovery would skip every one of them and its push
+	 * pages could never be paired - the observed second-client hang where
+	 * its releases are dropped ("no slot registered") and the guest waits
+	 * forever. Re-arm the discovery state with the session (ISSUE-0002).
+	 */
+	svc_pb_done = 0;
+	svc_pb_passes = 0;
+	svc_pb_idle = 0;
+	svc_pb_last = 0;
+	svc_pb_last_n = 0;
+	svc_gr_ring_done = 0;
+	svc_ring_n = 0;
+	/*
 	 * The learned tables are session state too: the primary VA->PA pin
 	 * table (consulted before every other source), the counter-derived
 	 * cache (serves a cached PA without re-proving it) and the hot-page