> Daniel Moore wrote:
> >
> > (still working on the stress tests - they seem to be doing the trick)
> >
> > I'm seeing some funny messages from xfs_zero_last_block
> >
> > => xfs_zero_last_block: unwritten?
> > => xfs_zero_last_block: We want DELWRI? not waiting?
> >
> > just before the page_buf panics
> >
> > => delalloc page 0xc106951c with no extent kernel BUG at page_buf.c:5275!
>
>
> This looks to be a problem in handling UNWRITTEN extents,
> and this one seems to be an UNWRITTEN at that contains EOF (hmm).
>
> Don't know the right fix to this one, but you can try avoiding
> this bug by creating the FS without support for unwritten:
> I believe the option is "-d unwritten=0" to mkfs_xfs.
>
>
> --------------------------------------------------------------------------
> Rajagopal Ananthanarayanan ("ananth")
> Member Technical Staff, SGI.
> --------------------------------------------------------------------------
We should really have unwritten extents turned off - probably should change
mkfs to default to off as well. We do not have code to handle them properly
yet.
Steve
|