[Bug] am33xx: Basic suspend resume support: Doesn't resume

[Bug] am33xx: Basic suspend resume support: Doesn't resume

[Christoph Fritz]

Hi,

 I'm having here a custom am335x board with DDR3 equipped trying to
setup suspend/resume support. I have tested this with the ancient
TI-PSP-3.2 Kernel and current 3.13-rc plus some pm patches.

Problem: After suspending the device, it doesn't resume (freeze):

echo "mem" > /sys/power/state
[  103.020845] PM: Syncing filesystems ... done.
[  103.055549] Freezing user space processes ... (elapsed 0.003 seconds) done.
[  103.066696] Freezing remaining freezable tasks ... (elapsed 0.002 seconds) done.
[  103.077506] PM: Sending message for entering DeepSleep mode
[  103.099896] PM: suspend of devices complete after 11.951 msecs
[  103.110894] PM: late suspend of devices complete after 4.659 msecs
[  103.122599] PM: noirq suspend of devices complete after 5.032 msecs

As a wakeup source I'm currently using gpio0_12.

Triggering the reset button (am335x-pin: WARMRSTn) in suspend-mode has
no effect, which is fine. But after changing the signal on the wakeup
source (here gpio0_12), triggering a reset works fine again.

So it seems to me that somewhere in the wakeup-process is a bug.

Could this be related to DDR3? Apart from that, this custom board here
is pretty similar to the ti-evm board wich works fine.

In oder to test, I've shrunken the kernel to minimal driver support to
sort out as much as possible segfaults in resume functions.

With a view to be able to debug this further, I just ordered an openocd
jtag adapter.

Any comments, ideas, tips, ..?

Thanks
 -- Christoph

Re: [Bug] am33xx: Basic suspend resume support: Doesn't resume

[Christoph Fritz]

On Thu, 2013-12-05 at 14:42 +0100, Christoph Fritz wrote:
> Hi,
> 
>  I'm having here a custom am335x board with DDR3 equipped trying to
> setup suspend/resume support. I have tested this with the ancient
> TI-PSP-3.2 Kernel and current 3.13-rc plus some pm patches.
> 
> Problem: After suspending the device, it doesn't resume (freeze):
> 
> echo "mem" > /sys/power/state
> [  103.020845] PM: Syncing filesystems ... done.
> [  103.055549] Freezing user space processes ... (elapsed 0.003 seconds) done.
> [  103.066696] Freezing remaining freezable tasks ... (elapsed 0.002 seconds) done.
> [  103.077506] PM: Sending message for entering DeepSleep mode
> [  103.099896] PM: suspend of devices complete after 11.951 msecs
> [  103.110894] PM: late suspend of devices complete after 4.659 msecs
> [  103.122599] PM: noirq suspend of devices complete after 5.032 msecs
> 
> As a wakeup source I'm currently using gpio0_12.
> 
> Triggering the reset button (am335x-pin: WARMRSTn) in suspend-mode has
> no effect, which is fine. But after changing the signal on the wakeup
> source (here gpio0_12), triggering a reset works fine again.
> 
> So it seems to me that somewhere in the wakeup-process is a bug.
> 
> Could this be related to DDR3? Apart from that, this custom board here
> is pretty similar to the ti-evm board wich works fine.
> 
> In oder to test, I've shrunken the kernel to minimal driver support to
> sort out as much as possible segfaults in resume functions.
> 
> With a view to be able to debug this further, I just ordered an openocd
> jtag adapter.
> 
> Any comments, ideas, tips, ..?

Update: It has definitely something to do with DDR3. I'll investigate
this further. By the way, is there a TI-Board with DDR3 and working
suspend-resume?

Thanks
  -- Christoph

Re: [Bug] am33xx: Basic suspend resume support: Doesn't resume

[Brian Murphy]

On 12/06/2013 12:43 AM, Christoph Fritz wrote:
> On Thu, 2013-12-05 at 14:42 +0100, Christoph Fritz wrote:
>> Hi,
>>
>>   I'm having here a custom am335x board with DDR3 equipped trying to
>> setup suspend/resume support. I have tested this with the ancient
>> TI-PSP-3.2 Kernel and current 3.13-rc plus some pm patches.
>>
>> Problem: After suspending the device, it doesn't resume (freeze):
>>
>> echo "mem" > /sys/power/state
>> [  103.020845] PM: Syncing filesystems ... done.
>> [  103.055549] Freezing user space processes ... (elapsed 0.003 seconds) done.
>> [  103.066696] Freezing remaining freezable tasks ... (elapsed 0.002 seconds) done.
>> [  103.077506] PM: Sending message for entering DeepSleep mode
>> [  103.099896] PM: suspend of devices complete after 11.951 msecs
>> [  103.110894] PM: late suspend of devices complete after 4.659 msecs
>> [  103.122599] PM: noirq suspend of devices complete after 5.032 msecs
>>
>> As a wakeup source I'm currently using gpio0_12.
>>
>> Triggering the reset button (am335x-pin: WARMRSTn) in suspend-mode has
>> no effect, which is fine. But after changing the signal on the wakeup
>> source (here gpio0_12), triggering a reset works fine again.
>>
>> So it seems to me that somewhere in the wakeup-process is a bug.
>>
>> Could this be related to DDR3? Apart from that, this custom board here
>> is pretty similar to the ti-evm board wich works fine.
>>
>> In oder to test, I've shrunken the kernel to minimal driver support to
>> sort out as much as possible segfaults in resume functions.
>>
>> With a view to be able to debug this further, I just ordered an openocd
>> jtag adapter.
>>
>> Any comments, ideas, tips, ..?
> Update: It has definitely something to do with DDR3. I'll investigate
> this further. By the way, is there a TI-Board with DDR3 and working
> suspend-resume?
>
I don't think so. I would like to get DDR3 (beaglebone black) and our 
custom board with
mDDR working. Any chance of you sharing what you have of patches 
somewhere so there
will be two working on a common base? (me and you)

/Brian