textsearch in module = BUG: scheduling while atomic

textsearch in module = BUG: scheduling while atomic

[Felipe Dias]

hi, i have a short module in kernel 2.6.22 with uses textsearch to
search strings in network packages, and i have this erros in dmesg:

--------------
BUG: scheduling while atomic: apache2/0x10000101/3710
 [<c02aefe6>] __sched_text_start+0x56/0x7c8
 [<c0119284>] __wake_up_common+0x35/0x53
 [<c01197f6>] __wake_up+0x32/0x43
 [<c028ba0f>] tcp_v4_send_check+0x86/0xbc
 [<c011bc98>] __cond_resched+0x18/0x29
 [<c02af7be>] cond_resched+0x26/0x31
 [<c015e711>] __kmalloc+0x38/0x6e
 [<f8f6e0fa>] bm_init+0x28/0x1c9 [ts_bm]
 [<c0256fe4>] __alloc_skb+0x47/0x104
 [<c01ba6da>] textsearch_prepare+0xf4/0x11a
 [<f91af451>] hook_test+0x4c/0xb6 [test_mod]
 [<c028d416>] tcp_v4_rcv+0x84c/0x8aa
 [<c026ef84>] nf_iterate+0x38/0x6a
 [<c0273f1c>] ip_rcv_finish+0x0/0x294
 [<c026f0c1>] nf_hook_slow+0x4d/0xb5
 [<c0273f1c>] ip_rcv_finish+0x0/0x294
 [<c027462f>] ip_rcv+0x20b/0x4bd
 [<c0273f1c>] ip_rcv_finish+0x0/0x294
 [<c025aee2>] netif_receive_skb+0x2ef/0x309
 [<c025ca16>] process_backlog+0x7c/0xe9
 [<c025cbe6>] net_rx_action+0x95/0x186
 [<c012469f>] __do_softirq+0x6c/0xcf
 [<c0124734>] do_softirq+0x32/0x36
 [<c0124900>] local_bh_enable+0x7b/0x89
 [<c025ced9>] dev_queue_xmit+0x202/0x221
 [<c0279686>] ip_output+0x269/0x2a3
 [<c0278d95>] ip_queue_xmit+0x358/0x39a
 [<f8cd15a9>] journal_end+0xba/0xc2 [reiserfs]
 [<f8cc2c16>] reiserfs_dirty_inode+0x78/0x7e [reiserfs]
 [<c0286f8f>] tcp_transmit_skb+0x618/0x652
 [<c0288906>] __tcp_push_pending_frames+0x71c/0x7cc
 [<f8cbe957>] reiserfs_file_write+0x1841/0x1871 [reiserfs]
 [<c0256fe4>] __alloc_skb+0x47/0x104
 [<c027ec8b>] tcp_sendmsg+0x91b/0x9f0
 [<c0258623>] skb_copy_datagram_iovec+0x53/0x1d0
 [<c027fcbb>] tcp_recvmsg+0x8e5/0x9f0
 [<c0251227>] sock_aio_write+0xf3/0xfb
 [<c0160e8a>] do_sync_readv_writev+0xc0/0x103
 [<c0130125>] autoremove_wake_function+0x0/0x33
 [<c0252bc2>] sys_getsockname+0x9f/0xb0
 [<c01b6856>] copy_from_user+0x23/0x4f
 [<c0160d40>] rw_copy_check_uvector+0x50/0xaa
 [<c0161566>] do_readv_writev+0xbc/0x187
 [<c0251134>] sock_aio_write+0x0/0xfb
 [<c0118731>] do_page_fault+0x273/0x516
 [<c016166e>] vfs_writev+0x3d/0x48
 [<c0161a6a>] sys_writev+0x41/0x67
 [<c0103d16>] sysenter_past_esp+0x5f/0x85
 =======================

My module part with find the string with textsearch:
-------------------------------
        int pos, err;
        struct ts_config *conf;
        struct ts_state state;

        list_for_each_entry_safe(entry, next, &head, list) {

                conf = textsearch_prepare("bm", entry->content,
strlen(entry->content), GFP_KERNEL, TS_AUTOLOAD);
                if (IS_ERR(conf))
                        err = PTR_ERR(conf);

                pos = textsearch_find_continuous(conf, &state,
sb->data, sb->len);

                if (pos != -1)
                        printk("%s: Contet: %s match!\n", MODULE_NAME,
entry->content);

                textsearch_destroy(conf);

        }
-------------------------------

What is this message ? somebody can help-me ?
Please send-me the correct mail list if this is wrong place to this question.

