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;
Organization: Red Hat UK Ltd. Registered Address: Red Hat UK Ltd, Amberley
        Place, 107-111 Peascod Street, Windsor, Berkshire, SI4 1TE, United
        Kingdom.
        Registered in England and Wales under Company Registration No. 3798903
From:   David Howells <dhowells@redhat.com>
In-Reply-To: <1822b768504.1d4e377e236061.5518350412857967240@siddh.me>
References: <1822b768504.1d4e377e236061.5518350412857967240@siddh.me> <20220723135447.199557-1-code@siddh.me> <Ytv/4Tljvlt0PJ2r@kroah.com>
To:     Siddh Raman Pant <code@siddh.me>
Cc:     dhowells@redhat.com, "Greg KH" <gregkh@linuxfoundation.org>,
        "Christophe JAILLET" <christophe.jaillet@wanadoo.fr>,
        "Eric Dumazet" <edumazet@google.com>,
        "Fabio M. De Francesco" <fmdefrancesco@gmail.com>,
        "linux-security-modules" <linux-security-module@vger.kernel.org>,
        "linux-kernel-mentees" 
        <linux-kernel-mentees@lists.linuxfoundation.org>,
        "linux-kernel" <linux-kernel@vger.kernel.org>,
        "syzbot+c70d87ac1d001f29a058" 
        <syzbot+c70d87ac1d001f29a058@syzkaller.appspotmail.com>
Subject: Re: [PATCH] kernel/watch_queue: Make pipe NULL while clearing watch_queue
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <3558069.1658933200.1@warthog.procyon.org.uk>
Content-Transfer-Encoding: quoted-printable
Date:   Wed, 27 Jul 2022 15:46:40 +0100
Message-ID: <3558070.1658933200@warthog.procyon.org.uk>
Precedence: bulk

Siddh Raman Pant <code@siddh.me> wrote:

> Greg KH <gregkh@linuxfoundation.org> wrote:

> > > -	spin_unlock_bh(&wqueue->lock);
> > >  	rcu_read_unlock();
> >
> > Also you now have a spinlock held when calling rcu_read_unlock(), are
> > you sure that's ok?

Worse, we have softirqs disabled still, which might cause problems for
rcu_read_unlock()?

> We logically should not do write operations in a read critical section, =
so the
> nulling of `wqueue->pipe->watch_queue` should happen after rcu_read_unlo=
ck().
> Also, since we already have a spinlock, we can use it to ensure the null=
ing.
> So I think it is okay.

Read/write locks are perhaps misnamed in this sense; they perhaps should b=
e
shared/exclusive.  But, yes, we *can* do certain write operations with the
lock held - if we're careful.  Locks are required if we need to pairs of
related memory accesses; if we're only making a single non-dependent write=
,
then we don't necessarily need a write lock.

However, you're referring to RCU read lock.  That's a very special lock th=
at
has to do with maintenance of persistence of objects without taking any ot=
her
lock.  The moment you drop that lock, anything you accessed under RCU prot=
ocol
rules should be considered to have evaporated.

Think of it more as a way to have a deferred destructor/deallocator.

So I would do:

+
+	/* Clearing the watch queue, so we should clean the associated pipe. */
+	if (wqueue->pipe) {
+		wqueue->pipe->watch_queue =3D NULL;
+		wqueue->pipe =3D NULL;
+	}
+
	spin_unlock_bh(&wqueue->lock);
 	rcu_read_unlock();
 }

However, since you're now changing wqueue->pipe whilst a notification is b=
eing
posted, you need a barrier in post_one_notification() to prevent the compi=
ler
from reloading the value:

	struct pipe_inode_info *pipe =3D READ_ONCE(wqueue->pipe);

David