Preempt-rt on 3.2.23 raspberry pi kernel

Preempt-rt on 3.2.23 raspberry pi kernel

[Jan Sommer]

Hello,

This probably is related to the this thread
http://comments.gmane.org/gmane.linux.rt.user/8424 but I don't know how to
answer to it as I have just subscribed to the mailinglist.
As Tom I used the 3.2.23 branch for the raspberry pi from bootc
(http://www.bootc.net/projects/raspberry-pi-kernel/) and applied the
rt-patch. Because Tom told about problems with the USB-driver I set up my
system to mount the rootfs via NFS.
The patched kernel boots up fine when using the voluntary, low latency and
basic preempt option but fails to boot after compiling it with the
preempt_full option.
You can find a screenshot (literally ;-) ) of the last kernel
output here: http://imageshack.us/photo/my-images/140/rpikernellog.jpg/

It complains about a certain part of the code within rtmutex.c.

As I'm fairly new to building kernel related things, does anyone has a
suggestion what to do? The problem occurs in both versions of the patch
(rt36 and rt37) and it does not matter if I want to mount the rootfs via
sd-card or NFS (probably because the error appears even before).
At the moment I don't have the chance to hook up a serial cable to the
serial port as I don't have level converters.


Best regards,

        Jan

Re: Preempt-rt on 3.2.23 raspberry pi kernel

[Nick Lott]

I have a complete kernel log of the same problem but on 3.16 with rt27.

http://www.raspberrypi.org/phpBB3/viewtopic.php?f=29&t=9311

I've not had time to go any further with this or update and test with a newer kerenel.

Hope this helps.

Nick Lott
http://www.brokentoaster.com/

On 29 Jul 2012, at 9:32PM, Jan Sommer wrote:

> Hello,
> 
> This probably is related to the this thread
> http://comments.gmane.org/gmane.linux.rt.user/8424 but I don't know how to
> answer to it as I have just subscribed to the mailinglist.
> As Tom I used the 3.2.23 branch for the raspberry pi from bootc
> (http://www.bootc.net/projects/raspberry-pi-kernel/) and applied the
> rt-patch. Because Tom told about problems with the USB-driver I set up my
> system to mount the rootfs via NFS.
> The patched kernel boots up fine when using the voluntary, low latency and
> basic preempt option but fails to boot after compiling it with the
> preempt_full option.
> You can find a screenshot (literally ;-) ) of the last kernel
> output here: http://imageshack.us/photo/my-images/140/rpikernellog.jpg/
> 
> It complains about a certain part of the code within rtmutex.c.
> 
> As I'm fairly new to building kernel related things, does anyone has a
> suggestion what to do? The problem occurs in both versions of the patch
> (rt36 and rt37) and it does not matter if I want to mount the rootfs via
> sd-card or NFS (probably because the error appears even before).
> At the moment I don't have the chance to hook up a serial cable to the
> serial port as I don't have level converters.
> 
> 
> Best regards,
> 
>       Jan
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Preempt-rt on 3.2.23 raspberry pi kernel

[Thomas Gleixner]

On Mon, 30 Jul 2012, Nick Lott wrote:

> I have a complete kernel log of the same problem but on 3.16 with rt27.
> 
> http://www.raspberrypi.org/phpBB3/viewtopic.php?f=29&t=9311
> 
> I've not had time to go any further with this or update and test
> with a newer kerenel.

Unfortunately verbose debug info is off, but it looks like you are
running into a recursive locking problem. You don't see it when you
switch back to !RT because the spinlocks become noops as this is an UP
machine.

Of course this proves, that nobody ever cared to boot a kernel with
DEBUG_SPINLOCK or even LOCKDEP enabled. That would have exposed such
problems. Hint: Documentation/SubmitChecklist and
Documentation/development-process/4.Coding exist for very good
reasons.

Thanks,

	tglx