On Wed, Apr 12, 2023 at 05:18:07AM -0700, Christoph Hellwig wrote:
> On Tue, Apr 11, 2023 at 04:31:59PM -0700, Darrick J. Wong wrote:
> > Unfortunately, the ondisk buffer is corrupt, but recovery just read the
> > buffer with no buffer ops specified:
> >
> > error = xfs_buf_read(mp->m_ddev_targp, buf_f->blf_blkno,
> > buf_f->blf_len, buf_flags, &bp, NULL);
>
> > +
> > + /*
> > + * We're skipping replay of this buffer log item due to the log
> > + * item LSN being behind the ondisk buffer. Verify the buffer
> > + * contents since we aren't going to run the write verifier.
> > + */
> > + if (bp->b_ops) {
> > + bp->b_ops->verify_read(bp);
> > + error = bp->b_error;
> > + }
>
> How do we end up with ops attached here if xfs_buf_read doesn't
> attach them? The buf type specific recover routines later attach
> ops, but this is called before we reach them.
The line of context above this hunk you cut out does exactly
that. i.e. this call:
xlog_recover_validate_buf_type(mp, bp, buf_f, NULLCOMMITLSN);
-Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
