1 // SPDX-License-Identifier: GPL-2.0-or-later
5 * An implementation of the DCCP protocol
6 * Copyright (c) 2005 Aristeu Sergio Rozanski Filho <aris@cathedrallabs.org>
7 * Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
8 * Copyright (c) 2005 Ian McDonald <ian.mcdonald@jandi.co.nz>
10 #include <linux/dccp.h>
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <asm/unaligned.h>
14 #include <linux/kernel.h>
15 #include <linux/skbuff.h>
22 u64
dccp_decode_value_var(const u8
*bf
, const u8 len
)
26 if (len
>= DCCP_OPTVAL_MAXLEN
)
27 value
+= ((u64
)*bf
++) << 40;
29 value
+= ((u64
)*bf
++) << 32;
31 value
+= ((u64
)*bf
++) << 24;
33 value
+= ((u64
)*bf
++) << 16;
35 value
+= ((u64
)*bf
++) << 8;
43 * dccp_parse_options - Parse DCCP options present in @skb
44 * @sk: client|server|listening dccp socket (when @dreq != NULL)
45 * @dreq: request socket to use during connection setup, or NULL
47 int dccp_parse_options(struct sock
*sk
, struct dccp_request_sock
*dreq
,
50 struct dccp_sock
*dp
= dccp_sk(sk
);
51 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
52 const u8 pkt_type
= DCCP_SKB_CB(skb
)->dccpd_type
;
53 unsigned char *options
= (unsigned char *)dh
+ dccp_hdr_len(skb
);
54 unsigned char *opt_ptr
= options
;
55 const unsigned char *opt_end
= (unsigned char *)dh
+
57 struct dccp_options_received
*opt_recv
= &dp
->dccps_options_received
;
58 unsigned char opt
, len
;
59 unsigned char *uninitialized_var(value
);
65 memset(opt_recv
, 0, sizeof(*opt_recv
));
68 while (opt_ptr
!= opt_end
) {
73 /* Check if this isn't a single byte option */
74 if (opt
> DCCPO_MAX_RESERVED
) {
75 if (opt_ptr
== opt_end
)
76 goto out_nonsensical_length
;
80 goto out_nonsensical_length
;
82 * Remove the type and len fields, leaving
89 if (opt_ptr
> opt_end
)
90 goto out_nonsensical_length
;
94 * CCID-specific options are ignored during connection setup, as
95 * negotiation may still be in progress (see RFC 4340, 10.3).
96 * The same applies to Ack Vectors, as these depend on the CCID.
98 if (dreq
!= NULL
&& (opt
>= DCCPO_MIN_RX_CCID_SPECIFIC
||
99 opt
== DCCPO_ACK_VECTOR_0
|| opt
== DCCPO_ACK_VECTOR_1
))
105 case DCCPO_MANDATORY
:
107 goto out_invalid_option
;
108 if (pkt_type
!= DCCP_PKT_DATA
)
111 case DCCPO_NDP_COUNT
:
113 goto out_invalid_option
;
115 opt_recv
->dccpor_ndp
= dccp_decode_value_var(value
, len
);
116 dccp_pr_debug("%s opt: NDP count=%llu\n", dccp_role(sk
),
117 (unsigned long long)opt_recv
->dccpor_ndp
);
119 case DCCPO_CHANGE_L
... DCCPO_CONFIRM_R
:
120 if (pkt_type
== DCCP_PKT_DATA
) /* RFC 4340, 6 */
123 goto out_invalid_option
;
124 rc
= dccp_feat_parse_options(sk
, dreq
, mandatory
, opt
,
125 *value
, value
+ 1, len
- 1);
127 goto out_featneg_failed
;
129 case DCCPO_TIMESTAMP
:
131 goto out_invalid_option
;
133 * RFC 4340 13.1: "The precise time corresponding to
134 * Timestamp Value zero is not specified". We use
135 * zero to indicate absence of a meaningful timestamp.
