Re: 2.6.16-rc5-mm1

[Nick Piggin <nickpiggin@yahoo.com.au>]

Andrew Morton wrote:
> Nick Piggin <nickpiggin@yahoo.com.au> wrote:
> 
>>Andrew Morton wrote:
>>
>>>Nick Piggin <nickpiggin@yahoo.com.au> wrote:
>>>
>>>
>>>>>>Feb 28 15:13:42 ltg01-sid kernel: BUG: warning at
>>>>>>/usr/src/linux-mm/fs/inotify.c:533/inotify_d_instantiate()
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [show_trace+13/15] show_trace+0xd/0xf
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [dump_stack+21/23] dump_stack+0x15/0x17
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [inotify_d_instantiate+47/98]
>>>>>>inotify_d_instantiate+0x2f/0x62
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [d_instantiate+70/114]
>>>>>>d_instantiate+0x46/0x72
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [ext3_add_nondir+44/64]
>>>>>>ext3_add_nondir+0x2c/0x40
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [ext3_link+163/217] ext3_link+0xa3/0xd9
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [vfs_link+292/379] vfs_link+0x124/0x17b
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [sys_linkat+157/218] sys_linkat+0x9d/0xda
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [sys_link+20/25] sys_link+0x14/0x19
>>>>>>Feb 28 15:13:42 ltg01-sid kernel:  [syscall_call+7/11] syscall_call+0x7/0xb
>>>>>>
>>>>>>Here is dmesg http://www.stardust.webpages.pl/files/mm/2.6.16-rc5-mm1/mm-dmesg
>>>>>>Here is config http://www.stardust.webpages.pl/files/mm/2.6.16-rc5-mm1/mm-config
>>>>>
>>>>>
>>>>>This patch is causing that warning
>>>>>inotify-lock-avoidance-with-parent-watch-status-in-dentry.patch
>>>>>
>>>>
>>>>The warning is harmless really. I guess it can be removed.
>>>
>>>
>>>?    How did DCACHE_INOTIFY_PARENT_WATCHED get set on that dentry?
>>>
>>
>>I guess it is because I don't clear it on +ve => -ve conversions.
>>We _could_ do that AFAIKS, by hooking into d_delete, but it would
>>have to be yet another fsnotify hook there because it will have to
>>be done under lock.
>>
> 
> 
> How could a -ve dentry become the source of a link?  We should be sure we
> know why this happened before ignoring it..
> 

The source of a link? That wouldn't get d_instantiate called on it, would it?
I figured this was the destination.

I can reproduce a similar looking oops by unlinking a PARENT_WATCHED file, then
touching it again.

-- 
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com