BUG in JFFS2

BUG in JFFS2

[Artem B. Bityuckiy]

Hello,

I've hit on bug in JFFS2, but I can not reproduce it easilly yet. I have 
  stess test which in general creates files, deletes them, creates 
directories and deletes them, and so on so on. When I run this test 
nightly, I see the following error when I come at work in the morning:

kernel BUG at 
/tester/linux/linux-2.6.8-1.521up-custom-nopreempt/fs/jffs2/gc.c:1164!
invalid operand: 0000 [#1]
DEBUG_PAGEALLOC
Modules linked in: jffs2 nandsim zlib_deflate mtdchar nand mtdpart 
nand_ids mtdcore nand_ecc nfsd exportfs md5 ipv6 autofs4d
CPU:    0
EIP:    0060:[<22aa58a0>]    Not tainted
EFLAGS: 00010206   (2.6.8-1.521up-custom-nopreempt)
EIP is at jffs2_garbage_collect_dnode+0x3f0/0x407 [jffs2]
eax: 003976f0   ebx: 00000000   ecx: 00397850   edx: 00000000
esi: 13fc3e00   edi: 13fc3e2c   ebp: 07198ea8   esp: 07198e08
ds: 007b   es: 007b   ss: 0068
Process mkfile (pid: 11938, threadinfo=07198000 task=11dbc9f0)
Stack: 00000010 00000fff 07198e3c 0220c6e7 00398000 07198e2c 00000246 
00000000
        00397000 003976f0 0902edf8 00000000 13fc3e40 000aa8f5 07198e48 
023660b4
        07198e7c 00000000 00000000 00000000 00000000 00000000 00000000 
00000000
Call Trace:
  [<02107c7a>] show_stack+0x7a/0x90
  [<02107df7>] show_registers+0x147/0x1f0
  [<0210801e>] die+0xee/0x270
  [<021084da>] do_invalid_op+0x9a/0xa0
  [<ffff3454>] 0xffff3454
  [<22aa4131>] jffs2_garbage_collect_live+0x2c1/0x4f0 [jffs2]
  [<22aa351c>] jffs2_garbage_collect_pass+0x93c/0x1290 [jffs2]
  [<22aab132>] jffs2_flush_wbuf_gc+0xa2/0x140 [jffs2]
  [<22a9a7ca>] jffs2_fsync+0x1a/0x20 [jffs2]
  [<0217d0c4>] sys_fsync+0xb4/0xf0
  [<ffff3200>] 0xffff3200
Code: 0f 0b 8c 04 b4 ed aa 22 eb c5 0f 0b 28 04 b4 ed aa 22 e9 d5


May be somebody may comment this without having the test which can 
easilly reproduce this bug?

P.S. My kernel is 2.6.8.1, UP, Preemption is disabled.

P.P.S. I observed this several times.

-- 
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: BUG in JFFS2

[David Woodhouse]

On Mon, 2004-11-15 at 11:27 +0300, Artem B. Bityuckiy wrote:
> Hello,
> 
> I've hit on bug in JFFS2, but I can not reproduce it easilly yet. I have 
>   stess test which in general creates files, deletes them, creates 
> directories and deletes them, and so on so on. When I run this test 
> nightly, I see the following error when I come at work in the morning:
> 
> kernel BUG at 
> /tester/linux/linux-2.6.8-1.521up-custom-nopreempt/fs/jffs2/gc.c:1164!

In my copy of v1.137 there's three instances of BUG_ON() around there,
but none at line 1164. They're at lines 116[123]. 

Start by changing it from a simple BUG_ON() to a printk saying precisely
what went wrong and giving all the other information which might be
relevant. If you can just trap into a debugger when it happens, that's
even better.

-- 
dwmw2

Re: BUG in JFFS2

[Artem B. Bityuckiy]

David Woodhouse wrote:
> On Mon, 2004-11-15 at 11:27 +0300, Artem B. Bityuckiy wrote:
> 
>>Hello,
>>
>>I've hit on bug in JFFS2, but I can not reproduce it easilly yet. I have 
>>  stess test which in general creates files, deletes them, creates 
>>directories and deletes them, and so on so on. When I run this test 
>>nightly, I see the following error when I come at work in the morning:
>>
>>kernel BUG at 
>>/tester/linux/linux-2.6.8-1.521up-custom-nopreempt/fs/jffs2/gc.c:1164!
> 
> 
> In my copy of v1.137 there's three instances of BUG_ON() around there,
> but none at line 1164. They're at lines 116[123]. 
I have the lastest MTD snapshot, v1.140. This is BUG_ON(end > 
JFFS2_F_I_SIZE(f));

> 
> Start by changing it from a simple BUG_ON() to a printk saying precisely
> what went wrong and giving all the other information which might be
> relevant. If you can just trap into a debugger when it happens, that's
> even better.
I'll do this as soos as I have done other pending JFFS2 tasks :-) I 
think I'll insert useful printkeyss there since now I have no configured 
environment to use debugger.

-- 
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.