2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Implementation of the Transmission Control Protocol(TCP).
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Mark Evans, <evansmp@uhura.aston.ac.uk>
11 * Corey Minyard <wf-rch!minyard@relay.EU.net>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Charles Hedrick, <hedrick@klinzhai.rutgers.edu>
14 * Linus Torvalds, <torvalds@cs.helsinki.fi>
15 * Alan Cox, <gw4pts@gw4pts.ampr.org>
16 * Matthew Dillon, <dillon@apollo.west.oic.com>
17 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
18 * Jorge Cwik, <jorge@laser.satlink.net>
22 * Changes: Pedro Roque : Retransmit queue handled by TCP.
23 * : Fragmentation on mtu decrease
24 * : Segment collapse on retransmit
27 * Linus Torvalds : send_delayed_ack
28 * David S. Miller : Charge memory using the right skb
29 * during syn/ack processing.
30 * David S. Miller : Output engine completely rewritten.
31 * Andrea Arcangeli: SYNACK carry ts_recent in tsecr.
32 * Cacophonix Gaul : draft-minshall-nagle-01
33 * J Hadi Salim : ECN support
37 #define pr_fmt(fmt) "TCP: " fmt
41 #include <linux/compiler.h>
42 #include <linux/gfp.h>
43 #include <linux/module.h>
45 /* People can turn this off for buggy TCP's found in printers etc. */
46 int sysctl_tcp_retrans_collapse __read_mostly = 1;
48 /* People can turn this on to work with those rare, broken TCPs that
49 * interpret the window field as a signed quantity.
51 int sysctl_tcp_workaround_signed_windows __read_mostly = 0;
53 /* Default TSQ limit of two TSO segments */
54 int sysctl_tcp_limit_output_bytes __read_mostly = 131072;
56 /* This limits the percentage of the congestion window which we
57 * will allow a single TSO frame to consume. Building TSO frames
58 * which are too large can cause TCP streams to be bursty.
60 int sysctl_tcp_tso_win_divisor __read_mostly = 3;
62 int sysctl_tcp_mtu_probing __read_mostly = 0;
63 int sysctl_tcp_base_mss __read_mostly = TCP_BASE_MSS;
65 /* By default, RFC2861 behavior. */
66 int sysctl_tcp_slow_start_after_idle __read_mostly = 1;
68 unsigned int sysctl_tcp_notsent_lowat __read_mostly = UINT_MAX;
69 EXPORT_SYMBOL(sysctl_tcp_notsent_lowat);
71 static bool tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle,
72 int push_one, gfp_t gfp);
74 /* Account for new data that has been sent to the network. */
75 static void tcp_event_new_data_sent(struct sock *sk, const struct sk_buff *skb)
77 struct inet_connection_sock *icsk = inet_csk(sk);
78 struct tcp_sock *tp = tcp_sk(sk);
79 unsigned int prior_packets = tp->packets_out;
81 tcp_advance_send_head(sk, skb);
82 tp->snd_nxt = TCP_SKB_CB(skb)->end_seq;
84 tp->packets_out += tcp_skb_pcount(skb);
85 if (!prior_packets || icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
86 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
90 NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPORIGDATASENT,
94 /* SND.NXT, if window was not shrunk.
95 * If window has been shrunk, what should we make? It is not clear at all.
96 * Using SND.UNA we will fail to open window, SND.NXT is out of window. :-(
97 * Anything in between SND.UNA...SND.UNA+SND.WND also can be already
98 * invalid. OK, let's make this for now:
100 static inline __u32 tcp_acceptable_seq(const struct sock *sk)
102 const struct tcp_sock *tp = tcp_sk(sk);
104 if (!before(tcp_wnd_end(tp), tp->snd_nxt))
107 return tcp_wnd_end(tp);
110 /* Calculate mss to advertise in SYN segment.
111 * RFC1122, RFC1063, draft-ietf-tcpimpl-pmtud-01 state that:
113 * 1. It is independent of path mtu.
114 * 2. Ideally, it is maximal possible segment size i.e. 65535-40.
115 * 3. For IPv4 it is reasonable to calculate it from maximal MTU of
116 * attached devices, because some buggy hosts are confused by
118 * 4. We do not make 3, we advertise MSS, calculated from first
119 * hop device mtu, but allow to raise it to ip_rt_min_advmss.
120 * This may be overridden via information stored in routing table.
121 * 5. Value 65535 for MSS is valid in IPv6 and means "as large as possible,
122 * probably even Jumbo".
124 static __u16 tcp_advertise_mss(struct sock *sk)
126 struct tcp_sock *tp = tcp_sk(sk);
127 const struct dst_entry *dst = __sk_dst_get(sk);
128 int mss = tp->advmss;
131 unsigned int metric = dst_metric_advmss(dst);
142 /* RFC2861. Reset CWND after idle period longer RTO to "restart window".
143 * This is the first part of cwnd validation mechanism. */
144 static void tcp_cwnd_restart(struct sock *sk, const struct dst_entry *dst)
146 struct tcp_sock *tp = tcp_sk(sk);
147 s32 delta = tcp_time_stamp - tp->lsndtime;
148 u32 restart_cwnd = tcp_init_cwnd(tp, dst);
149 u32 cwnd = tp->snd_cwnd;
151 tcp_ca_event(sk, CA_EVENT_CWND_RESTART);
153 tp->snd_ssthresh = tcp_current_ssthresh(sk);
154 restart_cwnd = min(restart_cwnd, cwnd);
156 while ((delta -= inet_csk(sk)->icsk_rto) > 0 && cwnd > restart_cwnd)
158 tp->snd_cwnd = max(cwnd, restart_cwnd);
159 tp->snd_cwnd_stamp = tcp_time_stamp;
160 tp->snd_cwnd_used = 0;
163 /* Congestion state accounting after a packet has been sent. */
164 static void tcp_event_data_sent(struct tcp_sock *tp,
167 struct inet_connection_sock *icsk = inet_csk(sk);
168 const u32 now = tcp_time_stamp;
169 const struct dst_entry *dst = __sk_dst_get(sk);
171 if (sysctl_tcp_slow_start_after_idle &&
172 (!tp->packets_out && (s32)(now - tp->lsndtime) > icsk->icsk_rto))
173 tcp_cwnd_restart(sk, __sk_dst_get(sk));
177 /* If it is a reply for ato after last received
178 * packet, enter pingpong mode.
180 if ((u32)(now - icsk->icsk_ack.lrcvtime) < icsk->icsk_ack.ato &&
181 (!dst || !dst_metric(dst, RTAX_QUICKACK)))
182 icsk->icsk_ack.pingpong = 1;
185 /* Account for an ACK we sent. */
186 static inline void tcp_event_ack_sent(struct sock *sk, unsigned int pkts)
188 tcp_dec_quickack_mode(sk, pkts);
189 inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
193 u32 tcp_default_init_rwnd(u32 mss)
195 /* Initial receive window should be twice of TCP_INIT_CWND to
196 * enable proper sending of new unsent data during fast recovery
197 * (RFC 3517, Section 4, NextSeg() rule (2)). Further place a
198 * limit when mss is larger than 1460.
200 u32 init_rwnd = TCP_INIT_CWND * 2;
203 init_rwnd = max((1460 * init_rwnd) / mss, 2U);
207 /* Determine a window scaling and initial window to offer.
208 * Based on the assumption that the given amount of space
209 * will be offered. Store the results in the tp structure.
210 * NOTE: for smooth operation initial space offering should
211 * be a multiple of mss if possible. We assume here that mss >= 1.
212 * This MUST be enforced by all callers.
214 void tcp_select_initial_window(int __space, __u32 mss,
215 __u32 *rcv_wnd, __u32 *window_clamp,
216 int wscale_ok, __u8 *rcv_wscale,
219 unsigned int space = (__space < 0 ? 0 : __space);
221 /* If no clamp set the clamp to the max possible scaled window */
222 if (*window_clamp == 0)
223 (*window_clamp) = (65535 << 14);
224 space = min(*window_clamp, space);
226 /* Quantize space offering to a multiple of mss if possible. */
228 space = (space / mss) * mss;
230 /* NOTE: offering an initial window larger than 32767
231 * will break some buggy TCP stacks. If the admin tells us
232 * it is likely we could be speaking with such a buggy stack
233 * we will truncate our initial window offering to 32K-1
234 * unless the remote has sent us a window scaling option,
235 * which we interpret as a sign the remote TCP is not
236 * misinterpreting the window field as a signed quantity.
238 if (sysctl_tcp_workaround_signed_windows)
239 (*rcv_wnd) = min(space, MAX_TCP_WINDOW);
245 /* Set window scaling on max possible window
246 * See RFC1323 for an explanation of the limit to 14
248 space = max_t(u32, sysctl_tcp_rmem[2], sysctl_rmem_max);
249 space = min_t(u32, space, *window_clamp);
250 while (space > 65535 && (*rcv_wscale) < 14) {
256 if (mss > (1 << *rcv_wscale)) {
257 if (!init_rcv_wnd) /* Use default unless specified otherwise */
258 init_rcv_wnd = tcp_default_init_rwnd(mss);
259 *rcv_wnd = min(*rcv_wnd, init_rcv_wnd * mss);
262 /* Set the clamp no higher than max representable value */
263 (*window_clamp) = min(65535U << (*rcv_wscale), *window_clamp);
265 EXPORT_SYMBOL(tcp_select_initial_window);
267 /* Chose a new window to advertise, update state in tcp_sock for the
268 * socket, and return result with RFC1323 scaling applied. The return
269 * value can be stuffed directly into th->window for an outgoing
272 static u16 tcp_select_window(struct sock *sk)
274 struct tcp_sock *tp = tcp_sk(sk);
275 u32 old_win = tp->rcv_wnd;
276 u32 cur_win = tcp_receive_window(tp);
277 u32 new_win = __tcp_select_window(sk);
279 /* Never shrink the offered window */
280 if (new_win < cur_win) {
281 /* Danger Will Robinson!
282 * Don't update rcv_wup/rcv_wnd here or else
283 * we will not be able to advertise a zero
284 * window in time. --DaveM
286 * Relax Will Robinson.
289 NET_INC_STATS(sock_net(sk),
290 LINUX_MIB_TCPWANTZEROWINDOWADV);
291 new_win = ALIGN(cur_win, 1 << tp->rx_opt.rcv_wscale);
293 tp->rcv_wnd = new_win;
294 tp->rcv_wup = tp->rcv_nxt;
296 /* Make sure we do not exceed the maximum possible
299 if (!tp->rx_opt.rcv_wscale && sysctl_tcp_workaround_signed_windows)
300 new_win = min(new_win, MAX_TCP_WINDOW);
302 new_win = min(new_win, (65535U << tp->rx_opt.rcv_wscale));
304 /* RFC1323 scaling applied */
305 new_win >>= tp->rx_opt.rcv_wscale;
307 /* If we advertise zero window, disable fast path. */
311 NET_INC_STATS(sock_net(sk),
312 LINUX_MIB_TCPTOZEROWINDOWADV);
313 } else if (old_win == 0) {
314 NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPFROMZEROWINDOWADV);
320 /* Packet ECN state for a SYN-ACK */
321 static inline void TCP_ECN_send_synack(const struct tcp_sock *tp, struct sk_buff *skb)
323 TCP_SKB_CB(skb)->tcp_flags &= ~TCPHDR_CWR;
324 if (!(tp->ecn_flags & TCP_ECN_OK))
325 TCP_SKB_CB(skb)->tcp_flags &= ~TCPHDR_ECE;
328 /* Packet ECN state for a SYN. */
329 static inline void TCP_ECN_send_syn(struct sock *sk, struct sk_buff *skb)
331 struct tcp_sock *tp = tcp_sk(sk);
334 if (sock_net(sk)->ipv4.sysctl_tcp_ecn == 1) {
335 TCP_SKB_CB(skb)->tcp_flags |= TCPHDR_ECE | TCPHDR_CWR;
336 tp->ecn_flags = TCP_ECN_OK;
340 static __inline__ void
341 TCP_ECN_make_synack(const struct request_sock *req, struct tcphdr *th)
343 if (inet_rsk(req)->ecn_ok)
347 /* Set up ECN state for a packet on a ESTABLISHED socket that is about to
350 static inline void TCP_ECN_send(struct sock *sk, struct sk_buff *skb,
353 struct tcp_sock *tp = tcp_sk(sk);
355 if (tp->ecn_flags & TCP_ECN_OK) {
356 /* Not-retransmitted data segment: set ECT and inject CWR. */
357 if (skb->len != tcp_header_len &&
358 !before(TCP_SKB_CB(skb)->seq, tp->snd_nxt)) {
360 if (tp->ecn_flags & TCP_ECN_QUEUE_CWR) {
361 tp->ecn_flags &= ~TCP_ECN_QUEUE_CWR;
362 tcp_hdr(skb)->cwr = 1;
363 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
366 /* ACK or retransmitted segment: clear ECT|CE */
367 INET_ECN_dontxmit(sk);
369 if (tp->ecn_flags & TCP_ECN_DEMAND_CWR)
370 tcp_hdr(skb)->ece = 1;
374 /* Constructs common control bits of non-data skb. If SYN/FIN is present,
375 * auto increment end seqno.
377 static void tcp_init_nondata_skb(struct sk_buff *skb, u32 seq, u8 flags)
379 struct skb_shared_info *shinfo = skb_shinfo(skb);
381 skb->ip_summed = CHECKSUM_PARTIAL;
384 TCP_SKB_CB(skb)->tcp_flags = flags;
385 TCP_SKB_CB(skb)->sacked = 0;
387 shinfo->gso_segs = 1;
388 shinfo->gso_size = 0;
389 shinfo->gso_type = 0;
391 TCP_SKB_CB(skb)->seq = seq;
392 if (flags & (TCPHDR_SYN | TCPHDR_FIN))
394 TCP_SKB_CB(skb)->end_seq = seq;
397 static inline bool tcp_urg_mode(const struct tcp_sock *tp)
399 return tp->snd_una != tp->snd_up;
402 #define OPTION_SACK_ADVERTISE (1 << 0)
403 #define OPTION_TS (1 << 1)
404 #define OPTION_MD5 (1 << 2)
405 #define OPTION_WSCALE (1 << 3)
406 #define OPTION_FAST_OPEN_COOKIE (1 << 8)
408 struct tcp_out_options {
409 u16 options; /* bit field of OPTION_* */
410 u16 mss; /* 0 to disable */
411 u8 ws; /* window scale, 0 to disable */
412 u8 num_sack_blocks; /* number of SACK blocks to include */
413 u8 hash_size; /* bytes in hash_location */
414 __u8 *hash_location; /* temporary pointer, overloaded */
415 __u32 tsval, tsecr; /* need to include OPTION_TS */
416 struct tcp_fastopen_cookie *fastopen_cookie; /* Fast open cookie */
419 /* Write previously computed TCP options to the packet.
