Received-SPF: pass (google.com: domain of linux-kernel+bounces-128471-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Precedence: bulk
MIME-Version: 1.0
References: <20240401073159.16668-1-andrea.righi@canonical.com> <20240401073159.16668-2-andrea.righi@canonical.com>
In-Reply-To: <20240401073159.16668-2-andrea.righi@canonical.com>
From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Date: Tue, 2 Apr 2024 10:58:33 -0700
Message-ID: <CAEf4BzbJE-Y72VqYXCDfwRXjhknVK+GHYjRfxdajzAEwjUkwUg@mail.gmail.com>
Subject: Re: [PATCH 1/2] libbpf: ringbuf: allow to consume up to a certain
 amount of items
To: Andrea Righi <andrea.righi@canonical.com>
Cc: Andrii Nakryiko <andrii@kernel.org>, Eduard Zingerman <eddyz87@gmail.com>, 
	Alexei Starovoitov <ast@kernel.org>, Daniel Borkmann <daniel@iogearbox.net>, 
	Martin KaFai Lau <martin.lau@linux.dev>, Song Liu <song@kernel.org>, 
	Yonghong Song <yonghong.song@linux.dev>, John Fastabend <john.fastabend@gmail.com>, 
	KP Singh <kpsingh@kernel.org>, Stanislav Fomichev <sdf@google.com>, Hao Luo <haoluo@google.com>, 
	Jiri Olsa <jolsa@kernel.org>, David Vernet <void@manifault.com>, Tejun Heo <tj@kernel.org>, 
	bpf@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Mon, Apr 1, 2024 at 12:32=E2=80=AFAM Andrea Righi <andrea.righi@canonica=
l.com> wrote:
>
> In some cases, instead of always consuming all items from ring buffers
> in a greedy way, we may want to consume up to a certain amount of items,
> for example when we need to copy items from the BPF ring buffer to a
> limited user buffer.
>
> This change allows to set an upper limit to the amount of items consumed
> from one or more ring buffers.
>
> Link: https://lore.kernel.org/lkml/20240310154726.734289-1-andrea.righi@c=
anonical.com/T
> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
> ---
>  tools/lib/bpf/ringbuf.c | 29 +++++++++++++++++------------
>  1 file changed, 17 insertions(+), 12 deletions(-)
>
> diff --git a/tools/lib/bpf/ringbuf.c b/tools/lib/bpf/ringbuf.c
> index aacb64278a01..81df535040d1 100644
> --- a/tools/lib/bpf/ringbuf.c
> +++ b/tools/lib/bpf/ringbuf.c
> @@ -231,7 +231,7 @@ static inline int roundup_len(__u32 len)
>         return (len + 7) / 8 * 8;
>  }
>
> -static int64_t ringbuf_process_ring(struct ring *r)
> +static int64_t ringbuf_process_ring(struct ring *r, int64_t max_items)
>  {
>         int *len_ptr, len, err;
>         /* 64-bit to avoid overflow in case of extreme application behavi=
or */
> @@ -264,7 +264,14 @@ static int64_t ringbuf_process_ring(struct ring *r)
>                                                           cons_pos);
>                                         return err;
>                                 }
> -                               cnt++;
> +                               if (++cnt >=3D max_items) {
> +                                       /* update consumer pos and return=
 the
> +                                        * total amount of items consumed=
.
> +                                        */
> +                                       smp_store_release(r->consumer_pos=
,
> +                                                         cons_pos);

Does this fit on a single line under 100 characters? If yes, please
keep it as a single line

but actually it seems cleaner to keep cnt++ intact, let
smp_store_release() below happen, and then check the exit condition.
Were you afraid to do unnecessary checks on discarded samples? I
wouldn't worry about that.

> +                                       goto done;
> +                               }
>                         }
>
>                         smp_store_release(r->consumer_pos, cons_pos);
> @@ -281,19 +288,18 @@ static int64_t ringbuf_process_ring(struct ring *r)
>   */
>  int ring_buffer__consume(struct ring_buffer *rb)
>  {
> -       int64_t err, res =3D 0;
> +       int64_t err, res =3D 0, max_items =3D INT_MAX;

I'm wondering if it might be better to have a convention that zero
means "no limit", which might allow the compiler to eliminate the
condition in ringbuf_process_ring altogether due to constant
propagation. WDYT?

>         int i;
>
>         for (i =3D 0; i < rb->ring_cnt; i++) {
>                 struct ring *ring =3D rb->rings[i];
>
> -               err =3D ringbuf_process_ring(ring);
> +               err =3D ringbuf_process_ring(ring, max_items);
>                 if (err < 0)
>                         return libbpf_err(err);
>                 res +=3D err;
> +               max_items -=3D err;
>         }
> -       if (res > INT_MAX)
> -               return INT_MAX;
>         return res;
>  }

[...]