Dear Artturi Alm, > On 07/02/13 19:28, minux wrote: > > On Mon, Jul 1, 2013 at 7:20 PM, Artturi Alm <artturi....@gmail.com> wrote: > >> So, i found out that the docs i have don't include > >> sd/mmc controller register definitions etc., only the overview. > >> which lead me to fix ehci and now it might boot, but > >> i cant figure out how to cross-build release, if it's even > > > > the cross-distrib target in Makefile.cross might help you (cross-gnu > > doesn't work, > > so getting a native toolchain cross compiled needs more work) > > > > alternatively, when i last tried to bring up the beagle port on > > Pandaboard ES, > > i used the lastest armish binary sets and although not all programs work > > (mainly > > due to using SWP on Cortex-A9) the system could boot into multi user > > successfully. > > > > after it boots up, you can rebuild /usr/src/lib/librthread, and replace > > the library > > so that perl runs and then you have a working system to do a full native > > build. > > > >> supported, and if not, what else can i do? > >> in short: i dont know how to build userland etc. to be placed > >> on the umass which attaches so i it would boot properly.. > >> i copied beagle/conf/RAMDISK into allwinner and it does build, > >> and is about 1mb bigger, but it panics on error opening disc. > >> and honestly, im not sure if it freezes now where it does because > >> there is no sd0b on that old stick, didnt bother with it since there's > >> no /bin etc. anyway, but userland is what i need i think :) > > Thanks, armish sets got it further, it does however still freeze quite > early, while running /etc/rc. > i was hoping to be able to stay within /allwinner, but now i think doing > so would mean waiting for devs to fix beagle on other boards than > Pandaboard ES, which i _guess_ to be pmap related issue atm.. > me trying to fix anything in arch/arm is like working blind, i have no > other boards than cubie to test on. > besides that, i have a cause of concern that this might be wasted effort > anyway... > > i spent some days again working on interrupt and timer handling to fix > errors from first revs that seemed to work without DEBUG, hiding the > errors away from dmesgs i've sent, now it doesn't print timeout delayed > warnings anymore nor > mutex held on enter panics or anything. > this made in china SoC has excellent docs avail, not. > now i don't wonder anymore that freebsd doesn't even try clearing > interrupts on A10, they enable/disable while masking/unmasking, not sure > yet about how linux does this, but i chose to use hwprios and response > register also, some hardcoded things still there to be cleaned..
Does it not use standard ARM GIC ? [...] Best regards, Marek Vasut
