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;
Date:   Thu, 23 Mar 2023 13:44:59 -0400
From:   Johannes Weiner <hannes@cmpxchg.org>
To:     Suren Baghdasaryan <surenb@google.com>
Cc:     Domenico Cerasuolo <cerasuolodomenico@gmail.com>,
        linux-kernel@vger.kernel.org, peterz@infradead.org,
        brauner@kernel.org, chris@chrisdown.name
Subject: Re: [PATCH v2 3/3] sched/psi: allow unprivileged polling of N*2s
 period
Message-ID: <20230323174459.GH739026@cmpxchg.org>
References: <20230323103350.40569-1-cerasuolodomenico@gmail.com>
 <20230323103350.40569-4-cerasuolodomenico@gmail.com>
 <CAJuCfpFh0iL0cE8V3dh-HkDx3-20Kyo+ZAt0vnqvOzA33HwnFg@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <CAJuCfpFh0iL0cE8V3dh-HkDx3-20Kyo+ZAt0vnqvOzA33HwnFg@mail.gmail.com>
Precedence: bulk

On Thu, Mar 23, 2023 at 09:55:11AM -0700, Suren Baghdasaryan wrote:
> On Thu, Mar 23, 2023 at 3:34 AM Domenico Cerasuolo
> > @@ -1254,16 +1262,19 @@ int psi_show(struct seq_file *m, struct psi_group *group, enum psi_res res)
> >  }
> >
> >  struct psi_trigger *psi_trigger_create(struct psi_group *group,
> > -                       char *buf, enum psi_res res)
> > +                       char *buf, enum psi_res res, struct file *file)
> >  {
> >         struct psi_trigger *t;
> >         enum psi_states state;
> >         u32 threshold_us;
> > +       bool privileged;
> >         u32 window_us;
> >
> >         if (static_branch_likely(&psi_disabled))
> >                 return ERR_PTR(-EOPNOTSUPP);
> >
> > +       privileged = cap_raised(file->f_cred->cap_effective, CAP_SYS_RESOURCE);
> 
> I missed one detail here. We are moving the cap check from open() to
> write(). That might break potential users which open the file from a
> process with that cap and then pass that FD to an unprivileged process
> to create the trigger by writing to that file. I'm not aware of any
> use of such a pattern but it is possible there are such users.
> With this change such users would have to delegate trigger creation to
> the privileged process too and the received FD would be used only for
> polling. IMHO that's a safer pattern because triggers are created by
> the privileged process.

Oh, it's checking file->f_cred, which is set up at open(). So if the
opener is privileged, the write can be delegated to an unprivileged
process.

But I agree that this is subtle and could use a comment.

This was a usecase specifically requested by Christian, actually.