Please attach /sys/class/drm/card0/error

I did some more investigation and upgraded to the latest packages again (intel-dri-10.2.2-1  linux-3.15.2-1  mesa-10.2.2-1  mesa-libgl-10.2.2-1 xf86-video-intel-2.99.912-1) which reproduces the issue and makes the display stay dark.

Then I only downgraded the kernel (3.15.2-1 => 3.14.5-1) and voilà: the system boots fine.

So the actual problem seems to be in the kernel, not in one of these: intel-dri-10.2.2-1 mesa-10.2.2-1  mesa-libgl-10.2.2-1 xf86-video-intel-2.99.912-1

(In reply to https://bugs.freedesktop.org/show_bug.cgi?id=80639#c1)
Thanks for your quick reply.
I could try to upgrade to the newer kernel that reproduces the issue and connect to the laptop via ssh in order to get /sys/class/drm/card0/error

I will do that tomorrow, as I do not have the time to do that now.

Created attachment 101958 [details]
/sys/class/drm/card0/error

(In reply to comment #1)
Here it is.
Connecting to the laptop via ssh worked like a charm.

This has the hallmarks of being invalid stolen. Can you try:

diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c
index 78fa532..2bce530 100644
--- a/drivers/gpu/drm/i915/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/i915_gem_stolen.c
@@ -330,6 +330,8 @@ int i915_gem_init_stolen(struct drm_device *dev)
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct resource *r;
 
+       return 0;
+
 #ifdef CONFIG_INTEL_IOMMU
        if (intel_iommu_gfx_mapped && INTEL_INFO(dev)->gen < 8) {
                DRM_INFO("DMAR active, disabling use of stolen memory\n");

Sure.
I compiled an Arch kernel yesterday, but made a mistake, so that the patch has not been applied to the sources. So I will do another compile today.

The kernel source was 3.15.2. To apply your patch, the package build description file (PKGBUILD) needs to be modified.

This was the first time I compiled a kernel for Arch linux. I managed the patch file to be read and considered valid by PKGBUILD, but did not see that the PKGBUILD script had to be modified at another place, so that the patch actually gets applied to the sources.

Ok, your patch worked.

I could boot the laptop and the display is working.
The log did not contain errors any more.

$ uname -r
3.15.2-1-custom

Created attachment 102200 [details]
screen before loading X

After loading Xorg screen gets blank and the message appears in syslog:
"kernel: [   12.428567] [drm] GPU HANG: ecode -1:0x00000000, reason: Command parser error, iir 0x00008000, action: continue" ...

I found the commit that introduced this behaviour: 1ad292b51e358c8b6e9b8966889c21f1fe705489 (git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git). In my case display was fixed after applying this:

--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -11038,7 +11038,6 @@ static void intel_init_display(struct drm_device *dev)
                        i9xx_update_primary_plane;
        } else {
                dev_priv->display.get_pipe_config = i9xx_get_pipe_config;
-               dev_priv->display.get_plane_config = i9xx_get_plane_config;
                dev_priv->display.crtc_mode_set = i9xx_crtc_mode_set;
                dev_priv->display.crtc_enable = i9xx_crtc_enable;
                dev_priv->display.crtc_disable = i9xx_crtc_disable;

A patch has been committed to the kernel sources 18 days ago and should be contained in v3.16.

https://github.com/torvalds/linux/commit/f1e1c2129b79cfdaf07bca37c5a10569fe021abe

I'll test the new kernel when it is available as an arch package.

(In reply to comment #11)
Actually the patch is contained in v3.15.6

Today I installed kernel 3.15.7-1-ARCH as an arch package and rebooted.

The laptop started normally, no GPU crash occured.

The log file did not contain errors any more.