hashent ^= (laddr->s6_addr32[3] ^ faddr->s6_addr32[3]);
hashent ^= hashent>>16;
hashent ^= hashent>>8;
- return (hashent & (tcp_ehash_size - 1));
+ return (hashent & (tcp_hashinfo.ehash_size - 1));
}
static __inline__ int tcp_v6_sk_hashfn(struct sock *sk)
return tcp_v6_hashfn(laddr, lport, faddr, fport);
}
-static inline int tcp_v6_bind_conflict(struct sock *sk,
- struct tcp_bind_bucket *tb)
+static inline int tcp_v6_bind_conflict(const struct sock *sk,
+ const struct inet_bind_bucket *tb)
{
- struct sock *sk2;
- struct hlist_node *node;
+ const struct sock *sk2;
+ const struct hlist_node *node;
/* We must walk the whole port owner list in this case. -DaveM */
sk_for_each_bound(sk2, node, &tb->owners) {
*/
static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
{
- struct tcp_bind_hashbucket *head;
- struct tcp_bind_bucket *tb;
+ struct inet_bind_hashbucket *head;
+ struct inet_bind_bucket *tb;
struct hlist_node *node;
int ret;
int remaining = (high - low) + 1;
int rover;
- spin_lock(&tcp_portalloc_lock);
- if (tcp_port_rover < low)
+ spin_lock(&tcp_hashinfo.portalloc_lock);
+ if (tcp_hashinfo.port_rover < low)
rover = low;
else
- rover = tcp_port_rover;
+ rover = tcp_hashinfo.port_rover;
do { rover++;
if (rover > high)
rover = low;
- head = &tcp_bhash[tcp_bhashfn(rover)];
+ head = &tcp_hashinfo.bhash[inet_bhashfn(rover, tcp_hashinfo.bhash_size)];
spin_lock(&head->lock);
- tb_for_each(tb, node, &head->chain)
+ inet_bind_bucket_for_each(tb, node, &head->chain)
if (tb->port == rover)
goto next;
break;
next:
spin_unlock(&head->lock);
} while (--remaining > 0);
- tcp_port_rover = rover;
- spin_unlock(&tcp_portalloc_lock);
-
- /* Exhausted local port range during search? */
+ tcp_hashinfo.port_rover = rover;
+ spin_unlock(&tcp_hashinfo.portalloc_lock);
+
+ /* Exhausted local port range during search? It is not
+ * possible for us to be holding one of the bind hash
+ * locks if this test triggers, because if 'remaining'
+ * drops to zero, we broke out of the do/while loop at
+ * the top level, not from the 'break;' statement.
+ */
ret = 1;
- if (remaining <= 0)
+ if (unlikely(remaining <= 0))
goto fail;
/* OK, here is the one we will use. */
snum = rover;
} else {
- head = &tcp_bhash[tcp_bhashfn(snum)];
+ head = &tcp_hashinfo.bhash[inet_bhashfn(snum, tcp_hashinfo.bhash_size)];
spin_lock(&head->lock);
- tb_for_each(tb, node, &head->chain)
+ inet_bind_bucket_for_each(tb, node, &head->chain)
if (tb->port == snum)
goto tb_found;
}
}
tb_not_found:
ret = 1;
- if (!tb && (tb = tcp_bucket_create(head, snum)) == NULL)
- goto fail_unlock;
+ if (tb == NULL) {
+ tb = inet_bind_bucket_create(tcp_hashinfo.bind_bucket_cachep, head, snum);
+ if (tb == NULL)
+ goto fail_unlock;
+ }
if (hlist_empty(&tb->owners)) {
if (sk->sk_reuse && sk->sk_state != TCP_LISTEN)
tb->fastreuse = 1;
tb->fastreuse = 0;
success:
- if (!tcp_sk(sk)->bind_hash)
- tcp_bind_hash(sk, tb, snum);
- BUG_TRAP(tcp_sk(sk)->bind_hash == tb);
+ if (!inet_sk(sk)->bind_hash)
+ inet_bind_hash(sk, tb, snum);
