Two cleanup/parsing nits: nvc_cfg NULL deref on OOM paths, sscanf %x into int

Aperta
#409 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
2/5
Tempo stimato
1-3 ore
Idoneità per principianti
58/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
c
Ambito
cli

Direzione di ricerca

Read the cleanup paths in src/cli/configure.c and src/cli/list.c, then inspect matches_pci_format in src/cli/common.c. Guard the nvc_cfg cleanup when it is NULL and use unsigned variables for the %x conversions; done means both issues are corrected and the project builds cleanly.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

needs-triage

Two smaller hardening notes from the same audit pass that reported #408 — bundling them in one issue since neither is urgent on its own.

1. nvc_cfg can be NULL on the OOM cleanup paths

src/cli/configure.c:481 and src/cli/list.c:293 both do:

if ((nvc = libnvc.context_new()) == NULL || (nvc_cfg = libnvc.config_new()) == NULL) { warn(...); goto fail; }
...
fail:
        free(nvc_cfg->imex.chans);   /* nvc_cfg may be NULL here */

If context_new() fails (calloc returns NULL under memory pressure — runc hooks do run under container memory limits), the fail: path dereferences nvc_cfg->imex.chans with nvc_cfg == NULL. The sibling cleanups all guard their NULLs (libnvc.shutdown(nvc) etc. handle it); these two spots don't. Fix is the obvious guard:

-        free(nvc_cfg->imex.chans);
+        if (nvc_cfg != NULL) {
+            free(nvc_cfg->imex.chans);
+        }

Haven't reproduced at runtime (would need calloc failure injection), but the deref is clear from the code.

2. sscanf("%x") into int in matches_pci_format

src/cli/common.c:18:

sscanf(gpu, "%x:%02x:%02x.0", &domainid, &busid, &deviceid);

with int domainid, busid, deviceid. The upstream build log itself carries the gcc warning — %x wants unsigned int *, not int *. Input is attacker-controlled (NVIDIA_VISIBLE_DEVICES → --device=), and values ≥ INT_MAX are undefined per the C standard. In practice glibc's scanf just folds the value, and sanitized runs with ffffffff:99:99.0, ffffffffffffffff:99:99.0, etc. all produced clean "unknown device" rejections — no UBSan/ASan report, so this is paper UB rather than a live fault on glibc. Fix is to make the three variables unsigned int.

Both spotted during a full make debug (ASan/UBSan) audit pass of v1.20.2 (c7727e5); the sanitized end-to-end configure run itself was clean, so these are the only cleanup/parsing nits left standing from that pass besides #408.

Lingua principale
C
Stelle
1.1k
Fork
278
Merge medio
3g 7h
PR unite (30g)
2

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di NVIDIA/libnvidia-container

Tutte le issue di NVIDIA/libnvidia-container

Issue simili

Altre issue su C

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.