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 u64
dccp_decode_value_var(const u8
*bf
, const u8 len
)
30 if (len
>= DCCP_OPTVAL_MAXLEN
)
31 value
+= ((u64
)*bf
++) << 40;
33 value
+= ((u64
)*bf
++) << 32;
35 value
+= ((u64
)*bf
++) << 24;
37 value
+= ((u64
)*bf
++) << 16;
39 value
+= ((u64
)*bf
++) << 8;
47 * dccp_parse_options - Parse DCCP options present in @skb
48 * @sk: client|server|listening dccp socket (when @dreq != NULL)
49 * @dreq: request socket to use during connection setup, or NULL
51 int dccp_parse_options(struct sock
*sk
, struct dccp_request_sock
*dreq
,
54 struct dccp_sock
*dp
= dccp_sk(sk
);
55 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
56 const u8 pkt_type
= DCCP_SKB_CB(skb
)->dccpd_type
;
57 unsigned char *options
= (unsigned char *)dh
+ dccp_hdr_len(skb
);
58 unsigned char *opt_ptr
= options
;
59 const unsigned char *opt_end
= (unsigned char *)dh
+
61 struct dccp_options_received
*opt_recv
= &dp
->dccps_options_received
;
62 unsigned char opt
, len
;
63 unsigned char *uninitialized_var(value
);
69 memset(opt_recv
, 0, sizeof(*opt_recv
));
72 while (opt_ptr
!= opt_end
) {
77 /* Check if this isn't a single byte option */
78 if (opt
> DCCPO_MAX_RESERVED
) {
79 if (opt_ptr
== opt_end
)
80 goto out_nonsensical_length
;
84 goto out_nonsensical_length
;
86 * Remove the type and len fields, leaving
93 if (opt_ptr
> opt_end
)
94 goto out_nonsensical_length
;
98 * CCID-specific options are ignored during connection setup, as
99 * negotiation may still be in progress (see RFC 4340, 10.3).
100 * The same applies to Ack Vectors, as these depend on the CCID.
102 if (dreq
!= NULL
&& (opt
>= DCCPO_MIN_RX_CCID_SPECIFIC
||
103 opt
== DCCPO_ACK_VECTOR_0
|| opt
== DCCPO_ACK_VECTOR_1
))
109 case DCCPO_MANDATORY
:
111 goto out_invalid_option
;
112 if (pkt_type
!= DCCP_PKT_DATA
)
115 case DCCPO_NDP_COUNT
:
117 goto out_invalid_option
;
119 opt_recv
->dccpor_ndp
= dccp_decode_value_var(value
, len
);
120 dccp_pr_debug("%s opt: NDP count=%llu\n", dccp_role(sk
),
121 (unsigned long long)opt_recv
->dccpor_ndp
);
123 case DCCPO_CHANGE_L
... DCCPO_CONFIRM_R
:
124 if (pkt_type
== DCCP_PKT_DATA
) /* RFC 4340, 6 */
126 rc
= dccp_feat_parse_options(sk
, dreq
, mandatory
, opt
,
127 *value
, value
+ 1, len
- 1);
129 goto out_featneg_failed
;
131 case DCCPO_TIMESTAMP
:
133 goto out_invalid_option
;
135 * RFC 4340 13.1: "The precise time corresponding to
136 * Timestamp Value zero is not specified". We use
137 * zero to indicate absence of a meaningful timestamp.
