End of January, another kernel update. This should fix DVI problems for a lot of people and also brings in some ARMv7 patches from mainline
- 2.6.31.12
- Updated to Freescale BSP 5.2.0 (lots of non-iMX515 stuff here too)
- Disabled SPDIF, accidentally enabled it in last release (doesn't work yet, use I2S instead)
- DVI sink support, should work better on DVI monitors now
- Better mode culling and use of the best monitor mode
- vmode= flag works again
- Patches from Pegatron
- DVI sink support (above) and CEA descriptor detection
- Altogether better handling of the SII9022 (still not up to spec though)
- PMIC reboot disabled (use mxc_reboot instead) should fix the shutdown-is-reboot problem
- max_res= flag will make the system use the biggest mode, 1280x720 or 1024x768 as fallbacks
- removed egregious hack from fbmem.c to adjust clocks and put it in a better place
- ARM tree patches
- remove that annoying migrate_pages warning on compile
- updates for ARMv6/ARMv7 prefetch abort handling and MMU handling
- updates to syscalls
- remove annoying unused variable warning
- ramzswap driver
- pulled from staging for 2.6.32
- updated ARM dcache handling if dirty bit set
- updated to compcache 0.6.2
The update is shipped as a patch against the 2.6.31 tree released in December (not mainline!), and supercedes the 2.6.31.9 patchset. Just extract the 2.6.31 tarball and then extract the patch tarball inside it and run patches/patch-kernel.sh (just like a Freescale BSP :)
You can find them at the Efika MX Kernel Platform Support page. All the problems reported by beta testers (DVI problems still, and the power off fix not working) are fixed.
My 1024x768 DVI display still does not work (see powerdeveloper.org for dmesg). "halt" still reboots the machine. There is also good news: sound quality got a lot better compared to previous releases.
ReplyDeleteYes the audio seems a little bit better for some reason. I saw a LOT of weird exception aborts when playing audio (some which were unhandled) and the new patches from 2.6.32/33 seem to have made them a lot better.
ReplyDeleteAnyone who wants to play with ramzswap, can do; just follow the docs (download from http://compcache.googlecode.com and read the README - Nitin graciously implemented a feature suggestion so you don't need any special tools to mount the very basic driver, which will compress swap in memory and to disk with module arguments again, just like old ramzswap used to :)
steev@gentoo noticed that there is a patch missing in the patch archive (and therefore I missed it for the release). Just rename the 933 to .patch and run the patch script or apply it manually and it'll apply cleanly. I don't think it's truly important, but it was an avenue of investigation for some problems I was having and is probably an important fix for someone :D