> Date: Mon, 21 Jan 2002 12:37:05 -0600 (CST)
> From: Eric Sandeen <sandeen@xxxxxxx>
> To: Luis Montes <luis.montes@xxxxxxxxx>
> Cc: <linux-xfs@xxxxxxxxxxx>
> Subject: Re: XFS kernel crash and corrupted FS
> On Mon, 21 Jan 2002, Luis Montes wrote:
>
> > The problem:
> > I was just finished installing DRI, and was running glxgears to test
> > it. Then when I tried to list a directory in one of the xfs
> > partitions it reported input/output error. I rebooted and the root
> > filesystem was corrupted beyond repair with the wnsuing kernel panic.
>
> How was this determined? Did you try xfs_repair on the filesystem? What
> was the output? Can you send the output from xfs_repair, if it's
> failing?
Yes I did try to repair it with xfs_repair. Checking some filesystems
it told me that there was valuable metadata and that I had to replay
it by mounting it, but then I couldn't and running xfs_repair -L in
some of them produced a HUGE amount of messages, that translated into
most of my files being scrambled into the repaired lost+found
directory. In one instance xfs just aborted with an assertion (I had
it saved, but then I also lost the second HD. As a result I lost all
the logs from xfs_repair, arrgh....) I was able to recover one
filesystem and dump a tar.gz to a zipdrive, all the rest were lost
:-(
>
> Also, are there any messages in your logs concerning a forced shutdown of
> the filesystem?
>
No, there are no messages in the log file ( and there is no log file
now anyway ;-)
> > Three are
> > damaged beyond repair (one actually seems to appear as an ext3!
> > filesystem)
>
> This probably means that the xfs superblock is damaged, and mount is
> finding an old ext3 signature from a previous mkfs.
>
> > But trying to mount one of the damaged fs's actually produces a
> > couple of "unable to handle kernel paging request" errors and the
> > system just freezes. It seems repeatable, three times I tried
> > mounting that filesystem and thre three times I got thar fault. Is
> > this something you guys might want to look more closely? It's been a
> > long long while since I had any kernel problems, and I just read in
> > the FAQ that I might first run ksymoops.
>
> Yes, the ksymoops output might shed more light on this.
>
> -Eric
>
>
Sorry, it just seems that the whole system was getting worst and
worst everytime I rebooted. Nothing ever appeard on syslog, and
mounting the filesystem that caused a kernel panic the day before was
now causing the machine to just reboot . One thing makes me feel
better, and it is that I had done a good backup to CD just the day
before. I'll just have to fix the hardware now, running a bios
update, one of those HD repair utilities from WD, check and double
check the DMA settings. Caviar drives have a rather spotty history
and maybe the bios isn't setting conservative enough timings? What I
don't understand is why I was able to get the system up an running
for some days, compile XFree86, mozilla, galeon, browse the net,
compile drivers... and then all of a sudden all hell broke loose,
and no amount of reformatting/reinstalling has been able to produce a
stable system ... I reinstalled from scratch repartitioning and
mkfs.xfs -ing all the partitions fresh, and even though the
installation itself went smooth, once I booted up the system went
immediately belly up and now (today in the early ours ...) the
machine wouldn't boot any partition on any disk! Lilo starts and
loads the kernel, but then no root filesystem is usable ... Oh
well, sorry for the ranting. Back to square one.
Thanks
Luis A. Montes
Physicist/Programmer
Quantum Design
e-mail: luis.montes@xxxxxxxxx
|