421 * Beware: Something in the Internet is very sensitive to the ordering of
422 * TCP options, we learned this through the hard way, so be careful here.
423 * Luckily we can at least blame others for their non-compliance but from
424 * inter-operability perspective it seems that we're somewhat stuck with
425 * the ordering which we have been using if we want to keep working with
426 * those broken things (not that it currently hurts anybody as there isn't
427 * particular reason why the ordering would need to be changed).
429 * At least SACK_PERM as the first option is known to lead to a disaster
430 * (but it may well be that other scenarios fail similarly).
432 static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp,
433 struct tcp_out_options *opts)
435 u16 options = opts->options; /* mungable copy */
437 if (unlikely(OPTION_MD5 & options)) {
438 *ptr++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
439 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
440 /* overload cookie hash location */
441 opts->hash_location = (__u8 *)ptr;
445 if (unlikely(opts->mss)) {
446 *ptr++ = htonl((TCPOPT_MSS << 24) |
447 (TCPOLEN_MSS << 16) |
451 if (likely(OPTION_TS & options)) {
452 if (unlikely(OPTION_SACK_ADVERTISE & options)) {
453 *ptr++ = htonl((TCPOPT_SACK_PERM << 24) |
454 (TCPOLEN_SACK_PERM << 16) |
455 (TCPOPT_TIMESTAMP << 8) |
457 options &= ~OPTION_SACK_ADVERTISE;
459 *ptr++ = htonl((TCPOPT_NOP << 24) |
461 (TCPOPT_TIMESTAMP << 8) |
464 *ptr++ = htonl(opts->tsval);
465 *ptr++ = htonl(opts->tsecr);
468 if (unlikely(OPTION_SACK_ADVERTISE & options)) {
469 *ptr++ = htonl((TCPOPT_NOP << 24) |
471 (TCPOPT_SACK_PERM << 8) |
475 if (unlikely(OPTION_WSCALE & options)) {
476 *ptr++ = htonl((TCPOPT_NOP << 24) |
477 (TCPOPT_WINDOW << 16) |
478 (TCPOLEN_WINDOW << 8) |
482 if (unlikely(opts->num_sack_blocks)) {
483 struct tcp_sack_block *sp = tp->rx_opt.dsack ?
484 tp->duplicate_sack : tp->selective_acks;
487 *ptr++ = htonl((TCPOPT_NOP << 24) |
490 (TCPOLEN_SACK_BASE + (opts->num_sack_blocks *
491 TCPOLEN_SACK_PERBLOCK)));
493 for (this_sack = 0; this_sack < opts->num_sack_blocks;
495 *ptr++ = htonl(sp[this_sack].start_seq);
496 *ptr++ = htonl(sp[this_sack].end_seq);
499 tp->rx_opt.dsack = 0;
502 if (unlikely(OPTION_FAST_OPEN_COOKIE & options)) {
503 struct tcp_fastopen_cookie *foc = opts->fastopen_cookie;
505 *ptr++ = htonl((TCPOPT_EXP << 24) |
506 ((TCPOLEN_EXP_FASTOPEN_BASE + foc->len) << 16) |
507 TCPOPT_FASTOPEN_MAGIC);
509 memcpy(ptr, foc->val, foc->len);
510 if ((foc->len & 3) == 2) {
511 u8 *align = ((u8 *)ptr) + foc->len;
512 align[0] = align[1] = TCPOPT_NOP;
514 ptr += (foc->len + 3) >> 2;
518 /* Compute TCP options for SYN packets. This is not the final
519 * network wire format yet.
521 static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb,
522 struct tcp_out_options *opts,
523 struct tcp_md5sig_key **md5)
525 struct tcp_sock *tp = tcp_sk(sk);
526 unsigned int remaining = MAX_TCP_OPTION_SPACE;
527 struct tcp_fastopen_request *fastopen = tp->fastopen_req;
529 #ifdef CONFIG_TCP_MD5SIG
530 *md5 = tp->af_specific->md5_lookup(sk, sk);
532 opts->options |= OPTION_MD5;
533 remaining -= TCPOLEN_MD5SIG_ALIGNED;
539 /* We always get an MSS option. The option bytes which will be seen in
540 * normal data packets should timestamps be used, must be in the MSS
541 * advertised. But we subtract them from tp->mss_cache so that
542 * calculations in tcp_sendmsg are simpler etc. So account for this
543 * fact here if necessary. If we don't do this correctly, as a
544 * receiver we won't recognize data packets as being full sized when we
545 * should, and thus we won't abide by the delayed ACK rules correctly.
546 * SACKs don't matter, we never delay an ACK when we have any of those
548 opts->mss = tcp_advertise_mss(sk);
549 remaining -= TCPOLEN_MSS_ALIGNED;
551 if (likely(sysctl_tcp_timestamps && *md5 == NULL)) {
552 opts->options |= OPTION_TS;
553 opts->tsval = TCP_SKB_CB(skb)->when + tp->tsoffset;
554 opts->tsecr = tp->rx_opt.ts_recent;
555 remaining -= TCPOLEN_TSTAMP_ALIGNED;
557 if (likely(sysctl_tcp_window_scaling)) {
558 opts->ws = tp->rx_opt.rcv_wscale;
559 opts->options |= OPTION_WSCALE;
560 remaining -= TCPOLEN_WSCALE_ALIGNED;
562 if (likely(sysctl_tcp_sack)) {
563 opts->options |= OPTION_SACK_ADVERTISE;
564 if (unlikely(!(OPTION_TS & opts->options)))
565 remaining -= TCPOLEN_SACKPERM_ALIGNED;
568 if (fastopen && fastopen->cookie.len >= 0) {
569 u32 need = TCPOLEN_EXP_FASTOPEN_BASE + fastopen->cookie.len;
570 need = (need + 3) & ~3U; /* Align to 32 bits */
571 if (remaining >= need) {
572 opts->options |= OPTION_FAST_OPEN_COOKIE;
573 opts->fastopen_cookie = &fastopen->cookie;
575 tp->syn_fastopen = 1;
579 return MAX_TCP_OPTION_SPACE - remaining;
582 /* Set up TCP options for SYN-ACKs. */
583 static unsigned int tcp_synack_options(struct sock *sk,
584 struct request_sock *req,
585 unsigned int mss, struct sk_buff *skb,
586 struct tcp_out_options *opts,
587 struct tcp_md5sig_key **md5,
588 struct tcp_fastopen_cookie *foc)
590 struct inet_request_sock *ireq = inet_rsk(req);
591 unsigned int remaining = MAX_TCP_OPTION_SPACE;
593 #ifdef CONFIG_TCP_MD5SIG
594 *md5 = tcp_rsk(req)->af_specific->md5_lookup(sk, req);
596 opts->options |= OPTION_MD5;
597 remaining -= TCPOLEN_MD5SIG_ALIGNED;
599 /* We can't fit any SACK blocks in a packet with MD5 + TS
600 * options. There was discussion about disabling SACK
601 * rather than TS in order to fit in better with old,
602 * buggy kernels, but that was deemed to be unnecessary.
604 ireq->tstamp_ok &= !ireq->sack_ok;
610 /* We always send an MSS option. */
612 remaining -= TCPOLEN_MSS_ALIGNED;
614 if (likely(ireq->wscale_ok)) {
615 opts->ws = ireq->rcv_wscale;
616 opts->options |= OPTION_WSCALE;
617 remaining -= TCPOLEN_WSCALE_ALIGNED;
619 if (likely(ireq->tstamp_ok)) {
620 opts->options |= OPTION_TS;
621 opts->tsval = TCP_SKB_CB(skb)->when;
622 opts->tsecr = req->ts_recent;
623 remaining -= TCPOLEN_TSTAMP_ALIGNED;
625 if (likely(ireq->sack_ok)) {
626 opts->options |= OPTION_SACK_ADVERTISE;
627 if (unlikely(!ireq->tstamp_ok))
628 remaining -= TCPOLEN_SACKPERM_ALIGNED;
630 if (foc != NULL && foc->len >= 0) {
631 u32 need = TCPOLEN_EXP_FASTOPEN_BASE + foc->len;
632 need = (need + 3) & ~3U; /* Align to 32 bits */
633 if (remaining >= need) {
634 opts->options |= OPTION_FAST_OPEN_COOKIE;
635 opts->fastopen_cookie = foc;
640 return MAX_TCP_OPTION_SPACE - remaining;
643 /* Compute TCP options for ESTABLISHED sockets. This is not the
644 * final wire format yet.
646 static unsigned int tcp_established_options(struct sock *sk, struct sk_buff *skb,
647 struct tcp_out_options *opts,
648 struct tcp_md5sig_key **md5)
650 struct tcp_skb_cb *tcb = skb ? TCP_SKB_CB(skb) : NULL;
651 struct tcp_sock *tp = tcp_sk(sk);
652 unsigned int size = 0;
653 unsigned int eff_sacks;
657 #ifdef CONFIG_TCP_MD5SIG
658 *md5 = tp->af_specific->md5_lookup(sk, sk);
659 if (unlikely(*md5)) {
660 opts->options |= OPTION_MD5;
661 size += TCPOLEN_MD5SIG_ALIGNED;
667 if (likely(tp->rx_opt.tstamp_ok)) {
668 opts->options |= OPTION_TS;
669 opts->tsval = tcb ? tcb->when + tp->tsoffset : 0;
670 opts->tsecr = tp->rx_opt.ts_recent;
671 size += TCPOLEN_TSTAMP_ALIGNED;
674 eff_sacks = tp->rx_opt.num_sacks + tp->rx_opt.dsack;
675 if (unlikely(eff_sacks)) {
676 const unsigned int remaining = MAX_TCP_OPTION_SPACE - size;
677 opts->num_sack_blocks =
678 min_t(unsigned int, eff_sacks,
679 (remaining - TCPOLEN_SACK_BASE_ALIGNED) /
680 TCPOLEN_SACK_PERBLOCK);
681 size += TCPOLEN_SACK_BASE_ALIGNED +
682 opts->num_sack_blocks * TCPOLEN_SACK_PERBLOCK;
689 /* TCP SMALL QUEUES (TSQ)
691 * TSQ goal is to keep small amount of skbs per tcp flow in tx queues (qdisc+dev)
692 * to reduce RTT and bufferbloat.
693 * We do this using a special skb destructor (tcp_wfree).
695 * Its important tcp_wfree() can be replaced by sock_wfree() in the event skb
696 * needs to be reallocated in a driver.
697 * The invariant being skb->truesize subtracted from sk->sk_wmem_alloc
699 * Since transmit from skb destructor is forbidden, we use a tasklet
700 * to process all sockets that eventually need to send more skbs.
701 * We use one tasklet per cpu, with its own queue of sockets.
704 struct tasklet_struct tasklet;
705 struct list_head head; /* queue of tcp sockets */
707 static DEFINE_PER_CPU(struct tsq_tasklet, tsq_tasklet);
709 static void tcp_tsq_handler(struct sock *sk)
711 if ((1 << sk->sk_state) &
712 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_CLOSING |
713 TCPF_CLOSE_WAIT | TCPF_LAST_ACK))
714 tcp_write_xmit(sk, tcp_current_mss(sk), tcp_sk(sk)->nonagle,
718 * One tasklet per cpu tries to send more skbs.
719 * We run in tasklet context but need to disable irqs when
720 * transferring tsq->head because tcp_wfree() might
721 * interrupt us (non NAPI drivers)
723 static void tcp_tasklet_func(unsigned long data)
725 struct tsq_tasklet *tsq = (struct tsq_tasklet *)data;
728 struct list_head *q, *n;
732 local_irq_save(flags);
733 list_splice_init(&tsq->head, &list);
734 local_irq_restore(flags);
736 list_for_each_safe(q, n, &list) {
737 tp = list_entry(q, struct tcp_sock, tsq_node);
738 list_del(&tp->tsq_node);
740 sk = (struct sock *)tp;
743 if (!sock_owned_by_user(sk)) {
746 /* defer the work to tcp_release_cb() */
747 set_bit(TCP_TSQ_DEFERRED, &tp->tsq_flags);
751 clear_bit(TSQ_QUEUED, &tp->tsq_flags);
756 #define TCP_DEFERRED_ALL ((1UL << TCP_TSQ_DEFERRED) | \
757 (1UL << TCP_WRITE_TIMER_DEFERRED) | \
758 (1UL << TCP_DELACK_TIMER_DEFERRED) | \
759 (1UL << TCP_MTU_REDUCED_DEFERRED))
761 * tcp_release_cb - tcp release_sock() callback
764 * called from release_sock() to perform protocol dependent
765 * actions before socket release.
767 void tcp_release_cb(struct sock *sk)
769 struct tcp_sock *tp = tcp_sk(sk);
770 unsigned long flags, nflags;
772 /* perform an atomic operation only if at least one flag is set */
774 flags = tp->tsq_flags;
775 if (!(flags & TCP_DEFERRED_ALL))
777 nflags = flags & ~TCP_DEFERRED_ALL;
778 } while (cmpxchg(&tp->tsq_flags, flags, nflags) != flags);
780 if (flags & (1UL << TCP_TSQ_DEFERRED))
783 /* Here begins the tricky part :
784 * We are called from release_sock() with :
786 * 2) sk_lock.slock spinlock held
787 * 3) socket owned by us (sk->sk_lock.owned == 1)
789 * But following code is meant to be called from BH handlers,
790 * so we should keep BH disabled, but early release socket ownership
792 sock_release_ownership(sk);
794 if (flags & (1UL << TCP_WRITE_TIMER_DEFERRED)) {
795 tcp_write_timer_handler(sk);
798 if (flags & (1UL << TCP_DELACK_TIMER_DEFERRED)) {
799 tcp_delack_timer_handler(sk);
802 if (flags & (1UL << TCP_MTU_REDUCED_DEFERRED)) {
803 inet_csk(sk)->icsk_af_ops->mtu_reduced(sk);
807 EXPORT_SYMBOL(tcp_release_cb);
809 void __init tcp_tasklet_init(void)
813 for_each_possible_cpu(i) {
814 struct tsq_tasklet *tsq = &per_cpu(tsq_tasklet, i);
816 INIT_LIST_HEAD(&tsq->head);
817 tasklet_init(&tsq->tasklet,
824 * Write buffer destructor automatically called from kfree_skb.
