diff options
| author | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-18 12:00:52 -0400 |
|---|---|---|
| committer | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-18 12:10:03 -0400 |
| commit | 04aad03adc45683d03fe1451b8f234061e6b88ff (patch) | |
| tree | bd354b50006a57bf54a302ba6a086fdca137f902 /kernel/gsp_sema_service.c | |
| parent | 2658fedd556f872f11dd3aeefa88ff6eb76c5342 (diff) | |
| download | open-iov-04aad03adc45683d03fe1451b8f234061e6b88ff.tar.gz open-iov-04aad03adc45683d03fe1451b8f234061e6b88ff.zip | |
vgsp: fail4 discrimination tooling - copy-out page witnesses + full-boot bigcopy resolution budget
Four guest CUDA clients fail all-bad while four pass; the aggregate
counters cannot separate 'engine result never produced' from 'produced
but written to a different guest page than the client's read resolves'.
Add the two per-page witnesses that correlate the write side with the
read side:
- gsp_phys_gr_copyout_dump (booter, default 1): per-changed-page
OUT-PAGE-WIT va=<device VA> hpa=<guest PA written> first8/last8=<bytes
shipped>, bounded 12 pages per launch, plus one RELOC-CENSUS line per
launch (gaps vs late-resolved words vs late pages).
- sema_bigcopy_log (sema service, default 24, was a hardcoded per-boot
budget): the svc_copy large-copy proof already logs both operands'
resolved VA -> pa/aperture/provenance and the first 16 bytes each side;
raising the budget past the boot's head makes the failing clients'
H2D/D2H resolutions visible.
Discrimination test for the next cycle: same device VA on OUT-PAGE-WIT
(copy-out HPA) and sema-bigcopy (D2H source pa) with different PAs proves
the produced-but-misplaced class; equal PAs redirects to the readback
content/engine-input path.
(cherry picked from commit 78ea8b7807231ac418a17b85c29ed1845e1b0062)
Diffstat (limited to 'kernel/gsp_sema_service.c')
| -rw-r--r-- | kernel/gsp_sema_service.c | 14 |
1 files changed, 13 insertions, 1 deletions
diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c index fb00685..b354437 100644 --- a/kernel/gsp_sema_service.c +++ b/kernel/gsp_sema_service.c @@ -5265,7 +5265,8 @@ static int svc_copy(struct nv_vgpu_device *vgpu, u64 dst, u64 src, size_t total, * these copies; this separates a wrong resolution from a wrong copy. */ if ((total >= (1u << 20) || - (dst_virt && dst_orig >= SVC_VAS_FLOOR)) && bigcopy_log < 24) { + (dst_virt && dst_orig >= SVC_VAS_FLOOR)) && + bigcopy_log < gsp_sema_bigcopy_log_max) { u8 sb[16] = {0}, db[16] = {0}; bigcopy_log++; @@ -10838,6 +10839,17 @@ bool gsp_sema_copy_dst_pin_guard = 1; * just-built client table is fresh), logged loudly. */ unsigned gsp_sema_copy_dst_guard_ms = 3000; + +/* + * COPYWIT fail4: per-boot budget for the sema-bigcopy resolution witnesses + * (svc_copy's large-copy proof: both operands' VA -> pa/aperture/provenance + * plus the first 16 bytes each side). The shipped default kept the boot's + * head only; a failing-client run happens minutes later, so the budget is a + * module parameter now - raise it for a discrimination cycle and correlate + * each D2H's resolved source PA with the booter's OUT-PAGE-WIT copy-out PA + * for the same device VA. 0 disables the lines. + */ +unsigned gsp_sema_bigcopy_log_max = 24; #define SVC_DST_GUARD_SLOTS 16 struct svc_dst_guard { u64 va_page; |
