[BUG] Native scope Close refuses retained children and transient cleanup stops serving #29
Labels
No labels
alpha-1
bug
ci
docs
duplicate
enhancement
help wanted
invalid
performance
phase-6
question
refactor
security
wontfix
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
erix/vfsd#29
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Suspected security vulnerabilities must be reported privately using SECURITY.md. This report concerns source-confirmed Close progress and transient cleanup handling; no runtime reproduction is claimed.
Summary
The native scope Close path refuses disposal while a retained child candidate exists, including the candidate held by a blocked advisory Wait. Separately, transient provider cleanup refusal propagates out of native serving to the runtime fail-stop path. Both prevent the intended original Close from completing and draining its own scope. Fair intake alone does not remove those restrictions.
Reproduction
Source-only trace at dd1a7b8a: admit an original acquisition scope and file candidate, independently publish a peer's conflicting interval, claim a blocking Wait, then queue Close with the endpoint-implied zero handle selector. The Close parser calls ensure_no_acquisitions and returns BUSY while that candidate remains. For a separate provider refusal, close_exact_object returns BUSY and the serving error reaches runtime's fail_stop.
The authored production-path controls are scope_close_retries_original_children_without_publishing_older_cached_results and closing_receiver_generation_is_revalidated_before_provider_retry. They use independent native owners and real scope/provider fixture state. Execution, frequency in an actual image and clean-image reproduction remain unmeasured under the current testing waiver.
Expected and actual behavior
Expected: dispose only the complete original scope and its own candidates, abandon older replies before fallible cleanup, retain the original Close/native owner and retry transient cleanup without clearing semantic obligations. Caller retirement cannot prove provider cleanup, and private revocation cannot restore any abandoned reply.
Actual prior source: retained children produce a permanent BUSY guard; provider BUSY/RETRY can stop serving rather than retain bounded original Close progress. The signed source repair removes that guard, consolidates scope disposal and retains an explicit Closing phase through transient refusal. Original association and receiver generation validate before every provider retry. Private revocation may suppress Close's reply while cleanup still finishes.
Environment and identity
Evidence and prior checks
Source inspection follows exact_dispatch, close_exact_object, retained original Call custody and runtime's error path. Repair adds original Closing progress, complete original child disposal, selective older-reply abandonment and validation before provider effects. Twelve fair-intake controls are authored and unrun, including retained cleanup through provider refusal, native caller loss, private reply abandonment and generation contradiction. The existing provider-refusal control now expects retained serving progress. Rustfmt applied without diagnostics. Tests, builds, lint, generated documentation, VM/profiler and CI observation remain deferred. No executed pass or warning-free build is claimed.
Tracking
Native serving feature, local stream sockets and concurrent mediation, Vfsd review, Docs review, Integration review, complete phase requirements.
Source repair is complete under the testing waiver; the wider mediator and guest-build requirements remain open.
erikinkinen referenced this issue2026-10-04 10:25:52 +02:00
erikinkinen referenced this issue from erix/posixd2026-10-04 10:25:52 +02:00
erikinkinen referenced this issue from erix/docs2026-10-04 10:25:52 +02:00
erikinkinen referenced this issue from erix/integration2026-10-04 10:25:52 +02:00