Linux 2.4.14 with XFS 1.0.2 patch (but without kdb patch) crashed on me
today. The on screen crash dump said something about kswapd.
I managed to salvage the following information from the screen before the
screen went black (courtesy of the screen blanker?). I didn't find any way
to get the display back so the rest of the information was lost:
---
Unable to handle kernel NULL pointer dereference at virtual address 00000018
printing eip:
c0137b20
*pde = 00000000
Oops: 0000
CPU: 1
EIP: 0010:[<c0137b20>]
[lost, something about kswapd in all of this]
---
The machine is a dual 600MHz PIII with 512 MB RAM and 1 GB of swap. I had
been running 2.4.14 for almost exactly two days when it crashed. There was
no significant load on the system at the time of the crash. Before this I
had been running 2.4.5 with no such problems (but other problems, such as
mysterious hangs, possibly related to the non-free OSS sound drivers or gcc
or XFS in order of suspicion).
I suppose this is not much use but perhaps it can serve as a data point. I'd
happily provide more information but with most of the crash dump lost I'm
not sure how. Please Cc: through email if you think there is some other info
that I could provide. Thanks.
28/11/01 04:36, Pascal Haakmat wrote:
> Linux 2.4.14 with XFS 1.0.2 patch (but without kdb patch) crashed on me
> today. The on screen crash dump said something about kswapd.
>
> I managed to salvage the following information from the screen before the
> screen went black (courtesy of the screen blanker?). I didn't find any way
> to get the display back so the rest of the information was lost:
[snip]
Some people suggested I should get the Oops information from
/var/log/messages. I should have mentioned that the Oops turns up nowhere in
the logfiles.
I did find a couple of (XFS-related?) Oopses that I hadn't noticed before
though. See the thread "XFS Oopses with 2.4.5 and 2.4.14?". I am starting
to wonder whether it is a good idea to run XFS. Thanks.