no self-refresh during S3 - any debug hints?

no self-refresh during S3 - any debug hints?

[Jeff Trull]

Hello ACPI experts,

I made a breakthrough in my s2r debugging last night when I realized
that the sooner I pressed the power button after suspend, the more
life the system exhibited post-resume.  In fact when I attempted to
resume immediately after entering suspend, my system had a successful
resume from memory for the first time ever.  Therefore, I have a new
hypothesis: DRAM self-refresh mode is not being properly entered.

I'm running kernel 2.6.20 on a laptop with an ATI IGP 320M Northbridge
and an Athlon XP.  DSDT has a _PTS but no _GTS.  I'm comfortable
recompiling the kernel.  Any suggestions as to how to proceed,
experiments to run, etc.?

Thanks for any help you can give,
Jeff

Re: no self-refresh during S3 - any debug hints?

[Matthew Garrett]

On Sat, May 31, 2008 at 12:15:50PM -0700, Jeff Trull wrote:

> I'm running kernel 2.6.20 on a laptop with an ATI IGP 320M Northbridge
> and an Athlon XP.  DSDT has a _PTS but no _GTS.  I'm comfortable
> recompiling the kernel.  Any suggestions as to how to proceed,
> experiments to run, etc.?

I've seen something similar with an HP nc4010, which also had an ATI IGP 
northbridge. Sadly, I wasn't able to work out why this was happening. 
It wasn't possible to get at chipset docs at the time. Now that AMD are 
being more open about their graphics docs, I'll see if I can manage any 
more.

-- 
Matthew Garrett | mjg59@srcf.ucam.org

Re: no self-refresh during S3 - any debug hints?

[Matthew Garrett]

On Sat, May 31, 2008 at 12:15:50PM -0700, Jeff Trull wrote:

> I'm running kernel 2.6.20 on a laptop with an ATI IGP 320M Northbridge
> and an Athlon XP.  DSDT has a _PTS but no _GTS.  I'm comfortable
> recompiling the kernel.  Any suggestions as to how to proceed,
> experiments to run, etc.?

Hi Jeff,

Could you attach the output of lspci -vxxx -s 0:0.0 ? Needs to be run as 
root.

-- 
Matthew Garrett | mjg59@srcf.ucam.org

Re: no self-refresh during S3 - any debug hints?

[Jeff Trull]

[-- Attachment #1: Type: text/plain, Size: 541 bytes --]

Delighted to!

On Thu, Jun 19, 2008 at 8:09 AM, Matthew Garrett <mjg59@srcf.ucam.org> wrote:
> On Sat, May 31, 2008 at 12:15:50PM -0700, Jeff Trull wrote:
>
>> I'm running kernel 2.6.20 on a laptop with an ATI IGP 320M Northbridge
>> and an Athlon XP.  DSDT has a _PTS but no _GTS.  I'm comfortable
>> recompiling the kernel.  Any suggestions as to how to proceed,
>> experiments to run, etc.?
>
> Hi Jeff,
>
> Could you attach the output of lspci -vxxx -s 0:0.0 ? Needs to be run as
> root.
>
> --
> Matthew Garrett | mjg59@srcf.ucam.org
>

[-- Attachment #2: lspci_vxx_000 --]
[-- Type: application/octet-stream, Size: 1141 bytes --]

00:00.0 Host bridge: ATI Technologies Inc AGP Bridge [IGP 320M] (rev 13)
	Flags: bus master, 66MHz, medium devsel, latency 64
	Memory at e0000000 (32-bit, prefetchable) [size=64M]
	Memory at dc400000 (32-bit, prefetchable) [size=4K]
	I/O ports at 1c10 [disabled] [size=4]
	Capabilities: [a0] AGP version 2.0
00: 02 10 b0 ca 06 00 30 02 13 00 00 06 00 40 00 00
10: 08 00 00 e0 08 00 40 dc 11 1c 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 a0 00 00 00 00 00 00 00 00 00 00 00
40: 00 00 00 00 0b 05 00 00 00 00 00 00 04 0d 00 00
50: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
60: 4d 0d a2 85 2d 0a 72 2c 00 00 00 00 d4 19 06 d5
70: e0 0d 20 31 84 84 84 84 00 00 10 00 00 00 00 00
80: 00 00 00 00 96 10 83 00 32 00 31 00 23 00 00 00
90: 00 00 00 00 00 00 05 00 00 00 e4 7f 84 00 00 00
a0: 02 00 20 00 17 02 00 0f 04 03 00 00 03 00 00 00
b0: 00 00 02 00 08 00 00 00 04 04 04 04 00 00 00 00
c0: 00 00 00 ff 5a 15 05 0f 00 1f ff 1f 00 00 84 00
d0: eb 27 0d 00 eb 27 4d 00 6a 00 60 10 00 00 00 ff
e0: 00 00 00 00 00 00 00 00 00 00 00 ff 00 00 00 ff
f0: 0b 00 09 00 00 00 00 ff 00 08 10 18 00 00 00 00

Re: no self-refresh during S3 - any debug hints?

[Len Brown]

On Sat, 31 May 2008, Jeff Trull wrote:

> I made a breakthrough in my s2r debugging last night when I realized
> that the sooner I pressed the power button after suspend, the more
> life the system exhibited post-resume.  In fact when I attempted to
> resume immediately after entering suspend, my system had a successful
> resume from memory for the first time ever.  Therefore, I have a new
> hypothesis: DRAM self-refresh mode is not being properly entered.
> 
> I'm running kernel 2.6.20 on a laptop with an ATI IGP 320M Northbridge
> and an Athlon XP.  DSDT has a _PTS but no _GTS.  I'm comfortable
> recompiling the kernel.  Any suggestions as to how to proceed,
> experiments to run, etc.?

