[NOTES] Kernel WG Weekly Status report for week ending 2011-08-26
dave.martin at linaro.org
Fri Aug 26 15:50:05 UTC 2011
On Fri, Aug 26, 2011 at 3:29 PM, Christian Robottom Reis
<kiko at linaro.org> wrote:
> On Fri, Aug 26, 2011 at 09:07:44AM -0500, Mounir Bsaibes wrote:
>> * Integrated patch set for Qualcomm Hexagon architecture
> Interesting. What exactly is this?
>> * The v6/v7 single kernel patches are now accepted by Russell; these should
>> allow linaro filesystems to work on omap2plus_defconfig kernels
> Can you reword this? I'm having a hard time understanding exactly what
> happened or what the outcome was here.
Previously, if you built a single kernel supporting v6 and v7
platforms, then some Thumb instructions (particular VFP/NEON) could
cause programs to die with an Illegal Instruction signal. This hits
all linaro filesystems, since they are built in Thumb-2 with VFP --
just booting would spray SIGILLs all over the console.
The patches allow appropriate handling of those instructions based on
the CPU the kernel is actually running on, instead of it being
hard-wired based on the minimum architecture the kernel was built for.
That allows some standard reference configurations used by upsteam
(e.g. omap2plus_defconfig) to be used with linaro filesystems. It
removes one obstacle to building a single kernel image across whole
SoC families, which should be of interest to SoC vendors. It's also
useful for kernel developers: kernel changes can be better tested
against upsteam, while still using the linaro filesystems.
Does that answer the question?
More information about the linaro-dev