Att.
Felipe Dias

Re: textsearch in module = BUG: scheduling while atomic

[Jiri Slaby]

On 11/04/2007 06:15 PM, Felipe Dias wrote:
> hi, i have a short module in kernel 2.6.22 with uses textsearch to
> search strings in network packages, and i have this erros in dmesg:
> 
> --------------
> BUG: scheduling while atomic: apache2/0x10000101/3710
>  [<c02aefe6>] __sched_text_start+0x56/0x7c8
>  [<c0119284>] __wake_up_common+0x35/0x53
>  [<c01197f6>] __wake_up+0x32/0x43
>  [<c028ba0f>] tcp_v4_send_check+0x86/0xbc
>  [<c011bc98>] __cond_resched+0x18/0x29
>  [<c02af7be>] cond_resched+0x26/0x31
>  [<c015e711>] __kmalloc+0x38/0x6e
>  [<f8f6e0fa>] bm_init+0x28/0x1c9 [ts_bm]
>  [<c0256fe4>] __alloc_skb+0x47/0x104
>  [<c01ba6da>] textsearch_prepare+0xf4/0x11a
>  [<f91af451>] hook_test+0x4c/0xb6 [test_mod]
>  [<c028d416>] tcp_v4_rcv+0x84c/0x8aa
>  [<c026ef84>] nf_iterate+0x38/0x6a
>  [<c0273f1c>] ip_rcv_finish+0x0/0x294
>  [<c026f0c1>] nf_hook_slow+0x4d/0xb5
>  [<c0273f1c>] ip_rcv_finish+0x0/0x294
>  [<c027462f>] ip_rcv+0x20b/0x4bd
>  [<c0273f1c>] ip_rcv_finish+0x0/0x294
>  [<c025aee2>] netif_receive_skb+0x2ef/0x309
>  [<c025ca16>] process_backlog+0x7c/0xe9
>  [<c025cbe6>] net_rx_action+0x95/0x186
>  [<c012469f>] __do_softirq+0x6c/0xcf
>  [<c0124734>] do_softirq+0x32/0x36
>  [<c0124900>] local_bh_enable+0x7b/0x89
>  [<c025ced9>] dev_queue_xmit+0x202/0x221
>  [<c0279686>] ip_output+0x269/0x2a3
>  [<c0278d95>] ip_queue_xmit+0x358/0x39a
>  [<f8cd15a9>] journal_end+0xba/0xc2 [reiserfs]
>  [<f8cc2c16>] reiserfs_dirty_inode+0x78/0x7e [reiserfs]
>  [<c0286f8f>] tcp_transmit_skb+0x618/0x652
>  [<c0288906>] __tcp_push_pending_frames+0x71c/0x7cc
>  [<f8cbe957>] reiserfs_file_write+0x1841/0x1871 [reiserfs]
>  [<c0256fe4>] __alloc_skb+0x47/0x104
>  [<c027ec8b>] tcp_sendmsg+0x91b/0x9f0
>  [<c0258623>] skb_copy_datagram_iovec+0x53/0x1d0
>  [<c027fcbb>] tcp_recvmsg+0x8e5/0x9f0
>  [<c0251227>] sock_aio_write+0xf3/0xfb
>  [<c0160e8a>] do_sync_readv_writev+0xc0/0x103
>  [<c0130125>] autoremove_wake_function+0x0/0x33
>  [<c0252bc2>] sys_getsockname+0x9f/0xb0
>  [<c01b6856>] copy_from_user+0x23/0x4f
>  [<c0160d40>] rw_copy_check_uvector+0x50/0xaa
>  [<c0161566>] do_readv_writev+0xbc/0x187
>  [<c0251134>] sock_aio_write+0x0/0xfb
>  [<c0118731>] do_page_fault+0x273/0x516
>  [<c016166e>] vfs_writev+0x3d/0x48
>  [<c0161a6a>] sys_writev+0x41/0x67
>  [<c0103d16>] sysenter_past_esp+0x5f/0x85
>  =======================
> 
> My module part with find the string with textsearch:
> -------------------------------
>         int pos, err;
>         struct ts_config *conf;
>         struct ts_state state;
> 
>         list_for_each_entry_safe(entry, next, &head, list) {
> 
>                 conf = textsearch_prepare("bm", entry->content,
> strlen(entry->content), GFP_KERNEL, TS_AUTOLOAD);

