Bug 11522 - S2ram is broken on laptops
Summary: S2ram is broken on laptops
Alias: None
Product: Power Management
Classification: Unclassified
Component: Hibernation/Suspend (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: power-management_other
Depends on:
Blocks: 7216 10492
  Show dependency tree
Reported: 2008-09-08 13:03 UTC by Clark Tompsett
Modified: 2009-06-23 08:04 UTC (History)
2 users (show)

See Also:
Kernel Version: 2.6.27-rc5
Regression: Yes
Bisected commit-id:

dmesg output from successful s2ram (54.74 KB, text/plain)
2008-09-08 13:11 UTC, Clark Tompsett
pm-suspend log (8.20 KB, text/plain)
2008-09-08 13:13 UTC, Clark Tompsett
correct log from 6772 showing good s2ram with 2.6.25 kernel (8.21 KB, text/plain)
2008-09-08 13:29 UTC, Clark Tompsett
dmesg log 6772 good suspend (44.59 KB, text/plain)
2008-09-08 13:32 UTC, Clark Tompsett
pm-suspend log 6772 2.6.27-rc5 bad (6.15 KB, text/plain)
2008-09-08 13:33 UTC, Clark Tompsett
lspci output (1.93 KB, text/plain)
2008-09-11 19:54 UTC, Clark Tompsett
git show output (1.55 KB, text/plain)
2008-09-23 17:36 UTC, Clark Tompsett
output from bisect (593 bytes, text/plain)
2008-09-28 16:55 UTC, Clark Tompsett
module list (4.45 KB, text/plain)
2008-09-28 16:56 UTC, Clark Tompsett

Description Clark Tompsett 2008-09-08 13:03:41 UTC
Latest working kernel version: 2.6.25
Earliest failing kernel version: 2.6.26
Hardware Environment: GW ml3109, hp dv6772 laptops
Software Environment: x86_64
Problem Description: Since the 2.6.26 kernels, both laptops will not suspend to disk or to ram.  Testing done with s2ram.  Related to bug 11368.  

Ml3109 (Intel celeron m 520) will suspend to ram but will not restore.  Using acpi_sleep=s3_beep results in no beep tones.  System cycles the cdrom twice and leave the hd light on with no activity.

dv6772 beeps heard but no system restore.

In both cases, there are no entries in the log or pm-suspend logs.

Steps to reproduce:

load a distribution with a 2.6.25 kernel on one or both laptops.  Attempt to s2ram and restore.  Laptop will restore from ram.

Load up the 2.6.27-rc5 kernel.  Try same test (without and with patch from 11368).  System will not restore from ram.
Comment 1 Clark Tompsett 2008-09-08 13:11:59 UTC
Created attachment 17682 [details]
dmesg output from successful s2ram

dmesg output from dv6772 showing successful recovery from s2ram on 2.6.25 kernel with patch from 11368.
Comment 2 Clark Tompsett 2008-09-08 13:13:03 UTC
Created attachment 17683 [details]
pm-suspend log

pm-suspend log from dv6772 that goes with the dmesg log.
Comment 3 Clark Tompsett 2008-09-08 13:14:36 UTC
logs attached are from the wrong machine.  Please delete them.
Comment 4 Clark Tompsett 2008-09-08 13:29:30 UTC
Created attachment 17684 [details]
correct log from 6772 showing good s2ram with 2.6.25 kernel

correct log from 6772 (amd tl-64) showing good s2ram with the 2.6.25 kernel
Comment 5 Clark Tompsett 2008-09-08 13:32:03 UTC
Created attachment 17685 [details]
dmesg log 6772 good suspend

dmesg log 6772 2.6.25 good suspend.
Comment 6 Clark Tompsett 2008-09-08 13:33:15 UTC
Created attachment 17686 [details]
pm-suspend log 6772 2.6.27-rc5 bad

pm suspend log showing system did not restore from s2ram 6772 2.6.27-rc5 kernel.  No binary blobs used in this test system.
Comment 7 Rafael J. Wysocki 2008-09-08 14:37:36 UTC
The fastest way to determine what broke suspend/resume on your boxes would be to carry out bisection of kernels between 2.6.25 and 2.6.26 using git-bisect.  Can you please do that?
Comment 8 Zhang Rui 2008-09-10 19:07:32 UTC
>+ /usr/sbin/s2ram -fa3 --vbe_post
does the problem still exist without using these parameters?
please attach the lspci output.
Comment 9 Clark Tompsett 2008-09-11 19:54:22 UTC
Created attachment 17739 [details]
lspci output

lspci output that you asked for.
Comment 10 Clark Tompsett 2008-09-11 19:56:19 UTC
bisect results -- 2.6.26-rc8 to rc9:

this broke s2ram 
Comment 11 Zhang Rui 2008-09-11 20:41:33 UTC
# git show 821b03ffac8851d6bc1d5530183d2eb25adae35d
fatal: bad object 821b03ffac8851d6bc1d5530183d2eb25adae35d

weird. :(
Comment 12 Rafael J. Wysocki 2008-09-12 08:33:39 UTC
(In reply to comment #10)
> bisect results -- 2.6.26-rc8 to rc9:
> 821b03ffac8851d6bc1d5530183d2eb25adae35d 
> this broke s2ram 

