[PATCH 2/4] net: allow sk_prot->release_cb() without sock lock held

[Jens Axboe <axboe@xxxxxxxxx>]

Add helpers that allow ->release_cb() to acquire the socket bh lock
when needed. For normal sockets, ->release_cb() is always invoked with
that lock held. For nolock sockets it will not be held, so provide an
easy way to acquire it when necessary.

Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
---
 include/net/sock.h    | 10 ++++++++++
 net/atm/common.c      |  5 ++++-
 net/ipv4/tcp_output.c |  2 ++
 net/mptcp/protocol.c  |  3 +++
 net/smc/af_smc.c      |  2 ++
 5 files changed, 21 insertions(+), 1 deletion(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index e8283a65b757..99fcc4d7eed9 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1696,6 +1696,16 @@ void release_sock(struct sock *sk);
 				SINGLE_DEPTH_NESTING)
 #define bh_unlock_sock(__sk)	spin_unlock(&((__sk)->sk_lock.slock))
 
+/* nolock helpers */
+#define bh_lock_sock_on_nolock(__sk)	do {			\
+	if ((__sk)->sk_no_lock)					\
+		spin_lock_bh(&(__sk)->sk_lock.slock);		\
+} while (0)
+#define bh_unlock_sock_on_nolock(__sk)	do {			\
+	if ((__sk)->sk_no_lock)					\
+		spin_unlock_bh(&(__sk)->sk_lock.slock);		\
+} while (0)
+
 bool __lock_sock_fast(struct sock *sk) __acquires(&sk->sk_lock.slock);
 
 /**
diff --git a/net/atm/common.c b/net/atm/common.c
index 1cfa9bf1d187..471363e929f6 100644
--- a/net/atm/common.c
+++ b/net/atm/common.c
@@ -126,8 +126,11 @@ static void vcc_release_cb(struct sock *sk)
 {
 	struct atm_vcc *vcc = atm_sk(sk);
 
-	if (vcc->release_cb)
+	if (vcc->release_cb) {
+		bh_lock_sock_on_nolock(sk);
 		vcc->release_cb(vcc);
+		bh_lock_sock_on_nolock(sk);
+	}
 }
 
 static struct proto vcc_proto = {
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 9ede847f4199..9f86ea63cbac 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -1100,6 +1100,7 @@ void tcp_release_cb(struct sock *sk)
 	 * But following code is meant to be called from BH handlers,
 	 * so we should keep BH disabled, but early release socket ownership
 	 */
+	bh_lock_sock_on_nolock(sk);
 	sock_release_ownership(sk);
 
 	if (flags & TCPF_WRITE_TIMER_DEFERRED) {
@@ -1114,6 +1115,7 @@ void tcp_release_cb(struct sock *sk)
 		inet_csk(sk)->icsk_af_ops->mtu_reduced(sk);
 		__sock_put(sk);
 	}
+	bh_unlock_sock_on_nolock(sk);
 }
 EXPORT_SYMBOL(tcp_release_cb);
 
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index 0cbea3b6d0a4..ae9078e8e137 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -3065,6 +3065,8 @@ static void mptcp_release_cb(struct sock *sk)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
 
+	bh_lock_sock_on_nolock(sk);
+
 	for (;;) {
 		unsigned long flags = (msk->cb_flags & MPTCP_FLAGS_PROCESS_CTX_NEED) |
 				      msk->push_pending;
@@ -3103,6 +3105,7 @@ static void mptcp_release_cb(struct sock *sk)
 		__mptcp_error_report(sk);
 
 	__mptcp_update_rmem(sk);
+	bh_unlock_sock_on_nolock(sk);
 }
 
 /* MP_JOIN client subflow must wait for 4th ack before sending any data:
diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index f0d118e9f155..3456dc6cd38b 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -201,10 +201,12 @@ static void smc_release_cb(struct sock *sk)
 {
 	struct smc_sock *smc = smc_sk(sk);
 
+	bh_lock_sock_on_nolock(sk);
 	if (smc->conn.tx_in_release_sock) {
 		smc_tx_pending(&smc->conn);
 		smc->conn.tx_in_release_sock = false;
 	}
+	bh_unlock_sock_on_nolock(sk);
 }
 
 struct proto smc_proto = {
-- 
2.35.1