Hi,
Just tested the imx51 Thumb-2 kernel on efikamx here -- it boots fine, but we only went as far as the initramfs.
Is anyone currently relying on CONFIG_LATENCYTOP? If so, we should investigate what the incompatibility is...
Cheers ---Dave
On Wed, Feb 23, 2011 at 6:11 PM, Dave Martin dave.martin@linaro.org wrote:
Hi there,
Talking with Loic just now, we thought it could be worth turning on Thumb-2 in the main omap kernel package after all, since we've at least seen it work on a few different boards now. It feels ready for wider testing.
Is it possible for you to move to a newer snapshot of the 2.6.38 linaro tree, or will that be problematic at this stage?
Since the imx51 kernel at least seems to build and boot in Thumb, it could be worth enabling a separate build in a PPA, but it's may not mature enough for everyone to use as the main kernel just yet. If you want to try this, you need CONFIG_ARM_UNWIND=y, CONFIG_LATENCYTOP=n and CONFIG_THUMB2_KERNEL=y
Cheers ---Dave
Dave,
Not sure if you noticed in the IRC logs so here is what happened for the packaged release.
Pulled latest linux-linaro-2.6.38 that has your patches.
Fixed arch/arm/mach-omap2/Makefile to allow smc instructions in arch/arm/mach-omap2/sleep34xx.S
Disabled CONFIG_BINFMT_AOUT.
The resulting kernel builds and boots but some modules have problems:
$ modprobe fat fat: unknown relocation: 102 FATAL: Error inserting vfat
So if we had deployed this kernel we would have images that are not upgradeable since we need access to the fat partition to install new uImage and uInitrd.
On the positive side we did get some testing and know there is at least one module relocation problem.
John
On Thu, Feb 24, 2011 at 4:40 AM, Dave Martin dave.martin@linaro.org wrote:
Hi,
Just tested the imx51 Thumb-2 kernel on efikamx here -- it boots fine, but we only went as far as the initramfs.
Is anyone currently relying on CONFIG_LATENCYTOP? If so, we should investigate what the incompatibility is...
Cheers ---Dave
On Wed, Feb 23, 2011 at 6:11 PM, Dave Martin dave.martin@linaro.org wrote:
Hi there,
Talking with Loic just now, we thought it could be worth turning on Thumb-2 in the main omap kernel package after all, since we've at least seen it work on a few different boards now. It feels ready for wider testing.
Is it possible for you to move to a newer snapshot of the 2.6.38 linaro tree, or will that be problematic at this stage?
Since the imx51 kernel at least seems to build and boot in Thumb, it could be worth enabling a separate build in a PPA, but it's may not mature enough for everyone to use as the main kernel just yet. If you want to try this, you need CONFIG_ARM_UNWIND=y, CONFIG_LATENCYTOP=n and CONFIG_THUMB2_KERNEL=y
Cheers ---Dave
On Thu, Feb 24, 2011 at 3:19 PM, John Rigby john.rigby@linaro.org wrote:
Dave,
Not sure if you noticed in the IRC logs so here is what happened for the packaged release.
Pulled latest linux-linaro-2.6.38 that has your patches.
Fixed arch/arm/mach-omap2/Makefile to allow smc instructions in arch/arm/mach-omap2/sleep34xx.S
Disabled CONFIG_BINFMT_AOUT.
The resulting kernel builds and boots but some modules have problems:
$ modprobe fat fat: unknown relocation: 102 FATAL: Error inserting vfat
So if we had deployed this kernel we would have images that are not upgradeable since we need access to the fat partition to install new uImage and uInitrd.
On the positive side we did get some testing and know there is at least one module relocation problem.
John
Thanks for trying it. I confess I hadn't tested modules much, since module support with Thumb-2 has been in the kernel for ages and is not arch-specific.
Can you give me a pointer to the exact config which was used, so I can try and reproduce it? Is there a build log?
IIUC, that relocation should not be ending up in modules: it's a short branch with range +/-2K, so fixup would rarely succeed.
Maybe there is a tools issue, or some missing processing step when .kos get generated...
Cheers ---Dave
Dave,
If you check out the omap-only branch of git://git.linaro.org/ubuntu/linux-linaro-natty.git.
Then fakeroot ./debian/rules clean fakeroot ./debian/rules editconfigs Turn on Thumb2
If you look in the output you will see the file in /tmp that was used to assemble the config.
