2.4.11 oops

2.4.11 oops

[Bob Matthews]

Linus,

I've received an oops while booting 2.4.11 on two different SMP
machines.  The kernel was SMP, HIGHMEM=64G with sym53c8xx, 3c59x,
eepro100, aic7xx and megaraid drivers statically linked.

Oops follows.  (Note: copied by hand, use with caution.)

ksymoops 2.4.0 on i686 2.4.2-2smp.  Options used
     -v /usr/src/linux/vmlinux (specified)
     -K (specified)
     -L (specified)
     -O (specified)
     -m /usr/src/linux/System.map (specified)

CPU:    1
EIP:    0010: [<c013d941>] Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010207
eax: 37e8ace4   ebx: 00000001     ecx: 00000001       edx: c02e1990
esi: f7dc6f44   edi: c02e1960     ebp: c2014000       esp: c2015c74
ds: 0018        es: 0018       ss: 0018
Process swapper (pid: 1, stackpage=c2015000)
Stack: 00000000 c02e1440 c02e1944 f7e7eaa4 c2014000 c013dab9 00000000
00000000
       f7e8ad0c e2015d5c c209ea0d 00000000 c014f3fc c2015e3c 00000124
00000202
       00000001 00000001 00000000 00000000 00000000 ffffffff 00000bb8
00000001
Call Trace: [<c013dab9>] [<c014f3fc>] [<c0111fe7>] [<c011965d>]
[<c014efd0>] [<c013e1ac>] [<c013e45d>] [<c013f3fe>] [<c0105ab0>]
[<c0106f6b>] [<c0110018>] [<c0105184>] [<c0105000>] [<c0105656>]
[<c0105070>]
Code: 0f 22 d8 eb 7d 8d 04 dd 00 00 00 00 ba 01 00 00 00 39 b8 40

>>EIP; c013d941 <exec_mmap+111/1f0>   <=====
Trace; c013dab9 <flush_old_exec+99/2f0>
Trace; c014f3fc <load_elf_binary+42c/a60>
Trace; c0111fe7 <schedule+407/630>
Trace; c011965d <__run_task_queue+5d/70>
Trace; c014efd0 <load_elf_binary+0/a60>
Trace; c013e1ac <search_binary_handler+8c/1c0>
Trace; c013e45d <do_execve+17d/1e0>
Trace; c013f3fe <getname+5e/a0>
Trace; c0105ab0 <sys_execve+30/60>
Trace; c0106f6b <system_call+33/38>
Trace; c0110018 <pin_2_irq+8/d0>
Trace; c0105184 <init+114/190>
Trace; c0105000 <_stext+0/0>
Trace; c0105656 <kernel_thread+26/30>
Trace; c0105070 <init+0/190>
Code;  c013d941 <exec_mmap+111/1f0>
00000000 <_EIP>:
Code;  c013d941 <exec_mmap+111/1f0>   <=====
   0:   0f 22 d8                  mov    %eax,%cr3   <=====
Code;  c013d944 <exec_mmap+114/1f0>
   3:   eb 7d                     jmp    82 <_EIP+0x82> c013d9c3
<exec_mmap+193/1f0>
Code;  c013d946 <exec_mmap+116/1f0>
   5:   8d 04 dd 00 00 00 00      lea    0x0(,%ebx,8),%eax
Code;  c013d94d <exec_mmap+11d/1f0>
   c:   ba 01 00 00 00            mov    $0x1,%edx
Code;  c013d952 <exec_mmap+122/1f0>
  11:   39 b8 40 00 00 00         cmp    %edi,0x40(%eax)

-- 
Bob Matthews
Red Hat, Inc.

Re: 2.4.11 oops

[Brian Gerst]

Bob Matthews wrote:
> 
> Linus,
> 
> I've received an oops while booting 2.4.11 on two different SMP
> machines.  The kernel was SMP, HIGHMEM=64G with sym53c8xx, 3c59x,
> eepro100, aic7xx and megaraid drivers statically linked.
>
> eax: 37e8ace4   ebx: 00000001     ecx: 00000001       edx: c02e1990
> Code;  c013d941 <exec_mmap+111/1f0>   <=====
>    0:   0f 22 d8                  mov    %eax,%cr3   <=====

What looks like happened here is that the pgd pointer isn't properly
aligned (it should be 32 byte aligned and 0x37e8ace4 is not).  Do you
have slab debugging turned on?  I think this has been fixed already in
the AC kernels.

--

				Brian Gerst

Re: 2.4.11 oops

[Linus Torvalds]

On Wed, 10 Oct 2001, Bob Matthews wrote:
>
> I've received an oops while booting 2.4.11 on two different SMP
> machines.  The kernel was SMP, HIGHMEM=64G with sym53c8xx, 3c59x,
> eepro100, aic7xx and megaraid drivers statically linked.

With CONFIG_SLAB_DEBUG on?

Slab debugging is not compatible with HIGHMEM=64G as-is, I think (and we
didn't use to expose that CONFIG option to normal users until recently).
Th ebug is in get_pgd_slow() in include/asm-i386/pgalloc.h, where it just
does a "kmalloc()" and expects the thing to be magically aligned.

Does anybody have (tested) patches for this?

		Linus

Re: 2.4.11 oops

[Alan Cox]

> Th ebug is in get_pgd_slow() in include/asm-i386/pgalloc.h, where it just
> does a "kmalloc()" and expects the thing to be magically aligned.
> 
> Does anybody have (tested) patches for this?

Ingo did patches for -ac a long time back. I've not submitted them since
it simply didnt seem an important matter when prioritising patches,

If you want them I can isolate them tomorrow

Re: 2.4.11 oops

[Bob Matthews]

Linus Torvalds wrote:
> 
> On Wed, 10 Oct 2001, Bob Matthews wrote:
> >
> > I've received an oops while booting 2.4.11 on two different SMP
> > machines.  The kernel was SMP, HIGHMEM=64G with sym53c8xx, 3c59x,
> > eepro100, aic7xx and megaraid drivers statically linked.
> 
> With CONFIG_SLAB_DEBUG on?
> 
> Slab debugging is not compatible with HIGHMEM=64G as-is

Ah yes, that was it.  I try again with mem alloc debugging turned off.

-- 
Bob Matthews
Red Hat, Inc.