xf86-video-intel v.2.2 and DRM

["Chavdar Ivanov" <ci4ic4%gmail.com@localhost>]

Hi,

Apologies for the cross-posting. I am not sure which is the best place.

-------------------begin quote---------
Date: Wed, 30 Jan 2008 13:40:33 +0000 (UTC)
From: Tobias Nygren <tnn%netbsd.org@localhost>
Subject: CVS commit: pkgsrc/x11/xf86-video-intel

Module Name:    pkgsrc
Committed By:   tnn
Date:           Wed Jan 30 13:40:33 UTC 2008

Modified Files:
       pkgsrc/x11/xf86-video-intel: Makefile distinfo

Log Message:
Update to xf86-video-intel-2.2.0.
Long list[1] of bugfixes. Fixes several bugs with i965, EXA, tv-out,
LCD detection, console switching, DRM, >4GB address space and more.
[1] http://lists.freedesktop.org/archives/xorg/2007-November/030246.html
-------------------------end quote-----------


broke *for me* Intel DRM.

Could anyone confirm:

a. that xf86-video-intel 2.2 works with DRM on NetBSD-current (i386,
just 1GB memory if that is of any significance; one of the messages
suggests yes).

&& b. Did [s]he use the unpatched DRM related kernel bits ( for me,
any attempt to run Intel DRM *without* Yorick Hardy's patches result
in a hard lock - only the power switch brings me back; luckily, they
still apply cleanly - with some minor modification at least - see
http://mail-index.netbsd.org/tech-kern/2007/10/08/0002.html ).

&& c. The combination of Xorg server (+options to compile),
MesaLib-dri, libdrm which lead to this.

I have been running generally fine modular-xorg-server 1.3 with the
above patches , but the xf86-video-intel was from the original source
as of maybe 7-8 months ago - not quite sure. Seeng the above message,
I decided to give it a try and lost it. The kernel part seems still
fine:

# dmesg|grep drm
i915drm0 at vga1
info: [drm] Intel i945GM (unit 0)
info: [drm] AGP at 0xd0000000 256MB
info: [drm] Initialized i915 1.5.0 20060119

Xorg.0.log also seems kosher with the exception of:

(EE) AIGLX error: drmMap of framebuffer failed (Invalid argument)(EE)
AIGLX: reverting to software rendering

which was not the case before. Last, glxinfo with LIBGL_DEBUG=verbose gives me:

libGL: XF86DRIGetClientDriverName: 1.9.0 i915 (screen 0)
libGL: OpenDriver: trying /usr/pkg/lib/dri/i915_dri.so
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is 4, (OK)
drmOpenByBusid: Searching for BusID pci:0000:00:02.0
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is 4, (OK)
drmOpenByBusid: drmOpenMinor returns 4
drmOpenByBusid: drmGetBusid reports pci:0000:00:02.0
libGL error: drmMap of framebuffer failed (Invalid argument)
libGL error: reverting to (slow) indirect rendering
...

I can't revert to the old xf86-video-i810, as the saved binary package
I have has got the .so file as just a soft link to the (manually
compiled) intel.so library, which was overwritten by the latest
package.

I certainly did quite q bit of googling for a selection of the above
messages; the result is, as expected, only Linux forum blurb pointing
to driver replacement and no idea why the above happens.

It is worth also mentioning that according to the announce message
quoted above the xf86-video-intel driver should be used with xorg 1.4;
downstream it may work, but various problems are expected... Is this
one of them? I *could* switch to 1.4, but that lead me (on a different
machine) to the need of modifying substantial pkgsrc part (all of
Gnome, actually, for now), so I would rather avoid it for now.

Basically, the question is, do we have DRM on Intel chips, and if yes,
what is the kosher way to achieve it.

----------------------------------------------------------------
/dev/random says:
        I don't suffer from insanity. I enjoy every minute of it.
----------------------------------------------------------------
Chavdar Ivanov   | Talbot Way, Small Heath Business Park
Delcam UK        | Birmingham B10 0HJ, United Kingdom
Customer Support | (+44)121-6831014
----------------------------------------------------------------