825 * We can't xmit new skbs from this context, as we might already
828 void tcp_wfree(struct sk_buff *skb)
830 struct sock *sk = skb->sk;
831 struct tcp_sock *tp = tcp_sk(sk);
833 if (test_and_clear_bit(TSQ_THROTTLED, &tp->tsq_flags) &&
834 !test_and_set_bit(TSQ_QUEUED, &tp->tsq_flags)) {
836 struct tsq_tasklet *tsq;
838 /* Keep a ref on socket.
839 * This last ref will be released in tcp_tasklet_func()
841 atomic_sub(skb->truesize - 1, &sk->sk_wmem_alloc);
843 /* queue this socket to tasklet queue */
844 local_irq_save(flags);
845 tsq = &__get_cpu_var(tsq_tasklet);
846 list_add(&tp->tsq_node, &tsq->head);
847 tasklet_schedule(&tsq->tasklet);
848 local_irq_restore(flags);
854 /* This routine actually transmits TCP packets queued in by
855 * tcp_do_sendmsg(). This is used by both the initial
856 * transmission and possible later retransmissions.
857 * All SKB's seen here are completely headerless. It is our
858 * job to build the TCP header, and pass the packet down to
859 * IP so it can do the same plus pass the packet off to the
862 * We are working here with either a clone of the original
863 * SKB, or a fresh unique copy made by the retransmit engine.
865 static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it,
868 const struct inet_connection_sock *icsk = inet_csk(sk);
869 struct inet_sock *inet;
871 struct tcp_skb_cb *tcb;
872 struct tcp_out_options opts;
873 unsigned int tcp_options_size, tcp_header_size;
874 struct tcp_md5sig_key *md5;
878 BUG_ON(!skb || !tcp_skb_pcount(skb));
881 skb_mstamp_get(&skb->skb_mstamp);
883 if (unlikely(skb_cloned(skb)))
884 skb = pskb_copy(skb, gfp_mask);
886 skb = skb_clone(skb, gfp_mask);
889 /* Our usage of tstamp should remain private */
890 skb->tstamp.tv64 = 0;
895 tcb = TCP_SKB_CB(skb);
896 memset(&opts, 0, sizeof(opts));
898 if (unlikely(tcb->tcp_flags & TCPHDR_SYN))
899 tcp_options_size = tcp_syn_options(sk, skb, &opts, &md5);
901 tcp_options_size = tcp_established_options(sk, skb, &opts,
903 tcp_header_size = tcp_options_size + sizeof(struct tcphdr);
905 if (tcp_packets_in_flight(tp) == 0)
906 tcp_ca_event(sk, CA_EVENT_TX_START);
908 /* if no packet is in qdisc/device queue, then allow XPS to select
911 skb->ooo_okay = sk_wmem_alloc_get(sk) == 0;
913 skb_push(skb, tcp_header_size);
914 skb_reset_transport_header(skb);
918 skb->destructor = tcp_wfree;
919 skb_set_hash_from_sk(skb, sk);
920 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
922 /* Build TCP header and checksum it. */
924 th->source = inet->inet_sport;
925 th->dest = inet->inet_dport;
926 th->seq = htonl(tcb->seq);
927 th->ack_seq = htonl(tp->rcv_nxt);
928 *(((__be16 *)th) + 6) = htons(((tcp_header_size >> 2) << 12) |
931 if (unlikely(tcb->tcp_flags & TCPHDR_SYN)) {
932 /* RFC1323: The window in SYN & SYN/ACK segments
935 th->window = htons(min(tp->rcv_wnd, 65535U));
937 th->window = htons(tcp_select_window(sk));
942 /* The urg_mode check is necessary during a below snd_una win probe */
943 if (unlikely(tcp_urg_mode(tp) && before(tcb->seq, tp->snd_up))) {
944 if (before(tp->snd_up, tcb->seq + 0x10000)) {
945 th->urg_ptr = htons(tp->snd_up - tcb->seq);
947 } else if (after(tcb->seq + 0xFFFF, tp->snd_nxt)) {
948 th->urg_ptr = htons(0xFFFF);
953 tcp_options_write((__be32 *)(th + 1), tp, &opts);
954 if (likely((tcb->tcp_flags & TCPHDR_SYN) == 0))
955 TCP_ECN_send(sk, skb, tcp_header_size);
957 #ifdef CONFIG_TCP_MD5SIG
958 /* Calculate the MD5 hash, as we have all we need now */
960 sk_nocaps_add(sk, NETIF_F_GSO_MASK);
961 tp->af_specific->calc_md5_hash(opts.hash_location,
966 icsk->icsk_af_ops->send_check(sk, skb);
968 if (likely(tcb->tcp_flags & TCPHDR_ACK))
969 tcp_event_ack_sent(sk, tcp_skb_pcount(skb));
971 if (skb->len != tcp_header_size)
972 tcp_event_data_sent(tp, sk);
974 if (after(tcb->end_seq, tp->snd_nxt) || tcb->seq == tcb->end_seq)
975 TCP_ADD_STATS(sock_net(sk), TCP_MIB_OUTSEGS,
976 tcp_skb_pcount(skb));
978 err = icsk->icsk_af_ops->queue_xmit(sk, skb, &inet->cork.fl);
979 if (likely(err <= 0))
984 return net_xmit_eval(err);
987 /* This routine just queues the buffer for sending.
989 * NOTE: probe0 timer is not checked, do not forget tcp_push_pending_frames,
990 * otherwise socket can stall.
992 static void tcp_queue_skb(struct sock *sk, struct sk_buff *skb)
994 struct tcp_sock *tp = tcp_sk(sk);
996 /* Advance write_seq and place onto the write_queue. */
997 tp->write_seq = TCP_SKB_CB(skb)->end_seq;
998 skb_header_release(skb);
999 tcp_add_write_queue_tail(sk, skb);
1000 sk->sk_wmem_queued += skb->truesize;
1001 sk_mem_charge(sk, skb->truesize);
1004 /* Initialize TSO segments for a packet. */
1005 static void tcp_set_skb_tso_segs(const struct sock *sk, struct sk_buff *skb,
1006 unsigned int mss_now)
1008 struct skb_shared_info *shinfo = skb_shinfo(skb);
1010 /* Make sure we own this skb before messing gso_size/gso_segs */
1011 WARN_ON_ONCE(skb_cloned(skb));
1013 if (skb->len <= mss_now || skb->ip_summed == CHECKSUM_NONE) {
1014 /* Avoid the costly divide in the normal
1017 shinfo->gso_segs = 1;
1018 shinfo->gso_size = 0;
1019 shinfo->gso_type = 0;
1021 shinfo->gso_segs = DIV_ROUND_UP(skb->len, mss_now);
1022 shinfo->gso_size = mss_now;
1023 shinfo->gso_type = sk->sk_gso_type;
1027 /* When a modification to fackets out becomes necessary, we need to check
1028 * skb is counted to fackets_out or not.
1030 static void tcp_adjust_fackets_out(struct sock *sk, const struct sk_buff *skb,
1033 struct tcp_sock *tp = tcp_sk(sk);
1035 if (!tp->sacked_out || tcp_is_reno(tp))
1038 if (after(tcp_highest_sack_seq(tp), TCP_SKB_CB(skb)->seq))
1039 tp->fackets_out -= decr;
1042 /* Pcount in the middle of the write queue got changed, we need to do various
1043 * tweaks to fix counters
1045 static void tcp_adjust_pcount(struct sock *sk, const struct sk_buff *skb, int decr)
1047 struct tcp_sock *tp = tcp_sk(sk);
1049 tp->packets_out -= decr;
1051 if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)
1052 tp->sacked_out -= decr;
1053 if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS)
1054 tp->retrans_out -= decr;
1055 if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST)
1056 tp->lost_out -= decr;
1058 /* Reno case is special. Sigh... */
1059 if (tcp_is_reno(tp) && decr > 0)
1060 tp->sacked_out -= min_t(u32, tp->sacked_out, decr);
1062 tcp_adjust_fackets_out(sk, skb, decr);
1064 if (tp->lost_skb_hint &&
1065 before(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(tp->lost_skb_hint)->seq) &&
1066 (tcp_is_fack(tp) || (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)))
1067 tp->lost_cnt_hint -= decr;
1069 tcp_verify_left_out(tp);
1072 static void tcp_fragment_tstamp(struct sk_buff *skb, struct sk_buff *skb2)
1074 struct skb_shared_info *shinfo = skb_shinfo(skb);
1076 if (unlikely(shinfo->tx_flags & SKBTX_ANY_TSTAMP) &&
1077 !before(shinfo->tskey, TCP_SKB_CB(skb2)->seq)) {
1078 struct skb_shared_info *shinfo2 = skb_shinfo(skb2);
1079 u8 tsflags = shinfo->tx_flags & SKBTX_ANY_TSTAMP;
1081 shinfo->tx_flags &= ~tsflags;
1082 shinfo2->tx_flags |= tsflags;
1083 swap(shinfo->tskey, shinfo2->tskey);
1087 /* Function to create two new TCP segments. Shrinks the given segment
1088 * to the specified size and appends a new segment with the rest of the
1089 * packet to the list. This won't be called frequently, I hope.
1090 * Remember, these are still headerless SKBs at this point.
1092 int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len,
1093 unsigned int mss_now, gfp_t gfp)
1095 struct tcp_sock *tp = tcp_sk(sk);
1096 struct sk_buff *buff;
1097 int nsize, old_factor;
1101 if (WARN_ON(len > skb->len))
1104 nsize = skb_headlen(skb) - len;
1108 if (skb_unclone(skb, gfp))
1111 /* Get a new skb... force flag on. */
1112 buff = sk_stream_alloc_skb(sk, nsize, gfp);
1114 return -ENOMEM; /* We'll just try again later. */
1116 sk->sk_wmem_queued += buff->truesize;
1117 sk_mem_charge(sk, buff->truesize);
1118 nlen = skb->len - len - nsize;
1119 buff->truesize += nlen;
1120 skb->truesize -= nlen;
1122 /* Correct the sequence numbers. */
1123 TCP_SKB_CB(buff)->seq = TCP_SKB_CB(skb)->seq + len;
1124 TCP_SKB_CB(buff)->end_seq = TCP_SKB_CB(skb)->end_seq;
1125 TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(buff)->seq;
1127 /* PSH and FIN should only be set in the second packet. */
1128 flags = TCP_SKB_CB(skb)->tcp_flags;
1129 TCP_SKB_CB(skb)->tcp_flags = flags & ~(TCPHDR_FIN | TCPHDR_PSH);
1130 TCP_SKB_CB(buff)->tcp_flags = flags;
1131 TCP_SKB_CB(buff)->sacked = TCP_SKB_CB(skb)->sacked;
1133 if (!skb_shinfo(skb)->nr_frags && skb->ip_summed != CHECKSUM_PARTIAL) {
1134 /* Copy and checksum data tail into the new buffer. */
1135 buff->csum = csum_partial_copy_nocheck(skb->data + len,
1136 skb_put(buff, nsize),
1141 skb->csum = csum_block_sub(skb->csum, buff->csum, len);
1143 skb->ip_summed = CHECKSUM_PARTIAL;
1144 skb_split(skb, buff, len);
1147 buff->ip_summed = skb->ip_summed;
1149 /* Looks stupid, but our code really uses when of
1150 * skbs, which it never sent before. --ANK
1152 TCP_SKB_CB(buff)->when = TCP_SKB_CB(skb)->when;
1153 buff->tstamp = skb->tstamp;
1154 tcp_fragment_tstamp(skb, buff);
1156 old_factor = tcp_skb_pcount(skb);
1158 /* Fix up tso_factor for both original and new SKB. */
1159 tcp_set_skb_tso_segs(sk, skb, mss_now);
1160 tcp_set_skb_tso_segs(sk, buff, mss_now);
1162 /* If this packet has been sent out already, we must
1163 * adjust the various packet counters.
1165 if (!before(tp->snd_nxt, TCP_SKB_CB(buff)->end_seq)) {
1166 int diff = old_factor - tcp_skb_pcount(skb) -
1167 tcp_skb_pcount(buff);
1170 tcp_adjust_pcount(sk, skb, diff);
1173 /* Link BUFF into the send queue. */
1174 skb_header_release(buff);
1175 tcp_insert_write_queue_after(skb, buff, sk);
1180 /* This is similar to __pskb_pull_head() (it will go to core/skbuff.c
1181 * eventually). The difference is that pulled data not copied, but
1182 * immediately discarded.
1184 static void __pskb_trim_head(struct sk_buff *skb, int len)
1186 struct skb_shared_info *shinfo;
1189 eat = min_t(int, len, skb_headlen(skb));
1191 __skb_pull(skb, eat);
1198 shinfo = skb_shinfo(skb);
1199 for (i = 0; i < shinfo->nr_frags; i++) {
1200 int size = skb_frag_size(&shinfo->frags[i]);
1203 skb_frag_unref(skb, i);
1206 shinfo->frags[k] = shinfo->frags[i];
1208 shinfo->frags[k].page_offset += eat;
1209 skb_frag_size_sub(&shinfo->frags[k], eat);
1215 shinfo->nr_frags = k;
1217 skb_reset_tail_pointer(skb);
1218 skb->data_len -= len;
1219 skb->len = skb->data_len;
1222 /* Remove acked data from a packet in the transmit queue. */
1223 int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
1225 if (skb_unclone(skb, GFP_ATOMIC))
1228 __pskb_trim_head(skb, len);
1230 TCP_SKB_CB(skb)->seq += len;
1231 skb->ip_summed = CHECKSUM_PARTIAL;
1233 skb->truesize -= len;
1234 sk->sk_wmem_queued -= len;
1235 sk_mem_uncharge(sk, len);
1236 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1238 /* Any change of skb->len requires recalculation of tso factor. */
1239 if (tcp_skb_pcount(skb) > 1)
1240 tcp_set_skb_tso_segs(sk, skb, tcp_skb_mss(skb));
1245 /* Calculate MSS not accounting any TCP options. */
1246 static inline int __tcp_mtu_to_mss(struct sock *sk, int pmtu)
1248 const struct tcp_sock *tp = tcp_sk(sk);
1249 const struct inet_connection_sock *icsk = inet_csk(sk);
1252 /* Calculate base mss without TCP options:
1253 It is MMS_S - sizeof(tcphdr) of rfc1122
1255 mss_now = pmtu - icsk->icsk_af_ops->net_header_len - sizeof(struct tcphdr);
1257 /* IPv6 adds a frag_hdr in case RTAX_FEATURE_ALLFRAG is set */
1258 if (icsk->icsk_af_ops->net_frag_header_len) {
1259 const struct dst_entry *dst = __sk_dst_get(sk);
1261 if (dst && dst_allfrag(dst))
1262 mss_now -= icsk->icsk_af_ops->net_frag_header_len;
1265 /* Clamp it (mss_clamp does not include tcp options) */
1266 if (mss_now > tp->rx_opt.mss_clamp)
1267 mss_now = tp->rx_opt.mss_clamp;
1269 /* Now subtract optional transport overhead */
1270 mss_now -= icsk->icsk_ext_hdr_len;
1272 /* Then reserve room for full set of TCP options and 8 bytes of data */
1278 /* Calculate MSS. Not accounting for SACKs here. */
1279 int tcp_mtu_to_mss(struct sock *sk, int pmtu)
1281 /* Subtract TCP options size, not including SACKs */
1282 return __tcp_mtu_to_mss(sk, pmtu) -
1283 (tcp_sk(sk)->tcp_header_len - sizeof(struct tcphdr));
1286 /* Inverse of above */
1287 int tcp_mss_to_mtu(struct sock *sk, int mss)
1289 const struct tcp_sock *tp = tcp_sk(sk);
1290 const struct inet_connection_sock *icsk = inet_csk(sk);
1294 tp->tcp_header_len +
1295 icsk->icsk_ext_hdr_len +
1296 icsk->icsk_af_ops->net_header_len;
1298 /* IPv6 adds a frag_hdr in case RTAX_FEATURE_ALLFRAG is set */
1299 if (icsk->icsk_af_ops->net_frag_header_len) {
1300 const struct dst_entry *dst = __sk_dst_get(sk);
1302 if (dst && dst_allfrag(dst))
1303 mtu += icsk->icsk_af_ops->net_frag_header_len;
1308 /* MTU probing init per socket */
1309 void tcp_mtup_init(struct sock *sk)
1311 struct tcp_sock *tp = tcp_sk(sk);
1312 struct inet_connection_sock *icsk = inet_csk(sk);
1314 icsk->icsk_mtup.enabled = sysctl_tcp_mtu_probing > 1;
1315 icsk->icsk_mtup.search_high = tp->rx_opt.mss_clamp + sizeof(struct tcphdr) +
1316 icsk->icsk_af_ops->net_header_len;
1317 icsk->icsk_mtup.search_low = tcp_mss_to_mtu(sk, sysctl_tcp_base_mss);
1318 icsk->icsk_mtup.probe_size = 0;
1320 EXPORT_SYMBOL(tcp_mtup_init);
1322 /* This function synchronize snd mss to current pmtu/exthdr set.
