On 11/16/26 20:28, [email protected] wrote:
Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Low] ovl: commit message does not state the actual fd leak in
   ovl_copyfile()
--

commit 19e71b9ba90a3111ae72b566162542bd1c41d68d
Author: Vladimir Riabchun <[email protected]>

ovl: Fix leaking fd references in ovl_copyfile

We lack of ovl_real_file, added in d66907b[1], which changed
logic so that the code needs no fdput anymore. We still need it.

Imported commit a1e0eb8[2] didn't bother with fdput cleanup,
because it relied on d66907b existence. We need to perform
a cleanup when rw_verify_area fails and put real_in/real_out fds.

[1]: https://github.com/torvalds/linux/commit/d66907b51ba0
[2]: https://github.com/torvalds/linux/commit/a1e0eb8f55cf

https://virtuozzo.atlassian.net/browse/VSTOR-142405
Fixes: d5a10d38e534 ("rh10: import RHEL10 kernel-6.12.0-211.75.1.el10")
Signed-off-by: Vladimir Riabchun <[email protected]>

[Severity: Low]
Could the message state the leak more directly? The sentence "the code
needs no fdput anymore. We still need it." contradicts itself, and the
message never names the path that leaks.

No contradiction. Code with d66907b needs no fdput, but this is not
our code.


As far as I can tell, the leak is in the cross-sb check in ovl_copyfile().
When rw_verify_area() fails, the code jumped to out_unlock. Both
ovl_real_fdget() calls had already succeeded, so this skipped fdput(real_in)
and fdput(real_out). When FDPUT_FPUT is set, that leaks the file
references.

The message could say this in one sentence. It could also name the
upstream commits by 12-character sha and subject instead of the short
shas and bare URLs.

The leak is obvious from the code. I'm not adding useless words to the
commit message.



--
Best regards, Riabchun Vladimir
Linux Kernel Developer, Virtuozzo

_______________________________________________
Devel mailing list
[email protected]
https://lists.openvz.org/mailman/listinfo/devel

Reply via email to