From: Desmond Cheong Zhi Xi desmondcheongzx@gmail.com
commit ba316be1b6a00db7126ed9a39f9bee434a508043 upstream.
struct sock.sk_timer should be used as a sock cleanup timer. However, SCO uses it to implement sock timeouts.
This causes issues because struct sock.sk_timer's callback is run in an IRQ context, and the timer callback function sco_sock_timeout takes a spin lock on the socket. However, other functions such as sco_conn_del and sco_conn_ready take the spin lock with interrupts enabled.
This inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} lock usage could lead to deadlocks as reported by Syzbot [1]: CPU0 ---- lock(slock-AF_BLUETOOTH-BTPROTO_SCO); <Interrupt> lock(slock-AF_BLUETOOTH-BTPROTO_SCO);
To fix this, we use delayed work to implement SCO sock timouts instead. This allows us to avoid taking the spin lock on the socket in an IRQ context, and corrects the misuse of struct sock.sk_timer.
As a note, cancel_delayed_work is used instead of cancel_delayed_work_sync in sco_sock_set_timer and sco_sock_clear_timer to avoid a deadlock. In the future, the call to bh_lock_sock inside sco_sock_timeout should be changed to lock_sock to synchronize with other functions using lock_sock. However, since sco_sock_set_timer and sco_sock_clear_timer are sometimes called under the locked socket (in sco_connect and __sco_sock_close), cancel_delayed_work_sync might cause them to sleep until an sco_sock_timeout that has started finishes running. But sco_sock_timeout would also sleep until it can grab the lock_sock.
Using cancel_delayed_work is fine because sco_sock_timeout does not change from run to run, hence there is no functional difference between: 1. waiting for a timeout to finish running before scheduling another timeout 2. scheduling another timeout while a timeout is running.
Link: https://syzkaller.appspot.com/bug?id=9089d89de0502e120f234ca0fc8a703f7368b31... [1] Reported-by: syzbot+2f6d7c28bb4bf7e82060@syzkaller.appspotmail.com Tested-by: syzbot+2f6d7c28bb4bf7e82060@syzkaller.appspotmail.com Signed-off-by: Desmond Cheong Zhi Xi desmondcheongzx@gmail.com Signed-off-by: Luiz Augusto von Dentz luiz.von.dentz@intel.com [OP: adjusted context for 4.14] Signed-off-by: Ovidiu Panait ovidiu.panait@windriver.com --- Note: these 2 fixes are part of CVE-2021-3640 patchset and are already present in 4.14+ stable branches. For this backport, small contextual adjustments were done to account for the old setup_timer() API usage.
net/bluetooth/sco.c | 35 +++++++++++++++++++++++++++++------ 1 file changed, 29 insertions(+), 6 deletions(-)
diff --git a/net/bluetooth/sco.c b/net/bluetooth/sco.c index a5cc8942fc3f..69d489f1f363 100644 --- a/net/bluetooth/sco.c +++ b/net/bluetooth/sco.c @@ -48,6 +48,8 @@ struct sco_conn { spinlock_t lock; struct sock *sk;
+ struct delayed_work timeout_work; + unsigned int mtu; };
@@ -73,9 +75,20 @@ struct sco_pinfo { #define SCO_CONN_TIMEOUT (HZ * 40) #define SCO_DISCONN_TIMEOUT (HZ * 2)
-static void sco_sock_timeout(unsigned long arg) +static void sco_sock_timeout(struct work_struct *work) { - struct sock *sk = (struct sock *)arg; + struct sco_conn *conn = container_of(work, struct sco_conn, + timeout_work.work); + struct sock *sk; + + sco_conn_lock(conn); + sk = conn->sk; + if (sk) + sock_hold(sk); + sco_conn_unlock(conn); + + if (!sk) + return;
BT_DBG("sock %p state %d", sk, sk->sk_state);
@@ -89,14 +102,21 @@ static void sco_sock_timeout(unsigned long arg)
static void sco_sock_set_timer(struct sock *sk, long timeout) { + if (!sco_pi(sk)->conn) + return; + BT_DBG("sock %p state %d timeout %ld", sk, sk->sk_state, timeout); - sk_reset_timer(sk, &sk->sk_timer, jiffies + timeout); + cancel_delayed_work(&sco_pi(sk)->conn->timeout_work); + schedule_delayed_work(&sco_pi(sk)->conn->timeout_work, timeout); }
static void sco_sock_clear_timer(struct sock *sk) { + if (!sco_pi(sk)->conn) + return; + BT_DBG("sock %p state %d", sk, sk->sk_state); - sk_stop_timer(sk, &sk->sk_timer); + cancel_delayed_work(&sco_pi(sk)->conn->timeout_work); }
/* ---- SCO connections ---- */ @@ -176,6 +196,9 @@ static void sco_conn_del(struct hci_conn *hcon, int err) sco_chan_del(sk, err); bh_unlock_sock(sk); sock_put(sk); + + /* Ensure no more work items will run before freeing conn. */ + cancel_delayed_work_sync(&conn->timeout_work); }
hcon->sco_data = NULL; @@ -190,6 +213,8 @@ static void __sco_chan_add(struct sco_conn *conn, struct sock *sk, sco_pi(sk)->conn = conn; conn->sk = sk;
+ INIT_DELAYED_WORK(&conn->timeout_work, sco_sock_timeout); + if (parent) bt_accept_enqueue(parent, sk, true); } @@ -466,8 +491,6 @@ static struct sock *sco_sock_alloc(struct net *net, struct socket *sock,
sco_pi(sk)->setting = BT_VOICE_CVSD_16BIT;
- setup_timer(&sk->sk_timer, sco_sock_timeout, (unsigned long)sk); - bt_sock_link(&sco_sk_list, sk); return sk; }
From: Desmond Cheong Zhi Xi desmondcheongzx@gmail.com
commit 49d8a5606428ca0962d09050a5af81461ff90fbb upstream.
