Re: sock lockup -> process in D state [Was: 2.6.26-rc5-mm1]

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 10 Jun 2008 00:33:01 +0200
Jiri Slaby <jirislaby@gmail.com> wrote:

> On 06/09/2008 02:39 PM, Andrew Morton wrote:
> >   ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.26-rc5/2.6.26-rc5-mm1/
> 
> I don't know how to reproduce it so far, but posting anyway:
> 
> httpd2-prefor D 00000000ffffffff     0  3697   2811
>   ffff810055bd7198 0000000000000046 ffff810055bd7160 ffff810055bd715c
>   ffffffff80728000 ffff810063896700 ffff81007d093380 ffff810063896980
>   00000001781e1700 00000001005585a6 ffff810063896980 0000000000000600
> Call Trace:
>   [<ffffffff8049d6fd>] lock_sock_nested+0x8d/0xd0
>   [<ffffffff8024e010>] ? autoremove_wake_function+0x0/0x40
>   [<ffffffff8049a00c>] sock_fasync+0x4c/0x180
>   [<ffffffff8049b4ad>] sock_close+0x1d/0x40
>   [<ffffffff802a6314>] __fput+0xc4/0x1a0
>   [<ffffffff802a640d>] fput+0x1d/0x30
>   [<ffffffff802a2b9b>] filp_close+0x5b/0x90
>   [<ffffffff802399e9>] put_files_struct+0x79/0xe0
>   [<ffffffff80239a9e>] exit_files+0x4e/0x60
>   [<ffffffff8023bc44>] do_exit+0x6e4/0x7f0
>   [<ffffffff8022cade>] ? __wake_up+0x4e/0x70
>   [<ffffffff8054bd18>] oops_end+0x88/0x90
>   [<ffffffff8020e0be>] die+0x5e/0x90
>   [<ffffffff8054c386>] do_trap+0x146/0x170
>   [<ffffffff8054e3e5>] ? atomic_notifier_call_chain+0x15/0x20
>   [<ffffffff8020e7b2>] do_invalid_op+0x92/0xb0
>   [<ffffffff80277ec7>] ? unlock_page+0x17/0x40
>   [<ffffffff8029ee95>] ? check_object+0x265/0x270
>   [<ffffffff8029e6c0>] ? init_object+0x50/0x90
>   [<ffffffff8054b739>] error_exit+0x0/0x51
>   [<ffffffff80277ec7>] ? unlock_page+0x17/0x40
>   [<ffffffff80277edd>] ? unlock_page+0x2d/0x40
>   [<ffffffff8028386d>] ? shrink_page_list+0x2fd/0x720
>   [<ffffffff80282ad3>] ? isolate_pages_global+0x1c3/0x270
>   [<ffffffff80283ed0>] ? shrink_list+0x240/0x5e0
>   [<ffffffff8029e6c0>] ? init_object+0x50/0x90
>   [<ffffffff802844c3>] ? shrink_zone+0x253/0x330
>   [<ffffffff80280060>] ? background_writeout+0x0/0xe0
>   [<ffffffff80285441>] ? try_to_free_pages+0x251/0x3e0
>   [<ffffffff80282910>] ? isolate_pages_global+0x0/0x270
>   [<ffffffff8027e8ff>] ? __alloc_pages_internal+0x20f/0x4e0
>   [<ffffffff802a0e04>] ? __slab_alloc+0x6d4/0x6f0
>   [<ffffffff804d1521>] ? sk_stream_alloc_skb+0x41/0x110
>   [<ffffffff804d1521>] ? sk_stream_alloc_skb+0x41/0x110
>   [<ffffffff802a2033>] ? __kmalloc_track_caller+0xc3/0xf0
>   [<ffffffff804a2b4e>] ? __alloc_skb+0x6e/0x150
>   [<ffffffff804d1521>] ? sk_stream_alloc_skb+0x41/0x110
>   [<ffffffff804d1960>] ? tcp_sendmsg+0x370/0xc30
>   [<ffffffff80499b07>] ? sock_aio_write+0x137/0x150
>   [<ffffffff802bb711>] ? touch_atime+0x31/0x140
>   [<ffffffff804999d0>] ? sock_aio_write+0x0/0x150
>   [<ffffffff802a4bcb>] ? do_sync_readv_writev+0xeb/0x130
>   [<ffffffff802aceb1>] ? pipe_read+0x4b1/0x4c0
>   [<ffffffff8024e010>] ? autoremove_wake_function+0x0/0x40
>   [<ffffffff802a4e41>] ? do_sync_read+0xf1/0x140
>   [<ffffffff802a49de>] ? rw_copy_check_uvector+0x7e/0x130
>   [<ffffffff802a538f>] ? do_readv_writev+0xcf/0x1e0
>   [<ffffffff802d985b>] ? sys_epoll_wait+0xcb/0x560
>   [<ffffffff802a54e0>] ? vfs_writev+0x40/0x60
>   [<ffffffff802a5550>] ? sys_writev+0x50/0xb0
>   [<ffffffff8020c42b>] ? system_call_after_swapgs+0x7b/0x80

Looks like it tried to go BUG then deadlocked.  Didn't it print BUG
stuff into the logs?

And I'd guess that you hit the unlock_page() BUG (Subject: Re:
2.6.26-rc5-mm1: kernel BUG at mm/filemap.c:575!).

This bug means that rc5-mm1 is toast, sorry.