Re: [PATCH v2 1/6] sched_clock: Expose struct clock_read_data

[Ahmed S. Darwish]

On Wed, Jul 15, 2020 at 10:14:43AM +0200, peterz@xxxxxxxxxxxxx wrote:
> On Wed, Jul 15, 2020 at 10:12:22AM +0200, Peter Zijlstra wrote:
> > On Wed, Jul 15, 2020 at 07:56:50AM +0200, Ahmed S. Darwish wrote:
> > > On Wed, Jul 15, 2020 at 10:05:07AM +0800, Leo Yan wrote:
> > > > From: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> > > >
> > > ...
> > > >
> > > > Provide struct clock_read_data and two (seqcount) helpers so that
> > > > architectures (arm64 in specific) can expose the numbers to userspace.
> > > >
> > > ...
> > > >
> > > > +struct clock_read_data *sched_clock_read_begin(unsigned int *seq)
> > > > +{
> > > > +	*seq = raw_read_seqcount(&cd.seq);
> > > > +	return cd.read_data + (*seq & 1);
> > > > +}
> > > > +
> > > ...
> > >
> > > Hmm, this seqcount_t is actually a latch seqcount. I know the original
> > > code also used raw_read_seqcount(), but while at it, let's use the
> > > proper read API for seqcount_t latchers: raw_read_seqcount_latch().
> > >
> > > raw_read_seqcount_latch() has no read memory barrier though, and a
> > > suspicious claim that READ_ONCE() pairs with an smp_wmb() (??). But if
> > > its implementation is wrong, let's fix it there instead.
> >
> > It's supposed to be a dependent load, so READ_ONCE() is sufficient.
> > Except, of course, the C standard has other ideas, so a compiler is
> > allowed to wreck that, but they mostly don't :-)
>
> Also see:
>
>   https://lkml.kernel.org/r/20200625085745.GD117543@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Oh, spot on.

Can we then please replace the raw_read_seqcount(), in the original
patch which started this discussion, with raw_read_seqcount_latch()?

I see that you already did something similar for timekeeping.c:
7fc26327b756 ("seqlock: Introduce raw_read_seqcount_latch()").

Confession time: I have an internal patch series which introduces
seqcount_latch_t. Having a standardized set of accessors for the
seqcount latch read and write paths will make everything much more
streamlined :-)

Thanks,

--
Ahmed S. Darwish
Linutronix GmbH