bootc: Handle ENOSYS from emulated syscalls in container lint instead of failing (#1481) #311

Activity

1. cgwalters-bot

   added

   target:bootc-dev

   P2

   on Oct 3, 2026
2. cgwalters-bot

   added this to Workstreamon Oct 3, 2026
3. cgwalters-bot

   added

   dispatchOpt in to the dispatcher's auto-dispatch of a devspace runOpt in to the dispatcher's auto-dispatch of a devspace run

   and removed

   dispatchOpt in to the dispatcher's auto-dispatch of a devspace runOpt in to the dispatcher's auto-dispatch of a devspace run

   on Oct 3, 2026
4. cgwalters-bot

   added a parent issue on Oct 3, 2026

   • Epic: bootc and ostree family (parked) #333
5. cgwalters-forge

   locked as resolved and limited conversation to collaborators on Oct 18, 2026
6. cgwalters-bot commented on Oct 22, 2026

   

   cgwalters-bot

   on Oct 22, 2026

   Last edited by cgwalters-bot

   CollaboratorAuthor

   More actions

   ✅ dispatch ended well. Result: #311 (comment)
7. cgwalters-bot commented on Oct 22, 2026

   

   cgwalters-bot

   on Oct 22, 2026

   CollaboratorAuthor

   More actions

   Triage (21 Oct 2026): valid and reproducible, but the brief points at the wrong syscall and the wrong repository. Needs the operator to pick a direction.

   What actually fails. The ENOSYS is from openat2, not set_robust_list. amnabbouti traced it in bootc#1481 (25 Aug 2026): QEMU linux-user 8.2.9 reports unsupported aarch64 syscall 437 (openat2) right before the error, and the same image passes under QEMU 10.2.2. QEMU gained openat2 in linux-user with commit 9651cead (Oct 2024), shipped in 9.2, motivated by the same symptom in bootc-image-builder#619. The original reporter's "fix" was a QEMU built from git master, which already had openat2; the set_robust_list stub was incidental (glibc ignores that call's result).

   Where the call lives. Not in bootc. Both failing lints reach cap-std-ext's direct openat2(RESOLVE_NO_XDEV | RESOLVE_BENEATH) in impl_open_dir_noxdev, which maps EXDEV to Ok(None) but passes ENOSYS through:

   • utf8 is a recursive lint run through walk_configuration().noxdev(), and cap-std-ext's walk_inner opens every child directory with that function.
   • var-tmpfiles calls open_dir_noxdev in crates/tmpfiles/src/lib.rs, and its path resolution goes through RootDir, whose open_beneath_rdonly is also a bare openat2(RESOLVE_IN_ROOT).

   cap-std-ext already tolerates ENOSYS from statx in is_mountpoint (returns Ok(None)), so there is precedent, but open_dir_noxdev has no fallback. cap-std#423 asks cap-primitives for a RESOLVE_NO_XDEV-aware manual resolver; sunfishcode said cap-std has never offered that, and cap-std#426 is about archiving cap-std, so nothing will come from there. bootc pins cap-std-ext 5.1.2 and cap-primitives 4.0.2; no ENOSYS handling exists in bootc's tree today.

   Why this is the operator's call. The brief ("skip the check with a clear warning on ENOSYS") conflicts with cgwalters's own upstream reply (5 Aug 2025), which prefers a stat-plus-dev/inode fallback over not running the lints, and the fallback belongs in coreos/cap-std-ext, which is not in this tracker's dispatch bounds (.github/agentic-job/dispatch.toml lists only the cgwalters-forge/bootc fork). The epic this sits under (#333) is parked. The options:

   1. Close as not planned, with a docs note. QEMU 9.2 and later implement openat2, so the environment fix exists; a line in the lint docs saying "under QEMU user-mode emulation older than 9.2, use --skip var-tmpfiles --skip utf8 or upgrade QEMU" covers the remaining users (Ubuntu 24.04 still ships 8.2). This is the cheapest option and matches cgwalters's "happy to take a patch for that" on documentation.
   2. Fix in cap-std-ext, then bump bootc. In impl_open_dir_noxdev and open_beneath_rdonly, on Errno::NOSYS fall back to a component-by-component openat(O_DIRECTORY | O_NOFOLLOW) with a statx MOUNT_ROOT check (or st_dev comparison) per component, returning Ok(None) on a crossing. That is one cap-std-ext PR plus a bootc dependency bump, and needs coreos/cap-std-ext added to the dispatch bounds or a run by hand.
   3. Do the brief as written in bootc. Catch ErrorKind::Unsupported or Errno::NOSYS in the lint runner and report the lint as skipped with a warning. Smallest change, but it silently weakens utf8 and var-tmpfiles on old emulators, which cgwalters argued against upstream.

   Recommendation: option 1, with option 2 if a code fix is wanted. I will not plan anything until one is chosen.

   Generated by dispatch · ◷