Re: [PATCH] percpu-refcount: Use normal instead of RCU-sched"

[Dennis Zhou <dennis@kernel.org>]

On Thu, Nov 07, 2019 at 06:24:34PM +0100, Sebastian Andrzej Siewior wrote:
> On 2019-11-07 11:55:19 [-0500], Dennis Zhou wrote:
> > On Thu, Nov 07, 2019 at 05:28:42PM +0100, Sebastian Andrzej Siewior wrote:
> > > > I just want to clarify a little bit. Is this patch aimed at fixing an
> > > > issue with RT kernels specifically? 
> > > 
> > > Due to the implications of preempt_disable() on RT kernels it fixes
> > > problems with RT kernels.
> > > 
> > 
> > Great, do you mind adding this explanation with what the implications
> > are in the commit message?
> 
> some RCU section here invoke callbacks which acquire spinlock_t locks.
> This does not work on RT with disabled preemption.
> 

Yeah, so adding a bit in the commit message about why it's an issue for
RT kernels with disabled preemption as I don't believe this is an issue
for non-RT kernels.


> > > > It'd also be nice to have the
> > > > numbers as well as if the kernel was RT or non-RT.
> > > 
> > > The benchmark was done on a CONFIG_PREEMPT kernel. As said in the commit
> > > log, the numbers were mostly the same, I can re-run the test and post
> > > numbers if you want them.
> > > This patch makes no difference on PREEMPT_NONE or PREEMPT_VOLUNTARY
> > > kernels.
> > > 
> > 
> > I think a more explicit explanation in the commit message would suffice.
> 
> What do you mean by "more explicit explanation"? The part with the
> numbers or that it makes no difference for PREEMPT_NONE and
> PREEMPT_VOLUNTARY?
> 

I just meant the above, the benchmarking is fine.

Thanks,
Dennis