There's no such commit in the Linus' tree.  Are you sure you're testing an unmodified kernel.org kernel?
Comment 13 Clark Tompsett 2008-09-12 12:29:00 UTC
command used to get the git tree:

$ git clone \
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git linux-2.6
Comment 14 Len Brown 2008-09-22 18:45:39 UTC
i don't see this commit either.  Clark, what do you see from:

$ git show 821b03ffac8851d6bc1d5530183d2eb25adae35d 
Comment 15 Clark Tompsett 2008-09-23 17:36:34 UTC
Created attachment 17980 [details]
git show output

output from get show as requested.
Comment 16 Zhang Rui 2008-09-26 19:38:59 UTC
you are probably running the linux-next tree.
please run "cat .git/config" to make sure.

If you're not using linus's git tree, please run the command in comment #13 again to get the latest mainline kernel and test again. :)
Comment 17 Clark Tompsett 2008-09-27 05:01:57 UTC
	repositoryformatversion = 0
	filemode = true
	bare = false
	logallrefupdates = true
[remote "origin"]
	url = git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
	fetch = +refs/heads/*:refs/remotes/origin/*
[branch "master"]
	remote = origin
	merge = refs/heads/master
Comment 18 Rafael J. Wysocki 2008-09-27 06:05:14 UTC
This commit is in the mainline and it is a merge with the networking tree.

Clark, please verify that commit 3d25802e3ba7c82457b5c12bbfeefe391d8a333e works for you.

If it does, please try bisection between commits 2fe195cfe3e53c144d247b2768e37732e8eae4d8 and 88a6f4ad76be425f47df7f892baf913bcd466fb3 .
Comment 19 Clark Tompsett 2008-09-27 21:43:38 UTC
I need information on how to do test specific commits.
Comment 20 Rafael J. Wysocki 2008-09-28 03:31:42 UTC
$ git bisect start
$ git bisect good 88a6f4ad76be425f47df7f892baf913bcd466fb3
$ git bisect bad 2fe195cfe3e53c144d247b2768e37732e8eae4d8

but first please verify that 88a6f4ad76be425f47df7f892baf913bcd466fb3 works for you and 2fe195cfe3e53c144d247b2768e37732e8eae4d8 fails.

If 2fe195cfe3e53c144d247b2768e37732e8eae4d8 works, then the merge itself has cause the problem to appear, so if you find that this is the case, please let me know.
Comment 21 Clark Tompsett 2008-09-28 16:55:32 UTC
Created attachment 18101 [details]
output from bisect

output from the bisect and this is what failed.  But, that driver is not used by my system(s).
Comment 22 Clark Tompsett 2008-09-28 16:56:29 UTC
Created attachment 18102 [details]
module list

This is a list of the modules used on the hp system I'm using for testing.
Comment 23 Zhang Rui 2008-12-08 19:46:58 UTC
does the problem still exist in the latest kernel release?
Comment 24 Clark Tompsett 2008-12-12 20:04:48 UTC
I'll have to get the current 2.6.28 kernel to test.  Currently with OpenSuSE 11.1 RC1, the gateway ml3109 (intel celeron 520) will suspend to ram but the hp dv6772 (amd tl84) does not.
Comment 25 Zhang Rui 2009-03-18 19:34:35 UTC
clark, any update? does 2.6.28 work for you on hp dv6772?
if not, please
1. set CONFIG_PM_DEBUG and rebuild your kernel
2. echo {freezer, devices, platform, processors, core} > /sys/power/pm_test
3. echo mem > /sys/power/state
4. does the system come back in a few seconds?
Comment 26 Clark Tompsett 2009-03-30 16:01:59 UTC
Tried the 2.6.29 kernel and it still will not return from s2ram mode.
Comment 27 Zhang Rui 2009-03-31 01:31:32 UTC
so you mean the test in comment #25 still failed?
If not, please do the test in comment #25.
note that you need to try "freezer/devices/platform/processor/core" one by one and see if all of them work, or which one fails.
Comment 28 Zhang Rui 2009-06-23 08:04:16 UTC
close this bug as there is no response from the bug reporter.
please reopen the bug if the problem still exists in the latest upstream kernel and you can provide the info requested in comment #27.

Note You need to log in before you can comment on or make changes to this bug.