1324 tp->rx_opt.user_mss is mss set by user by TCP_MAXSEG. It does NOT counts
1325 for TCP options, but includes only bare TCP header.
1327 tp->rx_opt.mss_clamp is mss negotiated at connection setup.
1328 It is minimum of user_mss and mss received with SYN.
1329 It also does not include TCP options.
1331 inet_csk(sk)->icsk_pmtu_cookie is last pmtu, seen by this function.
1333 tp->mss_cache is current effective sending mss, including
1334 all tcp options except for SACKs. It is evaluated,
1335 taking into account current pmtu, but never exceeds
1336 tp->rx_opt.mss_clamp.
1338 NOTE1. rfc1122 clearly states that advertised MSS
1339 DOES NOT include either tcp or ip options.
1341 NOTE2. inet_csk(sk)->icsk_pmtu_cookie and tp->mss_cache
1342 are READ ONLY outside this function. --ANK (980731)
1344 unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu)
1346 struct tcp_sock *tp = tcp_sk(sk);
1347 struct inet_connection_sock *icsk = inet_csk(sk);
1350 if (icsk->icsk_mtup.search_high > pmtu)
1351 icsk->icsk_mtup.search_high = pmtu;
1353 mss_now = tcp_mtu_to_mss(sk, pmtu);
1354 mss_now = tcp_bound_to_half_wnd(tp, mss_now);
1356 /* And store cached results */
1357 icsk->icsk_pmtu_cookie = pmtu;
1358 if (icsk->icsk_mtup.enabled)
1359 mss_now = min(mss_now, tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_low));
1360 tp->mss_cache = mss_now;
1364 EXPORT_SYMBOL(tcp_sync_mss);
1366 /* Compute the current effective MSS, taking SACKs and IP options,
1367 * and even PMTU discovery events into account.
1369 unsigned int tcp_current_mss(struct sock *sk)
1371 const struct tcp_sock *tp = tcp_sk(sk);
1372 const struct dst_entry *dst = __sk_dst_get(sk);
1374 unsigned int header_len;
1375 struct tcp_out_options opts;
1376 struct tcp_md5sig_key *md5;
1378 mss_now = tp->mss_cache;
1381 u32 mtu = dst_mtu(dst);
1382 if (mtu != inet_csk(sk)->icsk_pmtu_cookie)
1383 mss_now = tcp_sync_mss(sk, mtu);
1386 header_len = tcp_established_options(sk, NULL, &opts, &md5) +
1387 sizeof(struct tcphdr);
1388 /* The mss_cache is sized based on tp->tcp_header_len, which assumes
1389 * some common options. If this is an odd packet (because we have SACK
1390 * blocks etc) then our calculated header_len will be different, and
1391 * we have to adjust mss_now correspondingly */
1392 if (header_len != tp->tcp_header_len) {
1393 int delta = (int) header_len - tp->tcp_header_len;
1400 /* RFC2861, slow part. Adjust cwnd, after it was not full during one rto.
1401 * As additional protections, we do not touch cwnd in retransmission phases,
1402 * and if application hit its sndbuf limit recently.
1404 static void tcp_cwnd_application_limited(struct sock *sk)
1406 struct tcp_sock *tp = tcp_sk(sk);
1408 if (inet_csk(sk)->icsk_ca_state == TCP_CA_Open &&
1409 sk->sk_socket && !test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1410 /* Limited by application or receiver window. */
1411 u32 init_win = tcp_init_cwnd(tp, __sk_dst_get(sk));
1412 u32 win_used = max(tp->snd_cwnd_used, init_win);
1413 if (win_used < tp->snd_cwnd) {
1414 tp->snd_ssthresh = tcp_current_ssthresh(sk);
1415 tp->snd_cwnd = (tp->snd_cwnd + win_used) >> 1;
1417 tp->snd_cwnd_used = 0;
1419 tp->snd_cwnd_stamp = tcp_time_stamp;
1422 static void tcp_cwnd_validate(struct sock *sk, bool is_cwnd_limited)
1424 struct tcp_sock *tp = tcp_sk(sk);
1426 /* Track the maximum number of outstanding packets in each
1427 * window, and remember whether we were cwnd-limited then.
1429 if (!before(tp->snd_una, tp->max_packets_seq) ||
1430 tp->packets_out > tp->max_packets_out) {
1431 tp->max_packets_out = tp->packets_out;
1432 tp->max_packets_seq = tp->snd_nxt;
1433 tp->is_cwnd_limited = is_cwnd_limited;
1436 if (tcp_is_cwnd_limited(sk)) {
1437 /* Network is feed fully. */
1438 tp->snd_cwnd_used = 0;
1439 tp->snd_cwnd_stamp = tcp_time_stamp;
1441 /* Network starves. */
1442 if (tp->packets_out > tp->snd_cwnd_used)
1443 tp->snd_cwnd_used = tp->packets_out;
1445 if (sysctl_tcp_slow_start_after_idle &&
1446 (s32)(tcp_time_stamp - tp->snd_cwnd_stamp) >= inet_csk(sk)->icsk_rto)
1447 tcp_cwnd_application_limited(sk);
1451 /* Minshall's variant of the Nagle send check. */
1452 static bool tcp_minshall_check(const struct tcp_sock *tp)
1454 return after(tp->snd_sml, tp->snd_una) &&
1455 !after(tp->snd_sml, tp->snd_nxt);
1458 /* Update snd_sml if this skb is under mss
1459 * Note that a TSO packet might end with a sub-mss segment
1460 * The test is really :
1461 * if ((skb->len % mss) != 0)
1462 * tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
1463 * But we can avoid doing the divide again given we already have
1464 * skb_pcount = skb->len / mss_now
1466 static void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss_now,
1467 const struct sk_buff *skb)
1469 if (skb->len < tcp_skb_pcount(skb) * mss_now)
1470 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
1473 /* Return false, if packet can be sent now without violation Nagle's rules:
1474 * 1. It is full sized. (provided by caller in %partial bool)
1475 * 2. Or it contains FIN. (already checked by caller)
1476 * 3. Or TCP_CORK is not set, and TCP_NODELAY is set.
1477 * 4. Or TCP_CORK is not set, and all sent packets are ACKed.
1478 * With Minshall's modification: all sent small packets are ACKed.
1480 static bool tcp_nagle_check(bool partial, const struct tcp_sock *tp,
1484 ((nonagle & TCP_NAGLE_CORK) ||
1485 (!nonagle && tp->packets_out && tcp_minshall_check(tp)));
1487 /* Returns the portion of skb which can be sent right away */
1488 static unsigned int tcp_mss_split_point(const struct sock *sk,
1489 const struct sk_buff *skb,
1490 unsigned int mss_now,
1491 unsigned int max_segs,
1494 const struct tcp_sock *tp = tcp_sk(sk);
1495 u32 partial, needed, window, max_len;
1497 window = tcp_wnd_end(tp) - TCP_SKB_CB(skb)->seq;
1498 max_len = mss_now * max_segs;
1500 if (likely(max_len <= window && skb != tcp_write_queue_tail(sk)))
1503 needed = min(skb->len, window);
1505 if (max_len <= needed)
1508 partial = needed % mss_now;
1509 /* If last segment is not a full MSS, check if Nagle rules allow us
1510 * to include this last segment in this skb.
1511 * Otherwise, we'll split the skb at last MSS boundary
1513 if (tcp_nagle_check(partial != 0, tp, nonagle))
1514 return needed - partial;
1519 /* Can at least one segment of SKB be sent right now, according to the
1520 * congestion window rules? If so, return how many segments are allowed.
1522 static inline unsigned int tcp_cwnd_test(const struct tcp_sock *tp,
1523 const struct sk_buff *skb)
1525 u32 in_flight, cwnd;
1527 /* Don't be strict about the congestion window for the final FIN. */
1528 if ((TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) &&
1529 tcp_skb_pcount(skb) == 1)
1532 in_flight = tcp_packets_in_flight(tp);
1533 cwnd = tp->snd_cwnd;
1534 if (in_flight < cwnd)
1535 return (cwnd - in_flight);
1540 /* Initialize TSO state of a skb.
1541 * This must be invoked the first time we consider transmitting
1542 * SKB onto the wire.
1544 static int tcp_init_tso_segs(const struct sock *sk, struct sk_buff *skb,
1545 unsigned int mss_now)
1547 int tso_segs = tcp_skb_pcount(skb);
1549 if (!tso_segs || (tso_segs > 1 && tcp_skb_mss(skb) != mss_now)) {
1550 tcp_set_skb_tso_segs(sk, skb, mss_now);
1551 tso_segs = tcp_skb_pcount(skb);
1557 /* Return true if the Nagle test allows this packet to be
1560 static inline bool tcp_nagle_test(const struct tcp_sock *tp, const struct sk_buff *skb,
1561 unsigned int cur_mss, int nonagle)
1563 /* Nagle rule does not apply to frames, which sit in the middle of the
1564 * write_queue (they have no chances to get new data).
1566 * This is implemented in the callers, where they modify the 'nonagle'
1567 * argument based upon the location of SKB in the send queue.
1569 if (nonagle & TCP_NAGLE_PUSH)
1572 /* Don't use the nagle rule for urgent data (or for the final FIN). */
1573 if (tcp_urg_mode(tp) || (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN))
1576 if (!tcp_nagle_check(skb->len < cur_mss, tp, nonagle))
1582 /* Does at least the first segment of SKB fit into the send window? */
1583 static bool tcp_snd_wnd_test(const struct tcp_sock *tp,
1584 const struct sk_buff *skb,
1585 unsigned int cur_mss)
1587 u32 end_seq = TCP_SKB_CB(skb)->end_seq;
1589 if (skb->len > cur_mss)
1590 end_seq = TCP_SKB_CB(skb)->seq + cur_mss;
1592 return !after(end_seq, tcp_wnd_end(tp));
1595 /* This checks if the data bearing packet SKB (usually tcp_send_head(sk))
1596 * should be put on the wire right now. If so, it returns the number of
1597 * packets allowed by the congestion window.
1599 static unsigned int tcp_snd_test(const struct sock *sk, struct sk_buff *skb,
1600 unsigned int cur_mss, int nonagle)
1602 const struct tcp_sock *tp = tcp_sk(sk);
1603 unsigned int cwnd_quota;
1605 tcp_init_tso_segs(sk, skb, cur_mss);
1607 if (!tcp_nagle_test(tp, skb, cur_mss, nonagle))
1610 cwnd_quota = tcp_cwnd_test(tp, skb);
1611 if (cwnd_quota && !tcp_snd_wnd_test(tp, skb, cur_mss))
1617 /* Test if sending is allowed right now. */
1618 bool tcp_may_send_now(struct sock *sk)
1620 const struct tcp_sock *tp = tcp_sk(sk);
1621 struct sk_buff *skb = tcp_send_head(sk);
1624 tcp_snd_test(sk, skb, tcp_current_mss(sk),
1625 (tcp_skb_is_last(sk, skb) ?
1626 tp->nonagle : TCP_NAGLE_PUSH));
1629 /* Trim TSO SKB to LEN bytes, put the remaining data into a new packet
1630 * which is put after SKB on the list. It is very much like
1631 * tcp_fragment() except that it may make several kinds of assumptions
1632 * in order to speed up the splitting operation. In particular, we
1633 * know that all the data is in scatter-gather pages, and that the
1634 * packet has never been sent out before (and thus is not cloned).
1636 static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len,
1637 unsigned int mss_now, gfp_t gfp)
1639 struct sk_buff *buff;
1640 int nlen = skb->len - len;
1643 /* All of a TSO frame must be composed of paged data. */
1644 if (skb->len != skb->data_len)
1645 return tcp_fragment(sk, skb, len, mss_now, gfp);
1647 buff = sk_stream_alloc_skb(sk, 0, gfp);
1648 if (unlikely(buff == NULL))
1651 sk->sk_wmem_queued += buff->truesize;
1652 sk_mem_charge(sk, buff->truesize);
1653 buff->truesize += nlen;
1654 skb->truesize -= nlen;
1656 /* Correct the sequence numbers. */
1657 TCP_SKB_CB(buff)->seq = TCP_SKB_CB(skb)->seq + len;
1658 TCP_SKB_CB(buff)->end_seq = TCP_SKB_CB(skb)->end_seq;
1659 TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(buff)->seq;
1661 /* PSH and FIN should only be set in the second packet. */
1662 flags = TCP_SKB_CB(skb)->tcp_flags;
1663 TCP_SKB_CB(skb)->tcp_flags = flags & ~(TCPHDR_FIN | TCPHDR_PSH);
1664 TCP_SKB_CB(buff)->tcp_flags = flags;
1666 /* This packet was never sent out yet, so no SACK bits. */
1667 TCP_SKB_CB(buff)->sacked = 0;
1669 buff->ip_summed = skb->ip_summed = CHECKSUM_PARTIAL;
1670 skb_split(skb, buff, len);
1671 tcp_fragment_tstamp(skb, buff);
1673 /* Fix up tso_factor for both original and new SKB. */
1674 tcp_set_skb_tso_segs(sk, skb, mss_now);
1675 tcp_set_skb_tso_segs(sk, buff, mss_now);
1677 /* Link BUFF into the send queue. */
1678 skb_header_release(buff);
1679 tcp_insert_write_queue_after(skb, buff, sk);
1684 /* Try to defer sending, if possible, in order to minimize the amount
1685 * of TSO splitting we do. View it as a kind of TSO Nagle test.
