diff options
author | Arnaldo Carvalho de Melo <acme@mandriva.com> | 2005-08-10 03:03:31 -0400 |
---|---|---|
committer | David S. Miller <davem@sunset.davemloft.net> | 2005-08-29 18:56:18 -0400 |
commit | 6687e988d9aeaccad6774e6a8304f681f3ec0a03 (patch) | |
tree | ecd3d28f9989847aa1dcde4782de0210aeadc290 /net/ipv4/tcp_output.c | |
parent | 64ce207306debd7157f47282be94770407bec01c (diff) |
[ICSK]: Move TCP congestion avoidance members to icsk
This changeset basically moves tcp_sk()->{ca_ops,ca_state,etc} to inet_csk(),
minimal renaming/moving done in this changeset to ease review.
Most of it is just changes of struct tcp_sock * to struct sock * parameters.
With this we move to a state closer to two interesting goals:
1. Generalisation of net/ipv4/tcp_diag.c, becoming inet_diag.c, being used
for any INET transport protocol that has struct inet_hashinfo and are
derived from struct inet_connection_sock. Keeps the userspace API, that will
just not display DCCP sockets, while newer versions of tools can support
DCCP.
2. INET generic transport pluggable Congestion Avoidance infrastructure, using
the current TCP CA infrastructure with DCCP.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/ipv4/tcp_output.c')
-rw-r--r-- | net/ipv4/tcp_output.c | 36 |
1 files changed, 20 insertions, 16 deletions
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index f458eacb5ef2..267b0fcbfc9c 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c | |||
@@ -112,9 +112,9 @@ static void tcp_cwnd_restart(struct sock *sk, struct dst_entry *dst) | |||
112 | u32 restart_cwnd = tcp_init_cwnd(tp, dst); | 112 | u32 restart_cwnd = tcp_init_cwnd(tp, dst); |
113 | u32 cwnd = tp->snd_cwnd; | 113 | u32 cwnd = tp->snd_cwnd; |
114 | 114 | ||
115 | tcp_ca_event(tp, CA_EVENT_CWND_RESTART); | 115 | tcp_ca_event(sk, CA_EVENT_CWND_RESTART); |
116 | 116 | ||
117 | tp->snd_ssthresh = tcp_current_ssthresh(tp); | 117 | tp->snd_ssthresh = tcp_current_ssthresh(sk); |
118 | restart_cwnd = min(restart_cwnd, cwnd); | 118 | restart_cwnd = min(restart_cwnd, cwnd); |
119 | 119 | ||
120 | while ((delta -= inet_csk(sk)->icsk_rto) > 0 && cwnd > restart_cwnd) | 120 | while ((delta -= inet_csk(sk)->icsk_rto) > 0 && cwnd > restart_cwnd) |
@@ -265,6 +265,7 @@ static __inline__ u16 tcp_select_window(struct sock *sk) | |||
265 | static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb) | 265 | static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb) |
266 | { | 266 | { |
267 | if (skb != NULL) { | 267 | if (skb != NULL) { |
268 | const struct inet_connection_sock *icsk = inet_csk(sk); | ||
268 | struct inet_sock *inet = inet_sk(sk); | 269 | struct inet_sock *inet = inet_sk(sk); |
269 | struct tcp_sock *tp = tcp_sk(sk); | 270 | struct tcp_sock *tp = tcp_sk(sk); |
270 | struct tcp_skb_cb *tcb = TCP_SKB_CB(skb); | 271 | struct tcp_skb_cb *tcb = TCP_SKB_CB(skb); |
@@ -280,7 +281,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb) | |||
280 | #define SYSCTL_FLAG_SACK 0x4 | 281 | #define SYSCTL_FLAG_SACK 0x4 |
281 | 282 | ||
282 | /* If congestion control is doing timestamping */ | 283 | /* If congestion control is doing timestamping */ |
283 | if (tp->ca_ops->rtt_sample) | 284 | if (icsk->icsk_ca_ops->rtt_sample) |
284 | do_gettimeofday(&skb->stamp); | 285 | do_gettimeofday(&skb->stamp); |
285 | 286 | ||
286 | sysctl_flags = 0; | 287 | sysctl_flags = 0; |
@@ -308,7 +309,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb) | |||
308 | } | 309 | } |
309 | 310 | ||
310 | if (tcp_packets_in_flight(tp) == 0) | 311 | if (tcp_packets_in_flight(tp) == 0) |
