On Tue, Jul 10, 2001 at 01:58:50PM +0200, Seth Mos wrote:
> At 21:28 10-7-2001 +1000, Daniel Stone wrote:
> >On Tue, Jul 10, 2001 at 04:17:17AM -0500, Steve Lord wrote:
> > > As for what is happening to your file. This is not corruption per se, the
> > > updated inode size made it out to disk, but the data did not. If you do
> > > an xfs_bmap you will probably find there are are no extents in the file.
> > > (xfs_bmap .emacs). It is also not in in kernel buffer corruption problem,
> > > it is because the buffers were not written to disk before the crash, but
> > > the inode size was written in a transaction.
> > >
> > > Please try a later kernel as I have not seen reports like this in a
> > > while.
> >Let me add my voice - I haven't posted this as yet, because I've been trying
> >to track it down as to where it started, but I'm getting null-byte
> >corruption through entire files (my entire /etc/mailcap got replaced with
> >null bytes, as did the postinst's for a few packages). I'm tracking CVS, and
> >it appears to only have happened with the -pre2 sync, everything previous
> >seemed to work OK (though admittedly I haven't tried -pre1, and that was
> >most likely the culprit, being a large -ac sync). The affected partitions
> >are on IDE drives, not a VIA chipset or anything.
> The source must be with you if you are running -pre kernels.
> This breakage can be expected, the TAKE message of the checkout
> specifically mentioned that it did compile but was not tested.
I'm aware of this. Not only am I running a pre kernel, from CVS, but it said
that it wasn't tested. Fear.
Daniel Stone <daniel@xxxxxxxxx>
<Nuke> "can NE1 help me aim nuclear weaponz????? /MSG ME!!"