FS#12090 - kernels after 2.6.27-2-i686 fail to resume from s2ram

Attached to Project: Arch Linux
Opened by David (dsr) - Tuesday, 11 November 2008, 17:29 GMT
Last edited by Tobias Powalowski (tpowa) - Sunday, 01 March 2009, 22:04 GMT
Task Type Bug Report
Category Kernel
Status Closed
Assigned To Tobias Powalowski (tpowa)
Thomas Bächler (brain0)
Architecture i686
Severity Medium
Priority Normal
Reported Version None
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 5
Private No

Details

kernel26 2.6.27-2 works fine. However, 2.6.27.1-1, 2.6.27.4-1, and 2.6.27.5-1 fail to resume from suspend to RAM, both in the console and in X. The bug is not unique to pm-suspend; the kernels after 2.6.27-2 also fail to resume from `echo "mem" > /sys/power/state'. The unsolved thread at <http://bbs.archlinux.org/viewtopic.php?id=58219> might be describing a similar bug, but the problem there seems to deal with suspending rather than resuming. I'm using a first generation 13" MacBook, and I'm using coreutils 6.12-1, module-init-tools 3.4-1, and mkinitcpio 0.5.19-1.

Attached are the /proc/config.gz files for each kernel and the relevant portions from /var/log/message.log. The file messages-2.6.27-2.log contains the lines that were written to /var/log/messages.log when I successfully suspended and resumed with kernel26 2.6.27-2. The file messages-2.6.27.5-1.log contains the lines from /var/log/messages.log that deal with kernel26 2.6.27.5-1 (during which I suspended and failed to resume).
(application/octet-stream)    config-2.6.27-2.gz (21.9 KiB)
(application/octet-stream)    config-2.6.27.1-1.gz (21.9 KiB)
(application/octet-stream)    config-2.6.27.4-1.gz (21.9 KiB)
(application/octet-stream)    config-2.6.27.5-1.gz (21.9 KiB)
   messages-2.6.27-2.log (3 KiB)
   messages-2.6.27.5-1.log (34.6 KiB)
This task depends upon

Closed by  Tobias Powalowski (tpowa)
Sunday, 01 March 2009, 22:04 GMT
Reason for closing:  Fixed
Comment by Dale Blount (dale) - Thursday, 27 November 2008, 04:19 GMT
I've noticed this too on my amd64 still running arch i686.

[2008-11-24 22:36] upgraded kernel26 (2.6.27.1-1 -> 2.6.27.6-1)

Suspends just fine, won't resume. David, are you using nvidia binary drivers as well?

Comment by David (dsr) - Thursday, 27 November 2008, 05:28 GMT
Hey Dale,

No, I'm using xf86-video-intel for my Intel GMA 950. It certainly sounds like we're experiencing the same problem, so we can rule out graphics drivers as the culprit. Are you also using a MacBook (a later generation than mine with an Intel Core 2 Duo processor and an nvidia video card)?

I just tried kernel 2.6.27.6-1 like you did, and the problem persists as expected.
Comment by Thomas Bächler (brain0) - Thursday, 27 November 2008, 08:36 GMT
I had a weird problem as well (which I couldn't tie to any version of anything in particular), but it went away when I disabled the framebuffer (uvesafb or vesafb). Since then, resuming works again.
Comment by Dale Blount (dale) - Thursday, 27 November 2008, 14:15 GMT
Neither uvesafb or vesafb modules are loaded on my system.
Comment by Dale Blount (dale) - Thursday, 27 November 2008, 14:19 GMT
David, No, it's a whitebox amd64 3200+ on an ASUS board.
Comment by Dale Blount (dale) - Thursday, 27 November 2008, 14:23 GMT
I wonder if it's because IRQBALANCE was turned off in the kernel?
Comment by Dale Blount (dale) - Saturday, 29 November 2008, 18:41 GMT
2.6.27.7-1 doesn't fix this either.
Comment by Dale Blount (dale) - Saturday, 29 November 2008, 18:43 GMT
David, are you sure 2.6.27.1-1 is failing for you? I used that version for more than a month with my normal sucess rate of 80% or so.
Comment by David (dsr) - Saturday, 29 November 2008, 19:53 GMT
Dale, you're right. 2.6.27.1-1 just worked fine for me. My mistake. What do you mean 80% success rate? Does your computer fail to resume 20% of the time, forcing you to perform a hard reboot?
Comment by Dale Blount (dale) - Monday, 01 December 2008, 03:15 GMT
Yes, it used to fail about 20% of the time showing the same symptoms as now (no signal on vga, but fans spin up). I much preferred that to the current 100% failure rate.
Comment by paul (pwt) - Sunday, 07 December 2008, 18:32 GMT
I also believed there was something wrong. However, I had not tried to implement suspend to ram prior to kernel 2.6.27.3, so I could not tell if it worked before.

