Oops with raid5 and 2.6.18.3

arobinson74 · 02-20-2007, 07:22 PM

Here is the full dmesg log of the crash:

Code:

iret exception: 0000 [#1]
SMP
Modules linked in: ppdev lp button ac battery ipv6 dm_snapshot
dm_mirror dm_mod loop tsdev rtc psmouse parport_pc parport floppy
serio_raw pcspkr i2c_nforce2 snd_intel8x0 snd_ac97_codec snd_
ac97_bus snd_pcm snd_timer snd soundcore snd_page_alloc shpchp
pci_hotplug i2c_core nvidia_agp agpgart evdev ext3 jbd mbcache raid456
md_mod xor ide_cd cdrom ide_disk sd_mod generic 8139too
amd74xx ide_core sata_sil 8139cp mii sata_nv libata scsi_mod forcedeth
ehci_hcd ohci_hcd usbcore thermal processor fan
CPU:    0
EIP:    0060:[<de95b0cb>]    Not tainted VLI
EFLAGS: 00000216   (2.6.18-3-686 #1)
EIP is at copy_data+0xff/0x14b [raid456]
eax: ddcce000   ebx: 00001000   ecx: 0000000f   edx: c1f71000
esi: ddccefc4   edi: c1f71fc4   ebp: 00000000   esp: dd261e4c
ds: 007b   es: 007b   ss: 0068
Process md0_raid5 (pid: 1115, ti=dd260000 task=dd0ed550 task.ti=dd260000)
Stack: c1f71000 ddb55460 c1e377a0 00000000 ddcce000 00001000 c1f71000 00000000
      00000000 00000000 00000000 dd20c388 c1e377a0 dd20c354 de95d96d 0c649510
      00000000 c0116d0a 06323c4f 00000000 dd20c384 00000000 c13c52e0 0000e000
Call Trace:
 [<de95d96d>] handle_stripe+0x10da/0x2075 [raid456]
 [<c0116d0a>] find_busiest_group+0x177/0x46a
 [<c011669e>] __wake_up+0x2a/0x3d
 [<de95a72c>] __release_stripe+0x10c/0x110 [raid456]
 [<de95a751>] release_stripe+0x21/0x2e [raid456]
 [<de95ea15>] raid5d+0x10d/0x132 [raid456]
 [<de92d769>] md_thread+0xd7/0xed [md_mod]
 [<c012d961>] autoremove_wake_function+0x0/0x2d
 [<de92d692>] md_thread+0x0/0xed [md_mod]
 [<c012d893>] kthread+0xc2/0xef
 [<c012d7d1>] kthread+0x0/0xef
 [<c0101005>] kernel_thread_helper+0x5/0xb
Code: 8d 04 2f 01 4c 24 18 83 7c 24 0c 00 8b 54 24 18 8d 34 32 89 34
24 74 09 89 d9 89 c7 c1 e9 02 eb 0a 8b 3c 24 89 d9 89 c6 c1 e9 02 <f3>
a5 89 d9 83 e1 03 74 02 f3 a4 8b 44 24 18 ba 03 00
 00 00 e8
EIP: [<de95b0cb>] copy_data+0xff/0x14b [raid456] SS:ESP 0068:dd261e4c
 <6>note: md0_raid5[1115] exited with preempt_count 1

I was having instability with this machine before (slackware 10.1 with
2.6.10 kernel) while compiling code (especially the kernel). I just
rebuilt is as a debian box. It never died in the raid array code
before though, just in gcc.

I have tested the machine's ram with memtest86 (3 passes) and will
more thoroughly check it tonight. Besides bad RAM, does anyone have
any other ideas on what may be causing the issue?

Quakeboy02 · 02-20-2007, 08:38 PM

Is this an IDE RAID? My wife's computer drove me nuts with "memory errors" in Windows until I changed the 80-pin ribbon cable to the drive.

arobinson74 · 02-20-2007, 10:27 PM

SATA Raid. I have 5 drives. One is the operating system on an IDE. The Raid is SATA on 2 controllers, the NVidia on motherboard (2 ports) and one SATA controller card (2 ports). The array is made of 4 drives: 3 active and 1 spare.

arobinson74 · 02-20-2007, 10:35 PM

Another important fact (maybe) is that the debian 2.6.18-3 does not have kernel pre-empting turned on:

CONFIG_PREEMPT_NONE=y
# CONFIG_PREEMPT_VOLUNTARY is not set
# CONFIG_PREEMPT is not set
# CONFIG_PREEMPT_BKL is not set

Quakeboy02 · 02-20-2007, 11:30 PM

I'm outta suggestions, other than to make sure that you have the latest mdadm and the driver for your card. Which card is it, by the way? You might also try kernel 2.6.19.3.

arobinson74 · 02-21-2007, 09:48 AM

Update: I think that you can ignore this error in terms of raid. I am getting segmentation faults when I attempt to rebuild the kernel.

This is exactly the same problem I had with slackware 10.1 with the 2.6.10 kernel (I just rebuilt the computer using Debian). So I think it is a hardware issue. Memtest86 didn't show any errors after 35 passes so if I trust that program it isn't the RAM. So I'll have to check the CPU and motherboard.

Quakeboy02 · 02-21-2007, 12:08 PM

Quote:

I have 5 drives. One is the operating system on an IDE.

There is a good probability that your problem is with the IDE. Change the cable first.