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 <linux/kernel.h>
18 #include <linux/skbuff.h>
25 int sysctl_dccp_feat_sequence_window
= DCCPF_INITIAL_SEQUENCE_WINDOW
;
26 int sysctl_dccp_feat_rx_ccid
= DCCPF_INITIAL_CCID
;
27 int sysctl_dccp_feat_tx_ccid
= DCCPF_INITIAL_CCID
;
28 int sysctl_dccp_feat_ack_ratio
= DCCPF_INITIAL_ACK_RATIO
;
29 int sysctl_dccp_feat_send_ack_vector
= DCCPF_INITIAL_SEND_ACK_VECTOR
;
30 int sysctl_dccp_feat_send_ndp_count
= DCCPF_INITIAL_SEND_NDP_COUNT
;
32 void dccp_minisock_init(struct dccp_minisock
*dmsk
)
34 dmsk
->dccpms_sequence_window
= sysctl_dccp_feat_sequence_window
;
35 dmsk
->dccpms_rx_ccid
= sysctl_dccp_feat_rx_ccid
;
36 dmsk
->dccpms_tx_ccid
= sysctl_dccp_feat_tx_ccid
;
37 dmsk
->dccpms_ack_ratio
= sysctl_dccp_feat_ack_ratio
;
38 dmsk
->dccpms_send_ack_vector
= sysctl_dccp_feat_send_ack_vector
;
39 dmsk
->dccpms_send_ndp_count
= sysctl_dccp_feat_send_ndp_count
;
42 static u32
dccp_decode_value_var(const unsigned char *bf
, const u8 len
)
58 int dccp_parse_options(struct sock
*sk
, struct sk_buff
*skb
)
60 struct dccp_sock
*dp
= dccp_sk(sk
);
61 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
62 const u8 pkt_type
= DCCP_SKB_CB(skb
)->dccpd_type
;
63 u64 ackno
= DCCP_SKB_CB(skb
)->dccpd_ack_seq
;
64 unsigned char *options
= (unsigned char *)dh
+ dccp_hdr_len(skb
);
65 unsigned char *opt_ptr
= options
;
66 const unsigned char *opt_end
= (unsigned char *)dh
+
68 struct dccp_options_received
*opt_recv
= &dp
->dccps_options_received
;
69 unsigned char opt
, len
;
75 memset(opt_recv
, 0, sizeof(*opt_recv
));
78 while (opt_ptr
!= opt_end
) {
83 /* Check if this isn't a single byte option */
84 if (opt
> DCCPO_MAX_RESERVED
) {
85 if (opt_ptr
== opt_end
)
86 goto out_invalid_option
;
90 goto out_invalid_option
;
92 * Remove the type and len fields, leaving
99 if (opt_ptr
> opt_end
)
100 goto out_invalid_option
;
106 case DCCPO_MANDATORY
:
108 goto out_invalid_option
;
109 if (pkt_type
!= DCCP_PKT_DATA
)
112 case DCCPO_NDP_COUNT
:
114 goto out_invalid_option
;
116 opt_recv
->dccpor_ndp
= dccp_decode_value_var(value
, len
);
117 dccp_pr_debug("%s rx opt: NDP count=%d\n", dccp_role(sk
),
118 opt_recv
->dccpor_ndp
);
124 goto out_invalid_option
;
125 rc
= dccp_feat_change_recv(sk
, opt
, *value
, value
+ 1,
128 * When there is a change error, change_recv is
129 * responsible for dealing with it. i.e. reply with an
131 * If the change was mandatory, then we need to die.