+ BUG_TRAP(inet_sk(sk)->bind_hash == tb);
ret = 0;
fail_unlock:
BUG_TRAP(sk_unhashed(sk));
if (sk->sk_state == TCP_LISTEN) {
- list = &tcp_listening_hash[tcp_sk_listen_hashfn(sk)];
- lock = &tcp_lhash_lock;
- tcp_listen_wlock();
+ list = &tcp_hashinfo.listening_hash[inet_sk_listen_hashfn(sk)];
+ lock = &tcp_hashinfo.lhash_lock;
+ inet_listen_wlock(&tcp_hashinfo);
} else {
sk->sk_hashent = tcp_v6_sk_hashfn(sk);
- list = &tcp_ehash[sk->sk_hashent].chain;
- lock = &tcp_ehash[sk->sk_hashent].lock;
+ list = &tcp_hashinfo.ehash[sk->sk_hashent].chain;
+ lock = &tcp_hashinfo.ehash[sk->sk_hashent].lock;
write_lock(lock);
}
int score, hiscore;
hiscore=0;
- read_lock(&tcp_lhash_lock);
- sk_for_each(sk, node, &tcp_listening_hash[tcp_lhashfn(hnum)]) {
+ read_lock(&tcp_hashinfo.lhash_lock);
+ sk_for_each(sk, node, &tcp_hashinfo.listening_hash[inet_lhashfn(hnum)]) {
if (inet_sk(sk)->num == hnum && sk->sk_family == PF_INET6) {
struct ipv6_pinfo *np = inet6_sk(sk);
}
if (result)
sock_hold(result);
- read_unlock(&tcp_lhash_lock);
+ read_unlock(&tcp_hashinfo.lhash_lock);
return result;
}
struct in6_addr *daddr, u16 hnum,
int dif)
{
- struct tcp_ehash_bucket *head;
struct sock *sk;
- struct hlist_node *node;
- __u32 ports = TCP_COMBINED_PORTS(sport, hnum);
- int hash;
-
+ const struct hlist_node *node;
+ const __u32 ports = INET_COMBINED_PORTS(sport, hnum);
/* Optimize here for direct hit, only listening connections can
* have wildcards anyways.
*/
- hash = tcp_v6_hashfn(daddr, hnum, saddr, sport);
- head = &tcp_ehash[hash];
+ const int hash = tcp_v6_hashfn(daddr, hnum, saddr, sport);
+ struct inet_ehash_bucket *head = &tcp_hashinfo.ehash[hash];
+
read_lock(&head->lock);
sk_for_each(sk, node, &head->chain) {
/* For IPV6 do the cheaper port and family tests first. */
- if(TCP_IPV6_MATCH(sk, saddr, daddr, ports, dif))
+ if (INET6_MATCH(sk, saddr, daddr, ports, dif))
goto hit; /* You sunk my battleship! */
}
/* Must check for a TIME_WAIT'er before going to listener hash. */
- sk_for_each(sk, node, &(head + tcp_ehash_size)->chain) {
- /* FIXME: acme: check this... */
- struct tcp_tw_bucket *tw = (struct tcp_tw_bucket *)sk;
+ sk_for_each(sk, node, &(head + tcp_hashinfo.ehash_size)->chain) {
+ const struct inet_timewait_sock *tw = inet_twsk(sk);
if(*((__u32 *)&(tw->tw_dport)) == ports &&
sk->sk_family == PF_INET6) {
- if(ipv6_addr_equal(&tw->tw_v6_daddr, saddr) &&
- ipv6_addr_equal(&tw->tw_v6_rcv_saddr, daddr) &&
- (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif))
+ const struct tcp6_timewait_sock *tcp6tw = tcp6_twsk(sk);
+
+ if (ipv6_addr_equal(&tcp6tw->tw_v6_daddr, saddr) &&
+ ipv6_addr_equal(&tcp6tw->tw_v6_rcv_saddr, daddr) &&
+ (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif))
goto hit;
}
}
}
static int __tcp_v6_check_established(struct sock *sk, __u16 lport,
- struct tcp_tw_bucket **twp)
+ struct inet_timewait_sock **twp)
{
struct inet_sock *inet = inet_sk(sk);
struct ipv6_pinfo *np = inet6_sk(sk);
struct in6_addr *daddr = &np->rcv_saddr;