Before freeing struct sco_conn, all delayed timeout work should be cancelled. Otherwise, sco_sock_timeout could potentially use the sco_conn after it has been freed.
Additionally, sco_conn.timeout_work should be initialized when the connection is allocated, not when the channel is added. This is because an sco_conn can create channels with multiple sockets over its lifetime, which happens if sockets are released but the connection isn't deleted.
Fixes: ba316be1b6a0 ("Bluetooth: schedule SCO timeouts with delayed_work") Signed-off-by: Desmond Cheong Zhi Xi desmondcheongzx@gmail.com Signed-off-by: Luiz Augusto von Dentz luiz.von.dentz@intel.com [OP: adjusted context for 4.14] Signed-off-by: Ovidiu Panait ovidiu.panait@windriver.com --- net/bluetooth/sco.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/net/bluetooth/sco.c b/net/bluetooth/sco.c index 69d489f1f363..5c411118b30d 100644 --- a/net/bluetooth/sco.c +++ b/net/bluetooth/sco.c @@ -133,6 +133,7 @@ static struct sco_conn *sco_conn_add(struct hci_conn *hcon) return NULL;
spin_lock_init(&conn->lock); + INIT_DELAYED_WORK(&conn->timeout_work, sco_sock_timeout);
hcon->sco_data = conn; conn->hcon = hcon; @@ -196,11 +197,11 @@ static void sco_conn_del(struct hci_conn *hcon, int err) sco_chan_del(sk, err); bh_unlock_sock(sk); sock_put(sk); - - /* Ensure no more work items will run before freeing conn. */ - cancel_delayed_work_sync(&conn->timeout_work); }
+ /* Ensure no more work items will run before freeing conn. */ + cancel_delayed_work_sync(&conn->timeout_work); + hcon->sco_data = NULL; kfree(conn); } @@ -213,8 +214,6 @@ static void __sco_chan_add(struct sco_conn *conn, struct sock *sk, sco_pi(sk)->conn = conn; conn->sk = sk;
- INIT_DELAYED_WORK(&conn->timeout_work, sco_sock_timeout); - if (parent) bt_accept_enqueue(parent, sk, true); }
On Tue, Jan 18, 2022 at 06:32:39PM +0200, Ovidiu Panait wrote:
From: Desmond Cheong Zhi Xi desmondcheongzx@gmail.com
commit ba316be1b6a00db7126ed9a39f9bee434a508043 upstream.
struct sock.sk_timer should be used as a sock cleanup timer. However, SCO uses it to implement sock timeouts.
This causes issues because struct sock.sk_timer's callback is run in an IRQ context, and the timer callback function sco_sock_timeout takes a spin lock on the socket. However, other functions such as sco_conn_del and sco_conn_ready take the spin lock with interrupts enabled.
This inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} lock usage could lead to deadlocks as reported by Syzbot [1]: CPU0 ---- lock(slock-AF_BLUETOOTH-BTPROTO_SCO);
<Interrupt> lock(slock-AF_BLUETOOTH-BTPROTO_SCO);
To fix this, we use delayed work to implement SCO sock timouts instead. This allows us to avoid taking the spin lock on the socket in an IRQ context, and corrects the misuse of struct sock.sk_timer.
As a note, cancel_delayed_work is used instead of cancel_delayed_work_sync in sco_sock_set_timer and sco_sock_clear_timer to avoid a deadlock. In the future, the call to bh_lock_sock inside sco_sock_timeout should be changed to lock_sock to synchronize with other functions using lock_sock. However, since sco_sock_set_timer and sco_sock_clear_timer are sometimes called under the locked socket (in sco_connect and __sco_sock_close), cancel_delayed_work_sync might cause them to sleep until an sco_sock_timeout that has started finishes running. But sco_sock_timeout would also sleep until it can grab the lock_sock.
Using cancel_delayed_work is fine because sco_sock_timeout does not change from run to run, hence there is no functional difference between:
- waiting for a timeout to finish running before scheduling another
timeout 2. scheduling another timeout while a timeout is running.
Link: https://syzkaller.appspot.com/bug?id=9089d89de0502e120f234ca0fc8a703f7368b31... [1] Reported-by: syzbot+2f6d7c28bb4bf7e82060@syzkaller.appspotmail.com Tested-by: syzbot+2f6d7c28bb4bf7e82060@syzkaller.appspotmail.com Signed-off-by: Desmond Cheong Zhi Xi desmondcheongzx@gmail.com Signed-off-by: Luiz Augusto von Dentz luiz.von.dentz@intel.com [OP: adjusted context for 4.14] Signed-off-by: Ovidiu Panait ovidiu.panait@windriver.com
Note: these 2 fixes are part of CVE-2021-3640 patchset and are already present in 4.14+ stable branches. For this backport, small contextual adjustments were done to account for the old setup_timer() API usage.
Both now queued up, thanks.
greg k-h
linux-stable-mirror@lists.linaro.org