139 opt_val
= get_unaligned((__be32
*)value
);
140 if (unlikely(opt_val
== 0)) {
141 DCCP_WARN("Timestamp with zero value\n");
146 dreq
->dreq_timestamp_echo
= ntohl(opt_val
);
147 dreq
->dreq_timestamp_time
= dccp_timestamp();
149 opt_recv
->dccpor_timestamp
=
150 dp
->dccps_timestamp_echo
= ntohl(opt_val
);
151 dp
->dccps_timestamp_time
= dccp_timestamp();
153 dccp_pr_debug("%s rx opt: TIMESTAMP=%u, ackno=%llu\n",
154 dccp_role(sk
), ntohl(opt_val
),
156 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
157 /* schedule an Ack in case this sender is quiescent */
158 inet_csk_schedule_ack(sk
);
160 case DCCPO_TIMESTAMP_ECHO
:
161 if (len
!= 4 && len
!= 6 && len
!= 8)
162 goto out_invalid_option
;
164 opt_val
= get_unaligned((__be32
*)value
);
165 opt_recv
->dccpor_timestamp_echo
= ntohl(opt_val
);
167 dccp_pr_debug("%s rx opt: TIMESTAMP_ECHO=%u, len=%d, "
168 "ackno=%llu", dccp_role(sk
),
169 opt_recv
->dccpor_timestamp_echo
,
172 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
176 if (len
== 4) { /* no elapsed time included */
177 dccp_pr_debug_cat("\n");
181 if (len
== 6) { /* 2-byte elapsed time */
182 __be16 opt_val2
= get_unaligned((__be16
*)value
);
183 elapsed_time
= ntohs(opt_val2
);
184 } else { /* 4-byte elapsed time */
185 opt_val
= get_unaligned((__be32
*)value
);
186 elapsed_time
= ntohl(opt_val
);
189 dccp_pr_debug_cat(", ELAPSED_TIME=%u\n", elapsed_time
);
191 /* Give precedence to the biggest ELAPSED_TIME */
192 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
193 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
195 case DCCPO_ELAPSED_TIME
:
196 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 13.2 */
200 __be16 opt_val2
= get_unaligned((__be16
*)value
);
201 elapsed_time
= ntohs(opt_val2
);
202 } else if (len
== 4) {
203 opt_val
= get_unaligned((__be32
*)value
);
204 elapsed_time
= ntohl(opt_val
);
206 goto out_invalid_option
;
209 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
210 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
212 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n",
213 dccp_role(sk
), elapsed_time
);
215 case DCCPO_MIN_RX_CCID_SPECIFIC
... DCCPO_MAX_RX_CCID_SPECIFIC
:
216 if (ccid_hc_rx_parse_options(dp
->dccps_hc_rx_ccid
, sk
,
217 pkt_type
, opt
, value
, len
))
218 goto out_invalid_option
;
220 case DCCPO_ACK_VECTOR_0
:
221 case DCCPO_ACK_VECTOR_1
:
222 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 11.4 */
225 * Ack vectors are processed by the TX CCID if it is
226 * interested. The RX CCID need not parse Ack Vectors,
227 * since it is only interested in clearing old state.
230 case DCCPO_MIN_TX_CCID_SPECIFIC
... DCCPO_MAX_TX_CCID_SPECIFIC
:
231 if (ccid_hc_tx_parse_options(dp
->dccps_hc_tx_ccid
, sk
,
232 pkt_type
, opt
, value
, len
))
233 goto out_invalid_option
;
236 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
237 "implemented, ignoring", sk
, opt
, len
);
241 if (opt
!= DCCPO_MANDATORY
)
245 /* mandatory was the last byte in option list -> reset connection */
247 goto out_invalid_option
;
249 out_nonsensical_length
:
250 /* RFC 4340, 5.8: ignore option and all remaining option space */