I guess you want GFP_ATOMIC while calling it from (soft)irq.

regards,
-- 
Jiri Slaby (jirislaby@gmail.com)
Faculty of Informatics, Masaryk University

Re: textsearch in module = BUG: scheduling while atomic

[Felipe Dias]

Ow sorry... the warning:

BUG: scheduling while atomic: gnome-cups-icon/0x00000101/3827
 [<c02aefe6>] __sched_text_start+0x56/0x7c8
 [<c0130139>] autoremove_wake_function+0x14/0x33
 [<c0119284>] __wake_up_common+0x35/0x53
 [<c01197f6>] __wake_up+0x32/0x43
 [<c02af847>] wait_for_completion+0x6a/0x9f
 [<c011b6ae>] default_wake_function+0x0/0xc
 [<c012cdff>] call_usermodehelper_keys+0xad/0xc5
 [<c012ceec>] request_module+0xd5/0xe6
 [<c01b2762>] __next_cpu+0x12/0x21
 [<c0119e54>] find_busiest_group+0x1d6/0x55a
 [<c01ba672>] textsearch_prepare+0x8c/0x11a
 [<c017c11d>] __find_get_block_slow+0x127/0x131
 [<f9198451>] hook_test+0x4c/0xb6 [test_mod]
 [<c017c3f0>] __find_get_block+0x13f/0x149
 [<c026ef84>] nf_iterate+0x38/0x6a
 [<c0273f1c>] ip_rcv_finish+0x0/0x294
 [<c026f0c1>] nf_hook_slow+0x4d/0xb5
 [<c0273f1c>] ip_rcv_finish+0x0/0x294
 [<c027462f>] ip_rcv+0x20b/0x4bd
 [<c0273f1c>] ip_rcv_finish+0x0/0x294
 [<f8ccd2ad>] search_by_key+0x17f/0xe87 [reiserfs]
 [<c0127b32>] del_timer_sync+0xa/0x14
 [<c02afdfb>] schedule_timeout+0x79/0x8d
 [<c025aee2>] netif_receive_skb+0x2ef/0x309
 [<c026e201>] __nla_put+0xe/0x25
 [<c025ca16>] process_backlog+0x7c/0xe9
 [<c025cbe6>] net_rx_action+0x95/0x186
 [<c012469f>] __do_softirq+0x6c/0xcf
 [<c0124734>] do_softirq+0x32/0x36
 [<c0124900>] local_bh_enable+0x7b/0x89
 [<c025ced9>] dev_queue_xmit+0x202/0x221
 [<c0279686>] ip_output+0x269/0x2a3
 [<c0278d95>] ip_queue_xmit+0x358/0x39a
 [<c014b589>] __alloc_pages+0x64/0x2a8
 [<c015e2ee>] cache_alloc_refill+0x74/0x45f
 [<c028ba0f>] tcp_v4_send_check+0x86/0xbc
 [<c0286f8f>] tcp_transmit_skb+0x618/0x652
 [<c0256fe4>] __alloc_skb+0x47/0x104
 [<c0289319>] tcp_connect+0x2a3/0x31d
 [<c028c0cd>] tcp_v4_connect+0x493/0x5a4
 [<c0171eb3>] touch_atime+0x8b/0xac
 [<c0296f23>] inet_stream_connect+0x87/0x20d
 [<c01b6856>] copy_from_user+0x23/0x4f
 [<c02518b5>] sys_connect+0x82/0xad
 [<c02b0c6c>] _spin_lock_bh+0x8/0x18
 [<c02b0c6c>] _spin_lock_bh+0x8/0x18
 [<c02538c3>] release_sock+0x12/0x8c
 [<c027d938>] tcp_setsockopt+0x309/0x321
 [<c016fb25>] d_alloc+0x14b/0x17e
 [<c016f9d6>] d_instantiate+0x66/0x6a
 [<c02532a4>] sock_common_setsockopt+0x1d/0x22
 [<c0251814>] sys_setsockopt+0x88/0xa7
 [<c0252e33>] sys_socketcall+0x8f/0x242
 [<c0103d16>] sysenter_past_esp+0x5f/0x85
 =======================


