[BUG] Exact-entry effects precede status reply capacity admission #13

Closed
opened 2026-10-01 12:54:53 +02:00 by erikinkinen · 0 comments
Owner

Summary

FAT fixed-entry mutation and touch dispatch check status reply capacity after the operation. A too-small local reply buffer can therefore reject a response after changing media or consuming the retained entry. This is a source-confirmed reply-publication defect; no executed reproduction is claimed.

Reproduction

At dcdd48437fba7646e20e5462c314363ade9ee784, admit a private fixed-entry mkdir or touch grant, encode its valid V1 or V2 follow-up, and call provider dispatch with an output slice shorter than VFS_STATUS_RESPONSE_V1_SIZE. Inspect the dispatch ordering: it calls the effect before encoding the status. The new regression fixtures exercise every short output length on selected FAT12 and exFAT, then retry with a complete response buffer and inspect the created object. They are authored and unrun.

Expected and actual behavior

Expected: reject insufficient response capacity before media I/O, handle changes or one-shot consumption, leaving the grant usable for a retry.

Actual source ordering: the operation may complete before the status encoder reports insufficient capacity. No runtime output, exit status or frequency is asserted.

Environment and identity

  • Component/revision: fatd dcdd48437fba7646e20e5462c314363ade9ee784, fixed-entry dispatch in src/dispatch.rs and src/dispatch/timestamped.rs.
  • Integration source graph and image identity: not exercised; this report is limited to the provider source boundary.
  • Hardware, firmware and toolchain: no execution involved.
  • Route and authority: the existing private provider endpoint and already admitted fixed-entry operation; the defect is local reply-capacity ordering.
  • Last known working revision: unknown.

Evidence and prior checks

Source inspection identifies the effect-before-encode ordering in both mutation and touch families. The repair preflights complete status capacity before invoking either effect, preserving existing fail-stop handling. Authored controls compare output bytes, complete handle state and device event counts before retrying the unchanged request. Tests, builds, lint/format checks, VM/profiler execution and CI observation are deferred by user; none is represented as passing.

Tracking

The repair accompanies retained-entry migration, the FAT review, and Phase 6 tracking.

Source repair is signed and pushed in the FAT entry change. Authored controls cover every short status length for V1/V2 creation and touch on selected FAT12/exFAT, preserving media events and complete handle state before a successful-capacity retry. Closed as source-complete under the testing/CI waiver; no executed test, compiler or CI result is asserted.

## Summary FAT fixed-entry mutation and touch dispatch check status reply capacity after the operation. A too-small local reply buffer can therefore reject a response after changing media or consuming the retained entry. This is a source-confirmed reply-publication defect; no executed reproduction is claimed. ## Reproduction At `dcdd48437fba7646e20e5462c314363ade9ee784`, admit a private fixed-entry mkdir or touch grant, encode its valid V1 or V2 follow-up, and call provider dispatch with an output slice shorter than `VFS_STATUS_RESPONSE_V1_SIZE`. Inspect the dispatch ordering: it calls the effect before encoding the status. The new regression fixtures exercise every short output length on selected FAT12 and exFAT, then retry with a complete response buffer and inspect the created object. They are authored and unrun. ## Expected and actual behavior Expected: reject insufficient response capacity before media I/O, handle changes or one-shot consumption, leaving the grant usable for a retry. Actual source ordering: the operation may complete before the status encoder reports insufficient capacity. No runtime output, exit status or frequency is asserted. ## Environment and identity - Component/revision: fatd `dcdd48437fba7646e20e5462c314363ade9ee784`, fixed-entry dispatch in `src/dispatch.rs` and `src/dispatch/timestamped.rs`. - Integration source graph and image identity: not exercised; this report is limited to the provider source boundary. - Hardware, firmware and toolchain: no execution involved. - Route and authority: the existing private provider endpoint and already admitted fixed-entry operation; the defect is local reply-capacity ordering. - Last known working revision: unknown. ## Evidence and prior checks Source inspection identifies the effect-before-encode ordering in both mutation and touch families. The repair preflights complete status capacity before invoking either effect, preserving existing fail-stop handling. Authored controls compare output bytes, complete handle state and device event counts before retrying the unchanged request. Tests, builds, lint/format checks, VM/profiler execution and CI observation are deferred by user; none is represented as passing. ## Tracking The repair accompanies [retained-entry migration](https://git.erikinkinen.fi/erix/vfsd/issues/16), [the FAT review](https://git.erikinkinen.fi/erix/fatd/pulls/3), and [Phase 6 tracking](https://git.erikinkinen.fi/erix/integration/issues/65). Source repair is signed and pushed in [the FAT entry change](https://git.erikinkinen.fi/erix/fatd/commit/5a4e67af7653f6dbc4c24e4e4f319f3f2ad6353b). Authored controls cover every short status length for V1/V2 creation and touch on selected FAT12/exFAT, preserving media events and complete handle state before a successful-capacity retry. Closed as source-complete under the testing/CI waiver; no executed test, compiler or CI result is asserted.
erikinkinen added this to the Alpha 1 milestone 2026-10-30 09:34:00 +01:00
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
erix/fatd#13
No description provided.