Bug 33439

Resume is just ever so slightly snafu with 2.6.38 atm. Reverting

commit 01fe9dbde19a1a27b8ee63e2d964562962e1eb78
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sun Jan 16 19:37:30 2011 +0000

    drm/i915: Use ACPI OpRegion to determine lid status

has helped some people, but not my machines.

I should also mention that the problem does not occur when I use kernel 2.6.36, with the two patches mentioned here https://bugs.freedesktop.org/show_bug.cgi?id=29278#c63 . Basically they add a 300 ms delay to ironlake_panel_on and ironlake_panel_off in intel_dp.c . When I increase those delays to 900ms everything works as expected.

Works on my E6510, as Chris says maybe you're running into one of the other breakages?

I used echo mem > /sys/power/state to avoid getting into trouble with distro scripts (I always do this first to remove variables), but using the distro suspend menu also worked.

Hm, just saw a resume failure on drm-intel-next, but I didn't see it on edp-fixes-2.  Trying again.

Yeah, drm-intel-next resume seems broken, but the machine is hung; it's not just the panel not coming on (which is what I'd expect if it were an edp related failure), so this could be some other upstream problem.  Will debug.

I reverted the patch mentioned by Chris and recompiled drm-intel-next. Resuming from suspend is still broken, the machine is indeed hung. 

With edp-fixes-2, my machine resumed succesfully once out of 5 attempts. 



(In reply to comment #5)
> Yeah, drm-intel-next resume seems broken, but the machine is hung; it's not
> just the panel not coming on (which is what I'd expect if it were an edp
> related failure), so this could be some other upstream problem.  Will debug.

This is what an Ubuntu user reports on launchpad today:

"Today I installed kernel 2.6.38-997.201101241227 from drm-intel-next. It boots to X. However: 1) suspend and resume does not work, 2) when I try to reboot or shut down by clicking on the appropriate menu option it merely logs me out."

I have a dell e6510, full hd.  I have been using he edp-2 branch and it has been suspending and resuming with occasional problems (say 1 out 50 times it does not resume).  drm-intel-next however does not resume (consistently).

So I tried the edp-fixes-2 kernel again because everybody else seems to have success with it. I found out that resuming from suspend works if I use the menu option (it takes a while and I see what appear to be boot messages though). 

Resuming does NOT work when I suspend by closing the lid. Gnome is configured to suspend when I do that, and it suspends fine, but the display does not come on when I resume. The system is not hung because I can reboot the computer by typing my password and hitting the gnome keyboard shortcut. 


(In reply to comment #8)
> I have a dell e6510, full hd.  I have been using he edp-2 branch and it has
> been suspending and resuming with occasional problems (say 1 out 50 times it
> does not resume).  drm-intel-next however does not resume (consistently).

(In reply to comment #9)
> So I tried the edp-fixes-2 kernel again because everybody else seems to have
> success with it. I found out that resuming from suspend works if I use the menu
> option (it takes a while and I see what appear to be boot messages though). 
> 
> Resuming does NOT work when I suspend by closing the lid. Gnome is configured
> to suspend when I do that, and it suspends fine, but the display does not come
> on when I resume. The system is not hung because I can reboot the computer by
> typing my password and hitting the gnome keyboard shortcut. 


Is it the panel or the backlight?  With some kernels / patches it happened to me that the panel was on but the backlight was off.  Changing the brightness with fn+arrow keys happened to turn it on.

You might try pulling Linus's tree.  It has some suspend/resume fixes that get things working again for me.  The backlight doesn't come back, but everything else seems to come up fine (you can just echo a new backlight value into /sys/class/backlight/dell_backlight/brightness as a workaround until the backlight driver is fixed).

I ended up pulling Linus's tree into a copy of the drm-intel-next branch I had checked out, but I think Linus's tree by itself would also work.

(In reply to comment #11)
> You might try pulling Linus's tree.  It has some suspend/resume fixes that get
> things working again for me.  The backlight doesn't come back, but everything
> else seems to come up fine (you can just echo a new backlight value into
> /sys/class/backlight/dell_backlight/brightness as a workaround until the
> backlight driver is fixed).
> 
> I ended up pulling Linus's tree into a copy of the drm-intel-next branch I had
> checked out, but I think Linus's tree by itself would also work.

I have tested it a few hours ago.  The panel is not switched on on boot for me.  I have a e6510, full HD, 32 bits.

(In reply to comment #12)
> (In reply to comment #11)
> > You might try pulling Linus's tree.  It has some suspend/resume fixes that get
> > things working again for me.  The backlight doesn't come back, but everything
> > else seems to come up fine (you can just echo a new backlight value into
> > /sys/class/backlight/dell_backlight/brightness as a workaround until the
> > backlight driver is fixed).
> > 
> > I ended up pulling Linus's tree into a copy of the drm-intel-next branch I had
> > checked out, but I think Linus's tree by itself would also work.
> 
> I have tested it a few hours ago.  The panel is not switched on on boot for me.
>  I have a e6510, full HD, 32 bits.

