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_invalid_option
;
88 goto out_invalid_option
;
90 * Remove the type and len fields, leaving
97 if (opt_ptr
> opt_end
)
98 goto out_invalid_option
;
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).
112 if (dreq
!= NULL
&& opt
>= 128)
118 case DCCPO_MANDATORY
:
120 goto out_invalid_option
;
121 if (pkt_type
!= DCCP_PKT_DATA
)
124 case DCCPO_NDP_COUNT
:
126 goto out_invalid_option
;
128 opt_recv
->dccpor_ndp
= dccp_decode_value_var(value
, len
);
129 dccp_pr_debug("%s rx opt: NDP count=%d\n", dccp_role(sk
),
130 opt_recv
->dccpor_ndp
);
135 if (pkt_type
== DCCP_PKT_DATA
)
138 goto out_invalid_option
;
139 rc
= dccp_feat_change_recv(sk
, opt
, *value
, value
+ 1,
142 * When there is a change error, change_recv is
143 * responsible for dealing with it. i.e. reply with an
145 * If the change was mandatory, then we need to die.
148 goto out_invalid_option
;
150 case DCCPO_CONFIRM_L
:
152 case DCCPO_CONFIRM_R
:
153 if (pkt_type
== DCCP_PKT_DATA
)
155 if (len
< 2) /* FIXME this disallows empty confirm */
156 goto out_invalid_option
;
157 if (dccp_feat_confirm_recv(sk
, opt
, *value
,
159 goto out_invalid_option
;
161 case DCCPO_ACK_VECTOR_0
:
162 case DCCPO_ACK_VECTOR_1
:
163 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 11.4 */
166 if (dccp_msk(sk
)->dccpms_send_ack_vector
&&
167 dccp_ackvec_parse(sk
, skb
, &ackno
, opt
, value
, len
))
168 goto out_invalid_option
;
170 case DCCPO_TIMESTAMP
:
172 goto out_invalid_option
;
174 * RFC 4340 13.1: "The precise time corresponding to
175 * Timestamp Value zero is not specified". We use
176 * zero to indicate absence of a meaningful timestamp.
178 opt_val
= get_unaligned((__be32
*)value
);
179 if (unlikely(opt_val
== 0)) {
180 DCCP_WARN("Timestamp with zero value\n");
185 dreq
->dreq_timestamp_echo
= ntohl(opt_val
);
186 dreq
->dreq_timestamp_time
= dccp_timestamp();
188 opt_recv
->dccpor_timestamp
=
189 dp
->dccps_timestamp_echo
= ntohl(opt_val
);
190 dp
->dccps_timestamp_time
= dccp_timestamp();
192 dccp_pr_debug("%s rx opt: TIMESTAMP=%u, ackno=%llu\n",
193 dccp_role(sk
), ntohl(opt_val
),
195 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
197 case DCCPO_TIMESTAMP_ECHO
:
198 if (len
!= 4 && len
!= 6 && len
!= 8)
199 goto out_invalid_option
;
201 opt_val
= get_unaligned((__be32
*)value
);
202 opt_recv
->dccpor_timestamp_echo
= ntohl(opt_val
);
204 dccp_pr_debug("%s rx opt: TIMESTAMP_ECHO=%u, len=%d, "
205 "ackno=%llu", dccp_role(sk
),
206 opt_recv
->dccpor_timestamp_echo
,
209 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
213 if (len
== 4) { /* no elapsed time included */
214 dccp_pr_debug_cat("\n");
218 if (len
== 6) { /* 2-byte elapsed time */
219 __be16 opt_val2
= get_unaligned((__be16
*)value
);
220 elapsed_time
= ntohs(opt_val2
);
221 } else { /* 4-byte elapsed time */
222 opt_val
= get_unaligned((__be32
*)value
);
223 elapsed_time
= ntohl(opt_val
);
226 dccp_pr_debug_cat(", ELAPSED_TIME=%u\n", elapsed_time
);
228 /* Give precedence to the biggest ELAPSED_TIME */
229 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
230 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
232 case DCCPO_ELAPSED_TIME
:
233 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 13.2 */
237 __be16 opt_val2
= get_unaligned((__be16
*)value
);
238 elapsed_time
= ntohs(opt_val2
);
239 } else if (len
== 4) {
240 opt_val
= get_unaligned((__be32
*)value
);
241 elapsed_time
= ntohl(opt_val
);
243 goto out_invalid_option
;
246 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
247 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
249 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n",
250 dccp_role(sk
), elapsed_time
);
253 const u16 idx
= value
- options
;
255 if (ccid_hc_rx_parse_options(dp
->dccps_hc_rx_ccid
, sk
,
258 goto out_invalid_option
;
262 const u16 idx
= value
- options
;
264 if (ccid_hc_tx_parse_options(dp
->dccps_hc_tx_ccid
, sk
,
267 goto out_invalid_option
;
271 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
272 "implemented, ignoring", sk
