[PATCH net-next] net: __lock_sock() can be static

[PATCH net-next] net: __lock_sock() can be static

[Eric Dumazet]

After commit 6511882cdd82 ("mptcp: allocate fwd memory separately
on the rx and tx path") __lock_sock() can be static again.

Make sure __lock_sock() is not inlined, so that lock_sock_nested()
no longer needs a stack canary.

Add a noinline attribute on lock_sock_nested() so that calls
to lock_sock() from net/core/sock.c are not inlined,
none of them are fast path to deserve that:

 - sockopt_lock_sock()
 - sock_set_reuseport()
 - sock_set_reuseaddr()
 - sock_set_mark()
 - sock_set_keepalive()
 - sock_no_linger()
 - sock_bindtoindex()
 - sk_wait_data()
 - sock_set_rcvbuf()

$ scripts/bloat-o-meter -t vmlinux.old vmlinux
add/remove: 0/0 grow/shrink: 0/3 up/down: 0/-312 (-312)
Function                                     old     new   delta
__lock_sock                                  192     188      -4
__lock_sock_fast                             239      86    -153
lock_sock_nested                             227      72    -155
Total: Before=24888707, After=24888395, chg -0.00%

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/net/sock.h | 1 -
 net/core/sock.c    | 8 ++++----
 2 files changed, 4 insertions(+), 5 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 66b56288c1d3850439b2a0bed00be801d5770efa..55b61e4b0d8318887d527e919fc1103d78ac6d14 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1709,7 +1709,6 @@ static inline void lock_sock(struct sock *sk)
 	lock_sock_nested(sk, 0);
 }
 
-void __lock_sock(struct sock *sk);
 void __release_sock(struct sock *sk);
 void release_sock(struct sock *sk);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 693e6d80f501ef552aa58928f28b78a578169536..cfb2a6209946089669882cdbd5d1b36c53838989 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -3175,7 +3175,7 @@ bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
 }
 EXPORT_SYMBOL(sk_page_frag_refill);
 
-void __lock_sock(struct sock *sk)
+static void __lock_sock(struct sock *sk)
 	__releases(&sk->sk_lock.slock)
 	__acquires(&sk->sk_lock.slock)
 {
@@ -3774,14 +3774,14 @@ void sock_init_data(struct socket *sock, struct sock *sk)
 }
 EXPORT_SYMBOL(sock_init_data);
 
-void lock_sock_nested(struct sock *sk, int subclass)
+void noinline lock_sock_nested(struct sock *sk, int subclass)
 {
 	/* The sk_lock has mutex_lock() semantics here. */
 	mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
 
 	might_sleep();
 	spin_lock_bh(&sk->sk_lock.slock);
-	if (sock_owned_by_user_nocheck(sk))
+	if (unlikely(sock_owned_by_user_nocheck(sk)))
 		__lock_sock(sk);
 	sk->sk_lock.owned = 1;
 	spin_unlock_bh(&sk->sk_lock.slock);
@@ -3810,7 +3810,7 @@ bool __lock_sock_fast(struct sock *sk) __acquires(&sk->sk_lock.slock)
 	might_sleep();
 	spin_lock_bh(&sk->sk_lock.slock);
 
-	if (!sock_owned_by_user_nocheck(sk)) {
+	if (likely(!sock_owned_by_user_nocheck(sk))) {
 		/*
 		 * Fast path return with bottom halves disabled and
 		 * sock::sk_lock.slock held.
-- 
2.53.0.345.g96ddfc5eaa-goog

Re: [PATCH net-next] net: __lock_sock() can be static

[Kuniyuki Iwashima]

On Mon, Feb 23, 2026 at 1:27 AM Eric Dumazet <edumazet@google.com> wrote:
>
> After commit 6511882cdd82 ("mptcp: allocate fwd memory separately
> on the rx and tx path") __lock_sock() can be static again.
>
> Make sure __lock_sock() is not inlined, so that lock_sock_nested()
> no longer needs a stack canary.
>
> Add a noinline attribute on lock_sock_nested() so that calls
> to lock_sock() from net/core/sock.c are not inlined,
> none of them are fast path to deserve that:
>
>  - sockopt_lock_sock()
>  - sock_set_reuseport()
>  - sock_set_reuseaddr()
>  - sock_set_mark()
>  - sock_set_keepalive()
>  - sock_no_linger()
>  - sock_bindtoindex()
>  - sk_wait_data()
>  - sock_set_rcvbuf()
>
> $ scripts/bloat-o-meter -t vmlinux.old vmlinux
> add/remove: 0/0 grow/shrink: 0/3 up/down: 0/-312 (-312)
> Function                                     old     new   delta
> __lock_sock                                  192     188      -4
> __lock_sock_fast                             239      86    -153
> lock_sock_nested                             227      72    -155
> Total: Before=24888707, After=24888395, chg -0.00%
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>

Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>

Re: [PATCH net-next] net: __lock_sock() can be static

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Mon, 23 Feb 2026 09:27:15 +0000 you wrote:
> After commit 6511882cdd82 ("mptcp: allocate fwd memory separately
> on the rx and tx path") __lock_sock() can be static again.
> 
> Make sure __lock_sock() is not inlined, so that lock_sock_nested()
> no longer needs a stack canary.
> 
> Add a noinline attribute on lock_sock_nested() so that calls
> to lock_sock() from net/core/sock.c are not inlined,
> none of them are fast path to deserve that:
> 
> [...]

Here is the summary with links:
  - [net-next] net: __lock_sock() can be static
    https://git.kernel.org/netdev/net-next/c/2550def53bbf

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html