timewait sockets have a complex refcounting logic.
Once we realize it should be similar to established and
syn_recv sockets, we can use sk_nulls_del_node_init_rcu()
and remove inet_twsk_unhash()
In particular, deferred inet_twsk_put() added in commit
13475a30b66cd ("tcp: connect() race with timewait reuse")
looks unecessary : When removing a timewait socket from
ehash or bhash, caller must own a reference on the socket
anyway.
Signed-off-by: Eric Dumazet <eduma...@google.com>
---
include/net/inet_hashtables.h | 4 ++--
include/net/inet_timewait_sock.h | 6 ++----
net/ipv4/inet_hashtables.c | 31 ++++++++++-------------------
net/ipv4/inet_timewait_sock.c | 42 ++++++----------------------------------
net/ipv6/inet6_hashtables.c | 6 +-----
5 files changed, 21 insertions(+), 68 deletions(-)
diff --git a/include/net/inet_hashtables.h b/include/net/inet_hashtables.h
index b73c88a19dd4..b07d126694a7 100644
--- a/include/net/inet_hashtables.h
+++ b/include/net/inet_hashtables.h
@@ -205,8 +205,8 @@ void inet_put_port(struct sock *sk);
void inet_hashinfo_init(struct inet_hashinfo *h);
-int __inet_hash_nolisten(struct sock *sk, struct inet_timewait_sock *tw);
-int __inet_hash(struct sock *sk, struct inet_timewait_sock *tw);
+void __inet_hash_nolisten(struct sock *sk, struct sock *osk);
+void __inet_hash(struct sock *sk, struct sock *osk);
void inet_hash(struct sock *sk);
void inet_unhash(struct sock *sk);
diff --git a/include/net/inet_timewait_sock.h b/include/net/inet_timewait_sock.h
index 360c4802288d..96f52a4711c8 100644
--- a/include/net/inet_timewait_sock.h
+++ b/include/net/inet_timewait_sock.h
@@ -100,10 +100,8 @@ static inline struct inet_timewait_sock *inet_twsk(const
struct sock *sk)
void inet_twsk_free(struct inet_timewait_sock *tw);
void inet_twsk_put(struct inet_timewait_sock *tw);
-int inet_twsk_unhash(struct inet_timewait_sock *tw);
-
-int inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
- struct inet_hashinfo *hashinfo);
+void inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
+ struct inet_hashinfo *hashinfo);
struct inet_timewait_sock *inet_twsk_alloc(const struct sock *sk,
struct inet_timewait_death_row *dr,
diff --git a/net/ipv4/inet_hashtables.c b/net/ipv4/inet_hashtables.c
index 5f9b063bbe8a..e58840330da7 100644
--- a/net/ipv4/inet_hashtables.c
+++ b/net/ipv4/inet_hashtables.c
@@ -343,7 +343,6 @@ static int __inet_check_established(struct
inet_timewait_death_row *death_row,
struct sock *sk2;
const struct hlist_nulls_node *node;
struct inet_timewait_sock *tw = NULL;
- int twrefcnt = 0;
spin_lock(lock);
@@ -371,12 +370,10 @@ static int __inet_check_established(struct
inet_timewait_death_row *death_row,
WARN_ON(!sk_unhashed(sk));
__sk_nulls_add_node_rcu(sk, &head->chain);
if (tw) {
- twrefcnt = inet_twsk_unhash(tw);
+ sk_nulls_del_node_init_rcu((struct sock *)tw);
NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
}
spin_unlock(lock);
- if (twrefcnt)
- inet_twsk_put(tw);
sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
if (twp) {
@@ -384,7 +381,6 @@ static int __inet_check_established(struct
inet_timewait_death_row *death_row,
} else if (tw) {
/* Silly. Should hash-dance instead... */
inet_twsk_deschedule(tw);
-
inet_twsk_put(tw);
}
return 0;
@@ -403,13 +399,12 @@ static u32 inet_sk_port_offset(const struct sock *sk)
inet->inet_dport);
}
-int __inet_hash_nolisten(struct sock *sk, struct inet_timewait_sock *tw)
+void __inet_hash_nolisten(struct sock *sk, struct sock *osk)
{
struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
struct hlist_nulls_head *list;
struct inet_ehash_bucket *head;
spinlock_t *lock;
- int twrefcnt = 0;
WARN_ON(!sk_unhashed(sk));
@@ -420,23 +415,22 @@ int __inet_hash_nolisten(struct sock *sk, struct
inet_timewait_sock *tw)
spin_lock(lock);
__sk_nulls_add_node_rcu(sk, list);
- if (tw) {
- WARN_ON(sk->sk_hash != tw->tw_hash);
- twrefcnt = inet_twsk_unhash(tw);
+ if (osk) {
+ WARN_ON(sk->sk_hash != osk->sk_hash);
+ sk_nulls_del_node_init_rcu(osk);
}
spin_unlock(lock);
sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
- return twrefcnt;
}
EXPORT_SYMBOL_GPL(__inet_hash_nolisten);
-int __inet_hash(struct sock *sk, struct inet_timewait_sock *tw)
+void __inet_hash(struct sock *sk, struct sock *osk)
{
struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
struct inet_listen_hashbucket *ilb;
if (sk->sk_state != TCP_LISTEN)
- return __inet_hash_nolisten(sk, tw);
+ return __inet_hash_nolisten(sk, osk);
WARN_ON(!sk_unhashed(sk));
ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
@@ -445,7 +439,6 @@ int __inet_hash(struct sock *sk, struct inet_timewait_sock
*tw)
__sk_nulls_add_node_rcu(sk, &ilb->head);
sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
spin_unlock(&ilb->lock);
- return 0;
}
EXPORT_SYMBOL(__inet_hash);
@@ -492,7 +485,6 @@ int __inet_hash_connect(struct inet_timewait_death_row
*death_row,
struct inet_bind_bucket *tb;
int ret;
struct net *net = sock_net(sk);
- int twrefcnt = 1;
if (!snum) {
int i, remaining, low, high, port;
@@ -560,18 +552,15 @@ ok:
inet_bind_hash(sk, tb, port);
if (sk_unhashed(sk)) {
inet_sk(sk)->inet_sport = htons(port);
- twrefcnt += __inet_hash_nolisten(sk, tw);
+ __inet_hash_nolisten(sk, (struct sock *)tw);
}
if (tw)
- twrefcnt += inet_twsk_bind_unhash(tw, hinfo);
+ inet_twsk_bind_unhash(tw, hinfo);
spin_unlock(&head->lock);
if (tw) {
inet_twsk_deschedule(tw);
- while (twrefcnt) {
- twrefcnt--;
- inet_twsk_put(tw);
- }
+ inet_twsk_put(tw);
}
ret = 0;
diff --git a/net/ipv4/inet_timewait_sock.c b/net/ipv4/inet_timewait_sock.c
index 2ffbd16b79e0..92cd4d50404e 100644
--- a/net/ipv4/inet_timewait_sock.c
+++ b/net/ipv4/inet_timewait_sock.c
@@ -18,28 +18,6 @@
/**
- * inet_twsk_unhash - unhash a timewait socket from established hash
- * @tw: timewait socket
- *
- * unhash a timewait socket from established hash, if hashed.