311 | tcp_ca_event(tp, CA_EVENT_TX_START); | 312 | tcp_ca_event(sk, CA_EVENT_TX_START); |
312 | 313 | ||
313 | th = (struct tcphdr *) skb_push(skb, tcp_header_size); | 314 | th = (struct tcphdr *) skb_push(skb, tcp_header_size); |
314 | skb->h.th = th; | 315 | skb->h.th = th; |
@@ -366,7 +367,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb) | |||
366 | if (err <= 0) | 367 | if (err <= 0) |
367 | return err; | 368 | return err; |
368 | 369 | ||
369 | tcp_enter_cwr(tp); | 370 | tcp_enter_cwr(sk); |
370 | 371 | ||
371 | /* NET_XMIT_CN is special. It does not guarantee, | 372 | /* NET_XMIT_CN is special. It does not guarantee, |
372 | * that this packet is lost. It tells that device | 373 | * that this packet is lost. It tells that device |
@@ -905,12 +906,13 @@ static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len, | |||
905 | */ | 906 | */ |
906 | static int tcp_tso_should_defer(struct sock *sk, struct tcp_sock *tp, struct sk_buff *skb) | 907 | static int tcp_tso_should_defer(struct sock *sk, struct tcp_sock *tp, struct sk_buff *skb) |
907 | { | 908 | { |
909 | const struct inet_connection_sock *icsk = inet_csk(sk); | ||
908 | u32 send_win, cong_win, limit, in_flight; | 910 | u32 send_win, cong_win, limit, in_flight; |
909 | 911 | ||
910 | if (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) | 912 | if (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) |
911 | return 0; | 913 | return 0; |
912 | 914 | ||
913 | if (tp->ca_state != TCP_CA_Open) | 915 | if (icsk->icsk_ca_state != TCP_CA_Open) |
914 | return 0; | 916 | return 0; |
915 | 917 | ||
916 | in_flight = tcp_packets_in_flight(tp); | 918 | in_flight = tcp_packets_in_flight(tp); |
@@ -1287,6 +1289,7 @@ static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *skb, int m | |||
1287 | */ | 1289 | */ |
1288 | void tcp_simple_retransmit(struct sock *sk) | 1290 | void tcp_simple_retransmit(struct sock *sk) |
1289 | { | 1291 | { |
1292 | const struct inet_connection_sock *icsk = inet_csk(sk); | ||
1290 | struct tcp_sock *tp = tcp_sk(sk); | 1293 | struct tcp_sock *tp = tcp_sk(sk); |
1291 | struct sk_buff *skb; | 1294 | struct sk_buff *skb; |
1292 | unsigned int mss = tcp_current_mss(sk, 0); | 1295 | unsigned int mss = tcp_current_mss(sk, 0); |
@@ -1317,12 +1320,12 @@ void tcp_simple_retransmit(struct sock *sk) | |||
1317 | * in network, but units changed and effective | 1320 | * in network, but units changed and effective |
1318 | * cwnd/ssthresh really reduced now. | 1321 | * cwnd/ssthresh really reduced now. |
1319 | */ | 1322 | */ |
1320 | if (tp->ca_state != TCP_CA_Loss) { | 1323 | if (icsk->icsk_ca_state != TCP_CA_Loss) { |
1321 | tp->high_seq = tp->snd_nxt; | 1324 | tp->high_seq = tp->snd_nxt; |
1322 | tp->snd_ssthresh = tcp_current_ssthresh(tp); | 1325 | tp->snd_ssthresh = tcp_current_ssthresh(sk); |
1323 | tp->prior_ssthresh = 0; | 1326 | tp->prior_ssthresh = 0; |
1324 | tp->undo_marker = 0; | 1327 | tp->undo_marker = 0; |
1325 | tcp_set_ca_state(tp, TCP_CA_Loss); | 1328 | tcp_set_ca_state(sk, TCP_CA_Loss); |
1326 | } | 1329 | } |
1327 | tcp_xmit_retransmit_queue(sk); | 1330 | tcp_xmit_retransmit_queue(sk); |
1328 | } | 1331 | } |
@@ -1462,6 +1465,7 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb) | |||
1462 | */ | 1465 | */ |
1463 | void tcp_xmit_retransmit_queue(struct sock *sk) | 1466 | void tcp_xmit_retransmit_queue(struct sock *sk) |
1464 | { | 1467 | { |
1468 | const struct inet_connection_sock *icsk = inet_csk(sk); | ||
1465 | struct tcp_sock *tp = tcp_sk(sk); | 1469 | struct tcp_sock *tp = tcp_sk(sk); |
