ap4 (sh7372) regression in 3.3 / 3.4 with serial console - boot failures

ap4 (sh7372) regression in 3.3 / 3.4 with serial console - boot failures

[Guennadi Liakhovetski]

Hi all

I'm observing a regression with 3.3 and 3.4 kernels, 3.2 seems to be ok. 
The system (confirmed with mackerel and ap4evb) hangs during boot at 
random times. This can be directly after "freeing init memory," or some 
time during the OS start up, or immediately after the start up has 
completed. But if it managed to boot, it continues to run, I've never seen 
it hanging after I ssh onto the system. It also seems, that the 
probability to hang increases with time after the initial system power on. 
A simple short power off doesn't solve the problem.

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/

Re: ap4 (sh7372) regression in 3.3 / 3.4 with serial console - boot failures

[Paul Mundt]

On Wed, May 16, 2012 at 03:36:10PM +0200, Guennadi Liakhovetski wrote:
> I'm observing a regression with 3.3 and 3.4 kernels, 3.2 seems to be ok. 
> The system (confirmed with mackerel and ap4evb) hangs during boot at 
> random times. This can be directly after "freeing init memory," or some 
> time during the OS start up, or immediately after the start up has 
> completed. But if it managed to boot, it continues to run, I've never seen 
> it hanging after I ssh onto the system. It also seems, that the 
> probability to hang increases with time after the initial system power on. 
> A simple short power off doesn't solve the problem.
> 
So you experience random hangs at boot and you have a known good and
known bad version. If only there were some sort of a bisection tool one
could use..

Re: ap4 (sh7372) regression in 3.3 / 3.4 with serial console - boot failures

[Laurent Pinchart]

Hi Paul,

On Thursday 17 May 2012 09:46:04 Paul Mundt wrote:
> On Wed, May 16, 2012 at 03:36:10PM +0200, Guennadi Liakhovetski wrote:
> > I'm observing a regression with 3.3 and 3.4 kernels, 3.2 seems to be ok.
> > The system (confirmed with mackerel and ap4evb) hangs during boot at
> > random times. This can be directly after "freeing init memory," or some
> > time during the OS start up, or immediately after the start up has
> > completed. But if it managed to boot, it continues to run, I've never seen
> > it hanging after I ssh onto the system. It also seems, that the
> > probability to hang increases with time after the initial system power on.
> > A simple short power off doesn't solve the problem.
> 
> So you experience random hangs at boot and you have a known good and
> known bad version. If only there were some sort of a bisection tool one
> could use..

I'm experiencing the same issue on mackerel. I don't know at which kernel 
version the problems started in my case though.

Bisecting this would take quite a long time given the random nature of the 
problem. If there's no other option we'll of course need to go that way, but 
it certainly made sense to raise the matter on the list before investigating 
it in case someone was already working on a fix (or, better, had already 
written a fix :-)).

-- 
Regards,

Laurent Pinchart

Re: ap4 (sh7372) regression in 3.3 / 3.4 with serial console - boot failures

[Rafael J. Wysocki]

On Tuesday, May 22, 2012, Laurent Pinchart wrote:
> Hi Paul,
> 
> On Thursday 17 May 2012 09:46:04 Paul Mundt wrote:
> > On Wed, May 16, 2012 at 03:36:10PM +0200, Guennadi Liakhovetski wrote:
> > > I'm observing a regression with 3.3 and 3.4 kernels, 3.2 seems to be ok.
> > > The system (confirmed with mackerel and ap4evb) hangs during boot at
> > > random times. This can be directly after "freeing init memory," or some
> > > time during the OS start up, or immediately after the start up has
> > > completed. But if it managed to boot, it continues to run, I've never seen
> > > it hanging after I ssh onto the system. It also seems, that the
> > > probability to hang increases with time after the initial system power on.
> > > A simple short power off doesn't solve the problem.
> > 
> > So you experience random hangs at boot and you have a known good and
> > known bad version. If only there were some sort of a bisection tool one
> > could use..
> 
> I'm experiencing the same issue on mackerel. I don't know at which kernel 
> version the problems started in my case though.
> 
> Bisecting this would take quite a long time given the random nature of the 
> problem. If there's no other option we'll of course need to go that way, but 
> it certainly made sense to raise the matter on the list before investigating 
> it in case someone was already working on a fix (or, better, had already 
> written a fix :-)).

Well, why would they work on it if they didn't know that the issue was there?

Whatever.

Please try to add an extra pm_runtime_get() to the serial console initialization
to prevent it from being suspended and see if the problem is still reproducible
then.

Thanks,
Rafael