Received-SPF: pass (google.com: domain of linux-kernel+bounces-135741-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240408-hid-bpf-sleepable-v6-0-0499ddd91b94@kernel.org>
 <20240408-hid-bpf-sleepable-v6-1-0499ddd91b94@kernel.org> <65c249a6af45bfa5fe0f6c2331dcc1771a6f0b05.camel@gmail.com>
In-Reply-To: <65c249a6af45bfa5fe0f6c2331dcc1771a6f0b05.camel@gmail.com>
From: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Date: Mon, 8 Apr 2024 19:20:26 +0200
Message-ID: <CAO-hwJ+0erX3iJcOh9KBG3f01UiYvGk_Gx+-zyFc4Vb5LCcHxA@mail.gmail.com>
Subject: Re: [PATCH RFC bpf-next v6 1/6] bpf/helpers: introduce sleepable bpf_timers
To: Eduard Zingerman <eddyz87@gmail.com>
Cc: Benjamin Tissoires <bentiss@kernel.org>, Alexei Starovoitov <ast@kernel.org>, 
	Daniel Borkmann <daniel@iogearbox.net>, Andrii Nakryiko <andrii@kernel.org>, 
	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>, Mykola Lysenko <mykolal@fb.com>, Shuah Khan <shuah@kernel.org>, 
	bpf@vger.kernel.org, linux-kernel@vger.kernel.org, 
	linux-kselftest@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Mon, Apr 8, 2024 at 7:08=E2=80=AFPM Eduard Zingerman <eddyz87@gmail.com>=
 wrote:
>
> On Mon, 2024-04-08 at 10:09 +0200, Benjamin Tissoires wrote:
>
> [...]
>
> > diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
> > index 9234174ccb21..fd05d4358b31 100644
> > --- a/kernel/bpf/helpers.c
> > +++ b/kernel/bpf/helpers.c
> > @@ -1096,12 +1096,19 @@ const struct bpf_func_proto bpf_snprintf_proto =
=3D {
> >   * freeing the timers when inner map is replaced or deleted by user sp=
ace.
> >   */
> >  struct bpf_hrtimer {
> > -     struct hrtimer timer;
> > +     union {
> > +             struct hrtimer timer;
> > +             struct work_struct work;
> > +     };
> >       struct bpf_map *map;
> >       struct bpf_prog *prog;
> >       void __rcu *callback_fn;
> >       void *value;
> > -     struct rcu_head rcu;
> > +     union {
> > +             struct rcu_head rcu;
> > +             struct work_struct sync_work;
>
> Nit:
> I find this name very confusing, the field is used to cancel timer
> execution, is it a convention to call such things '...sync...'?
>
> > +     };
> > +     u64 flags;
> >  };
> >
>
> [...]
>
> > +static void bpf_timer_sync_work_cb(struct work_struct *work)
> > +{
> > +     struct bpf_hrtimer *t =3D container_of(work, struct bpf_hrtimer, =
sync_work);
> > +
> > +     cancel_work_sync(&t->work);
> > +
> > +     kfree_rcu(t, rcu);
>
> Sorry, I might be wrong, but this looks suspicious.
> The 'rcu' field of 'bpf_hrtimer' is defined as follows:
>
> struct bpf_hrtimer {
>         ...
>         union {
>                 struct rcu_head rcu;
>                 struct work_struct sync_work;
>         };
>         ...
> };
>
> And for sleepable timers the 'sync_work' field is set as follows:
>
> BPF_CALL_3(bpf_timer_init, struct bpf_timer_kern *, timer, struct bpf_map=
 *, map,
>            u64, flags)
> {
>         ...
>         INIT_WORK(&t->sync_work, bpf_timer_sync_work_cb);
>         ...
> }
>
> So, it looks like 'kfree_rcu' would be called for a non-rcu pointer.

That was my initial assumption too, but Alexei told me it was fine.
And I think he is correct because kfree_rcu doesn't need the rcu_head
to be initialized.

So in the end, we initialize the memory as a work_struct, and when
that work kicks in, we reuse that exact same memory as the rcu_head.
This is fine because that work will never be reused.

If I understand correctly, this is to save a few bytes as this is a
critical struct used in programs with a high rate usage, and every
byte counts.

Cheers,
Benjamin

>
> > +}
> > +
>
>