On Wed, 2014-06-11 at 15:51 +0100, Basil Eljuse wrote:
Hi Chris,
Yes this happens only with IKS runs as I can see from the detailed logs. Tixy, can you confirm if your config was also IKS?
It has HMP active and IKS support compiled in but not active, which should be same config as Linaro's LSK Android releases are made with, except I also had CONFIG_PROVE_LOCKING enabled.
I've just compiled without IKS and see the same problem.
I also tried again without the 2 new patches sent today and don't get the lockdep warning, and everything seems consistently reproducible to me.
Not sure which Android file system I have, think its a week or so old one from the CI jobs for the tip kernel (but only difference to LSK builds should be the kernel anyway, and I'm building my own one of those).
The kernel source I'm using is the latest lsk android branch: https://git.linaro.org/kernel/linux-linaro-stable.git/commit/ddb020c86c1de65... with the two new MP patches applied.
And the kernel config I'm using is that generated by running
ARCH=arm scripts/kconfig/merge_config.sh \ linaro/configs/linaro-base.conf \ linaro/configs/android.conf \ linaro/configs/big-LITTLE-MP.conf \ linaro/configs/big-LITTLE-IKS.conf \ linaro/configs/vexpress.conf \ linaro/configs/debug.conf
I've attached my boot logs, including the output of "cat /proc/config.gz" so we can see for definite the config used to build the kernel.
My personal scripts also set CONFIG_GATOR to =y rather than =m which is shown in these logs, but I tried again without that and see the same results.