Kernel oops

Linux home 2.4.20 #18 SMP Fri Sep 12 22:49:21 PDT 2003 i686 GNU/Linux

home:~# lvcreate --version
  LVM version:     1.95.15 (2003-01-10)
  Library version: 0.96.07-ioctl-cvs (2002-11-21)
  Driver version:  1.0.8

I created a script to take daily snapshots of my four logical volumes for backup purposes. Approximately every other time I run it the whole lvm system seems to freeze up. The first time it did this it crashed the whole computer by not allowing any writes to disk. The second two times the system didn't crash but I now have hung lvremove and lvcreate processes which appear to require a reboot to clear. strace on the processes shows nothing happening and kill -9 has no effect. In the dmesg I have lots of "invalidate: busy buffer" and then a kernel oops:

Unable to handle kernel NULL pointer dereference at virtual address

00000001
printing eip:
ce6a16d0
*pde = 00000000

Oops: 0002
CPU:    1
EIP:    0010:[<ce6a16d0>]    Tainted: P
EFLAGS: 00010202
eax: ce6a16d0   ebx: ce6a16a0   ecx: 00000010   edx: ce6a16d0
esi: 00000001   edi: 00000000   ebp: c023c230   esp: df6c3f3c
ds: 0018   es: 0018   ss: 0018
Process kcopyd (pid: 69, stackpage=df6c3000)
Stack: c023b83a ce6a16d0 00000001 00000001 00000b76 00000000 0000001b
00000000
       c7cc3c10 cda29004 00000010 c023a52f d41f3e8c c7c7b218 c023a4e0
c7c7b218
       c023c891 00000000 c7c7b218 cda29004 c031f858 c023c240 cda29004
c023c394
Call Trace:    [<c023b83a>] [<c023a52f>] [<c023a4e0>] [<c023c891>]

[<c023c240>]
[<c023c394>] [<c023c424>] [<c023c230>] [<c023c535>] [<c01075ee>]
[<c023c480>]

Code: d0 16 6a ce d0 16 6a ce 00 00 00 00 69 dd 63 00 8c ff 7c f9

This seems to be a reproduceable problem. I was under the impression that LVM was fairly stable by this point. The first crash happened in the middle of a backup that had been running for 48 hours which I will now have to restart. Not too happy. :( Uh oh...now the system can't write to disk again and I will have to reboot.

Tracy Reed
http://copilotconsulting.com

For various reasons (when I fix it, I'll Tell the Tale) I installed LVM2 today. It didn't solve my immediate problems, so I tried to go back to LVM1. However, the dev/vg_name/lv_name node had disappeared and even though both vgchange -ay, vgscan, lvscan and pvscan all reported seeing the lv fine, mount could not find it.

A reboot did not help. Panic lurked.

However, vgchange -an followed by vgchange -ay helped.

/ Rickard Olsson,IT-Konsult/
/ Telefon: +46 70 635 01 42/
/ http://www.webhackande.se/

If device nodes are missing, vgmknodes should help in LVM1.

--
AJ Lewis

Sistina Software Inc.             http://www.sistina.com
720 Washington Ave, Suite 200     Voice:  612-638-0500
Minneapolis, MN 55414             E-Mail: lewis_NOSPAM_(AT)sistina(DOT)com

Current GPG fingerprint = D638 9C01 D08F 29A4 5D50 D064 F6A6 0ED6 E386 E6EB Grab the key at: http://people.sistina.com/~lewis/gpg.html or one of the many keyservers out there...

Behind every good computer -- is a jumble of wires 'n stuff.

I didn't know that, but I had the opportunity to test it, as the /dev/f2/f2 LV went missing again after a reboot. vgmknodes reported success, but mount still did not find the node until after a vgchange -an / vgchange -ay combo. Weird.

As I mentioned earlier, I did install and run LVM2 on this machine before going back to LVM1 (even going back and forth a few times) and I suppose it's possible that this changed something in the metadata that LVM1 didn't like.

/ Rickard Olsson,IT-Konsult/
/ Telefon: +46 70 635 01 42/
/ http://www.webhackande.se/

On Mon, Sep 15, 2003 at 03:01:39AM -0700, Tracy R Reed spake thusly:

I was pretty tired and annoyed last night when I sent the above email and totally forgot to run the oops through ksymoops before sending. Can anyone tell me if the above LVM version should be relatively stable or is it an experimental/beta version? I need to get a stable working LVM on this box. Here is the oops after running through ksymoops:

Unable to handle kernel NULL pointer dereference at virtual address
00000001
ce6a16d0
*pde = 00000000
Oops: 0002
CPU:    1
EIP:    0010:[<ce6a16d0>]    Tainted: P

Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010202
eax: ce6a16d0 ebx: ce6a16a0 ecx: 00000010 edx: ce6a16d0
esi: 00000001 edi: 00000000 ebp: c023c230 esp: df6c3f3c
ds: 0018 es: 0018 ss: 0018
Process kcopyd (pid: 69, stackpage=df6c3000)

Stack: c023b83a ce6a16d0 00000001 00000001 00000b76 00000000 0000001b 00000000
       c7cc3c10 cda29004 00000010 c023a52f d41f3e8c c7c7b218 c023a4e0 c7c7b218
       c023c891 00000000 c7c7b218 cda29004 c031f858 c023c240 cda29004 c023c394
Call Trace:    [<c023b83a>] [<c023a52f>] [<c023a4e0>] [<c023c891>] [<c023c240>]
  [<c023c394>] [<c023c424>] [<c023c230>] [<c023c535>] [<c01075ee>] [<c023c480>]
Code: d0 16 6a ce d0 16 6a ce 00 00 00 00 69 dd 63 00 8c ff 7c f9

Trace; c023b83a <persistent_commit+fa/110>
Trace; c023a52f <copy_callback+3f/50>
Trace; c023a4e0 <commit_callback+0/10>
Trace; c023c891 <copy_complete+61/70>
Trace; c023c240 <run_complete_job+10/20>
Trace; c023c394 <process_jobs+64/d0>
Trace; c023c424 <do_work+24/80>
Trace; c023c230 <run_complete_job+0/20>
Trace; c023c535 <kcopyd+b5/f0>
Trace; c01075ee <arch_kernel_thread+2e/40>
Trace; c023c480 <kcopyd+0/f0>

Code; ce6a16d0 <_end+e2e5354/2050ace4>
00000000 <_EIP>:
Code; ce6a16d0 <_end+e2e5354/2050ace4> <=====
0: d0 16 rclb (%esi) <=====
Code; ce6a16d2 <_end+e2e5356/2050ace4>
2: 6a ce push $0xffffffce
Code; ce6a16d4 <_end+e2e5358/2050ace4>
4: d0 16 rclb (%esi)
Code; ce6a16d6 <_end+e2e535a/2050ace4>
6: 6a ce push $0xffffffce
Code; ce6a16d8 <_end+e2e535c/2050ace4>
8: 00 00 add %al,(%eax)
Code; ce6a16da <_end+e2e535e/2050ace4>
a: 00 00 add %al,(%eax)
Code; ce6a16dc <_end+e2e5360/2050ace4>
c: 69 dd 63 00 8c ff imul $0xff8c0063,%ebp,%ebx
Code; ce6a16e2 <_end+e2e5366/2050ace4>
12: 7c f9 jl d <_EIP+0xd>

Tracy Reed
http://copilotconsulting.com

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com