Re: 2.6.12-rc1-mm3

• Previous message: Christoph Hellwig: "Re: megaraid driver (proposed patch)"
• In reply to: Borislav Petkov: "Re: 2.6.12-rc1-mm3"
• Next in thread: Andrew Morton: "Re: 2.6.12-rc1-mm3"
• Reply: Andrew Morton: "Re: 2.6.12-rc1-mm3"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

To: Andrew Morton <akpm@osdl.org>
Date:	Thu, 31 Mar 2005 15:05:48 +0200

On Friday 25 March 2005 17:46, Borislav Petkov wrote:
> Hi Andrew,
>
> mm3 still not booting on my machine. Boot option 'nmi_watchdog=2' (my cpu
> is a dual core pentium 4 HT, 2.60 GHz) gets me a bit further in the boot
> process but it blocks there too.
>
> [output retyped from screen]:
> kernel: [ 4.109241] PM: Checking swsusp image.
> kernel: [ 4.109244] PM: Resume from disk failed.
> kernel: [ 4.112220] VFS: Mounted root (ext2 filesystem) readonly.
> kernel: [ 4.112465] Freeing unused kernel memory: 188k freed
> kernel: [ 4.142002] logips2pp: Detected unknown logitech mouse model 1
> kernel: [ 4.274620] input: PS/2 Logitech Mouse on isa0060/serio1
> <--- [point of previous blocks without boot option 'nmi_watchdog=2']--->
> INIT: version 2.86 booting
> Mounting a tmpfs over /dev... done.
> Creating initial device nodes... done.
> Setting parameters of disc: (none).
> Activating swap.
> kernel: [ 10.712648] Adding 976744k swap on /dev/hda2. Priority:-1
> extents:1 Checking root file system...
> fsck 1.36 (05-Feb-2005)
> /: clean, 127290/1831424 files, 898566/3662056 blocks
> [EOF]

Hi Andrew,

i finally got to run kdb within mm3 and I got a bit further but am not sure
whether I'm debugging in the right direction:

After booting with "kdb=early" I found out that the kernel blocks with the
partial message:

kmem_cache_create: Early error in slab task_struct
kernel BUG at mm/slab.c:1215
invalid operand: 0000 [#1]
PREEMPT SMP

and here all dies. After singlestepping through the code, I found out that
start_kernel calls at offset 0x14d fork_init, which calls at offset 0x39
kmem_cache_create. kmem_cache_create performs some initial checks:

   1205 /*
   1206 * Sanity checks... these are all serious usage bugs.
   1207 */
   1208 if ((!name) ||
   1209 in_interrupt() ||
   1210 (size < BYTES_PER_WORD) ||
   1211 (size > (1<<MAX_OBJ_ORDER)*PAGE_SIZE) ||
   1212 (dtor && !ctor)) {
   1213 printk(KERN_ERR "%s: Early error in slab %s\n",
   1214 __FUNCTION__, name);
   1215 BUG();
   1216 }

And after singlestepping a little bit more, I found out that the
in_interrupt() check returns true and printk is executed. Here's the
disassembled code:

00000b10 <kmem_cache_create>:
kmem_cache_create():
mm/slab.c:1201
     b10: 55 push %ebp
     b11: 89 e5 mov %esp,%ebp
     b13: 57 push %edi
     b14: 56 push %esi
     b15: 53 push %ebx
     b16: 83 ec 3c sub $0x3c,%esp
mm/slab.c:1208
     b19: 8b 55 08 mov 0x8(%ebp),%edx
     b1c: 85 d2 test %edx,%edx
     b1e: 0f 84 01 08 00 00 je 1325 <kmem_cache_create+0x815>
include/asm/thread_info.h:91
     b24: b8 00 f0 ff ff mov $0xfffff000,%eax
     b29: 21 e0 and %esp,%eax
include/asm/thread_info.h:89
     b2b: f7 40 14 00 ff ff 0f testl $0xfffff00,0x14(%eax)
     b32: 0f 85 ed 07 00 00 jne 1325 <kmem_cache_create+0x815> <---

this jump here executes and line 0x1325 is:

mm/slab.c:1213
    1325: c7 04 24 40 02 00 00 movl $0x240,(%esp)
    132c: 8b 45 08 mov 0x8(%ebp),%eax
    132f: 89 44 24 08 mov %eax,0x8(%esp)
    1333: b8 0d 00 00 00 mov $0xd,%eax
    1338: 89 44 24 04 mov %eax,0x4(%esp)
    133c: e8 fc ff ff ff call 133d <kmem_cache_create+0x82d>
mm/slab.c:1215
    1341: 0f 0b ud2a
    1343: bf 04 00 00 00 mov $0x4,%edi
    1348: 00 e9 add %ch,%cl
    134a: 13 f8 adc %eax,%edi
    134c: ff (bad)
    134d: ff c7 inc %edi

and BUG() is called.

Any suggestions or corrections will be greatly appreciated.

Regards,
Boris.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

• Previous message: Christoph Hellwig: "Re: megaraid driver (proposed patch)"
• In reply to: Borislav Petkov: "Re: 2.6.12-rc1-mm3"
• Next in thread: Andrew Morton: "Re: 2.6.12-rc1-mm3"
• Reply: Andrew Morton: "Re: 2.6.12-rc1-mm3"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]