struct in6_addr *saddr = &np->daddr;
int dif = sk->sk_bound_dev_if;
- u32 ports = TCP_COMBINED_PORTS(inet->dport, lport);
- int hash = tcp_v6_hashfn(daddr, inet->num, saddr, inet->dport);
- struct tcp_ehash_bucket *head = &tcp_ehash[hash];
+ const u32 ports = INET_COMBINED_PORTS(inet->dport, lport);
+ const int hash = tcp_v6_hashfn(daddr, inet->num, saddr, inet->dport);
+ struct inet_ehash_bucket *head = &tcp_hashinfo.ehash[hash];
struct sock *sk2;
- struct hlist_node *node;
- struct tcp_tw_bucket *tw;
+ const struct hlist_node *node;
+ struct inet_timewait_sock *tw;
write_lock(&head->lock);
/* Check TIME-WAIT sockets first. */
- sk_for_each(sk2, node, &(head + tcp_ehash_size)->chain) {
- tw = (struct tcp_tw_bucket*)sk2;
+ sk_for_each(sk2, node, &(head + tcp_hashinfo.ehash_size)->chain) {
+ const struct tcp6_timewait_sock *tcp6tw = tcp6_twsk(sk2);
+
+ tw = inet_twsk(sk2);
if(*((__u32 *)&(tw->tw_dport)) == ports &&
sk2->sk_family == PF_INET6 &&
- ipv6_addr_equal(&tw->tw_v6_daddr, saddr) &&
- ipv6_addr_equal(&tw->tw_v6_rcv_saddr, daddr) &&
+ ipv6_addr_equal(&tcp6tw->tw_v6_daddr, saddr) &&
+ ipv6_addr_equal(&tcp6tw->tw_v6_rcv_saddr, daddr) &&
sk2->sk_bound_dev_if == sk->sk_bound_dev_if) {
+ const struct tcp_timewait_sock *tcptw = tcp_twsk(sk2);
struct tcp_sock *tp = tcp_sk(sk);
- if (tw->tw_ts_recent_stamp &&
- (!twp || (sysctl_tcp_tw_reuse &&
- xtime.tv_sec -
- tw->tw_ts_recent_stamp > 1))) {
+ if (tcptw->tw_ts_recent_stamp &&
+ (!twp ||
+ (sysctl_tcp_tw_reuse &&
+ xtime.tv_sec - tcptw->tw_ts_recent_stamp > 1))) {
/* See comment in tcp_ipv4.c */
- tp->write_seq = tw->tw_snd_nxt + 65535 + 2;
+ tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2;
if (!tp->write_seq)
tp->write_seq = 1;
- tp->rx_opt.ts_recent = tw->tw_ts_recent;
- tp->rx_opt.ts_recent_stamp = tw->tw_ts_recent_stamp;
+ tp->rx_opt.ts_recent = tcptw->tw_ts_recent;
+ tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp;
sock_hold(sk2);
goto unique;
} else
/* And established part... */
sk_for_each(sk2, node, &head->chain) {
- if(TCP_IPV6_MATCH(sk2, saddr, daddr, ports, dif))
+ if (INET6_MATCH(sk2, saddr, daddr, ports, dif))
goto not_unique;
}
tcp_tw_deschedule(tw);
NET_INC_STATS_BH(LINUX_MIB_TIMEWAITRECYCLED);
- tcp_tw_put(tw);
+ inet_twsk_put(tw);
}
return 0;
static int tcp_v6_hash_connect(struct sock *sk)
{
unsigned short snum = inet_sk(sk)->num;
- struct tcp_bind_hashbucket *head;
- struct tcp_bind_bucket *tb;
+ struct inet_bind_hashbucket *head;
+ struct inet_bind_bucket *tb;
int ret;
if (!snum) {
static u32 hint;
u32 offset = hint + tcpv6_port_offset(sk);
struct hlist_node *node;
- struct tcp_tw_bucket *tw = NULL;
+ struct inet_timewait_sock *tw = NULL;
local_bh_disable();
for (i = 1; i <= range; i++) {
port = low + (i + offset) % range;
- head = &tcp_bhash[tcp_bhashfn(port)];
+ head = &tcp_hashinfo.bhash[inet_bhashfn(port, tcp_hashinfo.bhash_size)];
spin_lock(&head->lock);
/* Does not bother with rcv_saddr checks,
* because the established check is already
* unique enough.
