Problem with display remaining black after resume

From ThinkWiki
Jump to: navigation, search

There has been a problem encountered where the display stays black on resuming from suspend.

The symptom might have you think first that your system hang up, but you will realize that your ThinkPad works and you can even reset it via CtrlAltDel.

Affected Models

Affected Operating Systems

  • Linux (it's a kernel issue)
  • FreeBSD (6.x at least)
  • Windows XP

Solutions

Quick workaround for R61i, T23, maybe others

Try pressing CTRL+ALT+F1 to switch to text console. The backlight should come on normally. Press CTRL+ALT+F7 to return to X.

This solution is not working on R61i using Debian Squeeze. Upgrading the BIOS fixes the issue for R61i.

On a T23 using Ubuntu Feisty, pressing Fn+F7 (external/internal display change) once or twice brought the display back. After upgrading to Ubuntu Gutsy it doesn't work anymore, but pressing Fn+F3 (blank screen) and Fn (restore display) works.

Quick Workaround for R61 (at least 8918-5QG) using NVidia

Use Vesa driver instead of the proprietary NVidia driver.

Quick Workaround for T61 (at least 7662-CTO) using NVidia Quadro NVS 140

Try pressing Fn+F4 to get the OS suspend to RAM. Nothing on the screen will indicate that the OS is being suspended except for the Sleep LED. Wake up the OS by pressing the Fn key. This induces an additional 5-10 seconds of work. But this has consistently worked with no issues.

Pseudo-solution for R61

On an R61 running Fedora Core 9, the nv driver fails to turn the backlight on after resuming from a suspend to RAM. I fixed this by using the proprietary NVIDIA Linux drivers (v177.82).

Solution for ThinkPad Z60t

  • Display controller: Intel Corporation Mobile 915GM/GMS/910GML Express Graphics Controller (rev 03)
  • Distro: Fedora release 7 (Moonshine)
  • Kernel: Linux 2.6.22.5-76.fc7

The solution is straight forward - just to add configuration parameter for the default pm-utils package. Create file /etc/pm/config.d/config and put there one line DISPLAY_QUIRK_S3_BIOS="true", or execute following command:

echo DISPLAY_QUIRK_S3_BIOS=\"true\" >> /etc/pm/config.d/config

Semi-Solution for ThinkPad X60 with damaged system after s2ram usage

It happend when restarting a s2ram-session.

Symptom: Black screen with blinking "_" sign remaind. (without the ")

System status: HDD idle, fan running, everything else looks to wait for something to happen.

Semi-Solution: Booting with DVD-ROM and going through the installations menu, where you choose "other" and "boot a installed system" (something like that). Gladly it works, and OpenSuSE 10.1 comes up with 50% "failed" messages! I than shutdown properly, rebooted again and had 100% "done" again, with no other things affected.

Further: Repairing with the DVD-ROM crashed massivly(!), so I selected "boot a installed system" as final solution and it worked!

Unknown: Maybe the Solution for ThinkPads with 1400x1050 internal LCD and Intel 915GM will help, because X60s and X60 are very familiar. (Not tested so far.)

(If this Problem is not right here, please edit and move.)

Solution for ThinkPads with 1400x1050 internal LCD and Intel 915GM

see 1400x1050 on Intel 915GM.

Solution for ThinkPads with ATI graphic chips and Intel 915/945GM

Affected models include X41, X60s, X200s, R60, T60, and Z61t.

This soluton also applies to T42 with Intel 855 and ATI 9600 M10.

ATTENTION!
The use of acpi_sleep=s3_bios can cause subtle and pervasive system corruption in kernels after 2.6.32 or so. Its use should be avoided; all other alternatives should be considered first. (Details: acpi_sleep=s3_bios causes the kernel to call the BIOS POST code during resume-from-RAM. The kernel's resume path does not provide the execution environment expected by POST code, resulting in nasty side-effects.)

One solution may be to provide the acpi_sleep=s3_bios kernel parameter in your kernel parameter line.