1687 * This algorithm is from John Heffner.
1689 static bool tcp_tso_should_defer(struct sock *sk, struct sk_buff *skb,
1690 bool *is_cwnd_limited)
1692 struct tcp_sock *tp = tcp_sk(sk);
1693 const struct inet_connection_sock *icsk = inet_csk(sk);
1694 u32 send_win, cong_win, limit, in_flight;
1697 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
1700 if (icsk->icsk_ca_state != TCP_CA_Open)
1703 /* Defer for less than two clock ticks. */
1704 if (tp->tso_deferred &&
1705 (((u32)jiffies << 1) >> 1) - (tp->tso_deferred >> 1) > 1)
1708 in_flight = tcp_packets_in_flight(tp);
1710 BUG_ON(tcp_skb_pcount(skb) <= 1 || (tp->snd_cwnd <= in_flight));
1712 send_win = tcp_wnd_end(tp) - TCP_SKB_CB(skb)->seq;
1714 /* From in_flight test above, we know that cwnd > in_flight. */
1715 cong_win = (tp->snd_cwnd - in_flight) * tp->mss_cache;
1717 limit = min(send_win, cong_win);
1719 /* If a full-sized TSO skb can be sent, do it. */
1720 if (limit >= min_t(unsigned int, sk->sk_gso_max_size,
1721 tp->xmit_size_goal_segs * tp->mss_cache))
1724 /* Middle in queue won't get any more data, full sendable already? */
1725 if ((skb != tcp_write_queue_tail(sk)) && (limit >= skb->len))
1728 win_divisor = ACCESS_ONCE(sysctl_tcp_tso_win_divisor);
1730 u32 chunk = min(tp->snd_wnd, tp->snd_cwnd * tp->mss_cache);
1732 /* If at least some fraction of a window is available,
1735 chunk /= win_divisor;
1739 /* Different approach, try not to defer past a single
1740 * ACK. Receiver should ACK every other full sized
1741 * frame, so if we have space for more than 3 frames
1744 if (limit > tcp_max_tso_deferred_mss(tp) * tp->mss_cache)
1748 /* Ok, it looks like it is advisable to defer.
1749 * Do not rearm the timer if already set to not break TCP ACK clocking.
1751 if (!tp->tso_deferred)
1752 tp->tso_deferred = 1 | (jiffies << 1);
1754 if (cong_win < send_win && cong_win < skb->len)
1755 *is_cwnd_limited = true;
1760 tp->tso_deferred = 0;
1764 /* Create a new MTU probe if we are ready.
1765 * MTU probe is regularly attempting to increase the path MTU by
1766 * deliberately sending larger packets. This discovers routing
1767 * changes resulting in larger path MTUs.
1769 * Returns 0 if we should wait to probe (no cwnd available),
1770 * 1 if a probe was sent,
1773 static int tcp_mtu_probe(struct sock *sk)
1775 struct tcp_sock *tp = tcp_sk(sk);
1776 struct inet_connection_sock *icsk = inet_csk(sk);
1777 struct sk_buff *skb, *nskb, *next;
1784 /* Not currently probing/verifying,
1786 * have enough cwnd, and
1787 * not SACKing (the variable headers throw things off) */
1788 if (!icsk->icsk_mtup.enabled ||
1789 icsk->icsk_mtup.probe_size ||
1790 inet_csk(sk)->icsk_ca_state != TCP_CA_Open ||
1791 tp->snd_cwnd < 11 ||
1792 tp->rx_opt.num_sacks || tp->rx_opt.dsack)
1795 /* Very simple search strategy: just double the MSS. */
1796 mss_now = tcp_current_mss(sk);
1797 probe_size = 2 * tp->mss_cache;
1798 size_needed = probe_size + (tp->reordering + 1) * tp->mss_cache;
1799 if (probe_size > tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_high)) {
1800 /* TODO: set timer for probe_converge_event */
1804 /* Have enough data in the send queue to probe? */
1805 if (tp->write_seq - tp->snd_nxt < size_needed)
1808 if (tp->snd_wnd < size_needed)
1810 if (after(tp->snd_nxt + size_needed, tcp_wnd_end(tp)))
1813 /* Do we need to wait to drain cwnd? With none in flight, don't stall */
1814 if (tcp_packets_in_flight(tp) + 2 > tp->snd_cwnd) {
1815 if (!tcp_packets_in_flight(tp))
1821 /* We're allowed to probe. Build it now. */
1822 if ((nskb = sk_stream_alloc_skb(sk, probe_size, GFP_ATOMIC)) == NULL)
1824 sk->sk_wmem_queued += nskb->truesize;
1825 sk_mem_charge(sk, nskb->truesize);
1827 skb = tcp_send_head(sk);
1829 TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(skb)->seq;
1830 TCP_SKB_CB(nskb)->end_seq = TCP_SKB_CB(skb)->seq + probe_size;
1831 TCP_SKB_CB(nskb)->tcp_flags = TCPHDR_ACK;
1832 TCP_SKB_CB(nskb)->sacked = 0;
1834 nskb->ip_summed = skb->ip_summed;
1836 tcp_insert_write_queue_before(nskb, skb, sk);
1839 tcp_for_write_queue_from_safe(skb, next, sk) {
1840 copy = min_t(int, skb->len, probe_size - len);
1841 if (nskb->ip_summed)
1842 skb_copy_bits(skb, 0, skb_put(nskb, copy), copy);
1844 nskb->csum = skb_copy_and_csum_bits(skb, 0,
1845 skb_put(nskb, copy),
1848 if (skb->len <= copy) {
1849 /* We've eaten all the data from this skb.
1851 TCP_SKB_CB(nskb)->tcp_flags |= TCP_SKB_CB(skb)->tcp_flags;
1852 tcp_unlink_write_queue(skb, sk);
1853 sk_wmem_free_skb(sk, skb);
1855 TCP_SKB_CB(nskb)->tcp_flags |= TCP_SKB_CB(skb)->tcp_flags &
1856 ~(TCPHDR_FIN|TCPHDR_PSH);
1857 if (!skb_shinfo(skb)->nr_frags) {
1858 skb_pull(skb, copy);
1859 if (skb->ip_summed != CHECKSUM_PARTIAL)
1860 skb->csum = csum_partial(skb->data,
1863 __pskb_trim_head(skb, copy);
1864 tcp_set_skb_tso_segs(sk, skb, mss_now);
1866 TCP_SKB_CB(skb)->seq += copy;
1871 if (len >= probe_size)
1874 tcp_init_tso_segs(sk, nskb, nskb->len);
1876 /* We're ready to send. If this fails, the probe will
1877 * be resegmented into mss-sized pieces by tcp_write_xmit(). */
1878 TCP_SKB_CB(nskb)->when = tcp_time_stamp;
1879 if (!tcp_transmit_skb(sk, nskb, 1, GFP_ATOMIC)) {
1880 /* Decrement cwnd here because we are sending
1881 * effectively two packets. */
1883 tcp_event_new_data_sent(sk, nskb);
1885 icsk->icsk_mtup.probe_size = tcp_mss_to_mtu(sk, nskb->len);
1886 tp->mtu_probe.probe_seq_start = TCP_SKB_CB(nskb)->seq;
1887 tp->mtu_probe.probe_seq_end = TCP_SKB_CB(nskb)->end_seq;
1895 /* This routine writes packets to the network. It advances the
1896 * send_head. This happens as incoming acks open up the remote
1899 * LARGESEND note: !tcp_urg_mode is overkill, only frames between
1900 * snd_up-64k-mss .. snd_up cannot be large. However, taking into
1901 * account rare use of URG, this is not a big flaw.
1903 * Send at most one packet when push_one > 0. Temporarily ignore
1904 * cwnd limit to force at most one packet out when push_one == 2.
1906 * Returns true, if no segments are in flight and we have queued segments,
1907 * but cannot send anything now because of SWS or another problem.
1909 static bool tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle,
1910 int push_one, gfp_t gfp)
1912 struct tcp_sock *tp = tcp_sk(sk);
1913 struct sk_buff *skb;
1914 unsigned int tso_segs, sent_pkts;
1917 bool is_cwnd_limited = false;
1922 /* Do MTU probing. */
1923 result = tcp_mtu_probe(sk);
1926 } else if (result > 0) {
1931 while ((skb = tcp_send_head(sk))) {
1934 tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
1937 if (unlikely(tp->repair) && tp->repair_queue == TCP_SEND_QUEUE) {
1938 /* "when" is used as a start point for the retransmit timer */
1939 TCP_SKB_CB(skb)->when = tcp_time_stamp;
1940 goto repair; /* Skip network transmission */
1943 cwnd_quota = tcp_cwnd_test(tp, skb);
1945 is_cwnd_limited = true;
1947 /* Force out a loss probe pkt. */
1953 if (unlikely(!tcp_snd_wnd_test(tp, skb, mss_now)))
1956 if (tso_segs == 1) {
1957 if (unlikely(!tcp_nagle_test(tp, skb, mss_now,
1958 (tcp_skb_is_last(sk, skb) ?
1959 nonagle : TCP_NAGLE_PUSH))))
1963 tcp_tso_should_defer(sk, skb, &is_cwnd_limited))
1967 /* TCP Small Queues :
1968 * Control number of packets in qdisc/devices to two packets / or ~1 ms.
1970 * - better RTT estimation and ACK scheduling
1973 * Alas, some drivers / subsystems require a fair amount
1974 * of queued bytes to ensure line rate.
1975 * One example is wifi aggregation (802.11 AMPDU)
1977 limit = max_t(unsigned int, sysctl_tcp_limit_output_bytes,
1978 sk->sk_pacing_rate >> 10);
1980 if (atomic_read(&sk->sk_wmem_alloc) > limit) {
1981 set_bit(TSQ_THROTTLED, &tp->tsq_flags);
1982 /* It is possible TX completion already happened
1983 * before we set TSQ_THROTTLED, so we must
1984 * test again the condition.
1986 smp_mb__after_atomic();
1987 if (atomic_read(&sk->sk_wmem_alloc) > limit)
1992 if (tso_segs > 1 && !tcp_urg_mode(tp))
1993 limit = tcp_mss_split_point(sk, skb, mss_now,
1996 sk->sk_gso_max_segs),
1999 if (skb->len > limit &&
2000 unlikely(tso_fragment(sk, skb, limit, mss_now, gfp)))
2003 TCP_SKB_CB(skb)->when = tcp_time_stamp;
2005 if (unlikely(tcp_transmit_skb(sk, skb, 1, gfp)))
2009 /* Advance the send_head. This one is sent out.
2010 * This call will increment packets_out.
2012 tcp_event_new_data_sent(sk, skb);
2014 tcp_minshall_update(tp, mss_now, skb);
2015 sent_pkts += tcp_skb_pcount(skb);
2021 if (likely(sent_pkts)) {
2022 if (tcp_in_cwnd_reduction(sk))
2023 tp->prr_out += sent_pkts;
2025 /* Send one loss probe per tail loss episode. */
2027 tcp_schedule_loss_probe(sk);
2028 tcp_cwnd_validate(sk, is_cwnd_limited);
2031 return (push_one == 2) || (!tp->packets_out && tcp_send_head(sk));
2034 bool tcp_schedule_loss_probe(struct sock *sk)
2036 struct inet_connection_sock *icsk = inet_csk(sk);
2037 struct tcp_sock *tp = tcp_sk(sk);
2038 u32 timeout, tlp_time_stamp, rto_time_stamp;
2039 u32 rtt = usecs_to_jiffies(tp->srtt_us >> 3);
2041 if (WARN_ON(icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS))
2043 /* No consecutive loss probes. */
2044 if (WARN_ON(icsk->icsk_pending == ICSK_TIME_LOSS_PROBE)) {
2048 /* Don't do any loss probe on a Fast Open connection before 3WHS
2051 if (sk->sk_state == TCP_SYN_RECV)
2054 /* TLP is only scheduled when next timer event is RTO. */
2055 if (icsk->icsk_pending != ICSK_TIME_RETRANS)
2058 /* Schedule a loss probe in 2*RTT for SACK capable connections
2059 * in Open state, that are either limited by cwnd or application.
2061 if (sysctl_tcp_early_retrans < 3 || !tp->srtt_us || !tp->packets_out ||
2062 !tcp_is_sack(tp) || inet_csk(sk)->icsk_ca_state != TCP_CA_Open)
2065 if ((tp->snd_cwnd > tcp_packets_in_flight(tp)) &&
2069 /* Probe timeout is at least 1.5*rtt + TCP_DELACK_MAX to account
2070 * for delayed ack when there's one outstanding packet.
2073 if (tp->packets_out == 1)
2074 timeout = max_t(u32, timeout,
2075 (rtt + (rtt >> 1) + TCP_DELACK_MAX));
2076 timeout = max_t(u32, timeout, msecs_to_jiffies(10));
2078 /* If RTO is shorter, just schedule TLP in its place. */
2079 tlp_time_stamp = tcp_time_stamp + timeout;
2080 rto_time_stamp = (u32)inet_csk(sk)->icsk_timeout;
2081 if ((s32)(tlp_time_stamp - rto_time_stamp) > 0) {
2082 s32 delta = rto_time_stamp - tcp_time_stamp;
2087 inet_csk_reset_xmit_timer(sk, ICSK_TIME_LOSS_PROBE, timeout,
2092 /* Thanks to skb fast clones, we can detect if a prior transmit of
2093 * a packet is still in a qdisc or driver queue.
2094 * In this case, there is very little point doing a retransmit !
2095 * Note: This is called from BH context only.
