Skip to contentSkip to conversation

Kernel Oops

DISCUSSION Joël Cuissinat 2 participants 2 messages Oct 27, 2006

lore
Share & tools
lore
#0 lore raw

Hello,

I'm working for the french national education.
We work on a server for students since 3 years.
The solution is based on a Mandrake 9.1 but with a large number of updates.
Our kernel is : 2.4.32 and the xfstools are updated :
- xfsdump-2.2.42
- xfsprogs-2.8.11
- libxfs1-2.8.11
We made a tool to create students and about 1 time over 10 we have a kernel Oops.
With the same data we can reproduce the crash and it always append at the same place (the same student).
It was the same 2 years ago with older kernel and xfs releases but we hadn't enough student databases to test. Now, the project is growing and the number of crashes in schools too !

The user creation process is composed with :
mkdir, setfacl, chown, setquota for a pam-ldap authentification.

The Oops usually happends on the chown instruction but if we remove it, it happends on the setquota instruction which is one line behind.

Cheers,

Joël Cuissinat

Signature · 4 lines
Joël Cuissinat
Equipe EOLE
CETIAD Dijon, France
http://eole.orion.education.fr --
Oct 25 18:17:52 jscribe kernel: Unable to handle kernel NULL pointer dereference at virtual address 00000006                                               Oct 25 18:17:52 jscribe kernel:  printing eip:
Oct 25 18:17:52 jscribe kernel: c021de83                                                                                                                   Oct 25 18:17:52 jscribe kernel: *pgd = 0000000000000000
Oct 25 18:17:52 jscribe kernel: CPU:    0                                                                                                                  Oct 25 18:17:52 jscribe kernel: EIP:    0010:[xfs_trans_brelse+51/240]    Not tainted
Oct 25 18:17:52 jscribe kernel: EIP:    0010:[<c021de83>]    Not tainted                                                                                   Oct 25 18:17:52 jscribe kernel: EFLAGS: 00010246
Using defaults from ksymoops -t elf32-i386 -a i386
Oct 25 18:17:52 jscribe kernel: eax: 00000000   ebx: c59ed790   ecx: c59ed790   edx: 00000000                                                              Oct 25 18:17:52 jscribe kernel: esi: c5352e98   edi: ce520d80   ebp: ce520d80   esp: c6b97d2c
Warning (Oops_set_regs): garbage 'Oct 25 18:17:52 jscribe kernel: esi: c5352e98   edi: ce520d80   ebp: ce520d80   esp: c6b97d2c' at end of register line ignored
Oct 25 18:17:52 jscribe kernel: ds: 0018   es: 0018   ss: 0018                                                                                             Oct 25 18:17:52 jscribe kernel: Process chown (pid: 15389, stackpage=c6b97000)
Oct 25 18:17:52 jscribe kernel: Stack: 00000000 00000000 c5352e98 c0235f2c c5352e98 ce520d80 00000000 c5d7f198                                             Oct 25 18:17:52 jscribe kernel:        ce520d80 00000000 c5352e98 cff310a0 00000004 c0235fbf c5352e98 0000302d
Oct 25 18:17:52 jscribe kernel:        cff310a0 00000402 cf8b0980 cf8b0984 0000302d cf8b0c00 c023621a cf8b0c00                                             Oct 25 18:17:52 jscribe kernel: Call Trace:    [xfs_qm_dqread+188/208] [xfs_qm_idtodq+127/224] [xfs_qm_dqget+218/848] [xfs_qm_vop_dqalloc+245/624] [xfs_set
Oct 25 18:17:52 jscribe kernel:   [<c01d6d28>] [<c022d7f7>] [<c017f678>] [<c0165a11>] [<c0165a9e>] [<c012bf13>]                                            Oct 25 18:17:52 jscribe kernel:
Oct 25 18:17:52 jscribe kernel: Code: 8a 42 06 83 e0 01 84 c0 75 f2 f6 43 30 04 75 ec 52 56 e8 a6
                                                                                                                                                           >>EIP; c021de83 <xfs_trans_brelse+33/f0>   <=====
Quoted 1 line
Quoted 1 line

ebx; c59ed790 <_end+556a7ec/103890bc> >>ecx; c59ed790 <_end+556a7ec/103890bc>

                                                                                                                                                           Code;  c021de83 <xfs_trans_brelse+33/f0>
00000000 <_EIP>:                                                                                                                                           Code;  c021de83 <xfs_trans_brelse+33/f0>   <=====
   0:   8a 42 06                  mov    0x6(%edx),%al   <=====                                                                                            Code;  c021de86 <xfs_trans_brelse+36/f0>
   3:   83 e0 01                  and    $0x1,%eax                                                                                                         Code;  c021de89 <xfs_trans_brelse+39/f0>

6: 84 c0 test %al,%al
Code; c021de8b <xfs_trans_brelse+3b/f0>
8: 75 f2 jne fffffffc <_EIP+0xfffffffc>
Code; c021de8d <xfs_trans_brelse+3d/f0>
a: f6 43 30 04 testb $0x4,0x30(%ebx)
Code; c021de91 <xfs_trans_brelse+41/f0>
e: 75 ec jne fffffffc <_EIP+0xfffffffc>
Code; c021de93 <xfs_trans_brelse+43/f0>
10: 52 push %edx
Code; c021de94 <xfs_trans_brelse+44/f0>
11: 56 push %esi
Code; c021de95 <xfs_trans_brelse+45/f0>
12: e8 a6 00 00 00 call bd <_EIP+0xbd>

2 warnings and 1 error issued. Results may not be reliable.

