A random collection of notes on the Angstrom builds for Beagleboard
The revC MLO does not work correctly from NAND on revB boards and will cause the boot process to fail unless the user button is pressed (to cause booting from MMC) during the boot process.
Older uboots display a beagleboard splash. These are outdated and broken.
Koen is responsible for the distribution at http://www.angstrom-distribution.org/demo/beagleboard/.
USR0: heartbeat, driven by the Kernel. However, it is possible for the board to fail to boot yet the heartbeat to continue (for instance the "pixel clock problem").
USR1: MMC IO activity monitor.
Getting s-video working is a bit of a black art at the moment. Here is the script that tomba wrote that sets up svideo on fb2 (this requires the latest kernels with DSS2 support, such as the koen distribution).
ovl2=/sys/devices/platform/omapdss/overlay2 # check that display1 is really TV! tv=/sys/devices/platform/omapdss/display1 fb2=/sys/class/graphics/fb2 tvw=`cat $tv/timings | cut -d "," -f 2 | cut -d "/" -f 1` tvh=`cat $tv/timings | cut -d "," -f 3 | cut -d "/" -f 1` mem=$((tvw*tvh*4)) echo "0" > $tv/enabled echo "0" > $ovl2/enabled echo "" > $ovl2/manager # allocate memory for framebuffer echo $mem > $fb2/size fbset -fb /dev/fb2 -xres $tvw -vxres $tvw -yres $tvh -vyres $tvh -depth 32 echo "0,0" > $ovl2/position echo "$tvw,$tvh" > $ovl2/output_size echo "tv" > $ovl2/manager echo "1" > $ovl2/enabled echo "1" > $tv/enabled # output random pixels to tv cat /dev/urandom > /dev/fb2
For some reason, udev didn't create the devices for me. Check like so:
ls -l /dev/fb*
You should see:
lrwxrwxrwx 1 root root 3 Jun 13 2009 /dev/fb -> fb0 crw-rw---- 1 root video 29, 0 Jun 13 2009 /dev/fb0 crw-rw---- 1 root video 29, 1 Jun 13 2009 /dev/fb1 crw-rw---- 1 root video 29, 2 Jun 13 2009 /dev/fb2
If you only get two lines back issue the following commands as root:
mknod /dev/fb1 c 29 1 mknod /dev/fb2 c 29 2 chgrp video /dev/fb1 /dev/fb2 chmod 660 /dev/fb1 /dev/fb2
Various mailing list posts suggest a bootargs similar to below, although it didn't seem to make a difference for me. Your mileage may vary.
bootargs 'console=ttyS2,115200n8 root=/dev/mmcblk0p2 noinitrd omapfb.video_mode=720x576@50 init=/init rootfstype=ext3 rw rootdelay=2 video=omapfb:vram:2M,vram:4M,vram:4M omapdss.def_disp=dvi omapfb.mode=dvi:1024x768MR-24@60 omapfb.vram=0:4M,1:2M,2:4M'
DSS2 is the display driver that has been virtually rewritten. It includes a sysfs interface at /sys/class/grpahics and /sys/driver/platform/omapdss. See for instance  for more info - or just search the google mailing list for DSS2.
To test video playback, mplayer seems to work well.
wget http://www.beagleboard.org/uploads/HARRY.YUV mplayer -vo fbdev:/dev/fb2 HARRY.YUV
Or for a longer, more demanding test, try big buck bunny:
wget http://download.blender.org/peach/bigbuckbunny_movies/big_buck_bunny_480p_surround-fix.avi mplayer -nosound big_buck_bunny_480p_surround-fix.avi
wget ... mplayer -ao alsa -lavdopts lowres=1:fast -vo fbdev:/dev/fb2 big_buck_bunny_480p_surround-fix.avi
I'm fairly sure that mplayer is doing the video scaling in software. If it is possible to get it to use hardware scaling, the results should be msuch better.
Next is to investigate the gstreamer_openmax stuff that reportedly can playback big buck bunny using only 10% of the CPU.
James is an alternative build that is being used for the Beagleboard James project. Some developers say it has better display and devices support than the Koen distribution.
NOTE: James is based upon a snapshot of Koen's distribution and is not a separate distribution. Apparently there is some regression along the line. Frankly speaking James is in need of an update, but the author of James is too busy on other issues at the moment.