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 */
127 goto out_invalid_option
;
128 rc
= dccp_feat_parse_options(sk
, dreq
, mandatory
, opt
,
129 *value
, value
+ 1, len
- 1);
131 goto out_featneg_failed
;
133 case DCCPO_TIMESTAMP
:
135 goto out_invalid_option
;
137 * RFC 4340 13.1: "The precise time corresponding to
138 * Timestamp Value zero is not specified". We use
139 * zero to indicate absence of a meaningful timestamp.
141 opt_val
= get_unaligned((__be32
*)value
);
142 if (unlikely(opt_val
== 0)) {
143 DCCP_WARN("Timestamp with zero value\n");
148 dreq
->dreq_timestamp_echo
= ntohl(opt_val
);
149 dreq
->dreq_timestamp_time
= dccp_timestamp();
151 opt_recv
->dccpor_timestamp
=
152 dp
->dccps_timestamp_echo
= ntohl(opt_val
);
153 dp
->dccps_timestamp_time
= dccp_timestamp();
155 dccp_pr_debug("%s rx opt: TIMESTAMP=%u, ackno=%llu\n",
156 dccp_role(sk
), ntohl(opt_val
),
158 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
159 /* schedule an Ack in case this sender is quiescent */
160 inet_csk_schedule_ack(sk
);
162 case DCCPO_TIMESTAMP_ECHO
:
163 if (len
!= 4 && len
!= 6 && len
!= 8)
164 goto out_invalid_option
;
166 opt_val
= get_unaligned((__be32
*)value
);
167 opt_recv
->dccpor_timestamp_echo
= ntohl(opt_val
);
169 dccp_pr_debug("%s rx opt: TIMESTAMP_ECHO=%u, len=%d, "
170 "ackno=%llu", dccp_role(sk
),
171 opt_recv
->dccpor_timestamp_echo
,
174 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
178 if (len
== 4) { /* no elapsed time included */
179 dccp_pr_debug_cat("\n");
183 if (len
== 6) { /* 2-byte elapsed time */
184 __be16 opt_val2
= get_unaligned((__be16
*)value
);
185 elapsed_time
= ntohs(opt_val2
);
186 } else { /* 4-byte elapsed time */
187 opt_val
= get_unaligned((__be32
*)value
);
188 elapsed_time
= ntohl(opt_val
);
191 dccp_pr_debug_cat(", ELAPSED_TIME=%u\n", elapsed_time
);
193 /* Give precedence to the biggest ELAPSED_TIME */
194 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
195 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
197 case DCCPO_ELAPSED_TIME
:
198 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 13.2 */
202 __be16 opt_val2
= get_unaligned((__be16
*)value
);
203 elapsed_time
= ntohs(opt_val2
);
204 } else if (len
== 4) {
205 opt_val
= get_unaligned((__be32
*)value
);
206 elapsed_time
= ntohl(opt_val
);
208 goto out_invalid_option
;
211 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
212 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
214 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n",
215 dccp_role(sk
), elapsed_time
);
217 case DCCPO_MIN_RX_CCID_SPECIFIC
... DCCPO_MAX_RX_CCID_SPECIFIC
:
218 if (ccid_hc_rx_parse_options(dp
->dccps_hc_rx_ccid
, sk
,
219 pkt_type
, opt
, value
, len
))
220 goto out_invalid_option
;
222 case DCCPO_ACK_VECTOR_0
:
223 case DCCPO_ACK_VECTOR_1
:
224 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 11.4 */
227 * Ack vectors are processed by the TX CCID if it is
228 * interested. The RX CCID need not parse Ack Vectors,
229 * since it is only interested in clearing old state.