For grub this would look like this:

title           Linux, kernel 2.6.11-1-686
root            (hd0,0)
kernel          /boot/vmlinuz-2.6.11-1-686 root=/dev/hda1 ro acpi_sleep=s3_bios
initrd          /boot/initrd.img-2.6.11-1-686
savedefault
boot

For lilo it would look like this:

image=/boot/vmlinuz
    append="acpi_sleep=s3_bios"

The actual process of going to sleep is then managed through a sleep script; as a start, see the sleep.sh script in the Extreme Graphics 2 section below, but note the following comments:

In OpenSUSE 10.1 (at least on a T43p), it's necessary to override the default options for s2ram if you're using the newer ATI driver. This can be done putting SUSPEND2RAM_FORCE="yes" and SUSPEND2RAM_ACPI_SLEEP="3" in /etc/powersave/sleep.

In Ubuntu or Kubuntu, it may be necessary to modify /etc/default/acpi-support. In that file, make sure that ACPI_SLEEP is uncommented and set to true. With ATI chips, also make sure that SAVE_VBE_STATE is uncommented and set to true; with Intel chips, on the other hand, ensure that nothing is done with respect to VBE--no reposts, no state saves. Also commenting POST_VIDEO may help.

In Fedora, it may be necessary with the Intel chips to edit the resume_video() function in /etc/pm/functions-intel to comment out the VBE post and restore. (As of FC6 these seem to be pre-commented out.) Also, the laptop, after waking up, may go back to sleep immediately or whenever the AC adapter is disconnected. When this happens, it's caused by a bug in the HAL daemon that incorrectly reports certain ACPI events. This is a known problem and a simple workaround is described here.

NOTE!
It is possible this method will not work if the laptop is docked. It is also possible that the cited workaround for the HAL daemon bug will not work on some machines. A kludgier workaround in this event is to kill the HAL daemon on suspend. This necessitates the resuscitation of GPM upon resume.

Another solution is to use vbetool. If you are using Debian with the hibernate package, uncomment "EnableVbetool yes" in /etc/hibernate/hibernate.conf (or /etc/hibernate/ram.conf).



On T60 2007-CTO (Core2Duo 2Ghz, 2GB Ram, ATI X1400) the screen stayed blank after suspend-to-ram until I set vga=0 in lilo.conf.

Working config:

Linux 2.6.21.5
fglrx 8.37.6
debian etch:
 powersaved 0.14.0-5:
  UNLOAD_MODULES_BEFORE_SUSPEND2DISK="usb_storage ohci_hcd uhci_hcd ehci_hcd ipw3945 pcmcia yenta_socket rsrc_nonstatic pcmcia_core"
  UNLOAD_MODULES_BEFORE_SUSPEND2RAM="usb_storage ohci_hcd uhci_hcd ehci_hcd ipw3945 pcmcia yenta_socket rsrc_nonstatic pcmcia_core"   
 hibernate:
  SwitchToTextMode yes
 lilo.conf:
  vga=0

"EnableVbetool yes" and other suggestions didn't work for me.

For suspend-to-disk, don't load fglrx in initrd.


On T60-20076RG (Core2Duo 2GHz, ATI X1400) with OpenSUSE 11.1 and fglrx 8-12 the following had to be done to get suspend to RAM always resume:

  • Add S2RAM_QUIRKS_SOURCE="s2ram" to file /etc/pm/config.d/config
  • Create an executable script /etc/pm/sleep.d/00text containing:
#!/bin/bash

case "$1" in
        hibernate|suspend)
                /bin/chvt 1
                ;;
        thaw|resume)
                /bin/chvt 7
                ;;
esac

There seems to be a bug (Novell bugzilla) which makes it impossible for s2ram to switch to text console while suspending through pm-suspend. The script above forces console change. Along with setting s2ram as quirks source (which makes it correctly set acpi_bios before suspend - to s3_bios,s3_mode for T60 2007*) this can make resume work flawlessly despite using vesafb.

