Linux 3.0-rc4 intermittent failure to resume

Linux 3.0-rc4 intermittent failure to resume

[Brad Campbell]

G'day all,

I've recently moved from tuxonice to mainline uswsusp. I'm getting intermittent failure to resume 
problems. No message in dmesg that I can find, it just boots clean. Is there something I can do to 
get some debug info on this? It's really, really annoying.

About 75% of the time it works perfectly, and I can find no rhyme or reason to why it suddenly just 
decides to ignore the saved state.

32 bit userspace with a statically linked 64 bit s2disk, s2both & resume. 64 bit kernel.

Brad
-- 
Dolphins are so intelligent that within a few weeks they can
train Americans to stand at the edge of the pool and throw them
fish.

Re: Linux 3.0-rc4 intermittent failure to resume

[Rafael J. Wysocki]

On Thursday, June 23, 2011, Brad Campbell wrote:
> G'day all,
> 
> I've recently moved from tuxonice to mainline uswsusp. I'm getting intermittent failure to resume 
> problems. No message in dmesg that I can find, it just boots clean. Is there something I can do to 
> get some debug info on this? It's really, really annoying.
> 
> About 75% of the time it works perfectly, and I can find no rhyme or reason to why it suddenly just 
> decides to ignore the saved state.
> 
> 32 bit userspace with a statically linked 64 bit s2disk, s2both & resume. 64 bit kernel.

Do you use the i915 driver?

Rafael

Re: Linux 3.0-rc4 intermittent failure to resume

[Brad Campbell]

On 23/06/11 20:02, Rafael J. Wysocki wrote:
> On Thursday, June 23, 2011, Brad Campbell wrote:
>> G'day all,
>>
>> I've recently moved from tuxonice to mainline uswsusp. I'm getting intermittent failure to resume
>> problems. No message in dmesg that I can find, it just boots clean. Is there something I can do to
>> get some debug info on this? It's really, really annoying.
>>
>> About 75% of the time it works perfectly, and I can find no rhyme or reason to why it suddenly just
>> decides to ignore the saved state.
>>
>> 32 bit userspace with a statically linked 64 bit s2disk, s2both&  resume. 64 bit kernel.
>
> Do you use the i915 driver?
>

Indeed I do.

Re: Linux 3.0-rc4 intermittent failure to resume

[Brad Campbell]

On 24/06/11 10:41, Brad Campbell wrote:
> On 23/06/11 20:02, Rafael J. Wysocki wrote:
>> On Thursday, June 23, 2011, Brad Campbell wrote:
>>> G'day all,
>>>
>>> I've recently moved from tuxonice to mainline uswsusp. I'm getting intermittent failure to resume
>>> problems. No message in dmesg that I can find, it just boots clean. Is there something I can do to
>>> get some debug info on this? It's really, really annoying.
>>>
>>> About 75% of the time it works perfectly, and I can find no rhyme or reason to why it suddenly just
>>> decides to ignore the saved state.
>>>
>>> 32 bit userspace with a statically linked 64 bit s2disk, s2both&   resume. 64 bit kernel.
>> Do you use the i915 driver?
>>
> Indeed I do.
Just a follow up to this.
It looks like adding "early writeout = y" to my /etc/suspend.conf has worked around the problem.

If I had "shutdown method = reboot", then it would resume every time, but with "shutdown method = 
platform" it would fail to resume most of the time.
I'm running a Core 2 Duo with an Intel X18-M 160GB SSD on an intel AHCI controller.

Adding "early writeout = y" does not appear to change the progress statistics, or apparent speed of 
the suspend, but it has now survived 7 suspend/power off/resume cycles that it would never had 
survived before.

Odd ??

Re: Linux 3.0-rc4 intermittent failure to resume

[Brad Campbell]

On 28/06/11 10:39, Brad Campbell wrote:
> On 24/06/11 10:41, Brad Campbell wrote:
>> On 23/06/11 20:02, Rafael J. Wysocki wrote:
>>> On Thursday, June 23, 2011, Brad Campbell wrote:
>>>> G'day all,
>>>>
>>>> I've recently moved from tuxonice to mainline uswsusp. I'm getting intermittent failure to resume
>>>> problems. No message in dmesg that I can find, it just boots clean. Is there something I can do to
>>>> get some debug info on this? It's really, really annoying.
>>>>
>>>> About 75% of the time it works perfectly, and I can find no rhyme or reason to why it suddenly just
>>>> decides to ignore the saved state.
>>>>
>>>> 32 bit userspace with a statically linked 64 bit s2disk, s2both&    resume. 64 bit kernel.
>>> Do you use the i915 driver?
>>>
>> Indeed I do.
> Just a follow up to this.
> It looks like adding "early writeout = y" to my /etc/suspend.conf has worked around the problem.
>

Ok, so after _lots_ of investigating it was not early writeout. The 
culprit was/is the Ubuntu initramfs scripts.

