af_unix: Clean up unnecessary spin_lock(&sk->sk_peer_lock).
Subsystems (3)
cover af_unix: Clean up unnecessary spin_lock(&sk->sk_peer_lock).
Except for updating sk_peer_pid/sk_peer_cred, we do not need to hold sk_peer_lock during initialisation.
This series removes unnecessary spin_lock() and spin_unlock() in such places.
Kuniyuki Iwashima (2):
af_unix: Set sk_peer_pid/sk_peer_cred locklessly for new socket.
af_unix: Don't hold client's sk_peer_lock in copy_peercred().
net/unix/af_unix.c | 31 ++++++++++--------------------- 1 file changed, 10 insertions(+), 21 deletions(-)
Signature · 1 line
2.30.2
PATCH 1/2 af_unix: Set sk_peer_pid/sk_peer_cred locklessly for new socket.
init_peercred() is called in 3 places:
1. socketpair() : both sockets
2. connect() : child socket
3. listen() : listening socket
The first two need not hold sk_peer_lock because no one can touch the socket.
Let's set cred/pid without holding lock for the two cases and rename the old init_peercred() to update_peercred() to properly reflect the use case.
--- net/unix/af_unix.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
net/unix/af_unix.c +8 -3
@@ -684,6 +684,12 @@ static void unix_release_sock(struct sock *sk, int embrion) }static void init_peercred (struct sock *sk) +{ + sk->sk_peer_pid =get_pid (task_tgid (current)); + sk->sk_peer_cred =get_current_cred (); +} + +static void update_peercred (struct sock *sk) {const struct cred *old_cred;struct pid *old_pid;
@@ -691,8 +697,7 @@ static void init_peercred(struct sock *sk)spin_lock (&sk->sk_peer_lock); old_pid = sk->sk_peer_pid; old_cred = sk->sk_peer_cred; - sk->sk_peer_pid =get_pid (task_tgid (current)); - sk->sk_peer_cred =get_current_cred (); +init_peercred (sk);spin_unlock (&sk->sk_peer_lock);put_pid (old_pid);
@@ -743,7 +748,7 @@ static int unix_listen(struct socket *sock, int backlog) sk->sk_max_ack_backlog = backlog; sk->sk_state =TCP_LISTEN ;/* set credentials so connect can copy them */ -init_peercred (sk); +update_peercred (sk); err =0 ; out_unlock:
Signature · 1 line
2.30.2
PATCH 2/2 af_unix: Don't hold client's sk_peer_lock in copy_peercred().
In reply to Kuniyuki IwashimaWhen (AF_UNIX, SOCK_STREAM) socket connect()s to a listening socket, the listener's sk_peer_pid/sk_peer_cred are copied to the client in copy_peercred().
Then, two sk_peer_locks are held there; one is client's and another is listener's. However, we need not hold the client's lock.
The only place where the client's sk_peer_pid/sk_peer_cred are set is copy_peercred(), and once they are set, they never change.
Let's not hold client's sk_peer_lock in copy_peercred().
--- net/unix/af_unix.c | 20 ++------------------ 1 file changed, 2 insertions(+), 18 deletions(-)
net/unix/af_unix.c +2 -18
@@ -706,26 +706,10 @@ static void update_peercred(struct sock *sk)static void copy_peercred (struct sock *sk,struct sock *peersk) { -const struct cred *old_cred; -struct pid *old_pid; - -if (sk < peersk) { -spin_lock (&sk->sk_peer_lock); -spin_lock_nested (&peersk->sk_peer_lock,SINGLE_DEPTH_NESTING ); - }else { -spin_lock (&peersk->sk_peer_lock); -spin_lock_nested (&sk->sk_peer_lock,SINGLE_DEPTH_NESTING ); - } - old_pid = sk->sk_peer_pid; - old_cred = sk->sk_peer_cred; - sk->sk_peer_pid =get_pid (peersk->sk_peer_pid); +spin_lock (&peersk->sk_peer_lock); + sk->sk_peer_pid =get_pid (peersk->sk_peer_pid); sk->sk_peer_cred =get_cred (peersk->sk_peer_cred); - -spin_unlock (&sk->sk_peer_lock);spin_unlock (&peersk->sk_peer_lock); - -put_pid (old_pid); -put_cred (old_cred); }static int unix_listen (struct socket *sock,int backlog)
Signature · 1 line
2.30.2
PATCH 2/2 af_unix: Don't hold client's sk_peer_lock in copy_peercred().
From: Kuniyuki Iwashima <[email protected]>
Date: Sun, 28 Jan 2024 02:37:32 -0800
Quoted 24 lines
When (
AF_UNIX,SOCK_STREAM) socket connect()s to a listening socket, the listener's sk_peer_pid/sk_peer_cred are copied to the client incopy_peercred().Then, two
sk_peer_locksare held there; one is client's and another is listener's. However, we need not hold the client's lock.The only place where the client's sk_peer_pid/sk_peer_cred are set is
copy_peercred(), and once they are set, they never change.Let's not hold client's
sk_peer_lockincopy_peercred().Signed-off-by Kuniyuki Iwashima <[email protected]>--- net/unix/af_unix.c | 20 ++------------------ 1 file changed, 2 insertions(+), 18 deletions(-)net/unix/af_unix.c +0
-0@@ -706,26 +706,10 @@ static void update_peercred(struct sock *sk)static void copy_peercred (struct sock *sk,struct sock *peersk)
net/unix/af_unix.c static void update_peercred(struct sock *sk) · line 717
Quoted 10 lines
net/unix/af_unix.c +0
-9{ -const struct cred *old_cred; -struct pid *old_pid; - -if (sk < peersk) { -spin_lock (&sk->sk_peer_lock); -spin_lock_nested (&peersk->sk_peer_lock,SINGLE_DEPTH_NESTING ); - }else { -spin_lock (&peersk->sk_peer_lock); -spin_lock_nested (&sk->sk_peer_lock,SINGLE_DEPTH_NESTING );
Somehow I forgot about that the client's sk_peer_pid could be fetched at the same time by SO_PEERCRED/SO_PEERPIDFD, so the lock is needed.
net/unix/af_unix.c static void update_peercred(struct sock *sk) · line 720
Quoted 3 lines
net/unix/af_unix.c +0
-3- } - old_pid = sk->sk_peer_pid; - old_cred = sk->sk_peer_cred;
but old_pid/old_cred are always NULL, so this can be removed.
Also, the locking order can be fixed as
peer (TCP_LISTEN) -> sk (TCP_ESTABLISHED)
and we can write a lock_cmp_fn for sk_peer_lock.
I'll do so in v2, so
pw-bot: cr
Quoted 16 lines
net/unix/af_unix.c +2
-6- sk->sk_peer_pid =get_pid (peersk->sk_peer_pid); +spin_lock (&peersk->sk_peer_lock); + sk->sk_peer_pid =get_pid (peersk->sk_peer_pid); sk->sk_peer_cred =get_cred (peersk->sk_peer_cred); - -spin_unlock (&sk->sk_peer_lock);spin_unlock (&peersk->sk_peer_lock); - -put_pid (old_pid); -put_cred (old_cred); }static int unix_listen (struct socket *sock,int backlog)Signature · 1 line
2.30.2