Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230509163050.127d5123@rorschach.local.home> <20230515165707.hv65ekwp2djkjj5i@MacBook-Pro-8.local>
 <20230515192407.GA85@W11-BEAU-MD.localdomain> <20230517003628.aqqlvmzffj7fzzoj@MacBook-Pro-8.local>
 <CAHk-=whBKoovtifU2eCeyuBBee-QMcbxdXDLv0mu0k2DgxiaOw@mail.gmail.com>
 <CAHk-=wj1hh=ZUriY9pVFvD1MjqbRuzHc4yz=S2PCW7u3W0-_BQ@mail.gmail.com>
 <20230516222919.79bba667@rorschach.local.home> <CAHk-=wh_GEr4ehJKwMM3UA0-7CfNpVH7v_T-=1u+gq9VZD70mw@mail.gmail.com>
 <20230517172243.GA152@W11-BEAU-MD.localdomain> <CAHk-=whzzuNEW8UcV2_8OyuKcXPrk7-j_8GzOoroxz9JiZiD3w@mail.gmail.com>
 <20230517190750.GA366@W11-BEAU-MD.localdomain> <CAHk-=whTBvXJuoi_kACo3qi5WZUmRrhyA-_=rRFsycTytmB6qw@mail.gmail.com>
 <CAHk-=wi4w9bPKFFGwLULjJf9hnkL941+c4HbeEVKNzqH04wqDA@mail.gmail.com>
In-Reply-To: <CAHk-=wi4w9bPKFFGwLULjJf9hnkL941+c4HbeEVKNzqH04wqDA@mail.gmail.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Wed, 17 May 2023 13:08:59 -0700
Message-ID: <CAHk-=wicngggxVpbnrYHjRTwGE0WYscPRM+L2HO2BF8ia1EXgQ@mail.gmail.com>
Subject: Re: [PATCH] tracing/user_events: Run BPF program if attached
To:     Beau Belgrave <beaub@linux.microsoft.com>
Cc:     Steven Rostedt <rostedt@goodmis.org>,
        Alexei Starovoitov <alexei.starovoitov@gmail.com>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        linux-trace-kernel@vger.kernel.org,
        Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Andrii Nakryiko <andrii@kernel.org>, bpf <bpf@vger.kernel.org>,
        David Vernet <void@manifault.com>, dthaler@microsoft.com,
        brauner@kernel.org, hch@infradead.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Wed, May 17, 2023 at 12:36=E2=80=AFPM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> .. this is the patch that I think should go on top of it to fix the
> misleading "safe" and the incorrect RCU walk.
>
> NOTE! This adds that
>
>         lockdep_assert_held(&event_mutex);
>
> to user_event_enabler_update() too.

One more note: I think it would be really good to use different names
for the "links".

We have "mm->link", that is the list of mm's on the 'user_event_mms'
list, protected by the 'user_event_mms_lock' and RCU-walked.

And then we have 'enabler->link', which is the list of enables on the
'user_mm->enablers' list, protected by event_mutex, and _also_
occasionally RCU-walked.

And then we have 'validator->link', which isn't RCU-walked, and seems
to also be protected by the event_mutex (?).

This is all very confusing when looking at it as an outsider.
Particularly when you sometimes just see

        list_del_rcu(&mm->link);

and you have to figure "which 'link' are we talking about again?".

Also, I have to say, I found "mm->next" *really* confusing at first.

It turns out that "mm->next" is this list that is dynamically built up
by user_event_mm_get_all(), and is only a one-shot list that is valid
only as long as 'event_mutex' is held. But the only lock the code
*talks* about is the RCU lock, which does *not* protect that list, and
only exists as a way to walk that user_event_mms list without taking
any locks.

So user_event_enabler_update() actually relies on that 'event_mutex'
lock in another way than the obvious one that is about the
mm->enablers list that it *also* walks.

Again, that all seems to work, but it's *really* confusing how
"mm->next" always exists as a field, but is only usable and valid
while you hold that event_mutex and have called
user_event_mm_get_all() to gather the list.

I think both user_event_enabler_update() and user_event_mm_get_all()
should have a mention of how they require event_mutex and how that
->next list works.

Anyway, I still *think* the two patches I sent out are right, but I'm
just mentioning this confusion I had to deal with when trying to
decode what the rules were. Maybe all the above is obvious to
everybody else, but it took me a while to decipher (and maybe I
misread something).

             Linus