4 * An implementation of the DCCP protocol
5 * Copyright (c) 2005 Aristeu Sergio Rozanski Filho <aris@cathedrallabs.org>
6 * Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
7 * Copyright (c) 2005 Ian McDonald <ian.mcdonald@jandi.co.nz>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
14 #include <linux/dccp.h>
15 #include <linux/module.h>
16 #include <linux/types.h>
17 #include <asm/unaligned.h>
18 #include <linux/kernel.h>
19 #include <linux/skbuff.h>
26 int sysctl_dccp_feat_sequence_window = DCCPF_INITIAL_SEQUENCE_WINDOW;
27 int sysctl_dccp_feat_rx_ccid = DCCPF_INITIAL_CCID;
28 int sysctl_dccp_feat_tx_ccid = DCCPF_INITIAL_CCID;
29 int sysctl_dccp_feat_ack_ratio = DCCPF_INITIAL_ACK_RATIO;
30 int sysctl_dccp_feat_send_ack_vector = DCCPF_INITIAL_SEND_ACK_VECTOR;
31 int sysctl_dccp_feat_send_ndp_count = DCCPF_INITIAL_SEND_NDP_COUNT;
33 static u32 dccp_decode_value_var(const unsigned char *bf, const u8 len)
50 * dccp_parse_options - Parse DCCP options present in @skb
51 * @sk: client|server|listening dccp socket (when @dreq != NULL)
52 * @dreq: request socket to use during connection setup, or NULL
54 int dccp_parse_options(struct sock *sk, struct dccp_request_sock *dreq,
57 struct dccp_sock *dp = dccp_sk(sk);
58 const struct dccp_hdr *dh = dccp_hdr(skb);
59 const u8 pkt_type = DCCP_SKB_CB(skb)->dccpd_type;
60 u64 ackno = DCCP_SKB_CB(skb)->dccpd_ack_seq;
61 unsigned char *options = (unsigned char *)dh + dccp_hdr_len(skb);
62 unsigned char *opt_ptr = options;
63 const unsigned char *opt_end = (unsigned char *)dh +
65 struct dccp_options_received *opt_recv = &dp->dccps_options_received;
66 unsigned char opt, len;
73 memset(opt_recv, 0, sizeof(*opt_recv));
76 while (opt_ptr != opt_end) {
81 /* Check if this isn't a single byte option */
82 if (opt > DCCPO_MAX_RESERVED) {
83 if (opt_ptr == opt_end)
84 goto out_nonsensical_length;
88 goto out_nonsensical_length;
90 * Remove the type and len fields, leaving
97 if (opt_ptr > opt_end)
98 goto out_nonsensical_length;
102 * CCID-Specific Options (from RFC 4340, sec. 10.3):
104 * Option numbers 128 through 191 are for options sent from the
105 * HC-Sender to the HC-Receiver; option numbers 192 through 255
106 * are for options sent from the HC-Receiver to the HC-Sender.
108 * CCID-specific options are ignored during connection setup, as
109 * negotiation may still be in progress (see RFC 4340, 10.3).
110 * The same applies to Ack Vectors, as these depend on the CCID.
113 if (dreq != NULL && (opt >= 128 ||
114 opt == DCCPO_ACK_VECTOR_0 || opt == DCCPO_ACK_VECTOR_1))
120 case DCCPO_MANDATORY:
122 goto out_invalid_option;
123 if (pkt_type != DCCP_PKT_DATA)
126 case DCCPO_NDP_COUNT:
128 goto out_invalid_option;
130 opt_recv->dccpor_ndp = dccp_decode_value_var(value, len);
131 dccp_pr_debug("%s opt: NDP count=%llu\n", dccp_role(sk),
132 (unsigned long long)opt_recv->dccpor_ndp);
137 if (pkt_type == DCCP_PKT_DATA)
140 goto out_invalid_option;
141 rc = dccp_feat_change_recv(sk, opt, *value, value + 1,
144 * When there is a change error, change_recv is
145 * responsible for dealing with it. i.e. reply with an
147 * If the change was mandatory, then we need to die.
150 goto out_invalid_option;
152 case DCCPO_CONFIRM_L:
154 case DCCPO_CONFIRM_R:
155 if (pkt_type == DCCP_PKT_DATA)
157 if (len < 2) /* FIXME this disallows empty confirm */
158 goto out_invalid_option;
159 if (dccp_feat_confirm_recv(sk, opt, *value,
161 goto out_invalid_option;
163 case DCCPO_ACK_VECTOR_0:
164 case DCCPO_ACK_VECTOR_1:
165 if (dccp_packet_without_ack(skb)) /* RFC 4340, 11.4 */
168 if (dccp_msk(sk)->dccpms_send_ack_vector &&
169 dccp_ackvec_parse(sk, skb, &ackno, opt, value, len))
170 goto out_invalid_option;
172 case DCCPO_TIMESTAMP:
174 goto out_invalid_option;
176 * RFC 4340 13.1: "The precise time corresponding to
177 * Timestamp Value zero is not specified". We use
178 * zero to indicate absence of a meaningful timestamp.