Start by running the upstrteam kernel -- now 2.6.26, just out of 
principle.

Then, see Documentation/power/basic-pm-debugging.txt

and try the options in /sys/power/pm_test

If the cause is as you suspect, they should all work.
If any of them fail, then there is a problem elsewhere.

eg. the deepest test is core:

# echo core > /sys/power/pm_test
# echo mem > /sys/power/state

should work, because it does everything except actually leap into
the BIOS to enter self-refresh.

# echo none > /sys/power/pm_test

then disalbes this test mode.

-Len

Re: no self-refresh during S3 - any debug hints?

[Jeff Trull]

On Thu, Jun 19, 2008 at 12:02 PM, Len Brown <lenb@kernel.org> wrote:

>> I'm running kernel 2.6.20 on a laptop with an ATI IGP 320M Northbridge
>> and an Athlon XP.  DSDT has a _PTS but no _GTS.  I'm comfortable
>> recompiling the kernel.  Any suggestions as to how to proceed,
>> experiments to run, etc.?
>
> Start by running the upstrteam kernel -- now 2.6.26, just out of
> principle.
>
> Then, see Documentation/power/basic-pm-debugging.txt
>
> and try the options in /sys/power/pm_test

Thanks for the tips, Len.

I previously recompiled my kernel to implement a variant of Rafael's
patch from last November and was able to use it to determine that only
the actual S3 state enter/exit (i.e., the writes to the PM1*_CONTROL
in acpi_enter_sleep_state()) mattered - or, as you said, all the tests
worked.  I'd like to try 2.6.26 soon, but FWIW I did attempt 2.6.24
with the same results.

-Jeff

Re: no self-refresh during S3 - any debug hints?

[Pavel Machek]

On Thu 2008-06-19 15:02:12, Len Brown wrote:
> 
> 
> On Sat, 31 May 2008, Jeff Trull wrote:
> 
> > I made a breakthrough in my s2r debugging last night when I realized
> > that the sooner I pressed the power button after suspend, the more
> > life the system exhibited post-resume.  In fact when I attempted to
> > resume immediately after entering suspend, my system had a successful
> > resume from memory for the first time ever.  Therefore, I have a new
> > hypothesis: DRAM self-refresh mode is not being properly entered.

Hmm, or some other hardware... I ve seen machines where video ram was
not refreshed. If that piece of hw is critical enough, it will break
resume.

You could try with minimal system, with as little drivers as
possible... use acpi_sleep=s3_beep to determine where it fails.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: no self-refresh during S3 - any debug hints?

[Jeff Trull]

On Thu, Jun 19, 2008 at 1:10 PM, Pavel Machek <pavel@suse.cz> wrote:

>
> Hmm, or some other hardware... I ve seen machines where video ram was
> not refreshed. If that piece of hw is critical enough, it will break
> resume.
>
> You could try with minimal system, with as little drivers as
> possible... use acpi_sleep=s3_beep to determine where it fails.
>                                                                        Pavel

I don't have ready access to 2.6.26, but with 2.6.24 s3_beep produces
a (endless, loud) tone if I resume immediately after suspend, and
nothing at all if I wait 10 seconds first.

Jeff

Re: no self-refresh during S3 - any debug hints?

[Pavel Machek]

On Fri 2008-06-20 23:13:06, Jeff Trull wrote:
> On Thu, Jun 19, 2008 at 1:10 PM, Pavel Machek <pavel@suse.cz> wrote:
> 
> >
> > Hmm, or some other hardware... I ve seen machines where video ram was
> > not refreshed. If that piece of hw is critical enough, it will break
> > resume.
> >
> > You could try with minimal system, with as little drivers as
> > possible... use acpi_sleep=s3_beep to determine where it fails.
> 
> I don't have ready access to 2.6.26, but with 2.6.24 s3_beep produces
> a (endless, loud) tone if I resume immediately after suspend, and
> nothing at all if I wait 10 seconds first.

Hmm, okay, that confirms your theory :-(. Contacting BIOS vendor and
looking in DSDT for something strange turning off autorefresh is
something you could try...
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: no self-refresh during S3 - any debug hints?

[Jeff Trull]

Hi Matthew,

Did this data provide any insights?

If it seems like a hopeless case that's all right, I guess - I've moved on to 
a laptop with fully-supported hardware (courtesy Zareason) but if there's any 
chance of finding root cause I'll keep the old one for testing potential 
patches, etc.

Thanks and Regards,
Jeff Trull

On Thursday 19 June 2008 09:40:30 Jeff Trull wrote:
> Delighted to!
>
> On Thu, Jun 19, 2008 at 8:09 AM, Matthew Garrett <mjg59@srcf.ucam.org> 
wrote:
> > On Sat, May 31, 2008 at 12:15:50PM -0700, Jeff Trull wrote:
> >> I'm running kernel 2.6.20 on a laptop with an ATI IGP 320M Northbridge
> >> and an Athlon XP.  DSDT has a _PTS but no _GTS.  I'm comfortable
> >> recompiling the kernel.  Any suggestions as to how to proceed,
> >> experiments to run, etc.?
> >
> > Hi Jeff,
> >
> > Could you attach the output of lspci -vxxx -s 0:0.0 ? Needs to be run as
> > root.
> >
> > --
> > Matthew Garrett | mjg59@srcf.ucam.org