Thanks,
Felipe Dias

On 11/4/07, Jiri Slaby <jirislaby@gmail.com> wrote:
> On 11/04/2007 09:22 PM, Felipe Dias wrote:
> > Hi, thanks for your fast reply, I try, I put GFP_ATOMIC.
> > The error remains, but with less frequency. The sometimes appears, is
> > not where he does the checking. Any ideia ?
>
> Post the trace (the BUG warning) with the GFP_ATOMIC used.
>
> regards,
> --
> Jiri Slaby (jirislaby@gmail.com)
> Faculty of Informatics, Masaryk University
>

Re: textsearch in module = BUG: scheduling while atomic

[Jiri Slaby]

On 11/04/2007 11:05 PM, Felipe Dias wrote:
> Ow sorry... the warning:
> 
> BUG: scheduling while atomic: gnome-cups-icon/0x00000101/3827
>  [<c02aefe6>] __sched_text_start+0x56/0x7c8
>  [<c0130139>] autoremove_wake_function+0x14/0x33
>  [<c0119284>] __wake_up_common+0x35/0x53
>  [<c01197f6>] __wake_up+0x32/0x43
>  [<c02af847>] wait_for_completion+0x6a/0x9f
>  [<c011b6ae>] default_wake_function+0x0/0xc
>  [<c012cdff>] call_usermodehelper_keys+0xad/0xc5
>  [<c012ceec>] request_module+0xd5/0xe6

Seems like it is not inteded for using in atomic at all (you probably passed an
unknown algo here to the prepare function). You seem to have to use a workqueue
if it is possible or prepare the serach before the interrupt occurs. Post your
code if you want more appropriate answer.

>  [<c01b2762>] __next_cpu+0x12/0x21
>  [<c0119e54>] find_busiest_group+0x1d6/0x55a
>  [<c01ba672>] textsearch_prepare+0x8c/0x11a
>  [<c017c11d>] __find_get_block_slow+0x127/0x131
>  [<f9198451>] hook_test+0x4c/0xb6 [test_mod]
>  [<c017c3f0>] __find_get_block+0x13f/0x149
>  [<c026ef84>] nf_iterate+0x38/0x6a
>  [<c0273f1c>] ip_rcv_finish+0x0/0x294
>  [<c026f0c1>] nf_hook_slow+0x4d/0xb5
>  [<c0273f1c>] ip_rcv_finish+0x0/0x294
>  [<c027462f>] ip_rcv+0x20b/0x4bd
>  [<c0273f1c>] ip_rcv_finish+0x0/0x294
>  [<f8ccd2ad>] search_by_key+0x17f/0xe87 [reiserfs]
>  [<c0127b32>] del_timer_sync+0xa/0x14
>  [<c02afdfb>] schedule_timeout+0x79/0x8d
>  [<c025aee2>] netif_receive_skb+0x2ef/0x309
>  [<c026e201>] __nla_put+0xe/0x25
>  [<c025ca16>] process_backlog+0x7c/0xe9
>  [<c025cbe6>] net_rx_action+0x95/0x186
>  [<c012469f>] __do_softirq+0x6c/0xcf
>  [<c0124734>] do_softirq+0x32/0x36
>  [<c0124900>] local_bh_enable+0x7b/0x89
>  [<c025ced9>] dev_queue_xmit+0x202/0x221
>  [<c0279686>] ip_output+0x269/0x2a3
>  [<c0278d95>] ip_queue_xmit+0x358/0x39a
>  [<c014b589>] __alloc_pages+0x64/0x2a8
>  [<c015e2ee>] cache_alloc_refill+0x74/0x45f
>  [<c028ba0f>] tcp_v4_send_check+0x86/0xbc
>  [<c0286f8f>] tcp_transmit_skb+0x618/0x652
>  [<c0256fe4>] __alloc_skb+0x47/0x104
>  [<c0289319>] tcp_connect+0x2a3/0x31d
>  [<c028c0cd>] tcp_v4_connect+0x493/0x5a4
>  [<c0171eb3>] touch_atime+0x8b/0xac
>  [<c0296f23>] inet_stream_connect+0x87/0x20d
>  [<c01b6856>] copy_from_user+0x23/0x4f
>  [<c02518b5>] sys_connect+0x82/0xad
>  [<c02b0c6c>] _spin_lock_bh+0x8/0x18
>  [<c02b0c6c>] _spin_lock_bh+0x8/0x18
>  [<c02538c3>] release_sock+0x12/0x8c
>  [<c027d938>] tcp_setsockopt+0x309/0x321
>  [<c016fb25>] d_alloc+0x14b/0x17e
>  [<c016f9d6>] d_instantiate+0x66/0x6a
>  [<c02532a4>] sock_common_setsockopt+0x1d/0x22
>  [<c0251814>] sys_setsockopt+0x88/0xa7
>  [<c0252e33>] sys_socketcall+0x8f/0x242
>  [<c0103d16>] sysenter_past_esp+0x5f/0x85
>  =======================

