Unable to handle kernel NULL pointer kernel 2.4.21-pre3-ac4

Unable to handle kernel NULL pointer kernel 2.4.21-pre3-ac4

[Tupshin Harper]

Interesting bits:
KT400
LVM2(hence the ac tree)
reiserfs
happened during a lengthy mysql operation.

Jan 14 19:39:17 fussbudget kernel: Unable to handle kernel NULL pointer 
dereference at virtual address 00000004
Jan 14 19:39:17 fussbudget kernel: c0133c58
Jan 14 19:39:17 fussbudget kernel: *pde = 00000000
Jan 14 19:39:17 fussbudget kernel: Oops: 0002
Jan 14 19:39:17 fussbudget kernel: CPU:    0
Jan 14 19:39:17 fussbudget kernel: EIP: 
0010:[__free_pages_ok+600/640]    Tainted: PF
Jan 14 19:39:17 fussbudget kernel: EFLAGS: 00210246
Jan 14 19:39:17 fussbudget kernel: eax: 00000000   ebx: c10cfc00   ecx: 
c5662000   edx: c566205c
Jan 14 19:39:17 fussbudget kernel: esi: 00000000   edi: 00003b37   ebp: 
c0378350   esp: c5663e58
Jan 14 19:39:17 fussbudget kernel: ds: 0018   es: 0018   ss: 0018
Jan 14 19:39:17 fussbudget kernel: Process mysqld (pid: 2415, 
stackpage=c5663000)
Jan 14 19:39:17 fussbudget kernel: Stack: 00000001 00200282 caa14d40 
caa14d40 caa14d40 c10cfc00 c013f135 caa14d40
Jan 14 19:39:17 fussbudget kernel:        dd53a428 c10cfc00 00003b37 
c0378350 c0132e1f c10cfc00 000001d2 c5662000
Jan 14 19:39:17 fussbudget kernel:        00000200 000001d2 00000020 
00000020 000001d2 00000020 00000006 c0133061
Jan 14 19:39:17 fussbudget kernel: Call Trace: 
[try_to_free_buffers+133/240] [shrink_cache+527/768] 
[shrink_caches+97/176] [try_to_free_pages_zone+54/96] 
[balance_classzone+85/480]
Jan 14 19:39:17 fussbudget kernel: Code: 89 58 04 89 03 89 53 04 89 59 
5c 89 73 0c ff 41 68 eb c1 0f
Using defaults from ksymoops -t elf32-i386 -a i386


 >>ebx; c10cfc00 <_end+c944bc/205ca93c>
 >>ecx; c5662000 <_end+52268bc/205ca93c>
 >>edx; c566205c <_end+5226918/205ca93c>
 >>ebp; c0378350 <contig_page_data+b0/340>
 >>esp; c5663e58 <_end+5228714/205ca93c>

Code;  00000000 Before first symbol
00000000 <_EIP>:
Code;  00000000 Before first symbol
    0:   89 58 04                  mov    %ebx,0x4(%eax)
Code;  00000003 Before first symbol
    3:   89 03                     mov    %eax,(%ebx)
Code;  00000005 Before first symbol
    5:   89 53 04                  mov    %edx,0x4(%ebx)
Code;  00000008 Before first symbol
    8:   89 59 5c                  mov    %ebx,0x5c(%ecx)
Code;  0000000b Before first symbol
    b:   89 73 0c                  mov    %esi,0xc(%ebx)
Code;  0000000e Before first symbol
    e:   ff 41 68                  incl   0x68(%ecx)
Code;  00000011 Before first symbol
   11:   eb c1                     jmp    ffffffd4 <_EIP+0xffffffd4>
Code;  00000013 Before first symbol
   13:   0f 00 00                  sldtl  (%eax)

Re: Unable to handle kernel NULL pointer kernel 2.4.21-pre3-ac4

[Tupshin Harper]

FYI...the output I showed previously was on a tainted kernel(vmware 
modules), but after a fresh reboot and on the identical but untainted 
kernel, I got the same error while doing the same thing, namely 
converting a mysql myisam table to innodb. I did succesfully convert 
some tables without a problem, and the table that triggered the problem 
originally did succeed after rebooting.

-Tupshin

Re: Unable to handle kernel NULL pointer kernel 2.4.21-pre3-ac4

[Tupshin Harper]

OK....this definitely(well 99.5% chance) seems to be a problem that is 
in 2.4.21-pre3-ac4, but is not in 2.4.21-pre3, or in linus' bk tree. 
Also, this problem is not caused by the device-mapper patch which is the 
only reason I was trying the ac tree in the first place.

So, it's an ac specific problem separate from the device-mapper code:

Another (possibly related) hint is that at bootup, I get many(measured 
in the hundreds) reports of "ide: no cache flush required" which is in 
ide_cacheflush_p in drivers/ide/ide-disk.c, and is only present in the 
ac tree.

Does this seem like a likely culprit?

Hello...is this thing on...can anyone hear me? ;-)

-Tupshin

Re: Unable to handle kernel NULL pointer kernel 2.4.21-pre3-ac4

[Tupshin Harper]

Bertrand VIEILLE [Bébert] wrote:

>Hello !!
>
>I have the same problem with -ac tree.
>
>Alan Cox said he suspected several things to induce this oops:
>
>* Guess #1 is reverting mm/shmem.c.
>* Guess #2 is reverting the buffer cache changes.
>* Guess #3 is new IDE + highmem
>
no highmem enabled here.

>* Guess #4 is quota related (are people seeing the problem with quota
>disabled ?)
>
I do have quota enabled, so I'll try it without just to double check 
your results.

>
>Personnally, I answered him, I dont'have quota enabled, so Guess #4 doesn't
>exist anymore.
>
>  
>
It's looking like shmem or buffer cache.

The call trace that I posted certainly looks more like the buffer cache, 
but obviously doesn't eliminate shmem as the culprit. Is there an easy 
way to back out one or the other of these changes? I'm happy to do some 
testing.

-Tupshin