aboutsummaryrefslogtreecommitdiffstats
path: root/net/dccp
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-10-23 14:47:02 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2010-10-23 14:47:02 -0400
commit5f05647dd81c11a6a165ccc8f0c1370b16f3bcb0 (patch)
tree7851ef1c93aa1aba7ef327ca4b75fd35e6d10f29 /net/dccp
parent02f36038c568111ad4fc433f6fa760ff5e38fab4 (diff)
parentec37a48d1d16c30b655ac5280209edf52a6775d4 (diff)
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6: (1699 commits) bnx2/bnx2x: Unsupported Ethtool operations should return -EINVAL. vlan: Calling vlan_hwaccel_do_receive() is always valid. tproxy: use the interface primary IP address as a default value for --on-ip tproxy: added IPv6 support to the socket match cxgb3: function namespace cleanup tproxy: added IPv6 support to the TPROXY target tproxy: added IPv6 socket lookup function to nf_tproxy_core be2net: Changes to use only priority codes allowed by f/w tproxy: allow non-local binds of IPv6 sockets if IP_TRANSPARENT is enabled tproxy: added tproxy sockopt interface in the IPV6 layer tproxy: added udp6_lib_lookup function tproxy: added const specifiers to udp lookup functions tproxy: split off ipv6 defragmentation to a separate module l2tp: small cleanup nf_nat: restrict ICMP translation for embedded header can: mcp251x: fix generation of error frames can: mcp251x: fix endless loop in interrupt handler if CANINTF_MERRF is set can-raw: add msg_flags to distinguish local traffic 9p: client code cleanup rds: make local functions/variables static ... Fix up conflicts in net/core/dev.c, drivers/net/pcmcia/smc91c92_cs.c and drivers/net/wireless/ath/ath9k/debug.c as per David
Diffstat (limited to 'net/dccp')
-rw-r--r--net/dccp/ccid.h52
-rw-r--r--net/dccp/ccids/Kconfig31
-rw-r--r--net/dccp/ccids/ccid2.c289
-rw-r--r--net/dccp/ccids/ccid2.h35
-rw-r--r--net/dccp/ccids/ccid3.c256
-rw-r--r--net/dccp/ccids/ccid3.h51
-rw-r--r--net/dccp/ccids/lib/loss_interval.c2
-rw-r--r--net/dccp/ccids/lib/packet_history.c39
-rw-r--r--net/dccp/ccids/lib/packet_history.h22
-rw-r--r--net/dccp/ccids/lib/tfrc.h1
-rw-r--r--net/dccp/ccids/lib/tfrc_equation.c14
-rw-r--r--net/dccp/dccp.h46
-rw-r--r--net/dccp/feat.c10
-rw-r--r--net/dccp/feat.h1
-rw-r--r--net/dccp/input.c20
-rw-r--r--net/dccp/ipv4.c10
-rw-r--r--net/dccp/ipv6.c10
-rw-r--r--net/dccp/minisocks.c30
-rw-r--r--net/dccp/options.c31
-rw-r--r--net/dccp/output.c20
-rw-r--r--net/dccp/proto.c50
21 files changed, 410 insertions, 610 deletions
diff --git a/net/dccp/ccid.h b/net/dccp/ccid.h
index 6df6f8ac9636..117fb093dcaf 100644
--- a/net/dccp/ccid.h
+++ b/net/dccp/ccid.h
@@ -62,22 +62,18 @@ struct ccid_operations {
62 void (*ccid_hc_tx_exit)(struct sock *sk); 62 void (*ccid_hc_tx_exit)(struct sock *sk);
63 void (*ccid_hc_rx_packet_recv)(struct sock *sk, 63 void (*ccid_hc_rx_packet_recv)(struct sock *sk,
64 struct sk_buff *skb); 64 struct sk_buff *skb);
65 int (*ccid_hc_rx_parse_options)(struct sock *sk, 65 int (*ccid_hc_rx_parse_options)(struct sock *sk, u8 pkt,
66 unsigned char option, 66 u8 opt, u8 *val, u8 len);
67 unsigned char len, u16 idx,
68 unsigned char* value);
69 int (*ccid_hc_rx_insert_options)(struct sock *sk, 67 int (*ccid_hc_rx_insert_options)(struct sock *sk,
70 struct sk_buff *skb); 68 struct sk_buff *skb);
71 void (*ccid_hc_tx_packet_recv)(struct sock *sk, 69 void (*ccid_hc_tx_packet_recv)(struct sock *sk,
72 struct sk_buff *skb); 70 struct sk_buff *skb);
73 int (*ccid_hc_tx_parse_options)(struct sock *sk, 71 int (*ccid_hc_tx_parse_options)(struct sock *sk, u8 pkt,
74 unsigned char option, 72 u8 opt, u8 *val, u8 len);
75 unsigned char len, u16 idx,
76 unsigned char* value);
77 int (*ccid_hc_tx_send_packet)(struct sock *sk, 73 int (*ccid_hc_tx_send_packet)(struct sock *sk,
78 struct sk_buff *skb); 74 struct sk_buff *skb);
79 void (*ccid_hc_tx_packet_sent)(struct sock *sk, 75 void (*ccid_hc_tx_packet_sent)(struct sock *sk,
80 int more, unsigned int len); 76 unsigned int len);
81 void (*ccid_hc_rx_get_info)(struct sock *sk, 77 void (*ccid_hc_rx_get_info)(struct sock *sk,
82 struct tcp_info *info); 78 struct tcp_info *info);
83 void (*ccid_hc_tx_get_info)(struct sock *sk, 79 void (*ccid_hc_tx_get_info)(struct sock *sk,
@@ -148,10 +144,10 @@ static inline int ccid_hc_tx_send_packet(struct ccid *ccid, struct sock *sk,
148} 144}
149 145
150static inline void ccid_hc_tx_packet_sent(struct ccid *ccid, struct sock *sk, 146static inline void ccid_hc_tx_packet_sent(struct ccid *ccid, struct sock *sk,
151 int more, unsigned int len) 147 unsigned int len)
152{ 148{
153 if (ccid->ccid_ops->ccid_hc_tx_packet_sent != NULL) 149 if (ccid->ccid_ops->ccid_hc_tx_packet_sent != NULL)
154 ccid->ccid_ops->ccid_hc_tx_packet_sent(sk, more, len); 150 ccid->ccid_ops->ccid_hc_tx_packet_sent(sk, len);
155} 151}
156 152
157static inline void ccid_hc_rx_packet_recv(struct ccid *ccid, struct sock *sk, 153static inline void ccid_hc_rx_packet_recv(struct ccid *ccid, struct sock *sk,
@@ -168,27 +164,31 @@ static inline void ccid_hc_tx_packet_recv(struct ccid *ccid, struct sock *sk,
168 ccid->ccid_ops->ccid_hc_tx_packet_recv(sk, skb); 164 ccid->ccid_ops->ccid_hc_tx_packet_recv(sk, skb);
169} 165}
170 166
167/**
168 * ccid_hc_tx_parse_options - Parse CCID-specific options sent by the receiver
169 * @pkt: type of packet that @opt appears on (RFC 4340, 5.1)
170 * @opt: the CCID-specific option type (RFC 4340, 5.8 and 10.3)
171 * @val: value of @opt
172 * @len: length of @val in bytes
173 */
171static inline int ccid_hc_tx_parse_options(struct ccid *ccid, struct sock *sk, 174static inline int ccid_hc_tx_parse_options(struct ccid *ccid, struct sock *sk,
172 unsigned char option, 175 u8 pkt, u8 opt, u8 *val, u8 len)
173 unsigned char len, u16 idx,
174 unsigned char* value)
175{ 176{
176 int rc = 0; 177 if (ccid->ccid_ops->ccid_hc_tx_parse_options == NULL)
177 if (ccid->ccid_ops->ccid_hc_tx_parse_options != NULL) 178 return 0;
178 rc = ccid->ccid_ops->ccid_hc_tx_parse_options(sk, option, len, idx, 179 return ccid->ccid_ops->ccid_hc_tx_parse_options(sk, pkt, opt, val, len);
179 value);
180 return rc;
181} 180}
182 181
182/**
183 * ccid_hc_rx_parse_options - Parse CCID-specific options sent by the sender
184 * Arguments are analogous to ccid_hc_tx_parse_options()
185 */
183static inline int ccid_hc_rx_parse_options(struct ccid *ccid, struct sock *sk, 186static inline int ccid_hc_rx_parse_options(struct ccid *ccid, struct sock *sk,
184 unsigned char option, 187 u8 pkt, u8 opt, u8 *val, u8 len)
185 unsigned char len, u16 idx,
186 unsigned char* value)
187{ 188{
188 int rc = 0; 189 if (ccid->ccid_ops->ccid_hc_rx_parse_options == NULL)
189 if (ccid->ccid_ops->ccid_hc_rx_parse_options != NULL) 190 return 0;
190 rc = ccid->ccid_ops->ccid_hc_rx_parse_options(sk, option, len, idx, value); 191 return ccid->ccid_ops->ccid_hc_rx_parse_options(sk, pkt, opt, val, len);
191 return rc;
192} 192}
193 193
194static inline int ccid_hc_rx_insert_options(struct ccid *ccid, struct sock *sk, 194static inline int ccid_hc_rx_insert_options(struct ccid *ccid, struct sock *sk,
diff --git a/net/dccp/ccids/Kconfig b/net/dccp/ccids/Kconfig
index 8408398cd44e..0581143cb800 100644
--- a/net/dccp/ccids/Kconfig
+++ b/net/dccp/ccids/Kconfig
@@ -47,37 +47,6 @@ config IP_DCCP_CCID3_DEBUG
47 47
48 If in doubt, say N. 48 If in doubt, say N.
49 49
50config IP_DCCP_CCID3_RTO
51 int "Use higher bound for nofeedback timer"
52 default 100
53 depends on IP_DCCP_CCID3 && EXPERIMENTAL
54 ---help---
55 Use higher lower bound for nofeedback timer expiration.
56
57 The TFRC nofeedback timer normally expires after the maximum of 4
58 RTTs and twice the current send interval (RFC 3448, 4.3). On LANs
59 with a small RTT this can mean a high processing load and reduced
60 performance, since then the nofeedback timer is triggered very
61 frequently.
62
63 This option enables to set a higher lower bound for the nofeedback
64 value. Values in units of milliseconds can be set here.
65
66 A value of 0 disables this feature by enforcing the value specified
67 in RFC 3448. The following values have been suggested as bounds for
68 experimental use:
69 * 16-20ms to match the typical multimedia inter-frame interval
70 * 100ms as a reasonable compromise [default]
71 * 1000ms corresponds to the lower TCP RTO bound (RFC 2988, 2.4)
72
73 The default of 100ms is a compromise between a large value for
74 efficient DCCP implementations, and a small value to avoid disrupting
75 the network in times of congestion.
76
77 The purpose of the nofeedback timer is to slow DCCP down when there
78 is serious network congestion: experimenting with larger values should
79 therefore not be performed on WANs.
80
81config IP_DCCP_TFRC_LIB 50config IP_DCCP_TFRC_LIB
82 def_bool y if IP_DCCP_CCID3 51 def_bool y if IP_DCCP_CCID3
83 52
diff --git a/net/dccp/ccids/ccid2.c b/net/dccp/ccids/ccid2.c
index 9b3ae9922be1..d850e291f87c 100644
--- a/net/dccp/ccids/ccid2.c
+++ b/net/dccp/ccids/ccid2.c
@@ -25,59 +25,14 @@
25 */ 25 */
26#include <linux/slab.h> 26#include <linux/slab.h>
27#include "../feat.h" 27#include "../feat.h"
28#include "../ccid.h"
29#include "../dccp.h"
30#include "ccid2.h" 28#include "ccid2.h"
31 29
32 30
33#ifdef CONFIG_IP_DCCP_CCID2_DEBUG 31#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
34static int ccid2_debug; 32static int ccid2_debug;
35#define ccid2_pr_debug(format, a...) DCCP_PR_DEBUG(ccid2_debug, format, ##a) 33#define ccid2_pr_debug(format, a...) DCCP_PR_DEBUG(ccid2_debug, format, ##a)
36
37static void ccid2_hc_tx_check_sanity(const struct ccid2_hc_tx_sock *hc)
38{
39 int len = 0;
40 int pipe = 0;
41 struct ccid2_seq *seqp = hc->tx_seqh;
42
43 /* there is data in the chain */
44 if (seqp != hc->tx_seqt) {
45 seqp = seqp->ccid2s_prev;
46 len++;
47 if (!seqp->ccid2s_acked)
48 pipe++;
49
50 while (seqp != hc->tx_seqt) {
51 struct ccid2_seq *prev = seqp->ccid2s_prev;
52
53 len++;
54 if (!prev->ccid2s_acked)
55 pipe++;
56
57 /* packets are sent sequentially */
58 BUG_ON(dccp_delta_seqno(seqp->ccid2s_seq,
59 prev->ccid2s_seq ) >= 0);
60 BUG_ON(time_before(seqp->ccid2s_sent,
61 prev->ccid2s_sent));
62
63 seqp = prev;
64 }
65 }
66
67 BUG_ON(pipe != hc->tx_pipe);
68 ccid2_pr_debug("len of chain=%d\n", len);
69
70 do {
71 seqp = seqp->ccid2s_prev;
72 len++;
73 } while (seqp != hc->tx_seqh);
74
75 ccid2_pr_debug("total len=%d\n", len);
76 BUG_ON(len != hc->tx_seqbufc * CCID2_SEQBUF_LEN);
77}
78#else 34#else
79#define ccid2_pr_debug(format, a...) 35#define ccid2_pr_debug(format, a...)