If you want to build the packaged kernel then go follow along...
## Make a couple of changes in order to cross compile: ## turn off tools (actually do_tools is already off for omap) sed -i -e 's/do_tools.*=.*/do_tools = false/' debian.linaro/rules.d/armel.mk ## turn off dh_strip sed -i -e 's/dh_strip.*pkghdr.*/echo skipping dh_strip for cross build/' debian/rules.d/2-binary-arch.mk Then debuild: debuild -eCROSS_COMPILE=arm-linux-gnueabi- -nc -ns -aarmel -b
If you have any questions please ask.
John
On Fri, Feb 25, 2011 at 4:07 AM, Dave Martin dave.martin@linaro.org wrote:
On Thu, Feb 24, 2011 at 3:19 PM, John Rigby john.rigby@linaro.org wrote:
Dave,
Not sure if you noticed in the IRC logs so here is what happened for the packaged release.
Pulled latest linux-linaro-2.6.38 that has your patches.
Fixed arch/arm/mach-omap2/Makefile to allow smc instructions in arch/arm/mach-omap2/sleep34xx.S
Disabled CONFIG_BINFMT_AOUT.
The resulting kernel builds and boots but some modules have problems:
$ modprobe fat fat: unknown relocation: 102 FATAL: Error inserting vfat
So if we had deployed this kernel we would have images that are not upgradeable since we need access to the fat partition to install new uImage and uInitrd.
On the positive side we did get some testing and know there is at least one module relocation problem.
John
Thanks for trying it. I confess I hadn't tested modules much, since module support with Thumb-2 has been in the kernel for ages and is not arch-specific.
Can you give me a pointer to the exact config which was used, so I can try and reproduce it? Is there a build log?
IIUC, that relocation should not be ending up in modules: it's a short branch with range +/-2K, so fixup would rarely succeed.
Maybe there is a tools issue, or some missing processing step when .kos get generated...
Cheers ---Dave
On Thu, 24 Feb 2011, John Rigby wrote:
The resulting kernel builds and boots but some modules have problems:
$ modprobe fat fat: unknown relocation: 102 FATAL: Error inserting vfat
A workaround for what appears to be a binutils bug has been merged in linaro-2.6.38. So the Thumb2 kernel testing may resume on trusted targets.
Nicolas
Hi,
On Mon, Feb 28, 2011 at 9:19 PM, Nicolas Pitre nicolas.pitre@linaro.org wrote:
On Thu, 24 Feb 2011, John Rigby wrote:
The resulting kernel builds and boots but some modules have problems:
$ modprobe fat fat: unknown relocation: 102 FATAL: Error inserting vfat
A workaround for what appears to be a binutils bug has been merged in linaro-2.6.38. So the Thumb2 kernel testing may resume on trusted targets.
Thanks for merging it.
It's a bit ugly to include turn off compiler optimisations to work around this though, so we might encounter upstream to that patch.
In any case, we still need someone to take a look at the possible tools issue -- CC'ing linaro-toolchain in case people aren't aware:
https://bugs.launchpad.net/binutils-linaro/+bug/725126
Cheers ---Dave
On Wed, Mar 2, 2011 at 12:35 AM, Dave Martin dave.martin@linaro.org wrote:
Hi,
On Mon, Feb 28, 2011 at 9:19 PM, Nicolas Pitre nicolas.pitre@linaro.org wrote:
On Thu, 24 Feb 2011, John Rigby wrote:
The resulting kernel builds and boots but some modules have problems:
$ modprobe fat fat: unknown relocation: 102 FATAL: Error inserting vfat
A workaround for what appears to be a binutils bug has been merged in linaro-2.6.38. So the Thumb2 kernel testing may resume on trusted targets.
Thanks for merging it.
It's a bit ugly to include turn off compiler optimisations to work around this though, so we might encounter upstream to that patch.
In any case, we still need someone to take a look at the possible tools issue -- CC'ing linaro-toolchain in case people aren't aware:
Hi Dave. We're not working on binutils at the moment. There's a binutils-linaro project but it was created a bit too enthusiastically. Could you make sure this is reported upstream?
-- Michael
-
Dave Martin -
John Rigby -
Loïc Minier -
Michael Hope -
Nicolas Pitre