Created attachment 116472 [details]
dmesg output

Created attachment 116473 [details]
xrandr.log

Here is a video showing the bug while pressing the Prt Scrn button:
https://drive.google.com/file/d/0B7D2Y0QXFND2aHFvcXdSelQyQVE/view?usp=sharing

Is it always a fullscreen glitch?

(In reply to Chris Wilson from comment #4)
> Is it always a fullscreen glitch?

Yes.

Pretty sure that is a display underrun. Could you please build a vanilla upstream kernel, say drm-intel-fixes from http://cgit.freedesktop.org/drm-intel/log/?h=drm-intel-fixes (or the latest Linus tag) and let's proceed from there?

OK I will try this ASAP. But don't expect any news from me for some weeks.

Ok, a couple of quick(er) tests to try then:

echo > /etc/X11/xorg.conf.d/20-intel.conf <<EOF
Section "Device"
  Identifier "igfx"
  Driver "intel"
  Option "AccelMethod" "blt"
  #Option "AccelMethod" "none"
  #Option "AccelMethod" "uxa"
EndSection
EOF

then switch between blt/none/uxa for a couple of test runs. Other tests would be running at different modes, or even enabling Option "TearFree" in the above conf snippet.

Actually, I remember thinking it is probably the HWCursor playing up when it changes size - part of the reason why I want to investigate a new kernel (or you can try 3.19).

Unfortunately there is no simple option to disable the HWCursor in xorg.conf for Intel, probably should add one, but if you want to experiment you can try

diff --git a/src/sna/sna_display.c b/src/sna/sna_display.c
index 965c35c..d1102fe 100644
--- a/src/sna/sna_display.c
+++ b/src/sna/sna_display.c
@@ -5698,6 +5698,8 @@ sna_cursor_pre_init(struct sna *sna)
        if (sna->mode.num_real_crtc == 0)
                return;
 
+       return;
+
 #define LOCAL_IOCTL_GET_CAP    DRM_IOWR(0x0c, struct local_get_cap)
 #define DRM_CAP_CURSOR_WIDTH   8
 #define DRM_CAP_CURSOR_HEIGHT  9

or

diff --git a/src/sna/sna_display.c b/src/sna/sna_display.c
index 965c35c..3c373bc 100644
--- a/src/sna/sna_display.c
+++ b/src/sna/sna_display.c
@@ -5705,6 +5705,7 @@ sna_cursor_pre_init(struct sna *sna)
 #define I915_PARAM_HAS_COHERENT_PHYS_GTT 29
 
        sna->cursor.max_size = 64;
+       return;
 
        cap.value = 0;
        cap.name = DRM_CAP_CURSOR_WIDTH;

the first disables the HW cursor entirely, the second disables cursor resizing.

I will try the xorg.conf modification, and also install a fresher kernel 4.0 from ELRepo and in last case *try* to build a vanilla kernel with your modification.
But sorry I already doesn't have access to this computer so I can't try anything at this moment.

Here is the results of the tests I just executed:

kernel 3.10 (default centos) + no xorg.conf (default centos) = glitch
kernel 3.10 (default centos) + xorg.conf "blt"  = glitch
kernel 3.10 (default centos) + xorg.conf "none" = glitch
kernel 3.10 (default centos) + xorg.conf "uxa"  = FIXED!

kernel 4.1 (elrepo) + no xorg.conf (default centos) = glitch
kernel 4.1 (elrepo) + xorg.conf "blt"  = glitch
kernel 4.1 (elrepo) + xorg.conf "none" = glitch
kernel 4.1 (elrepo) + xorg.conf "uxa"  = FIXED!

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/xorg/driver/xf86-video-intel/issues/54.