Skip to contentSkip to conversation

Jinxed VAIO wreckage - current state of affairs

DISCUSSION Thomas Gleixner 6 participants 13 messages Jun 9, 2007 – Jun 16, 2007

lore
Share & tools
lore
#0 lore raw

Andrew's jinxed VAIO breaks with the high resolution timer updates in a very strange way. Andrew identified the following patch as the culprit:

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/broken-out/clockevents-fix-resume-logic.patch

This makes no sense at all. The patch just moves the timer restart to a different (later) place in the code and does exactly the same thing as the current code does.

On resume the VAIO is stuck in the following place:

<Andrews debug session>

We finish swsusp_save() and a few other functions then we go

        hibernate
        ->platform_finish
          ->acpi_hibernation_finish
            ->acpi_leave_sleep_state
              ->acpi_evaluate_object

and there it dies, in this call:

status = acpi_evaluate_object(NULL, METHOD_NAME__WAK, &arg_list, NULL);

I wonder how your patch caused that?

<debugs further>

OK, it gets to the last statement in acpi_evaluate_object():

        return_ACPI_STATUS(status);

but doesn't hit the printk on return to the caller, acpi_leave_sleep_state().

</Andrews debug session>

Some data points:

This happens only, when the local apic timer is used. With PIT the resume works fine.

I back ported the full high res stuff to 2.6.20. On 2.6.20 the VAIO survives that patch.

Can the suspend/resume and ACPI wizards please give some hint how to track this 100% reproducible wreckage down.

Thanks,

tglx

#1 lore raw
Quoted 40 lines · Thomas Gleixner

Andrew's jinxed VAIO breaks with the high resolution timer updates in a very strange way. Andrew identified the following patch as the culprit:

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/broken-out/clockevents-fix-resume-logic.patch

This makes no sense at all. The patch just moves the timer restart to a different (later) place in the code and does exactly the same thing as the current code does.

On resume the VAIO is stuck in the following place:

<Andrews debug session>

We finish swsusp_save() and a few other functions then we go

        hibernate
        ->platform_finish
          ->acpi_hibernation_finish
            ->acpi_leave_sleep_state
              ->acpi_evaluate_object

and there it dies, in this call:

status = acpi_evaluate_object(NULL, METHOD_NAME__WAK, &arg_list, NULL);

I wonder how your patch caused that?

<debugs further>

OK, it gets to the last statement in acpi_evaluate_object():

        return_ACPI_STATUS(status);

but doesn't hit the printk on return to the caller, acpi_leave_sleep_state().

</Andrews debug session>

Some data points:

Quoted 8 lines

This happens only, when the local apic timer is used. With PIT the resume works fine.

I back ported the full high res stuff to 2.6.20. On 2.6.20 the VAIO survives that patch.

Can the suspend/resume and ACPI wizards please give some hint how to track this 100% reproducible wreckage down.

Hmm, how's 2.6.22-rc4-mm2 doing on the Vaio?

There are a couple of patches in there that might help in theory, this series in particular:

swsusp-remove-incorrect-code-from-userc.patch swsusp-remove-code-duplication-between-diskc-and-userc.patch swsusp-introduce-restore-platform-operations.patch swsusp-fix-hibernation-code-ordering.patch swsusp-remove-code-duplication-between-diskc-and-userc-fix.patch

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#2 lore raw
Quoted 5 lines · Rafael J. Wysocki
Quoted 2 lines

Can the suspend/resume and ACPI wizards please give some hint how to track this 100% reproducible wreckage down.

Hmm, how's 2.6.22-rc4-mm2 doing on the Vaio?

Quoted 8 lines

There are a couple of patches in there that might help in theory, this series in particular:

swsusp-remove-incorrect-code-from-userc.patch swsusp-remove-code-duplication-between-diskc-and-userc.patch swsusp-introduce-restore-platform-operations.patch swsusp-fix-hibernation-code-ordering.patch swsusp-remove-code-duplication-between-diskc-and-userc-fix.patch

I'm working on a hrt version against -mm, so we can figure this out.

tglx

#3 lore raw
In reply to Rafael J. Wysocki
Quoted 1 line · On Sat, 9 Jun 2007 22:59:49 +0200 "Rafael J. Wysocki" <[email protected]> wrote:

Hmm, how's 2.6.22-rc4-mm2 doing on the Vaio?

People would have heard if it was busted ;)

Have seen occasional hangs in e100 resume-from-RAM, and occasional all-black-and-dead symptoms after resume-from-RAM, but it seems to work at least 90% of the time.

#4 lore raw
Quoted 9 lines · Andrew Morton
Quoted 1 line · On Sat, 9 Jun 2007 22:59:49 +0200 "Rafael J. Wysocki" <[email protected]> wrote:

Hmm, how's 2.6.22-rc4-mm2 doing on the Vaio?

People would have heard if it was busted ;)

Have seen occasional hangs in e100 resume-from-RAM, and occasional all-black-and-dead symptoms after resume-from-RAM, but it seems to work at least 90% of the time.