134 goto out_invalid_option
;
136 case DCCPO_CONFIRM_L
:
138 case DCCPO_CONFIRM_R
:
140 goto out_invalid_option
;
141 if (dccp_feat_confirm_recv(sk
, opt
, *value
,
143 goto out_invalid_option
;
145 case DCCPO_ACK_VECTOR_0
:
146 case DCCPO_ACK_VECTOR_1
:
147 if (pkt_type
== DCCP_PKT_DATA
)
150 if (dccp_msk(sk
)->dccpms_send_ack_vector
&&
151 dccp_ackvec_parse(sk
, skb
, &ackno
, opt
, value
, len
))
152 goto out_invalid_option
;
154 case DCCPO_TIMESTAMP
:
156 goto out_invalid_option
;
158 opt_recv
->dccpor_timestamp
= ntohl(*(__be32
*)value
);
160 dp
->dccps_timestamp_echo
= opt_recv
->dccpor_timestamp
;
161 dccp_timestamp(sk
, &dp
->dccps_timestamp_time
);
163 dccp_pr_debug("%s rx opt: TIMESTAMP=%u, ackno=%llu\n",
164 dccp_role(sk
), opt_recv
->dccpor_timestamp
,
166 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
168 case DCCPO_TIMESTAMP_ECHO
:
169 if (len
!= 4 && len
!= 6 && len
!= 8)
170 goto out_invalid_option
;
172 opt_recv
->dccpor_timestamp_echo
= ntohl(*(__be32
*)value
);
174 dccp_pr_debug("%s rx opt: TIMESTAMP_ECHO=%u, len=%d, "
175 "ackno=%llu", dccp_role(sk
),
176 opt_recv
->dccpor_timestamp_echo
,
179 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
183 dccp_pr_debug_cat("\n");
188 elapsed_time
= ntohs(*(__be16
*)(value
+ 4));
190 elapsed_time
= ntohl(*(__be32
*)(value
+ 4));
192 dccp_pr_debug_cat(", ELAPSED_TIME=%d\n", elapsed_time
);
194 /* Give precedence to the biggest ELAPSED_TIME */
195 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
196 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
198 case DCCPO_ELAPSED_TIME
:
199 if (len
!= 2 && len
!= 4)
200 goto out_invalid_option
;
202 if (pkt_type
== DCCP_PKT_DATA
)
206 elapsed_time
= ntohs(*(__be16
*)value
);
208 elapsed_time
= ntohl(*(__be32
*)value
);
210 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
211 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
213 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n",
214 dccp_role(sk
), elapsed_time
);
217 * From RFC 4340, sec. 10.3:
219 * Option numbers 128 through 191 are for
220 * options sent from the HC-Sender to the
221 * HC-Receiver; option numbers 192 through 255
222 * are for options sent from the HC-Receiver to
226 const u16 idx
= value
- options
;
228 if (ccid_hc_rx_parse_options(dp
->dccps_hc_rx_ccid
, sk
,
231 goto out_invalid_option
;
235 const u16 idx
= value
- options
;
237 if (ccid_hc_tx_parse_options(dp
->dccps_hc_tx_ccid
, sk
,
240 goto out_invalid_option
;
244 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
245 "implemented, ignoring", sk
, opt
, len
);
249 if (opt
!= DCCPO_MANDATORY
)
253 /* mandatory was the last byte in option list -> reset connection */
255 goto out_invalid_option
;
260 DCCP_INC_STATS_BH(DCCP_MIB_INVALIDOPT
);
261 DCCP_SKB_CB(skb
)->dccpd_reset_code
= DCCP_RESET_CODE_OPTION_ERROR
;
262 DCCP_WARN("DCCP(%p): invalid option %d, len=%d", sk
, opt
, len
);
266 EXPORT_SYMBOL_GPL(dccp_parse_options
);
268 static void dccp_encode_value_var(const u32 value
, unsigned char *to
,
269 const unsigned int len
)
272 *to
++ = (value
& 0xFF000000) >> 24;
274 *to
++ = (value
& 0xFF0000) >> 16;
276 *to
++ = (value
& 0xFF00) >> 8;
278 *to
++ = (value
& 0xFF);
281 static inline int dccp_ndp_len(const int ndp
)
283 return likely(ndp
<= 0xFF) ? 1 : ndp
<= 0xFFFF ? 2 : 3;
286 int dccp_insert_option(struct sock
*sk
, struct sk_buff
*skb
,
287 const unsigned char option
,
288 const void *value
, const unsigned char len
)
292 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
+ 2 > DCCP_MAX_OPT_LEN
)
