Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210908184905.163787-1-posk@google.com> <20210908184905.163787-3-posk@google.com>
 <CAG48ez2LyLNkH4iVbeKJUuH=oh57WECkKYCW+G9mtheoh7Fsvg@mail.gmail.com>
In-Reply-To: <CAG48ez2LyLNkH4iVbeKJUuH=oh57WECkKYCW+G9mtheoh7Fsvg@mail.gmail.com>
From:   Jann Horn <jannh@google.com>
Date:   Thu, 9 Sep 2021 03:16:17 +0200
Message-ID: <CAG48ez3rRReCVpuhQu_-f_a0Yn=cUBqu1aNSG0JtHK8mjqzFsw@mail.gmail.com>
Subject: Re: [PATCH 2/4 v0.5] sched/umcg: RFC: add userspace atomic helpers
To:     Peter Oskolkov <posk@posk.io>
Cc:     Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        linux-kernel@vger.kernel.org, linux-api@vger.kernel.org,
        Paul Turner <pjt@google.com>, Ben Segall <bsegall@google.com>,
        Peter Oskolkov <posk@google.com>,
        Andrei Vagin <avagin@google.com>,
        Thierry Delisle <tdelisle@uwaterloo.ca>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Sep 9, 2021 at 1:38 AM Jann Horn <jannh@google.com> wrote:
> On Wed, Sep 8, 2021 at 8:49 PM Peter Oskolkov <posk@posk.io> wrote:
> > Add helper functions to work atomically with userspace 32/64 bit values -
> > there are some .*futex.* named helpers, but they are not exactly
> > what is needed for UMCG; I haven't found what else I could use, so I
> > rolled these.
[...]
> You of course won't be able to simply traverse userspace pointers in
> such a situation, only access the specific userspace object that
> you've prepared beforehand, but luckily it looks like:
>
>  * idle_server_tid_ptr is only accessed in get_idle_server()
>    -> which is used from process_waking_worker()
>      -> which runs in sleepable context
>  * idle_workers_ptr is accessed from:
>    -> enqueue_idle_worker
>      -> which is also used from process_waking_worker()

Ah, I guess I got that wrong: process_waking_worker() is sleepable,
but it might be holding the mmap lock, so it can't fault, right? Which
means this would actually be problematic...