2097 static bool skb_still_in_host_queue(const struct sock *sk,
2098 const struct sk_buff *skb)
2100 const struct sk_buff *fclone = skb + 1;
2102 if (unlikely(skb->fclone == SKB_FCLONE_ORIG &&
2103 fclone->fclone == SKB_FCLONE_CLONE)) {
2104 NET_INC_STATS_BH(sock_net(sk),
2105 LINUX_MIB_TCPSPURIOUS_RTX_HOSTQUEUES);
2111 /* When probe timeout (PTO) fires, send a new segment if one exists, else
2112 * retransmit the last segment.
2114 void tcp_send_loss_probe(struct sock *sk)
2116 struct tcp_sock *tp = tcp_sk(sk);
2117 struct sk_buff *skb;
2119 int mss = tcp_current_mss(sk);
2122 if (tcp_send_head(sk) != NULL) {
2123 err = tcp_write_xmit(sk, mss, TCP_NAGLE_OFF, 2, GFP_ATOMIC);
2127 /* At most one outstanding TLP retransmission. */
2128 if (tp->tlp_high_seq)
2131 /* Retransmit last segment. */
2132 skb = tcp_write_queue_tail(sk);
2136 if (skb_still_in_host_queue(sk, skb))
2139 pcount = tcp_skb_pcount(skb);
2140 if (WARN_ON(!pcount))
2143 if ((pcount > 1) && (skb->len > (pcount - 1) * mss)) {
2144 if (unlikely(tcp_fragment(sk, skb, (pcount - 1) * mss, mss,
2147 skb = tcp_write_queue_tail(sk);
2150 if (WARN_ON(!skb || !tcp_skb_pcount(skb)))
2153 err = __tcp_retransmit_skb(sk, skb);
2155 /* Record snd_nxt for loss detection. */
2157 tp->tlp_high_seq = tp->snd_nxt;
2160 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
2161 inet_csk(sk)->icsk_rto,
2165 NET_INC_STATS_BH(sock_net(sk),
2166 LINUX_MIB_TCPLOSSPROBES);
2169 /* Push out any pending frames which were held back due to
2170 * TCP_CORK or attempt at coalescing tiny packets.
2171 * The socket must be locked by the caller.
2173 void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
2176 /* If we are closed, the bytes will have to remain here.
2177 * In time closedown will finish, we empty the write queue and
2178 * all will be happy.
2180 if (unlikely(sk->sk_state == TCP_CLOSE))
2183 if (tcp_write_xmit(sk, cur_mss, nonagle, 0,
2184 sk_gfp_atomic(sk, GFP_ATOMIC)))
2185 tcp_check_probe_timer(sk);
2188 /* Send _single_ skb sitting at the send head. This function requires
2189 * true push pending frames to setup probe timer etc.
2191 void tcp_push_one(struct sock *sk, unsigned int mss_now)
2193 struct sk_buff *skb = tcp_send_head(sk);
2195 BUG_ON(!skb || skb->len < mss_now);
2197 tcp_write_xmit(sk, mss_now, TCP_NAGLE_PUSH, 1, sk->sk_allocation);
2200 /* This function returns the amount that we can raise the
2201 * usable window based on the following constraints
2203 * 1. The window can never be shrunk once it is offered (RFC 793)
2204 * 2. We limit memory per socket
2207 * "the suggested [SWS] avoidance algorithm for the receiver is to keep
2208 * RECV.NEXT + RCV.WIN fixed until:
2209 * RCV.BUFF - RCV.USER - RCV.WINDOW >= min(1/2 RCV.BUFF, MSS)"
2211 * i.e. don't raise the right edge of the window until you can raise
2212 * it at least MSS bytes.
2214 * Unfortunately, the recommended algorithm breaks header prediction,
2215 * since header prediction assumes th->window stays fixed.
2217 * Strictly speaking, keeping th->window fixed violates the receiver
2218 * side SWS prevention criteria. The problem is that under this rule
2219 * a stream of single byte packets will cause the right side of the
2220 * window to always advance by a single byte.
2222 * Of course, if the sender implements sender side SWS prevention
2223 * then this will not be a problem.
2225 * BSD seems to make the following compromise:
2227 * If the free space is less than the 1/4 of the maximum
2228 * space available and the free space is less than 1/2 mss,
2229 * then set the window to 0.
2230 * [ Actually, bsd uses MSS and 1/4 of maximal _window_ ]
2231 * Otherwise, just prevent the window from shrinking
2232 * and from being larger than the largest representable value.
2234 * This prevents incremental opening of the window in the regime
2235 * where TCP is limited by the speed of the reader side taking
2236 * data out of the TCP receive queue. It does nothing about
2237 * those cases where the window is constrained on the sender side
2238 * because the pipeline is full.
2240 * BSD also seems to "accidentally" limit itself to windows that are a
2241 * multiple of MSS, at least until the free space gets quite small.
2242 * This would appear to be a side effect of the mbuf implementation.
2243 * Combining these two algorithms results in the observed behavior
2244 * of having a fixed window size at almost all times.
2246 * Below we obtain similar behavior by forcing the offered window to
2247 * a multiple of the mss when it is feasible to do so.
2249 * Note, we don't "adjust" for TIMESTAMP or SACK option bytes.
2250 * Regular options like TIMESTAMP are taken into account.
2252 u32 __tcp_select_window(struct sock *sk)
2254 struct inet_connection_sock *icsk = inet_csk(sk);
2255 struct tcp_sock *tp = tcp_sk(sk);
2256 /* MSS for the peer's data. Previous versions used mss_clamp
2257 * here. I don't know if the value based on our guesses
2258 * of peer's MSS is better for the performance. It's more correct
2259 * but may be worse for the performance because of rcv_mss
2260 * fluctuations. --SAW 1998/11/1
2262 int mss = icsk->icsk_ack.rcv_mss;
2263 int free_space = tcp_space(sk);
2264 int allowed_space = tcp_full_space(sk);
2265 int full_space = min_t(int, tp->window_clamp, allowed_space);
2268 if (mss > full_space)
2271 if (free_space < (full_space >> 1)) {
2272 icsk->icsk_ack.quick = 0;
2274 if (sk_under_memory_pressure(sk))
2275 tp->rcv_ssthresh = min(tp->rcv_ssthresh,
2278 /* free_space might become our new window, make sure we don't
2279 * increase it due to wscale.
2281 free_space = round_down(free_space, 1 << tp->rx_opt.rcv_wscale);
2283 /* if free space is less than mss estimate, or is below 1/16th
2284 * of the maximum allowed, try to move to zero-window, else
2285 * tcp_clamp_window() will grow rcv buf up to tcp_rmem[2], and
2286 * new incoming data is dropped due to memory limits.
2287 * With large window, mss test triggers way too late in order
2288 * to announce zero window in time before rmem limit kicks in.
2290 if (free_space < (allowed_space >> 4) || free_space < mss)
2294 if (free_space > tp->rcv_ssthresh)
2295 free_space = tp->rcv_ssthresh;
2297 /* Don't do rounding if we are using window scaling, since the
2298 * scaled window will not line up with the MSS boundary anyway.
2300 window = tp->rcv_wnd;
2301 if (tp->rx_opt.rcv_wscale) {
2302 window = free_space;
2304 /* Advertise enough space so that it won't get scaled away.
2305 * Import case: prevent zero window announcement if
2306 * 1<<rcv_wscale > mss.
2308 if (((window >> tp->rx_opt.rcv_wscale) << tp->rx_opt.rcv_wscale) != window)
2309 window = (((window >> tp->rx_opt.rcv_wscale) + 1)
2310 << tp->rx_opt.rcv_wscale);
2312 /* Get the largest window that is a nice multiple of mss.
2313 * Window clamp already applied above.
2314 * If our current window offering is within 1 mss of the
2315 * free space we just keep it. This prevents the divide
2316 * and multiply from happening most of the time.
2317 * We also don't do any window rounding when the free space
2320 if (window <= free_space - mss || window > free_space)
2321 window = (free_space / mss) * mss;
2322 else if (mss == full_space &&
2323 free_space > window + (full_space >> 1))
2324 window = free_space;
2330 /* Collapses two adjacent SKB's during retransmission. */
2331 static void tcp_collapse_retrans(struct sock *sk, struct sk_buff *skb)
2333 struct tcp_sock *tp = tcp_sk(sk);
2334 struct sk_buff *next_skb = tcp_write_queue_next(sk, skb);
2335 int skb_size, next_skb_size;
2337 skb_size = skb->len;
2338 next_skb_size = next_skb->len;
2340 BUG_ON(tcp_skb_pcount(skb) != 1 || tcp_skb_pcount(next_skb) != 1);
2342 tcp_highest_sack_combine(sk, next_skb, skb);
2344 tcp_unlink_write_queue(next_skb, sk);
2346 skb_copy_from_linear_data(next_skb, skb_put(skb, next_skb_size),
2349 if (next_skb->ip_summed == CHECKSUM_PARTIAL)
2350 skb->ip_summed = CHECKSUM_PARTIAL;
2352 if (skb->ip_summed != CHECKSUM_PARTIAL)
2353 skb->csum = csum_block_add(skb->csum, next_skb->csum, skb_size);
2355 /* Update sequence range on original skb. */
2356 TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(next_skb)->end_seq;
2358 /* Merge over control information. This moves PSH/FIN etc. over */
2359 TCP_SKB_CB(skb)->tcp_flags |= TCP_SKB_CB(next_skb)->tcp_flags;
2361 /* All done, get rid of second SKB and account for it so
2362 * packet counting does not break.
2364 TCP_SKB_CB(skb)->sacked |= TCP_SKB_CB(next_skb)->sacked & TCPCB_EVER_RETRANS;
2366 /* changed transmit queue under us so clear hints */
2367 tcp_clear_retrans_hints_partial(tp);
2368 if (next_skb == tp->retransmit_skb_hint)
2369 tp->retransmit_skb_hint = skb;
2371 tcp_adjust_pcount(sk, next_skb, tcp_skb_pcount(next_skb));
2373 sk_wmem_free_skb(sk, next_skb);
2376 /* Check if coalescing SKBs is legal. */
2377 static bool tcp_can_collapse(const struct sock *sk, const struct sk_buff *skb)
2379 if (tcp_skb_pcount(skb) > 1)
2381 /* TODO: SACK collapsing could be used to remove this condition */
2382 if (skb_shinfo(skb)->nr_frags != 0)
2384 if (skb_cloned(skb))
2386 if (skb == tcp_send_head(sk))
2388 /* Some heurestics for collapsing over SACK'd could be invented */
2389 if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)
2395 /* Collapse packets in the retransmit queue to make to create
2396 * less packets on the wire. This is only done on retransmission.
2398 static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *to,
2401 struct tcp_sock *tp = tcp_sk(sk);
2402 struct sk_buff *skb = to, *tmp;
2405 if (!sysctl_tcp_retrans_collapse)
2407 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_SYN)
2410 tcp_for_write_queue_from_safe(skb, tmp, sk) {
2411 if (!tcp_can_collapse(sk, skb))
2423 /* Punt if not enough space exists in the first SKB for
2424 * the data in the second
2426 if (skb->len > skb_availroom(to))
2429 if (after(TCP_SKB_CB(skb)->end_seq, tcp_wnd_end(tp)))
2432 tcp_collapse_retrans(sk, to);
2436 /* This retransmits one SKB. Policy decisions and retransmit queue
2437 * state updates are done by the caller. Returns non-zero if an
2438 * error occurred which prevented the send.
2440 int __tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
2442 struct tcp_sock *tp = tcp_sk(sk);
2443 struct inet_connection_sock *icsk = inet_csk(sk);
2444 unsigned int cur_mss;
2447 /* Inconslusive MTU probe */
2448 if (icsk->icsk_mtup.probe_size) {
2449 icsk->icsk_mtup.probe_size = 0;
2452 /* Do not sent more than we queued. 1/4 is reserved for possible
2453 * copying overhead: fragmentation, tunneling, mangling etc.
2455 if (atomic_read(&sk->sk_wmem_alloc) >
2456 min(sk->sk_wmem_queued + (sk->sk_wmem_queued >> 2), sk->sk_sndbuf))
2459 if (skb_still_in_host_queue(sk, skb))
2462 if (before(TCP_SKB_CB(skb)->seq, tp->snd_una)) {
2463 if (before(TCP_SKB_CB(skb)->end_seq, tp->snd_una))
2465 if (tcp_trim_head(sk, skb, tp->snd_una - TCP_SKB_CB(skb)->seq))
2469 if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk))
2470 return -EHOSTUNREACH; /* Routing failure or similar. */
2472 cur_mss = tcp_current_mss(sk);
2474 /* If receiver has shrunk his window, and skb is out of
2475 * new window, do not retransmit it. The exception is the
2476 * case, when window is shrunk to zero. In this case
2477 * our retransmit serves as a zero window probe.
2479 if (!before(TCP_SKB_CB(skb)->seq, tcp_wnd_end(tp)) &&
2480 TCP_SKB_CB(skb)->seq != tp->snd_una)
2483 if (skb->len > cur_mss) {
2484 if (tcp_fragment(sk, skb, cur_mss, cur_mss, GFP_ATOMIC))
2485 return -ENOMEM; /* We'll try again later. */
2487 int oldpcount = tcp_skb_pcount(skb);
2489 if (unlikely(oldpcount > 1)) {
2490 if (skb_unclone(skb, GFP_ATOMIC))
2492 tcp_init_tso_segs(sk, skb, cur_mss);
2493 tcp_adjust_pcount(sk, skb, oldpcount - tcp_skb_pcount(skb));
2497 tcp_retrans_try_collapse(sk, skb, cur_mss);
2499 /* Make a copy, if the first transmission SKB clone we made
2500 * is still in somebody's hands, else make a clone.
2502 TCP_SKB_CB(skb)->when = tcp_time_stamp;
2504 /* make sure skb->data is aligned on arches that require it
2505 * and check if ack-trimming & collapsing extended the headroom
2506 * beyond what csum_start can cover.
2508 if (unlikely((NET_IP_ALIGN && ((unsigned long)skb->data & 3)) ||
2509 skb_headroom(skb) >= 0xFFFF)) {
2510 struct sk_buff *nskb = __pskb_copy(skb, MAX_TCP_HEADER,
2512 err = nskb ? tcp_transmit_skb(sk, nskb, 0, GFP_ATOMIC) :
2515 err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
2519 TCP_SKB_CB(skb)->sacked |= TCPCB_EVER_RETRANS;
2520 /* Update global TCP statistics. */
2521 TCP_INC_STATS(sock_net(sk), TCP_MIB_RETRANSSEGS);
2522 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_SYN)
2523 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPSYNRETRANS);
2524 tp->total_retrans++;
2529 int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
2531 struct tcp_sock *tp = tcp_sk(sk);
2532 int err = __tcp_retransmit_skb(sk, skb);
2535 #if FASTRETRANS_DEBUG > 0
2536 if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) {
2537 net_dbg_ratelimited("retrans_out leaked\n");
2540 if (!tp->retrans_out)
2541 tp->lost_retrans_low = tp->snd_nxt;
2542 TCP_SKB_CB(skb)->sacked |= TCPCB_RETRANS;
2543 tp->retrans_out += tcp_skb_pcount(skb);
2545 /* Save stamp of the first retransmit. */
2546 if (!tp->retrans_stamp)
2547 tp->retrans_stamp = TCP_SKB_CB(skb)->when;
2549 /* snd_nxt is stored to detect loss of retransmitted segment,
2550 * see tcp_input.c tcp_sacktag_write_queue().