80#define ccid2_hc_tx_check_sanity(hc)
81#endif 36#endif
82 37
83static int ccid2_hc_tx_alloc_seq(struct ccid2_hc_tx_sock *hc) 38static int ccid2_hc_tx_alloc_seq(struct ccid2_hc_tx_sock *hc)
@@ -156,19 +111,10 @@ static void ccid2_change_l_ack_ratio(struct sock *sk, u32 val)
156 dp->dccps_l_ack_ratio = val; 111 dp->dccps_l_ack_ratio = val;
157} 112}
158 113
159static void ccid2_change_srtt(struct ccid2_hc_tx_sock *hc, long val)
160{
161 ccid2_pr_debug("change SRTT to %ld\n", val);
162 hc->tx_srtt = val;
163}
164
165static void ccid2_start_rto_timer(struct sock *sk);
166
167static void ccid2_hc_tx_rto_expire(unsigned long data) 114static void ccid2_hc_tx_rto_expire(unsigned long data)
168{ 115{
169 struct sock *sk = (struct sock *)data; 116 struct sock *sk = (struct sock *)data;
170 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk); 117 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
171 long s;
172 118
173 bh_lock_sock(sk); 119 bh_lock_sock(sk);
174 if (sock_owned_by_user(sk)) { 120 if (sock_owned_by_user(sk)) {
@@ -178,23 +124,19 @@ static void ccid2_hc_tx_rto_expire(unsigned long data)
178 124
179 ccid2_pr_debug("RTO_EXPIRE\n"); 125 ccid2_pr_debug("RTO_EXPIRE\n");
180 126
181 ccid2_hc_tx_check_sanity(hc);
182
183 /* back-off timer */ 127 /* back-off timer */
184 hc->tx_rto <<= 1; 128 hc->tx_rto <<= 1;
129 if (hc->tx_rto > DCCP_RTO_MAX)
130 hc->tx_rto = DCCP_RTO_MAX;
185 131
186 s = hc->tx_rto / HZ; 132 sk_reset_timer(sk, &hc->tx_rtotimer, jiffies + hc->tx_rto);
187 if (s > 60)
188 hc->tx_rto = 60 * HZ;
189
190 ccid2_start_rto_timer(sk);
191 133
192 /* adjust pipe, cwnd etc */ 134 /* adjust pipe, cwnd etc */
193 hc->tx_ssthresh = hc->tx_cwnd / 2; 135 hc->tx_ssthresh = hc->tx_cwnd / 2;
194 if (hc->tx_ssthresh < 2) 136 if (hc->tx_ssthresh < 2)
195 hc->tx_ssthresh = 2; 137 hc->tx_ssthresh = 2;
196 hc->tx_cwnd = 1; 138 hc->tx_cwnd = 1;
197 hc->tx_pipe = 0; 139 hc->tx_pipe = 0;
198 140
199 /* clear state about stuff we sent */ 141 /* clear state about stuff we sent */
200 hc->tx_seqt = hc->tx_seqh; 142 hc->tx_seqt = hc->tx_seqh;
@@ -204,23 +146,12 @@ static void ccid2_hc_tx_rto_expire(unsigned long data)
204 hc->tx_rpseq = 0; 146 hc->tx_rpseq = 0;
205 hc->tx_rpdupack = -1; 147 hc->tx_rpdupack = -1;
206 ccid2_change_l_ack_ratio(sk, 1); 148 ccid2_change_l_ack_ratio(sk, 1);
207 ccid2_hc_tx_check_sanity(hc);
208out: 149out:
209 bh_unlock_sock(sk); 150 bh_unlock_sock(sk);
210 sock_put(sk); 151 sock_put(sk);
211} 152}
212 153
213static void ccid2_start_rto_timer(struct sock *sk) 154static void ccid2_hc_tx_packet_sent(struct sock *sk, unsigned int len)
214{
215 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
216
217 ccid2_pr_debug("setting RTO timeout=%ld\n", hc->tx_rto);
218
219 BUG_ON(timer_pending(&hc->tx_rtotimer));
220 sk_reset_timer(sk, &hc->tx_rtotimer, jiffies + hc->tx_rto);
221}
222
223static void ccid2_hc_tx_packet_sent(struct sock *sk, int more, unsigned int len)
224{ 155{
225 struct dccp_sock *dp = dccp_sk(sk); 156 struct dccp_sock *dp = dccp_sk(sk);
226 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk); 157 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
@@ -230,7 +161,7 @@ static void ccid2_hc_tx_packet_sent(struct sock *sk, int more, unsigned int len)
230 161
231 hc->tx_seqh->ccid2s_seq = dp->dccps_gss; 162 hc->tx_seqh->ccid2s_seq = dp->dccps_gss;
232 hc->tx_seqh->ccid2s_acked = 0; 163 hc->tx_seqh->ccid2s_acked = 0;
233 hc->tx_seqh->ccid2s_sent = jiffies; 164 hc->tx_seqh->ccid2s_sent = ccid2_time_stamp;
234 165
235 next = hc->tx_seqh->ccid2s_next; 166 next = hc->tx_seqh->ccid2s_next;
236 /* check if we need to alloc more space */ 167 /* check if we need to alloc more space */
@@ -296,23 +227,20 @@ static void ccid2_hc_tx_packet_sent(struct sock *sk, int more, unsigned int len)
296 } 227 }
297#endif 228#endif
298 229
299 /* setup RTO timer */ 230 sk_reset_timer(sk, &hc->tx_rtotimer, jiffies + hc->tx_rto);
300 if (!timer_pending(&hc->tx_rtotimer))
301 ccid2_start_rto_timer(sk);
302 231
303#ifdef CONFIG_IP_DCCP_CCID2_DEBUG 232#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
304 do { 233 do {
305 struct ccid2_seq *seqp = hc->tx_seqt; 234 struct ccid2_seq *seqp = hc->tx_seqt;
306 235
307 while (seqp != hc->tx_seqh) { 236 while (seqp != hc->tx_seqh) {
308 ccid2_pr_debug("out seq=%llu acked=%d time=%lu\n", 237 ccid2_pr_debug("out seq=%llu acked=%d time=%u\n",
309 (unsigned long long)seqp->ccid2s_seq, 238 (unsigned long long)seqp->ccid2s_seq,
310 seqp->ccid2s_acked, seqp->ccid2s_sent); 239 seqp->ccid2s_acked, seqp->ccid2s_sent);
311 seqp = seqp->ccid2s_next; 240 seqp = seqp->ccid2s_next;
312 } 241 }
313 } while (0); 242 } while (0);
314 ccid2_pr_debug("=========\n"); 243 ccid2_pr_debug("=========\n");
315 ccid2_hc_tx_check_sanity(hc);
316#endif 244#endif
317} 245}
318 246
@@ -378,17 +306,87 @@ out_invalid_option:
378 return -1; 306 return -1;
379} 307}
380 308
381static void ccid2_hc_tx_kill_rto_timer(struct sock *sk) 309/**
310 * ccid2_rtt_estimator - Sample RTT and compute RTO using RFC2988 algorithm
311 * This code is almost identical with TCP's tcp_rtt_estimator(), since
312 * - it has a higher sampling frequency (recommended by RFC 1323),
313 * - the RTO does not collapse into RTT due to RTTVAR going towards zero,
314 * - it is simple (cf. more complex proposals such as Eifel timer or research
315 * which suggests that the gain should be set according to window size),
316 * - in tests it was found to work well with CCID2 [gerrit].
317 */
318static void ccid2_rtt_estimator(struct sock *sk, const long mrtt)
382{ 319{
383 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk); 320 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
321 long m = mrtt ? : 1;
384 322
385 sk_stop_timer(sk, &hc->tx_rtotimer); 323 if (hc->tx_srtt == 0) {
386 ccid2_pr_debug("deleted RTO timer\n"); 324 /* First measurement m */
325 hc->tx_srtt = m << 3;
326 hc->tx_mdev = m << 1;
327
328 hc->tx_mdev_max = max(hc->tx_mdev, tcp_rto_min(sk));
329 hc->tx_rttvar = hc->tx_mdev_max;
330
331 hc->tx_rtt_seq = dccp_sk(sk)->dccps_gss;
332 } else {
333 /* Update scaled SRTT as SRTT += 1/8 * (m - SRTT) */
334 m -= (hc->tx_srtt >> 3);
335 hc->tx_srtt += m;
336
337 /* Similarly, update scaled mdev with regard to |m| */
338 if (m < 0) {
339 m = -m;
340 m -= (hc->tx_mdev >> 2);
341 /*
342 * This neutralises RTO increase when RTT < SRTT - mdev
343 * (see P. Sarolahti, A. Kuznetsov,"Congestion Control
344 * in Linux TCP", USENIX 2002, pp. 49-62).
345 */
346 if (m > 0)
347 m >>= 3;
348 } else {
349 m -= (hc->tx_mdev >> 2);
350 }
351 hc->tx_mdev += m;
352
353 if (hc->tx_mdev > hc->tx_mdev_max) {
354 hc->tx_mdev_max = hc->tx_mdev;
355 if (hc->tx_mdev_max > hc->tx_rttvar)
356 hc->tx_rttvar = hc->tx_mdev_max;
357 }
358
359 /*
360 * Decay RTTVAR at most once per flight, exploiting that
361 * 1) pipe <= cwnd <= Sequence_Window = W (RFC 4340, 7.5.2)
362 * 2) AWL = GSS-W+1 <= GAR <= GSS (RFC 4340, 7.5.1)
363 * GAR is a useful bound for FlightSize = pipe.
364 * AWL is probably too low here, as it over-estimates pipe.
365 */
366 if (after48(dccp_sk(sk)->dccps_gar, hc->tx_rtt_seq)) {
367 if (hc->tx_mdev_max < hc->tx_rttvar)
368 hc->tx_rttvar -= (hc->tx_rttvar -
369 hc->tx_mdev_max) >> 2;
370 hc->tx_rtt_seq = dccp_sk(sk)->dccps_gss;
371 hc->tx_mdev_max = tcp_rto_min(sk);
372 }
373 }
374
375 /*
376 * Set RTO from SRTT and RTTVAR
377 * As in TCP, 4 * RTTVAR >= TCP_RTO_MIN, giving a minimum RTO of 200 ms.
378 * This agrees with RFC 4341, 5:
379 * "Because DCCP does not retransmit data, DCCP does not require
380 * TCP's recommended minimum timeout of one second".
381 */
382 hc->tx_rto = (hc->tx_srtt >> 3) + hc->tx_rttvar;
383
384 if (hc->tx_rto > DCCP_RTO_MAX)
385 hc->tx_rto = DCCP_RTO_MAX;
387} 386}
388 387
389static inline void ccid2_new_ack(struct sock *sk, 388static void ccid2_new_ack(struct sock *sk, struct ccid2_seq *seqp,
390 struct ccid2_seq *seqp, 389 unsigned int *maxincr)
391 unsigned int *maxincr)
392{ 390{
393 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk); 391 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
394 392
@@ -402,93 +400,27 @@ static inline void ccid2_new_ack(struct sock *sk,
402 hc->tx_cwnd += 1; 400 hc->tx_cwnd += 1;
403 hc->tx_packets_acked = 0; 401 hc->tx_packets_acked = 0;
404 } 402 }
405 403 /*
406 /* update RTO */ 404 * FIXME: RTT is sampled several times per acknowledgment (for each
407 if (hc->tx_srtt == -1 || 405 * entry in the Ack Vector), instead of once per Ack (as in TCP SACK).
408 time_after(jiffies, hc->tx_lastrtt + hc->tx_srtt)) { 406 * This causes the RTT to be over-estimated, since the older entries
409 unsigned long r = (long)jiffies - (long)seqp->ccid2s_sent; 407 * in the Ack Vector have earlier sending times.
410 int s; 408 * The cleanest solution is to not use the ccid2s_sent field at all
411 409 * and instead use DCCP timestamps: requires changes in other places.
412 /* first measurement */ 410 */
413 if (hc->tx_srtt == -1) { 411 ccid2_rtt_estimator(sk, ccid2_time_stamp - seqp->ccid2s_sent);
414 ccid2_pr_debug("R: %lu Time=%lu seq=%llu\n",
415 r, jiffies,
416 (unsigned long long)seqp->ccid2s_seq);
417 ccid2_change_srtt(hc, r);
418 hc->tx_rttvar = r >> 1;
419 } else {
420 /* RTTVAR */
421 long tmp = hc->tx_srtt - r;
422 long srtt;
423
424 if (tmp < 0)
425 tmp *= -1;
426
427 tmp >>= 2;
428 hc->tx_rttvar *= 3;
429 hc->tx_rttvar >>= 2;
430 hc->tx_rttvar += tmp;
431
432 /* SRTT */
433 srtt = hc->tx_srtt;
434 srtt *= 7;
435 srtt >>= 3;
436 tmp = r >> 3;
437 srtt += tmp;
438 ccid2_change_srtt(hc, srtt);
439 }
440 s = hc->tx_rttvar << 2;
441 /* clock granularity is 1 when based on jiffies */
442 if (!s)
443 s = 1;
444 hc->tx_rto = hc->tx_srtt + s;
445
446 /* must be at least a second */
447 s = hc->tx_rto / HZ;
448 /* DCCP doesn't require this [but I like it cuz my code sux] */
449#if 1
450 if (s < 1)
451 hc->tx_rto = HZ;
452#endif
453 /* max 60 seconds */
454 if (s > 60)
455 hc->tx_rto = HZ * 60;
456
457 hc->tx_lastrtt = jiffies;
458
459 ccid2_pr_debug("srtt: %ld rttvar: %ld rto: %ld (HZ=%d) R=%lu\n",
460 hc->tx_srtt, hc->tx_rttvar,
461 hc->tx_rto, HZ, r);
462 }
463
464 /* we got a new ack, so re-start RTO timer */
465 ccid2_hc_tx_kill_rto_timer(sk);
466 ccid2_start_rto_timer(sk);
467}
468
469static void ccid2_hc_tx_dec_pipe(struct sock *sk)
470{
471 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
472
473 if (hc->tx_pipe == 0)
474 DCCP_BUG("pipe == 0");
475 else
476 hc->tx_pipe--;
477
478 if (hc->tx_pipe == 0)
479 ccid2_hc_tx_kill_rto_timer(sk);
480} 412}
481 413
482static void ccid2_congestion_event(struct sock *sk, struct ccid2_seq *seqp) 414static void ccid2_congestion_event(struct sock *sk, struct ccid2_seq *seqp)
483{ 415{
484 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk); 416 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
485 417
486 if (time_before(seqp->ccid2s_sent, hc->tx_last_cong)) { 418 if ((s32)(seqp->ccid2s_sent - hc->tx_last_cong) < 0) {
487 ccid2_pr_debug("Multiple losses in an RTT---treating as one\n"); 419 ccid2_pr_debug("Multiple losses in an RTT---treating as one\n");
488 return; 420 return;
489 } 421 }
490 422
491 hc->tx_last_cong = jiffies; 423 hc->tx_last_cong = ccid2_time_stamp;
492 424
493 hc->tx_cwnd = hc->tx_cwnd / 2 ? : 1U; 425 hc->tx_cwnd = hc->tx_cwnd / 2 ? : 1U;
494 hc->tx_ssthresh = max(hc->tx_cwnd, 2U); 426 hc->tx_ssthresh = max(hc->tx_cwnd, 2U);
@@ -510,7 +442,6 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
510 int done = 0; 442 int done = 0;
511 unsigned int maxincr = 0; 443 unsigned int maxincr = 0;
512 444
513 ccid2_hc_tx_check_sanity(hc);
514 /* check reverse path congestion */ 445 /* check reverse path congestion */
515 seqno = DCCP_SKB_CB(skb)->dccpd_seq; 446 seqno = DCCP_SKB_CB(skb)->dccpd_seq;
516 447
@@ -620,7 +551,7 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
620 seqp->ccid2s_acked = 1; 551 seqp->ccid2s_acked = 1;
621 ccid2_pr_debug("Got ack for %llu\n", 552 ccid2_pr_debug("Got ack for %llu\n",
622 (unsigned long long)seqp->ccid2s_seq); 553 (unsigned long long)seqp->ccid2s_seq);
623 ccid2_hc_tx_dec_pipe(sk); 554 hc->tx_pipe--;
624 } 555 }
625 if (seqp == hc->tx_seqt) { 556 if (seqp == hc->tx_seqt) {
626 done = 1; 557 done = 1;
@@ -677,7 +608,7 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
677 * one ack vector. 608 * one ack vector.
