<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html> <head> <meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type"> </head> <body bgcolor="#ffffff" text="#000000"> Jordan, So with the NoAccel True option placed inside the card section, my graphics work. :) Everything isn't perfect though.  (As I'm sure you suspect)  If I try to terminate the X server the graphics are stalled (hung).  I can still access the system via remote shell. If I say inside X, and run my desktop, I can use it! :) XFCE4 is working just fine for me. Anyway, I now have a high res screen with many colors (don't know what color depth I have) Top is reporting a light system load with the drivers & Xorg (before the CPU was pegged) (Note: I'm running make also on Firefox) :)  Love FreeBSD...   PID USERNAME     VCSW  IVCSW   READ  WRITE  FAULT  TOTAL PERCENT COMMAN last pid:  1041;  load averages:  0.02,  0.13,  0.08    up 0+00:06:48  09:57:11 41 processes:  1 running, 40 sleeping CPU states:  0.8% user,  0.0% nice,  1.2% system,  1.2% interrupt, 96.9% idle Mem: 37M Active, 26M Inact, 38M Wired, 260K Cache, 33M Buf, 125M Free Swap: 470M Total, 470M Free   PID USERNAME    THR PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND  1034 root          1  96    0  6256K  3292K select   0:00  0.36% sshd  1033 root          1   4    0  2580K  1700K sbwait   0:01  0.05% fetch   838 root          1  96    0 28200K 14520K select   0:10  0.00% Xorg   881 root          1  96    0 22812K 14720K select   0:02  0.00% Terminal   874 root          3  20    0 20260K 13108K kserel   0:02  0.00% xfdesktop   870 root          1  96    0 18968K 11424K select   0:01  0.00% xfce4-panel   976 root          1   8    0  1328K  1216K wait     0:01  0.00% make   903 root          1   8    0  1264K  1144K wait     0:01  0.00% make   868 root          1  96    0 13480K  9540K select   0:01  0.00% xfwm4   875 root          1  96    0 17496K 10120K select   0:00  0.00% xfce4-menu-pl   858 root          1  96    0 12380K  7324K select   0:00  0.00% xfce4-session   866 root          1  96    0 18472K  9104K select   0:00  0.00% xfce-mcs-mana   513 root          1  96    0  1340K   808K select   0:00  0.00% moused   872 root          1  96    0 12752K  8152K select   0:00  0.00% Thunar   883 root          1  20    0  4688K  2868K pause    0:00  0.00% csh   832 root          1  20    0  4688K  2800K pause    0:00  0.00% csh  1037 root          1  20    0  4824K  2936K pause    0:00  0.00% csh I'm including my Xorg.0.log file.  It appears that the VESA drivers are loading.  Is this correct?  I guess turning off the acceleration enables the probing to work. I will go ahead and add in the debug code you have asked for.  At least I have some usable graphics.  I'm using XFCE in high res mode, and I'm building other GUI apps to test. Later I will build and run Xperf and see what kind of results that shows. Thank you so much guys!  I hope I can contribute something! :) Larry Jordan Crouse wrote: <blockquote cite="mid20071220165511.GE25794@cosmic.amd.com" type="cite"> <pre wrap="">On 20/12/07 02:42 -0800, Larry Maloney wrote: </pre> <blockquote type="cite"> <pre wrap="">Sorry, I was wrong. The driver loads, I get an X back ground and the X cursor, but the session is frozen. I can't load any graphic applications inside X. </pre> </blockquote> <pre wrap=""> That is most likely because the GPU is stuck in an infinite loop. Easy way to check is to add option "NoAccel" "true" to the card section in your xorg.conf. If that is the case, then the problem is probably in the MSR code, which I feared might give us problems in BSD. put copious ErrorF messages through the LXReadMSR and LXWriteMSR functions - the GeodeReadMSR and LXWRiteMSR functions should return error on your machine (since BSD doesn't have /dev/cpu/0/msr), but the LX_MSR_READ and LX_MSR_WRITE functions should work. One thing to check is to make sure that the inline assembly is correct and that you have I/O permission in BSD (you should, but you never know). if there is a sane way to read MSRs without the VSA I/O on BSD, then lets put that into the GeodeReadMSR and GeodeWriteMSR functions bounded the appropriate #ifdefs, which is the more correct solution (as you can tell, I know nothing of the internal workings of BSD). The problem _might_ be somewhere else, but to be honest with you, every time this failure case has come up, then this has been the problem. Jordan </pre> </blockquote> <pre class="moz-signature" cols="72">-- ============================== Larry P. Maloney CEO/CTO SoftNet Solutions Phone: 650-669-1942 Web: <a class="moz-txt-link-abbreviated" href="http://www.softnetsolution.com">www.softnetsolution.com</a></pre> </body> </html>