180 opt_val = get_unaligned((__be32 *)value);
181 if (unlikely(opt_val == 0)) {
182 DCCP_WARN("Timestamp with zero value\n");
187 dreq->dreq_timestamp_echo = ntohl(opt_val);
188 dreq->dreq_timestamp_time = dccp_timestamp();
190 opt_recv->dccpor_timestamp =
191 dp->dccps_timestamp_echo = ntohl(opt_val);
192 dp->dccps_timestamp_time = dccp_timestamp();
194 dccp_pr_debug("%s rx opt: TIMESTAMP=%u, ackno=%llu\n",
195 dccp_role(sk), ntohl(opt_val),
197 DCCP_SKB_CB(skb)->dccpd_ack_seq);
199 case DCCPO_TIMESTAMP_ECHO:
200 if (len != 4 && len != 6 && len != 8)
201 goto out_invalid_option;
203 opt_val = get_unaligned((__be32 *)value);
204 opt_recv->dccpor_timestamp_echo = ntohl(opt_val);
206 dccp_pr_debug("%s rx opt: TIMESTAMP_ECHO=%u, len=%d, "
207 "ackno=%llu", dccp_role(sk),
208 opt_recv->dccpor_timestamp_echo,
211 DCCP_SKB_CB(skb)->dccpd_ack_seq);
215 if (len == 4) { /* no elapsed time included */
216 dccp_pr_debug_cat("\n");
220 if (len == 6) { /* 2-byte elapsed time */
221 __be16 opt_val2 = get_unaligned((__be16 *)value);
222 elapsed_time = ntohs(opt_val2);
223 } else { /* 4-byte elapsed time */
224 opt_val = get_unaligned((__be32 *)value);
225 elapsed_time = ntohl(opt_val);
228 dccp_pr_debug_cat(", ELAPSED_TIME=%u\n", elapsed_time);
230 /* Give precedence to the biggest ELAPSED_TIME */
231 if (elapsed_time > opt_recv->dccpor_elapsed_time)
232 opt_recv->dccpor_elapsed_time = elapsed_time;
234 case DCCPO_ELAPSED_TIME:
235 if (dccp_packet_without_ack(skb)) /* RFC 4340, 13.2 */
239 __be16 opt_val2 = get_unaligned((__be16 *)value);
240 elapsed_time = ntohs(opt_val2);
241 } else if (len == 4) {
242 opt_val = get_unaligned((__be32 *)value);
243 elapsed_time = ntohl(opt_val);
245 goto out_invalid_option;
248 if (elapsed_time > opt_recv->dccpor_elapsed_time)
249 opt_recv->dccpor_elapsed_time = elapsed_time;
251 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n",
252 dccp_role(sk), elapsed_time);
255 const u16 idx = value - options;
257 if (ccid_hc_rx_parse_options(dp->dccps_hc_rx_ccid, sk,
260 goto out_invalid_option;
264 const u16 idx = value - options;
266 if (ccid_hc_tx_parse_options(dp->dccps_hc_tx_ccid, sk,
269 goto out_invalid_option;
273 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
274 "implemented, ignoring", sk, opt, len);
278 if (opt != DCCPO_MANDATORY)
282 /* mandatory was the last byte in option list -> reset connection */
284 goto out_invalid_option;
286 out_nonsensical_length:
287 /* RFC 4340, 5.8: ignore option and all remaining option space */
291 DCCP_INC_STATS_BH(DCCP_MIB_INVALIDOPT);
292 DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_OPTION_ERROR;
293 DCCP_WARN("DCCP(%p): invalid option %d, len=%d", sk, opt, len);
294 DCCP_SKB_CB(skb)->dccpd_reset_data[0] = opt;
295 DCCP_SKB_CB(skb)->dccpd_reset_data[1] = len > 0 ? value[0] : 0;
296 DCCP_SKB_CB(skb)->dccpd_reset_data[2] = len > 1 ? value[1] : 0;
300 EXPORT_SYMBOL_GPL(dccp_parse_options);
302 static void dccp_encode_value_var(const u32 value, unsigned char *to,
303 const unsigned int len)
306 *to++ = (value & 0xFF000000) >> 24;
308 *to++ = (value & 0xFF0000) >> 16;
310 *to++ = (value & 0xFF00) >> 8;
312 *to++ = (value & 0xFF);
315 static inline u8 dccp_ndp_len(const u64 ndp)
317 if (likely(ndp <= 0xFF))
319 return likely(ndp <= USHORT_MAX) ? 2 : (ndp <= UINT_MAX ? 4 : 6);
322 int dccp_insert_option(struct sock *sk, struct sk_buff *skb,
323 const unsigned char option,
