On Tue, 18 Apr 2000, Cyrille Chepelov (home) wrote:
> Got something *very* similar recently. I posted it to the list, nobody
> cared to reply ; not precise enough, probably.
>
> Can you characterise when does your kswapd bang (how much memory your hog
> ate). Can you also run your hog while in a text mode console, to see if
> kswapd "oopses" anything there before it dies. Last but not least, are the
> symptoms the same with -pre5 (I found kswapd still crashed, but behaved
> much better wrt CPU usage).
Doh! Of course you (and mikeg, thanks as well) are right. Silly me missed
an oops in the logs. Here it is (argh, klogd messed it up, disassembly
added by hand):
[Oops from 2.3.99-pre3 - kswapd died after some repeated heavy swapping,
running a mem hog.]
Unable to handle kernel NULL pointer dereference at virtual address 000000a4
printing eip:
c015d6cc
*pde = 00000000
Oops: 0000
CPU: 0
EIP: 0010:[shm_swap_core+76/224]
EFLAGS: 00010282
eax: 00000000 ebx: c020a940 ecx: 00000000 edx: 006c1c05
esi: c460e030 edi: c020a940 ebp: ceeeb860 esp: cffd3f5c
ds: 0018 es: 0018 ss: 0018
Process kswapd (pid: 2, stackpage=cffd3000)
Stack: ceeeb860 00000000 c020a940 c020a940 00000000 c020a940 00000000 c015d8d8
ceeeb860 00000000 00757300 c020a940 cffd3fa0 cffd3fa4 0000000a 00000006
00000004 00000001 00000006 00757300 c012e57b 00000006 00000004 c020a940
Call Trace: [shm_swap+264/336] [do_try_to_free_pages+123/176] [kswapd+131/304] [empty_bad_page+0/4096] [kernel_thread+35/48]
Code: 8b b8 a4 00 00 00 89 7c 24 14 3b bb a4 00 00 00 75 30 29 f8
0: 8b b8 a4 00 00 00 movl 0xa4(%eax),%edi <===
6: 89 7c 24 14 movl %edi,0x14(%esp,1)
a: 3b bb a4 00 00 00 cmpl 0xa4(%ebx),%edi
10: 75 30 jne 00000042
12: 29 f8 subl %edi,%eax
So we died in ipc/shm.c. Umm..., if I traced the disassembly right, it
looks like page_map->zone was NULL. I saw a few shm fixes going into
-pre4, but at a quick glance I can't tell if they're related. Christoph,
does this one look familiar to you?
Regards,
Daniel.
-- GNU/Linux Audio Mechanics - http://www.glame.de GPG Key ID 89BF7E2B - http://www.keyserver.net- 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/
This archive was generated by hypermail 2b29 : Sun Apr 23 2000 - 21:00:13 EST