#define SYNC_INIT 1
#endif
-int sysctl_tcp_tw_recycle;
-int sysctl_tcp_max_tw_buckets = NR_FILE*2;
-
int sysctl_tcp_syncookies = SYNC_INIT;
int sysctl_tcp_abort_on_overflow;
-static void tcp_tw_schedule(struct inet_timewait_sock *tw, int timeo);
+struct inet_timewait_death_row tcp_death_row = {
+ .sysctl_max_tw_buckets = NR_FILE * 2,
+ .period = TCP_TIMEWAIT_LEN / INET_TWDR_TWKILL_SLOTS,
+ .death_lock = SPIN_LOCK_UNLOCKED,
+ .hashinfo = &tcp_hashinfo,
+ .tw_timer = TIMER_INITIALIZER(inet_twdr_hangman, 0,
+ (unsigned long)&tcp_death_row),
+ .twkill_work = __WORK_INITIALIZER(tcp_death_row.twkill_work,
+ inet_twdr_twkill_work,
+ &tcp_death_row),
+/* Short-time timewait calendar */
+
+ .twcal_hand = -1,
+ .twcal_timer = TIMER_INITIALIZER(inet_twdr_twcal_tick, 0,
+ (unsigned long)&tcp_death_row),
+};
+
+EXPORT_SYMBOL_GPL(tcp_death_row);
static __inline__ int tcp_in_window(u32 seq, u32 end_seq, u32 s_win, u32 e_win)
{
return (seq == e_win && seq == end_seq);
}
-/* New-style handling of TIME_WAIT sockets. */
-
-int tcp_tw_count;
-
/*
* * Main purpose of TIME-WAIT state is to close connection gracefully,
* when one of ends sits in LAST-ACK or CLOSING retransmitting FIN
if (!th->fin ||
TCP_SKB_CB(skb)->end_seq != tcptw->tw_rcv_nxt + 1) {
kill_with_rst:
- tcp_tw_deschedule(tw);
+ inet_twsk_deschedule(tw, &tcp_death_row);
inet_twsk_put(tw);
return TCP_TW_RST;
}
* do not undertsnad recycling in any case, it not
* a big problem in practice. --ANK */
if (tw->tw_family == AF_INET &&
- sysctl_tcp_tw_recycle && tcptw->tw_ts_recent_stamp &&
+ tcp_death_row.sysctl_tw_recycle && tcptw->tw_ts_recent_stamp &&
tcp_v4_tw_remember_stamp(tw))
- tcp_tw_schedule(tw, tw->tw_timeout);
+ inet_twsk_schedule(tw, &tcp_death_row, tw->tw_timeout,
+ TCP_TIMEWAIT_LEN);
else
- tcp_tw_schedule(tw, TCP_TIMEWAIT_LEN);
+ inet_twsk_schedule(tw, &tcp_death_row, TCP_TIMEWAIT_LEN,
+ TCP_TIMEWAIT_LEN);
return TCP_TW_ACK;
}
*/
if (sysctl_tcp_rfc1337 == 0) {
kill:
- tcp_tw_deschedule(tw);
+ inet_twsk_deschedule(tw, &tcp_death_row);
inet_twsk_put(tw);
return TCP_TW_SUCCESS;
}
}
- tcp_tw_schedule(tw, TCP_TIMEWAIT_LEN);
+ inet_twsk_schedule(tw, &tcp_death_row, TCP_TIMEWAIT_LEN,
+ TCP_TIMEWAIT_LEN);
if (tmp_opt.saw_tstamp) {
tcptw->tw_ts_recent = tmp_opt.rcv_tsval;
* Do not reschedule in the last case.