295 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
+ 2;
297 to
= skb_push(skb
, len
+ 2);
301 memcpy(to
, value
, len
);
305 EXPORT_SYMBOL_GPL(dccp_insert_option
);
307 static int dccp_insert_option_ndp(struct sock
*sk
, struct sk_buff
*skb
)
309 struct dccp_sock
*dp
= dccp_sk(sk
);
310 int ndp
= dp
->dccps_ndp_count
;
312 if (dccp_non_data_packet(skb
))
313 ++dp
->dccps_ndp_count
;
315 dp
->dccps_ndp_count
= 0;
319 const int ndp_len
= dccp_ndp_len(ndp
);
320 const int len
= ndp_len
+ 2;
322 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
325 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
327 ptr
= skb_push(skb
, len
);
328 *ptr
++ = DCCPO_NDP_COUNT
;
330 dccp_encode_value_var(ndp
, ptr
, ndp_len
);
336 static inline int dccp_elapsed_time_len(const u32 elapsed_time
)
338 return elapsed_time
== 0 ? 0 : elapsed_time
<= 0xFFFF ? 2 : 4;
341 int dccp_insert_option_elapsed_time(struct sock
*sk
, struct sk_buff
*skb
,
344 const int elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
345 const int len
= 2 + elapsed_time_len
;
348 if (elapsed_time_len
== 0)
351 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
354 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
356 to
= skb_push(skb
, len
);
357 *to
++ = DCCPO_ELAPSED_TIME
;
360 if (elapsed_time_len
== 2) {
361 const __be16 var16
= htons((u16
)elapsed_time
);
362 memcpy(to
, &var16
, 2);
364 const __be32 var32
= htonl(elapsed_time
);
365 memcpy(to
, &var32
, 4);
371 EXPORT_SYMBOL_GPL(dccp_insert_option_elapsed_time
);
373 void dccp_timestamp(const struct sock
*sk
, struct timeval
*tv
)
375 const struct dccp_sock
*dp
= dccp_sk(sk
);
378 tv
->tv_sec
-= dp
->dccps_epoch
.tv_sec
;
379 tv
->tv_usec
-= dp
->dccps_epoch
.tv_usec
;
381 while (tv
->tv_usec
< 0) {
383 tv
->tv_usec
+= USEC_PER_SEC
;
387 EXPORT_SYMBOL_GPL(dccp_timestamp
);
389 int dccp_insert_option_timestamp(struct sock
*sk
, struct sk_buff
*skb
)
394 dccp_timestamp(sk
, &tv
);
395 now
= htonl(timeval_usecs(&tv
) / 10);
396 /* yes this will overflow but that is the point as we want a
397 * 10 usec 32 bit timer which mean it wraps every 11.9 hours */
399 return dccp_insert_option(sk
, skb
, DCCPO_TIMESTAMP
, &now
, sizeof(now
));
402 EXPORT_SYMBOL_GPL(dccp_insert_option_timestamp
);
404 static int dccp_insert_option_timestamp_echo(struct sock
*sk
,
407 struct dccp_sock
*dp
= dccp_sk(sk
);
411 int len
, elapsed_time_len
;
414 dccp_timestamp(sk
, &now
);
415 elapsed_time
= timeval_delta(&now
, &dp
->dccps_timestamp_time
) / 10;
416 elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
417 len
= 6 + elapsed_time_len
;
419 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
422 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
424 to
= skb_push(skb
, len
);
425 *to
++ = DCCPO_TIMESTAMP_ECHO
;
428 tstamp_echo
= htonl(dp
->dccps_timestamp_echo
);
429 memcpy(to
, &tstamp_echo
, 4);
432 if (elapsed_time_len
== 2) {
433 const __be16 var16
= htons((u16
)elapsed_time
);
434 memcpy(to
, &var16
, 2);
435 } else if (elapsed_time_len
== 4) {
436 const __be32 var32
= htonl(elapsed_time
);
437 memcpy(to
, &var32
, 4);
440 dp
->dccps_timestamp_echo
= 0;
441 dp
->dccps_timestamp_time
.tv_sec
= 0;
442 dp
->dccps_timestamp_time
.tv_usec
= 0;
446 static int dccp_insert_feat_opt(struct sk_buff
*skb
, u8 type
, u8 feat
,
451 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
+ 3 > DCCP_MAX_OPT_LEN
) {
452 DCCP_WARN("packet too small for feature %d option!\n", feat
);
456 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
+ 3;
458 to
= skb_push(skb
, len
+ 3);
