linux-kernel - Re: [PATCH v2 6/8] binder: allow freeze notification for dead nodes

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAH5fLgjnyKtXsnPbvCFz64BBRqvWPwh6reM-myWA9AEBKFhcJg@mail.gmail.com>
Date: Fri, 27 Sep 2024 18:15:40 +0200
From: Alice Ryhl <aliceryhl@...gle.com>
To: Yu-Ting Tseng <yutingtseng@...gle.com>
Cc: Carlos Llamas <cmllamas@...gle.com>, Greg Kroah-Hartman <gregkh@...uxfoundation.org>, 
	Arve Hjønnevåg <arve@...roid.com>, 
	Todd Kjos <tkjos@...roid.com>, Martijn Coenen <maco@...roid.com>, 
	Joel Fernandes <joel@...lfernandes.org>, Christian Brauner <brauner@...nel.org>, 
	Suren Baghdasaryan <surenb@...gle.com>, linux-kernel@...r.kernel.org, kernel-team@...roid.com, 
	stable@...r.kernel.org
Subject: Re: [PATCH v2 6/8] binder: allow freeze notification for dead nodes

On Fri, Sep 27, 2024 at 6:13 PM Yu-Ting Tseng <yutingtseng@...gle.com> wrote:
>
> On Fri, Sep 27, 2024 at 12:19 AM Alice Ryhl <aliceryhl@...gle.com> wrote:
> >
> > On Fri, Sep 27, 2024 at 1:37 AM Carlos Llamas <cmllamas@...gle.com> wrote:
> > >
> > > Alice points out that binder_request_freeze_notification() should not
> > > return EINVAL when the relevant node is dead [1]. The node can die at
> > > any point even if the user input is valid. Instead, allow the request
> > > to be allocated but skip the initial notification for dead nodes. This
> > > avoids propagating unnecessary errors back to userspace.
> > >
> > > Fixes: d579b04a52a1 ("binder: frozen notification")
> > > Cc: stable@...r.kernel.org
> > > Suggested-by: Alice Ryhl <aliceryhl@...gle.com>
> > > Link: https://lore.kernel.org/all/CAH5fLghapZJ4PbbkC8V5A6Zay-_sgTzwVpwqk6RWWUNKKyJC_Q@mail.gmail.com/ [1]
> > > Signed-off-by: Carlos Llamas <cmllamas@...gle.com>
> > > ---
> > >  drivers/android/binder.c | 28 +++++++++++++---------------
> > >  1 file changed, 13 insertions(+), 15 deletions(-)
> > >
> > > diff --git a/drivers/android/binder.c b/drivers/android/binder.c
> > > index 73dc6cbc1681..415fc9759249 100644
> > > --- a/drivers/android/binder.c
> > > +++ b/drivers/android/binder.c
> > > @@ -3856,7 +3856,6 @@ binder_request_freeze_notification(struct binder_proc *proc,
> > >  {
> > >         struct binder_ref_freeze *freeze;
> > >         struct binder_ref *ref;
> > > -       bool is_frozen;
> > >
> > >         freeze = kzalloc(sizeof(*freeze), GFP_KERNEL);
> > >         if (!freeze)
> > > @@ -3872,32 +3871,31 @@ binder_request_freeze_notification(struct binder_proc *proc,
> > >         }
> > >
> > >         binder_node_lock(ref->node);
> > > -
> > > -       if (ref->freeze || !ref->node->proc) {
> > > -               binder_user_error("%d:%d invalid BC_REQUEST_FREEZE_NOTIFICATION %s\n",
> > > -                                 proc->pid, thread->pid,
> > > -                                 ref->freeze ? "already set" : "dead node");
> > > +       if (ref->freeze) {
> > > +               binder_user_error("%d:%d BC_REQUEST_FREEZE_NOTIFICATION already set\n",
> > > +                                 proc->pid, thread->pid);
> > >                 binder_node_unlock(ref->node);
> > >                 binder_proc_unlock(proc);
> > >                 kfree(freeze);
> > >                 return -EINVAL;
> > >         }
> > > -       binder_inner_proc_lock(ref->node->proc);
> > > -       is_frozen = ref->node->proc->is_frozen;
> > > -       binder_inner_proc_unlock(ref->node->proc);
> > >
> > >         binder_stats_created(BINDER_STAT_FREEZE);
> > >         INIT_LIST_HEAD(&freeze->work.entry);
> > >         freeze->cookie = handle_cookie->cookie;
> > >         freeze->work.type = BINDER_WORK_FROZEN_BINDER;
> > > -       freeze->is_frozen = is_frozen;
> > > -
> > >         ref->freeze = freeze;
> > >
> > > -       binder_inner_proc_lock(proc);
> > > -       binder_enqueue_work_ilocked(&ref->freeze->work, &proc->todo);
> > > -       binder_wakeup_proc_ilocked(proc);
> > > -       binder_inner_proc_unlock(proc);
> > > +       if (ref->node->proc) {
> > > +               binder_inner_proc_lock(ref->node->proc);
> > > +               freeze->is_frozen = ref->node->proc->is_frozen;
> > > +               binder_inner_proc_unlock(ref->node->proc);
> > > +
> > > +               binder_inner_proc_lock(proc);
> > > +               binder_enqueue_work_ilocked(&freeze->work, &proc->todo);
> > > +               binder_wakeup_proc_ilocked(proc);
> > > +               binder_inner_proc_unlock(proc);
> >
> > This is not a problem with your change ... but, why exactly are we
> > scheduling the BINDER_WORK_FROZEN_BINDER right after creating it? For
> > death notications, we only schedule it immediately if the process is
> > dead. So shouldn't we only schedule it if the process is not frozen?
> >
> > And if the answer is that frozen notifications are always sent
> > immediately to notify about the current state, then we should also
> > send one for a dead process ... maybe. I guess a dead process is not
> > frozen?
> Yes this is to immediately notify about the current state (frozen or
> unfrozen). A dead process is in neither state so it feels more correct
> not to send either?

Okay.

On the other hand, I can easily imagine userspace code being written
with the assumption that it'll always get a notification immediately.
That would probably result in deadlocks in the edge case where the
process happens to be dead.

Alice