*/
if (paws_reject || th->ack)
- tcp_tw_schedule(tw, TCP_TIMEWAIT_LEN);
+ inet_twsk_schedule(tw, &tcp_death_row, TCP_TIMEWAIT_LEN,
+ TCP_TIMEWAIT_LEN);
/* Send ACK. Note, we do not put the bucket,
* it will be released by caller.
const struct tcp_sock *tp = tcp_sk(sk);
int recycle_ok = 0;
- if (sysctl_tcp_tw_recycle && tp->rx_opt.ts_recent_stamp)
+ if (tcp_death_row.sysctl_tw_recycle && tp->rx_opt.ts_recent_stamp)
recycle_ok = tp->af_specific->remember_stamp(sk);
- if (tcp_tw_count < sysctl_tcp_max_tw_buckets)
+ if (tcp_death_row.tw_count < tcp_death_row.sysctl_max_tw_buckets)
tw = inet_twsk_alloc(sk, state);
if (tw != NULL) {
struct tcp_timewait_sock *tcptw = tcp_twsk((struct sock *)tw);
- const int rto = (tp->rto << 2) - (tp->rto >> 1);
+ const struct inet_connection_sock *icsk = inet_csk(sk);
+ const int rto = (icsk->icsk_rto << 2) - (icsk->icsk_rto >> 1);
tw->tw_rcv_wscale = tp->rx_opt.rcv_wscale;
tcptw->tw_rcv_nxt = tp->rcv_nxt;
timeo = TCP_TIMEWAIT_LEN;
}
- tcp_tw_schedule(tw, timeo);
+ inet_twsk_schedule(tw, &tcp_death_row, timeo,
+ TCP_TIMEWAIT_LEN);
inet_twsk_put(tw);
} else {
/* Sorry, if we're out of memory, just CLOSE this
tcp_done(sk);
}
-/* Kill off TIME_WAIT sockets once their lifetime has expired. */
-static int tcp_tw_death_row_slot;
-
-static void tcp_twkill(unsigned long);
-
-/* TIME_WAIT reaping mechanism. */
-#define TCP_TWKILL_SLOTS 8 /* Please keep this a power of 2. */
-#define TCP_TWKILL_PERIOD (TCP_TIMEWAIT_LEN/TCP_TWKILL_SLOTS)
-
-#define TCP_TWKILL_QUOTA 100
-
-static struct hlist_head tcp_tw_death_row[TCP_TWKILL_SLOTS];
-static DEFINE_SPINLOCK(tw_death_lock);
-static struct timer_list tcp_tw_timer = TIMER_INITIALIZER(tcp_twkill, 0, 0);
-static void twkill_work(void *);
-static DECLARE_WORK(tcp_twkill_work, twkill_work, NULL);
-static u32 twkill_thread_slots;
-
-/* Returns non-zero if quota exceeded. */
-static int tcp_do_twkill_work(int slot, unsigned int quota)
-{
- struct inet_timewait_sock *tw;
- struct hlist_node *node;
- unsigned int killed;
- int ret;
-
- /* NOTE: compare this to previous version where lock
- * was released after detaching chain. It was racy,
- * because tw buckets are scheduled in not serialized context
- * in 2.3 (with netfilter), and with softnet it is common, because
- * soft irqs are not sequenced.
- */
- killed = 0;
- ret = 0;
-rescan:
- inet_twsk_for_each_inmate(tw, node, &tcp_tw_death_row[slot]) {
- __inet_twsk_del_dead_node(tw);
- spin_unlock(&tw_death_lock);
- __inet_twsk_kill(tw, &tcp_hashinfo);
- inet_twsk_put(tw);
- killed++;
- spin_lock(&tw_death_lock);
- if (killed > quota) {
- ret = 1;
- break;
- }
-
- /* While we dropped tw_death_lock, another cpu may have
- * killed off the next TW bucket in the list, therefore
- * do a fresh re-read of the hlist head node with the
- * lock reacquired. We still use the hlist traversal
- * macro in order to get the prefetches.
