Re: Boot regression on Mackerel with current mainline

[Chris Ball <cjb@laptop.org>]

Hi Guennadi,

On Wed, Jul 04 2012, Guennadi Liakhovetski wrote:
> On Tue, 3 Jul 2012, Rafael J. Wysocki wrote:
>
>> On Tuesday, July 03, 2012, Guennadi Liakhovetski wrote:
>> > Hi Rafael
>> > 
>> > On Tue, 3 Jul 2012, Rafael J. Wysocki wrote:
>> > 
>> > > Hi,
>> > > 
>> > > We seem to have a functional regression on Mackerel with 3.5-rc5 which
>> > > seems to be related to SDHI.  Namely, the SDHI driver appears to wait forever
>> > > for the devices to appear and commenting out all of the sdhi devices in the
>> > > board file works around the problem.
>> > > 
>> > > I can investigate some more later today if need be.
>> > 
>> > Have you checked the logs? Are you seeing any of these in the kernel log:
>> > 
>> > genirq: Threaded irq requested with handler=NULL and !ONESHOT for irq 
>> 
>> I see this only once per boot, but then sh_mobile_sdhi is saying "timeout
>> waiting for hardware interrupt" pretty much continuously.
>
> Yes, that's the one. Ok, let's fix it in the driver. In fact, a fix has 
> already been submitted, but for 3.6. I hoped, 3.5 would resort to a softer 
> form of the ONESHOT failing, so, I didn't push it for -rc. I think, the 
> easiest would be to apply these two patches:
>
> "mmc: tmio: use MMC opcode defines instead of numbers"
> http://git.kernel.org/?p=linux/kernel/git/cjb/mmc.git;a=commit;hY8cd9cc217b0aa46089a3838a8b1c37b97aa08b
>
> and
>
> "mmc: extend and rename cd-gpio helpers to handle more slot GPIO functions"
> http://git.kernel.org/?p=linux/kernel/git/cjb/mmc.git;a=commit;hØe60fb51effc83e6f53de191a8440f46e5a3f2c
>
> The latter one looks big because of a rename, but with "-M" it's pretty 
> trivial. Chris, do you agree?

Is the minimal fix to just add "| IRQF_ONESHOT" to the request_threaded_irq
argument?  If so, I'd rather push just that for 3.5, and I can take care of
having that change appear simultaneously in linux-next so that there aren't
any conflicts.  Is that okay with you?

Thanks,

- Chris.
-- 
Chris Ball   <cjb@laptop.org>   <http://printf.net/>
One Laptop Per Child