1466 | struct sk_buff *skb; | 1470 | struct sk_buff *skb; |
1467 | int packet_cnt = tp->lost_out; | 1471 | int packet_cnt = tp->lost_out; |
@@ -1485,7 +1489,7 @@ void tcp_xmit_retransmit_queue(struct sock *sk) | |||
1485 | if (!(sacked&(TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))) { | 1489 | if (!(sacked&(TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))) { |
1486 | if (tcp_retransmit_skb(sk, skb)) | 1490 | if (tcp_retransmit_skb(sk, skb)) |
1487 | return; | 1491 | return; |
1488 | if (tp->ca_state != TCP_CA_Loss) | 1492 | if (icsk->icsk_ca_state != TCP_CA_Loss) |
1489 | NET_INC_STATS_BH(LINUX_MIB_TCPFASTRETRANS); | 1493 | NET_INC_STATS_BH(LINUX_MIB_TCPFASTRETRANS); |
1490 | else | 1494 | else |
1491 | NET_INC_STATS_BH(LINUX_MIB_TCPSLOWSTARTRETRANS); | 1495 | NET_INC_STATS_BH(LINUX_MIB_TCPSLOWSTARTRETRANS); |
@@ -1507,7 +1511,7 @@ void tcp_xmit_retransmit_queue(struct sock *sk) | |||
1507 | /* OK, demanded retransmission is finished. */ | 1511 | /* OK, demanded retransmission is finished. */ |
1508 | 1512 | ||
1509 | /* Forward retransmissions are possible only during Recovery. */ | 1513 | /* Forward retransmissions are possible only during Recovery. */ |
1510 | if (tp->ca_state != TCP_CA_Recovery) | 1514 | if (icsk->icsk_ca_state != TCP_CA_Recovery) |
1511 | return; | 1515 | return; |
1512 | 1516 | ||
1513 | /* No forward retransmissions in Reno are possible. */ | 1517 | /* No forward retransmissions in Reno are possible. */ |
@@ -2028,7 +2032,7 @@ void tcp_send_probe0(struct sock *sk) | |||
2028 | 2032 | ||
2029 | if (tp->packets_out || !sk->sk_send_head) { | 2033 | if (tp->packets_out || !sk->sk_send_head) { |
2030 | /* Cancel probe timer, if it is not required. */ | 2034 | /* Cancel probe timer, if it is not required. */ |
2031 | tp->probes_out = 0; | 2035 | icsk->icsk_probes_out = 0; |
2032 | icsk->icsk_backoff = 0; | 2036 | icsk->icsk_backoff = 0; |
2033 | return; | 2037 | return; |
2034 | } | 2038 | } |
@@ -2036,19 +2040,19 @@ void tcp_send_probe0(struct sock *sk) | |||
2036 | if (err <= 0) { | 2040 | if (err <= 0) { |
2037 | if (icsk->icsk_backoff < sysctl_tcp_retries2) | 2041 | if (icsk->icsk_backoff < sysctl_tcp_retries2) |
2038 | icsk->icsk_backoff++; | 2042 | icsk->icsk_backoff++; |
2039 | tp->probes_out++; | 2043 | icsk->icsk_probes_out++; |
2040 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, | 2044 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
2041 | min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX), | 2045 | min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX), |
2042 | TCP_RTO_MAX); | 2046 | TCP_RTO_MAX); |
2043 | } else { | 2047 | } else { |
2044 | /* If packet was not sent due to local congestion, | 2048 | /* If packet was not sent due to local congestion, |
2045 | * do not backoff and do not remember probes_out. | 2049 | * do not backoff and do not remember icsk_probes_out. |
2046 | * Let local senders to fight for local resources. | 2050 | * Let local senders to fight for local resources. |
2047 | * | 2051 | * |
2048 | * Use accumulated backoff yet. | 2052 | * Use accumulated backoff yet. |
2049 | */ | 2053 | */ |
2050 | if (!tp->probes_out) | 2054 | if (!icsk->icsk_probes_out) |
2051 | tp->probes_out=1; | 2055 | icsk->icsk_probes_out = 1; |
2052 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, | 2056 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
2053 | min(icsk->icsk_rto << icsk->icsk_backoff, | 2057 | min(icsk->icsk_rto << icsk->icsk_backoff, |
2054 | TCP_RESOURCE_PROBE_INTERVAL), | 2058 | TCP_RESOURCE_PROBE_INTERVAL), |