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;
From:   Dylan Yudaken <dylany@fb.com>
To:     <viro@zeniv.linux.org.uk>, <bcrl@kvack.org>
CC:     <linux-fsdevel@vger.kernel.org>, <linux-aio@kvack.org>,
        <linux-kernel@vger.kernel.org>, Jens Axboe <axboe@kernel.dk>,
        <Kernel-team@fb.com>, Dylan Yudaken <dylany@fb.com>
Subject: [PATCH] eventfd: guard wake_up in eventfd fs calls as well
Date:   Tue, 16 Aug 2022 06:59:59 -0700
Message-ID: <20220816135959.1490641-1-dylany@fb.com>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain
Precedence: bulk

Guard wakeups that the user can trigger, and that may end up triggering a
call back into eventfd_signal. This is in addition to the current approac=
h
that only guards in eventfd_signal.

Rename in_eventfd_signal -> in_eventfd at the same time to reflect this.

Without this there would be a deadlock in the following code using libaio=
:

int main()
{
	struct io_context *ctx =3D NULL;
	struct iocb iocb;
	struct iocb *iocbs[] =3D { &iocb };
	int evfd;
        uint64_t val =3D 1;

	evfd =3D eventfd(0, EFD_CLOEXEC);
	assert(!io_setup(2, &ctx));
	io_prep_poll(&iocb, evfd, POLLIN);
	io_set_eventfd(&iocb, evfd);
	assert(1 =3D=3D io_submit(ctx, 1, iocbs));
        write(evfd, &val, 8);
}

Signed-off-by: Dylan Yudaken <dylany@fb.com>
---
 fs/eventfd.c            | 10 +++++++---
 include/linux/eventfd.h |  2 +-
 include/linux/sched.h   |  2 +-
 3 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/fs/eventfd.c b/fs/eventfd.c
index 3627dd7d25db..c0ffee99ad23 100644
--- a/fs/eventfd.c
+++ b/fs/eventfd.c
@@ -69,17 +69,17 @@ __u64 eventfd_signal(struct eventfd_ctx *ctx, __u64 n=
)
 	 * it returns false, the eventfd_signal() call should be deferred to a
 	 * safe context.
 	 */
-	if (WARN_ON_ONCE(current->in_eventfd_signal))
+	if (WARN_ON_ONCE(current->in_eventfd))
 		return 0;
=20
 	spin_lock_irqsave(&ctx->wqh.lock, flags);
-	current->in_eventfd_signal =3D 1;
+	current->in_eventfd =3D 1;
 	if (ULLONG_MAX - ctx->count < n)
 		n =3D ULLONG_MAX - ctx->count;
 	ctx->count +=3D n;
 	if (waitqueue_active(&ctx->wqh))
 		wake_up_locked_poll(&ctx->wqh, EPOLLIN);
-	current->in_eventfd_signal =3D 0;
+	current->in_eventfd =3D 0;
 	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
=20
 	return n;
@@ -253,8 +253,10 @@ static ssize_t eventfd_read(struct kiocb *iocb, stru=
ct iov_iter *to)
 		__set_current_state(TASK_RUNNING);
 	}
 	eventfd_ctx_do_read(ctx, &ucnt);
+	current->in_eventfd =3D 1;
 	if (waitqueue_active(&ctx->wqh))
 		wake_up_locked_poll(&ctx->wqh, EPOLLOUT);
+	current->in_eventfd =3D 0;
 	spin_unlock_irq(&ctx->wqh.lock);
 	if (unlikely(copy_to_iter(&ucnt, sizeof(ucnt), to) !=3D sizeof(ucnt)))
 		return -EFAULT;
@@ -301,8 +303,10 @@ static ssize_t eventfd_write(struct file *file, cons=
t char __user *buf, size_t c
 	}
 	if (likely(res > 0)) {
 		ctx->count +=3D ucnt;
+		current->in_eventfd =3D 1;
 		if (waitqueue_active(&ctx->wqh))
 			wake_up_locked_poll(&ctx->wqh, EPOLLIN);
+		current->in_eventfd =3D 0;
 	}
 	spin_unlock_irq(&ctx->wqh.lock);
=20
diff --git a/include/linux/eventfd.h b/include/linux/eventfd.h
index 305d5f19093b..30eb30d6909b 100644
--- a/include/linux/eventfd.h
+++ b/include/linux/eventfd.h
@@ -46,7 +46,7 @@ void eventfd_ctx_do_read(struct eventfd_ctx *ctx, __u64=
 *cnt);
=20
 static inline bool eventfd_signal_allowed(void)
 {
-	return !current->in_eventfd_signal;
+	return !current->in_eventfd;
 }
=20
 #else /* CONFIG_EVENTFD */
diff --git a/include/linux/sched.h b/include/linux/sched.h
index e7b2f8a5c711..8d82d6d32670 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -936,7 +936,7 @@ struct task_struct {
 #endif
 #ifdef CONFIG_EVENTFD
 	/* Recursion prevention for eventfd_signal() */
-	unsigned			in_eventfd_signal:1;
+	unsigned			in_eventfd:1;
 #endif
 #ifdef CONFIG_IOMMU_SVA
 	unsigned			pasid_activated:1;
--=20
2.30.2