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  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]
Date:   Wed, 21 Jul 2021 17:38:31 +0800
From:   Desmond Cheong Zhi Xi <desmondcheongzx@...il.com>
To:     marcel@...tmann.org, johan.hedberg@...il.com, luiz.dentz@...il.com,
        davem@...emloft.net, kuba@...nel.org, matthieu.baerts@...sares.net,
        stefan@...enfreihafen.org
Cc:     Desmond Cheong Zhi Xi <desmondcheongzx@...il.com>,
        linux-bluetooth@...r.kernel.org, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, skhan@...uxfoundation.org,
        gregkh@...uxfoundation.org,
        linux-kernel-mentees@...ts.linuxfoundation.org,
        syzbot+2f6d7c28bb4bf7e82060@...kaller.appspotmail.com
Subject: [PATCH v3 1/2] Bluetooth: fix inconsistent lock state in SCO

Syzbot reported an inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} lock
usage in sco_conn_del and sco_sock_timeout that could lead to
deadlocks.

This inconsistent lock state can also happen in sco_conn_ready,
rfcomm_connect_ind, and bt_accept_enqueue.

The issue is that these functions take a spin lock on the socket with
interrupts enabled, but sco_sock_timeout takes the lock in an IRQ
context. This could lead to deadlocks:

       CPU0
       ----
  lock(slock-AF_BLUETOOTH-BTPROTO_SCO);
  <Interrupt>
    lock(slock-AF_BLUETOOTH-BTPROTO_SCO);

 *** DEADLOCK ***

We fix this by ensuring that local bh is disabled before calling
bh_lock_sock.

After doing this, we additionally need to protect sco_conn_lock by
disabling local bh.

This is necessary because sco_conn_del makes a call to sco_chan_del
while holding on to the sock lock, and sco_chan_del itself makes a
call to sco_conn_lock. If sco_conn_lock is held elsewhere with
interrupts enabled, there could still be a
slock-AF_BLUETOOTH-BTPROTO_SCO --> &conn->lock#2 lock inversion as
follows:

        CPU0                    CPU1
        ----                    ----
   lock(&conn->lock#2);
                                local_irq_disable();
                                lock(slock-AF_BLUETOOTH-BTPROTO_SCO);
                                lock(&conn->lock#2);
   <Interrupt>
     lock(slock-AF_BLUETOOTH-BTPROTO_SCO);

  *** DEADLOCK ***

Although sco_conn_del disables local bh before calling sco_chan_del,
we can still wrap the calls to sco_conn_lock in sco_chan_del, with
local_bh_disable/enable as this pair of functions are reentrant.

Reported-by: syzbot+2f6d7c28bb4bf7e82060@...kaller.appspotmail.com
Tested-by: syzbot+2f6d7c28bb4bf7e82060@...kaller.appspotmail.com
Signed-off-by: Desmond Cheong Zhi Xi <desmondcheongzx@...il.com>
---
 net/bluetooth/sco.c | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/net/bluetooth/sco.c b/net/bluetooth/sco.c
index 3bd41563f118..34f3419c3330 100644
--- a/net/bluetooth/sco.c
+++ b/net/bluetooth/sco.c
@@ -140,10 +140,12 @@ static void sco_chan_del(struct sock *sk, int err)
 	BT_DBG("sk %p, conn %p, err %d", sk, conn, err);
 
 	if (conn) {
+		local_bh_disable();
 		sco_conn_lock(conn);
 		conn->sk = NULL;
 		sco_pi(sk)->conn = NULL;
 		sco_conn_unlock(conn);
+		local_bh_enable();
 
 		if (conn->hcon)
 			hci_conn_drop(conn->hcon);
@@ -167,16 +169,22 @@ static void sco_conn_del(struct hci_conn *hcon, int err)
 	BT_DBG("hcon %p conn %p, err %d", hcon, conn, err);
 
 	/* Kill socket */
+	local_bh_disable();
 	sco_conn_lock(conn);
 	sk = conn->sk;
 	sco_conn_unlock(conn);
+	local_bh_enable();
 
 	if (sk) {
 		sock_hold(sk);
+
+		local_bh_disable();
 		bh_lock_sock(sk);
 		sco_sock_clear_timer(sk);
 		sco_chan_del(sk, err);
 		bh_unlock_sock(sk);
+		local_bh_enable();
+
 		sco_sock_kill(sk);
 		sock_put(sk);
 	}
@@ -202,6 +210,7 @@ static int sco_chan_add(struct sco_conn *conn, struct sock *sk,
 {
 	int err = 0;
 
+	local_bh_disable();
 	sco_conn_lock(conn);
 	if (conn->sk)
 		err = -EBUSY;
@@ -209,6 +218,7 @@ static int sco_chan_add(struct sco_conn *conn, struct sock *sk,
 		__sco_chan_add(conn, sk, parent);
 
 	sco_conn_unlock(conn);
+	local_bh_enable();
 	return err;
 }
 
@@ -303,9 +313,11 @@ static void sco_recv_frame(struct sco_conn *conn, struct sk_buff *skb)
 {
 	struct sock *sk;
 
+	local_bh_disable();
 	sco_conn_lock(conn);
 	sk = conn->sk;
 	sco_conn_unlock(conn);
+	local_bh_enable();
 
 	if (!sk)
 		goto drop;
@@ -420,10 +432,12 @@ static void __sco_sock_close(struct sock *sk)
 		if (sco_pi(sk)->conn->hcon) {
 			sk->sk_state = BT_DISCONN;
 			sco_sock_set_timer(sk, SCO_DISCONN_TIMEOUT);
+			local_bh_disable();
 			sco_conn_lock(sco_pi(sk)->conn);
 			hci_conn_drop(sco_pi(sk)->conn->hcon);
 			sco_pi(sk)->conn->hcon = NULL;
 			sco_conn_unlock(sco_pi(sk)->conn);
+			local_bh_enable();
 		} else
 			sco_chan_del(sk, ECONNRESET);
 		break;
@@ -1084,21 +1098,26 @@ static void sco_conn_ready(struct sco_conn *conn)
 
 	if (sk) {
 		sco_sock_clear_timer(sk);
+		local_bh_disable();
 		bh_lock_sock(sk);
 		sk->sk_state = BT_CONNECTED;
 		sk->sk_state_change(sk);
 		bh_unlock_sock(sk);
+		local_bh_enable();
 	} else {
+		local_bh_disable();
 		sco_conn_lock(conn);
 
 		if (!conn->hcon) {
 			sco_conn_unlock(conn);
+			local_bh_enable();
 			return;
 		}
 
 		parent = sco_get_sock_listen(&conn->hcon->src);
 		if (!parent) {
 			sco_conn_unlock(conn);
+			local_bh_enable();
 			return;
 		}
 
@@ -1109,6 +1128,7 @@ static void sco_conn_ready(struct sco_conn *conn)
 		if (!sk) {
 			bh_unlock_sock(parent);
 			sco_conn_unlock(conn);
+			local_bh_enable();
 			return;
 		}
 
@@ -1131,6 +1151,7 @@ static void sco_conn_ready(struct sco_conn *conn)
 		bh_unlock_sock(parent);
 
 		sco_conn_unlock(conn);
+		local_bh_enable();
 	}
 }
 
-- 
2.25.1

Powered by blists - more mailing lists