- * ehash lock must be held by caller.
- * Returns 1 if caller should call inet_twsk_put() after lock release.
- */
-int inet_twsk_unhash(struct inet_timewait_sock *tw)
-{
- if (hlist_nulls_unhashed(&tw->tw_node))
- return 0;
-
- hlist_nulls_del_rcu(&tw->tw_node);
- sk_nulls_node_init(&tw->tw_node);
- /*
- * We cannot call inet_twsk_put() ourself under lock,
- * caller must call it for us.
- */
- return 1;
-}
-
-/**
* inet_twsk_bind_unhash - unhash a timewait socket from bind hash
* @tw: timewait socket
* @hashinfo: hashinfo pointer
@@ -48,35 +26,29 @@ int inet_twsk_unhash(struct inet_timewait_sock *tw)
* bind hash lock must be held by caller.
* Returns 1 if caller should call inet_twsk_put() after lock release.
*/
-int inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
+void inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
struct inet_hashinfo *hashinfo)
{
struct inet_bind_bucket *tb = tw->tw_tb;
if (!tb)
- return 0;
+ return;
__hlist_del(&tw->tw_bind_node);
tw->tw_tb = NULL;
inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
- /*
- * We cannot call inet_twsk_put() ourself under lock,
- * caller must call it for us.
- */
- return 1;
+ __sock_put((struct sock *)tw);
}
/* Must be called with locally disabled BHs. */
static void inet_twsk_kill(struct inet_timewait_sock *tw)
{
struct inet_hashinfo *hashinfo = tw->tw_dr->hashinfo;
- struct inet_bind_hashbucket *bhead;
- int refcnt;
- /* Unlink from established hashes. */
spinlock_t *lock = inet_ehash_lockp(hashinfo, tw->tw_hash);
+ struct inet_bind_hashbucket *bhead;
spin_lock(lock);
- refcnt = inet_twsk_unhash(tw);
+ sk_nulls_del_node_init_rcu((struct sock *)tw);
spin_unlock(lock);
/* Disassociate with bind bucket. */
@@ -84,11 +56,9 @@ static void inet_twsk_kill(struct inet_timewait_sock *tw)
hashinfo->bhash_size)];
spin_lock(&bhead->lock);
- refcnt += inet_twsk_bind_unhash(tw, hashinfo);
+ inet_twsk_bind_unhash(tw, hashinfo);
spin_unlock(&bhead->lock);
- BUG_ON(refcnt >= atomic_read(&tw->tw_refcnt));
- atomic_sub(refcnt, &tw->tw_refcnt);
atomic_dec(&tw->tw_dr->tw_count);
inet_twsk_put(tw);
}
diff --git a/net/ipv6/inet6_hashtables.c b/net/ipv6/inet6_hashtables.c
index b4fd96de97e6..a237398aa2b4 100644
--- a/net/ipv6/inet6_hashtables.c
+++ b/net/ipv6/inet6_hashtables.c
@@ -207,7 +207,6 @@ static int __inet6_check_established(struct
inet_timewait_death_row *death_row,
struct sock *sk2;
const struct hlist_nulls_node *node;
struct inet_timewait_sock *tw = NULL;
- int twrefcnt = 0;
spin_lock(lock);
@@ -234,12 +233,10 @@ static int __inet6_check_established(struct
inet_timewait_death_row *death_row,
WARN_ON(!sk_unhashed(sk));
__sk_nulls_add_node_rcu(sk, &head->chain);
if (tw) {
- twrefcnt = inet_twsk_unhash(tw);
+ sk_nulls_del_node_init_rcu((struct sock *)tw);
NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
}
spin_unlock(lock);
- if (twrefcnt)
- inet_twsk_put(tw);
sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
if (twp) {
@@ -247,7 +244,6 @@ static int __inet6_check_established(struct
inet_timewait_death_row *death_row,
} else if (tw) {
/* Silly. Should hash-dance instead... */
inet_twsk_deschedule(tw);
-
inet_twsk_put(tw);
}
return 0;
--
2.4.3.573.g4eafbef
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
