diff options
| -rw-r--r-- | kernel/gsp_sema_service.c | 34 | ||||
| -rw-r--r-- | tools/guest-diag/guest_pin_stream.py | 26 |
2 files changed, 60 insertions, 0 deletions
diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c index 24cb034..e86c463 100644 --- a/kernel/gsp_sema_service.c +++ b/kernel/gsp_sema_service.c @@ -21375,6 +21375,40 @@ static bool svc_gr_request(struct nv_vgpu_device *vgpu, return false; } /* + * r586: the defer cap expired with argument pages still unresolved - + * fail-closed like the r583 dstpend cap. The historical path + * forwarded the launch with the partial set and the engine read + * stale/zero operands (r585 boot saxpy: n_data=969 dropped=7 + * partial=1 -> forwarded at the cap -> bad=1047527 csum=130523290; + * r584 matmul: n_data=2 dropped=18 -> r564 degenerate refusal -> + * no-write). Refuse the forward, author the completion in software, + * consume; the dropped pages were already demanded through the pin + * channel (r585) so the retry path had its chance first. + */ + if (svc_fwd_req.gr_arg_partial && !svc_gr_data_stalled && + svc_gr_defer_expired(svc_gr_cand.push_va)) { + svc.fwd_gr_refused++; + svc.fwd_gr_closed++; + if (svc_gr_cand.rel_va) { + svc_gr_sw_release(vgpu, ch->h_object, + svc_gr_cand.rel_va, + svc_gr_cand.rel_payload); + svc_gr_cand.rel_done = true; + } + svc_gr_cand.offered = true; + svc_gr_cand.valid = false; + if (svc.fwd_gr_closed_log < 32) { + svc.fwd_gr_closed_log++; + pr_warn("vgsp-work: GR forward REFUSED at defer cap h=0x%x push=0x%llx (argument pages still unresolved: %u; fail-closed, QMD release authored in software)\n", + ch->h_object, svc_gr_cand.push_va, + svc_fwd_req.gr_dropped); + } + clear_bit_unlock(SVC_FWD_INFLIGHT, &svc_fwd_state); + mutex_unlock(&svc_fwd_lock); + return false; /* no busy-defer: the pump consumes; the + * candidate is owned and sw-completed */ + } + /* * m150: the fill found high argument pointers that no source can * translate (the second-client zeros class). Do NOT forward a launch * whose kernel would dereference unresolved pointers: leave the diff --git a/tools/guest-diag/guest_pin_stream.py b/tools/guest-diag/guest_pin_stream.py index b56f79d..175ed23 100644 --- a/tools/guest-diag/guest_pin_stream.py +++ b/tools/guest-diag/guest_pin_stream.py @@ -1819,6 +1819,32 @@ def main(): except Exception: ok = False got = [] + # r586: a demanded VA outside the CE + # pushbuffer window (the GRFILL staging pages, + # high client VAs) is a plain client mapping + # page: resolve it through the client's own + # pagemap. Without this the demand for those + # VAs was a silent no-op (resolve_va_by_content + # rejects anything outside PB_CE_LO..PB_CE_HI) + # and the fills kept dropping them. + if not ok and not got and pid: + try: + pfd = os.open('/proc/%d/pagemap' % pid, + os.O_RDONLY) + try: + sub = {} + vpg = wv & ~0xFFF + scan_range(pfd, vpg, vpg + 0x1000, + sub) + finally: + os.close(pfd) + if sub: + for v, p in sub.items(): + got.append('0x%x,0x%x,1\n' + % (v, p)) + ok = True + except OSError: + pass if ok and got: lines.extend(got) _want_state["done"].add(wv) |