- */
- goto rescan;
- }
-
- tcp_tw_count -= killed;
- NET_ADD_STATS_BH(LINUX_MIB_TIMEWAITED, killed);
-
- return ret;
-}
-
-static void tcp_twkill(unsigned long dummy)
-{
- int need_timer, ret;
-
- spin_lock(&tw_death_lock);
-
- if (tcp_tw_count == 0)
- goto out;
-
- need_timer = 0;
- ret = tcp_do_twkill_work(tcp_tw_death_row_slot, TCP_TWKILL_QUOTA);
- if (ret) {
- twkill_thread_slots |= (1 << tcp_tw_death_row_slot);
- mb();
- schedule_work(&tcp_twkill_work);
- need_timer = 1;
- } else {
- /* We purged the entire slot, anything left? */
- if (tcp_tw_count)
- need_timer = 1;
- }
- tcp_tw_death_row_slot =
- ((tcp_tw_death_row_slot + 1) & (TCP_TWKILL_SLOTS - 1));
- if (need_timer)
- mod_timer(&tcp_tw_timer, jiffies + TCP_TWKILL_PERIOD);
-out:
- spin_unlock(&tw_death_lock);
-}
-
-extern void twkill_slots_invalid(void);
-
-static void twkill_work(void *dummy)
-{
- int i;
-
- if ((TCP_TWKILL_SLOTS - 1) > (sizeof(twkill_thread_slots) * 8))
- twkill_slots_invalid();
-
- while (twkill_thread_slots) {
- spin_lock_bh(&tw_death_lock);
- for (i = 0; i < TCP_TWKILL_SLOTS; i++) {
- if (!(twkill_thread_slots & (1 << i)))
- continue;
-
- while (tcp_do_twkill_work(i, TCP_TWKILL_QUOTA) != 0) {
- if (need_resched()) {
- spin_unlock_bh(&tw_death_lock);
- schedule();
- spin_lock_bh(&tw_death_lock);
- }
- }
-
- twkill_thread_slots &= ~(1 << i);
- }
- spin_unlock_bh(&tw_death_lock);
- }
-}
-
-/* These are always called from BH context. See callers in
- * tcp_input.c to verify this.
- */
-
-/* This is for handling early-kills of TIME_WAIT sockets. */
-void tcp_tw_deschedule(struct inet_timewait_sock *tw)
-{
- spin_lock(&tw_death_lock);
- if (inet_twsk_del_dead_node(tw)) {
- inet_twsk_put(tw);
- if (--tcp_tw_count == 0)
- del_timer(&tcp_tw_timer);
- }
- spin_unlock(&tw_death_lock);
- __inet_twsk_kill(tw, &tcp_hashinfo);
-}
-
-/* Short-time timewait calendar */
-
-static int tcp_twcal_hand = -1;
-static int tcp_twcal_jiffie;
-static void tcp_twcal_tick(unsigned long);
-static struct timer_list tcp_twcal_timer =
- TIMER_INITIALIZER(tcp_twcal_tick, 0, 0);
-static struct hlist_head tcp_twcal_row[TCP_TW_RECYCLE_SLOTS];
-
-static void tcp_tw_schedule(struct inet_timewait_sock *tw, const int timeo)
-{
- struct hlist_head *list;
- int slot;
-
- /* timeout := RTO * 3.5
- *
- * 3.5 = 1+2+0.5 to wait for two retransmits.
- *
- * RATIONALE: if FIN arrived and we entered TIME-WAIT state,
- * our ACK acking that FIN can be lost. If N subsequent retransmitted
- * FINs (or previous seqments) are lost (probability of such event
- * is p^(N+1), where p is probability to lose single packet and
- * time to detect the loss is about RTO*(2^N - 1) with exponential
- * backoff). Normal timewait length is calculated so, that we
- * waited at least for one retransmitted FIN (maximal RTO is 120sec).
- * [ BTW Linux. following BSD, violates this requirement waiting
- * only for 60sec, we should wait at least for 240 secs.
- * Well, 240 consumes too much of resources 8)
- * ]
- * This interval is not reduced to catch old duplicate and
- * responces to our wandering segments living for two MSLs.
- * However, if we use PAWS to detect
- * old duplicates, we can reduce the interval to bounds required
- * by RTO, rather than MSL. So, if peer understands PAWS, we
- * kill tw bucket after 3.5*RTO (it is important that this number
- * is greater than TS tick!) and detect old duplicates with help
- * of PAWS.