regards,
-- 
Jiri Slaby (jirislaby@gmail.com)
Faculty of Informatics, Masaryk University

Re: textsearch in module = BUG: scheduling while atomic

[Jiri Slaby]

On 11/04/2007 11:16 PM, Jiri Slaby wrote:
> On 11/04/2007 11:05 PM, Felipe Dias wrote:
>> Ow sorry... the warning:
>>
>> BUG: scheduling while atomic: gnome-cups-icon/0x00000101/3827
>>  [<c02aefe6>] __sched_text_start+0x56/0x7c8
>>  [<c0130139>] autoremove_wake_function+0x14/0x33
>>  [<c0119284>] __wake_up_common+0x35/0x53
>>  [<c01197f6>] __wake_up+0x32/0x43
>>  [<c02af847>] wait_for_completion+0x6a/0x9f
>>  [<c011b6ae>] default_wake_function+0x0/0xc
>>  [<c012cdff>] call_usermodehelper_keys+0xad/0xc5
>>  [<c012ceec>] request_module+0xd5/0xe6
> 
> Seems like it is not inteded for using in atomic at all (you probably passed an
> unknown algo here to the prepare function). You seem to have to use a workqueue
> if it is possible or prepare the serach before the interrupt occurs.

Or pass 0 instead of TS_AUTOLOAD.

Re: textsearch in module = BUG: scheduling while atomic

[Felipe Dias]

I change the algo, from "bm" to "kmp" and probably resolve. I will
make more tests and if error occour i post latter...

Thanks so much.
Felipe Dias

On 11/4/07, Jiri Slaby <jirislaby@gmail.com> wrote:
> On 11/04/2007 11:16 PM, Jiri Slaby wrote:
> > On 11/04/2007 11:05 PM, Felipe Dias wrote:
> >> Ow sorry... the warning:
> >>
> >> BUG: scheduling while atomic: gnome-cups-icon/0x00000101/3827
> >>  [<c02aefe6>] __sched_text_start+0x56/0x7c8
> >>  [<c0130139>] autoremove_wake_function+0x14/0x33
> >>  [<c0119284>] __wake_up_common+0x35/0x53
> >>  [<c01197f6>] __wake_up+0x32/0x43
> >>  [<c02af847>] wait_for_completion+0x6a/0x9f
> >>  [<c011b6ae>] default_wake_function+0x0/0xc
> >>  [<c012cdff>] call_usermodehelper_keys+0xad/0xc5
> >>  [<c012ceec>] request_module+0xd5/0xe6
> >
> > Seems like it is not inteded for using in atomic at all (you probably passed an
> > unknown algo here to the prepare function). You seem to have to use a workqueue
> > if it is possible or prepare the serach before the interrupt occurs.
>
> Or pass 0 instead of TS_AUTOLOAD.
>

Re: textsearch in module = BUG: scheduling while atomic

[Jiri Slaby]

On 11/04/2007 11:57 PM, Felipe Dias wrote:
> I change the algo, from "bm" to "kmp" and probably resolve. I will
> make more tests and if error occour i post latter...

Anyway, you should do what I suggested before -- pass 0 instead of TS_AUTOLOAD.
If somebody doesn't have kmp module loaded it will spit out the warnings again,
because the kernel will try to load it (and it's not atomic operation).

regards,
-- 
Jiri Slaby (jirislaby@gmail.com)
Faculty of Informatics, Masaryk University