254 DCCP_INC_STATS_BH(DCCP_MIB_INVALIDOPT
);
255 rc
= DCCP_RESET_CODE_OPTION_ERROR
;
257 DCCP_WARN("DCCP(%p): Option %d (len=%d) error=%u\n", sk
, opt
, len
, rc
);
258 DCCP_SKB_CB(skb
)->dccpd_reset_code
= rc
;
259 DCCP_SKB_CB(skb
)->dccpd_reset_data
[0] = opt
;
260 DCCP_SKB_CB(skb
)->dccpd_reset_data
[1] = len
> 0 ? value
[0] : 0;
261 DCCP_SKB_CB(skb
)->dccpd_reset_data
[2] = len
> 1 ? value
[1] : 0;
265 EXPORT_SYMBOL_GPL(dccp_parse_options
);
267 void dccp_encode_value_var(const u64 value
, u8
*to
, const u8 len
)
269 if (len
>= DCCP_OPTVAL_MAXLEN
)
270 *to
++ = (value
& 0xFF0000000000ull
) >> 40;
272 *to
++ = (value
& 0xFF00000000ull
) >> 32;
274 *to
++ = (value
& 0xFF000000) >> 24;
276 *to
++ = (value
& 0xFF0000) >> 16;
278 *to
++ = (value
& 0xFF00) >> 8;
280 *to
++ = (value
& 0xFF);
283 static inline u8
dccp_ndp_len(const u64 ndp
)
285 if (likely(ndp
<= 0xFF))
287 return likely(ndp
<= USHRT_MAX
) ? 2 : (ndp
<= UINT_MAX
? 4 : 6);
290 int dccp_insert_option(struct sk_buff
*skb
, const unsigned char option
,
291 const void *value
, const unsigned char len
)
295 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
+ 2 > DCCP_MAX_OPT_LEN
)
298 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
+ 2;
300 to
= skb_push(skb
, len
+ 2);
304 memcpy(to
, value
, len
);
308 EXPORT_SYMBOL_GPL(dccp_insert_option
);
310 static int dccp_insert_option_ndp(struct sock
*sk
, struct sk_buff
*skb
)
312 struct dccp_sock
*dp
= dccp_sk(sk
);
313 u64 ndp
= dp
->dccps_ndp_count
;
315 if (dccp_non_data_packet(skb
))
316 ++dp
->dccps_ndp_count
;
318 dp
->dccps_ndp_count
= 0;
322 const int ndp_len
= dccp_ndp_len(ndp
);
323 const int len
= ndp_len
+ 2;
325 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
328 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
330 ptr
= skb_push(skb
, len
);
331 *ptr
++ = DCCPO_NDP_COUNT
;
333 dccp_encode_value_var(ndp
, ptr
, ndp_len
);
339 static inline int dccp_elapsed_time_len(const u32 elapsed_time
)
341 return elapsed_time
== 0 ? 0 : elapsed_time
<= 0xFFFF ? 2 : 4;
344 /* FIXME: This function is currently not used anywhere */
345 int dccp_insert_option_elapsed_time(struct sk_buff
*skb
, u32 elapsed_time
)
347 const int elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
348 const int len
= 2 + elapsed_time_len
;
351 if (elapsed_time_len
== 0)
354 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
357 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
359 to
= skb_push(skb
, len
);
360 *to
++ = DCCPO_ELAPSED_TIME
;
363 if (elapsed_time_len
== 2) {
364 const __be16 var16
= htons((u16
)elapsed_time
);
365 memcpy(to
, &var16
, 2);
367 const __be32 var32
= htonl(elapsed_time
);
368 memcpy(to
, &var32
, 4);
374 EXPORT_SYMBOL_GPL(dccp_insert_option_elapsed_time
);
376 static int dccp_insert_option_timestamp(struct sk_buff
*skb
)
378 __be32 now
= htonl(dccp_timestamp());
379 /* yes this will overflow but that is the point as we want a
380 * 10 usec 32 bit timer which mean it wraps every 11.9 hours */
382 return dccp_insert_option(skb
, DCCPO_TIMESTAMP
, &now
, sizeof(now
));
385 static int dccp_insert_option_timestamp_echo(struct dccp_sock
*dp
,
386 struct dccp_request_sock
*dreq
,