(I meant I tested Linus' tree, not the drm-intel-next branch)

I have install drm-intel-next today on my E6410 and when I press suspend to RAM it goes to plymouth screen and that's it. When I move mouse then I get back to X. Is this the right bug to report?

(In reply to comment #14)
> I have install drm-intel-next today on my E6410 and when I press suspend to RAM
> it goes to plymouth screen and that's it. When I move mouse then I get back to
> X. Is this the right bug to report?

In this case, it is unlikely to be the i915.ko that is preventing your system from suspending.

I've installed drm-intel-next (commit 1a3665c81df32b23c38d4ba8a74761551d5673b1) today, and it boots to X.  It also suspends and resume successfully so far BUT resuming does not switches on the backlight.  However, changing the brightness does the trick.  I have not tried hibernating.


Cheers!

(In reply to comment #16)
> I've installed drm-intel-next (commit 1a3665c81df32b23c38d4ba8a74761551d5673b1)
> today, and it boots to X.  It also suspends and resume successfully so far BUT
> resuming does not switches on the backlight.  However, changing the brightness
> does the trick.  I have not tried hibernating.


Did some further tests -- apparently that version does not handle output through the VGA very well: when I activated VGA output the internal panel is swithed off and I found no way of switching it on again.  In the text login console (VT1) the following message appears: 

[drm:indel_disable_transcoder] *ERROR* failed to disable transcoder

which I do not know whether it's related or not with switching off the panel.

Best.

Today I finally installed 2.6.38-rc2 from drm-intel-next, head 1a3665. Resuming from suspend works, but the display brightness is very low. Adjusting the brightness does not help, I have to unplug the powercord to bring things back to normal.

(In reply to comment #18)
> Today I finally installed 2.6.38-rc2 from drm-intel-next, head 1a3665. Resuming
> from suspend works, but the display brightness is very low. Adjusting the
> brightness does not help, I have to unplug the powercord to bring things back
> to normal.

Weird.  My biggest problem with that version is that connecting an external projector to the VGA output (something I need to make at least once per week) does not not work, since the internal panel is shut down and I cannot bring it back to life again wo. rebooting.  I suppose that this is something which comes from some other commit, but it is a big inconvenient for me.

I need VGA and HDMI output every once in a while, too, but I can't test it right now because I'm travelling. Maybe you can open a new bug report for this issue? If you can let me know the bug number I'll report my findings there when I'm back home in 2 weeks. 

(In reply to comment #19)
> (In reply to comment #18)
> > Today I finally installed 2.6.38-rc2 from drm-intel-next, head 1a3665. Resuming
> > from suspend works, but the display brightness is very low. Adjusting the
> > brightness does not help, I have to unplug the powercord to bring things back
> > to normal.
> 
> Weird.  My biggest problem with that version is that connecting an external
> projector to the VGA output (something I need to make at least once per week)
> does not not work, since the internal panel is shut down and I cannot bring it
> back to life again wo. rebooting.  I suppose that this is something which comes
> from some other commit, but it is a big inconvenient for me.

I have the e6510, i5, 1920x1050 display, 2.6.37-rc8-g1a7ed43

Resume fails about one time in five.  When it does, screen is blank but backlit.  UI events appear to be discarded, except the disk runs for a
bit after I poke the power button momentarily.

This last time, it came up fully with original screen contents, then
the windows blanked (preparatory to redraw), and before redraw could
complete, it switched back to console and then blanked, backlit.
I forgot to try adjusting brightness.

Created attachment 43777 [details]
dmesg after black screen, plug & unplug remote monitor

Dell e6510, i5, 1920x1080, kernel is clean drm-intel-next 569e898.

Machine boots and, usually, resumes normally, but after inactivity 
timeout or some resume attempts, screen remains black but backlit.
Plugging and unplugging an external monitor sometimes brings it 
back. Or, X freezes, as this time, but ssh still works.

The same issue on Dell e4310 and still in 2.6.38 (while other issues are all resolved).

Still a problem?  My E6510 has been behaving well at least...

(In reply to comment #24)
> Still a problem?  My E6510 has been behaving well at least...

Works for me with 2.6.39.1 (at last!). That's the first kernel version I've ever tried that has been able to turn on my display at all without manual patching (though I admit I previously gave up back around 2.6.37).

Great, thanks for the update.

I just tried 2.6.39.1 (ubuntu mainline build) on my E6510, and the problem still seems to be there...

I've done a couple more tests and it seems that resuming works sometimes, but not consistently. Switching to VT1 and back gives a bunch of "unable to connect to X server" errors, but then the screen does turn on.
Seems like this display is a nightmare... 

(In reply to comment #27)
> I just tried 2.6.39.1 (ubuntu mainline build) on my E6510, and the problem
> still seems to be there...