I am running a customized Redhat 2.4.20-18.7 kernel (updated
3ware driver, i2c and lm_sensors) on a dual Athlon server
(Tyan S2460S, 2GB, dual MP2000) with two 3ware IDE raid adapter
(7810 with 40GB Raid1 and 200GB Raid1, 7850 with 840GB Raid5).
This setup seemed to run stable while using reiserfs on the
Raid5 (heavy load for about three weeks). Since I wasn't happy
with the performance of the Raid5 array, I switched over to xfs.
I took the 2.4.20-18.9 kernel src rpm from the XFS 1.2 Redhat 9
installer iso, extracted the xfs specific patches and applied
them to my customized 2.4.20-18.7 kernel source area.
The conversion to xfs went fine and performance was
much better then with reiserfs. After about a weeks
running under heavy load, the machine crashed with
the following error :
kernel BUG at rmap.c:398!
invalid operand: 0000
CPU: 1
EIP: 0010:[<c013e73b>] Tainted: PF
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010246
eax: 00000000 ebx: f2ef5163 ecx: fffe3000 edx: 000f2ef5
esi: 00350000 edi: c1000030 ebp: fffe3d40 esp: c44c5f48
ds: 0018 es: 0018 ss: 0018
Process kswapd (pid: 5, stackpage=c44c5000)
Stack: 014c5c00 00000000 c1030ce8 c1030ce8 c1030ce8 00000006 c03265c0 ffffffff
c013e86a 014c5c00 c1030ce8 c0139a00 c03b7c20 00000000 c03265c0 c1030ce8
c1030ce8 c03265c0 00000000 c0134c94 00000000 0000003f c03265c0 c0326f78
Call Trace: [<c013e86a>] (0xc44c5f68))
[<c0139a00>] (0xc44c5f74))
[<c0134c94>] (0xc44c5f94))
[<c01370c9>] (0xc44c5fac))
[<c0137d5b>] (0xc44c5fcc))
[<c0105000>] (0xc44c5fe8))
[<c0105706>] (0xc44c5ff0))
[<c0137900>] (0xc44c5ff8))
Running the output through ksymoops gives :
>>EIP; c013e73b <try_to_unmap_one+13b/200> <=====
>>ebx; f2ef5163 <_end+32af4047/38523ee4>
>>edi; c1000030 <_end+bfef14/38523ee4>
>>esp; c44c5f48 <_end+40c4e2c/38523ee4>
Trace; c013e86a <try_to_unmap+6a/190>
Trace; c0139a00 <add_to_swap+60/80>
Trace; c0134c94 <launder_page+5a4/9b0>
Trace; c01370c9 <rebalance_dirty_zone+99/100>
Trace; c0137d5b <kswapd+45b/4b0>
Trace; c0105000 <_stext+0/0>
Trace; c0105706 <arch_kernel_thread+26/30>
Trace; c0137900 <kswapd+0/4b0>
This doesn't look like xfs caused the problem, but before I post a bugreport
to bugzilla.redhat.com, I would like to know if the problem might be related
to the kernel core patches I applied to get xfs to work.
Dirk
|