[Intel-gfx] [PATCH] EXA pixmap management support (for GTT mapping)

[Alan W. Irwin]

Eric said:

> I'd really like to see us drop XAA at this point.  It's not like it really
works on newer hardware, anyway.  And if we could unifdef EXA, I'd be even
happier -- it really makes a mess of a bunch of our code.

Please don't forget http://bugs.freedesktop.org/show_bug.cgi?id=14464.

I am dating myself here, but remember when Linux users were proud of keeping
their desktop systems up for many months in a row and the huge chuckle we
got over the 49.7-day windows crash (due to an overflowing uptime counter)
which wasn't fixed for a long time because so few windows machines ever
reached 49.7 days of operations without a reboot?

Part of the reason Linux users could go for so many months without a reboot
was XAA was rock-solid.  I used XAA exclusively until 10 months ago, and it
just never locked up with my usual light desktop use (standard 2D KDE-3.x
desktop, many xterms, some konqueror sessions, some repeated runs of a 3D
game which uses DRI, foobillard).

EXA just isn't in that stability class yet.  I have been keeping a diary of
the X lockups I get with EXA go I can provide good information for the above
bug report.  Here is the complete (excluding times when I deliberately took
the machine down) summary of my recent X uptimes.

lockup after ~6 hours
lockup after 15 days
lockup after 18 days
lockup after 3 days

My current test has achieved 15 days (this time with DRI turned off which
makes foobillard unplayable), but I don't have a lot of hope it will get
much higher than that from the above experience and because none of the
above lockups occurred when foobillard was being run.  Probably some sort of
karma is involved here as payback for all my laughing at Windows instability
years ago.  :-) Frankly, I would be very happy to reach 50 days of X uptime
now!

I sympathize with the desire of the Intel developers to drop XAA since it
sounds like nobody wants to maintain it any more.  And with cleaner less
complicated code, it should be easier to make EXA rock solid. Nevertheless,
let me emphasize my first sentence by repeating it:

Please don't forget http://bugs.freedesktop.org/show_bug.cgi?id=14464.

I don't have the coding skills to help directly with debugging this
important issue, but I plan to help indirectly by reporting results at the
above URL as my series of long-term EXA tests continue. Note, my tests are
done with kernel and X packages taken from Debian unstable
(linux-image-2.6.26-1-amd64, version 2.6.26-8; xserver-xorg, version
1:7.3+18; and xserver-xorg-video-intel, version 2:2.3.2-2+lenny4) so I am
currently testing essentially the kernel, X and Intel driver versions that
will be in the forthcoming Debian release.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________