678 */ 609 */
679 ccid2_congestion_event(sk, seqp); 610 ccid2_congestion_event(sk, seqp);
680 ccid2_hc_tx_dec_pipe(sk); 611 hc->tx_pipe--;
681 } 612 }
682 if (seqp == hc->tx_seqt) 613 if (seqp == hc->tx_seqt)
683 break; 614 break;
@@ -695,7 +626,11 @@ static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
695 hc->tx_seqt = hc->tx_seqt->ccid2s_next; 626 hc->tx_seqt = hc->tx_seqt->ccid2s_next;
696 } 627 }
697 628
698 ccid2_hc_tx_check_sanity(hc); 629 /* restart RTO timer if not all outstanding data has been acked */
630 if (hc->tx_pipe == 0)
631 sk_stop_timer(sk, &hc->tx_rtotimer);
632 else
633 sk_reset_timer(sk, &hc->tx_rtotimer, jiffies + hc->tx_rto);
699} 634}
700 635
701static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk) 636static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk)
@@ -707,12 +642,8 @@ static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk)
707 /* RFC 4341, 5: initialise ssthresh to arbitrarily high (max) value */ 642 /* RFC 4341, 5: initialise ssthresh to arbitrarily high (max) value */
708 hc->tx_ssthresh = ~0U; 643 hc->tx_ssthresh = ~0U;
709 644
710 /* 645 /* Use larger initial windows (RFC 4341, section 5). */
711 * RFC 4341, 5: "The cwnd parameter is initialized to at most four 646 hc->tx_cwnd = rfc3390_bytes_to_packets(dp->dccps_mss_cache);
712 * packets for new connections, following the rules from [RFC3390]".
713 * We need to convert the bytes of RFC3390 into the packets of RFC 4341.
714 */
715 hc->tx_cwnd = clamp(4380U / dp->dccps_mss_cache, 2U, 4U);
716 647
717 /* Make sure that Ack Ratio is enabled and within bounds. */ 648 /* Make sure that Ack Ratio is enabled and within bounds. */
718 max_ratio = DIV_ROUND_UP(hc->tx_cwnd, 2); 649 max_ratio = DIV_ROUND_UP(hc->tx_cwnd, 2);
@@ -723,15 +654,11 @@ static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk)
723 if (ccid2_hc_tx_alloc_seq(hc)) 654 if (ccid2_hc_tx_alloc_seq(hc))
724 return -ENOMEM; 655 return -ENOMEM;
725 656
726 hc->tx_rto = 3 * HZ; 657 hc->tx_rto = DCCP_TIMEOUT_INIT;
727 ccid2_change_srtt(hc, -1);
728 hc->tx_rttvar = -1;
729 hc->tx_rpdupack = -1; 658 hc->tx_rpdupack = -1;
730 hc->tx_last_cong = jiffies; 659 hc->tx_last_cong = ccid2_time_stamp;
731 setup_timer(&hc->tx_rtotimer, ccid2_hc_tx_rto_expire, 660 setup_timer(&hc->tx_rtotimer, ccid2_hc_tx_rto_expire,
732 (unsigned long)sk); 661 (unsigned long)sk);
733
734 ccid2_hc_tx_check_sanity(hc);
735 return 0; 662 return 0;
736} 663}
737 664
@@ -740,7 +667,7 @@ static void ccid2_hc_tx_exit(struct sock *sk)
740 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk); 667 struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
741 int i; 668 int i;
742 669
743 ccid2_hc_tx_kill_rto_timer(sk); 670 sk_stop_timer(sk, &hc->tx_rtotimer);
744 671
745 for (i = 0; i < hc->tx_seqbufc; i++) 672 for (i = 0; i < hc->tx_seqbufc; i++)
746 kfree(hc->tx_seqbuf[i]); 673 kfree(hc->tx_seqbuf[i]);
diff --git a/net/dccp/ccids/ccid2.h b/net/dccp/ccids/ccid2.h
index 1ec6a30103bb..9731c2dc1487 100644
--- a/net/dccp/ccids/ccid2.h
+++ b/net/dccp/ccids/ccid2.h
@@ -18,18 +18,23 @@
18#ifndef _DCCP_CCID2_H_ 18#ifndef _DCCP_CCID2_H_
19#define _DCCP_CCID2_H_ 19#define _DCCP_CCID2_H_
20 20
21#include <linux/dccp.h>
22#include <linux/timer.h> 21#include <linux/timer.h>
23#include <linux/types.h> 22#include <linux/types.h>
24#include "../ccid.h" 23#include "../ccid.h"
24#include "../dccp.h"
25
26/*
27 * CCID-2 timestamping faces the same issues as TCP timestamping.
28 * Hence we reuse/share as much of the code as possible.
29 */
30#define ccid2_time_stamp tcp_time_stamp
31
25/* NUMDUPACK parameter from RFC 4341, p. 6 */ 32/* NUMDUPACK parameter from RFC 4341, p. 6 */
26#define NUMDUPACK 3 33#define NUMDUPACK 3
27 34
28struct sock;
29
30struct ccid2_seq { 35struct ccid2_seq {
31 u64 ccid2s_seq; 36 u64 ccid2s_seq;
32 unsigned long ccid2s_sent; 37 u32 ccid2s_sent;
33 int ccid2s_acked; 38 int ccid2s_acked;
34 struct ccid2_seq *ccid2s_prev; 39 struct ccid2_seq *ccid2s_prev;
35 struct ccid2_seq *ccid2s_next; 40 struct ccid2_seq *ccid2s_next;
@@ -42,7 +47,12 @@ struct ccid2_seq {
42 * struct ccid2_hc_tx_sock - CCID2 TX half connection 47 * struct ccid2_hc_tx_sock - CCID2 TX half connection
43 * @tx_{cwnd,ssthresh,pipe}: as per RFC 4341, section 5 48 * @tx_{cwnd,ssthresh,pipe}: as per RFC 4341, section 5
44 * @tx_packets_acked: Ack counter for deriving cwnd growth (RFC 3465) 49 * @tx_packets_acked: Ack counter for deriving cwnd growth (RFC 3465)
45 * @tx_lastrtt: time RTT was last measured 50 * @tx_srtt: smoothed RTT estimate, scaled by 2^3
51 * @tx_mdev: smoothed RTT variation, scaled by 2^2
52 * @tx_mdev_max: maximum of @mdev during one flight
53 * @tx_rttvar: moving average/maximum of @mdev_max
54 * @tx_rto: RTO value deriving from SRTT and RTTVAR (RFC 2988)
55 * @tx_rtt_seq: to decay RTTVAR at most once per flight
46 * @tx_rpseq: last consecutive seqno 56 * @tx_rpseq: last consecutive seqno
47 * @tx_rpdupack: dupacks since rpseq 57 * @tx_rpdupack: dupacks since rpseq
48 */ 58 */
@@ -55,14 +65,19 @@ struct ccid2_hc_tx_sock {
55 int tx_seqbufc; 65 int tx_seqbufc;
56 struct ccid2_seq *tx_seqh; 66 struct ccid2_seq *tx_seqh;
57 struct ccid2_seq *tx_seqt; 67 struct ccid2_seq *tx_seqt;
58 long tx_rto; 68
59 long tx_srtt; 69 /* RTT measurement: variables/principles are the same as in TCP */
60 long tx_rttvar; 70 u32 tx_srtt,
61 unsigned long tx_lastrtt; 71 tx_mdev,
72 tx_mdev_max,
73 tx_rttvar,
74 tx_rto;
75 u64 tx_rtt_seq:48;
62 struct timer_list tx_rtotimer; 76 struct timer_list tx_rtotimer;
77
63 u64 tx_rpseq; 78 u64 tx_rpseq;
64 int tx_rpdupack; 79 int tx_rpdupack;
65 unsigned long tx_last_cong; 80 u32 tx_last_cong;
66 u64 tx_high_ack; 81 u64 tx_high_ack;
67}; 82};
68 83
diff --git a/net/dccp/ccids/ccid3.c b/net/dccp/ccids/ccid3.c
index 95f752986497..3060a60ed5ab 100644
--- a/net/dccp/ccids/ccid3.c
+++ b/net/dccp/ccids/ccid3.c
@@ -54,7 +54,6 @@ static const char *ccid3_tx_state_name(enum ccid3_hc_tx_states state)
54 [TFRC_SSTATE_NO_SENT] = "NO_SENT", 54 [TFRC_SSTATE_NO_SENT] = "NO_SENT",
55 [TFRC_SSTATE_NO_FBACK] = "NO_FBACK", 55 [TFRC_SSTATE_NO_FBACK] = "NO_FBACK",
56 [TFRC_SSTATE_FBACK] = "FBACK", 56 [TFRC_SSTATE_FBACK] = "FBACK",
57 [TFRC_SSTATE_TERM] = "TERM",
58 }; 57 };
59 58
60 return ccid3_state_names[state]; 59 return ccid3_state_names[state];
@@ -91,19 +90,16 @@ static inline u64 rfc3390_initial_rate(struct sock *sk)
91 return scaled_div(w_init << 6, hc->tx_rtt); 90 return scaled_div(w_init << 6, hc->tx_rtt);
92} 91}
93 92
94/* 93/**
95 * Recalculate t_ipi and delta (should be called whenever X changes) 94 * ccid3_update_send_interval - Calculate new t_ipi = s / X_inst
95 * This respects the granularity of X_inst (64 * bytes/second).
96 */ 96 */
97static void ccid3_update_send_interval(struct ccid3_hc_tx_sock *hc) 97static void ccid3_update_send_interval(struct ccid3_hc_tx_sock *hc)
98{ 98{
99 /* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */
100 hc->tx_t_ipi = scaled_div32(((u64)hc->tx_s) << 6, hc->tx_x); 99 hc->tx_t_ipi = scaled_div32(((u64)hc->tx_s) << 6, hc->tx_x);
101 100
102 /* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */ 101 ccid3_pr_debug("t_ipi=%u, s=%u, X=%u\n", hc->tx_t_ipi,
103 hc->tx_delta = min_t(u32, hc->tx_t_ipi / 2, TFRC_OPSYS_HALF_TIME_GRAN); 102 hc->tx_s, (unsigned)(hc->tx_x >> 6));
104
105 ccid3_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n", hc->tx_t_ipi,
106 hc->tx_delta, hc->tx_s, (unsigned)(hc->tx_x >> 6));
107} 103}
108 104
109static u32 ccid3_hc_tx_idle_rtt(struct ccid3_hc_tx_sock *hc, ktime_t now) 105static u32 ccid3_hc_tx_idle_rtt(struct ccid3_hc_tx_sock *hc, ktime_t now)
@@ -211,16 +207,19 @@ static void ccid3_hc_tx_no_feedback_timer(unsigned long data)
211 ccid3_pr_debug("%s(%p, state=%s) - entry\n", dccp_role(sk), sk, 207 ccid3_pr_debug("%s(%p, state=%s) - entry\n", dccp_role(sk), sk,
212 ccid3_tx_state_name(hc->tx_state)); 208 ccid3_tx_state_name(hc->tx_state));
213 209
210 /* Ignore and do not restart after leaving the established state */
211 if ((1 << sk->sk_state) & ~(DCCPF_OPEN | DCCPF_PARTOPEN))
212 goto out;
213
214 /* Reset feedback state to "no feedback received" */
214 if (hc->tx_state == TFRC_SSTATE_FBACK) 215 if (hc->tx_state == TFRC_SSTATE_FBACK)
215 ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK); 216 ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK);
216 else if (hc->tx_state != TFRC_SSTATE_NO_FBACK)
217 goto out;
218 217
219 /* 218 /*
220 * Determine new allowed sending rate X as per draft rfc3448bis-00, 4.4 219 * Determine new allowed sending rate X as per draft rfc3448bis-00, 4.4
220 * RTO is 0 if and only if no feedback has been received yet.