Solution for ThinkPads with Intel Extreme Graphics 2

NOTE!
On X40s/X41s - even with Intel Extreme Graphics - and for R52s with Intel Graphics Media Accelerator 900 the solution for ATI graphics chips above is reported to work. In this case, make sure no changes to VBE are made, especially no state saves and no reposts.

The following solution should work on 865G, 865GV, 855GM, 855GME, 852GME chipsets.

  • First of all, do not use the acpi_sleep=s3_bios kernel parameter.
  • Second, completely remove framebuffer support from your kernel. If it's built as modules, it is important that they do not get loaded at all.
  • Before suspending, change to a console and safe the video state with # cat /proc/bus/pci/00/02.0 > /tmp/video_state.
  • On resume, restore the video state with # cat /tmp/video_state > /proc/bus/pci/00/02.0 and change back to X.
  • For Debian Etch 4.0 on R50e just make following changes to /etc/default/acpi-support:
#SAVE_VBE_STATE=true
#VBESTATE=/var/lib/acpi-support/vbestate
#POST_VIDEO=true
SAVE_VIDEO_PCI_STATE=true
  • For a R50e the only thing needed to make suspend to ram work in Ubuntu 6.06 is adding
Option  "VBERestore" "yes"

to the Device section in your /etc/X11/xorg.conf, and the example script below.

The following example /etc/acpi/actions/sleep.sh script shows how to integrate the according lines.

#!/bin/bash

# change to console 1
FGCONSOLE=`fgconsole`
chvt 6

# safe video state
cat /proc/bus/pci/00/02.0 > /tmp/video_state

# sync filesystem
sync

# sync hardware clock with system time
hwclock --systohc

# go to sleep
echo -n 3 > /proc/acpi/sleep

# waking up
# restore system clock
hwclock --hctosys

# restore video state
cat /tmp/video_state > /proc/bus/pci/00/02.0

# change back to X
chvt $FGCONSOLE

# clean up behind us
rm /tmp/video_state

With Ubuntu 6.10 on a R51 (2887-32G) I just (as none of the other tricks above) had to add fb=false to the kernel line in /etc/grub/menu.lst and edit /etc/defaults/acpi-support this way:

SAVE_VBE_STATE=false
POST_VIDEO=false
SAVE_VIDEO_PCI_STATE=true
USE_DPMS=false
DOUBLE_CONSOLE_SWITCH=false

Solution for ThinkPads with Intel I830 Chipset

The following solution worked for me on an X30 with I830M chipset with kernel >= 2.6.16.

it *almost* works on my T400S (intel mobile 4 integrated graphics) and kernel lenny-2.6.30-amd64, but see below for better solution. (sometimes, it doesn't wake up at all, probably because some kernel modules unrelated to video don't like this suspend method. once, i had to restart gdm. i am using a weird setup if two gdm sessions on two virtual terminals, don't know if that's required to reproduce the latter problem.)

this works with vesafb and also with intelfb frambuffer support.

The following example /etc/acpi/actions/sleep.sh script shows how to integrate the according lines.

#!/bin/bash

FGCONSOLE=`fgconsole`
chvt 8
sync
hwclock --systohc

echo -n "mem" > /sys/power/state

hwclock --hctosys
vbetool post

if [ "$FGCONSOLE" -ge "7" ] ; then
  chvt $FGCONSOLE
else
  chvt 7
  chvt $FGCONSOLE
fi

If it still doesn't work try to add

Option  "ForceEnablePipeA" "true"

to the Device section in your /etc/X11/xorg.conf.

Solution for ThinkPads with ATI graphic (and possibly other) chips and FreeBSD

The FreeBSD acpi(4) manpage mentions a tunable parameter, "hw.acpi.reset_video":

   hw.acpi.reset_video
            Reset the video adapter from real mode during the resume path.
            Some systems need this help, others have display problems if it
            is enabled.  Default is 0 (disabled).

This tunable can be set by adding the following line to your FreeBSD machine's /boot/loader.conf file:

   hw.acpi.reset_video="1"

And rebooting your machine. Hopefully, the next time you resume from a suspend, you'll see your video again. This solution doesn't appear to be specific to ATI hardware in any way, so I presume it would be helpful for video chipsets other than ATI, as well.

