From 7c657876b63cb1d8a2ec06f8fc6c37bb8412e66c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 9 Aug 2005 20:14:34 -0700 Subject: [DCCP]: Initial implementation Development to this point was done on a subversion repository at: http://oops.ghostprotocols.net:81/cgi-bin/viewcvs.cgi/dccp-2.6/ This repository will be kept at this site for the foreseable future, so that interested parties can see the history of this code, attributions, etc. If I ever decide to take this offline I'll provide the full history at some other suitable place. Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 406 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 406 insertions(+) create mode 100644 net/dccp/output.c (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c new file mode 100644 index 000000000000..22ca2910d4f2 --- /dev/null +++ b/net/dccp/output.c @@ -0,0 +1,406 @@ +/* + * net/dccp/output.c + * + * An implementation of the DCCP protocol + * Arnaldo Carvalho de Melo + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#include +#include +#include + +#include + +#include "ccid.h" +#include "dccp.h" + +static inline void dccp_event_ack_sent(struct sock *sk) +{ + inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK); +} + +/* + * All SKB's seen here are completely headerless. It is our + * job to build the DCCP header, and pass the packet down to + * IP so it can do the same plus pass the packet off to the + * device. + */ +int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) +{ + if (likely(skb != NULL)) { + const struct inet_sock *inet = inet_sk(sk); + struct dccp_sock *dp = dccp_sk(sk); + struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); + struct dccp_hdr *dh; + /* XXX For now we're using only 48 bits sequence numbers */ + const int dccp_header_size = sizeof(*dh) + + sizeof(struct dccp_hdr_ext) + + dccp_packet_hdr_len(dcb->dccpd_type); + int err, set_ack = 1; + u64 ackno = dp->dccps_gsr; + + /* + * FIXME: study DCCP_PKT_SYNC[ACK] to see what is the right thing + * to do here... + */ + dccp_inc_seqno(&dp->dccps_gss); + + dcb->dccpd_seq = dp->dccps_gss; + dccp_insert_options(sk, skb); + + switch (dcb->dccpd_type) { + case DCCP_PKT_DATA: + set_ack = 0; + break; + case DCCP_PKT_SYNC: + case DCCP_PKT_SYNCACK: + ackno = dcb->dccpd_seq; + break; + } + + skb->h.raw = skb_push(skb, dccp_header_size); + dh = dccp_hdr(skb); + /* Data packets are not cloned as they are never retransmitted */ + if (skb_cloned(skb)) + skb_set_owner_w(skb, sk); + + /* Build DCCP header and checksum it. */ + memset(dh, 0, dccp_header_size); + dh->dccph_type = dcb->dccpd_type; + dh->dccph_sport = inet->sport; + dh->dccph_dport = inet->dport; + dh->dccph_doff = (dccp_header_size + dcb->dccpd_opt_len) / 4; + dh->dccph_ccval = dcb->dccpd_ccval; + /* XXX For now we're using only 48 bits sequence numbers */ + dh->dccph_x = 1; + + dp->dccps_awh = dp->dccps_gss; + dccp_hdr_set_seq(dh, dp->dccps_gss); + if (set_ack) + dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), ackno); + + switch (dcb->dccpd_type) { + case DCCP_PKT_REQUEST: + dccp_hdr_request(skb)->dccph_req_service = dcb->dccpd_service; + break; + case DCCP_PKT_RESET: + dccp_hdr_reset(skb)->dccph_reset_code = dcb->dccpd_reset_code; + break; + } + + dh->dccph_checksum = dccp_v4_checksum(skb); + + if (dcb->dccpd_type == DCCP_PKT_ACK || + dcb->dccpd_type == DCCP_PKT_DATAACK) + dccp_event_ack_sent(sk); + + DCCP_INC_STATS(DCCP_MIB_OUTSEGS); + + err = ip_queue_xmit(skb, 0); + if (err <= 0) + return err; + + /* NET_XMIT_CN is special. It does not guarantee, + * that this packet is lost. It tells that device + * is about to start to drop packets or already + * drops some packets of the same priority and + * invokes us to send less aggressively. + */ + return err == NET_XMIT_CN ? 