[GH-ISSUE #77] stress: eintr, fifo and unix tests fail instead of UNSUP on nfs and msdos #71

New issue

Closed

opened 2026-11-03 22:03:43 -07:00 by gitea-mirror · 1 comment

gitea-mirror commented

2026-11-03 22:03:43 -07:00

Owner

Originally created by @raynorpat on GitHub (Oct 20, 2026).
Original GitHub issue: https://github.com/RhapsodiOS/RhapsodiOS/issues/77

First all-category runs of the suite on NFS and msdos (combined retest of ffs-perf f68d4802c). Each fails 3 of 3:

misc eintr reports flock-child-failed on nfs and msdos, where flock returns EOPNOTSUPP.
misc fifo reports mkfifo: EINVAL on msdos, which has no fifos.
net unix times out on msdos. It has no socket nodes, so the child's connect fails and the parent blocks in accept.

These are filesystem limits, not bugs: like the existing flock/fcntl, link and rename cases, they should end as UNSUP with the exact errno (st_unsup), and the child's failure must not leave the parent blocked.

Originally created by @raynorpat on GitHub (Oct 20, 2026). Original GitHub issue: https://github.com/RhapsodiOS/RhapsodiOS/issues/77 First all-category runs of the suite on NFS and msdos (combined retest of ffs-perf f68d4802c). Each fails 3 of 3: - `misc eintr` reports `flock-child-failed` on nfs and msdos, where `flock` returns EOPNOTSUPP. - `misc fifo` reports `mkfifo: EINVAL` on msdos, which has no fifos. - `net unix` times out on msdos. It has no socket nodes, so the child's connect fails and the parent blocks in accept. These are filesystem limits, not bugs: like the existing flock/fcntl, link and rename cases, they should end as UNSUP with the exact errno (`st_unsup`), and the child's failure must not leave the parent blocked.

gitea-mirror closed this issue

2026-11-03 22:03:44 -07:00

gitea-mirror commented

2026-11-03 22:03:45 -07:00

Author

Owner

@raynorpat commented on GitHub (Oct 20, 2026):

Fixed by 4610ecd7f, merged into ffs-perf in a9fbc781a. No kernel change.

misc eintr runs the same probe_flock as misc flock, so flock: EOPNOTSUPP makes it UNSUP.
probe_fifo also accepts EINVAL, which is what msdosfs mkfifo returns.
net unix gets a probe (unix_probe): it binds, listens and connects a path socket under -d. ENOTSOCK, EOPNOTSUPP or ENOSYS make the test UNSUP. msdosfs makes the node a regular file, so connect gets ENOTSOCK (uipc_usrreq.c :478).
t_unix's parent now connects before it accepts and forks, so a failed connect is a FAIL, not a parent blocked in accept.
ENOTSOCK was added to the errno names; the README lists the new UNSUP cases.

Verified on the combined kernel (cksum 1504968375), seed 75, --cats "misc net":

Test	nfs	msdos	ufs
misc eintr	FAIL → UNSUP flock: EOPNOTSUPP	FAIL → UNSUP flock: EOPNOTSUPP	PASS
misc fifo	PASS	FAIL → UNSUP mkfifo: EINVAL	PASS
net unix	PASS	FAIL (timeout) → UNSUP connect: ENOTSOCK	PASS

No other result changed. msdos fsck is clean, and make check-misc and check-net pass.

@raynorpat commented on GitHub (Oct 20, 2026): Fixed by 4610ecd7f, merged into ffs-perf in a9fbc781a. No kernel change. - `misc eintr` runs the same `probe_flock` as `misc flock`, so `flock: EOPNOTSUPP` makes it UNSUP. - `probe_fifo` also accepts EINVAL, which is what msdosfs `mkfifo` returns. - `net unix` gets a probe (`unix_probe`): it binds, listens and connects a path socket under `-d`. ENOTSOCK, EOPNOTSUPP or ENOSYS make the test UNSUP. msdosfs makes the node a regular file, so connect gets ENOTSOCK (`uipc_usrreq.c` :478). - `t_unix`'s parent now connects before it accepts and forks, so a failed connect is a FAIL, not a parent blocked in accept. - ENOTSOCK was added to the errno names; the README lists the new UNSUP cases. **Verified** on the combined kernel (cksum 1504968375), seed 75, `--cats "misc net"`: | Test | nfs | msdos | ufs | |---|---|---|---| | misc eintr | FAIL → UNSUP flock: EOPNOTSUPP | FAIL → UNSUP flock: EOPNOTSUPP | PASS | | misc fifo | PASS | FAIL → UNSUP mkfifo: EINVAL | PASS | | net unix | PASS | FAIL (timeout) → UNSUP connect: ENOTSOCK | PASS | No other result changed. msdos fsck is clean, and `make check-misc` and `check-net` pass.