Re: [PATCH bpf-next 2/2] libbpf, xsk: add libbpf_smp_store_release libbpf_smp_load_acquire

["Björn Töpel" <bjorn.topel@gmail.com>]

On Tue, 2 Mar 2021 at 10:25, Daniel Borkmann <daniel@iogearbox.net> wrote:
>

[...]

> > I wonder if it's possible to cook a LKMM litmus test for this...?
>
> That would be amazing! :-)
>

With the help of Paul and Alan [1] (Thanks!) I've cooked 8 litmus
tests for this [2].

The litmus tests is based on a one entry ring-buffer, and there are
two scenarios. The ring is full, i.e. the producer has written an
entry, so the consumer has to go first. The ring is empty, i.e. the
producer has to go first. There is one test for each permutation:
barrier only, acqrel only, acqrel+barrier, barrier+acqrel.

According to these tests the code in this series is correct. Now, for
the v2 some more wording/explanations are needed. Do you think I
should include the litmus tests in the patch, or just refer to them?
Paste parts of them into the cover?

> (Another option which can be done independently could be to update [0] with outlining a
>   pairing scenario as we have here describing the forward/backward compatibility on the
>   barriers used, I think that would be quite useful as well.)
>
>    [0] Documentation/memory-barriers.txt
>

Yeah, I agree. There is some information on it though in the "SMP
BARRIER PAIRING" section:
--8<--
General barriers pair with each other, though they also pair with most
other types of barriers, albeit without multicopy atomicity.  An acquire
barrier pairs with a release barrier, but both may also pair with other
barriers, including of course general barriers.  A write barrier pairs
with a data dependency barrier, a control dependency, an acquire barrier,
a release barrier, a read barrier, or a general barrier.  Similarly a
read barrier, control dependency, or a data dependency barrier pairs
with a write barrier, an acquire barrier, a release barrier, or a
general barrier:
-->8--

And there's the tools/memory-model/Documentation/cheatsheet.txt.

That being said; In this case more is more. :-D


Björn

[1] https://lore.kernel.org/lkml/CAJ+HfNhxWFeKnn1aZw-YJmzpBuCaoeGkXXKn058GhY-6ZBDtZA@mail.gmail.com/
[2] https://github.com/bjoto/litmus-xsk/commit/0db0dc426a7e1248f83e21f10f9e840f970f4cb7

> >> Would also be great to get Will's ACK on that when you have a v2. :)
> >
> > Yup! :-)
> >
> >
> > Björn
> >
> >
> >> Thanks,
> >> Daniel
> >>
> >>> [1] https://lore.kernel.org/bpf/20200316184423.GA14143@willie-the-truck/
>