*/
- tb_for_each(tb, node, &head->chain) {
+ inet_bind_bucket_for_each(tb, node, &head->chain) {
if (tb->port == port) {
BUG_TRAP(!hlist_empty(&tb->owners));
if (tb->fastreuse >= 0)
}
}
- tb = tcp_bucket_create(head, port);
+ tb = inet_bind_bucket_create(tcp_hashinfo.bind_bucket_cachep, head, port);
if (!tb) {
spin_unlock(&head->lock);
break;
hint += i;
/* Head lock still held and bh's disabled */
- tcp_bind_hash(sk, tb, port);
+ inet_bind_hash(sk, tb, port);
if (sk_unhashed(sk)) {
inet_sk(sk)->sport = htons(port);
__tcp_v6_hash(sk);
if (tw) {
tcp_tw_deschedule(tw);
- tcp_tw_put(tw);
+ inet_twsk_put(tw);
}
ret = 0;
goto out;
}
- head = &tcp_bhash[tcp_bhashfn(snum)];
- tb = tcp_sk(sk)->bind_hash;
+ head = &tcp_hashinfo.bhash[inet_bhashfn(snum, tcp_hashinfo.bhash_size)];
+ tb = inet_sk(sk)->bind_hash;
spin_lock_bh(&head->lock);
if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) {
}
if (sk->sk_state == TCP_TIME_WAIT) {
- tcp_tw_put((struct tcp_tw_bucket*)sk);
+ inet_twsk_put((struct inet_timewait_sock *)sk);
return;
}
static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
- struct tcp_tw_bucket *tw = (struct tcp_tw_bucket *)sk;
+ struct inet_timewait_sock *tw = inet_twsk(sk);
+ const struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
- tcp_v6_send_ack(skb, tw->tw_snd_nxt, tw->tw_rcv_nxt,
- tw->tw_rcv_wnd >> tw->tw_rcv_wscale, tw->tw_ts_recent);
+ tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
+ tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
+ tcptw->tw_ts_recent);
- tcp_tw_put(tw);
+ inet_twsk_put(tw);
}
static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req)
bh_lock_sock(nsk);
return nsk;
}
- tcp_tw_put((struct tcp_tw_bucket*)nsk);
+ inet_twsk_put((struct inet_timewait_sock *)nsk);
return NULL;
}
newnp->mcast_oif = tcp_v6_iif(skb);
newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
- /* Charge newly allocated IPv6 socket. Though it is mapped,
- * it is IPv6 yet.
