anarsoul's blog

I spent some days getting qemu to emulate some Z2 hardware, and got some results :) Almost all hw is emulated, it lacks only voltage regulator and WiFi, but it's not easy to implement Marvell's WiFi chip, because there's no datasheet, and anyway it makes no sense. Also screen rotation is not supported (yet). Stock bootloader - blob - (with slight modification) and stock kernel works. Blob puts Z2 into deepsleep on reset, so I hacked it not to do this. How to compile qemu: git clone git://github.com/anarsoul/qemu-z2.git cd qemu-z2 ./configure --audio-drv-list=alsa --target-list=arm-softmmu make -j4 arm-softmmu/qemu-system-arm binary is what you need. So, how to hack blob: put your blob image (first 64kb of NOR) somewhere convert it to hex representation with xxd: xxd mtdblock1.bin >mtdblock1.hex apply patch onto mtdblock1.hex convert hex back to bin: mv mtdblock1.bin mtdblock1.bin.bak; xxd -r mtdblock1.hex >mtdblock1.bin I don't pu...

Some days ago I made a driver (named pxa270-voltage) to set cpu core voltage according to its frequency, it should improve battery life (but I did not perform testing). freq-voltage table was taken from cpufreq-2xx driver. If anyone can test and report if it really improves battery life - it would be nice. Here's new patchset, precompiled binaries can be taken here , patches still should be applied ontop of Marex's tree Also new patchset fixes race condition in pxa2xx-spi suspend handler. P.S. This driver is just temporary solution until someone fix cpufreq on pxa270.

Recently I did some stuff for Z2 kernel: fixed bug in libertas (already submitted patches upstream) added proper suspend/resume support for libertas fixed wrong coefficient to convert adc value to battery voltage fixed bug in Z2 ASoC driver (with detecting Z2 soundcard) make Z2 resume only on power button and AC plug Patches agains Marex's tree you can find here . cpufreq and cpu voltage regulator do not work, so don't enable them in kernel config.

Last week I was busy getting PXA27x overlay working, as it was pretty broken in mainline kernel. There were two problems: overlay driver somehow corrupted memory on second access to it, and actually it does not work. So I got following messages on dmesg, and of course no visible overlay plane: [ 93.679574] overlay1fb_disable: timeout disabling overlay1 [ 95.601537] BUG: Bad page state in process sh pfn:a1b60 [ 95.601645] page:c0456c00 count:0 mapcount:0 mapping: (null) index:0x0 [ 95.601698] page flags: 0x200(arch_1) First bug was successfully fixed by Russel King, here you can find his patch. It fixed memory corruption, but overlay still did not work. Problem was in selecting planes Z-ordering in wrong time, so overlay was under main plane, and main plane has no transparency bit (rgb565 colorspace), so there was no way to make it visible. I fixed this issue, added some cosmetical fixes, and produced patch . To use 2nd overlay (one with yuv colospace available) I created mp...