CONFIG_CPU_IDLE broken for N900

CONFIG_CPU_IDLE broken for N900

[Ameya Palande]

Hi,

When I enable CONFIG_CPU_IDLE and boot kernel, it hangs around the point
when it switches to user space. I am able to reproduce this for 2.6.35
on linux-omap and linux-2.6 trees.

Has anyone else noticed the same thing for omap3 based boards?

Cheers,
Ameya.

Re: CONFIG_CPU_IDLE broken for N900

[Ameya Palande]

On Tue, 2010-08-10 at 14:21 +0200, Palande Ameya (Nokia-MS/Helsinki)
wrote:
> Hi,
> 
> When I enable CONFIG_CPU_IDLE and boot kernel, it hangs around the point
> when it switches to user space. I am able to reproduce this for 2.6.35
> on linux-omap and linux-2.6 trees.
> 
> Has anyone else noticed the same thing for omap3 based boards?

Broken on linux-omap-pm as well!

Cheers,
Ameya.

Re: CONFIG_CPU_IDLE broken for N900

[Jarkko Nikula]

On Tue, 10 Aug 2010 15:35:18 +0300
Ameya Palande <ameya.palande@nokia.com> wrote:

> > When I enable CONFIG_CPU_IDLE and boot kernel, it hangs around the point
> > when it switches to user space. I am able to reproduce this for 2.6.35
> > on linux-omap and linux-2.6 trees.
> > 
> > Has anyone else noticed the same thing for omap3 based boards?
> 
> Broken on linux-omap-pm as well!
> 
I'm able to reproduce with 2.6.35-rc1 but not with l-o head. It seems
that the commit below fixes this.

commit c786bcf12e092802496f1bc440b2b189144612a5
Author: Kevin Hilman <khilman@deeprootsystems.com>
Date:   Thu Jun 3 17:13:22 2010 +0000

    OMAP3: PM: fix IO daisy chain enable to use PM_WKEN reg


-- 
Jarkko

Re: CONFIG_CPU_IDLE broken for N900

[Ameya Palande]

Hi Jarkko,

On Tue, 2010-08-10 at 15:03 +0200, ext Jarkko Nikula wrote:
> On Tue, 10 Aug 2010 15:35:18 +0300
> Ameya Palande <ameya.palande@nokia.com> wrote:
> 
> > > When I enable CONFIG_CPU_IDLE and boot kernel, it hangs around the point
> > > when it switches to user space. I am able to reproduce this for 2.6.35
> > > on linux-omap and linux-2.6 trees.
> > > 
> > > Has anyone else noticed the same thing for omap3 based boards?
> > 
> > Broken on linux-omap-pm as well!
> > 
> I'm able to reproduce with 2.6.35-rc1 but not with l-o head. It seems
> that the commit below fixes this.
> 
> commit c786bcf12e092802496f1bc440b2b189144612a5
> Author: Kevin Hilman <khilman@deeprootsystems.com>
> Date:   Thu Jun 3 17:13:22 2010 +0000
> 
>     OMAP3: PM: fix IO daisy chain enable to use PM_WKEN reg

Just tested it on linux-omap with omap3_defconfig and found that it is
broken. When kernel bootup is complete, and it switches to user space,
device hangs. When I disable CPU IDLE, it works..

Cheers,
Ameya.

Re: CONFIG_CPU_IDLE broken for N900

[Jarkko Nikula]

On Tue, 10 Aug 2010 16:43:16 +0300
Ameya Palande <ameya.palande@nokia.com> wrote:

> > I'm able to reproduce with 2.6.35-rc1 but not with l-o head. It seems
> > that the commit below fixes this.
> > 
> > commit c786bcf12e092802496f1bc440b2b189144612a5
> > Author: Kevin Hilman <khilman@deeprootsystems.com>
> > Date:   Thu Jun 3 17:13:22 2010 +0000
> > 
> >     OMAP3: PM: fix IO daisy chain enable to use PM_WKEN reg
> 
> Just tested it on linux-omap with omap3_defconfig and found that it is
> broken. When kernel bootup is complete, and it switches to user space,
> device hangs. When I disable CPU IDLE, it works..
> 
And I was too fast, problem is there in l-o head but it seems to hang in
different places for me. Above commit removed "Wake up daisy chain
activation failed." error that was slowing down the boot but this issue
seems to be different.

I can boot up to Debian console login but system hangs during user
space boot if I send something over serial. This happend with
omap3_defconfig in l-o head using N900. However, the same kernel works
fine on Beagle. Can this be some silicon dependent issue?


-- 
Jarkko

Re: CONFIG_CPU_IDLE broken for N900

[Kevin Hilman]

Ameya Palande <ameya.palande@nokia.com> writes:

> When I enable CONFIG_CPU_IDLE and boot kernel, it hangs around the point
> when it switches to user space. I am able to reproduce this for 2.6.35
> on linux-omap and linux-2.6 trees.
>
> Has anyone else noticed the same thing for omap3 based boards?

I was able to reproduce on my n900 and found the problem.  I believe
this is the same problem that has been haunting the Overo folks in that
their console (on UART2 like n900) was not responding after suspend.

The patch below fixes the problem for me, and will post an official
version shortly.

Tested with l-o master (omap3_defconfig + CONFIG_CPU_IDLE=y) as well
with current PM branch.  PM branch (specifically pm-fixes sub-branch)
has been updated to include this fix.

Kevin


diff --git a/arch/arm/mach-omap2/pm34xx.c b/arch/arm/mach-omap2/pm34xx.c
index fb4994a..7b03426 100644
--- a/arch/arm/mach-omap2/pm34xx.c
+++ b/arch/arm/mach-omap2/pm34xx.c
@@ -480,7 +480,9 @@ void omap_sram_idle(void)
 	}
 
 	/* Disable IO-PAD and IO-CHAIN wakeup */
-	if (omap3_has_io_wakeup() && core_next_state < PWRDM_POWER_ON) {
+	if (omap3_has_io_wakeup() &&
+	    (per_next_state < PWRDM_POWER_ON ||
+	     core_next_state < PWRDM_POWER_ON)) {
 		prm_clear_mod_reg_bits(OMAP3430_EN_IO_MASK, WKUP_MOD, PM_WKEN);
 		omap3_disable_io_chain();
 	}
-- 
1.7.2.1