2552 TCP_SKB_CB(skb)->ack_seq = tp->snd_nxt;
2553 } else if (err != -EBUSY) {
2554 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPRETRANSFAIL);
2557 if (tp->undo_retrans < 0)
2558 tp->undo_retrans = 0;
2559 tp->undo_retrans += tcp_skb_pcount(skb);
2563 /* Check if we forward retransmits are possible in the current
2564 * window/congestion state.
2566 static bool tcp_can_forward_retransmit(struct sock *sk)
2568 const struct inet_connection_sock *icsk = inet_csk(sk);
2569 const struct tcp_sock *tp = tcp_sk(sk);
2571 /* Forward retransmissions are possible only during Recovery. */
2572 if (icsk->icsk_ca_state != TCP_CA_Recovery)
2575 /* No forward retransmissions in Reno are possible. */
2576 if (tcp_is_reno(tp))
2579 /* Yeah, we have to make difficult choice between forward transmission
2580 * and retransmission... Both ways have their merits...
2582 * For now we do not retransmit anything, while we have some new
2583 * segments to send. In the other cases, follow rule 3 for
2584 * NextSeg() specified in RFC3517.
2587 if (tcp_may_send_now(sk))
2593 /* This gets called after a retransmit timeout, and the initially
2594 * retransmitted data is acknowledged. It tries to continue
2595 * resending the rest of the retransmit queue, until either
2596 * we've sent it all or the congestion window limit is reached.
2597 * If doing SACK, the first ACK which comes back for a timeout
2598 * based retransmit packet might feed us FACK information again.
2599 * If so, we use it to avoid unnecessarily retransmissions.
2601 void tcp_xmit_retransmit_queue(struct sock *sk)
2603 const struct inet_connection_sock *icsk = inet_csk(sk);
2604 struct tcp_sock *tp = tcp_sk(sk);
2605 struct sk_buff *skb;
2606 struct sk_buff *hole = NULL;
2609 int fwd_rexmitting = 0;
2611 if (!tp->packets_out)
2615 tp->retransmit_high = tp->snd_una;
2617 if (tp->retransmit_skb_hint) {
2618 skb = tp->retransmit_skb_hint;
2619 last_lost = TCP_SKB_CB(skb)->end_seq;
2620 if (after(last_lost, tp->retransmit_high))
2621 last_lost = tp->retransmit_high;
2623 skb = tcp_write_queue_head(sk);
2624 last_lost = tp->snd_una;
2627 tcp_for_write_queue_from(skb, sk) {
2628 __u8 sacked = TCP_SKB_CB(skb)->sacked;
2630 if (skb == tcp_send_head(sk))
2632 /* we could do better than to assign each time */
2634 tp->retransmit_skb_hint = skb;
2636 /* Assume this retransmit will generate
2637 * only one packet for congestion window
2638 * calculation purposes. This works because
2639 * tcp_retransmit_skb() will chop up the
2640 * packet to be MSS sized and all the
2641 * packet counting works out.
2643 if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
2646 if (fwd_rexmitting) {
2648 if (!before(TCP_SKB_CB(skb)->seq, tcp_highest_sack_seq(tp)))
2650 mib_idx = LINUX_MIB_TCPFORWARDRETRANS;
2652 } else if (!before(TCP_SKB_CB(skb)->seq, tp->retransmit_high)) {
2653 tp->retransmit_high = last_lost;
2654 if (!tcp_can_forward_retransmit(sk))
2656 /* Backtrack if necessary to non-L'ed skb */
2664 } else if (!(sacked & TCPCB_LOST)) {
2665 if (hole == NULL && !(sacked & (TCPCB_SACKED_RETRANS|TCPCB_SACKED_ACKED)))
2670 last_lost = TCP_SKB_CB(skb)->end_seq;
2671 if (icsk->icsk_ca_state != TCP_CA_Loss)
2672 mib_idx = LINUX_MIB_TCPFASTRETRANS;
2674 mib_idx = LINUX_MIB_TCPSLOWSTARTRETRANS;
2677 if (sacked & (TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))
2680 if (tcp_retransmit_skb(sk, skb))
2683 NET_INC_STATS_BH(sock_net(sk), mib_idx);
2685 if (tcp_in_cwnd_reduction(sk))
2686 tp->prr_out += tcp_skb_pcount(skb);
2688 if (skb == tcp_write_queue_head(sk))
2689 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
2690 inet_csk(sk)->icsk_rto,
2695 /* Send a fin. The caller locks the socket for us. This cannot be
2696 * allowed to fail queueing a FIN frame under any circumstances.
2698 void tcp_send_fin(struct sock *sk)
2700 struct tcp_sock *tp = tcp_sk(sk);
2701 struct sk_buff *skb = tcp_write_queue_tail(sk);
2704 /* Optimization, tack on the FIN if we have a queue of
2705 * unsent frames. But be careful about outgoing SACKS
2708 mss_now = tcp_current_mss(sk);
2710 if (tcp_send_head(sk) != NULL) {
2711 TCP_SKB_CB(skb)->tcp_flags |= TCPHDR_FIN;
2712 TCP_SKB_CB(skb)->end_seq++;
2715 /* Socket is locked, keep trying until memory is available. */
2717 skb = alloc_skb_fclone(MAX_TCP_HEADER,
2724 /* Reserve space for headers and prepare control bits. */
2725 skb_reserve(skb, MAX_TCP_HEADER);
2726 /* FIN eats a sequence byte, write_seq advanced by tcp_queue_skb(). */
2727 tcp_init_nondata_skb(skb, tp->write_seq,
2728 TCPHDR_ACK | TCPHDR_FIN);
2729 tcp_queue_skb(sk, skb);
2731 __tcp_push_pending_frames(sk, mss_now, TCP_NAGLE_OFF);
2734 /* We get here when a process closes a file descriptor (either due to
2735 * an explicit close() or as a byproduct of exit()'ing) and there
2736 * was unread data in the receive queue. This behavior is recommended
2737 * by RFC 2525, section 2.17. -DaveM
2739 void tcp_send_active_reset(struct sock *sk, gfp_t priority)
2741 struct sk_buff *skb;
2743 /* NOTE: No TCP options attached and we never retransmit this. */
2744 skb = alloc_skb(MAX_TCP_HEADER, priority);
2746 NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTFAILED);
2750 /* Reserve space for headers and prepare control bits. */
2751 skb_reserve(skb, MAX_TCP_HEADER);
2752 tcp_init_nondata_skb(skb, tcp_acceptable_seq(sk),
2753 TCPHDR_ACK | TCPHDR_RST);
2755 TCP_SKB_CB(skb)->when = tcp_time_stamp;
2756 if (tcp_transmit_skb(sk, skb, 0, priority))
2757 NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTFAILED);
2759 TCP_INC_STATS(sock_net(sk), TCP_MIB_OUTRSTS);
2762 /* Send a crossed SYN-ACK during socket establishment.
2763 * WARNING: This routine must only be called when we have already sent
2764 * a SYN packet that crossed the incoming SYN that caused this routine
2765 * to get called. If this assumption fails then the initial rcv_wnd
2766 * and rcv_wscale values will not be correct.
2768 int tcp_send_synack(struct sock *sk)
2770 struct sk_buff *skb;
2772 skb = tcp_write_queue_head(sk);
2773 if (skb == NULL || !(TCP_SKB_CB(skb)->tcp_flags & TCPHDR_SYN)) {
2774 pr_debug("%s: wrong queue state\n", __func__);
2777 if (!(TCP_SKB_CB(skb)->tcp_flags & TCPHDR_ACK)) {
2778 if (skb_cloned(skb)) {
2779 struct sk_buff *nskb = skb_copy(skb, GFP_ATOMIC);
2782 tcp_unlink_write_queue(skb, sk);
2783 skb_header_release(nskb);
2784 __tcp_add_write_queue_head(sk, nskb);
2785 sk_wmem_free_skb(sk, skb);
2786 sk->sk_wmem_queued += nskb->truesize;
2787 sk_mem_charge(sk, nskb->truesize);
2791 TCP_SKB_CB(skb)->tcp_flags |= TCPHDR_ACK;
2792 TCP_ECN_send_synack(tcp_sk(sk), skb);
2794 TCP_SKB_CB(skb)->when = tcp_time_stamp;
2795 return tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
2799 * tcp_make_synack - Prepare a SYN-ACK.
2800 * sk: listener socket
2801 * dst: dst entry attached to the SYNACK
2802 * req: request_sock pointer
2804 * Allocate one skb and build a SYNACK packet.
2805 * @dst is consumed : Caller should not use it again.
2807 struct sk_buff *tcp_make_synack(struct sock *sk, struct dst_entry *dst,
2808 struct request_sock *req,
2809 struct tcp_fastopen_cookie *foc)
2811 struct tcp_out_options opts;
2812 struct inet_request_sock *ireq = inet_rsk(req);
2813 struct tcp_sock *tp = tcp_sk(sk);
2815 struct sk_buff *skb;
2816 struct tcp_md5sig_key *md5;
2817 int tcp_header_size;
2820 skb = sock_wmalloc(sk, MAX_TCP_HEADER, 1, GFP_ATOMIC);
2821 if (unlikely(!skb)) {
2825 /* Reserve space for headers. */
2826 skb_reserve(skb, MAX_TCP_HEADER);
2828 skb_dst_set(skb, dst);
2829 security_skb_owned_by(skb, sk);
2831 mss = dst_metric_advmss(dst);
2832 if (tp->rx_opt.user_mss && tp->rx_opt.user_mss < mss)
2833 mss = tp->rx_opt.user_mss;
2835 memset(&opts, 0, sizeof(opts));
2836 #ifdef CONFIG_SYN_COOKIES
2837 if (unlikely(req->cookie_ts))
2838 TCP_SKB_CB(skb)->when = cookie_init_timestamp(req);
2841 TCP_SKB_CB(skb)->when = tcp_time_stamp;
2842 tcp_header_size = tcp_synack_options(sk, req, mss, skb, &opts, &md5,
2845 skb_push(skb, tcp_header_size);
2846 skb_reset_transport_header(skb);
2849 memset(th, 0, sizeof(struct tcphdr));
2852 TCP_ECN_make_synack(req, th);
2853 th->source = htons(ireq->ir_num);
2854 th->dest = ireq->ir_rmt_port;
2855 /* Setting of flags are superfluous here for callers (and ECE is
2856 * not even correctly set)
2858 tcp_init_nondata_skb(skb, tcp_rsk(req)->snt_isn,
2859 TCPHDR_SYN | TCPHDR_ACK);
2861 th->seq = htonl(TCP_SKB_CB(skb)->seq);
2862 /* XXX data is queued and acked as is. No buffer/window check */
2863 th->ack_seq = htonl(tcp_rsk(req)->rcv_nxt);
2865 /* RFC1323: The window in SYN & SYN/ACK segments is never scaled. */
2866 th->window = htons(min(req->rcv_wnd, 65535U));
2867 tcp_options_write((__be32 *)(th + 1), tp, &opts);
2868 th->doff = (tcp_header_size >> 2);
2869 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_OUTSEGS);
2871 #ifdef CONFIG_TCP_MD5SIG
2872 /* Okay, we have all we need - do the md5 hash if needed */
2874 tcp_rsk(req)->af_specific->calc_md5_hash(opts.hash_location,
2875 md5, NULL, req, skb);
2881 EXPORT_SYMBOL(tcp_make_synack);
2883 /* Do all connect socket setups that can be done AF independent. */
2884 static void tcp_connect_init(struct sock *sk)
2886 const struct dst_entry *dst = __sk_dst_get(sk);
2887 struct tcp_sock *tp = tcp_sk(sk);
2890 /* We'll fix this up when we get a response from the other end.
2891 * See tcp_input.c:tcp_rcv_state_process case TCP_SYN_SENT.
2893 tp->tcp_header_len = sizeof(struct tcphdr) +
2894 (sysctl_tcp_timestamps ? TCPOLEN_TSTAMP_ALIGNED : 0);
2896 #ifdef CONFIG_TCP_MD5SIG
2897 if (tp->af_specific->md5_lookup(sk, sk) != NULL)
2898 tp->tcp_header_len += TCPOLEN_MD5SIG_ALIGNED;
2901 /* If user gave his TCP_MAXSEG, record it to clamp */
2902 if (tp->rx_opt.user_mss)
2903 tp->rx_opt.mss_clamp = tp->rx_opt.user_mss;
2906 tcp_sync_mss(sk, dst_mtu(dst));
2908 if (!tp->window_clamp)
2909 tp->window_clamp = dst_metric(dst, RTAX_WINDOW);
2910 tp->advmss = dst_metric_advmss(dst);
2911 if (tp->rx_opt.user_mss && tp->rx_opt.user_mss < tp->advmss)
2912 tp->advmss = tp->rx_opt.user_mss;
2914 tcp_initialize_rcv_mss(sk);
2916 /* limit the window selection if the user enforce a smaller rx buffer */
2917 if (sk->sk_userlocks & SOCK_RCVBUF_LOCK &&
2918 (tp->window_clamp > tcp_full_space(sk) || tp->window_clamp == 0))
2919 tp->window_clamp = tcp_full_space(sk);
2921 tcp_select_initial_window(tcp_full_space(sk),
2922 tp->advmss - (tp->rx_opt.ts_recent_stamp ? tp->tcp_header_len - sizeof(struct tcphdr) : 0),
2925 sysctl_tcp_window_scaling,
2927 dst_metric(dst, RTAX_INITRWND));
2929 tp->rx_opt.rcv_wscale = rcv_wscale;
2930 tp->rcv_ssthresh = tp->rcv_wnd;
2933 sock_reset_flag(sk, SOCK_DONE);
2936 tp->snd_una = tp->write_seq;
2937 tp->snd_sml = tp->write_seq;
2938 tp->snd_up = tp->write_seq;
2939 tp->snd_nxt = tp->write_seq;
2941 if (likely(!tp->repair))
2944 tp->rcv_tstamp = tcp_time_stamp;
2945 tp->rcv_wup = tp->rcv_nxt;
2946 tp->copied_seq = tp->rcv_nxt;
2948 inet_csk(sk)->icsk_rto = TCP_TIMEOUT_INIT;
2949 inet_csk(sk)->icsk_retransmits = 0;
2950 tcp_clear_retrans(tp);
2953 static void tcp_connect_queue_skb(struct sock *sk, struct sk_buff *skb)
2955 struct tcp_sock *tp = tcp_sk(sk);
2956 struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
2958 tcb->end_seq += skb->len;
2959 skb_header_release(skb);
2960 __tcp_add_write_queue_tail(sk, skb);
2961 sk->sk_wmem_queued += skb->truesize;
2962 sk_mem_charge(sk, skb->truesize);
2963 tp->write_seq = tcb->end_seq;
2964 tp->packets_out += tcp_skb_pcount(skb);
2967 /* Build and send a SYN with data and (cached) Fast Open cookie. However,
2968 * queue a data-only packet after the regular SYN, such that regular SYNs
2969 * are retransmitted on timeouts. Also if the remote SYN-ACK acknowledges
2970 * only the SYN sequence, the data are retransmitted in the first ACK.
