reboot/suspend regression: hx4700 freezes under linux-3.3-rc1

reboot/suspend regression: hx4700 freezes under linux-3.3-rc1

[Paul Parsons]

Both reboot (via reboot(RB_AUTOBOOT)) and suspend (via apm --suspend) freeze my hx4700 under linux-3.3-rc1. They both worked under linux-3.2.

I presume that at least the reboot freeze is related to the restart changes back in November. Can anyone suggest anything obvious I should check? If not I'll attempt to get console output on to the screen, which may or may not help.

reboot/suspend regression: hx4700 freezes under linux-3.3-rc1

[Russell King - ARM Linux]

On Tue, Jan 31, 2012 at 05:28:10PM +0000, Paul Parsons wrote:
> Both reboot (via reboot(RB_AUTOBOOT)) and suspend (via apm --suspend)
> freeze my hx4700 under linux-3.3-rc1. They both worked under linux-3.2.
> 
> I presume that at least the reboot freeze is related to the restart
> changes back in November. Can anyone suggest anything obvious I
> should check? If not I'll attempt to get console output on to the
> screen, which may or may not help.

I'd also suggest trying to bisect them to find out exactly which
commit caused the problem.

You should be able to just do a bisect between v3.2-rc1 and
f88b8979d26615ce68772cebc85c3b556571afca
ARM: restart: remove the now empty arch_reset()

However, you might need 28f9a9294a37ddbc1d821782afc8754eed4bce4e
applied to be able to build.

reboot/suspend regression: hx4700 freezes under linux-3.3-rc1

[Paul Parsons]

--- On Tue, 31/1/12, Russell King - ARM Linux <linux@arm.linux.org.uk> wrote:
> On Tue, Jan 31, 2012 at 05:28:10PM
> +0000, Paul Parsons wrote:
> > Both reboot (via reboot(RB_AUTOBOOT)) and suspend (via
> apm --suspend)
> > freeze my hx4700 under linux-3.3-rc1. They both worked
> under linux-3.2.
> > 
> > I presume that at least the reboot freeze is related to
> the restart
> > changes back in November. Can anyone suggest anything
> obvious I
> > should check? If not I'll attempt to get console output
> on to the
> > screen, which may or may not help.
> 
> I'd also suggest trying to bisect them to find out exactly
> which
> commit caused the problem.

OK, I found the reboot bug:

register_syscore_ops(&pxa_gpio_syscore_ops);

is called in two different places. This corrupts syscore_ops_list and
causes syscore_shutdown() to enter an infinite loop.

The fix is to remove the extra call added since 3.2:

--- clean-3.3-rc1/drivers/gpio/gpio-pxa.c	2012-01-19 23:04:48.000000000 +0000
+++ linux-3.3-rc1/drivers/gpio/gpio-pxa.c	2012-02-01 15:37:39.060989857 +0000
@@ -606,10 +606,3 @@
 	.suspend	= pxa_gpio_suspend,
 	.resume		= pxa_gpio_resume,
 };
-
-static int __init pxa_gpio_sysinit(void)
-{
-	register_syscore_ops(&pxa_gpio_syscore_ops);
-	return 0;
-}
-postcore_initcall(pxa_gpio_sysinit);

The suspend bug is alas unrelated.