Re: [PATCH bpf] bpf: zero-fill re-used per-cpu map element

From: David Verbeiren <david.verbeiren@tessares.net>
To: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Cc: bpf <bpf@vger.kernel.org>, Networking <netdev@vger.kernel.org>,
	Matthieu Baerts <matthieu.baerts@tessares.net>
Subject: Re: [PATCH bpf] bpf: zero-fill re-used per-cpu map element
Date: Tue, 27 Oct 2020 15:48:28 +0100	[thread overview]
Message-ID: <CAHzPrnF0yZY8rk6_qMS55_=gLCKwHq1s7LaRtSqGy823gtwLMA@mail.gmail.com> (raw)
In-Reply-To: <CAEf4BzZaJaYw0tB0R+q3qoQX7=qy3T9jvzf5q=TH++t66wNd-w@mail.gmail.com>

On Mon, Oct 26, 2020 at 11:48 PM Andrii Nakryiko
<andrii.nakryiko@gmail.com> wrote:
>
> On Fri, Oct 23, 2020 at 8:48 AM David Verbeiren
> <david.verbeiren@tessares.net> wrote:
> > [...]
> > +                       if (!onallcpus)
> > +                               for_each_possible_cpu(cpu)
> > +                                       memset((void *)per_cpu_ptr(pptr, cpu),
> > +                                              0, size);
>
> Technically, you don't have to memset() for the current CPU, right?
> Don't know if extra check is cheaper than avoiding one memset() call,
> though.

I thought about that as well but, because it depends on the 'size',
I decided to keep it simple. However, taking into account your other
comments, I think there is a possibility to combine it all nicely in a
separate function.

> But regardless, this 6 level nesting looks pretty bad, maybe move the
> for_each_possible_cpu() loop into a helper function?
>
> Also, does the per-CPU LRU hashmap need the same treatment?
I think it does. Good catch!

Thanks for your feedback. v2 is coming.