Re: DRM update for testing

On Wed, 2005-12-21 at 17:26 +0100, martinko wrote:
> Eric Anholt wrote:
> > I've got a patch at:
> > http://people.freebsd.org/~anholt/dri/drm-sys-20051123.diff
> > for testing, which merges DRM CVS into FreeBSD.  I've done basic tests
> > of a recent DRM CVS with WITNESS on AGP/PCI Matrox, AGP/PCI Rage 128,
> > AGP Radeon r100/r200/r300, AGP Savage4 (new!), SiS, and 3dfx, and this
> > version should be equivalent to what I tested.  I'll do another pass of
> > testing before I commit, but I'd also like to get some wider testing so
> > more panics can show up if they exist.  This particular diff has only
> > been compile-tested with LINT on amd64 so far.
> > 
> > Known issues:
> > - Radeon PCIGART mode hung for me.  I've heard this is caused by a
> > change in X.Org upstream which itself worked around a different hang, so
> > those of you not using xorg-server-snap wouldn't hit it hopefully.
> > - Savage3d untested iirc
> > - Savage4 PCI doesn't work (needs some love in the memory mapping
> > department, I think).
> > 
> > However, savage support has experimental and will only be on by default
> > with the next release of X.Org, so I feel OK with integrating it into
> > -current.
> > 
> > So, for testing: If you get instant reboots while starting/using X, that
> > almost always means a kernel panic.  This is new behavior in the last
> > year or so afaik (where a number of us have issues dumping while in X).
> > However, I've hooked up serial consoles and got backtraces fine, which
> > hopefully you can do if you hit issues, as well. 
> > 
> > Also, note that this brings in a new enough r300 driver for running Mesa
> > CVS's r300 driver, which hasn't been the case for a while.  It doesn't
> > bring in the via driver, which jakeb ported but which got overcome by
> > new linux-specific code (volunteers to work on this in DRM CVS would be
> > excellent!) in CVS.  Also, the i915 (i830 through i915 integrated
> > graphics) driver should be pretty quick for someone with a bit of newbus
> > knowledge to finish off, or it could be a chance someone who cared to
> > learn more about the FreeBSD kernel :)
> > 
> 
> eric,
> 
> i've got only 64MB video memory, while :
> 
> $ dmesg | g drm
> drm0: <ATI Radeon RV300 Mobility 9600 M10> port 0xc800-0xc8ff mem 
> 0xd0000000-0xd7ffffff,0xff8f0000-0xff8fffff irq 5 at device 0.0 on pci1
> info: [drm] AGP at 0xe0000000 256MB
> info: [drm] Initialized radeon 1.19.0 20050911
> 
> note, that i'm running 6-stable and cvsuped recently just to discover 
> that radeon driver is finally attaching.

Yes, your AGP aperture is a limit on system memory usable by AGP.
Totally unrelated to your video card's framebuffer size.

-- 
Eric Anholt                                     eta_at_lclark.edu
http://people.freebsd.org/~anholt/              anholt_at_FreeBSD.org