[Bug 281033] [msdosfs] rm -f fails

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=281033

Stefan Eßer <se@FreeBSD.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |se@FreeBSD.org
           Assignee|fs@FreeBSD.org              |se@FreeBSD.org
             Status|New                         |Open

--- Comment #1 from Stefan Eßer <se@FreeBSD.org> ---
The patch that has been discussed as GitHub pull request 1329
(https://github.com/freebsd/freebsd-src/pull/1329) will be attached to this PR.

It has an undesirable side effect: the mv command returns "No such file or
directory" instead of "Invalid argument" when the target file name is invalid:

# touch /boot/efi/A.DAT
# mv /boot/efi/A.DAT /boot/efi/B*.DAT
mv: rename /boot/efi/A.DAT to /boot/efi/B*.DAT: No such file or directory

This is caused by namei() being called with the target filename and that
invokes the patched msdosfs_lookup() and thus gets and returns the changed
error code.

But since the patch affects namei() and functions that call it, more vnode
operations will return ENOENT instead of ```EINVAL`` with this patch (when
given an invalid MSDOS file name on an MSDOSFS file system).

The rename(2) man page mentions both EINVAL and ENOENT, but neither description
mentions the case of an invalid target file name:

[ENOENT] A component of the from path does not exist, or a path prefix of to
does not exist.

[EINVAL] The from argument is a parent directory of to, or an attempt is made
to rename ‘.’ or ‘..’.

Maybe the case of a target name not allowed on some file system should be
added, e.g. one of:

[EINVAL] The from argument is a parent directory of to, or an attempt is made
to rename ‘.’ or ‘..’, or the target filename is invalid.

[ENOENT] A component of the from path does not exist, a path prefix of to does
not exist, or the target filename is invalid.

If changing the return code of rename() (and possibly other system calls that
return the error code of a failed namei() call unmodiified) is acceptable for
this specific error case, I'll commit the patch to -CURRENT, but I'm not sure
whether POLA will allow merging to -STABLE.

-- 
You are receiving this mail because:
You are the assignee for the bug.