If this entry doesn't help you, you might consider searching in the FreeBSD-Mobile email-list archive for more insight.



FOOTNOTES [Δ]
  1. If you have this problem with R50e and the above solution doesn't work, try switching to console first. An example sleep script can be found here.

Solution using s2ram for Intel 915/945GM

Just using the "s2ram -f -p" command from the uswsusp package will work from within X, at least on a Z61e. On X60s it is enough to issue the "s2ram" command and it works. On X61 "s2ram -f -a 1" can work properly. Best idea seems to be to put this into the corresponding acpi script:

% cat /etc/acpi/sleep.sh 
#!/bin/sh
test -f /usr/share/acpi-support/power-funcs || exit 0
test -f /usr/sbin/s2ram || exit 0
rmmod usb_storage
rmmod uhci_hcd
rmmod ehci_hcd
/usr/sbin/s2ram -f -a 1 -m
modprobe uhci_hcd
modprobe ehci_hcd
modprobe usb_storage

Source: http://d.hatena.ne.jp/conceal-rs/20080309/1205083315 Works on my X61.

T400S (intel mobile 4 integrated graphics) and kernel lenny-2.6.30-amd64: seems good so far (around 5 suspends without problems).

Solution using DOUBLE_CONSOLE_SWITCH

By setting the following in /etc/default/acpi-support the display comes back on X61s using Intel chipset:

    DOUBLE_CONSOLE_SWITCH=true

Fedora 8 doesn't have DOUBLE_CONSOLE_SWITCH, but it works when one does: First, add option "VBERestore" "true" to /etc/X11/xorg.conf

   Section "Device"
       Identifier  "Videocard0"
       Driver      "intel"
       Option      "VBERestore"  "true" 
   EndSection

Then suspends with

   pm-suspend --quirk-vbemode-restore --quirk-s3-bios

Solution for nvidia-drivers-180* series

The proprietary NVidia drivers of the 180 series introduce several problems with suspend to ram:

  • Situation 1: Suspend from console, no X-Server running:

You might need to use vbetool to save and restore the vbestate. When using hibernate-script, this can be done by setting the following config variables:

EnableVbetool yes
RestoreVbeStateFrom /var/lib/vbetool/vbestate
VbetoolPost yes

You may need to run mkdir -p /var/lib/vbetool && vbetool vbestate save > /var/lib/vbetool/vbestate first.

On newer distributions, you might need to not use vbetool. On a Ubuntu Hardy with Linux 2.6.24, and probably on other Debian-based distributions, edit '/etc/default/acpi-support' and set 'SAVE_VBE_STATE=false'.

  • Situation 2: Suspend from running X-Server:

You cannot use vbetool or any other quirks, since it seems to confuse the nvidia X driver. That means you should enter S3 simply by doing echo mem > /sys/power/state.

If you have your hotkeys handled by acpid, you might differentiate those two cases by checking for a running X process in your hotkey handler (i.e. /etc/acpid/default.sh):

pgrep -x X > /dev/null \          # checks for running process with name "X"
 && echo mem > /sys/power/state   # if found, do plain S3 suspend
 || hibernate-ram                 # otherwise, run quirked script

Furthermore, it seems to be a good idea to use the 180 series with a 2.6.28* kernel.

It has been reported that acpi_sleep=S3_bios should be used instead of acpi_sleep=S3_mode as a boot option.

See, http://www.nvnews.net/vbulletin/showthread.php?t=123303&highlight=suspend&page=6


It might also help to put blacklist intel_agp in /etc/modprobe.d/blacklist

Finally, it seems to depend on precise model nr. See https://bugs.launchpad.net/ubuntu/+bug/235284 for a discussion and patch.

Quirk workaround for T410s

The vbe post quirk in pm-utils works on a T410s 29123KC.

On Debian sid pm-utils script is hacked to apply '--quirk-vbe-post' when kms is in place.