On Mon, 2021-02-22 at 11:43 +0100, Miklos Szeredi wrote:
> On Sun, Feb 21, 2021 at 2:30 AM Marios Makassikis
> <mmakassikis@xxxxxxxxxx> wrote:
> > Hello,
> >
> > I hope this is the correct list to report this bug I've been
> > seeing.
> >
> > Background: I am testing a kernel SMB server implementation
> > ("ksmbd": https://github.com/cifsd-team/cifsd).
> >
> > As part of my tests, I tried having a Windows client store a backup
> > on a SMB
> > share that is backed by an NTFS formatted disk. In doing so, the
> > kernel
> > reports a BUG and locks up (either immediately, or after a few
> > minutes).
>
> Seems like fi->write_files list gets corrupted.
>
> Is list debugging turned on?
>
> Can you get a crashdump, and see if the rest of the fi structure is
> okay?
>
> Thanks,
> Miklos
Hello Miklos,
I managed to get a crashdump, but couldn't get the crash utility to
work (I fed it the kernel and the crashdump but it exited without any
error message and I didn't get a chance to figure out what was wrong).
Meanwhile, a fix was committed on ksmbd with which I cannot reproduce
the issue [1]. Previously, the i_mutex lock was not held in
set_file_basic_info(): I'm guessing this is what caused the list
corruption.
Marios
[1]
https://github.com/cifsd-team/cifsd/commit/5e929125e519acaf48abc4c42f8389caa26c4d5a