464 memcpy(to
, val
, len
);
466 dccp_pr_debug("%s(%s (%d), ...), length %d\n",
467 dccp_feat_typename(type
),
468 dccp_feat_name(feat
), feat
, len
);
472 static int dccp_insert_options_feat(struct sock
*sk
, struct sk_buff
*skb
)
474 struct dccp_sock
*dp
= dccp_sk(sk
);
475 struct dccp_minisock
*dmsk
= dccp_msk(sk
);
476 struct dccp_opt_pend
*opt
, *next
;
479 /* confirm any options [NN opts] */
480 list_for_each_entry_safe(opt
, next
, &dmsk
->dccpms_conf
, dccpop_node
) {
481 dccp_insert_feat_opt(skb
, opt
->dccpop_type
,
482 opt
->dccpop_feat
, opt
->dccpop_val
,
484 /* fear empty confirms */
486 kfree(opt
->dccpop_val
);
489 INIT_LIST_HEAD(&dmsk
->dccpms_conf
);
491 /* see which features we need to send */
492 list_for_each_entry(opt
, &dmsk
->dccpms_pending
, dccpop_node
) {
493 /* see if we need to send any confirm */
494 if (opt
->dccpop_sc
) {
495 dccp_insert_feat_opt(skb
, opt
->dccpop_type
+ 1,
497 opt
->dccpop_sc
->dccpoc_val
,
498 opt
->dccpop_sc
->dccpoc_len
);
500 BUG_ON(!opt
->dccpop_sc
->dccpoc_val
);
501 kfree(opt
->dccpop_sc
->dccpoc_val
);
502 kfree(opt
->dccpop_sc
);
503 opt
->dccpop_sc
= NULL
;
506 /* any option not confirmed, re-send it */
507 if (!opt
->dccpop_conf
) {
508 dccp_insert_feat_opt(skb
, opt
->dccpop_type
,
509 opt
->dccpop_feat
, opt
->dccpop_val
,
516 * If this is the master listening sock, we don't set a timer on it. It
517 * should be fine because if the dude doesn't receive our RESPONSE
518 * [which will contain the CHANGE] he will send another REQUEST which
519 * will "retrnasmit" the change.
521 if (change
&& dp
->dccps_role
!= DCCP_ROLE_LISTEN
) {
522 dccp_pr_debug("reset feat negotiation timer %p\n", sk
);
524 /* XXX don't reset the timer on re-transmissions. I.e. reset it
525 * only when sending new stuff i guess. Currently the timer
526 * never backs off because on re-transmission it just resets it!
528 inet_csk_reset_xmit_timer(sk
, ICSK_TIME_RETRANS
,
529 inet_csk(sk
)->icsk_rto
, DCCP_RTO_MAX
);
535 int dccp_insert_options(struct sock
*sk
, struct sk_buff
*skb
)
537 struct dccp_sock
*dp
= dccp_sk(sk
);
538 struct dccp_minisock
*dmsk
= dccp_msk(sk
);
540 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
542 if (dmsk
->dccpms_send_ndp_count
&&
543 dccp_insert_option_ndp(sk
, skb
))
546 if (!dccp_packet_without_ack(skb
)) {
547 if (dmsk
->dccpms_send_ack_vector
&&
548 dccp_ackvec_pending(dp
->dccps_hc_rx_ackvec
) &&
549 dccp_insert_option_ackvec(sk
, skb
))
552 if (dp
->dccps_timestamp_echo
!= 0 &&
553 dccp_insert_option_timestamp_echo(sk
, skb
))
557 if (dp
->dccps_hc_rx_insert_options
) {
558 if (ccid_hc_rx_insert_options(dp
->dccps_hc_rx_ccid
, sk
, skb
))
560 dp
->dccps_hc_rx_insert_options
= 0;
563 /* Feature negotiation */
564 /* Data packets can't do feat negotiation */
565 if (DCCP_SKB_CB(skb
)->dccpd_type
!= DCCP_PKT_DATA
&&
566 DCCP_SKB_CB(skb
)->dccpd_type
!= DCCP_PKT_DATAACK
&&
567 dccp_insert_options_feat(sk
, skb
))
571 * Obtain RTT sample from Request/Response exchange.
572 * This is currently used in CCID 3 initialisation.
574 if (DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_REQUEST
&&
575 dccp_insert_option_timestamp(sk
, skb
))
578 /* XXX: insert other options when appropriate */
580 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
!= 0) {
581 /* The length of all options has to be a multiple of 4 */
582 int padding
= DCCP_SKB_CB(skb
)->dccpd_opt_len
% 4;
585 padding
= 4 - padding
;
586 memset(skb_push(skb
, padding
), 0, padding
);
587 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= padding
;