391 u32 elapsed_time
, elapsed_time_len
, len
;
394 elapsed_time
= dccp_timestamp() - dreq
->dreq_timestamp_time
;
395 tstamp_echo
= htonl(dreq
->dreq_timestamp_echo
);
396 dreq
->dreq_timestamp_echo
= 0;
398 elapsed_time
= dccp_timestamp() - dp
->dccps_timestamp_time
;
399 tstamp_echo
= htonl(dp
->dccps_timestamp_echo
);
400 dp
->dccps_timestamp_echo
= 0;
403 elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
404 len
= 6 + elapsed_time_len
;
406 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
409 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
411 to
= skb_push(skb
, len
);
412 *to
++ = DCCPO_TIMESTAMP_ECHO
;
415 memcpy(to
, &tstamp_echo
, 4);
418 if (elapsed_time_len
== 2) {
419 const __be16 var16
= htons((u16
)elapsed_time
);
420 memcpy(to
, &var16
, 2);
421 } else if (elapsed_time_len
== 4) {
422 const __be32 var32
= htonl(elapsed_time
);
423 memcpy(to
, &var32
, 4);
429 static int dccp_insert_option_ackvec(struct sock
*sk
, struct sk_buff
*skb
)
431 struct dccp_sock
*dp
= dccp_sk(sk
);
432 struct dccp_ackvec
*av
= dp
->dccps_hc_rx_ackvec
;
433 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
434 const u16 buflen
= dccp_ackvec_buflen(av
);
435 /* Figure out how many options do we need to represent the ackvec */
436 const u8 nr_opts
= DIV_ROUND_UP(buflen
, DCCP_SINGLE_OPT_MAXLEN
);
437 u16 len
= buflen
+ 2 * nr_opts
;
439 const unsigned char *tail
, *from
;
442 if (dcb
->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
) {
443 DCCP_WARN("Lacking space for %u bytes on %s packet\n", len
,
444 dccp_packet_name(dcb
->dccpd_type
));
448 * Since Ack Vectors are variable-length, we can not always predict
449 * their size. To catch exception cases where the space is running out
450 * on the skb, a separate Sync is scheduled to carry the Ack Vector.
452 if (len
> DCCPAV_MIN_OPTLEN
&&
453 len
+ dcb
->dccpd_opt_len
+ skb
->len
> dp
->dccps_mss_cache
) {
454 DCCP_WARN("No space left for Ack Vector (%u) on skb (%u+%u), "
455 "MPS=%u ==> reduce payload size?\n", len
, skb
->len
,
456 dcb
->dccpd_opt_len
, dp
->dccps_mss_cache
);
457 dp
->dccps_sync_scheduled
= 1;
460 dcb
->dccpd_opt_len
+= len
;
462 to
= skb_push(skb
, len
);
464 from
= av
->av_buf
+ av
->av_buf_head
;
465 tail
= av
->av_buf
+ DCCPAV_MAX_ACKVEC_LEN
;
467 for (i
= 0; i
< nr_opts
; ++i
) {
470 if (len
> DCCP_SINGLE_OPT_MAXLEN
)
471 copylen
= DCCP_SINGLE_OPT_MAXLEN
;
474 * RFC 4340, 12.2: Encode the Nonce Echo for this Ack Vector via
475 * its type; ack_nonce is the sum of all individual buf_nonce's.
477 nonce
^= av
->av_buf_nonce
[i
];
479 *to
++ = DCCPO_ACK_VECTOR_0
+ av
->av_buf_nonce
[i
];
482 /* Check if buf_head wraps */
483 if (from
+ copylen
> tail
) {
484 const u16 tailsize
= tail
- from
;
486 memcpy(to
, from
, tailsize
);
493 memcpy(to
, from
, copylen
);
499 * Each sent Ack Vector is recorded in the list, as per A.2 of RFC 4340.
501 if (dccp_ackvec_update_records(av
, dcb
->dccpd_seq
, nonce
))
507 * dccp_insert_option_mandatory - Mandatory option (5.8.2)
508 * Note that since we are using skb_push, this function needs to be called
509 * _after_ inserting the option it is supposed to influence (stack order).