2971 * If cookie is not cached or other error occurs, falls back to send a
2972 * regular SYN with Fast Open cookie request option.
2974 static int tcp_send_syn_data(struct sock *sk, struct sk_buff *syn)
2976 struct tcp_sock *tp = tcp_sk(sk);
2977 struct tcp_fastopen_request *fo = tp->fastopen_req;
2978 int syn_loss = 0, space, i, err = 0, iovlen = fo->data->msg_iovlen;
2979 struct sk_buff *syn_data = NULL, *data;
2980 unsigned long last_syn_loss = 0;
2982 tp->rx_opt.mss_clamp = tp->advmss; /* If MSS is not cached */
2983 tcp_fastopen_cache_get(sk, &tp->rx_opt.mss_clamp, &fo->cookie,
2984 &syn_loss, &last_syn_loss);
2985 /* Recurring FO SYN losses: revert to regular handshake temporarily */
2987 time_before(jiffies, last_syn_loss + (60*HZ << syn_loss))) {
2988 fo->cookie.len = -1;
2992 if (sysctl_tcp_fastopen & TFO_CLIENT_NO_COOKIE)
2993 fo->cookie.len = -1;
2994 else if (fo->cookie.len <= 0)
2997 /* MSS for SYN-data is based on cached MSS and bounded by PMTU and
2998 * user-MSS. Reserve maximum option space for middleboxes that add
2999 * private TCP options. The cost is reduced data space in SYN :(
3001 if (tp->rx_opt.user_mss && tp->rx_opt.user_mss < tp->rx_opt.mss_clamp)
3002 tp->rx_opt.mss_clamp = tp->rx_opt.user_mss;
3003 space = __tcp_mtu_to_mss(sk, inet_csk(sk)->icsk_pmtu_cookie) -
3004 MAX_TCP_OPTION_SPACE;
3006 space = min_t(size_t, space, fo->size);
3008 /* limit to order-0 allocations */
3009 space = min_t(size_t, space, SKB_MAX_HEAD(MAX_TCP_HEADER));
3011 syn_data = skb_copy_expand(syn, MAX_TCP_HEADER, space,
3013 if (syn_data == NULL)
3016 for (i = 0; i < iovlen && syn_data->len < space; ++i) {
3017 struct iovec *iov = &fo->data->msg_iov[i];
3018 unsigned char __user *from = iov->iov_base;
3019 int len = iov->iov_len;
3021 if (syn_data->len + len > space)
3022 len = space - syn_data->len;
3023 else if (i + 1 == iovlen)
3024 /* No more data pending in inet_wait_for_connect() */
3027 if (skb_add_data(syn_data, from, len))
3031 /* Queue a data-only packet after the regular SYN for retransmission */
3032 data = pskb_copy(syn_data, sk->sk_allocation);
3035 TCP_SKB_CB(data)->seq++;
3036 TCP_SKB_CB(data)->tcp_flags &= ~TCPHDR_SYN;
3037 TCP_SKB_CB(data)->tcp_flags = (TCPHDR_ACK|TCPHDR_PSH);
3038 tcp_connect_queue_skb(sk, data);
3039 fo->copied = data->len;
3041 /* syn_data is about to be sent, we need to take current time stamps
3042 * for the packets that are in write queue : SYN packet and DATA
3044 skb_mstamp_get(&syn->skb_mstamp);
3045 data->skb_mstamp = syn->skb_mstamp;
3047 if (tcp_transmit_skb(sk, syn_data, 0, sk->sk_allocation) == 0) {
3048 tp->syn_data = (fo->copied > 0);
3049 NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPORIGDATASENT);
3055 /* Send a regular SYN with Fast Open cookie request option */
3056 if (fo->cookie.len > 0)
3058 err = tcp_transmit_skb(sk, syn, 1, sk->sk_allocation);
3060 tp->syn_fastopen = 0;
3061 kfree_skb(syn_data);
3063 fo->cookie.len = -1; /* Exclude Fast Open option for SYN retries */
3067 /* Build a SYN and send it off. */
3068 int tcp_connect(struct sock *sk)
3070 struct tcp_sock *tp = tcp_sk(sk);
3071 struct sk_buff *buff;
3074 tcp_connect_init(sk);
3076 if (unlikely(tp->repair)) {
3077 tcp_finish_connect(sk, NULL);
3081 buff = alloc_skb_fclone(MAX_TCP_HEADER + 15, sk->sk_allocation);
3082 if (unlikely(buff == NULL))
3085 /* Reserve space for headers. */
3086 skb_reserve(buff, MAX_TCP_HEADER);
3088 tcp_init_nondata_skb(buff, tp->write_seq++, TCPHDR_SYN);
3089 tp->retrans_stamp = TCP_SKB_CB(buff)->when = tcp_time_stamp;
3090 tcp_connect_queue_skb(sk, buff);
3091 TCP_ECN_send_syn(sk, buff);
3093 /* Send off SYN; include data in Fast Open. */
3094 err = tp->fastopen_req ? tcp_send_syn_data(sk, buff) :
3095 tcp_transmit_skb(sk, buff, 1, sk->sk_allocation);
3096 if (err == -ECONNREFUSED)
3099 /* We change tp->snd_nxt after the tcp_transmit_skb() call
3100 * in order to make this packet get counted in tcpOutSegs.
3102 tp->snd_nxt = tp->write_seq;
3103 tp->pushed_seq = tp->write_seq;
3104 TCP_INC_STATS(sock_net(sk), TCP_MIB_ACTIVEOPENS);
3106 /* Timer for repeating the SYN until an answer. */
3107 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
3108 inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
3111 EXPORT_SYMBOL(tcp_connect);
3113 /* Send out a delayed ack, the caller does the policy checking
3114 * to see if we should even be here. See tcp_input.c:tcp_ack_snd_check()
3117 void tcp_send_delayed_ack(struct sock *sk)
3119 struct inet_connection_sock *icsk = inet_csk(sk);
3120 int ato = icsk->icsk_ack.ato;
3121 unsigned long timeout;
3123 if (ato > TCP_DELACK_MIN) {
3124 const struct tcp_sock *tp = tcp_sk(sk);
3125 int max_ato = HZ / 2;
3127 if (icsk->icsk_ack.pingpong ||
3128 (icsk->icsk_ack.pending & ICSK_ACK_PUSHED))
3129 max_ato = TCP_DELACK_MAX;
3131 /* Slow path, intersegment interval is "high". */
3133 /* If some rtt estimate is known, use it to bound delayed ack.
3134 * Do not use inet_csk(sk)->icsk_rto here, use results of rtt measurements
3138 int rtt = max_t(int, usecs_to_jiffies(tp->srtt_us >> 3),
3145 ato = min(ato, max_ato);
3148 /* Stay within the limit we were given */
3149 timeout = jiffies + ato;
3151 /* Use new timeout only if there wasn't a older one earlier. */
3152 if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) {
3153 /* If delack timer was blocked or is about to expire,
3156 if (icsk->icsk_ack.blocked ||
3157 time_before_eq(icsk->icsk_ack.timeout, jiffies + (ato >> 2))) {
3162 if (!time_before(timeout, icsk->icsk_ack.timeout))
3163 timeout = icsk->icsk_ack.timeout;
3165 icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER;
3166 icsk->icsk_ack.timeout = timeout;
3167 sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout);
3170 /* This routine sends an ack and also updates the window. */
3171 void tcp_send_ack(struct sock *sk)
3173 struct sk_buff *buff;
3175 /* If we have been reset, we may not send again. */
3176 if (sk->sk_state == TCP_CLOSE)
3179 /* We are not putting this on the write queue, so
3180 * tcp_transmit_skb() will set the ownership to this
3183 buff = alloc_skb(MAX_TCP_HEADER, sk_gfp_atomic(sk, GFP_ATOMIC));
3185 inet_csk_schedule_ack(sk);
3186 inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
3187 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
3188 TCP_DELACK_MAX, TCP_RTO_MAX);
3192 /* Reserve space for headers and prepare control bits. */
3193 skb_reserve(buff, MAX_TCP_HEADER);
3194 tcp_init_nondata_skb(buff, tcp_acceptable_seq(sk), TCPHDR_ACK);
3196 /* Send it off, this clears delayed acks for us. */
3197 TCP_SKB_CB(buff)->when = tcp_time_stamp;
3198 tcp_transmit_skb(sk, buff, 0, sk_gfp_atomic(sk, GFP_ATOMIC));
3201 /* This routine sends a packet with an out of date sequence
3202 * number. It assumes the other end will try to ack it.
3204 * Question: what should we make while urgent mode?
3205 * 4.4BSD forces sending single byte of data. We cannot send
3206 * out of window data, because we have SND.NXT==SND.MAX...
3208 * Current solution: to send TWO zero-length segments in urgent mode:
3209 * one is with SEG.SEQ=SND.UNA to deliver urgent pointer, another is
3210 * out-of-date with SND.UNA-1 to probe window.
3212 static int tcp_xmit_probe_skb(struct sock *sk, int urgent)
3214 struct tcp_sock *tp = tcp_sk(sk);
3215 struct sk_buff *skb;
3217 /* We don't queue it, tcp_transmit_skb() sets ownership. */
3218 skb = alloc_skb(MAX_TCP_HEADER, sk_gfp_atomic(sk, GFP_ATOMIC));
3222 /* Reserve space for headers and set control bits. */
3223 skb_reserve(skb, MAX_TCP_HEADER);
3224 /* Use a previous sequence. This should cause the other
3225 * end to send an ack. Don't queue or clone SKB, just
3228 tcp_init_nondata_skb(skb, tp->snd_una - !urgent, TCPHDR_ACK);
3229 TCP_SKB_CB(skb)->when = tcp_time_stamp;
3230 return tcp_transmit_skb(sk, skb, 0, GFP_ATOMIC);
3233 void tcp_send_window_probe(struct sock *sk)
3235 if (sk->sk_state == TCP_ESTABLISHED) {
3236 tcp_sk(sk)->snd_wl1 = tcp_sk(sk)->rcv_nxt - 1;
3237 tcp_xmit_probe_skb(sk, 0);
3241 /* Initiate keepalive or window probe from timer. */
3242 int tcp_write_wakeup(struct sock *sk)
3244 struct tcp_sock *tp = tcp_sk(sk);
3245 struct sk_buff *skb;
3247 if (sk->sk_state == TCP_CLOSE)
3250 if ((skb = tcp_send_head(sk)) != NULL &&
3251 before(TCP_SKB_CB(skb)->seq, tcp_wnd_end(tp))) {
3253 unsigned int mss = tcp_current_mss(sk);
3254 unsigned int seg_size = tcp_wnd_end(tp) - TCP_SKB_CB(skb)->seq;
3256 if (before(tp->pushed_seq, TCP_SKB_CB(skb)->end_seq))
3257 tp->pushed_seq = TCP_SKB_CB(skb)->end_seq;
3259 /* We are probing the opening of a window
3260 * but the window size is != 0
3261 * must have been a result SWS avoidance ( sender )
3263 if (seg_size < TCP_SKB_CB(skb)->end_seq - TCP_SKB_CB(skb)->seq ||
3265 seg_size = min(seg_size, mss);
3266 TCP_SKB_CB(skb)->tcp_flags |= TCPHDR_PSH;
3267 if (tcp_fragment(sk, skb, seg_size, mss, GFP_ATOMIC))
3269 } else if (!tcp_skb_pcount(skb))
3270 tcp_set_skb_tso_segs(sk, skb, mss);
3272 TCP_SKB_CB(skb)->tcp_flags |= TCPHDR_PSH;
3273 TCP_SKB_CB(skb)->when = tcp_time_stamp;
3274 err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
3276 tcp_event_new_data_sent(sk, skb);
3279 if (between(tp->snd_up, tp->snd_una + 1, tp->snd_una + 0xFFFF))
3280 tcp_xmit_probe_skb(sk, 1);
3281 return tcp_xmit_probe_skb(sk, 0);
3285 /* A window probe timeout has occurred. If window is not closed send
3286 * a partial packet else a zero probe.
3288 void tcp_send_probe0(struct sock *sk)
3290 struct inet_connection_sock *icsk = inet_csk(sk);
3291 struct tcp_sock *tp = tcp_sk(sk);
3294 err = tcp_write_wakeup(sk);
3296 if (tp->packets_out || !tcp_send_head(sk)) {
3297 /* Cancel probe timer, if it is not required. */
3298 icsk->icsk_probes_out = 0;
3299 icsk->icsk_backoff = 0;
3304 if (icsk->icsk_backoff < sysctl_tcp_retries2)
3305 icsk->icsk_backoff++;
3306 icsk->icsk_probes_out++;
3307 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
3308 min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX),
3311 /* If packet was not sent due to local congestion,
3312 * do not backoff and do not remember icsk_probes_out.
3313 * Let local senders to fight for local resources.
3315 * Use accumulated backoff yet.
3317 if (!icsk->icsk_probes_out)
3318 icsk->icsk_probes_out = 1;
3319 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
3320 min(icsk->icsk_rto << icsk->icsk_backoff,
3321 TCP_RESOURCE_PROBE_INTERVAL),
3326 int tcp_rtx_synack(struct sock *sk, struct request_sock *req)
3328 const struct tcp_request_sock_ops *af_ops = tcp_rsk(req)->af_specific;
3332 res = af_ops->send_synack(sk, NULL, &fl, req, 0, NULL);
3334 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS);
3335 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPSYNRETRANS);
3339 EXPORT_SYMBOL(tcp_rtx_synack);