232 case DCCPO_MIN_TX_CCID_SPECIFIC
... DCCPO_MAX_TX_CCID_SPECIFIC
:
233 if (ccid_hc_tx_parse_options(dp
->dccps_hc_tx_ccid
, sk
,
234 pkt_type
, opt
, value
, len
))
235 goto out_invalid_option
;
238 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
239 "implemented, ignoring", sk
, opt
, len
);
243 if (opt
!= DCCPO_MANDATORY
)
247 /* mandatory was the last byte in option list -> reset connection */
249 goto out_invalid_option
;
251 out_nonsensical_length
:
252 /* RFC 4340, 5.8: ignore option and all remaining option space */
256 DCCP_INC_STATS(DCCP_MIB_INVALIDOPT
);
257 rc
= DCCP_RESET_CODE_OPTION_ERROR
;
259 DCCP_WARN("DCCP(%p): Option %d (len=%d) error=%u\n", sk
, opt
, len
, rc
);
260 DCCP_SKB_CB(skb
)->dccpd_reset_code
= rc
;
261 DCCP_SKB_CB(skb
)->dccpd_reset_data
[0] = opt
;
262 DCCP_SKB_CB(skb
)->dccpd_reset_data
[1] = len
> 0 ? value
[0] : 0;
263 DCCP_SKB_CB(skb
)->dccpd_reset_data
[2] = len
> 1 ? value
[1] : 0;
267 EXPORT_SYMBOL_GPL(dccp_parse_options
);
269 void dccp_encode_value_var(const u64 value
, u8
*to
, const u8 len
)
271 if (len
>= DCCP_OPTVAL_MAXLEN
)
272 *to
++ = (value
& 0xFF0000000000ull
) >> 40;
274 *to
++ = (value
& 0xFF00000000ull
) >> 32;
276 *to
++ = (value
& 0xFF000000) >> 24;
278 *to
++ = (value
& 0xFF0000) >> 16;
280 *to
++ = (value
& 0xFF00) >> 8;
282 *to
++ = (value
& 0xFF);
285 static inline u8
dccp_ndp_len(const u64 ndp
)
287 if (likely(ndp
<= 0xFF))
289 return likely(ndp
<= USHRT_MAX
) ? 2 : (ndp
<= UINT_MAX
? 4 : 6);
292 int dccp_insert_option(struct sk_buff
*skb
, const unsigned char option
,
293 const void *value
, const unsigned char len
)
297 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
+ 2 > DCCP_MAX_OPT_LEN
)
300 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
+ 2;
302 to
= skb_push(skb
, len
+ 2);
306 memcpy(to
, value
, len
);
310 EXPORT_SYMBOL_GPL(dccp_insert_option
);
312 static int dccp_insert_option_ndp(struct sock
*sk
, struct sk_buff
*skb
)
314 struct dccp_sock
*dp
= dccp_sk(sk
);
315 u64 ndp
= dp
->dccps_ndp_count
;
317 if (dccp_non_data_packet(skb
))
318 ++dp
->dccps_ndp_count
;
320 dp
->dccps_ndp_count
= 0;
324 const int ndp_len
= dccp_ndp_len(ndp
);
325 const int len
= ndp_len
+ 2;
327 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
330 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
332 ptr
= skb_push(skb
, len
);
333 *ptr
++ = DCCPO_NDP_COUNT
;
335 dccp_encode_value_var(ndp
, ptr
, ndp_len
);
341 static inline int dccp_elapsed_time_len(const u32 elapsed_time
)
343 return elapsed_time
== 0 ? 0 : elapsed_time
<= 0xFFFF ? 2 : 4;
346 static int dccp_insert_option_timestamp(struct sk_buff
*skb
)
348 __be32 now
= htonl(dccp_timestamp());
349 /* yes this will overflow but that is the point as we want a
350 * 10 usec 32 bit timer which mean it wraps every 11.9 hours */
352 return dccp_insert_option(skb
, DCCPO_TIMESTAMP
, &now
, sizeof(now
));
355 static int dccp_insert_option_timestamp_echo(struct dccp_sock
*dp
,
356 struct dccp_request_sock
*dreq
,
361 u32 elapsed_time
, elapsed_time_len
, len
;
364 elapsed_time
= dccp_timestamp() - dreq