- */
- slot = (timeo + (1<<TCP_TW_RECYCLE_TICK) - 1) >> TCP_TW_RECYCLE_TICK;
-
- spin_lock(&tw_death_lock);
-
- /* Unlink it, if it was scheduled */
- if (inet_twsk_del_dead_node(tw))
- tcp_tw_count--;
- else
- atomic_inc(&tw->tw_refcnt);
-
- if (slot >= TCP_TW_RECYCLE_SLOTS) {
- /* Schedule to slow timer */
- if (timeo >= TCP_TIMEWAIT_LEN) {
- slot = TCP_TWKILL_SLOTS-1;
- } else {
- slot = (timeo + TCP_TWKILL_PERIOD-1) / TCP_TWKILL_PERIOD;
- if (slot >= TCP_TWKILL_SLOTS)
- slot = TCP_TWKILL_SLOTS-1;
- }
- tw->tw_ttd = jiffies + timeo;
- slot = (tcp_tw_death_row_slot + slot) & (TCP_TWKILL_SLOTS - 1);
- list = &tcp_tw_death_row[slot];
- } else {
- tw->tw_ttd = jiffies + (slot << TCP_TW_RECYCLE_TICK);
-
- if (tcp_twcal_hand < 0) {
- tcp_twcal_hand = 0;
- tcp_twcal_jiffie = jiffies;
- tcp_twcal_timer.expires = tcp_twcal_jiffie + (slot<<TCP_TW_RECYCLE_TICK);
- add_timer(&tcp_twcal_timer);
- } else {
- if (time_after(tcp_twcal_timer.expires, jiffies + (slot<<TCP_TW_RECYCLE_TICK)))
- mod_timer(&tcp_twcal_timer, jiffies + (slot<<TCP_TW_RECYCLE_TICK));
- slot = (tcp_twcal_hand + slot)&(TCP_TW_RECYCLE_SLOTS-1);
- }
- list = &tcp_twcal_row[slot];
- }
-
- hlist_add_head(&tw->tw_death_node, list);
-
- if (tcp_tw_count++ == 0)
- mod_timer(&tcp_tw_timer, jiffies+TCP_TWKILL_PERIOD);
- spin_unlock(&tw_death_lock);
-}
-
-void tcp_twcal_tick(unsigned long dummy)
-{
- int n, slot;
- unsigned long j;
- unsigned long now = jiffies;
- int killed = 0;
- int adv = 0;
-
- spin_lock(&tw_death_lock);
- if (tcp_twcal_hand < 0)
- goto out;
-
- slot = tcp_twcal_hand;
- j = tcp_twcal_jiffie;
-
- for (n=0; n<TCP_TW_RECYCLE_SLOTS; n++) {
- if (time_before_eq(j, now)) {
- struct hlist_node *node, *safe;
- struct inet_timewait_sock *tw;
-
- inet_twsk_for_each_inmate_safe(tw, node, safe,
- &tcp_twcal_row[slot]) {
- __inet_twsk_del_dead_node(tw);
- __inet_twsk_kill(tw, &tcp_hashinfo);
- inet_twsk_put(tw);
- killed++;
- }
- } else {
- if (!adv) {
- adv = 1;
- tcp_twcal_jiffie = j;
- tcp_twcal_hand = slot;
- }
-
- if (!hlist_empty(&tcp_twcal_row[slot])) {
- mod_timer(&tcp_twcal_timer, j);
- goto out;
- }
- }
- j += (1<<TCP_TW_RECYCLE_TICK);
- slot = (slot+1)&(TCP_TW_RECYCLE_SLOTS-1);
- }
- tcp_twcal_hand = -1;
-
-out:
- if ((tcp_tw_count -= killed) == 0)
- del_timer(&tcp_tw_timer);
- NET_ADD_STATS_BH(LINUX_MIB_TIMEWAITKILLED, killed);
- spin_unlock(&tw_death_lock);
-}
-
/* This is not only more efficient than what we used to do, it eliminates
* a lot of code duplication between IPv4/IPv6 SYN recv processing. -DaveM
*
*/
struct sock *tcp_create_openreq_child(struct sock *sk, struct request_sock *req, struct sk_buff *skb)
{
- /* allocate the newsk from the same slab of the master sock,
- * if not, at sk_free time we'll try to free it from the wrong
- * slabcache (i.e. is it TCPv4 or v6?), this is handled thru sk->sk_prot -acme */
- struct sock *newsk = sk_alloc(PF_INET, GFP_ATOMIC, sk->sk_prot, 0);
+ struct sock *newsk = inet_csk_clone(sk, req, GFP_ATOMIC);
- if(newsk != NULL) {
- struct inet_request_sock *ireq = inet_rsk(req);