511 int dccp_insert_option_mandatory(struct sk_buff
*skb
)
513 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
>= DCCP_MAX_OPT_LEN
)
516 DCCP_SKB_CB(skb
)->dccpd_opt_len
++;
517 *skb_push(skb
, 1) = DCCPO_MANDATORY
;
522 * dccp_insert_fn_opt - Insert single Feature-Negotiation option into @skb
523 * @type: %DCCPO_CHANGE_L, %DCCPO_CHANGE_R, %DCCPO_CONFIRM_L, %DCCPO_CONFIRM_R
524 * @feat: one out of %dccp_feature_numbers
525 * @val: NN value or SP array (preferred element first) to copy
526 * @len: true length of @val in bytes (excluding first element repetition)
527 * @repeat_first: whether to copy the first element of @val twice
528 * The last argument is used to construct Confirm options, where the preferred
529 * value and the preference list appear separately (RFC 4340, 6.3.1). Preference
530 * lists are kept such that the preferred entry is always first, so we only need
531 * to copy twice, and avoid the overhead of cloning into a bigger array.
533 int dccp_insert_fn_opt(struct sk_buff
*skb
, u8 type
, u8 feat
,
534 u8
*val
, u8 len
, bool repeat_first
)
538 /* take the `Feature' field and possible repetition into account */
539 if (len
> (DCCP_SINGLE_OPT_MAXLEN
- 2)) {
540 DCCP_WARN("length %u for feature %u too large\n", len
, feat
);
544 if (unlikely(val
== NULL
|| len
== 0))
545 len
= repeat_first
= 0;
546 tot_len
= 3 + repeat_first
+ len
;
548 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ tot_len
> DCCP_MAX_OPT_LEN
) {
549 DCCP_WARN("packet too small for feature %d option!\n", feat
);
552 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= tot_len
;
554 to
= skb_push(skb
, tot_len
);
562 memcpy(to
, val
, len
);
566 /* The length of all options needs to be a multiple of 4 (5.8) */
567 static void dccp_insert_option_padding(struct sk_buff
*skb
)
569 int padding
= DCCP_SKB_CB(skb
)->dccpd_opt_len
% 4;
572 padding
= 4 - padding
;
573 memset(skb_push(skb
, padding
), 0, padding
);
574 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= padding
;
578 int dccp_insert_options(struct sock
*sk
, struct sk_buff
*skb
)
580 struct dccp_sock
*dp
= dccp_sk(sk
);
582 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
584 if (dp
->dccps_send_ndp_count
&& dccp_insert_option_ndp(sk
, skb
))
587 if (DCCP_SKB_CB(skb
)->dccpd_type
!= DCCP_PKT_DATA
) {
589 /* Feature Negotiation */
590 if (dccp_feat_insert_opts(dp
, NULL
, skb
))
593 if (DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_REQUEST
) {
595 * Obtain RTT sample from Request/Response exchange.
596 * This is currently used for TFRC initialisation.
598 if (dccp_insert_option_timestamp(skb
))
601 } else if (dccp_ackvec_pending(sk
) &&
602 dccp_insert_option_ackvec(sk
, skb
)) {
607 if (dp
->dccps_hc_rx_insert_options
) {
608 if (ccid_hc_rx_insert_options(dp
->dccps_hc_rx_ccid
, sk
, skb
))
610 dp
->dccps_hc_rx_insert_options
= 0;
613 if (dp
->dccps_timestamp_echo
!= 0 &&
614 dccp_insert_option_timestamp_echo(dp
, NULL
, skb
))
617 dccp_insert_option_padding(skb
);
621 int dccp_insert_options_rsk(struct dccp_request_sock
*dreq
, struct sk_buff
*skb
)
623 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
625 if (dccp_feat_insert_opts(NULL
, dreq
, skb
))
628 /* Obtain RTT sample from Response/Ack exchange (used by TFRC). */
629 if (dccp_insert_option_timestamp(skb
))
632 if (dreq
->dreq_timestamp_echo
!= 0 &&
633 dccp_insert_option_timestamp_echo(NULL
, dreq
, skb
))
636 dccp_insert_option_padding(skb
);