I'm not sure if I caught a complete instance of this in "syslog", but
here's what I've been seeing with 3.3.0-rc[1-3] kernels. Shortly after
booting multi-user, logging in, and attempting a package build, the
"fun" starts: the following message scrolls repeatedly (possibly with
different details) across the console. To date, I've had to hit the
reset switch to recover, which is why I'm not sure whether I caught a
complete syslog entry.
BUG: scheduling while atomic: kjournald/943/0x00000002
Modules linked in: snd_seq_midi snd_emu10k1_synth snd_emux_synth snd_seq_virmidi snd_seq_midi_event snd_seq_midi_emul snd_seq snd_emu10k1 snd_rawmidi snd_ac97_codec ac97_bus snd_pcm snd_seq_device snd_timer snd_page_alloc snd_util_mem snd_hwdep snd soundcore af_packet ipv6 bsd_comp ppp_generic slhc usbhid uhci_hcd ehci_hcd usbcore usb_common binfmt_misc
Pid: 943, comm: kjournald Not tainted 3.3.0-rc3 #1
Call Trace:
[<c1269968>] ? __schedule+0x57/0x3df
[<c113e0de>] ? cfq_exit_icq+0x2f/0x39
[<c126a0be>] ? io_schedule+0x2f/0x3e
[<c10a72b4>] ? sleep_on_buffer+0x5/0x8
[<c1268acb>] ? __wait_on_bit+0x33/0x58
[<c10a72af>] ? unmap_underlying_metadata+0x29/0x29
[<c1268ba2>] ? out_of_line_wait_on_bit+0xb2/0xba
[<c10a72af>] ? unmap_underlying_metadata+0x29/0x29
[<c1036655>] ? autoremove_wake_function+0x2d/0x2d
[<c10a7283>] ? __wait_on_buffer+0x1c/0x1f
[<c10e9800>] ? journal_commit_transaction+0x9cf/0xf5f
[<c100309c>] ? do_IRQ+0x5c/0x6c
[<c103d2ec>] ? __wake_up+0x2f/0x5b
[<c103d2fa>] ? __wake_up+0x3d/0x5b
[<c1036817>] ? prepare_to_wait+0x44/0x60
[<c10ebe34>] ? kjournald+0xd1/0x206
[<c1036628>] ? wake_up_bit+0x5c/0x5c
Processor is an AMD K6-III/450. Disks are SCSI. Controller is an
Adaptec AHA-2930U2.
--Bob Tracy
On Mon, Feb 20, 2012 at 11:41:59PM -0600, Bob Tracy wrote:
> I'm not sure if I caught a complete instance of this in "syslog", but
> here's what I've been seeing with 3.3.0-rc[1-3] kernels. Shortly after
> booting multi-user, logging in, and attempting a package build, the
> "fun" starts: the following message scrolls repeatedly (possibly with
> different details) across the console. To date, I've had to hit the
> reset switch to recover, which is why I'm not sure whether I caught a
> complete syslog entry.
Hmm.. I've taken a quick look at fs/jbd/commit.c, and I don't see
anything obvious. Can you try compiling with LOCKDEP compiled and see
if you can reproduce it, so we can see what spin lock is being held?
Thanks,
- Ted
On Tue, Feb 21, 2012 at 07:42:40AM -0500, Ted Ts'o wrote:
> On Mon, Feb 20, 2012 at 11:41:59PM -0600, Bob Tracy wrote:
> > I'm not sure if I caught a complete instance of this in "syslog", but
> > here's what I've been seeing with 3.3.0-rc[1-3] kernels. Shortly after
> > booting multi-user, logging in, and attempting a package build, the
> > "fun" starts: the following message scrolls repeatedly (possibly with
> > different details) across the console. To date, I've had to hit the
> > reset switch to recover, which is why I'm not sure whether I caught a
> > complete syslog entry.
>
> Hmm.. I've taken a quick look at fs/jbd/commit.c, and I don't see
> anything obvious. Can you try compiling with LOCKDEP compiled and see
> if you can reproduce it, so we can see what spin lock is being held?
Was waiting for more specific instructions with respect to kernel config
options and how to use LOCKDEP (not familiar with it), and decided to
give 3.3.0-rc4 a try in the meantime. Did NOT see the kjournald issue.
I guess I'll label this a "watch" item for now, unless something that
changed between -rc3 and -rc4 can be identified as the probable fix.
Thanks.
--Bob