0 : err; + } + return -ENOBUFS; +} + +unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu) +{ + struct dccp_sock *dp = dccp_sk(sk); + int mss_now; + + /* + * FIXME: we really should be using the af_specific thing to support IPv6. + * mss_now = pmtu - tp->af_specific->net_header_len - sizeof(struct dccp_hdr) - sizeof(struct dccp_hdr_ext); + */ + mss_now = pmtu - sizeof(struct iphdr) - sizeof(struct dccp_hdr) - sizeof(struct dccp_hdr_ext); + + /* Now subtract optional transport overhead */ + mss_now -= dp->dccps_ext_header_len; + + /* + * FIXME: this should come from the CCID infrastructure, where, say, + * TFRC will say it wants TIMESTAMPS, ELAPSED time, etc, for now lets + * put a rough estimate for NDP + TIMESTAMP + TIMESTAMP_ECHO + ELAPSED + * TIME + TFRC_OPT_LOSS_EVENT_RATE + TFRC_OPT_RECEIVE_RATE + padding to + * make it a multiple of 4 + */ + + mss_now -= ((5 + 6 + 10 + 6 + 6 + 6 + 3) / 4) * 4; + + /* And store cached results */ + dp->dccps_pmtu_cookie = pmtu; + dp->dccps_mss_cache = mss_now; + + return mss_now; +} + +int dccp_retransmit_skb(struct sock *sk, struct sk_buff *skb) +{ + if (inet_sk_rebuild_header(sk) != 0) + return -EHOSTUNREACH; /* Routing failure or similar. */ + + return dccp_transmit_skb(sk, (skb_cloned(skb) ? + pskb_copy(skb, GFP_ATOMIC): + skb_clone(skb, GFP_ATOMIC))); +} + +struct sk_buff *dccp_make_response(struct sock *sk, struct dst_entry *dst, + struct request_sock *req) +{ + struct dccp_hdr *dh; + const int dccp_header_size = sizeof(struct dccp_hdr) + + sizeof(struct dccp_hdr_ext) + + sizeof(struct dccp_hdr_response); + struct sk_buff *skb = sock_wmalloc(sk, MAX_HEADER + DCCP_MAX_OPT_LEN + + dccp_header_size, 1, + GFP_ATOMIC); + if (skb == NULL) + return NULL; + + /* Reserve space for headers. */ + skb_reserve(skb, MAX_HEADER + DCCP_MAX_OPT_LEN + dccp_header_size); + + skb->dst = dst_clone(dst); + skb->csum = 0; + + DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESPONSE; + DCCP_SKB_CB(skb)->dccpd_seq = dccp_rsk(req)->dreq_iss; + dccp_insert_options(sk, skb); + + skb->h.raw = skb_push(skb, dccp_header_size); + + dh = dccp_hdr(skb); + memset(dh, 0, dccp_header_size); + + dh->dccph_sport = inet_sk(sk)->sport; + dh->dccph_dport = inet_rsk(req)->rmt_port; + dh->dccph_doff = (dccp_header_size + DCCP_SKB_CB(skb)->dccpd_opt_len) / 4; + dh->dccph_type = DCCP_PKT_RESPONSE; + dh->dccph_x = 1; + dccp_hdr_set_seq(dh, dccp_rsk(req)->dreq_iss); + dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dccp_rsk(req)->dreq_isr); + + dh->dccph_checksum = dccp_v4_checksum(skb); + + DCCP_INC_STATS(DCCP_MIB_OUTSEGS); + return skb; +} + +struct sk_buff *dccp_make_reset(struct sock *sk, struct dst_entry *dst, + const enum dccp_reset_codes code) + +{ + struct dccp_hdr *dh; + struct dccp_sock *dp = dccp_sk(sk); + const int dccp_header_size = sizeof(struct dccp_hdr) + + sizeof(struct dccp_hdr_ext) + + sizeof(struct dccp_hdr_reset); + struct sk_buff *skb = sock_wmalloc(sk, MAX_HEADER + DCCP_MAX_OPT_LEN + + dccp_header_size, 1, + GFP_ATOMIC); + if (skb == NULL) + return NULL; + + /* Reserve space for headers. */ + skb_reserve(skb, MAX_HEADER + DCCP_MAX_OPT_LEN + dccp_header_size); + + skb->dst = dst_clone(dst); + skb->csum = 0; + + dccp_inc_seqno(&dp->dccps_gss); + + DCCP_SKB_CB(skb)->dccpd_reset_code = code; + DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESET; + DCCP_SKB_CB(skb)->dccpd_seq = dp->dccps_gss; + dccp_insert_options(sk, skb); + + skb->h.raw = skb_push(skb, dccp_header_size); + + dh = dccp_hdr(skb); + memset(dh, 0, dccp_header_size); + + dh->dccph_sport = inet_sk(sk)->sport; + dh->dccph_dport = inet_sk(sk)->dport; + dh->dccph_doff = (dccp_header_size + DCCP_SKB_CB(skb)->dccpd_opt_len) / 4; + dh->dccph_type = DCCP_PKT_RESET; + dh->dccph_x = 1; + dccp_hdr_set_seq(dh, dp->dccps_gss); + dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dp->dccps_gsr); + + dccp_hdr_reset(skb)->dccph_reset_code = code; + + dh->dccph_checksum = dccp_v4_checksum(skb); + + DCCP_INC_STATS(DCCP_MIB_OUTSEGS); + return skb; +} + +/* + * Do all connect socket setups that can be done AF independent. + */ +static inline void dccp_connect_init(struct sock *sk) +{ + struct dst_entry *dst = __sk_dst_get(sk); + struct inet_connection_sock *icsk = inet_csk(sk); + + sk->sk_err = 0; + sock_reset_flag(sk, SOCK_DONE); + + dccp_sync_mss(sk, dst_mtu(dst)); + + /* + * FIXME: set dp->{dccps_swh,dccps_swl}, with + * something like dccp_inc_seq + */ + + icsk->icsk_retransmits = 0; +} + +int dccp_connect(struct sock *sk) +{ + struct sk_buff *skb; + struct inet_connection_sock *icsk = inet_csk(sk); + + dccp_connect_init(sk); + + skb = alloc_skb(MAX_DCCP_HEADER + 15, sk->sk_allocation); + if (unlikely(skb == NULL)) + return -ENOBUFS; + + /* Reserve space for headers. */ + skb_reserve(skb, MAX_DCCP_HEADER); + + DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_REQUEST; + /* FIXME: set service to something meaningful, coming + * from userspace*/ + DCCP_SKB_CB(skb)->dccpd_service = 0; + skb->csum = 0; + skb_set_owner_w(skb, sk); + + BUG_TRAP(sk->sk_send_head == NULL); + sk->sk_send_head = skb; + dccp_transmit_skb(sk, skb_clone(skb, GFP_KERNEL)); + DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS); + + /* Timer for repeating the REQUEST until an answer. */ + inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, icsk->icsk_rto, TCP_RTO_MAX); + return 0; +} + +void dccp_send_ack(struct sock *sk) +{ + /* If we have been reset, we may not send again. */ + if (sk->sk_state != DCCP_CLOSED) { + struct sk_buff *skb = alloc_skb(MAX_DCCP_HEADER, GFP_ATOMIC); + + if (skb == NULL) { + inet_csk_schedule_ack(sk); + inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN; + inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, TCP_DELACK_MAX, TCP_RTO_MAX); + return; + } + + /* Reserve space for headers */ + skb_reserve(skb, MAX_DCCP_HEADER); + skb->csum = 0; + DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_ACK; + skb_set_owner_w(skb, sk); + dccp_transmit_skb(sk, skb); + } +} + +EXPORT_SYMBOL_GPL(dccp_send_ack); + +void dccp_send_delayed_ack(struct sock *sk) +{ + struct inet_connection_sock *icsk = inet_csk(sk); + /* + * FIXME: tune this timer. elapsed time fixes the skew, so no problem + * with using 2s, and active senders also piggyback the ACK into a + * DATAACK packet, so this is really for quiescent senders. + */ + unsigned long timeout = jiffies + 2 * HZ; + + /* Use new timeout only if there wasn't a older one earlier. */ + if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) { + /* If delack timer was blocked or is about to expire, + * send ACK now. + * + * FIXME: check the "about to expire" part + */ + if (icsk->icsk_ack.blocked) { + dccp_send_ack(sk); + return; + } + + if (!time_before(timeout, icsk->icsk_ack.timeout)) + timeout = icsk->icsk_ack.timeout; + } + icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER; + icsk->icsk_ack.timeout = timeout; + sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout); +} + +void dccp_send_sync(struct sock *sk, u64 seq) +{ + /* + * We are not putting this on the write queue, so + * dccp_transmit_skb() will set the ownership to this + * sock. + */ + struct sk_buff *skb = alloc_skb(MAX_DCCP_HEADER, GFP_ATOMIC); + + if (skb == NULL) + /* FIXME: how to make sure the sync is