[2.6.10-rc2] In-kernel swsusp broken

[2.6.10-rc2] In-kernel swsusp broken

[Joshua Kwan]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

It seems that as of 2.6.10-rc2 my laptop no longer is able to suspend to
disk with an echo disk > /sys/power/state. It goes to freeing memory,
and then blanks the screen. Usually, the screen will turn on again, it
will write data to the swap partition then power off by itself, but
instead, the screen stays off, the CPU starts going wild, and a hard
reset is necessary. It used to work in 2.6.10-rc1.

Is there any help I can provide with this? It looks like I will have to
binary search between rc1 and rc2, but I won't have time for that until
next week...

Thanks

- --
Joshua Kwan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
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=peP5
-----END PGP SIGNATURE-----

Re: [2.6.10-rc2] In-kernel swsusp broken

[Joshua Kwan]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Andrew Morton wrote:
| Joshua, any progress on this one?
|
| Thanks.

Yes, I've run through the BK snapshots and have discovered that
something between 2.6.10-rc1-bk14 and 2.6.10-rc1-bk15 killed swsusp for
me. I'm reviewing the interdiff now, which I've put up for reading
convenience at

http://people.debian.org/~joshk/bk14-bk15.diff.bz2

Stay tuned.. I might find something!

- --
Joshua Kwan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
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=fwzq
-----END PGP SIGNATURE-----

Re: [2.6.10-rc2] In-kernel swsusp broken

[Joshua Kwan]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Andrew Morton wrote:
|>Stay tuned.. I might find something!
|
|
| Cool, thanks.  I'll hold my breath ;)

Got it:

http://linux.bkbits.net:8080/linux-2.6/cset@1.2026.53.35

Reverting the bit of this that affects pci-driver.c fixes bk15 for me.
2.6.10-rc2 with this reverted also works.

Anything else you need? Could it be that a certain driver has a bad
suspend function which is causing the hang? I'll try and come up with
some debugging code to figure out where the hang is happening but I
first have to disable PM for my framebuffer without breaking other things.

Hope this helps.

- --
Joshua Kwan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
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=6DhP
-----END PGP SIGNATURE-----

-- 
Joshua Kwan

Re: [2.6.10-rc2] In-kernel swsusp broken

[Joshua Kwan]

[-- Attachment #1: Type: text/plain, Size: 1264 bytes --]

[CC:ing Jaroslav, solution involving ALSA code is attached.]

Andrew Morton wrote:
> Yes, if you could apply something like the below and see if we can see
> which driver is causing the problem?  Make sure that CONFIG_PCI_NAMES is
> set...

Yeah, did that (didn't use yours), and apparently all the suspend 
functions return 0 but it died after pcibios_set_master on several 
devices - the last call succeeded (it was my intel8x0 sound card) and 
thereafter the hang occurred.

I noticed a recent -mm patch added

pci_disable_device(chip->pci);

in sound/pci/intel8x0.c -> intel8x0_suspend. Looks like that didn't 
help; I also had to add a pci_save_state(chip->pci) above that for 
suspend to work, mimicking the way USB seemed to handle suspending, and 
that worked! Had to do that for intel8x0m.c as well. I looked for other 
instances of disable_device in sound/pci and saw many 
pci_set_power_state(pdev, 3). Would that have a similar effect to saving 
the state? Should I have used that instead? Well, you be the judge...

I'm not really sure what to send a diff against, but I figure -rc3 is 
best so that other people can take advantage of the fix without having 
to patch to -mm first.

Signed-off-by: Joshua Kwan <joshk@triplehelix.org>

-- 
Joshua Kwan

[-- Attachment #2: intel8x0-save-pci-state.patch --]
[-- Type: text/x-patch, Size: 778 bytes --]

--- linux-2.6.9/sound/pci/intel8x0.c~	2004-12-03 23:44:28.000000000 -0800
+++ linux-2.6.9/sound/pci/intel8x0.c	2004-12-03 23:44:36.000000000 -0800
@@ -2279,6 +2279,8 @@
 	for (i = 0; i < 3; i++)
 		if (chip->ac97[i])
 			snd_ac97_suspend(chip->ac97[i]);
+	pci_save_state(chip->pci);
+	pci_disable_device(chip->pci);
 	snd_power_change_state(card, SNDRV_CTL_POWER_D3hot);
 	return 0;
 }
--- linux-2.6.9/sound/pci/intel8x0m.c~	2004-12-03 23:44:42.000000000 -0800
+++ linux-2.6.9/sound/pci/intel8x0m.c	2004-12-03 23:44:57.000000000 -0800
@@ -1091,6 +1091,8 @@
 		snd_pcm_suspend_all(chip->pcm[i]);
 	if (chip->ac97)
 		snd_ac97_suspend(chip->ac97);
+	pci_save_state(chip->pci);
+	pci_disable_device(chip->pci);
 	snd_power_change_state(card, SNDRV_CTL_POWER_D3hot);
 	return 0;
 }