On Fri, Dec 6, 2024 at 9:37 PM David Disseldorp <ddiss@xxxxxxx> wrote:
>
> On Fri, 6 Dec 2024 16:35:18 +0900, Namjae Jeon wrote:
> ...
> > > 300 WARN_ON_ONCE(ia_valid & ATTR_MTIME);
> > > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^--- here.
>
> I should mention, fstests generic/215 atop a 6.13.0-rc1 cifs.ko SMB3
> mount was the trigger for this ksmbd warning.
Okay, I have reproduced it. Thanks. I will take a look.
>
> > > Looking at smb2pdu.c:set_file_basic_info() it's easy enough to see how
> > > we can get here with !ATTR_CTIME alongside ATTR_MTIME.
> > >
> > > The following patch avoids the warning, but I'm not familiar with this
> > > code-path, so please let me know whether or not it makes sense:
> > mtime and atime will probably not be updated.
>
> Unless I'm missing something, this patched ksmbd still triggers mtime
> update via the setattr_copy_mgtime()->(ia_valid & ATTR_MTIME_SET) path.
If FS_MGTIME flag is not set in the fstype of local filesystem, mtime
is not updated with only ATTR_MTIME_SET.
Thanks.
>
> > I will change it so that ATTR_CTIME is also set when changing mtime.
>
> That should also work. I was turned off that path due to the
> 64e7875560270 ("ksmbd: fix oops from fuse driver") ATTR_CTIME changes.
>
> Thanks, David
