Issues with all kernels after 3.3.7

[u.kleine-koenig@pengutronix.de (Uwe Kleine-König)]

Hello,

On Fri, Aug 17, 2012 at 05:30:27PM +0000, Alan M Butler wrote:
> On 17 August 2012 14:01, Alan M Butler <alanbutty12@gmail.com> wrote:
> > On 17 August 2012 13:26, Uwe Kleine-K?nig
> > <u.kleine-koenig@pengutronix.de> wrote:
> >> On Fri, Aug 17, 2012 at 01:47:06PM +0100, alan butler wrote:
> >>> I have been trying all kernels after 3.3.7 (except for the 3.5 series)
> >>> on my ix2-200
> >>> and have found that on every kernel 3.3.8, 3.4, 3.4.2, 3.4.4 =>3.4.7
> >>> and now even on 3.6 rc1
> >>> and the next dated today 17 august that when i use git for example:
> >>>
> >>> git clone git://git.videolan.org/x264
> >>>
> >>> my system just crash's / hangs and even through the serial port i get
> >>> no response.
> >>> This also happens when i try to access a web interface for example the
> >>> serviio java webui.
> >>>
> >>> The second issue i have noticed is that my raid 0 array hangs / pauses
> >>> when being mounted
> >>> at system startup. (for approximatly 30 seconds maybe more maybe less
> >>> im not certain).
> >>> But again on the 3.3.7 kernel there was no issue and no hang.
> >>>
> >>> If i return to 3.3.7 everything works fine with no other modifications
> >>> to the os or kernels.
> >>> I am using debian wheezy at the moment but also tried debian squeeze
> >>> so it does not
> >>> seem to be related to the specific linux version just the kernel.
> >> Can you bisect your problem? Doing that between 3.3.7 and 3.3.8 seems to
> >> be the obvious range to test.
> >>
> >> You can also try to enable the various debugging options like
> >>
> >>         CONFIG_DETECT_HUNG_TASK
> >>         CONFIG_PROVE_LOCKING
> >>         CONFIG_DEBUG_ATOMIC_SLEEP
> >>         CONFIG_MAGIC_SYSRQ
> >>
> >> or try https://lkml.org/lkml/2012/5/26/83.
> >>
> >> Best regards
> >> Uwe
> >>
> >> --
> >> Pengutronix e.K.                           | Uwe Kleine-K?nig            |
> >> Industrial Linux Solutions                 | http://www.pengutronix.de/  |
> >
> > i enabled the options you said there and i see alot of the following
> > popping up while connected through serial:
> >
> >  BUG: sleeping function called from invalid context at
> > include/linux/freezer.h:46
> > [  126.896958] in_atomic(): 0, irqs_disabled(): 128, pid: 2180, name:
> > console-kit-dae
> > [  126.904566] no locks held by console-kit-dae/2180.
> > [  126.909378] irq event stamp: 27643
> > [  126.912797] hardirqs last  enabled at (27642): [<c03781ac>]
> > _raw_spin_unlock_irqrestore+0x3c/0x5c
> > [  126.921735] hardirqs last disabled at (27643): [<c00090cc>]
> > ret_fast_syscall+0xc/0x38
> > [  126.929618] softirqs last  enabled at (25565): [<c001bc20>]
> > irq_exit+0x54/0xb8
> > [  126.936890] softirqs last disabled at (25558): [<c001bc20>]
> > irq_exit+0x54/0xb8
> > [  126.944185] [<c000ea8c>] (unwind_backtrace+0x0/0xe0) from
> > [<c000b414>] (do_signal+0x84/0x5c0)
> > [  126.952764] [<c000b414>] (do_signal+0x84/0x5c0) from [<c000be0c>]
> > (do_notify_resume+0x18/0x60)
> > [  126.961430] [<c000be0c>] (do_notify_resume+0x18/0x60) from
> > [<c0009120>] (work_pending+0x24/0x28)
> >
> > there seems to be alot more of them when i have serviio upnp server running.
> 
> After a little testing with those config options enabled that you
> sujested iv found that the problem with git first appears in the 3.4.1
> kernel.
> For example:
> in 3.4.0 kernel i can use 'git clone git://git.videolan.org/x264' successfully.
> 
> From 3.4.1 kernel on  I can not use 'git clone
> git://git.videolan.org/x264' the system hangs / crashes with no output
> at all.
> 
> the other issue the hang / stall while mounting my etx4 raid 0 is
> actualy much more recent than i remembered i have tested each kernel
> from 3.4.0 all the way to 3.4.9 with the config options enabled as
> sujested before and the stall first starts in kernel 3.4.8 and the
> following bug keeps popping up repeatedly until the raid is mounted
> and then anytime a disk is accessed it seems. I was certain it had
> been popping up before 3.4.8.
> 
> The following is one of what pops up:
> 
>  BUG: sleeping function called from invalid context at
> include/linux/freezer.h:46
> in_atomic(): 0, irqs_disabled(): 128, pid: 2166, name: minissdpd
> no locks held by minissdpd/2166.
>  irq event stamp: 2081
> hardirqs last  enabled at (2080): [<c02f8dc8>] _raw_spin_unlock_irq+0x24/0x4c
> hardirqs last disabled at (2081): [<c00090cc>] ret_fast_syscall+0xc/0x38
> softirqs last  enabled at (0): [<c0016b38>] copy_process+0x3f8/0xfe8
> softirqs last disabled at (0): [<  (null)>]   (null)
> [<c000eae4>] (unwind_backtrace+0x0/0xe0) from [<c000b5a0>]
> (do_signal+0x84/0x554)
> [<c000b5a0>] (do_signal+0x84/0x554) from [<c000bec0>]
> (do_notify_resume+0x18/0x60)
> [<c000bec0>] (do_notify_resume+0x18/0x60) from [<c0009120>]
> (work_pending+0x24/0x28)
I think this is an unrelated issue that I think is fixed in later
kernels. So I'd disable CONFIG_DEBUG_ATOMIC_SLEEP for further testing.
Can you try a bisection, i.e.

	git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
	cd linus
	git remote add -f stable git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git
	git bisect v3.4.1 v3.4

and test if the kernel that is checked out then results in a freeze.
Depending on the test result either do:

	git bisect good # i.e. problem doesn't occur

or

	git bisect bad # problem is reproducible

Repeat that until git points out the first bad commit and report that.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |