I think bero's workaround that Michael Hope suggested, pass -mno-unaligned-access to the kernel via KFLAGS may fix this. The change we made on Android was:
http://review.android.git.linaro.org/#patch,sidebyside,1285,1,tasks/kernel.m...
On 7 December 2011 23:05, john stultz johnstul@us.ibm.com wrote:
Zach reminded me that this month is compressed, so a linaro+android kernel would be needed immediately for 11.12. As Andrey is just ramping up in taking over for the Linaro Android kernel maintenance, I wanted to just get a kernel out, using the older kernel workflow, so that we had something current for 11.12.
Anyway. This is straight from Andy Green's linaro-androidization-tracking branch, with a few small build fixes added on that I found in my testing and the base android_*_defconfig files.
You can find the tree here: git://android.git.linaro.org/kernel/linaro-android.git linaro-android-3.2-agreen-rebase
The current sha is tagged as: linux-linaro-3.2-2011.12-0-android-0
Known issues: There seems to be something in the androidization branch that is causing problems on beagle xm and origen. In my testing beagle xm kernel ends up hanging in mid-boot(after ~4 seconds). And the orgien board doesn't show anything past "Uncompressing Linux... done, booting the kernel". If I drop the androidization patches and go back to the v3.2-rc4 base, both kernels boot until the Android userland environment starts and falls over because the android features are missing. I mucked about for awhile on both of these tonight, but wasn't able to solve either of them, so I'd appreciate any help trying to narrow down what is wrong on Origen (beagle is apparently lower priority).
Andy, one issue with the re-factored android patch tree: Its not very bisect-able. If I jump back to a topic branch, frequently there are missing dependencies that keep it from building. Any thoughts on how we can better chase down these sorts of issues?
thanks -john