You doing than I am on my S580p.

if AHCI is loaded, damn thing will not turn off. Goofy part is, another Sony S80p at work does NOT have this probelm - same bios, same drive firmware.

Suspend to disk mostly works - sometimes when you return, the screen is kinda wonky.

Suspend to ram - going in appears to work, coming out it's dead.

thomas

#5 lore raw
In reply to Andrew Morton
Quoted 5 lines · Andrew Morton
Quoted 1 line · On Sat, 9 Jun 2007 22:59:49 +0200 "Rafael J. Wysocki" <[email protected]> wrote:

Hmm, how's 2.6.22-rc4-mm2 doing on the Vaio?

People would have heard if it was busted ;)

Just found a brown paperbag bug in the resume patch logic. Sigh, I was staring at that code for month without noticing.

Andrew,

can you please test

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/patch-2.6.22-rc4-hrt9.patch

Thanks,

tglx

#7 lore raw
In reply to Thomas Gleixner
Quoted 7 lines · On Fri, 15 Jun 2007 14:05:57 +0200 Thomas Gleixner wrote:
Quoted 5 lines · Andrew Morton
Quoted 1 line · On Sat, 9 Jun 2007 22:59:49 +0200 "Rafael J. Wysocki" <[email protected]> wrote:

Hmm, how's 2.6.22-rc4-mm2 doing on the Vaio?

People would have heard if it was busted ;)

Quoted 8 lines

Just found a brown paperbag bug in the resume patch logic. Sigh, I was staring at that code for month without noticing.

Andrew,

can you please test

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/patch-2.6.22-rc4-hrt9.patch

bad URL, not found.

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

#8 lore raw
Quoted 12 lines · Randy Dunlap
Quoted 10 lines · On Fri, 15 Jun 2007 14:05:57 +0200 Thomas Gleixner wrote:
Quoted 5 lines · Andrew Morton
Quoted 1 line · On Sat, 9 Jun 2007 22:59:49 +0200 "Rafael J. Wysocki" <[email protected]> wrote:

Hmm, how's 2.6.22-rc4-mm2 doing on the Vaio?

People would have heard if it was busted ;)

Just found a brown paperbag bug in the resume patch logic. Sigh, I was staring at that code for month without noticing.

Quoted 7 lines
Quoted 5 lines

Andrew,

can you please test

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/patch-2.6.22-rc4-hrt9.patch

bad URL, not found.

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/patch-2.6.22-rc4-hrt10.patch

tglx

#9 lore raw
Quoted 15 lines · On Fri, 15 Jun 2007 17:31:06 +0200 Thomas Gleixner <[email protected]> wrote:
Quoted 14 lines · Randy Dunlap
Quoted 12 lines · On Fri, 15 Jun 2007 14:05:57 +0200 Thomas Gleixner wrote:
Quoted 5 lines · Andrew Morton
Quoted 1 line · On Sat, 9 Jun 2007 22:59:49 +0200 "Rafael J. Wysocki" <[email protected]> wrote:

Hmm, how's 2.6.22-rc4-mm2 doing on the Vaio?

People would have heard if it was busted ;)

Just found a brown paperbag bug in the resume patch logic. Sigh, I was staring at that code for month without noticing.

Andrew,

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

can you please test

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/patch-2.6.22-rc4-hrt9.patch

bad URL, not found.

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/patch-2.6.22-rc4-hrt10.patch

That suspends and resumes OK.

What was the bug?

#11 lore raw
Quoted 7 lines · Thomas Gleixner
Quoted 6 lines · Andrew Morton
Quoted 3 lines

http://www.tglx.de/projects/hrtimers/2.6.22-rc4/patch-2.6.22-rc4-hrt10.patch

That suspends and resumes OK.

Quoted 7 lines
Quoted 1 line

What was the bug?

A stupid state check, which prevented the PIT to be setup again. So the box got stuck waiting for a timer to expire.

Still I can not explain, why this resulted in this strange "disappear in the return instruction" behavior.

I put up a fixed patch series against rc4-mm to:

http://www.tglx.de/projects/hrtimers/2.6.22-rc4-mm2/patch-2.6.22-rc4-mm2-hrt3.patches.tar.bz2

tglx

#12 lore raw
Quoted 6 lines · On Sat, 16 Jun 2007 08:53:33 +0200 Thomas Gleixner <[email protected]> wrote:
Quoted 2 lines

Still I can not explain, why this resulted in this strange "disappear in the return instruction" behavior.

I put up a fixed patch series against rc4-mm to:

http://www.tglx.de/projects/hrtimers/2.6.22-rc4-mm2/patch-2.6.22-rc4-mm2-hrt3.patches.tar.bz2

I expect that everyone's forgotten what they do. It wouldn't hurt to send them all out in the usual fashion.

btw, I have a huge backlog here (almost two days' worth!) and I'll be only intermittently up for a couple of weeks. There will be some delays, sorry.