Sorry for being late, my router / mail server has hardware problems. :-(
A bit offtopic: I am looking for a forum/mailing list/people who are work with this processor, especially with its 2d video accel on linux. I have problems, NSC didn't respond, and I stucked... :-(
you can split a list off for geode, but we're interested here, and I would welcome a discussion on this list -- it's not a high traffic list.
I know, it's a refreshment after lkml, or, say, comp.lang.tcl. :-)
So, my problem is not related to linuxbios (maybe there will be one time, if we decide to use this processor for our product).
I have an Advantech PCM-5820 board, with GX1 and CS5530, and I play with it to test it's hardware bitblt feature. I have patched the 2.4.17 kernel with NSC's geodefb driver, compiled a user-space GAL lib. I wrote a program to move a part of the screen, the usleep a short interval (aka scrolling a region). I know that this is not nice as the GX1 spec said that I should not use the fb while the bitblt engine is working, and because I have fb console, the console writes into it (it do not know about the bitblt engine), but there was no problem, the center of the screen scrolled like a dream, just when I typed something, there were 1-1 pixel errors.
The problem was when I decided to check the memory bandwidth, and issued a dd if=/dev/mem of=/dev/null bs=1024 count=16384 command. Sometimes (about every 2nd time) it completely freezes the machine. No oops, no kernel panic, just freeze to death. :-( And I do not understand what is going on, because if I do not scroll, there is no such problem, and the bitblt operation just means to write some values into some (processor-handled) IO registers.
What could be the problem? How to debug? What 'dd if=/dev/mem ...' exactly do? Many questions, and I have not answers. What a philosophical work is ours... :-)
Thanks: Circum