On the occasions it failed to resume (which were many) it turns out 
/bin/resume was being called prior to udev having created /dev/snapshot.

I put a simple wait loop in the resume script to wait for the creation 
of /dev/snapshot and I've not had a fail to resume since.

Curse you asynchronous dynamic device creation <shakes fist>

Regards,
Brad

Re: Linux 3.0-rc4 intermittent failure to resume

[Brad Campbell]

On 28/06/11 10:39, Brad Campbell wrote:
> On 24/06/11 10:41, Brad Campbell wrote:
>> On 23/06/11 20:02, Rafael J. Wysocki wrote:
>>> On Thursday, June 23, 2011, Brad Campbell wrote:
>>>> G'day all,
>>>>
>>>> I've recently moved from tuxonice to mainline uswsusp. I'm getting intermittent failure to resume
>>>> problems. No message in dmesg that I can find, it just boots clean. Is there something I can do to
>>>> get some debug info on this? It's really, really annoying.
>>>>
>>>> About 75% of the time it works perfectly, and I can find no rhyme or reason to why it suddenly just
>>>> decides to ignore the saved state.
>>>>
>>>> 32 bit userspace with a statically linked 64 bit s2disk, s2both&    resume. 64 bit kernel.
>>> Do you use the i915 driver?
>>>
>> Indeed I do.
> Just a follow up to this.
> It looks like adding "early writeout = y" to my /etc/suspend.conf has worked around the problem.
>

Ok, so after _lots_ of investigating it was not early writeout. The 
culprit was/is the Ubuntu initramfs scripts.

On the occasions it failed to resume (which were many) it turns out 
/bin/resume was being called prior to udev having created /dev/snapshot.

I put a simple wait loop in the resume script to wait for the creation 
of /dev/snapshot and I've not had a fail to resume since.

Curse you asynchronous dynamic device creation <shakes fist>

Regards,
Brad

Re: Linux 3.0-rc4 intermittent failure to resume

[Rafael J. Wysocki]

On Wednesday, June 29, 2011, Brad Campbell wrote:
> On 28/06/11 10:39, Brad Campbell wrote:
> > On 24/06/11 10:41, Brad Campbell wrote:
> >> On 23/06/11 20:02, Rafael J. Wysocki wrote:
> >>> On Thursday, June 23, 2011, Brad Campbell wrote:
> >>>> G'day all,
> >>>>
> >>>> I've recently moved from tuxonice to mainline uswsusp. I'm getting intermittent failure to resume
> >>>> problems. No message in dmesg that I can find, it just boots clean. Is there something I can do to
> >>>> get some debug info on this? It's really, really annoying.
> >>>>
> >>>> About 75% of the time it works perfectly, and I can find no rhyme or reason to why it suddenly just
> >>>> decides to ignore the saved state.
> >>>>
> >>>> 32 bit userspace with a statically linked 64 bit s2disk, s2both&    resume. 64 bit kernel.
> >>> Do you use the i915 driver?
> >>>
> >> Indeed I do.
> > Just a follow up to this.
> > It looks like adding "early writeout = y" to my /etc/suspend.conf has worked around the problem.
> >
> 
> Ok, so after _lots_ of investigating it was not early writeout. The 
> culprit was/is the Ubuntu initramfs scripts.
> 
> On the occasions it failed to resume (which were many) it turns out 
> /bin/resume was being called prior to udev having created /dev/snapshot.
> 
> I put a simple wait loop in the resume script to wait for the creation 
> of /dev/snapshot and I've not had a fail to resume since.
> 
> Curse you asynchronous dynamic device creation <shakes fist>

Thank you very  much for nailing this down, I was afraid there were some
obscure bug lurking in the kernel code.

Take care,
Rafael

Re: Linux 3.0-rc4 intermittent failure to resume

[Brad Campbell]

On 30/06/11 03:34, Rafael J. Wysocki wrote:
> On Wednesday, June 29, 2011, Brad Campbell wrote:

>> On the occasions it failed to resume (which were many) it turns out
>> /bin/resume was being called prior to udev having created /dev/snapshot.
>>
>> I put a simple wait loop in the resume script to wait for the creation
>> of /dev/snapshot and I've not had a fail to resume since.
>>
>> Curse you asynchronous dynamic device creation<shakes fist>
>
> Thank you very  much for nailing this down, I was afraid there were some
> obscure bug lurking in the kernel code.

Yeah, no worries.

I set up a loop to do automated suspend/resume cycles.
I managed over 100 cycles while compiling 2 kernels in the background.

Over the 100 cycles the kernel locked up twice, but both times were in 
very early boot. I power cycled the machine and it resumed with no 
problems, so the issue is not with the suspend/resume code.