324 const void *value, const unsigned char len)
328 if (DCCP_SKB_CB(skb)->dccpd_opt_len + len + 2 > DCCP_MAX_OPT_LEN)
331 DCCP_SKB_CB(skb)->dccpd_opt_len += len + 2;
333 to = skb_push(skb, len + 2);
337 memcpy(to, value, len);
341 EXPORT_SYMBOL_GPL(dccp_insert_option);
343 static int dccp_insert_option_ndp(struct sock *sk, struct sk_buff *skb)
345 struct dccp_sock *dp = dccp_sk(sk);
346 u64 ndp = dp->dccps_ndp_count;
348 if (dccp_non_data_packet(skb))
349 ++dp->dccps_ndp_count;
351 dp->dccps_ndp_count = 0;
355 const int ndp_len = dccp_ndp_len(ndp);
356 const int len = ndp_len + 2;
358 if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
361 DCCP_SKB_CB(skb)->dccpd_opt_len += len;
363 ptr = skb_push(skb, len);
364 *ptr++ = DCCPO_NDP_COUNT;
366 dccp_encode_value_var(ndp, ptr, ndp_len);
372 static inline int dccp_elapsed_time_len(const u32 elapsed_time)
374 return elapsed_time == 0 ? 0 : elapsed_time <= 0xFFFF ? 2 : 4;
377 int dccp_insert_option_elapsed_time(struct sock *sk, struct sk_buff *skb,
380 const int elapsed_time_len = dccp_elapsed_time_len(elapsed_time);
381 const int len = 2 + elapsed_time_len;
384 if (elapsed_time_len == 0)
387 if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
390 DCCP_SKB_CB(skb)->dccpd_opt_len += len;
392 to = skb_push(skb, len);
393 *to++ = DCCPO_ELAPSED_TIME;
396 if (elapsed_time_len == 2) {
397 const __be16 var16 = htons((u16)elapsed_time);
398 memcpy(to, &var16, 2);
400 const __be32 var32 = htonl(elapsed_time);
401 memcpy(to, &var32, 4);
407 EXPORT_SYMBOL_GPL(dccp_insert_option_elapsed_time);
409 int dccp_insert_option_timestamp(struct sock *sk, struct sk_buff *skb)
411 __be32 now = htonl(dccp_timestamp());
412 /* yes this will overflow but that is the point as we want a
413 * 10 usec 32 bit timer which mean it wraps every 11.9 hours */
415 return dccp_insert_option(sk, skb, DCCPO_TIMESTAMP, &now, sizeof(now));
418 EXPORT_SYMBOL_GPL(dccp_insert_option_timestamp);
420 static int dccp_insert_option_timestamp_echo(struct dccp_sock *dp,
421 struct dccp_request_sock *dreq,
426 u32 elapsed_time, elapsed_time_len, len;
429 elapsed_time = dccp_timestamp() - dreq->dreq_timestamp_time;
430 tstamp_echo = htonl(dreq->dreq_timestamp_echo);
431 dreq->dreq_timestamp_echo = 0;
433 elapsed_time = dccp_timestamp() - dp->dccps_timestamp_time;
434 tstamp_echo = htonl(dp->dccps_timestamp_echo);
435 dp->dccps_timestamp_echo = 0;
438 elapsed_time_len = dccp_elapsed_time_len(elapsed_time);
439 len = 6 + elapsed_time_len;
441 if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
444 DCCP_SKB_CB(skb)->dccpd_opt_len += len;
446 to = skb_push(skb, len);
447 *to++ = DCCPO_TIMESTAMP_ECHO;
450 memcpy(to, &tstamp_echo, 4);
453 if (elapsed_time_len == 2) {
454 const __be16 var16 = htons((u16)elapsed_time);
455 memcpy(to, &var16, 2);
456 } else if (elapsed_time_len == 4) {
457 const __be32 var32 = htonl(elapsed_time);
458 memcpy(to, &var32, 4);
464 static int dccp_insert_feat_opt(struct sk_buff *skb, u8 type, u8 feat,
469 if (DCCP_SKB_CB(skb)->dccpd_opt_len + len + 3 > DCCP_MAX_OPT_LEN) {
470 DCCP_WARN("packet too small for feature %d option!\n", feat);
474 DCCP_SKB_CB(skb)->dccpd_opt_len += len + 3;
476 to = skb_push(skb, len + 3);
482 memcpy(to, val, len);
484 dccp_pr_debug("%s(%s (%d), ...), length %d\n",
485 dccp_feat_typename(type),
486 dccp_feat_name(feat), feat, len);
490 static int dccp_insert_options_feat(struct sock *sk, struct sk_buff *skb)