+ if (newsk != NULL) {
+ const struct inet_request_sock *ireq = inet_rsk(req);
struct tcp_request_sock *treq = tcp_rsk(req);
- struct inet_sock *newinet = inet_sk(newsk);
+ struct inet_connection_sock *newicsk = inet_csk(sk);
struct tcp_sock *newtp;
- struct sk_filter *filter;
-
- memcpy(newsk, sk, sizeof(struct tcp_sock));
- newsk->sk_state = TCP_SYN_RECV;
-
- /* SANITY */
- sk_node_init(&newsk->sk_node);
- newinet->bind_hash = NULL;
-
- /* Clone the TCP header template */
- newinet->dport = ireq->rmt_port;
-
- sock_lock_init(newsk);
- bh_lock_sock(newsk);
-
- rwlock_init(&newsk->sk_dst_lock);
- newsk->sk_dst_cache = NULL;
- atomic_set(&newsk->sk_rmem_alloc, 0);
- skb_queue_head_init(&newsk->sk_receive_queue);
- atomic_set(&newsk->sk_wmem_alloc, 0);
- skb_queue_head_init(&newsk->sk_write_queue);
- atomic_set(&newsk->sk_omem_alloc, 0);
- newsk->sk_wmem_queued = 0;
- newsk->sk_forward_alloc = 0;
-
- sock_reset_flag(newsk, SOCK_DONE);
- newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
- newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
- newsk->sk_send_head = NULL;
- rwlock_init(&newsk->sk_callback_lock);
- skb_queue_head_init(&newsk->sk_error_queue);
- newsk->sk_write_space = sk_stream_write_space;
-
- if ((filter = newsk->sk_filter) != NULL)
- sk_filter_charge(newsk, filter);
-
- if (unlikely(xfrm_sk_clone_policy(newsk))) {
- /* It is still raw copy of parent, so invalidate
- * destructor and make plain sk_free() */
- newsk->sk_destruct = NULL;
- sk_free(newsk);
- return NULL;
- }
/* Now setup tcp_sock */
newtp = tcp_sk(newsk);
newtp->pred_flags = 0;
newtp->rcv_nxt = treq->rcv_isn + 1;
- newtp->snd_nxt = treq->snt_isn + 1;
- newtp->snd_una = treq->snt_isn + 1;
- newtp->snd_sml = treq->snt_isn + 1;
+ newtp->snd_nxt = newtp->snd_una = newtp->snd_sml = treq->snt_isn + 1;
tcp_prequeue_init(newtp);
tcp_init_wl(newtp, treq->snt_isn, treq->rcv_isn);
- newtp->retransmits = 0;
- newtp->backoff = 0;
newtp->srtt = 0;
newtp->mdev = TCP_TIMEOUT_INIT;
- newtp->rto = TCP_TIMEOUT_INIT;
+ newicsk->icsk_rto = TCP_TIMEOUT_INIT;
newtp->packets_out = 0;
newtp->left_out = 0;
newtp->frto_counter = 0;
newtp->frto_highmark = 0;
- newtp->ca_ops = &tcp_reno;
+ newicsk->icsk_ca_ops = &tcp_reno;
- tcp_set_ca_state(newtp, TCP_CA_Open);
+ tcp_set_ca_state(newsk, TCP_CA_Open);
tcp_init_xmit_timers(newsk);
skb_queue_head_init(&newtp->out_of_order_queue);
newtp->rcv_wup = treq->rcv_isn + 1;
newtp->rx_opt.dsack = 0;
newtp->rx_opt.eff_sacks = 0;
- newtp->probes_out = 0;
newtp->rx_opt.num_sacks = 0;
newtp->urg_data = 0;
- /* Deinitialize accept_queue to trap illegal accesses. */
- memset(&newtp->accept_queue, 0, sizeof(newtp->accept_queue));
-
- /* Back to base struct sock members. */
- newsk->sk_err = 0;
- newsk->sk_priority = 0;
- atomic_set(&newsk->sk_refcnt, 2);
-
- /*
- * Increment the counter in the same struct proto as the master
- * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
- * is the same as sk->sk_prot->socks, as this field was copied
- * with memcpy), same rationale as the first comment in this
- * function.
