diff options
| author | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-23 16:59:57 -0400 |
|---|---|---|
| committer | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-23 16:59:57 -0400 |
| commit | 0f73f4e69344eab4ffebaf310ba2d937b4471ea9 (patch) | |
| tree | 911e726666a7d74ec6747c8623d8bfddc7748579 /kernel | |
| parent | 127d41aae5a68f07c87785193cdf17469412ba1b (diff) | |
| download | open-iov-0f73f4e69344eab4ffebaf310ba2d937b4471ea9.tar.gz open-iov-0f73f4e69344eab4ffebaf310ba2d937b4471ea9.zip | |
vgsp-sema r563c: the stall watchdog gates on ACTIVE withheld work, not cumulative counters
r566 evidence: the dump fired 4x with done=8929 and opdef=215/7/5 frozen
while the ring was EMPTY (every channel last_get == put - the clients were
waiting, no entry pending) and the forced flush had nothing to deliver:
is a lifetime counter (215 holds applied/expired long ago),
not proof of a current hold. A real stall has either a pending translation
(svc_pending_n) or an ACTIVE opdef slot with releases recorded
(e->active && e->n); the dump (and with it the self-heal arm) now requires
one of those. The r564-class true stalls (active slots holding releases)
still dump and still heal.
Diffstat (limited to 'kernel')
| -rw-r--r-- | kernel/gsp_sema_service.c | 24 |
1 files changed, 22 insertions, 2 deletions
diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c index 25ae31f..d47ec08 100644 --- a/kernel/gsp_sema_service.c +++ b/kernel/gsp_sema_service.c @@ -25108,8 +25108,28 @@ static void svc_stall_watch_check(void) svc_stall_watch_last = jiffies; /* re-arm the window */ if (svc_stall_watch_dumps >= SVC_STALL_WATCH_DUMPS) return; - if (!svc_pending_n && !svc.opdef_held) - return; /* idle boot, not a stall */ + /* + * r563c: gate on ACTIVE withheld work, not on the cumulative counters. + * r566 evidence: the dump fired four times with done=8929 and + * opdef=215/7/5 frozen - but the ring was EMPTY (every channel + * last_get == put: the clients were simply waiting and no entry was + * pending), and the forced flush had nothing to deliver. `opdef_held` + * is a lifetime counter (215 holds applied/expired long ago), not + * proof of a current hold. A real stall has either a pending + * translation (svc_pending_n) or an ACTIVE opdef slot with releases + * recorded (e->active && e->n) - only then is there something the + * pump is withholding. + */ + { + bool work = svc_pending_n != 0; + u32 wi; + + for (wi = 0; wi < SVC_OPDEF_SLOTS && !work; wi++) + if (svc_opdef[wi].active && svc_opdef[wi].n) + work = true; + if (!work) + return; + } svc_stall_watch_dumps++; svc_stall_watch_dump(); /* |