On Thu, 3 Dec 2020 11:18:44 +0300 Denis Kirjanov wrote: > in the case of a socket which is already bound to an adress > there is no sense to create a path in the next attempts
> diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c > index 41c3303c3357..489d49a1739c 100644 > --- a/net/unix/af_unix.c > +++ b/net/unix/af_unix.c > @@ -1029,6 +1029,16 @@ static int unix_bind(struct socket *sock, struct > sockaddr *uaddr, int addr_len) > goto out; > } > > + /* check if we're already bound to a path */ > + err = mutex_lock_interruptible(&u->bindlock); > + if (err) > + goto out; > + if (u->addr) > + err = -EINVAL; > + mutex_unlock(&u->bindlock); This, like v1, is not atomic with the creation, two threads can pass this check and then try to assign to u->addr. Naive question - can't we add the removal of the unnecessary node in the error path, in the kernel? > + if (err) > + goto out; > + > err = unix_mkname(sunaddr, addr_len, &hash); > if (err < 0) > goto out; > @@ -1049,10 +1059,6 @@ static int unix_bind(struct socket *sock, struct > sockaddr *uaddr, int addr_len) > if (err) > goto out_put; > > - err = -EINVAL; > - if (u->addr) > - goto out_up; > - > err = -ENOMEM; > addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL); > if (!addr)
