[BUG] Directory workflow stalls during initial-shell cleanup #132

Open
opened 2026-10-15 17:29:04 +02:00 by erikinkinen · 1 comment
Owner

Summary

The initial-shell directory workflow stalls during cleanup on the updated source graph. It never reaches DIRECTORY_WORKFLOW_OK or the expected successful exit.

Reproduction

Run the unchanged appliance-disk-image-initial-exsh-coreutils-directory-script-positive scenario at Integration 2ca698797761a50901c20d812bdfc3c0efaf8a10. The original release profile uses one QEMU CPU, a writable single-virtio GPT disk, and the development runtime configuration.

Expected and actual behavior

The script deliberately requires rm on a directory and rmdir on a file to fail, then continues with permitted operations. The observed rm: filesystem operation failed is therefore expected.

The trace subsequently records process 0x1f exiting with code 5 and Exsh exiting with 0xe5. Rootd's hangup call, operation 1668 on slot 2636, receives native status 3 and no reply. Rootd reports CLEANUP_UNCERTAIN and exits with 0xd1; the original 45-second serial watchdog ends the run.

Environment and identity

Runtime Integration 975a2e3c, Rootd 635892a8, Vfsd fe7b078a, TTYd 56f9cdf2, Exsh 8540bcf4, Coreutils 78df1388, and Kernel 6f2cc519. Authority remains the scenario's exact create/unlink grants. No last working revision is established.

Evidence and prior checks

One guest execution failed, with its complete serial log, selected images, source bindings and native artifacts captured. Runner retirement succeeded; guest cleanup did not. Delegated page-fault messages alone do not establish a fault cause. The precise service failure remains unidentified.

Tracking

Related: multi-PTY cleanup failure, Vfsd repair. A shared cause has not been established.

  • Identify the first failing service operation from the actual selected artifacts.
  • Repair the cause with a focused regression.
  • Pass the unchanged directory scenario with confirmed guest cleanup.
## Summary The initial-shell directory workflow stalls during cleanup on the updated source graph. It never reaches `DIRECTORY_WORKFLOW_OK` or the expected successful exit. ## Reproduction Run the unchanged `appliance-disk-image-initial-exsh-coreutils-directory-script-positive` scenario at Integration `2ca698797761a50901c20d812bdfc3c0efaf8a10`. The original release profile uses one QEMU CPU, a writable single-virtio GPT disk, and the development runtime configuration. ## Expected and actual behavior The script deliberately requires `rm` on a directory and `rmdir` on a file to fail, then continues with permitted operations. The observed `rm: filesystem operation failed` is therefore expected. The trace subsequently records process `0x1f` exiting with code 5 and Exsh exiting with `0xe5`. Rootd's hangup call, operation 1668 on slot 2636, receives native status 3 and no reply. Rootd reports `CLEANUP_UNCERTAIN` and exits with `0xd1`; the original 45-second serial watchdog ends the run. ## Environment and identity Runtime Integration `975a2e3c`, Rootd `635892a8`, Vfsd `fe7b078a`, TTYd `56f9cdf2`, Exsh `8540bcf4`, Coreutils `78df1388`, and Kernel `6f2cc519`. Authority remains the scenario's exact create/unlink grants. No last working revision is established. ## Evidence and prior checks One guest execution failed, with its complete serial log, selected images, source bindings and native artifacts captured. Runner retirement succeeded; guest cleanup did not. Delegated page-fault messages alone do not establish a fault cause. The precise service failure remains unidentified. ## Tracking Related: [multi-PTY cleanup failure](https://git.erikinkinen.fi/erix/integration/issues/125), [Vfsd repair](https://git.erikinkinen.fi/erix/vfsd/issues/33). A shared cause has not been established. - [ ] Identify the first failing service operation from the actual selected artifacts. - [ ] Repair the cause with a focused regression. - [ ] Pass the unchanged directory scenario with confirmed guest cleanup.
Author
Owner

The unchanged directory scenario on Integration f467304f, using Rootd ff0bd1ef and Kernel e463d945, reaches the workflow and successful-exit markers. It still fails its original checker because Kernel process-exit output splits the required permission-denied diagnostic; Kernel #55 tracks that defect. All host processes retired and the complete failed capture is preserved. This result does not establish the cause of the earlier service-exit and cleanup failure, so this issue remains open.

The unchanged directory scenario on [Integration f467304f](https://git.erikinkinen.fi/erix/integration/commit/f467304fc8f99f4b7d6cd7e23a1254d08e510cef), using Rootd ff0bd1ef and Kernel e463d945, reaches the workflow and successful-exit markers. It still fails its original checker because Kernel process-exit output splits the required permission-denied diagnostic; [Kernel #55](https://git.erikinkinen.fi/erix/kernel/issues/55) tracks that defect. All host processes retired and the complete failed capture is preserved. This result does not establish the cause of the earlier service-exit and cleanup failure, so this issue remains open.
erikinkinen added this to the Alpha 1 milestone 2026-10-30 09:34:17 +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/integration#132
No description provided.