- *
- * This _changes_ the previous behaviour, where
- * tcp_create_openreq_child always was incrementing the
- * equivalent to tcp_prot->socks (inet_sock_nr), so this have
- * to be taken into account in all callers. -acme
- */
- sk_refcnt_debug_inc(newsk);
-
- atomic_inc(&tcp_sockets_allocated);
if (sock_flag(newsk, SOCK_KEEPOPEN))
- tcp_reset_keepalive_timer(newsk,
- keepalive_time_when(newtp));
- newsk->sk_socket = NULL;
- newsk->sk_sleep = NULL;
+ inet_csk_reset_keepalive_timer(newsk,
+ keepalive_time_when(newtp));
newtp->rx_opt.tstamp_ok = ireq->tstamp_ok;
if((newtp->rx_opt.sack_ok = ireq->sack_ok) != 0) {
newtp->tcp_header_len = sizeof(struct tcphdr);
}
if (skb->len >= TCP_MIN_RCVMSS+newtp->tcp_header_len)
- newtp->ack.last_seg_size = skb->len-newtp->tcp_header_len;
+ newicsk->icsk_ack.last_seg_size = skb->len - newtp->tcp_header_len;
newtp->rx_opt.mss_clamp = req->mss;
TCP_ECN_openreq_child(newtp, req);
if (newtp->ecn_flags&TCP_ECN_OK)
does sequence test, SYN is truncated, and thus we consider
it a bare ACK.
- If tp->defer_accept, we silently drop this bare ACK. Otherwise,
- we create an established connection. Both ends (listening sockets)
- accept the new incoming connection and try to talk to each other. 8-)
+ If icsk->icsk_accept_queue.rskq_defer_accept, we silently drop this
+ bare ACK. Otherwise, we create an established connection. Both
+ ends (listening sockets) accept the new incoming connection and try
+ to talk to each other. 8-)
Note: This case is both harmless, and rare. Possibility is about the
same as us discovering intelligent life on another plant tomorrow.
return NULL;
/* If TCP_DEFER_ACCEPT is set, drop bare ACK. */
- if (tp->defer_accept && TCP_SKB_CB(skb)->end_seq == tcp_rsk(req)->rcv_isn + 1) {
+ if (inet_csk(sk)->icsk_accept_queue.rskq_defer_accept &&
+ TCP_SKB_CB(skb)->end_seq == tcp_rsk(req)->rcv_isn + 1) {
inet_rsk(req)->acked = 1;
return NULL;
}
if (child == NULL)
goto listen_overflow;
- tcp_synq_unlink(tp, req, prev);
- tcp_synq_removed(sk, req);
+ inet_csk_reqsk_queue_unlink(sk, req, prev);
+ inet_csk_reqsk_queue_removed(sk, req);
- tcp_acceptq_queue(sk, req, child);
+ inet_csk_reqsk_queue_add(sk, req, child);
return child;
listen_overflow:
if (!(flg & TCP_FLAG_RST))
req->rsk_ops->send_reset(skb);
- tcp_synq_drop(sk, req, prev);
+ inet_csk_reqsk_queue_drop(sk, req, prev);
return NULL;
}
EXPORT_SYMBOL(tcp_child_process);
EXPORT_SYMBOL(tcp_create_openreq_child);
EXPORT_SYMBOL(tcp_timewait_state_process);
-EXPORT_SYMBOL(tcp_tw_deschedule);
projects / linux-2.6 / blobdiff