221 */ 221 */
222 if (hc->tx_t_rto == 0 || /* no feedback received yet */ 222 if (hc->tx_t_rto == 0 || hc->tx_p == 0) {
223 hc->tx_p == 0) {
224 223
225 /* halve send rate directly */ 224 /* halve send rate directly */
226 hc->tx_x = max(hc->tx_x / 2, 225 hc->tx_x = max(hc->tx_x / 2,
@@ -256,7 +255,7 @@ static void ccid3_hc_tx_no_feedback_timer(unsigned long data)
256 * Set new timeout for the nofeedback timer. 255 * Set new timeout for the nofeedback timer.
257 * See comments in packet_recv() regarding the value of t_RTO. 256 * See comments in packet_recv() regarding the value of t_RTO.
258 */ 257 */
259 if (unlikely(hc->tx_t_rto == 0)) /* no feedback yet */ 258 if (unlikely(hc->tx_t_rto == 0)) /* no feedback received yet */
260 t_nfb = TFRC_INITIAL_TIMEOUT; 259 t_nfb = TFRC_INITIAL_TIMEOUT;
261 else 260 else
262 t_nfb = max(hc->tx_t_rto, 2 * hc->tx_t_ipi); 261 t_nfb = max(hc->tx_t_rto, 2 * hc->tx_t_ipi);
@@ -290,8 +289,7 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
290 if (unlikely(skb->len == 0)) 289 if (unlikely(skb->len == 0))
291 return -EBADMSG; 290 return -EBADMSG;
292 291
293 switch (hc->tx_state) { 292 if (hc->tx_state == TFRC_SSTATE_NO_SENT) {
294 case TFRC_SSTATE_NO_SENT:
295 sk_reset_timer(sk, &hc->tx_no_feedback_timer, (jiffies + 293 sk_reset_timer(sk, &hc->tx_no_feedback_timer, (jiffies +
296 usecs_to_jiffies(TFRC_INITIAL_TIMEOUT))); 294 usecs_to_jiffies(TFRC_INITIAL_TIMEOUT)));
297 hc->tx_last_win_count = 0; 295 hc->tx_last_win_count = 0;
@@ -326,27 +324,22 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
326 ccid3_update_send_interval(hc); 324 ccid3_update_send_interval(hc);
327 325
328 ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK); 326 ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK);
329 break; 327
330 case TFRC_SSTATE_NO_FBACK: 328 } else {
331 case TFRC_SSTATE_FBACK:
332 delay = ktime_us_delta(hc->tx_t_nom, now); 329 delay = ktime_us_delta(hc->tx_t_nom, now);
333 ccid3_pr_debug("delay=%ld\n", (long)delay); 330 ccid3_pr_debug("delay=%ld\n", (long)delay);
334 /* 331 /*
335 * Scheduling of packet transmissions [RFC 3448, 4.6] 332 * Scheduling of packet transmissions (RFC 5348, 8.3)
336 * 333 *
337 * if (t_now > t_nom - delta) 334 * if (t_now > t_nom - delta)
338 * // send the packet now 335 * // send the packet now
339 * else 336 * else
340 * // send the packet in (t_nom - t_now) milliseconds. 337 * // send the packet in (t_nom - t_now) milliseconds.
341 */ 338 */
342 if (delay - (s64)hc->tx_delta >= 1000) 339 if (delay >= TFRC_T_DELTA)
343 return (u32)delay / 1000L; 340 return (u32)delay / USEC_PER_MSEC;
344 341
345 ccid3_hc_tx_update_win_count(hc, now); 342 ccid3_hc_tx_update_win_count(hc, now);
346 break;
347 case TFRC_SSTATE_TERM:
348 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
349 return -EINVAL;
350 } 343 }
351 344
352 /* prepare to send now (add options etc.) */ 345 /* prepare to send now (add options etc.) */
@@ -358,8 +351,7 @@ static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
358 return 0; 351 return 0;
359} 352}
360 353
361static void ccid3_hc_tx_packet_sent(struct sock *sk, int more, 354static void ccid3_hc_tx_packet_sent(struct sock *sk, unsigned int len)
362 unsigned int len)
363{ 355{
364 struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk); 356 struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk);
365 357
@@ -372,48 +364,34 @@ static void ccid3_hc_tx_packet_sent(struct sock *sk, int more,
372static void ccid3_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb) 364static void ccid3_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
373{ 365{
374 struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk); 366 struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk);
375 struct ccid3_options_received *opt_recv; 367 struct tfrc_tx_hist_entry *acked;
376 ktime_t now; 368 ktime_t now;
377 unsigned long t_nfb; 369 unsigned long t_nfb;
378 u32 pinv, r_sample; 370 u32 r_sample;
379 371
380 /* we are only interested in ACKs */ 372 /* we are only interested in ACKs */
381 if (!(DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK || 373 if (!(DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK ||
382 DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_DATAACK)) 374 DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_DATAACK))
383 return; 375 return;
384 /* ... and only in the established state */
385 if (hc->tx_state != TFRC_SSTATE_FBACK &&
386 hc->tx_state != TFRC_SSTATE_NO_FBACK)
387 return;
388
389 opt_recv = &hc->tx_options_received;
390 now = ktime_get_real();
391
392 /* Estimate RTT from history if ACK number is valid */
393 r_sample = tfrc_tx_hist_rtt(hc->tx_hist,
394 DCCP_SKB_CB(skb)->dccpd_ack_seq, now);
395 if (r_sample == 0) {
396 DCCP_WARN("%s(%p): %s with bogus ACK-%llu\n", dccp_role(sk), sk,
397 dccp_packet_name(DCCP_SKB_CB(skb)->dccpd_type),
398 (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq);
399 return;
400 }
401
402 /* Update receive rate in units of 64 * bytes/second */
403 hc->tx_x_recv = opt_recv->ccid3or_receive_rate;
404 hc->tx_x_recv <<= 6;
405
406 /* Update loss event rate (which is scaled by 1e6) */
407 pinv = opt_recv->ccid3or_loss_event_rate;
408 if (pinv == ~0U || pinv == 0) /* see RFC 4342, 8.5 */
409 hc->tx_p = 0;
410 else /* can not exceed 100% */
411 hc->tx_p = scaled_div(1, pinv);
412 /* 376 /*
413 * Validate new RTT sample and update moving average 377 * Locate the acknowledged packet in the TX history.
378 *
379 * Returning "entry not found" here can for instance happen when
380 * - the host has not sent out anything (e.g. a passive server),
381 * - the Ack is outdated (packet with higher Ack number was received),
382 * - it is a bogus Ack (for a packet not sent on this connection).
414 */ 383 */
415 r_sample = dccp_sample_rtt(sk, r_sample); 384 acked = tfrc_tx_hist_find_entry(hc->tx_hist, dccp_hdr_ack_seq(skb));
385 if (acked == NULL)
386 return;
387 /* For the sake of RTT sampling, ignore/remove all older entries */
388 tfrc_tx_hist_purge(&acked->next);
389
390 /* Update the moving average for the RTT estimate (RFC 3448, 4.3) */
391 now = ktime_get_real();
392 r_sample = dccp_sample_rtt(sk, ktime_us_delta(now, acked->stamp));
416 hc->tx_rtt = tfrc_ewma(hc->tx_rtt, r_sample, 9); 393 hc->tx_rtt = tfrc_ewma(hc->tx_rtt, r_sample, 9);
394
417 /* 395 /*
418 * Update allowed sending rate X as per draft rfc3448bis-00, 4.2/3 396 * Update allowed sending rate X as per draft rfc3448bis-00, 4.2/3
419 */ 397 */
@@ -461,13 +439,12 @@ done_computing_x:
461 sk->sk_write_space(sk); 439 sk->sk_write_space(sk);
462 440
463 /* 441 /*
464 * Update timeout interval for the nofeedback timer. 442 * Update timeout interval for the nofeedback timer. In order to control
465 * We use a configuration option to increase the lower bound. 443 * rate halving on networks with very low RTTs (<= 1 ms), use per-route
466 * This can help avoid triggering the nofeedback timer too 444 * tunable RTAX_RTO_MIN value as the lower bound.
467 * often ('spinning') on LANs with small RTTs.
468 */ 445 */
469 hc->tx_t_rto = max_t(u32, 4 * hc->tx_rtt, (CONFIG_IP_DCCP_CCID3_RTO * 446 hc->tx_t_rto = max_t(u32, 4 * hc->tx_rtt,
470 (USEC_PER_SEC / 1000))); 447 USEC_PER_SEC/HZ * tcp_rto_min(sk));
471 /* 448 /*
472 * Schedule no feedback timer to expire in 449 * Schedule no feedback timer to expire in
473 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi) 450 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
@@ -482,66 +459,41 @@ done_computing_x:
482 jiffies + usecs_to_jiffies(t_nfb)); 459 jiffies + usecs_to_jiffies(t_nfb));
483} 460}
484 461
485static int ccid3_hc_tx_parse_options(struct sock *sk, unsigned char option, 462static int ccid3_hc_tx_parse_options(struct sock *sk, u8 packet_type,
486 unsigned char len, u16 idx, 463 u8 option, u8 *optval, u8 optlen)
487 unsigned char *value)
488{ 464{
489 int rc = 0;
490 const struct dccp_sock *dp = dccp_sk(sk);
491 struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk); 465 struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk);
492 struct ccid3_options_received *opt_recv;
493 __be32 opt_val; 466 __be32 opt_val;
494 467
495 opt_recv = &hc->tx_options_received;
496
497 if (opt_recv->ccid3or_seqno != dp->dccps_gsr) {
498 opt_recv->ccid3or_seqno = dp->dccps_gsr;
499 opt_recv->ccid3or_loss_event_rate = ~0;
500 opt_recv->ccid3or_loss_intervals_idx = 0;
501 opt_recv->ccid3or_loss_intervals_len = 0;
502 opt_recv->ccid3or_receive_rate = 0;
503 }
504
505 switch (option) { 468 switch (option) {
469 case TFRC_OPT_RECEIVE_RATE:
506 case TFRC_OPT_LOSS_EVENT_RATE: 470 case TFRC_OPT_LOSS_EVENT_RATE:
507 if (unlikely(len != 4)) { 471 /* Must be ignored on Data packets, cf. RFC 4342 8.3 and 8.5 */
508 DCCP_WARN("%s(%p), invalid len %d " 472 if (packet_type == DCCP_PKT_DATA)
509 "for TFRC_OPT_LOSS_EVENT_RATE\n", 473 break;
510 dccp_role(sk), sk, len); 474 if (unlikely(optlen != 4)) {
511 rc = -EINVAL; 475 DCCP_WARN("%s(%p), invalid len %d for %u\n",
512 } else { 476 dccp_role(sk), sk, optlen, option);
513 opt_val = get_unaligned((__be32 *)value); 477 return -EINVAL;
514 opt_recv->ccid3or_loss_event_rate = ntohl(opt_val);
515 ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
516 dccp_role(sk), sk,
517 opt_recv->ccid3or_loss_event_rate);
518 } 478 }
519 break; 479 opt_val = ntohl(get_unaligned((__be32 *)optval));
520 case TFRC_OPT_LOSS_INTERVALS: 480
521 opt_recv->ccid3or_loss_intervals_idx = idx; 481 if (option == TFRC_OPT_RECEIVE_RATE) {
522 opt_recv->ccid3or_loss_intervals_len = len; 482 /* Receive Rate is kept in units of 64 bytes/second */
