[Cluster-devel] Mounting lock_nolock file systems?

[Andreas Gruenbacher]

On Wed, Jun 22, 2022 at 4:32 PM Christoph Hellwig <hch at lst.de> wrote:
> On Tue, Jun 21, 2022 at 02:58:57PM +0200, Christoph Hellwig wrote:
> > Sorry, actually it was the pagecache for-next branch from willy.  Looks
> > like mainline itself is fine.
> >
> > I'll try to get the superblock information from the pagecache branch
> > once I find a little time, chasing a bunch of other bugs in the meantime.
>
> I bisected it down to:
>
> commit 1abe0e8c19c514827408ba7e7e84969b6f2e784f
> Author: Matthew Wilcox (Oracle) <willy at infradead.org>
> Date:   Wed May 18 14:41:39 2022 -0400
>
>     gfs: Check PageUptodate instead of PageError
>
>     This is the correct flag to test to know if the read completed
>     successfully.
>
>     Signed-off-by: Matthew Wilcox (Oracle) <willy at infradead.org>
>
> but I don't have any explanation how it caused that breakage yet.

gfs2_find_jhead() uses gfs2_end_log_read() as the end_io function of
the bios it submits, and gfs2_end_log_read() doesn't set the pages it
reads uptodate. That should be fixed; it doesn't make much sense.

Willy, can you remove the above patch from the pagecache tree? We can
put it in the gfs2 tree after that gfs2_end_log_read() fix.

(Side note: it's gfs2, not gfs.)

Thanks,
Andreas