#1 lore raw

Do you have a test case that can reproduce this?

If so please file a bug
http://oss.sgi.com/bugzilla
and attach the test program.

Include any other relevant info.

Quoted 65 lines · Joël Cuissinat

Hello,

I'm working for the french national education.
We work on a server for students since 3 years.
The solution is based on a Mandrake 9.1 but with a large number of updates. Our kernel is : 2.4.32 and the xfstools are updated :
- xfsdump-2.2.42
- xfsprogs-2.8.11
- libxfs1-2.8.11
We made a tool to create students and about 1 time over 10 we have a kernel Oops.
With the same data we can reproduce the crash and it always append at the same place (the same student).
It was the same 2 years ago with older kernel and xfs releases but we hadn't enough student databases to test. Now, the project is growing and the number of crashes in schools too !

The user creation process is composed with :
mkdir, setfacl, chown, setquota for a pam-ldap authentification.

The Oops usually happends on the chown instruction but if we remove it, it happends on the setquota instruction which is one line behind.

Cheers,

Joël Cuissinat

plain text document attachment (ksymoops.txt)

Oct 25 18:17:52 jscribe kernel: Unable to handle kernel NULL pointer dereference at virtual address 00000006                                               Oct 25 18:17:52 jscribe kernel:  printing eip:
Oct 25 18:17:52 jscribe kernel: c021de83                                                                                                                   Oct 25 18:17:52 jscribe kernel: *pgd = 0000000000000000
Oct 25 18:17:52 jscribe kernel: CPU:    0                                                                                                                  Oct 25 18:17:52 jscribe kernel: EIP:    0010:[xfs_trans_brelse+51/240]    Not tainted
Oct 25 18:17:52 jscribe kernel: EIP:    0010:[<c021de83>]    Not tainted                                                                                   Oct 25 18:17:52 jscribe kernel: EFLAGS: 00010246
Using defaults from ksymoops -t elf32-i386 -a i386
Oct 25 18:17:52 jscribe kernel: eax: 00000000   ebx: c59ed790   ecx: c59ed790   edx: 00000000                                                              Oct 25 18:17:52 jscribe kernel: esi: c5352e98   edi: ce520d80   ebp: ce520d80   esp: c6b97d2c
Warning (Oops_set_regs): garbage 'Oct 25 18:17:52 jscribe kernel: esi: c5352e98   edi: ce520d80   ebp: ce520d80   esp: c6b97d2c' at end of register line ignored
Oct 25 18:17:52 jscribe kernel: ds: 0018   es: 0018   ss: 0018                                                                                             Oct 25 18:17:52 jscribe kernel: Process chown (pid: 15389, stackpage=c6b97000)
Oct 25 18:17:52 jscribe kernel: Stack: 00000000 00000000 c5352e98 c0235f2c c5352e98 ce520d80 00000000 c5d7f198                                             Oct 25 18:17:52 jscribe kernel:        ce520d80 00000000 c5352e98 cff310a0 00000004 c0235fbf c5352e98 0000302d
Oct 25 18:17:52 jscribe kernel:        cff310a0 00000402 cf8b0980 cf8b0984 0000302d cf8b0c00 c023621a cf8b0c00                                             Oct 25 18:17:52 jscribe kernel: Call Trace:    [xfs_qm_dqread+188/208] [xfs_qm_idtodq+127/224] [xfs_qm_dqget+218/848] [xfs_qm_vop_dqalloc+245/624] [xfs_set
Oct 25 18:17:52 jscribe kernel:   [<c01d6d28>] [<c022d7f7>] [<c017f678>] [<c0165a11>] [<c0165a9e>] [<c012bf13>]                                            Oct 25 18:17:52 jscribe kernel:
Oct 25 18:17:52 jscribe kernel: Code: 8a 42 06 83 e0 01 84 c0 75 f2 f6 43 30 04 75 ec 52 56 e8 a6
                                                                                                                                                           >>EIP; c021de83 <xfs_trans_brelse+33/f0>   <=====
Quoted 1 line
Quoted 1 line

ebx; c59ed790 <_end+556a7ec/103890bc> >>ecx; c59ed790 <_end+556a7ec/103890bc>

                                                                                                                                                           Code;  c021de83 <xfs_trans_brelse+33/f0>
00000000 <_EIP>:                                                                                                                                           Code;  c021de83 <xfs_trans_brelse+33/f0>   <=====
   0:   8a 42 06                  mov    0x6(%edx),%al   <=====                                                                                            Code;  c021de86 <xfs_trans_brelse+36/f0>
   3:   83 e0 01                  and    $0x1,%eax                                                                                                         Code;  c021de89 <xfs_trans_brelse+39/f0>

6: 84 c0 test %al,%al
Code; c021de8b <xfs_trans_brelse+3b/f0>
8: 75 f2 jne fffffffc <_EIP+0xfffffffc>
Code; c021de8d <xfs_trans_brelse+3d/f0>
a: f6 43 30 04 testb $0x4,0x30(%ebx)
Code; c021de91 <xfs_trans_brelse+41/f0>
e: 75 ec jne fffffffc <_EIP+0xfffffffc>
Code; c021de93 <xfs_trans_brelse+43/f0>
10: 52 push %edx
Code; c021de94 <xfs_trans_brelse+44/f0>
11: 56 push %esi
Code; c021de95 <xfs_trans_brelse+45/f0>
12: e8 a6 00 00 00 call bd <_EIP+0xbd>

2 warnings and 1 error issued. Results may not be reliable.

Signature · 1 line
Russell Cattelan <[email protected]>