, opt
, len
);
276 if (opt
!= DCCPO_MANDATORY
)
280 /* mandatory was the last byte in option list -> reset connection */
282 goto out_invalid_option
;
287 DCCP_INC_STATS_BH(DCCP_MIB_INVALIDOPT
);
288 DCCP_SKB_CB(skb
)->dccpd_reset_code
= DCCP_RESET_CODE_OPTION_ERROR
;
289 DCCP_WARN("DCCP(%p): invalid option %d, len=%d", sk
, opt
, len
);
293 EXPORT_SYMBOL_GPL(dccp_parse_options
);
295 static void dccp_encode_value_var(const u32 value
, unsigned char *to
,
296 const unsigned int len
)
299 *to
++ = (value
& 0xFF000000) >> 24;
301 *to
++ = (value
& 0xFF0000) >> 16;
303 *to
++ = (value
& 0xFF00) >> 8;
305 *to
++ = (value
& 0xFF);
308 static inline int dccp_ndp_len(const int ndp
)
310 return likely(ndp
<= 0xFF) ? 1 : ndp
<= 0xFFFF ? 2 : 3;
313 int dccp_insert_option(struct sock
*sk
, struct sk_buff
*skb
,
314 const unsigned char option
,
315 const void *value
, const unsigned char len
)
319 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
+ 2 > DCCP_MAX_OPT_LEN
)
322 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
+ 2;
324 to
= skb_push(skb
, len
+ 2);
328 memcpy(to
, value
, len
);
332 EXPORT_SYMBOL_GPL(dccp_insert_option
);
334 static int dccp_insert_option_ndp(struct sock
*sk
, struct sk_buff
*skb
)
336 struct dccp_sock
*dp
= dccp_sk(sk
);
337 int ndp
= dp
->dccps_ndp_count
;
339 if (dccp_non_data_packet(skb
))
340 ++dp
->dccps_ndp_count
;
342 dp
->dccps_ndp_count
= 0;
346 const int ndp_len
= dccp_ndp_len(ndp
);
347 const int len
= ndp_len
+ 2;
349 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
352 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
354 ptr
= skb_push(skb
, len
);
355 *ptr
++ = DCCPO_NDP_COUNT
;
357 dccp_encode_value_var(ndp
, ptr
, ndp_len
);
363 static inline int dccp_elapsed_time_len(const u32 elapsed_time
)
365 return elapsed_time
== 0 ? 0 : elapsed_time
<= 0xFFFF ? 2 : 4;
368 int dccp_insert_option_elapsed_time(struct sock
*sk
, struct sk_buff
*skb
,
371 const int elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
372 const int len
= 2 + elapsed_time_len
;
375 if (elapsed_time_len
== 0)
378 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
381 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
383 to
= skb_push(skb
, len
);
384 *to
++ = DCCPO_ELAPSED_TIME
;
387 if (elapsed_time_len
== 2) {
388 const __be16 var16
= htons((u16
)elapsed_time
);
389 memcpy(to
, &var16
, 2);
391 const __be32 var32
= htonl(elapsed_time
);
392 memcpy(to
, &var32
, 4);
398 EXPORT_SYMBOL_GPL(dccp_insert_option_elapsed_time
);
400 int dccp_insert_option_timestamp(struct sock
*sk
, struct sk_buff
*skb
)
402 __be32 now
= htonl(dccp_timestamp());
403 /* yes this will overflow but that is the point as we want a
404 * 10 usec 32 bit timer which mean it wraps every 11.9 hours */
406 return dccp_insert_option(sk
, skb
, DCCPO_TIMESTAMP
, &now
, sizeof(now
));
409 EXPORT_SYMBOL_GPL(dccp_insert_option_timestamp
);
411 static int dccp_insert_option_timestamp_echo(struct dccp_sock
*dp
,
412 struct dccp_request_sock
*dreq
,
417 u32 elapsed_time
, elapsed_time_len
, len
;
420 elapsed_time
= dccp_timestamp() - dreq
->dreq_timestamp_time
;
421 tstamp_echo
= htonl(dreq
->dreq_timestamp_echo
);
422 dreq
->dreq_timestamp_echo
= 0;
424 elapsed_time
= dccp_timestamp() - dp
->dccps_timestamp_time
;
425 tstamp_echo
= htonl(dp
->dccps_timestamp_echo
);
426 dp
->dccps_timestamp_echo
= 0;
429 elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
430 len
= 6 + elapsed_time_len
;
432 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
435 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
437 to
= skb_push(skb
, len
);
438 *to
++ = DCCPO_TIMESTAMP_ECHO
;
441 memcpy(to
, &tstamp_echo
, 4);
444 if (elapsed_time_len
== 2) {
445 const __be16 var16
= htons((u16
)elapsed_time
);
446 memcpy(to
, &var16
, 2);
447 } else if (elapsed_time_len
== 4) {
448 const __be32 var32
= htonl(elapsed_time
);
449 memcpy(to
, &var32
, 4);
455 static int dccp_insert_feat_opt(struct sk_buff
*skb
, u8 type
, u8 feat
,
460 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
