From: Eric Dumazet edumazet@google.com Date: Mon, 7 Apr 2025 13:21:11 +0200
diff --git a/net/core/sock.c b/net/core/sock.c index 323892066def..d426c5f8e20f 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -2130,6 +2130,8 @@ int sk_getsockopt(struct sock *sk, int level, int optname, */ static inline void sock_lock_init(struct sock *sk) {
sk_owner_clear(sk);if (sk->sk_kern_sock) sock_lock_init_class_and_name( sk,@@ -2324,6 +2326,8 @@ static void __sk_destruct(struct rcu_head *head) __netns_tracker_free(net, &sk->ns_tracker, false); net_passive_dec(net); }
sk_owner_put(sk);I am not convinced that the socket lock can be used after this point, now or in the future.
sk_prot_free(sk->sk_prot_creator, sk);}
Maybe move this in sk_prot_free() instead ?
Makes sense. Will move sk_owner_put() to sk_prot_free() in v3.
Thanks!
diff --git a/net/core/sock.c b/net/core/sock.c index 323892066def..9ab149d1584c 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -2226,6 +2226,9 @@ static void sk_prot_free(struct proto *prot, struct sock *sk) cgroup_sk_free(&sk->sk_cgrp_data); mem_cgroup_sk_free(sk); security_sk_free(sk);
sk_owner_put(sk);if (slab != NULL) kmem_cache_free(slab, sk); else