What I just found is that my BIOS was no set up to enable suspend to ram. I just enabled it and now it works fine with kernel 2.26.27.7.

I would tus advise you to check your BIOS settings, just to be sure...
Comment by Dale Blount (dale) - Sunday, 07 December 2008, 19:13 GMT
There is no such option in my bios.
Comment by David (dsr) - Sunday, 07 December 2008, 19:25 GMT
My MacBook doesn't even use BIOS. :-) And I don't feel comfortable messing with the EFI settings. Besides, suspend to ram works fine with 2.6.27.1-1.
Comment by Dale Blount (dale) - Monday, 08 December 2008, 16:08 GMT
Hibernate appears to work on my HP laptop with x86-64 2.6.27-7, so it must be only broken on certain hardware or in i686 only.
Comment by paul (pwt) - Monday, 08 December 2008, 19:00 GMT
I forgot to mention that it works for me on an i686 install. It must be certain hardwares only.
Comment by Dale Blount (dale) - Tuesday, 09 December 2008, 04:20 GMT
Still broken in 2.6.27-8 from testing.
Comment by Dale Blount (dale) - Tuesday, 09 December 2008, 04:32 GMT
I'm guessing this is related: http://lists-archives.org/linux-kernel/19755613-hangs-on-wakeup-from-s2ram.html
Comment by Konni (konni) - Thursday, 11 December 2008, 13:47 GMT
for me suspend works fine until 2.6.27-7.
the new 2.6.27-8 breaks it, on resume all i get is a black screen instantly.
Comment by Dale Blount (dale) - Saturday, 13 December 2008, 05:49 GMT
I've just spent a few hours trying to debug with PM_TRACE... http://www.kernel.org/doc/Documentation/power/s2ram.txt

I can't get anything to print out after the magic number and my RTC never changes. Does this work for anyone?
Comment by paul (pwt) - Wednesday, 17 December 2008, 14:35 GMT
Works for me (TM). AsRock Motherboard, no proprietary video drivers, i686.
Comment by Dale Blount (dale) - Thursday, 18 December 2008, 18:56 GMT
This is supposed to be fixed in 2.6.27.10.
Comment by Konni (konni) - Monday, 22 December 2008, 17:14 GMT
just updated to 2.6.27.10 - no change.
Comment by Dale Blount (dale) - Tuesday, 23 December 2008, 04:18 GMT
No change for me either...
Comment by Dale Blount (dale) - Saturday, 27 December 2008, 21:43 GMT
Appears to be fixed by 2.6.28 for me... whohoo!
Comment by David (dsr) - Sunday, 28 December 2008, 00:29 GMT
2.6.28 from [testing] solves for me too, yay!
Comment by Konni (konni) - Monday, 29 December 2008, 23:46 GMT
  • Field changed: Percent Complete (100% → 0%)
2.6.28 still doesn't fix this for me.
Comment by Dale Blount (dale) - Tuesday, 30 December 2008, 03:00 GMT
Konni, as yours didn't start breaking until 2.6.27.7, it's probably not the same bug.
Comment by Konni (konni) - Tuesday, 30 December 2008, 07:33 GMT
looks like you're right.
guess its caused by some other thing, but i got the same results you had.
Comment by Konni (konni) - Tuesday, 30 December 2008, 15:35 GMT
looks like you're right.
guess its caused by some other thing, but i got the same results you had.
Comment by Konni (konni) - Monday, 12 January 2009, 15:38 GMT
updated to 2.6.28-3 and it seems to work... yay!
(might have worked in 2.6.28-2 as well, didnt check)

Loading...