+ 3 > DCCP_MAX_OPT_LEN
) {
461 DCCP_WARN("packet too small for feature %d option!\n", feat
);
465 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
+ 3;
467 to
= skb_push(skb
, len
+ 3);
473 memcpy(to
, val
, len
);
475 dccp_pr_debug("%s(%s (%d), ...), length %d\n",
476 dccp_feat_typename(type
),
477 dccp_feat_name(feat
), feat
, len
);
481 static int dccp_insert_options_feat(struct sock
*sk
, struct sk_buff
*skb
)
483 struct dccp_sock
*dp
= dccp_sk(sk
);
484 struct dccp_minisock
*dmsk
= dccp_msk(sk
);
485 struct dccp_opt_pend
*opt
, *next
;
488 /* confirm any options [NN opts] */
489 list_for_each_entry_safe(opt
, next
, &dmsk
->dccpms_conf
, dccpop_node
) {
490 dccp_insert_feat_opt(skb
, opt
->dccpop_type
,
491 opt
->dccpop_feat
, opt
->dccpop_val
,
493 /* fear empty confirms */
495 kfree(opt
->dccpop_val
);
498 INIT_LIST_HEAD(&dmsk
->dccpms_conf
);
500 /* see which features we need to send */
501 list_for_each_entry(opt
, &dmsk
->dccpms_pending
, dccpop_node
) {
502 /* see if we need to send any confirm */
503 if (opt
->dccpop_sc
) {
504 dccp_insert_feat_opt(skb
, opt
->dccpop_type
+ 1,
506 opt
->dccpop_sc
->dccpoc_val
,
507 opt
->dccpop_sc
->dccpoc_len
);
509 BUG_ON(!opt
->dccpop_sc
->dccpoc_val
);
510 kfree(opt
->dccpop_sc
->dccpoc_val
);
511 kfree(opt
->dccpop_sc
);
512 opt
->dccpop_sc
= NULL
;
515 /* any option not confirmed, re-send it */
516 if (!opt
->dccpop_conf
) {
517 dccp_insert_feat_opt(skb
, opt
->dccpop_type
,
518 opt
->dccpop_feat
, opt
->dccpop_val
,
525 * If this is the master listening sock, we don't set a timer on it. It
526 * should be fine because if the dude doesn't receive our RESPONSE
527 * [which will contain the CHANGE] he will send another REQUEST which
528 * will "retrnasmit" the change.
530 if (change
&& dp
->dccps_role
!= DCCP_ROLE_LISTEN
) {
531 dccp_pr_debug("reset feat negotiation timer %p\n", sk
);
533 /* XXX don't reset the timer on re-transmissions. I.e. reset it
534 * only when sending new stuff i guess. Currently the timer
535 * never backs off because on re-transmission it just resets it!
537 inet_csk_reset_xmit_timer(sk
, ICSK_TIME_RETRANS
,
538 inet_csk(sk
)->icsk_rto
, DCCP_RTO_MAX
);
544 /* The length of all options needs to be a multiple of 4 (5.8) */
545 static void dccp_insert_option_padding(struct sk_buff
*skb
)
547 int padding
= DCCP_SKB_CB(skb
)->dccpd_opt_len
% 4;
550 padding
= 4 - padding
;
551 memset(skb_push(skb
, padding
), 0, padding
);
552 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= padding
;
556 int dccp_insert_options(struct sock
*sk
, struct sk_buff
*skb
)
558 struct dccp_sock
*dp
= dccp_sk(sk
);
559 struct dccp_minisock
*dmsk
= dccp_msk(sk
);
561 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
563 if (dmsk
->dccpms_send_ndp_count
&&
564 dccp_insert_option_ndp(sk
, skb
))
567 if (!dccp_packet_without_ack(skb
)) {
568 if (dmsk
->dccpms_send_ack_vector
&&
569 dccp_ackvec_pending(dp
->dccps_hc_rx_ackvec
) &&
570 dccp_insert_option_ackvec(sk
, skb
))
574 if (dp
->dccps_hc_rx_insert_options
) {
575 if (ccid_hc_rx_insert_options(dp
->dccps_hc_rx_ccid
, sk
, skb
))
577 dp
->dccps_hc_rx_insert_options
= 0;
580 /* Feature negotiation */
581 /* Data packets can't do feat negotiation */
582 if (DCCP_SKB_CB(skb
)->dccpd_type
!= DCCP_PKT_DATA
&&
583 DCCP_SKB_CB(skb
)->dccpd_type
!= DCCP_PKT_DATAACK
&&
584 dccp_insert_options_feat(sk
, skb
))
588 * Obtain RTT sample from Request/Response exchange.
589 * This is currently used in CCID 3 initialisation.
591 if (DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_REQUEST
&&
592 dccp_insert_option_timestamp(sk
, skb
))
595 if (dp
->dccps_timestamp_echo
!= 0 &&
596 dccp_insert_option_timestamp_echo(dp
, NULL
, skb
))
599 dccp_insert_option_padding(skb
);
603 int dccp_insert_options_rsk(struct dccp_request_sock
*dreq
, struct sk_buff
*skb
)
605 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
607 if (dreq
->dreq_timestamp_echo
!= 0 &&
608 dccp_insert_option_timestamp_echo(NULL
, dreq
, skb
))
611 dccp_insert_option_padding(skb
);