523 ccid3_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n", 483 hc->tx_x_recv = opt_val;
524 dccp_role(sk), sk, 484 hc->tx_x_recv <<= 6;
525 opt_recv->ccid3or_loss_intervals_idx, 485
526 opt_recv->ccid3or_loss_intervals_len);
527 break;
528 case TFRC_OPT_RECEIVE_RATE:
529 if (unlikely(len != 4)) {
530 DCCP_WARN("%s(%p), invalid len %d "
531 "for TFRC_OPT_RECEIVE_RATE\n",
532 dccp_role(sk), sk, len);
533 rc = -EINVAL;
534 } else {
535 opt_val = get_unaligned((__be32 *)value);
536 opt_recv->ccid3or_receive_rate = ntohl(opt_val);
537 ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n", 486 ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n",
538 dccp_role(sk), sk, 487 dccp_role(sk), sk, opt_val);
539 opt_recv->ccid3or_receive_rate); 488 } else {
489 /* Update the fixpoint Loss Event Rate fraction */
490 hc->tx_p = tfrc_invert_loss_event_rate(opt_val);
491
492 ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
493 dccp_role(sk), sk, opt_val);
540 } 494 }
541 break;
542 } 495 }
543 496 return 0;
544 return rc;
545} 497}
546 498
547static int ccid3_hc_tx_init(struct ccid *ccid, struct sock *sk) 499static int ccid3_hc_tx_init(struct ccid *ccid, struct sock *sk)
@@ -559,42 +511,36 @@ static void ccid3_hc_tx_exit(struct sock *sk)
559{ 511{
560 struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk); 512 struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk);
561 513
562 ccid3_hc_tx_set_state(sk, TFRC_SSTATE_TERM);
563 sk_stop_timer(sk, &hc->tx_no_feedback_timer); 514 sk_stop_timer(sk, &hc->tx_no_feedback_timer);
564
565 tfrc_tx_hist_purge(&hc->tx_hist); 515 tfrc_tx_hist_purge(&hc->tx_hist);
566} 516}
567 517
568static void ccid3_hc_tx_get_info(struct sock *sk, struct tcp_info *info) 518static void ccid3_hc_tx_get_info(struct sock *sk, struct tcp_info *info)
569{ 519{
570 struct ccid3_hc_tx_sock *hc; 520 info->tcpi_rto = ccid3_hc_tx_sk(sk)->tx_t_rto;
571 521 info->tcpi_rtt = ccid3_hc_tx_sk(sk)->tx_rtt;
572 /* Listen socks doesn't have a private CCID block */
573 if (sk->sk_state == DCCP_LISTEN)
574 return;
575
576 hc = ccid3_hc_tx_sk(sk);
577 info->tcpi_rto = hc->tx_t_rto;
578 info->tcpi_rtt = hc->tx_rtt;
579} 522}
580 523
581static int ccid3_hc_tx_getsockopt(struct sock *sk, const int optname, int len, 524static int ccid3_hc_tx_getsockopt(struct sock *sk, const int optname, int len,
582 u32 __user *optval, int __user *optlen) 525 u32 __user *optval, int __user *optlen)
583{ 526{
584 const struct ccid3_hc_tx_sock *hc; 527 const struct ccid3_hc_tx_sock *hc = ccid3_hc_tx_sk(sk);
528 struct tfrc_tx_info tfrc;
585 const void *val; 529 const void *val;
586 530
587 /* Listen socks doesn't have a private CCID block */
588 if (sk->sk_state == DCCP_LISTEN)
589 return -EINVAL;
590
591 hc = ccid3_hc_tx_sk(sk);
592 switch (optname) { 531 switch (optname) {
593 case DCCP_SOCKOPT_CCID_TX_INFO: 532 case DCCP_SOCKOPT_CCID_TX_INFO:
594 if (len < sizeof(hc->tx_tfrc)) 533 if (len < sizeof(tfrc))
595 return -EINVAL; 534 return -EINVAL;
596 len = sizeof(hc->tx_tfrc); 535 tfrc.tfrctx_x = hc->tx_x;
597 val = &hc->tx_tfrc; 536 tfrc.tfrctx_x_recv = hc->tx_x_recv;
537 tfrc.tfrctx_x_calc = hc->tx_x_calc;
538 tfrc.tfrctx_rtt = hc->tx_rtt;
539 tfrc.tfrctx_p = hc->tx_p;
540 tfrc.tfrctx_rto = hc->tx_t_rto;
541 tfrc.tfrctx_ipi = hc->tx_t_ipi;
542 len = sizeof(tfrc);
543 val = &tfrc;
598 break; 544 break;
599 default: 545 default:
600 return -ENOPROTOOPT; 546 return -ENOPROTOOPT;
@@ -624,7 +570,6 @@ static const char *ccid3_rx_state_name(enum ccid3_hc_rx_states state)
624 static const char *const ccid3_rx_state_names[] = { 570 static const char *const ccid3_rx_state_names[] = {
625 [TFRC_RSTATE_NO_DATA] = "NO_DATA", 571 [TFRC_RSTATE_NO_DATA] = "NO_DATA",
626 [TFRC_RSTATE_DATA] = "DATA", 572 [TFRC_RSTATE_DATA] = "DATA",
627 [TFRC_RSTATE_TERM] = "TERM",
628 }; 573 };
629 574
630 return ccid3_rx_state_names[state]; 575 return ccid3_rx_state_names[state];
@@ -650,14 +595,9 @@ static void ccid3_hc_rx_send_feedback(struct sock *sk,
650{ 595{
651 struct ccid3_hc_rx_sock *hc = ccid3_hc_rx_sk(sk); 596 struct ccid3_hc_rx_sock *hc = ccid3_hc_rx_sk(sk);
652 struct dccp_sock *dp = dccp_sk(sk); 597 struct dccp_sock *dp = dccp_sk(sk);
653 ktime_t now; 598 ktime_t now = ktime_get_real();
654 s64 delta = 0; 599 s64 delta = 0;
655 600
656 if (unlikely(hc->rx_state == TFRC_RSTATE_TERM))
657 return;
658
659 now = ktime_get_real();
660
661 switch (fbtype) { 601 switch (fbtype) {
662 case CCID3_FBACK_INITIAL: 602 case CCID3_FBACK_INITIAL:
663 hc->rx_x_recv = 0; 603 hc->rx_x_recv = 0;
@@ -701,14 +641,12 @@ static void ccid3_hc_rx_send_feedback(struct sock *sk,
701 641
702static int ccid3_hc_rx_insert_options(struct sock *sk, struct sk_buff *skb) 642static int ccid3_hc_rx_insert_options(struct sock *sk, struct sk_buff *skb)
703{ 643{
704 const struct ccid3_hc_rx_sock *hc; 644 const struct ccid3_hc_rx_sock *hc = ccid3_hc_rx_sk(sk);
705 __be32 x_recv, pinv; 645 __be32 x_recv, pinv;
706 646
707 if (!(sk->sk_state == DCCP_OPEN || sk->sk_state == DCCP_PARTOPEN)) 647 if (!(sk->sk_state == DCCP_OPEN || sk->sk_state == DCCP_PARTOPEN))
708 return 0; 648 return 0;
709 649
710 hc = ccid3_hc_rx_sk(sk);
711
712 if (dccp_packet_without_ack(skb)) 650 if (dccp_packet_without_ack(skb))
713 return 0; 651 return 0;
714 652
@@ -749,10 +687,11 @@ static u32 ccid3_first_li(struct sock *sk)
749 x_recv = scaled_div32(hc->rx_bytes_recv, delta); 687 x_recv = scaled_div32(hc->rx_bytes_recv, delta);
750 if (x_recv == 0) { /* would also trigger divide-by-zero */ 688 if (x_recv == 0) { /* would also trigger divide-by-zero */
751 DCCP_WARN("X_recv==0\n"); 689 DCCP_WARN("X_recv==0\n");
752 if ((x_recv = hc->rx_x_recv) == 0) { 690 if (hc->rx_x_recv == 0) {
753 DCCP_BUG("stored value of X_recv is zero"); 691 DCCP_BUG("stored value of X_recv is zero");
754 return ~0U; 692 return ~0U;
755 } 693 }
694 x_recv = hc->rx_x_recv;
756 } 695 }
757 696
758 fval = scaled_div(hc->rx_s, hc->rx_rtt); 697 fval = scaled_div(hc->rx_s, hc->rx_rtt);
@@ -862,46 +801,31 @@ static void ccid3_hc_rx_exit(struct sock *sk)
862{ 801{
863 struct ccid3_hc_rx_sock *hc = ccid3_hc_rx_sk(sk); 802 struct ccid3_hc_rx_sock *hc = ccid3_hc_rx_sk(sk);
864 803
865 ccid3_hc_rx_set_state(sk, TFRC_RSTATE_TERM);
866
867 tfrc_rx_hist_purge(&hc->rx_hist); 804 tfrc_rx_hist_purge(&hc->rx_hist);
868 tfrc_lh_cleanup(&hc->rx_li_hist); 805 tfrc_lh_cleanup(&hc->rx_li_hist);
869} 806}
870 807
871static void ccid3_hc_rx_get_info(struct sock *sk, struct tcp_info *info) 808static void ccid3_hc_rx_get_info(struct sock *sk, struct tcp_info *info)
872{ 809{
873 const struct ccid3_hc_rx_sock *hc; 810 info->tcpi_ca_state = ccid3_hc_rx_sk(sk)->rx_state;
874
875 /* Listen socks doesn't have a private CCID block */
876 if (sk->sk_state == DCCP_LISTEN)
877 return;
878
879 hc = ccid3_hc_rx_sk(sk);
880 info->tcpi_ca_state = hc->rx_state;
881 info->tcpi_options |= TCPI_OPT_TIMESTAMPS; 811 info->tcpi_options |= TCPI_OPT_TIMESTAMPS;
882 info->tcpi_rcv_rtt = hc->rx_rtt; 812 info->tcpi_rcv_rtt = ccid3_hc_rx_sk(sk)->rx_rtt;
883} 813}
884 814
885static int ccid3_hc_rx_getsockopt(struct sock *sk, const int optname, int len, 815static int ccid3_hc_rx_getsockopt(struct sock *sk, const int optname, int len,
886 u32 __user *optval, int __user *optlen) 816 u32 __user *optval, int __user *optlen)
887{ 817{
888 const struct ccid3_hc_rx_sock *hc; 818 const struct ccid3_hc_rx_sock *hc = ccid3_hc_rx_sk(sk);
889 struct tfrc_rx_info rx_info; 819 struct tfrc_rx_info rx_info;
890 const void *val; 820 const void *val;
891 821
892 /* Listen socks doesn't have a private CCID block */
893 if (sk->sk_state == DCCP_LISTEN)
894 return -EINVAL;
895
896 hc = ccid3_hc_rx_sk(sk);
897 switch (optname) { 822 switch (optname) {
898 case DCCP_SOCKOPT_CCID_RX_INFO: 823 case DCCP_SOCKOPT_CCID_RX_INFO:
899 if (len < sizeof(rx_info)) 824 if (len < sizeof(rx_info))
900 return -EINVAL; 825 return -EINVAL;
901 rx_info.tfrcrx_x_recv = hc->rx_x_recv; 826 rx_info.tfrcrx_x_recv = hc->rx_x_recv;
902 rx_info.tfrcrx_rtt = hc->rx_rtt; 827 rx_info.tfrcrx_rtt = hc->rx_rtt;
903 rx_info.tfrcrx_p = hc->rx_pinv == 0 ? ~0U : 828 rx_info.tfrcrx_p = tfrc_invert_loss_event_rate(hc->rx_pinv);
904 scaled_div(1, hc->rx_pinv);
905 len = sizeof(rx_info); 829 len = sizeof(rx_info);
906 val = &rx_info; 830 val = &rx_info;
907 break; 831 break;
diff --git a/net/dccp/ccids/ccid3.h b/net/dccp/ccids/ccid3.h
index 032635776653..1a9933c29672 100644
--- a/net/dccp/ccids/ccid3.h
+++ b/net/dccp/ccids/ccid3.h
@@ -42,35 +42,36 @@
42#include "lib/tfrc.h" 42#include "lib/tfrc.h"
43#include "../ccid.h" 43#include "../ccid.h"
44 44
45/* Two seconds as per RFC 3448 4.2 */ 45/* Two seconds as per RFC 5348, 4.2 */
46#define TFRC_INITIAL_TIMEOUT (2 * USEC_PER_SEC) 46#define TFRC_INITIAL_TIMEOUT (2 * USEC_PER_SEC)
47 47
48/* In usecs - half the scheduling granularity as per RFC3448 4.6 */
49#define TFRC_OPSYS_HALF_TIME_GRAN (USEC_PER_SEC / (2 * HZ))
50
51/* Parameter t_mbi from [RFC 3448, 4.3]: backoff interval in seconds */ 48/* Parameter t_mbi from [RFC 3448, 4.3]: backoff interval in seconds */
52#define TFRC_T_MBI 64 49#define TFRC_T_MBI 64
53 50
51/*
52 * The t_delta parameter (RFC 5348, 8.3): delays of less than %USEC_PER_MSEC are
53 * rounded down to 0, since sk_reset_timer() here uses millisecond granularity.
54 * Hence we can use a constant t_delta = %USEC_PER_MSEC when HZ >= 500. A coarse
55 * resolution of HZ < 500 means that the error is below one timer tick (t_gran)
56 * when using the constant t_delta = t_gran / 2 = %USEC_PER_SEC / (2 * HZ).
