Fw: [Bugme-new] [Bug 2133] New: bad: scheduling while atomic!

Fw: [Bugme-new] [Bug 2133] New: bad: scheduling while atomic!

[Andrew Morton]

This is possibly a locking bug in the ipv6 code: someone somewhere
forgot to undo a lock and the task ended up in atomic state.
Possibly it is specific to x86_64.

Has anyone run the TAHI suite on x86?



Begin forwarded message:

Date: Wed, 18 Feb 2004 06:31:34 -0800
From: bugme-daemon@osdl.org
To: bugme-new@lists.osdl.org
Subject: [Bugme-new] [Bug 2133] New: bad: scheduling while atomic!


http://bugme.osdl.org/show_bug.cgi?id=2133

           Summary: bad: scheduling while atomic!
    Kernel Version: 2.6.3
            Status: NEW
          Severity: high
             Owner: other_other@kernel-bugs.osdl.org
         Submitter: michal-bugzilla@logix.cz


Hardware Environment: AMD64
Software Environment: kernel 2.6.3
Problem Description: On my AMD64 machine kernel 2.6.3 I'm getting these messages
while running the TAHI testsuite:

kernel: BUG: dst underflow 0: fffffff802cde24
kernel: last message repeated 6 times
kernel: last message repeated 9 times

kernel: bad: scheduling while atomic!
kernel: Call Trace:{fffff8012fd39}{schedule+73} {fffff8010f7e0}{default_idle+0}

kernel: bad: scheduling while atomic!
kernel: Call Trace:<ffffffff8012fd39>{schedule+73}
<ffffffff80131609>{autoremove_wake_function+9}
<ffffffff801302f8>{__wake_up_common+56} <ffffffff801a7a07>{do_get_write_access+807}
<ffffffff801302b0>{default_wake_function+0}
<ffffffff801a7ea2>{journal_get_undo_access+50}
<ffffffff80196bac>{ext3_try_to_allocate+60} <ffffffff80196ec4>{ext3_new_block+436}
<ffffffff801993f1>{ext3_alloc_block+17} <ffffffff80199805>{ext3_alloc_branch+85}
<ffffffff80199ece>{ext3_get_block_handle+558}
<ffffffff8016a6f4>{alloc_buffer_head+36}
<ffffffff80167fe6>{create_buffers+102} <ffffffff80168e1d>{__block_prepare_write+365}
<ffffffff80199fd0>{ext3_get_block+0} <ffffffff801697fa>{block_prepare_write+26}
<ffffffff8019a5ea>{ext3_prepare_write+106}
<ffffffff8014d62a>{generic_file_aio_write_nolock+1274}
<ffffffff8026e21f>{skb_copy_datagram_iovec+79}
<ffffffff8014da35>{generic_file_write_nolock+149}
<ffffffff8020ec09>{opost_block+425} <ffffffff8011171a>{sysret_careful+13}
<ffffffff80131600>{autoremove_wake_function+0} <ffffffff80210b97>{write_chan+551}
<ffffffff801302b0>{default_wake_function+0}
<ffffffff801302b0>{default_wake_function+0}
<ffffffff8014dc1c>{generic_file_writev+60} <ffffffff80166413>{do_readv_writev+435}
<ffffffff80165e00>{do_sync_write+0} <ffffffff80166659>{sys_writev+73}
<ffffffff801116b0>{system_call+124}

kernel: bad: scheduling while atomic!
kernel: Call Trace:<ffffffff8012fd39>{schedule+73}
<ffffffff801a0f2d>{__ext3_journal_stop+45}
<ffffffff8019a83d>{ext3_ordered_commit_write+189}
<ffffffff8014d8ba>{generic_file_aio_write_nolock+1930}
<ffffffff8026e21f>{skb_copy_datagram_iovec+79}
<ffffffff8014da35>{generic_file_write_nolock+149}
<ffffffff8020ec09>{opost_block+425} <ffffffff80131600>{autoremove_wake_function+0}
<ffffffff80210b97>{write_chan+551} <ffffffff801302b0>{default_wake_function+0}
<ffffffff801302b0>{default_wake_function+0}
<ffffffff8014dc1c>{generic_file_writev+60}
<ffffffff80166413>{do_readv_writev+435} <ffffffff80165e00>{do_sync_write+0}
<ffffffff80166659>{sys_writev+73} <ffffffff801116b0>{system_call+124} 

kernel: Warning: kfree_skb on hard IRG ffffff802c0d4f
kernel: bad: scheduling while atomic!
kernel: Call Trace:{fffff8012fd39}{schedule+73} {fffff8011171a}{sysret_careful+13}

The last message is repeating *many* times per second and so makes the system
unusable.

Steps to reproduce: running tests from 'nd' directory of TAHI testsuite, during
test #21 it breaks, all following tests fail.

Strange conjunction: at the time of the message the screen is sleeping, so I hit
"Alt" to wake it up and surprise - some tests start PASSing again! (at least the
simple ones, like 'ping')

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

Re: Fw: [Bugme-new] [Bug 2133] New: bad: scheduling while atomic!

[Nivedita Singhvi]

Andrew Morton rote:

> This is possibly a locking bug in the ipv6 code: someone somewhere
> forgot to undo a lock and the task ended up in atomic state.
> Possibly it is specific to x86_64.
> 
> Has anyone run the TAHI suite on x86?

Yep, we're running TAHI on x86, but haven't seen this.
Not sure we're on 2.6.3 yet, though.

We'll check it out..

thanks,
Nivedita