sent? */ + return; + + /* Reserve space for headers and prepare control bits. */ + skb_reserve(skb, MAX_DCCP_HEADER); + skb->csum = 0; + DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_SYNC; + DCCP_SKB_CB(skb)->dccpd_seq = seq; + + skb_set_owner_w(skb, sk); + dccp_transmit_skb(sk, skb); +} + +/* Send a DCCP_PKT_CLOSE/CLOSEREQ. The caller locks the socket for us. This cannot be + * allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under any circumstances. + */ +void dccp_send_close(struct sock *sk) +{ + struct dccp_sock *dp = dccp_sk(sk); + struct sk_buff *skb; + + /* Socket is locked, keep trying until memory is available. */ + for (;;) { + skb = alloc_skb(sk->sk_prot->max_header, GFP_KERNEL); + if (skb != NULL) + break; + yield(); + } + + /* Reserve space for headers and prepare control bits. */ + skb_reserve(skb, sk->sk_prot->max_header); + skb->csum = 0; + DCCP_SKB_CB(skb)->dccpd_type = dp->dccps_role == DCCP_ROLE_CLIENT ? DCCP_PKT_CLOSE : DCCP_PKT_CLOSEREQ; + + skb_set_owner_w(skb, sk); + dccp_transmit_skb(sk, skb); + + ccid_hc_rx_exit(dp->dccps_hc_rx_ccid, sk); + ccid_hc_tx_exit(dp->dccps_hc_tx_ccid, sk); +} -- cgit v1.2.2 From 95b81ef794278c835b321f6376b0522cd5df59b7 Mon Sep 17 00:00:00 2001 From: Yoshifumi Nishida Date: Tue, 9 Aug 2005 20:15:35 -0700 Subject: [DCCP]: Fix checksum routines Signed-off-by: Yoshifumi Nishida Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index 22ca2910d4f2..4945eaa9d1a4 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -93,7 +93,8 @@ int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) break; } - dh->dccph_checksum = dccp_v4_checksum(skb); + dh->dccph_checksum = dccp_v4_checksum(skb, inet->saddr, + inet->daddr); if (dcb->dccpd_type == DCCP_PKT_ACK || dcb->dccpd_type == DCCP_PKT_DATAACK) @@ -193,7 +194,8 @@ struct sk_buff *dccp_make_response(struct sock *sk, struct dst_entry *dst, dccp_hdr_set_seq(dh, dccp_rsk(req)->dreq_iss); dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dccp_rsk(req)->dreq_isr); - dh->dccph_checksum = dccp_v4_checksum(skb); + dh->dccph_checksum = dccp_v4_checksum(skb, inet_rsk(req)->loc_addr, + inet_rsk(req)->rmt_addr); DCCP_INC_STATS(DCCP_MIB_OUTSEGS); return skb; @@ -242,7 +244,8 @@ struct sk_buff *dccp_make_reset(struct sock *sk, struct dst_entry *dst, dccp_hdr_reset(skb)->dccph_reset_code = code; - dh->dccph_checksum = dccp_v4_checksum(skb); + dh->dccph_checksum = dccp_v4_checksum(skb, inet_sk(sk)->saddr, + inet_sk(sk)->daddr); DCCP_INC_STATS(DCCP_MIB_OUTSEGS); return skb; -- cgit v1.2.2 From 27258ee54f8cd4a43d09319aa5448145afc2cb8d Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 9 Aug 2005 20:30:56 -0700 Subject: [DCCP]: Introduce dccp_write_xmit from code in dccp_sendmsg This way it gets closer to the TCP flow, where congestion window checks are done, it seems we can map ccid_hc_tx_send_packet in dccp_write_xmit to tcp_snd_wnd_test in tcp_write_xmit, a CCID2 decision should just fit in here as well... Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 38 +++++++++++++++++++++++++++++++++++++- 1 file changed, 37 insertions(+), 1 deletion(-) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index 4945eaa9d1a4..50292c0605fb 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -148,6 +148,41 @@ unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu) return mss_now; } +int dccp_write_xmit(struct sock *sk, struct sk_buff *skb, const int len) +{ + const struct dccp_sock *dp = dccp_sk(sk); + int err = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb, len); + + if (err == 0) { + const struct dccp_ackpkts *ap = dp->dccps_hc_rx_ackpkts; + struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); + + if (sk->sk_state == DCCP_PARTOPEN) { + /* See 8.1.5. Handshake Completion */ + inet_csk_schedule_ack(sk); + inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, + inet_csk(sk)->icsk_rto, + DCCP_RTO_MAX); + dcb->dccpd_type = DCCP_PKT_DATAACK; + /* + * FIXME: we really should have a + * dccps_ack_pending or use icsk. + */ + } else if (inet_csk_ack_scheduled(sk) || + (dp->dccps_options.dccpo_send_ack_vector && + ap->dccpap_buf_ackno != DCCP_MAX_SEQNO + 1 && + ap->dccpap_ack_seqno == DCCP_MAX_SEQNO + 1)) + dcb->dccpd_type = DCCP_PKT_DATAACK; + else + dcb->dccpd_type = DCCP_PKT_DATA; + + err = dccp_transmit_skb(sk, skb); + ccid_hc_tx_packet_sent(dp->dccps_hc_tx_ccid, sk, 0, len); + } + + return err; +} + int dccp_retransmit_skb(struct sock *sk, struct sk_buff *skb) { if (inet_sk_rebuild_header(sk) != 0) @@ -299,7 +334,8 @@ int dccp_connect(struct sock *sk) DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS); /* Timer for repeating the REQUEST until an answer. */ - inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, icsk->icsk_rto, TCP_RTO_MAX); + inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, + icsk->icsk_rto, DCCP_RTO_MAX); return 0; } -- cgit v1.2.2 From 7690af3fff7633e40b1b9950eb8489129251d074 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 13 Aug 2005 20:34:54 -0300 Subject: [DCCP]: Just reflow the source code to fit in 80 columns Andrew Morton should be happy now 8) Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 44 +++++++++++++++++++++++++++++--------------- 1 file changed, 29 insertions(+), 15 deletions(-) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index 50292c0605fb..dcc061bed924 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -40,13 +40,13 @@ int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) /* XXX For now we're using only 48 bits sequence numbers */ const int dccp_header_size = sizeof(*dh) + sizeof(struct dccp_hdr_ext) + - dccp_packet_hdr_len(dcb->dccpd_type); + dccp_packet_hdr_len(dcb->dccpd_type); int err, set_ack = 1; u64 ackno = dp->dccps_gsr; /* - * FIXME: study DCCP_PKT_SYNC[ACK] to see what is the right thing - * to do here... + * FIXME: study DCCP_PKT_SYNC[ACK] to see what is the right + * thing to do here... */ dccp_inc_seqno(&dp->dccps_gss); @@ -65,7 +65,9 @@ int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) skb->h.raw = skb_push(skb, dccp_header_size); dh = dccp_hdr(skb); - /* Data packets are not cloned as they are never retransmitted */ + /* + * Data packets are not cloned as they are never retransmitted + */ if (skb_cloned(skb)) skb_set_owner_w(skb, sk); @@ -86,10 +88,12 @@ int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) switch (dcb->dccpd_type) { case DCCP_PKT_REQUEST: - dccp_hdr_request(skb)->dccph_req_service = dcb->dccpd_service; + dccp_hdr_request(skb)->dccph_req_service = + dcb->dccpd_service; break; case DCCP_PKT_RESET: - dccp_hdr_reset(skb)->dccph_reset_code = dcb->dccpd_reset_code; + dccp_hdr_reset(skb)->dccph_reset_code = + dcb->dccpd_reset_code; break; } @@ -123,10 +127,13 @@ unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu) int mss_now; /* - * FIXME: we really should be using the af_specific thing to support IPv6. - * mss_now = pmtu - tp->af_specific->net_header_len - sizeof(struct dccp_hdr) - sizeof(struct dccp_hdr_ext); + * FIXME: we really should be using the af_specific thing to support + * IPv6. + * mss_now = pmtu - tp->af_specific->net_header_len - + * sizeof(struct dccp_hdr) - sizeof(struct dccp_hdr_ext); */ - mss_now = pmtu - sizeof(struct iphdr) - sizeof(struct dccp_hdr) - sizeof(struct dccp_hdr_ext); + mss_now = pmtu - sizeof(struct iphdr) - sizeof(struct dccp_hdr) - + sizeof(struct dccp_hdr_ext); /* Now