+ /*
+ * No need to charge this sock to the relevant IPv6 refcnt debug socks count
+ * here, tcp_create_openreq_child now does this for us, see the comment in
+ * that function for the gory details. -acme
*/
-#ifdef INET_REFCNT_DEBUG
- atomic_inc(&inet6_sock_nr);
-#endif
/* It is tricky place. Until this moment IPv4 tcp
worked with IPv6 af_tcp.af_specific.
if (newsk == NULL)
goto out;
- /* Charge newly allocated IPv6 socket */
-#ifdef INET_REFCNT_DEBUG
- atomic_inc(&inet6_sock_nr);
-#endif
+ /*
+ * No need to charge this sock to the relevant IPv6 refcnt debug socks
+ * count here, tcp_create_openreq_child now does this for us, see the
+ * comment in that function for the gory details. -acme
+ */
ip6_dst_store(newsk, dst, NULL);
newsk->sk_route_caps = dst->dev->features &
newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
__tcp_v6_hash(newsk);
- tcp_inherit_port(sk, newsk);
+ inet_inherit_port(&tcp_hashinfo, sk, newsk);
return newsk;
do_time_wait:
if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
- tcp_tw_put((struct tcp_tw_bucket *) sk);
+ inet_twsk_put((struct inet_timewait_sock *)sk);
goto discard_it;
}
if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
TCP_INC_STATS_BH(TCP_MIB_INERRS);
- tcp_tw_put((struct tcp_tw_bucket *) sk);
+ inet_twsk_put((struct inet_timewait_sock *)sk);
goto discard_it;
}
- switch(tcp_timewait_state_process((struct tcp_tw_bucket *)sk,
- skb, th, skb->len)) {
+ switch (tcp_timewait_state_process((struct inet_timewait_sock *)sk,
+ skb, th)) {
case TCP_TW_SYN:
{
struct sock *sk2;
sk2 = tcp_v6_lookup_listener(&skb->nh.ipv6h->daddr, ntohs(th->dest), tcp_v6_iif(skb));
if (sk2 != NULL) {
- tcp_tw_deschedule((struct tcp_tw_bucket *)sk);
- tcp_tw_put((struct tcp_tw_bucket *)sk);
+ tcp_tw_deschedule((struct inet_timewait_sock *)sk);
+ inet_twsk_put((struct inet_timewait_sock *)sk);
sk = sk2;
goto process;
}
static struct tcp_func ipv6_mapped = {
.queue_xmit = ip_queue_xmit,
.send_check = tcp_v4_send_check,
- .rebuild_header = tcp_v4_rebuild_header,
+ .rebuild_header = inet_sk_rebuild_header,
.conn_request = tcp_v6_conn_request,
.syn_recv_sock = tcp_v6_syn_recv_sock,
.remember_stamp = tcp_v4_remember_stamp,
*/
tp->snd_ssthresh = 0x7fffffff;
tp->snd_cwnd_clamp = ~0;
- tp->mss_cache_std = tp->mss_cache = 536;
+ tp->mss_cache = 536;
tp->reordering = sysctl_tcp_reordering;
sk->sk_state = TCP_CLOSE;
tp->af_specific = &ipv6_specific;
-
+ tp->ca_ops = &tcp_init_congestion_ops;
sk->sk_write_space = sk_stream_write_space;
sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
}
static void get_timewait6_sock(struct seq_file *seq,
- struct tcp_tw_bucket *tw, int i)
+ struct inet_timewait_sock *tw, int i)
{
struct in6_addr *dest, *src;
__u16 destp, srcp;
+ struct tcp6_timewait_sock *tcp6tw = tcp6_twsk((struct sock *)tw);
int ttd = tw->tw_ttd - jiffies;
if (ttd < 0)
ttd = 0;
- dest = &tw->tw_v6_daddr;
- src = &tw->tw_v6_rcv_saddr;
+ dest = &tcp6tw->tw_v6_daddr;
+ src = &tcp6tw->tw_v6_rcv_saddr;
destp = ntohs(tw->tw_dport);
srcp = ntohs(tw->tw_sport);
.sysctl_rmem = sysctl_tcp_rmem,
.max_header = MAX_TCP_HEADER,
.obj_size = sizeof(struct tcp6_sock),
+ .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
.rsk_prot = &tcp6_request_sock_ops,
};
git://bbs.cooldavid.org / net-next-2.6.git / blobdiff