segmentation fault in TOP command

[Jose Barroca <jose.barroca@netcabo.pt>]

Dear all,

I'm trying to deal with a peculiar problem that came up the other day.
I've searched the net, posted in newbie groups, but to no avail. So,
perhaps you can lend a hand:

Using a 2.6.12.12 straight from kernel.org:
- I experience loss of responsiveness (mouse, keyboard, music) during
r/w intensive operations, such as lengthy computations in matlab
(exceeding my RAM), or a simple system update using Debian's dselect.
Mouse clicks and keypresses don't get lost, but xmms may skip the
tracks. And all this happens intermitently during the mentioned r-w op.
== This did not happen with previous kernels ==

- I had a case of FS corruption, which I could not trace. I use ext3,
only one partition for the complete debian system. I keep my data in
other partitions. Reason is a small disk in a laptop. This corruption
made itself visible after a reboot, when I called top to check why bash
was taking so long to complete a directory name (TAB press):

rdrs@abafado:~$ top
Segmentation fault

Other outputs included: "can't execute binary file", "attempt to access
beyond end of device"

I ran e2fsck -vc to get a read-only badblock scan, but the latter came
out clean. I had a lot of illegal inodes, though. This ext3 partition
was never accessed by other OSs.
== I use top on a dayly basis, so corruption happenned not long ago.
There were no power outages, but the previous kernel (2.6.11) had
NLS_DEFAULT=iso8859-1, whereas the new one has NLS_DEFAULT=utf8 ==

---
And, to sum up, I've been through a MEMTEST86, an E2FSCK, don't think
the machine was cracked (not even literally speaking), and ran SMARTCTL
-a /dev/hda.

This one had an interesting output: there was indication of an error
happening some 197 days ago. I could decipher the remaining info. Also,
the REALLOACTED_SECTOR_CT has a very high number, though it is labelled
"PRE_FAIL".


I'm quite at a loss on what to do now - where should I start looking?
And even if I simply replace "top", is that even possible, or advisable?

Eagerly waiting for your answers,

Jose






-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/