subtract optional transport overhead */ mss_now -= dp->dccps_ext_header_len; @@ -223,7 +230,8 @@ struct sk_buff *dccp_make_response(struct sock *sk, struct dst_entry *dst, dh->dccph_sport = inet_sk(sk)->sport; dh->dccph_dport = inet_rsk(req)->rmt_port; - dh->dccph_doff = (dccp_header_size + DCCP_SKB_CB(skb)->dccpd_opt_len) / 4; + dh->dccph_doff = (dccp_header_size + + DCCP_SKB_CB(skb)->dccpd_opt_len) / 4; dh->dccph_type = DCCP_PKT_RESPONSE; dh->dccph_x = 1; dccp_hdr_set_seq(dh, dccp_rsk(req)->dreq_iss); @@ -271,7 +279,8 @@ struct sk_buff *dccp_make_reset(struct sock *sk, struct dst_entry *dst, dh->dccph_sport = inet_sk(sk)->sport; dh->dccph_dport = inet_sk(sk)->dport; - dh->dccph_doff = (dccp_header_size + DCCP_SKB_CB(skb)->dccpd_opt_len) / 4; + dh->dccph_doff = (dccp_header_size + + DCCP_SKB_CB(skb)->dccpd_opt_len) / 4; dh->dccph_type = DCCP_PKT_RESET; dh->dccph_x = 1; dccp_hdr_set_seq(dh, dp->dccps_gss); @@ -348,7 +357,9 @@ void dccp_send_ack(struct sock *sk) if (skb == NULL) { inet_csk_schedule_ack(sk); inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN; - inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, TCP_DELACK_MAX, TCP_RTO_MAX); + inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, + TCP_DELACK_MAX, + DCCP_RTO_MAX); return; } @@ -416,8 +427,10 @@ void dccp_send_sync(struct sock *sk, u64 seq) dccp_transmit_skb(sk, skb); } -/* Send a DCCP_PKT_CLOSE/CLOSEREQ. The caller locks the socket for us. This cannot be - * allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under any circumstances. +/* + * Send a DCCP_PKT_CLOSE/CLOSEREQ. The caller locks the socket for us. This + * cannot be allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under + * any circumstances. */ void dccp_send_close(struct sock *sk) { @@ -435,7 +448,8 @@ void dccp_send_close(struct sock *sk) /* Reserve space for headers and prepare control bits. */ skb_reserve(skb, sk->sk_prot->max_header); skb->csum = 0; - DCCP_SKB_CB(skb)->dccpd_type = dp->dccps_role == DCCP_ROLE_CLIENT ? DCCP_PKT_CLOSE : DCCP_PKT_CLOSEREQ; + DCCP_SKB_CB(skb)->dccpd_type = dp->dccps_role == DCCP_ROLE_CLIENT ? + DCCP_PKT_CLOSE : DCCP_PKT_CLOSEREQ; skb_set_owner_w(skb, sk); dccp_transmit_skb(sk, skb); -- cgit v1.2.2 From e92ae93a8aa66aea12935420cb22d4df1c18d023 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 17 Aug 2005 03:10:59 -0300 Subject: [DCCP]: Send SYNCACK packets in response to SYNC packets Also fix step 6 when receiving SYNC or SYNCACK packets, i.e. we were not using the updated swl. Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index dcc061bed924..384fd0920983 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -404,7 +404,8 @@ void dccp_send_delayed_ack(struct sock *sk) sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout); } -void dccp_send_sync(struct sock *sk, u64 seq) +void dccp_send_sync(struct sock *sk, const u64 seq, + const enum dccp_pkt_type pkt_type) { /* * We are not putting this on the write queue, so @@ -420,7 +421,7 @@ void dccp_send_sync(struct sock *sk, u64 seq) /* Reserve space for headers and prepare control bits. */ skb_reserve(skb, MAX_DCCP_HEADER); skb->csum = 0; - DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_SYNC; + DCCP_SKB_CB(skb)->dccpd_type = pkt_type; DCCP_SKB_CB(skb)->dccpd_seq = seq; skb_set_owner_w(skb, sk); -- cgit v1.2.2 From 24117727b753426d85ba09671c24854834f81b2c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sun, 21 Aug 2005 05:40:16 -0300 Subject: [DCCP]: Fix ackno setting in SYNC/SYNCACK packets Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index 384fd0920983..708fc3c0a969 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -44,15 +44,8 @@ int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) int err, set_ack = 1; u64 ackno = dp->dccps_gsr; - /* - * FIXME: study DCCP_PKT_SYNC[ACK] to see what is the right - * thing to do here... - */ dccp_inc_seqno(&dp->dccps_gss); - dcb->dccpd_seq = dp->dccps_gss; - dccp_insert_options(sk, skb); - switch (dcb->dccpd_type) { case DCCP_PKT_DATA: set_ack = 0; @@ -62,6 +55,9 @@ int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) ackno = dcb->dccpd_seq; break; } + + dcb->dccpd_seq = dp->dccps_gss; + dccp_insert_options(sk, skb); skb->h.raw = skb_push(skb, dccp_header_size); dh = dccp_hdr(skb); -- cgit v1.2.2 From 7ad07e7cf343181002c10c39d3f57a88e4903d4f Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 23 Aug 2005 21:50:06 -0700 Subject: [DCCP]: Implement the CLOSING timer So that we retransmit CLOSE/CLOSEREQ packets till they elicit an answer or we hit a timeout. Most of the machinery uses TCP approaches, this code has to be polished & audited, but this is better than we had before. Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index 708fc3c0a969..630ca7741022 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -96,8 +96,7 @@ int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) dh->dccph_checksum = dccp_v4_checksum(skb, inet->saddr, inet->daddr); - if (dcb->dccpd_type == DCCP_PKT_ACK || - dcb->dccpd_type == DCCP_PKT_DATAACK) + if (set_ack) dccp_event_ack_sent(sk); DCCP_INC_STATS(DCCP_MIB_OUTSEGS); @@ -429,18 +428,15 @@ void dccp_send_sync(struct sock *sk, const u64 seq, * cannot be allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under * any circumstances. */ -void dccp_send_close(struct sock *sk) +void dccp_send_close(struct sock *sk, const int active) { struct dccp_sock *dp = dccp_sk(sk); struct sk_buff *skb; + const unsigned int prio = active ? GFP_KERNEL : GFP_ATOMIC; - /* Socket is locked, keep trying until memory is available. */ - for (;;) { - skb = alloc_skb(sk->sk_prot->max_header, GFP_KERNEL); - if (skb != NULL) - break; - yield(); - } + skb = alloc_skb(sk->sk_prot->max_header, prio); + if (skb == NULL) + return; /* Reserve space for headers and prepare control bits. */ skb_reserve(skb, sk->sk_prot->max_header); @@ -449,7 +445,12 @@ void dccp_send_close(struct sock *sk) DCCP_PKT_CLOSE : DCCP_PKT_CLOSEREQ; skb_set_owner_w(skb, sk); - dccp_transmit_skb(sk, skb); + if (active) { + BUG_TRAP(sk->sk_send_head == NULL); + sk->sk_send_head = skb; + dccp_transmit_skb(sk, skb_clone(skb, prio)); + } else + dccp_transmit_skb(sk, skb); ccid_hc_rx_exit(dp->dccps_hc_rx_ccid, sk); ccid_hc_tx_exit(dp->dccps_hc_tx_ccid, sk); -- cgit v1.2.2 From a4beb1b64f5846e216bf2c439022df480151902a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 23 Aug 2005 21:50:45 -0700 Subject: [DCCP]: Send a DATAACK packet when we have a TIMESTAMP_ECHO pending Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 1 + 1 file changed, 1 insertion(+) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index 630ca7741022..f96dedd3ad5e 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -171,6 +171,7 @@ int dccp_write_xmit(struct sock *sk, struct sk_buff *skb, const int len) * dccps_ack_pending or use icsk. */ } else if (inet_csk_ack_scheduled(sk) || + dp->dccps_timestamp_echo != 0 || (dp->dccps_options.dccpo_send_ack_vector && ap->dccpap_buf_ackno != DCCP_MAX_SEQNO + 1 && ap->dccpap_ack_seqno == DCCP_MAX_SEQNO + 1)) -- cgit v1.2.2 From d6809c12b3334a929c39bf08ea63bd819e0500f7 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 27 Aug 2005 03:06:35 -0300 Subject: [DCCP]: Introduce dccp_wait_for_ccid and use it in dccp_write_xmit This is not quite what I think we should have long term but improves performance for now, so lets use it till we get CCID3 working well, then we can think about using sk_write_queue, perhaps using some ideas from Juwen Lai's old stack for 2.4.20. Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 61 +++++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 59 insertions(+), 2 deletions(-) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index f96dedd3ad5e..116f6db5678d 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -150,14 +150,71 @@ unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu) return mss_now; } -int dccp_write_xmit(struct sock *sk, struct sk_buff *skb, const int len) +/** + * dccp_wait_for_ccid - Wait for ccid to tell us we can send a packet + * @sk: socket to wait for + * @timeo: for how long + */ +static int dccp_wait_for_ccid(struct sock *sk, struct sk_buff *skb, + long *timeo) +{ + struct dccp_sock *dp = dccp_sk(sk); + DEFINE_WAIT(wait); + long delay; + int rc; + + while (1) { + prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE); + + if (sk->sk_err || (sk->sk_shutdown & SEND_SHUTDOWN)) + goto do_error; + if (!*timeo) + goto do_nonblock; + if (signal_pending(current)) + goto do_interrupted; + + rc = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb, + skb->len); + if (rc <= 0) + break; + delay = msecs_to_jiffies(rc); + if (delay > *timeo || delay < 0) + goto do_nonblock; + + sk->sk_write_pending++; + release_sock(sk); + *timeo -= schedule_timeout(delay); + lock_sock(sk); + sk->sk_write_pending--; + } +out: + finish_wait(sk->sk_sleep, &wait); + return rc; + +do_error: + rc = -EPIPE; + goto out; +do_nonblock: + rc = -EAGAIN; + goto out; +do_interrupted: + rc = sock_intr_errno(*timeo); + goto out; +} + +int dccp_write_xmit(struct sock *sk, struct sk_buff *skb, long *timeo) { const struct dccp_sock *dp = dccp_sk(sk); - int err = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb, len); + int err = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb, + skb->len); + + if (err > 0) + err = dccp_wait_for_ccid(sk, skb, timeo); if (err == 0) { const struct dccp_ackpkts *ap = dp->dccps_hc_rx_ackpkts; struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); + const int len = skb->len; if (sk->sk_state == DCCP_PARTOPEN) { /* See 8.1.5. Handshake Completion */ -- cgit v1.2.2 From c530cfb1ce1e8f230744c3f3bd86771f50725053 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 29 Aug 2005 02:15:54 -0300 Subject: [CCID3]: Call sk->sk_write_space(sk) when receiving a feedback packet This makes the send rate calculations behave way more closely to what is specified, with the jitter previously seen on x and x_recv disappearing completely on non lossy setups. This resembles the tcp_data_snd_check code, that possibly we'll end up using in DCCP as well, perhaps moving this code to inet_connection_sock. For now I'm doing the simplest implementation tho. Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: David S. Miller --- net/dccp/output.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'net/dccp/output.c') diff --git a/net/dccp/output.c b/net/dccp/output.c index 116f6db5678d..28de157a4326 100644 --- a/net/dccp/output.c +++ b/net/dccp/output.c @@ -150,6 +150,19 @@ unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu) return mss_now; } +void dccp_write_space(struct sock *sk) +{ + read_lock(&sk->sk_callback_lock); + + if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) + wake_up_interruptible(sk->sk_sleep); + /* Should agree with poll, otherwise some programs break */ + if (sock_writeable(sk)) + sk_wake_async(sk, 2, POLL_OUT); + + read_unlock(&sk->sk_callback_lock); +} + /** * dccp_wait_for_ccid - Wait for ccid to tell us we can send a packet * @sk: socket to wait for -- cgit v1.2.2