57 */
58#if (HZ >= 500)
59# define TFRC_T_DELTA USEC_PER_MSEC
60#else
61# define TFRC_T_DELTA (USEC_PER_SEC / (2 * HZ))
62#endif
63
54enum ccid3_options { 64enum ccid3_options {
55 TFRC_OPT_LOSS_EVENT_RATE = 192, 65 TFRC_OPT_LOSS_EVENT_RATE = 192,
56 TFRC_OPT_LOSS_INTERVALS = 193, 66 TFRC_OPT_LOSS_INTERVALS = 193,
57 TFRC_OPT_RECEIVE_RATE = 194, 67 TFRC_OPT_RECEIVE_RATE = 194,
58}; 68};
59 69
60struct ccid3_options_received {
61 u64 ccid3or_seqno:48,
62 ccid3or_loss_intervals_idx:16;
63 u16 ccid3or_loss_intervals_len;
64 u32 ccid3or_loss_event_rate;
65 u32 ccid3or_receive_rate;
66};
67
68/* TFRC sender states */ 70/* TFRC sender states */
69enum ccid3_hc_tx_states { 71enum ccid3_hc_tx_states {
70 TFRC_SSTATE_NO_SENT = 1, 72 TFRC_SSTATE_NO_SENT = 1,
71 TFRC_SSTATE_NO_FBACK, 73 TFRC_SSTATE_NO_FBACK,
72 TFRC_SSTATE_FBACK, 74 TFRC_SSTATE_FBACK,
73 TFRC_SSTATE_TERM,
74}; 75};
75 76
76/** 77/**
@@ -90,19 +91,16 @@ enum ccid3_hc_tx_states {
90 * @tx_no_feedback_timer: Handle to no feedback timer 91 * @tx_no_feedback_timer: Handle to no feedback timer
91 * @tx_t_ld: Time last doubled during slow start 92 * @tx_t_ld: Time last doubled during slow start
92 * @tx_t_nom: Nominal send time of next packet 93 * @tx_t_nom: Nominal send time of next packet
93 * @tx_delta: Send timer delta (RFC 3448, 4.6) in usecs
94 * @tx_hist: Packet history 94 * @tx_hist: Packet history
95 * @tx_options_received: Parsed set of retrieved options
96 */ 95 */
97struct ccid3_hc_tx_sock { 96struct ccid3_hc_tx_sock {
98 struct tfrc_tx_info tx_tfrc; 97 u64 tx_x;
99#define tx_x tx_tfrc.tfrctx_x 98 u64 tx_x_recv;
100#define tx_x_recv tx_tfrc.tfrctx_x_recv 99 u32 tx_x_calc;
101#define tx_x_calc tx_tfrc.tfrctx_x_calc 100 u32 tx_rtt;
102#define tx_rtt tx_tfrc.tfrctx_rtt 101 u32 tx_p;
103#define tx_p tx_tfrc.tfrctx_p 102 u32 tx_t_rto;
104#define tx_t_rto tx_tfrc.tfrctx_rto 103 u32 tx_t_ipi;
105#define tx_t_ipi tx_tfrc.tfrctx_ipi
106 u16 tx_s; 104 u16 tx_s;
107 enum ccid3_hc_tx_states tx_state:8; 105 enum ccid3_hc_tx_states tx_state:8;
108 u8 tx_last_win_count; 106 u8 tx_last_win_count;
@@ -110,9 +108,7 @@ struct ccid3_hc_tx_sock {
110 struct timer_list tx_no_feedback_timer; 108 struct timer_list tx_no_feedback_timer;
111 ktime_t tx_t_ld; 109 ktime_t tx_t_ld;
112 ktime_t tx_t_nom; 110 ktime_t tx_t_nom;
113 u32 tx_delta;
114 struct tfrc_tx_hist_entry *tx_hist; 111 struct tfrc_tx_hist_entry *tx_hist;
115 struct ccid3_options_received tx_options_received;
116}; 112};
117 113
118static inline struct ccid3_hc_tx_sock *ccid3_hc_tx_sk(const struct sock *sk) 114static inline struct ccid3_hc_tx_sock *ccid3_hc_tx_sk(const struct sock *sk)
@@ -126,21 +122,16 @@ static inline struct ccid3_hc_tx_sock *ccid3_hc_tx_sk(const struct sock *sk)
126enum ccid3_hc_rx_states { 122enum ccid3_hc_rx_states {
127 TFRC_RSTATE_NO_DATA = 1, 123 TFRC_RSTATE_NO_DATA = 1,
128 TFRC_RSTATE_DATA, 124 TFRC_RSTATE_DATA,
129 TFRC_RSTATE_TERM = 127,
130}; 125};
131 126
132/** 127/**
133 * struct ccid3_hc_rx_sock - CCID3 receiver half-connection socket 128 * struct ccid3_hc_rx_sock - CCID3 receiver half-connection socket
134 * @rx_x_recv: Receiver estimate of send rate (RFC 3448 4.3)
135 * @rx_rtt: Receiver estimate of rtt (non-standard)
136 * @rx_p: Current loss event rate (RFC 3448 5.4)
137 * @rx_last_counter: Tracks window counter (RFC 4342, 8.1) 129 * @rx_last_counter: Tracks window counter (RFC 4342, 8.1)
138 * @rx_state: Receiver state, one of %ccid3_hc_rx_states 130 * @rx_state: Receiver state, one of %ccid3_hc_rx_states
139 * @rx_bytes_recv: Total sum of DCCP payload bytes 131 * @rx_bytes_recv: Total sum of DCCP payload bytes
140 * @rx_x_recv: Receiver estimate of send rate (RFC 3448, sec. 4.3) 132 * @rx_x_recv: Receiver estimate of send rate (RFC 3448, sec. 4.3)
141 * @rx_rtt: Receiver estimate of RTT 133 * @rx_rtt: Receiver estimate of RTT
142 * @rx_tstamp_last_feedback: Time at which last feedback was sent 134 * @rx_tstamp_last_feedback: Time at which last feedback was sent
143 * @rx_tstamp_last_ack: Time at which last feedback was sent
144 * @rx_hist: Packet history (loss detection + RTT sampling) 135 * @rx_hist: Packet history (loss detection + RTT sampling)
145 * @rx_li_hist: Loss Interval database 136 * @rx_li_hist: Loss Interval database
146 * @rx_s: Received packet size in bytes 137 * @rx_s: Received packet size in bytes
diff --git a/net/dccp/ccids/lib/loss_interval.c b/net/dccp/ccids/lib/loss_interval.c
index 8fc3cbf79071..497723c4d4bb 100644
--- a/net/dccp/ccids/lib/loss_interval.c
+++ b/net/dccp/ccids/lib/loss_interval.c
@@ -116,7 +116,7 @@ u8 tfrc_lh_update_i_mean(struct tfrc_loss_hist *lh, struct sk_buff *skb)
116 cur->li_length = len; 116 cur->li_length = len;
117 tfrc_lh_calc_i_mean(lh); 117 tfrc_lh_calc_i_mean(lh);
118 118
119 return (lh->i_mean < old_i_mean); 119 return lh->i_mean < old_i_mean;
120} 120}
121 121
122/* Determine if `new_loss' does begin a new loss interval [RFC 4342, 10.2] */ 122/* Determine if `new_loss' does begin a new loss interval [RFC 4342, 10.2] */
diff --git a/net/dccp/ccids/lib/packet_history.c b/net/dccp/ccids/lib/packet_history.c
index 3a4f414e94a0..de8fe294bf0b 100644
--- a/net/dccp/ccids/lib/packet_history.c
+++ b/net/dccp/ccids/lib/packet_history.c
@@ -38,18 +38,6 @@
38#include "packet_history.h" 38#include "packet_history.h"
39#include "../../dccp.h" 39#include "../../dccp.h"
40 40
41/**
42 * tfrc_tx_hist_entry - Simple singly-linked TX history list
43 * @next: next oldest entry (LIFO order)
44 * @seqno: sequence number of this entry
45 * @stamp: send time of packet with sequence number @seqno
46 */
47struct tfrc_tx_hist_entry {
48 struct tfrc_tx_hist_entry *next;
49 u64 seqno;
50 ktime_t stamp;
51};
52
53/* 41/*
54 * Transmitter History Routines 42 * Transmitter History Routines
55 */ 43 */
@@ -71,15 +59,6 @@ void tfrc_tx_packet_history_exit(void)
71 } 59 }
72} 60}
73 61
74static struct tfrc_tx_hist_entry *
75 tfrc_tx_hist_find_entry(struct tfrc_tx_hist_entry *head, u64 seqno)
76{
77 while (head != NULL && head->seqno != seqno)
78 head = head->next;
79
80 return head;
81}
82
83int tfrc_tx_hist_add(struct tfrc_tx_hist_entry **headp, u64 seqno) 62int tfrc_tx_hist_add(struct tfrc_tx_hist_entry **headp, u64 seqno)
84{ 63{
85 struct tfrc_tx_hist_entry *entry = kmem_cache_alloc(tfrc_tx_hist_slab, gfp_any()); 64 struct tfrc_tx_hist_entry *entry = kmem_cache_alloc(tfrc_tx_hist_slab, gfp_any());
@@ -107,24 +86,6 @@ void tfrc_tx_hist_purge(struct tfrc_tx_hist_entry **headp)
107 *headp = NULL; 86 *headp = NULL;
108} 87}
109 88
110u32 tfrc_tx_hist_rtt(struct tfrc_tx_hist_entry *head, const u64 seqno,
111 const ktime_t now)
112{
113 u32 rtt = 0;
114 struct tfrc_tx_hist_entry *packet = tfrc_tx_hist_find_entry(head, seqno);
115
116 if (packet != NULL) {
117 rtt = ktime_us_delta(now, packet->stamp);
118 /*
119 * Garbage-collect older (irrelevant) entries:
120 */
121 tfrc_tx_hist_purge(&packet->next);
122 }
123
124 return rtt;
125}
126
127
128/* 89/*
129 * Receiver History Routines 90 * Receiver History Routines
130 */ 91 */
diff --git a/net/dccp/ccids/lib/packet_history.h b/net/dccp/ccids/lib/packet_history.h
index 7df6c5299999..7ee4a9d9d335 100644
--- a/net/dccp/ccids/lib/packet_history.h
+++ b/net/dccp/ccids/lib/packet_history.h
@@ -40,12 +40,28 @@
40#include <linux/slab.h> 40#include <linux/slab.h>
41#include "tfrc.h" 41#include "tfrc.h"
42 42
43struct tfrc_tx_hist_entry; 43/**
44 * tfrc_tx_hist_entry - Simple singly-linked TX history list
45 * @next: next oldest entry (LIFO order)
46 * @seqno: sequence number of this entry
47 * @stamp: send time of packet with sequence number @seqno
48 */
49struct tfrc_tx_hist_entry {
50 struct tfrc_tx_hist_entry *next;
51 u64 seqno;
52 ktime_t stamp;
53};
54
55static inline struct tfrc_tx_hist_entry *
56 tfrc_tx_hist_find_entry(struct tfrc_tx_hist_entry *head, u64 seqno)
57{
58 while (head != NULL && head->seqno != seqno)
59 head = head->next;
60 return head;
61}
44 62
45extern int tfrc_tx_hist_add(struct tfrc_tx_hist_entry **headp, u64 seqno); 63extern int tfrc_tx_hist_add(struct tfrc_tx_hist_entry **headp, u64 seqno);
46extern void tfrc_tx_hist_purge(struct tfrc_tx_hist_entry **headp); 64extern void tfrc_tx_hist_purge(struct tfrc_tx_hist_entry **headp);
47extern u32 tfrc_tx_hist_rtt(struct tfrc_tx_hist_entry *head,
48 const u64 seqno, const ktime_t now);
49 65
50/* Subtraction a-b modulo-16, respects circular wrap-around */ 66/* Subtraction a-b modulo-16, respects circular wrap-around */
51#define SUB16(a, b) (((a) + 16 - (b)) & 0xF) 67#define SUB16(a, b) (((a) + 16 - (b)) & 0xF)
diff --git a/net/dccp/ccids/lib/tfrc.h b/net/dccp/ccids/lib/tfrc.h
index 01bb48e96c2e..f8ee3f549770 100644
--- a/net/dccp/ccids/lib/tfrc.h
+++ b/net/dccp/ccids/lib/tfrc.h
@@ -57,6 +57,7 @@ static inline u32 tfrc_ewma(const u32 avg, const u32 newval, const u8 weight)
57 57
58extern u32 tfrc_calc_x(u16 s, u32 R, u32 p); 58extern u32 tfrc_calc_x(u16 s, u32 R, u32 p);
59extern u32 tfrc_calc_x_reverse_lookup(u32 fvalue); 59extern u32 tfrc_calc_x_reverse_lookup(u32 fvalue);
60extern u32 tfrc_invert_loss_event_rate(u32 loss_event_rate);
60 61
61extern int tfrc_tx_packet_history_init(void); 62extern int tfrc_tx_packet_history_init(void);
62extern void tfrc_tx_packet_history_exit(void); 63extern void tfrc_tx_packet_history_exit(void);
diff --git a/net/dccp/ccids/lib/tfrc_equation.c b/net/dccp/ccids/lib/tfrc_equation.c
index 22ca1cf0eb55..a052a4377e26 100644
--- a/net/dccp/ccids/lib/tfrc_equation.c
+++ b/net/dccp/ccids/lib/tfrc_equation.c
@@ -687,3 +687,17 @@ u32 tfrc_calc_x_reverse_lookup(u32 fvalue)
687 index = tfrc_binsearch(fvalue, 0); 687 index = tfrc_binsearch(fvalue, 0);
688 return (index + 1) * 1000000 / TFRC_CALC_X_ARRSIZE; 688 return (index + 1) * 1000000 / TFRC_CALC_X_ARRSIZE;
689} 689}
690
691/**
692 * tfrc_invert_loss_event_rate - Compute p so that 10^6 corresponds to 100%
693 * When @loss_event_rate is large, there is a chance that p is truncated to 0.
694 * To avoid re-entering slow-start in that case, we set p = TFRC_SMALLEST_P > 0.
695 */
696u32 tfrc_invert_loss_event_rate(u32 loss_event_rate)
697{
698 if (loss_event_rate == UINT_MAX) /* see RFC 4342, 8.5 */
699 return 0;
700 if (unlikely(loss_event_rate == 0)) /* map 1/0 into 100% */
701 return 1000000;
702 return max_t(u32, scaled_div(1, loss_event_rate), TFRC_SMALLEST_P);
703}
diff --git a/net/dccp/dccp.h b/net/dccp/dccp.h
index 3ccef1b70fee..3eb264b60823 100644
--- a/net/dccp/dccp.h
+++ b/net/dccp/dccp.h
@@ -153,18 +153,27 @@ static inline u64 max48(const u64 seq1, const u64 seq2)
153} 153}
154 154
155/** 155/**
156 * dccp_loss_free - Evaluates condition for data loss from RFC 4340, 7.7.1 156 * dccp_loss_count - Approximate the number of lost data packets in a burst loss
157 * @s1: start sequence number 157 * @s1: last known sequence number before the loss ('hole')
158 * @s2: end sequence number 158 * @s2: first sequence number seen after the 'hole'
159 * @ndp: NDP count on packet with sequence number @s2 159 * @ndp: NDP count on packet with sequence number @s2
160 * Returns true if the sequence range s1...s2 has no data loss.
161 */ 160 */
162static inline bool dccp_loss_free(const u64 s1, const u64 s2, const u64 ndp) 161static inline u64 dccp_loss_count(const u64 s1, const u64 s2, const u64 ndp)
163{ 162{
164 s64 delta = dccp_delta_seqno(s1, s2); 163 s64 delta = dccp_delta_seqno(s1, s2);
165 164
166 WARN_ON(delta < 0); 165 WARN_ON(delta < 0);
167 return (u64)delta <= ndp + 1; 166 delta -= ndp + 1;
167
168 return delta > 0 ? delta : 0;
169}
170
171/**
172 * dccp_loss_free - Evaluate condition for data loss from RFC 4340, 7.7.1
173 */
174static inline bool dccp_loss_free(const u64 s1, const u64 s2, const u64 ndp)
175{
176 return dccp_loss_count(s1, s2, ndp) == 0;
168} 177}
169 178
170enum { 179enum {
@@ -246,7 +255,6 @@ static inline void dccp_clear_xmit_timers(struct sock *sk)
246extern unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu); 255extern unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu);
247 256
248extern const char *dccp_packet_name(const int type); 257extern const char *dccp_packet_name(const int type);
249extern const char *dccp_state_name(const int state);
250 258
251extern void dccp_set_state(struct sock *sk, const int state); 259extern void dccp_set_state(struct sock *sk, const int state);
252extern void dccp_done(struct sock *sk); 260extern void dccp_done(struct sock *sk);
@@ -415,6 +423,23 @@ static inline void dccp_update_gsr(struct sock *sk, u64 seq)
415 dp->dccps_gsr = seq; 423 dp->dccps_gsr = seq;
416 /* Sequence validity window depends on remote Sequence Window (7.5.1) */ 424 /* Sequence validity window depends on remote Sequence Window (7.5.1) */
417 dp->dccps_swl = SUB48(ADD48(dp->dccps_gsr, 1), dp->dccps_r_seq_win / 4); 425 dp->dccps_swl = SUB48(ADD48(dp->dccps_gsr, 1), dp->dccps_r_seq_win / 4);
426 /*
427 * Adjust SWL so that it is not below ISR. In contrast to RFC 4340,
428 * 7.5.1 we perform this check beyond the initial handshake: W/W' are
429 * always > 32, so for the first W/W' packets in the lifetime of a
430 * connection we always have to adjust SWL.
