Hey Nicolas, In working on my 2.6.37 based merge tree of the linaro and android kernels, I've been seeing a crash or hang at init time on my beagleboard xm.
I reproduced it with your linaro-2.6.37 tree, and I managed to bisect it down to the following commit:
ff8ea0c44f666883ef1e5545c33f858557b02043 ARM: 6384/1: Remove the domain switching on ARMv6k/v7 CPUs
Reverting this patch allows the linaro-2.6.37 kernel to boot on a linaro disk image, and also allows my dev/linaro.android branch to boot using an android disk image.
Is anyone else seeing this issue?
thanks -john
On Wed, 26 Jan 2011, john stultz wrote:
Hey Nicolas, In working on my 2.6.37 based merge tree of the linaro and android kernels, I've been seeing a crash or hang at init time on my beagleboard xm.
I reproduced it with your linaro-2.6.37 tree, and I managed to bisect it down to the following commit:
ff8ea0c44f666883ef1e5545c33f858557b02043 ARM: 6384/1: Remove the domain switching on ARMv6k/v7 CPUs
Reverting this patch allows the linaro-2.6.37 kernel to boot on a linaro disk image, and also allows my dev/linaro.android branch to boot using an android disk image.
Is anyone else seeing this issue?
Nope. And strangely enough, I don't see that commit SHA1 anywhere in all the linaro branches nor in mainline. Here, that commit tag line can be found in commit 247055aa21ffef1c49dd64710d5e94c2aee19b58 instead.
Are you sure your tree is genuine?
Nicolas
On Wed, 2011-01-26 at 22:46 -0500, Nicolas Pitre wrote:
On Wed, 26 Jan 2011, john stultz wrote:
Hey Nicolas, In working on my 2.6.37 based merge tree of the linaro and android kernels, I've been seeing a crash or hang at init time on my beagleboard xm.
I reproduced it with your linaro-2.6.37 tree, and I managed to bisect it down to the following commit:
ff8ea0c44f666883ef1e5545c33f858557b02043 ARM: 6384/1: Remove the domain switching on ARMv6k/v7 CPUs
Reverting this patch allows the linaro-2.6.37 kernel to boot on a linaro disk image, and also allows my dev/linaro.android branch to boot using an android disk image.
Is anyone else seeing this issue?
Nope. And strangely enough, I don't see that commit SHA1 anywhere in all the linaro branches nor in mainline. Here, that commit tag line can be found in commit 247055aa21ffef1c49dd64710d5e94c2aee19b58 instead.
Are you sure your tree is genuine?
Ah. Sorry for the confusion on that bit. 247055aa21ffef1c49dd64710d5e94c2aee19b58 is the change in question.
I was testing in an android environment, and needed the android patches to boot. Since I at first assumed it was some mis-merge caused by me of the linaro-2.6.37 tree into the android tree, I re-based the entire linaro-2.6.37 branch (yes, all 1000+ patches) on top of the android tree so I could bisect down the cause of the breakage (starting at a known good point with the android tree). The rebase changed the hash.
Only once I found it the problem and noticed it didn't seem to be connected to the android changes, did I end up testing with a linaro disk image, which also showed the problem.
Hmm. So you are not seeing the same issue? Even on a beagleboard xm? Its getting late here, so I'll reconfirm I see the problem with a base linaro image and kernel tomorrow and send you my config.
thanks -john
On Wed, 26 Jan 2011, john stultz wrote:
Only once I found it the problem and noticed it didn't seem to be connected to the android changes, did I end up testing with a linaro disk image, which also showed the problem.
Hmm. So you are not seeing the same issue? Even on a beagleboard xm? Its getting late here, so I'll reconfirm I see the problem with a base linaro image and kernel tomorrow and send you my config.
Yes, I would be highly interested in your config. I didn't test on the XM lately but this patch has been part of the Linaro kernel tree for months now and no one reported any issues related to it. It is also part of mainline now and no one seems to have trouble with it there either.
Nicolas
On Thu, 2011-01-27 at 09:21 -0500, Nicolas Pitre wrote:
On Wed, 26 Jan 2011, john stultz wrote:
Only once I found it the problem and noticed it didn't seem to be connected to the android changes, did I end up testing with a linaro disk image, which also showed the problem.
Hmm. So you are not seeing the same issue? Even on a beagleboard xm? Its getting late here, so I'll reconfirm I see the problem with a base linaro image and kernel tomorrow and send you my config.
Yes, I would be highly interested in your config. I didn't test on the XM lately but this patch has been part of the Linaro kernel tree for months now and no one reported any issues related to it. It is also part of mainline now and no one seems to have trouble with it there either.
Bah! Now I can't reproduce it with the linaro-2.6.37 tree either. I must have flubbed something in my testing (copied over the wrong uImage, most likely).
So maybe it is something in the android tree that's causing trouble with that patch. I'll keep digging.
Sorry for the noise then.
thanks -john
-
john stultz -
Nicolas Pitre