137 opt_val
= get_unaligned((__be32
*)value
);
138 if (unlikely(opt_val
== 0)) {
139 DCCP_WARN("Timestamp with zero value\n");
144 dreq
->dreq_timestamp_echo
= ntohl(opt_val
);
145 dreq
->dreq_timestamp_time
= dccp_timestamp();
147 opt_recv
->dccpor_timestamp
=
148 dp
->dccps_timestamp_echo
= ntohl(opt_val
);
149 dp
->dccps_timestamp_time
= dccp_timestamp();
151 dccp_pr_debug("%s rx opt: TIMESTAMP=%u, ackno=%llu\n",
152 dccp_role(sk
), ntohl(opt_val
),
154 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
155 /* schedule an Ack in case this sender is quiescent */
156 inet_csk_schedule_ack(sk
);
158 case DCCPO_TIMESTAMP_ECHO
:
159 if (len
!= 4 && len
!= 6 && len
!= 8)
160 goto out_invalid_option
;
162 opt_val
= get_unaligned((__be32
*)value
);
163 opt_recv
->dccpor_timestamp_echo
= ntohl(opt_val
);
165 dccp_pr_debug("%s rx opt: TIMESTAMP_ECHO=%u, len=%d, "
166 "ackno=%llu", dccp_role(sk
),
167 opt_recv
->dccpor_timestamp_echo
,
170 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
174 if (len
== 4) { /* no elapsed time included */
175 dccp_pr_debug_cat("\n");
179 if (len
== 6) { /* 2-byte elapsed time */
180 __be16 opt_val2
= get_unaligned((__be16
*)value
);
181 elapsed_time
= ntohs(opt_val2
);
182 } else { /* 4-byte elapsed time */
183 opt_val
= get_unaligned((__be32
*)value
);
184 elapsed_time
= ntohl(opt_val
);
187 dccp_pr_debug_cat(", ELAPSED_TIME=%u\n", elapsed_time
);
189 /* Give precedence to the biggest ELAPSED_TIME */
190 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
191 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
193 case DCCPO_ELAPSED_TIME
:
194 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 13.2 */
198 __be16 opt_val2
= get_unaligned((__be16
*)value
);
199 elapsed_time
= ntohs(opt_val2
);
200 } else if (len
== 4) {
201 opt_val
= get_unaligned((__be32
*)value
);
202 elapsed_time
= ntohl(opt_val
);
204 goto out_invalid_option
;
207 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
208 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
210 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n",
211 dccp_role(sk
), elapsed_time
);
213 case DCCPO_MIN_RX_CCID_SPECIFIC
... DCCPO_MAX_RX_CCID_SPECIFIC
:
214 if (ccid_hc_rx_parse_options(dp
->dccps_hc_rx_ccid
, sk
,
215 pkt_type
, opt
, value
, len
))
216 goto out_invalid_option
;
218 case DCCPO_ACK_VECTOR_0
:
219 case DCCPO_ACK_VECTOR_1
:
220 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 11.4 */
223 * Ack vectors are processed by the TX CCID if it is
224 * interested. The RX CCID need not parse Ack Vectors,
225 * since it is only interested in clearing old state.
228 case DCCPO_MIN_TX_CCID_SPECIFIC
... DCCPO_MAX_TX_CCID_SPECIFIC
:
229 if (ccid_hc_tx_parse_options(dp
->dccps_hc_tx_ccid
, sk
,
230 pkt_type
, opt
, value
, len
))
231 goto out_invalid_option
;
234 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
235 "implemented, ignoring", sk
, opt
, len
);
239 if (opt
!= DCCPO_MANDATORY
)
243 /* mandatory was the last byte in option list -> reset connection */
245 goto out_invalid_option
;
247 out_nonsensical_length
:
248 /* RFC 4340, 5.8: ignore option and all remaining option space */
252 DCCP_INC_STATS(DCCP_MIB_INVALIDOPT
);
253 rc
= DCCP_RESET_CODE_OPTION_ERROR
;
255 DCCP_WARN("DCCP(%p): Option %d (len=%d) error=%u\n", sk
, opt
, len
, rc
);