431 * A second reason why we are doing this is that the window depends on
432 * the feature-remote value of Sequence Window: nothing stops the peer
433 * from updating this value while we are busy adjusting SWL for the
434 * first W packets (we would have to count from scratch again then).
435 * Therefore it is safer to always make sure that the Sequence Window
436 * is not artificially extended by a peer who grows SWL downwards by
437 * continually updating the feature-remote Sequence-Window.
438 * If sequence numbers wrap it is bad luck. But that will take a while
439 * (48 bit), and this measure prevents Sequence-number attacks.
440 */
441 if (before48(dp->dccps_swl, dp->dccps_isr))
442 dp->dccps_swl = dp->dccps_isr;
418 dp->dccps_swh = ADD48(dp->dccps_gsr, (3 * dp->dccps_r_seq_win) / 4); 443 dp->dccps_swh = ADD48(dp->dccps_gsr, (3 * dp->dccps_r_seq_win) / 4);
419} 444}
420 445
@@ -425,14 +450,16 @@ static inline void dccp_update_gss(struct sock *sk, u64 seq)
425 dp->dccps_gss = seq; 450 dp->dccps_gss = seq;
426 /* Ack validity window depends on local Sequence Window value (7.5.1) */ 451 /* Ack validity window depends on local Sequence Window value (7.5.1) */
427 dp->dccps_awl = SUB48(ADD48(dp->dccps_gss, 1), dp->dccps_l_seq_win); 452 dp->dccps_awl = SUB48(ADD48(dp->dccps_gss, 1), dp->dccps_l_seq_win);
453 /* Adjust AWL so that it is not below ISS - see comment above for SWL */
454 if (before48(dp->dccps_awl, dp->dccps_iss))
455 dp->dccps_awl = dp->dccps_iss;
428 dp->dccps_awh = dp->dccps_gss; 456 dp->dccps_awh = dp->dccps_gss;
429} 457}
430 458
431static inline int dccp_ack_pending(const struct sock *sk) 459static inline int dccp_ack_pending(const struct sock *sk)
432{ 460{
433 const struct dccp_sock *dp = dccp_sk(sk); 461 const struct dccp_sock *dp = dccp_sk(sk);
434 return dp->dccps_timestamp_echo != 0 || 462 return (dp->dccps_hc_rx_ackvec != NULL &&
435 (dp->dccps_hc_rx_ackvec != NULL &&
436 dccp_ackvec_pending(dp->dccps_hc_rx_ackvec)) || 463 dccp_ackvec_pending(dp->dccps_hc_rx_ackvec)) ||
437 inet_csk_ack_scheduled(sk); 464 inet_csk_ack_scheduled(sk);
438} 465}
@@ -449,7 +476,6 @@ extern int dccp_insert_options_rsk(struct dccp_request_sock*, struct sk_buff*);
449extern int dccp_insert_option_elapsed_time(struct sk_buff *skb, u32 elapsed); 476extern int dccp_insert_option_elapsed_time(struct sk_buff *skb, u32 elapsed);
450extern u32 dccp_timestamp(void); 477extern u32 dccp_timestamp(void);
451extern void dccp_timestamping_init(void); 478extern void dccp_timestamping_init(void);
452extern int dccp_insert_option_timestamp(struct sk_buff *skb);
453extern int dccp_insert_option(struct sk_buff *skb, unsigned char option, 479extern int dccp_insert_option(struct sk_buff *skb, unsigned char option,
454 const void *value, unsigned char len); 480 const void *value, unsigned char len);
455 481
diff --git a/net/dccp/feat.c b/net/dccp/feat.c
index df7dd26cf07e..568def952722 100644
--- a/net/dccp/feat.c
+++ b/net/dccp/feat.c
@@ -730,16 +730,6 @@ int dccp_feat_register_sp(struct sock *sk, u8 feat, u8 is_local,
730 0, list, len); 730 0, list, len);
731} 731}
732 732
733/* Analogous to dccp_feat_register_sp(), but for non-negotiable values */
734int dccp_feat_register_nn(struct sock *sk, u8 feat, u64 val)
735{
736 /* any changes must be registered before establishing the connection */
737 if (sk->sk_state != DCCP_CLOSED)
738 return -EISCONN;
739 if (dccp_feat_type(feat) != FEAT_NN)
740 return -EINVAL;
741 return __feat_register_nn(&dccp_sk(sk)->dccps_featneg, feat, 0, val);
742}
743 733
744/* 734/*
745 * Tracking features whose value depend on the choice of CCID 735 * Tracking features whose value depend on the choice of CCID
diff --git a/net/dccp/feat.h b/net/dccp/feat.h
index f96721619def..e56a4e5e634e 100644
--- a/net/dccp/feat.h
+++ b/net/dccp/feat.h
@@ -111,7 +111,6 @@ extern int dccp_feat_init(struct sock *sk);
111extern void dccp_feat_initialise_sysctls(void); 111extern void dccp_feat_initialise_sysctls(void);
112extern int dccp_feat_register_sp(struct sock *sk, u8 feat, u8 is_local, 112extern int dccp_feat_register_sp(struct sock *sk, u8 feat, u8 is_local,
113 u8 const *list, u8 len); 113 u8 const *list, u8 len);
114extern int dccp_feat_register_nn(struct sock *sk, u8 feat, u64 val);
115extern int dccp_feat_parse_options(struct sock *, struct dccp_request_sock *, 114extern int dccp_feat_parse_options(struct sock *, struct dccp_request_sock *,
116 u8 mand, u8 opt, u8 feat, u8 *val, u8 len); 115 u8 mand, u8 opt, u8 feat, u8 *val, u8 len);
117extern int dccp_feat_clone_list(struct list_head const *, struct list_head *); 116extern int dccp_feat_clone_list(struct list_head const *, struct list_head *);
diff --git a/net/dccp/input.c b/net/dccp/input.c
index 10c957a88f4f..265985370fa1 100644
--- a/net/dccp/input.c
+++ b/net/dccp/input.c
@@ -259,7 +259,7 @@ static int dccp_check_seqno(struct sock *sk, struct sk_buff *skb)
259 sysctl_dccp_sync_ratelimit))) 259 sysctl_dccp_sync_ratelimit)))
260 return 0; 260 return 0;
261 261
262 DCCP_WARN("DCCP: Step 6 failed for %s packet, " 262 DCCP_WARN("Step 6 failed for %s packet, "
263 "(LSWL(%llu) <= P.seqno(%llu) <= S.SWH(%llu)) and " 263 "(LSWL(%llu) <= P.seqno(%llu) <= S.SWH(%llu)) and "
264 "(P.ackno %s or LAWL(%llu) <= P.ackno(%llu) <= S.AWH(%llu), " 264 "(P.ackno %s or LAWL(%llu) <= P.ackno(%llu) <= S.AWH(%llu), "
265 "sending SYNC...\n", dccp_packet_name(dh->dccph_type), 265 "sending SYNC...\n", dccp_packet_name(dh->dccph_type),
@@ -441,20 +441,14 @@ static int dccp_rcv_request_sent_state_process(struct sock *sk,
441 kfree_skb(sk->sk_send_head); 441 kfree_skb(sk->sk_send_head);
442 sk->sk_send_head = NULL; 442 sk->sk_send_head = NULL;
443 443
444 dp->dccps_isr = DCCP_SKB_CB(skb)->dccpd_seq;
445 dccp_update_gsr(sk, dp->dccps_isr);
446 /* 444 /*
447 * SWL and AWL are initially adjusted so that they are not less than 445 * Set ISR, GSR from packet. ISS was set in dccp_v{4,6}_connect
448 * the initial Sequence Numbers received and sent, respectively: 446 * and GSS in dccp_transmit_skb(). Setting AWL/AWH and SWL/SWH
449 * SWL := max(GSR + 1 - floor(W/4), ISR), 447 * is done as part of activating the feature values below, since
450 * AWL := max(GSS - W' + 1, ISS). 448 * these settings depend on the local/remote Sequence Window
451 * These adjustments MUST be applied only at the beginning of the 449 * features, which were undefined or not confirmed until now.
452 * connection.
453 *
454 * AWL was adjusted in dccp_v4_connect -acme
455 */ 450 */
456 dccp_set_seqno(&dp->dccps_swl, 451 dp->dccps_gsr = dp->dccps_isr = DCCP_SKB_CB(skb)->dccpd_seq;
457 max48(dp->dccps_swl, dp->dccps_isr));
458 452
459 dccp_sync_mss(sk, icsk->icsk_pmtu_cookie); 453 dccp_sync_mss(sk, icsk->icsk_pmtu_cookie);
460 454
diff --git a/net/dccp/ipv4.c b/net/dccp/ipv4.c
index d4a166f0f391..3f69ea114829 100644
--- a/net/dccp/ipv4.c
+++ b/net/dccp/ipv4.c
@@ -392,7 +392,7 @@ struct sock *dccp_v4_request_recv_sock(struct sock *sk, struct sk_buff *skb,
392 392
393 newsk = dccp_create_openreq_child(sk, req, skb); 393 newsk = dccp_create_openreq_child(sk, req, skb);
394 if (newsk == NULL) 394 if (newsk == NULL)
395 goto exit; 395 goto exit_nonewsk;
396 396
397 sk_setup_caps(newsk, dst); 397 sk_setup_caps(newsk, dst);
398 398
@@ -409,16 +409,20 @@ struct sock *dccp_v4_request_recv_sock(struct sock *sk, struct sk_buff *skb,
409 409
410 dccp_sync_mss(newsk, dst_mtu(dst)); 410 dccp_sync_mss(newsk, dst_mtu(dst));
411 411
412 if (__inet_inherit_port(sk, newsk) < 0) {
413 sock_put(newsk);
414 goto exit;
415 }
412 __inet_hash_nolisten(newsk, NULL); 416 __inet_hash_nolisten(newsk, NULL);
413 __inet_inherit_port(sk, newsk);
414 417
415 return newsk; 418 return newsk;
416 419
417exit_overflow: 420exit_overflow:
418 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); 421 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
422exit_nonewsk:
423 dst_release(dst);
419exit: 424exit:
420 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); 425 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
421 dst_release(dst);
422 return NULL; 426 return NULL;
423} 427}
424 428
diff --git a/net/dccp/ipv6.c b/net/dccp/ipv6.c
index 6e3f32575df7..dca711df9b60 100644
--- a/net/dccp/ipv6.c
+++ b/net/dccp/ipv6.c
@@ -564,7 +564,7 @@ static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
564 564
565 newsk = dccp_create_openreq_child(sk, req, skb); 565 newsk = dccp_create_openreq_child(sk, req, skb);
566 if (newsk == NULL) 566 if (newsk == NULL)
567 goto out; 567 goto out_nonewsk;
568 568
569 /* 569 /*
570 * No need to charge this sock to the relevant IPv6 refcnt debug socks 570 * No need to charge this sock to the relevant IPv6 refcnt debug socks
@@ -632,18 +632,22 @@ static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
632 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; 632 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
633 newinet->inet_rcv_saddr = LOOPBACK4_IPV6; 633 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
634 634
635 if (__inet_inherit_port(sk, newsk) < 0) {
636 sock_put(newsk);
637 goto out;
638 }
635 __inet6_hash(newsk, NULL); 639 __inet6_hash(newsk, NULL);
636 __inet_inherit_port(sk, newsk);
637 640
638 return newsk; 641 return newsk;
639 642
640out_overflow: 643out_overflow:
641 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); 644 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
645out_nonewsk:
646 dst_release(dst);
642out: 647out:
643 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); 648 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
644 if (opt != NULL && opt != np->opt) 649 if (opt != NULL && opt != np->opt)
645 sock_kfree_s(sk, opt, opt->tot_len); 650 sock_kfree_s(sk, opt, opt->tot_len);
646 dst_release(dst);
647 return NULL; 651 return NULL;
648} 652}
649 653
diff --git a/net/dccp/minisocks.c b/net/dccp/minisocks.c
index 128b089d3aef..d7041a0963af 100644
--- a/net/dccp/minisocks.c
+++ b/net/dccp/minisocks.c
@@ -121,30 +121,18 @@ struct sock *dccp_create_openreq_child(struct sock *sk,
121 * 121 *
122 * Choose S.ISS (initial seqno) or set from Init Cookies 122 * Choose S.ISS (initial seqno) or set from Init Cookies
123 * Initialize S.GAR := S.ISS 123 * Initialize S.GAR := S.ISS
124 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies 124 * Set S.ISR, S.GSR from packet (or Init Cookies)
125 */ 125 *
126 newdp->dccps_gar = newdp->dccps_iss = dreq->dreq_iss; 126 * Setting AWL/AWH and SWL/SWH happens as part of the feature
127 dccp_update_gss(newsk, dreq->dreq_iss); 127 * activation below, as these windows all depend on the local
128 128 * and remote Sequence Window feature values (7.5.2).
129 newdp->dccps_isr = dreq->dreq_isr;
130 dccp_update_gsr(newsk, dreq->dreq_isr);
131
132 /*
133 * SWL and AWL are initially adjusted so that they are not less than
134 * the initial Sequence Numbers received and sent, respectively:
135 * SWL := max(GSR + 1 - floor(W/4), ISR),
136 * AWL := max(GSS - W' + 1, ISS).
137 * These adjustments MUST be applied only at the beginning of the
138 * connection.
