Re: [ping][PATCH][s390][ppc] Add REQUEUE_PI support

["Ryan S. Arnold" <ryan dot arnold at gmail dot com>]

On Mon, Jan 14, 2013 at 9:27 AM, Siddhesh Poyarekar <siddhesh@redhat.com> wrote:
> Ping!
>
> On Tue, Jan 08, 2013 at 06:57:02PM +0530, Siddhesh Poyarekar wrote:
>> Hi,
>>
>> Support for priority inheritance in pthread condition variables using
>> FUTEX_WAIT_REQUEUE_PI and FUTEX_CMP_REQUEUE_PI have been in x86 code
>> since some time, but the generic code was never updated to make this
>> available for other architectures.  Attached patch modifies the
>> generic code to make use of FUTEX_*REQUEUE_PI for condition variables
>> backed by PI-aware mutexes whenever the underlying architecture
>> supports it and also adds support for powerpc and s390.  I have
>> verified that there are no regressions in the testsuite due to this
>> patch on s390x and ppc64.  Once this goes in, other architectures can
>> add this support by simply defining lll_futex_wait_requeue_pi,
>> lll_futex_timed_wait_requeue_pi and lll_futex_cmp_requeue_pi, similar
>> to the way I have done in s390 and ppc and compiling against a kernel
>> newer than 2.6.31.  I'll send out a separate email on ports once the
>> patch is in.

I don't really understand this code base well enough to give you a
worthwhile review.

Does the test-suite cover all of the corner cases that this code exposes?

In principle I don't see a problem with the code, and I don't want to
be overly conservative.  I think it can go in if there are no
regressions and I'll have my team try to stress the mechanism a bit.

Ryan