Things then went fine until about 2.1.132, when I noticed that a simple
"make -j2" (building egcs or gimp for example) has a 30% chance of killing
the system.
The symptoms are:
1. the mousecursor (X) stops. The keyboard leds still work
the next thing I do is either Magic-SysRq-Sync+Unmount or Ctrl-AltBackspace,
to get out of X (which shows that something happens, i.e. the windows unmap).
2. the machine is then dead.
3. totally dead (no disk/led/screen activity)
needless to say, magic-sysrq sync&unmount didn't work, either
4. the syslog is empty
I tried this even without running X, almost same symptoms (well,
Sysrq+unmount displays:
Jan 28 00:24:44 cerebro kernel: SysRq: Emergency Remount R/O
Jan 28 00:24:44 cerebro kernel: SysRq: Emergency Sync
and, as you can see, THESE messages actually make it into the syslog file,
although the kernel doesn't seem to know anything about my filesystems
anymore!)
This behaviour was the same when I used the (non-hacked) original version of
apm.c, so this is not the source of problems, either.
A friend of mine has the same problem with two of his linux boxes. The same
hardware works fine with 2.1.130 (and, I think 131).
What makes me wonder is that this is exactly the same behaviour as with
my original bug-report. Is the mm-system in 2.1 soo flaky that linux just
halts in low-memory situations?? At the moment, 2.2.0 is totally unusable
for me (and my friend) (well, I still use it, but I avoid parallel makes
as hell)
vmstat 1, btw, shows that free goes down to ~2000, then buff to around ~2000,
and then cache to ~12000 when teh machien halts.
BTW, I can NOT reliably reproduce this, but it happens about once per 10 hours
(sometimes directly after rebooting, though)
My system;
MB ASUS P2L97 DS (SMP 2 CPU + adaptec AIC-7880)
192MB ram, no swap, no memory overcommitment
-----==- |
----==-- _ |
---==---(_)__ __ ____ __ Marc Lehmann +--
--==---/ / _ \/ // /\ \/ / pcg@goof.com |e|
-=====/_/_//_/\_,_/ /_/\_\ XX11-RIPE --+
The choice of a GNU generation |
|
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/