139 */ 129 */
140 dccp_set_seqno(&newdp->dccps_swl, 130 newdp->dccps_gss = newdp->dccps_iss = dreq->dreq_iss;
141 max48(newdp->dccps_swl, newdp->dccps_isr)); 131 newdp->dccps_gar = newdp->dccps_iss;
142 dccp_set_seqno(&newdp->dccps_awl, 132 newdp->dccps_gsr = newdp->dccps_isr = dreq->dreq_isr;
143 max48(newdp->dccps_awl, newdp->dccps_iss));
144 133
145 /* 134 /*
146 * Activate features after initialising the sequence numbers, 135 * Activate features: initialise CCIDs, sequence windows etc.
147 * since CCID initialisation may depend on GSS, ISR, ISS etc.
148 */ 136 */
149 if (dccp_feat_activate_values(newsk, &dreq->dreq_featneg)) { 137 if (dccp_feat_activate_values(newsk, &dreq->dreq_featneg)) {
150 /* It is still raw copy of parent, so invalidate 138 /* It is still raw copy of parent, so invalidate
diff --git a/net/dccp/options.c b/net/dccp/options.c
index bfda087bd90d..cd3061813009 100644
--- a/net/dccp/options.c
+++ b/net/dccp/options.c
@@ -96,18 +96,11 @@ int dccp_parse_options(struct sock *sk, struct dccp_request_sock *dreq,
96 } 96 }
97 97
98 /* 98 /*
99 * CCID-Specific Options (from RFC 4340, sec. 10.3):
100 *
101 * Option numbers 128 through 191 are for options sent from the
102 * HC-Sender to the HC-Receiver; option numbers 192 through 255
103 * are for options sent from the HC-Receiver to the HC-Sender.
104 *
105 * CCID-specific options are ignored during connection setup, as 99 * CCID-specific options are ignored during connection setup, as
106 * negotiation may still be in progress (see RFC 4340, 10.3). 100 * negotiation may still be in progress (see RFC 4340, 10.3).
107 * The same applies to Ack Vectors, as these depend on the CCID. 101 * The same applies to Ack Vectors, as these depend on the CCID.
108 *
109 */ 102 */
110 if (dreq != NULL && (opt >= 128 || 103 if (dreq != NULL && (opt >= DCCPO_MIN_RX_CCID_SPECIFIC ||
111 opt == DCCPO_ACK_VECTOR_0 || opt == DCCPO_ACK_VECTOR_1)) 104 opt == DCCPO_ACK_VECTOR_0 || opt == DCCPO_ACK_VECTOR_1))
112 goto ignore_option; 105 goto ignore_option;
113 106
@@ -170,6 +163,8 @@ int dccp_parse_options(struct sock *sk, struct dccp_request_sock *dreq,
170 dccp_role(sk), ntohl(opt_val), 163 dccp_role(sk), ntohl(opt_val),
171 (unsigned long long) 164 (unsigned long long)
172 DCCP_SKB_CB(skb)->dccpd_ack_seq); 165 DCCP_SKB_CB(skb)->dccpd_ack_seq);
166 /* schedule an Ack in case this sender is quiescent */
167 inet_csk_schedule_ack(sk);
173 break; 168 break;
174 case DCCPO_TIMESTAMP_ECHO: 169 case DCCPO_TIMESTAMP_ECHO:
175 if (len != 4 && len != 6 && len != 8) 170 if (len != 4 && len != 6 && len != 8)
@@ -226,23 +221,15 @@ int dccp_parse_options(struct sock *sk, struct dccp_request_sock *dreq,
226 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n", 221 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n",
227 dccp_role(sk), elapsed_time); 222 dccp_role(sk), elapsed_time);
228 break; 223 break;
229 case 128 ... 191: { 224 case DCCPO_MIN_RX_CCID_SPECIFIC ... DCCPO_MAX_RX_CCID_SPECIFIC:
230 const u16 idx = value - options;
231
232 if (ccid_hc_rx_parse_options(dp->dccps_hc_rx_ccid, sk, 225 if (ccid_hc_rx_parse_options(dp->dccps_hc_rx_ccid, sk,
233 opt, len, idx, 226 pkt_type, opt, value, len))
234 value) != 0)
235 goto out_invalid_option; 227 goto out_invalid_option;
236 }
237 break; 228 break;
238 case 192 ... 255: { 229 case DCCPO_MIN_TX_CCID_SPECIFIC ... DCCPO_MAX_TX_CCID_SPECIFIC:
239 const u16 idx = value - options;
240
241 if (ccid_hc_tx_parse_options(dp->dccps_hc_tx_ccid, sk, 230 if (ccid_hc_tx_parse_options(dp->dccps_hc_tx_ccid, sk,
242 opt, len, idx, 231 pkt_type, opt, value, len))
243 value) != 0)
244 goto out_invalid_option; 232 goto out_invalid_option;
245 }
246 break; 233 break;
247 default: 234 default:
248 DCCP_CRIT("DCCP(%p): option %d(len=%d) not " 235 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
@@ -384,7 +371,7 @@ int dccp_insert_option_elapsed_time(struct sk_buff *skb, u32 elapsed_time)
384 371
385EXPORT_SYMBOL_GPL(dccp_insert_option_elapsed_time); 372EXPORT_SYMBOL_GPL(dccp_insert_option_elapsed_time);
386 373
387int dccp_insert_option_timestamp(struct sk_buff *skb) 374static int dccp_insert_option_timestamp(struct sk_buff *skb)
388{ 375{
389 __be32 now = htonl(dccp_timestamp()); 376 __be32 now = htonl(dccp_timestamp());
390 /* yes this will overflow but that is the point as we want a 377 /* yes this will overflow but that is the point as we want a
@@ -393,8 +380,6 @@ int dccp_insert_option_timestamp(struct sk_buff *skb)
393 return dccp_insert_option(skb, DCCPO_TIMESTAMP, &now, sizeof(now)); 380 return dccp_insert_option(skb, DCCPO_TIMESTAMP, &now, sizeof(now));
394} 381}
395 382
396EXPORT_SYMBOL_GPL(dccp_insert_option_timestamp);
397
398static int dccp_insert_option_timestamp_echo(struct dccp_sock *dp, 383static int dccp_insert_option_timestamp_echo(struct dccp_sock *dp,
399 struct dccp_request_sock *dreq, 384 struct dccp_request_sock *dreq,
400 struct sk_buff *skb) 385 struct sk_buff *skb)
diff --git a/net/dccp/output.c b/net/dccp/output.c
index aadbdb58758b..a988fe9ffcba 100644
--- a/net/dccp/output.c
+++ b/net/dccp/output.c
@@ -304,7 +304,7 @@ void dccp_write_xmit(struct sock *sk, int block)
304 dcb->dccpd_type = DCCP_PKT_DATA; 304 dcb->dccpd_type = DCCP_PKT_DATA;
305 305
306 err = dccp_transmit_skb(sk, skb); 306 err = dccp_transmit_skb(sk, skb);
307 ccid_hc_tx_packet_sent(dp->dccps_hc_tx_ccid, sk, 0, len); 307 ccid_hc_tx_packet_sent(dp->dccps_hc_tx_ccid, sk, len);
308 if (err) 308 if (err)
309 DCCP_BUG("err=%d after ccid_hc_tx_packet_sent", 309 DCCP_BUG("err=%d after ccid_hc_tx_packet_sent",
310 err); 310 err);
@@ -474,8 +474,9 @@ int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code)
474/* 474/*
475 * Do all connect socket setups that can be done AF independent. 475 * Do all connect socket setups that can be done AF independent.
476 */ 476 */
477static inline void dccp_connect_init(struct sock *sk) 477int dccp_connect(struct sock *sk)
478{ 478{
479 struct sk_buff *skb;
479 struct dccp_sock *dp = dccp_sk(sk); 480 struct dccp_sock *dp = dccp_sk(sk);
480 struct dst_entry *dst = __sk_dst_get(sk); 481 struct dst_entry *dst = __sk_dst_get(sk);
481 struct inet_connection_sock *icsk = inet_csk(sk); 482 struct inet_connection_sock *icsk = inet_csk(sk);
@@ -485,22 +486,12 @@ static inline void dccp_connect_init(struct sock *sk)
485 486
486 dccp_sync_mss(sk, dst_mtu(dst)); 487 dccp_sync_mss(sk, dst_mtu(dst));
487 488
488 /* Initialise GAR as per 8.5; AWL/AWH are set in dccp_transmit_skb() */
489 dp->dccps_gar = dp->dccps_iss;
490
491 icsk->icsk_retransmits = 0;
492}
493
494int dccp_connect(struct sock *sk)
495{
496 struct sk_buff *skb;
497 struct inet_connection_sock *icsk = inet_csk(sk);
498
499 /* do not connect if feature negotiation setup fails */ 489 /* do not connect if feature negotiation setup fails */
500 if (dccp_feat_finalise_settings(dccp_sk(sk))) 490 if (dccp_feat_finalise_settings(dccp_sk(sk)))
501 return -EPROTO; 491 return -EPROTO;
502 492
503 dccp_connect_init(sk); 493 /* Initialise GAR as per 8.5; AWL/AWH are set in dccp_transmit_skb() */
494 dp->dccps_gar = dp->dccps_iss;
504 495
505 skb = alloc_skb(sk->sk_prot->max_header, sk->sk_allocation); 496 skb = alloc_skb(sk->sk_prot->max_header, sk->sk_allocation);
506 if (unlikely(skb == NULL)) 497 if (unlikely(skb == NULL))
@@ -516,6 +507,7 @@ int dccp_connect(struct sock *sk)
516 DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS); 507 DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS);
517 508
518 /* Timer for repeating the REQUEST until an answer. */ 509 /* Timer for repeating the REQUEST until an answer. */
510 icsk->icsk_retransmits = 0;
519 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, 511 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
520 icsk->icsk_rto, DCCP_RTO_MAX); 512 icsk->icsk_rto, DCCP_RTO_MAX);
521 return 0; 513 return 0;
diff --git a/net/dccp/proto.c b/net/dccp/proto.c
index 096250d1323b..7e5fc04eb6d1 100644
--- a/net/dccp/proto.c
+++ b/net/dccp/proto.c
@@ -50,6 +50,30 @@ EXPORT_SYMBOL_GPL(dccp_hashinfo);
50/* the maximum queue length for tx in packets. 0 is no limit */ 50/* the maximum queue length for tx in packets. 0 is no limit */
51int sysctl_dccp_tx_qlen __read_mostly = 5; 51int sysctl_dccp_tx_qlen __read_mostly = 5;
52 52
53#ifdef CONFIG_IP_DCCP_DEBUG
54static const char *dccp_state_name(const int state)
55{
56 static const char *const dccp_state_names[] = {
57 [DCCP_OPEN] = "OPEN",
58 [DCCP_REQUESTING] = "REQUESTING",
59 [DCCP_PARTOPEN] = "PARTOPEN",
60 [DCCP_LISTEN] = "LISTEN",
61 [DCCP_RESPOND] = "RESPOND",
62 [DCCP_CLOSING] = "CLOSING",
63 [DCCP_ACTIVE_CLOSEREQ] = "CLOSEREQ",
64 [DCCP_PASSIVE_CLOSE] = "PASSIVE_CLOSE",
65 [DCCP_PASSIVE_CLOSEREQ] = "PASSIVE_CLOSEREQ",
66 [DCCP_TIME_WAIT] = "TIME_WAIT",
67 [DCCP_CLOSED] = "CLOSED",
68 };
69
70 if (state >= DCCP_MAX_STATES)
71 return "INVALID STATE!";
72 else
73 return dccp_state_names[state];
74}
75#endif
76
53void dccp_set_state(struct sock *sk, const int state) 77void dccp_set_state(struct sock *sk, const int state)
54{ 78{
55 const int oldstate = sk->sk_state; 79 const int oldstate = sk->sk_state;
@@ -146,30 +170,6 @@ const char *dccp_packet_name(const int type)
146 170
147EXPORT_SYMBOL_GPL(dccp_packet_name); 171EXPORT_SYMBOL_GPL(dccp_packet_name);
148 172
149const char *dccp_state_name(const int state)
150{
151 static const char *const dccp_state_names[] = {
152 [DCCP_OPEN] = "OPEN",
153 [DCCP_REQUESTING] = "REQUESTING",
154 [DCCP_PARTOPEN] = "PARTOPEN",
155 [DCCP_LISTEN] = "LISTEN",
156 [DCCP_RESPOND] = "RESPOND",
157 [DCCP_CLOSING] = "CLOSING",
158 [DCCP_ACTIVE_CLOSEREQ] = "CLOSEREQ",
159 [DCCP_PASSIVE_CLOSE] = "PASSIVE_CLOSE",
160 [DCCP_PASSIVE_CLOSEREQ] = "PASSIVE_CLOSEREQ",
161 [DCCP_TIME_WAIT] = "TIME_WAIT",
162 [DCCP_CLOSED] = "CLOSED",
163 };
164
165 if (state >= DCCP_MAX_STATES)
166 return "INVALID STATE!";
167 else
168 return dccp_state_names[state];
169}
170
171EXPORT_SYMBOL_GPL(dccp_state_name);
172
173int dccp_init_sock(struct sock *sk, const __u8 ctl_sock_initialized) 173int dccp_init_sock(struct sock *sk, const __u8 ctl_sock_initialized)
174{ 174{
175 struct dccp_sock *dp = dccp_sk(sk); 175 struct dccp_sock *dp = dccp_sk(sk);
@@ -944,7 +944,7 @@ void dccp_close(struct sock *sk, long timeout)
944 944
945 if (data_was_unread) { 945 if (data_was_unread) {
946 /* Unread data was tossed, send an appropriate Reset Code */ 946 /* Unread data was tossed, send an appropriate Reset Code */
947 DCCP_WARN("DCCP: ABORT -- %u bytes unread\n", data_was_unread); 947 DCCP_WARN("ABORT with %u bytes unread\n", data_was_unread);
948 dccp_send_reset(sk, DCCP_RESET_CODE_ABORTED); 948 dccp_send_reset(sk, DCCP_RESET_CODE_ABORTED);
949 dccp_set_state(sk, DCCP_CLOSED); 949 dccp_set_state(sk, DCCP_CLOSED);
950 } else if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) { 950 } else if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-28 16:39:58 -0500