->dreq_timestamp_time
;
365 tstamp_echo
= htonl(dreq
->dreq_timestamp_echo
);
366 dreq
->dreq_timestamp_echo
= 0;
368 elapsed_time
= dccp_timestamp() - dp
->dccps_timestamp_time
;
369 tstamp_echo
= htonl(dp
->dccps_timestamp_echo
);
370 dp
->dccps_timestamp_echo
= 0;
373 elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
374 len
= 6 + elapsed_time_len
;
376 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
379 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
381 to
= skb_push(skb
, len
);
382 *to
++ = DCCPO_TIMESTAMP_ECHO
;
385 memcpy(to
, &tstamp_echo
, 4);
388 if (elapsed_time_len
== 2) {
389 const __be16 var16
= htons((u16
)elapsed_time
);
390 memcpy(to
, &var16
, 2);
391 } else if (elapsed_time_len
== 4) {
392 const __be32 var32
= htonl(elapsed_time
);
393 memcpy(to
, &var32
, 4);
399 static int dccp_insert_option_ackvec(struct sock
*sk
, struct sk_buff
*skb
)
401 struct dccp_sock
*dp
= dccp_sk(sk
);
402 struct dccp_ackvec
*av
= dp
->dccps_hc_rx_ackvec
;
403 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
404 const u16 buflen
= dccp_ackvec_buflen(av
);
405 /* Figure out how many options do we need to represent the ackvec */
406 const u8 nr_opts
= DIV_ROUND_UP(buflen
, DCCP_SINGLE_OPT_MAXLEN
);
407 u16 len
= buflen
+ 2 * nr_opts
;
409 const unsigned char *tail
, *from
;
412 if (dcb
->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
) {
413 DCCP_WARN("Lacking space for %u bytes on %s packet\n", len
,
414 dccp_packet_name(dcb
->dccpd_type
));
418 * Since Ack Vectors are variable-length, we can not always predict
419 * their size. To catch exception cases where the space is running out
420 * on the skb, a separate Sync is scheduled to carry the Ack Vector.
422 if (len
> DCCPAV_MIN_OPTLEN
&&
423 len
+ dcb
->dccpd_opt_len
+ skb
->len
> dp
->dccps_mss_cache
) {
424 DCCP_WARN("No space left for Ack Vector (%u) on skb (%u+%u), "
425 "MPS=%u ==> reduce payload size?\n", len
, skb
->len
,
426 dcb
->dccpd_opt_len
, dp
->dccps_mss_cache
);
427 dp
->dccps_sync_scheduled
= 1;
430 dcb
->dccpd_opt_len
+= len
;
432 to
= skb_push(skb
, len
);
434 from
= av
->av_buf
+ av
->av_buf_head
;
435 tail
= av
->av_buf
+ DCCPAV_MAX_ACKVEC_LEN
;
437 for (i
= 0; i
< nr_opts
; ++i
) {
440 if (len
> DCCP_SINGLE_OPT_MAXLEN
)
441 copylen
= DCCP_SINGLE_OPT_MAXLEN
;
444 * RFC 4340, 12.2: Encode the Nonce Echo for this Ack Vector via
445 * its type; ack_nonce is the sum of all individual buf_nonce's.
447 nonce
^= av
->av_buf_nonce
[i
];
449 *to
++ = DCCPO_ACK_VECTOR_0
+ av
->av_buf_nonce
[i
];
452 /* Check if buf_head wraps */
453 if (from
+ copylen
> tail
) {
454 const u16 tailsize
= tail
- from
;
456 memcpy(to
, from
, tailsize
);
463 memcpy(to
, from
, copylen
);
469 * Each sent Ack Vector is recorded in the list, as per A.2 of RFC 4340.
471 if (dccp_ackvec_update_records(av
, dcb
->dccpd_seq
, nonce
))
477 * dccp_insert_option_mandatory - Mandatory option (5.8.2)
478 * Note that since we are using skb_push, this function needs to be called
479 * _after_ inserting the option it is supposed to influence (stack order).
