yesterday I updated my xorg-server from 18.104.22.1681-r2 (stable xorg-server version from the gentoo portage repo) to 1.7.2.
Consequently I had to update xf86-video-intel (from 2.9.1 to git master). And here the problems began.
Some more information about the system:
kernel: gentoo-sources-2.6.31-r6 (vanilla sources with some light patches)
mesa: git master
libdrm: git master
gfx: Intel Corporation Mobile 915GM/GMS/910GML Express Graphics Controller
KMS: activated (and working)
The problems appearing now were:
(i) Extreme input/display lag: In gdm e.g. it takes about 3 seconds for an input to appear on the screen.
The problem however is not really the input, but a delay when the screen content is refreshed - somehow it's taking just too long.
What I mean is: I can just login normally (blind typing) and everything works, it's just that the visual output is extremely delayed.
The same applies to window activities inside XFCE. Sometimes window contents don't appear at all when a new dialogue box is opened. However when moving the box or doing other activities the visual content suddenly updated. Very weird...
(ii) Opening a restored session in Seamonkey (Mozilla suite) sometimes crashes X, sometimes only hangs it - with Magic SysRq still working. Switching to console didn't work when I produced a hang.
The hangs and crashes described in (ii) are accompanied from these errors in dmesg:
[drm:i915_gem_object_bind_to_gtt] *ERROR* Invalid object alignment requested 4096
I managed to avoid (ii) by upgrading my kernel to drm-next git master. But (i) remains.
The issue seems to a regression of the intel driver, and not of xorg-server. Sadly I couldn't downgrade intel to 2.9.1 with xorg-server-1.7.2 since I doesn't build then.
But I did this:
a) Downgrade xorg-server to 1.7.1 - no change in (i)
b) Downgrade intel to 2.9.1 (compiles on 1.7.1) -> issue gone
So looks like a regression between 2.9.1 and git master. I want to emphasize that I have NO fancy desktop effects like compiz or stuff activated. XFCE is NOT using a compositing.
This may be dup with bug#25409. You can firstly try reverting that commit to see if that works.
Yep, looks like it's a duplicate. I just updated xf86-video-intel and commit 415aab474edd1425034981306718afd8506445f1 has also fixed this issue!
*** This bug has been marked as a duplicate of bug 25409 ***