[PATCH v1] powerpc64le: Revert "Optimized strcmp for power10"

[Sachin Monga]

Hey Paul

Thanks for sharing. I'll go through it.


As Carlos told, our first priority is to fix the CVE by reverting.

We can take the improvement and rework a bit easy with the 2nd phase.


Regards:
Sachin.


On 17/06/25 2:29 am, Carlos O'Donell wrote:
> On 6/16/25 4:19 PM, Paul Murphy wrote:
>> Sachin,
>>
>> Here is a patch which I think fixes the various vector register ABI
>> violations for the power10 optimizations. The fix should be
>> straightforward. I tested this on a P10 machine and didn't see any new
>> failures. It's been a few years since I've posted a patch here; I
>> recommend double checking my work. I missed this when I reviewed it.
>
> Paul, Thank you for posting this!
>
> Please post this in a distinct thread for review if you're looking
> to contribute the patch upstream.
>
> https://sourceware.org/glibc/wiki/Contribution%20checklist
>
> As part of the glibc security team I'm going to go ahead with the
> revert since I'd like the CVE fixed ASAP.
>
> Sachin and I spoke briefly this morning in the glibc upstream patch
> queue review meeting and we agreed to continue down the path of
> reverting the problematic code.
>
> Please feel free to work with Sachin getting improved and fixed
> versions posted, reviewed and accepted. We can do this with more time
> once the CVE is resolved.
>
> I wonder if we cant get a generic test framework for testing if
> caller preserved registers are actually preserved? :-)
>