492 struct dccp_sock *dp = dccp_sk(sk);
493 struct dccp_minisock *dmsk = dccp_msk(sk);
494 struct dccp_opt_pend *opt, *next;
497 /* confirm any options [NN opts] */
498 list_for_each_entry_safe(opt, next, &dmsk->dccpms_conf, dccpop_node) {
499 dccp_insert_feat_opt(skb, opt->dccpop_type,
500 opt->dccpop_feat, opt->dccpop_val,
502 /* fear empty confirms */
504 kfree(opt->dccpop_val);
507 INIT_LIST_HEAD(&dmsk->dccpms_conf);
509 /* see which features we need to send */
510 list_for_each_entry(opt, &dmsk->dccpms_pending, dccpop_node) {
511 /* see if we need to send any confirm */
512 if (opt->dccpop_sc) {
513 dccp_insert_feat_opt(skb, opt->dccpop_type + 1,
515 opt->dccpop_sc->dccpoc_val,
516 opt->dccpop_sc->dccpoc_len);
518 BUG_ON(!opt->dccpop_sc->dccpoc_val);
519 kfree(opt->dccpop_sc->dccpoc_val);
520 kfree(opt->dccpop_sc);
521 opt->dccpop_sc = NULL;
524 /* any option not confirmed, re-send it */
525 if (!opt->dccpop_conf) {
526 dccp_insert_feat_opt(skb, opt->dccpop_type,
527 opt->dccpop_feat, opt->dccpop_val,
534 * If this is the master listening sock, we don't set a timer on it. It
535 * should be fine because if the dude doesn't receive our RESPONSE
536 * [which will contain the CHANGE] he will send another REQUEST which
537 * will "retrnasmit" the change.
539 if (change && dp->dccps_role != DCCP_ROLE_LISTEN) {
540 dccp_pr_debug("reset feat negotiation timer %p\n", sk);
542 /* XXX don't reset the timer on re-transmissions. I.e. reset it
543 * only when sending new stuff i guess. Currently the timer
544 * never backs off because on re-transmission it just resets it!
546 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
547 inet_csk(sk)->icsk_rto, DCCP_RTO_MAX);
553 /* The length of all options needs to be a multiple of 4 (5.8) */
554 static void dccp_insert_option_padding(struct sk_buff *skb)
556 int padding = DCCP_SKB_CB(skb)->dccpd_opt_len % 4;
559 padding = 4 - padding;
560 memset(skb_push(skb, padding), 0, padding);
561 DCCP_SKB_CB(skb)->dccpd_opt_len += padding;
565 int dccp_insert_options(struct sock *sk, struct sk_buff *skb)
567 struct dccp_sock *dp = dccp_sk(sk);
568 struct dccp_minisock *dmsk = dccp_msk(sk);
570 DCCP_SKB_CB(skb)->dccpd_opt_len = 0;
572 if (dmsk->dccpms_send_ndp_count &&
573 dccp_insert_option_ndp(sk, skb))
576 if (!dccp_packet_without_ack(skb)) {
577 if (dmsk->dccpms_send_ack_vector &&
578 dccp_ackvec_pending(dp->dccps_hc_rx_ackvec) &&
579 dccp_insert_option_ackvec(sk, skb))
583 if (dp->dccps_hc_rx_insert_options) {
584 if (ccid_hc_rx_insert_options(dp->dccps_hc_rx_ccid, sk, skb))
586 dp->dccps_hc_rx_insert_options = 0;
589 /* Feature negotiation */
590 /* Data packets can't do feat negotiation */
591 if (DCCP_SKB_CB(skb)->dccpd_type != DCCP_PKT_DATA &&
592 DCCP_SKB_CB(skb)->dccpd_type != DCCP_PKT_DATAACK &&
593 dccp_insert_options_feat(sk, skb))
597 * Obtain RTT sample from Request/Response exchange.
598 * This is currently used in CCID 3 initialisation.
600 if (DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_REQUEST &&
601 dccp_insert_option_timestamp(sk, skb))
604 if (dp->dccps_timestamp_echo != 0 &&
605 dccp_insert_option_timestamp_echo(dp, NULL, skb))
608 dccp_insert_option_padding(skb);
612 int dccp_insert_options_rsk(struct dccp_request_sock *dreq, struct sk_buff *skb)
614 DCCP_SKB_CB(skb)->dccpd_opt_len = 0;
616 if (dreq->dreq_timestamp_echo != 0 &&
617 dccp_insert_option_timestamp_echo(NULL, dreq, skb))
620 dccp_insert_option_padding(skb);