481 int dccp_insert_option_mandatory(struct sk_buff
*skb
)
483 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
>= DCCP_MAX_OPT_LEN
)
486 DCCP_SKB_CB(skb
)->dccpd_opt_len
++;
487 *skb_push(skb
, 1) = DCCPO_MANDATORY
;
492 * dccp_insert_fn_opt - Insert single Feature-Negotiation option into @skb
493 * @type: %DCCPO_CHANGE_L, %DCCPO_CHANGE_R, %DCCPO_CONFIRM_L, %DCCPO_CONFIRM_R
494 * @feat: one out of %dccp_feature_numbers
495 * @val: NN value or SP array (preferred element first) to copy
496 * @len: true length of @val in bytes (excluding first element repetition)
497 * @repeat_first: whether to copy the first element of @val twice
499 * The last argument is used to construct Confirm options, where the preferred
500 * value and the preference list appear separately (RFC 4340, 6.3.1). Preference
501 * lists are kept such that the preferred entry is always first, so we only need
502 * to copy twice, and avoid the overhead of cloning into a bigger array.
504 int dccp_insert_fn_opt(struct sk_buff
*skb
, u8 type
, u8 feat
,
505 u8
*val
, u8 len
, bool repeat_first
)
509 /* take the `Feature' field and possible repetition into account */
510 if (len
> (DCCP_SINGLE_OPT_MAXLEN
- 2)) {
511 DCCP_WARN("length %u for feature %u too large\n", len
, feat
);
515 if (unlikely(val
== NULL
|| len
== 0))
516 len
= repeat_first
= false;
517 tot_len
= 3 + repeat_first
+ len
;
519 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ tot_len
> DCCP_MAX_OPT_LEN
) {
520 DCCP_WARN("packet too small for feature %d option!\n", feat
);
523 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= tot_len
;
525 to
= skb_push(skb
, tot_len
);
533 memcpy(to
, val
, len
);
537 /* The length of all options needs to be a multiple of 4 (5.8) */
538 static void dccp_insert_option_padding(struct sk_buff
*skb
)
540 int padding
= DCCP_SKB_CB(skb
)->dccpd_opt_len
% 4;
543 padding
= 4 - padding
;
544 memset(skb_push(skb
, padding
), 0, padding
);
545 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= padding
;
549 int dccp_insert_options(struct sock
*sk
, struct sk_buff
*skb
)
551 struct dccp_sock
*dp
= dccp_sk(sk
);
553 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
555 if (dp
->dccps_send_ndp_count
&& dccp_insert_option_ndp(sk
, skb
))
558 if (DCCP_SKB_CB(skb
)->dccpd_type
!= DCCP_PKT_DATA
) {
560 /* Feature Negotiation */
561 if (dccp_feat_insert_opts(dp
, NULL
, skb
))
564 if (DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_REQUEST
) {
566 * Obtain RTT sample from Request/Response exchange.
567 * This is currently used for TFRC initialisation.
569 if (dccp_insert_option_timestamp(skb
))
572 } else if (dccp_ackvec_pending(sk
) &&
573 dccp_insert_option_ackvec(sk
, skb
)) {
578 if (dp
->dccps_hc_rx_insert_options
) {
579 if (ccid_hc_rx_insert_options(dp
->dccps_hc_rx_ccid
, sk
, skb
))
581 dp
->dccps_hc_rx_insert_options
= 0;
584 if (dp
->dccps_timestamp_echo
!= 0 &&
585 dccp_insert_option_timestamp_echo(dp
, NULL
, skb
))
588 dccp_insert_option_padding(skb
);
592 int dccp_insert_options_rsk(struct dccp_request_sock
*dreq
, struct sk_buff
*skb
)
594 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
596 if (dccp_feat_insert_opts(NULL
, dreq
, skb
))
599 /* Obtain RTT sample from Response/Ack exchange (used by TFRC). */
600 if (dccp_insert_option_timestamp(skb
))
603 if (dreq
->dreq_timestamp_echo
!= 0 &&
604 dccp_insert_option_timestamp_echo(NULL
, dreq
, skb
))
607 dccp_insert_option_padding(skb
);