256 DCCP_SKB_CB(skb
)->dccpd_reset_code
= rc
;
257 DCCP_SKB_CB(skb
)->dccpd_reset_data
[0] = opt
;
258 DCCP_SKB_CB(skb
)->dccpd_reset_data
[1] = len
> 0 ? value
[0] : 0;
259 DCCP_SKB_CB(skb
)->dccpd_reset_data
[2] = len
> 1 ? value
[1] : 0;
263 EXPORT_SYMBOL_GPL(dccp_parse_options
);
265 void dccp_encode_value_var(const u64 value
, u8
*to
, const u8 len
)
267 if (len
>= DCCP_OPTVAL_MAXLEN
)
268 *to
++ = (value
& 0xFF0000000000ull
) >> 40;
270 *to
++ = (value
& 0xFF00000000ull
) >> 32;
272 *to
++ = (value
& 0xFF000000) >> 24;
274 *to
++ = (value
& 0xFF0000) >> 16;
276 *to
++ = (value
& 0xFF00) >> 8;
278 *to
++ = (value
& 0xFF);
281 static inline u8
dccp_ndp_len(const u64 ndp
)
283 if (likely(ndp
<= 0xFF))
285 return likely(ndp
<= USHRT_MAX
) ? 2 : (ndp
<= UINT_MAX
? 4 : 6);
288 int dccp_insert_option(struct sk_buff
*skb
, const unsigned char option
,
289 const void *value
, const unsigned char len
)
293 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
+ 2 > DCCP_MAX_OPT_LEN
)
296 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
+ 2;
298 to
= skb_push(skb
, len
+ 2);
302 memcpy(to
, value
, len
);
306 EXPORT_SYMBOL_GPL(dccp_insert_option
);
308 static int dccp_insert_option_ndp(struct sock
*sk
, struct sk_buff
*skb
)
310 struct dccp_sock
*dp
= dccp_sk(sk
);
311 u64 ndp
= dp
->dccps_ndp_count
;
313 if (dccp_non_data_packet(skb
))
314 ++dp
->dccps_ndp_count
;
316 dp
->dccps_ndp_count
= 0;
320 const int ndp_len
= dccp_ndp_len(ndp
);
321 const int len
= ndp_len
+ 2;
323 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
326 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
328 ptr
= skb_push(skb
, len
);
329 *ptr
++ = DCCPO_NDP_COUNT
;
331 dccp_encode_value_var(ndp
, ptr
, ndp_len
);
337 static inline int dccp_elapsed_time_len(const u32 elapsed_time
)
339 return elapsed_time
== 0 ? 0 : elapsed_time
<= 0xFFFF ? 2 : 4;
342 static int dccp_insert_option_timestamp(struct sk_buff
*skb
)
344 __be32 now
= htonl(dccp_timestamp());
345 /* yes this will overflow but that is the point as we want a
346 * 10 usec 32 bit timer which mean it wraps every 11.9 hours */
348 return dccp_insert_option(skb
, DCCPO_TIMESTAMP
, &now
, sizeof(now
));
351 static int dccp_insert_option_timestamp_echo(struct dccp_sock
*dp
,
352 struct dccp_request_sock
*dreq
,
357 u32 elapsed_time
, elapsed_time_len
, len
;
360 elapsed_time
= dccp_timestamp() - dreq
->dreq_timestamp_time
;
361 tstamp_echo
= htonl(dreq
->dreq_timestamp_echo
);
362 dreq
->dreq_timestamp_echo
= 0;
364 elapsed_time
= dccp_timestamp() - dp
->dccps_timestamp_time
;
365 tstamp_echo
= htonl(dp
->dccps_timestamp_echo
);
366 dp
->dccps_timestamp_echo
= 0;
369 elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
370 len
= 6 + elapsed_time_len
;
372 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
375 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
377 to
= skb_push(skb
, len
);
378 *to
++ = DCCPO_TIMESTAMP_ECHO
;
381 memcpy(to
, &tstamp_echo
, 4);
384 if (elapsed_time_len
== 2) {
385 const __be16 var16
= htons((u16
)elapsed_time
);
386 memcpy(to
, &var16
, 2);
387 } else if (elapsed_time_len
== 4) {
388 const __be32 var32
= htonl(elapsed_time
);
389 memcpy(to
, &var32
, 4);
395 static int dccp_insert_option_ackvec(struct sock
*sk
, struct sk_buff
*skb
)
397 struct dccp_sock
*dp
= dccp_sk(sk
);
398 struct dccp_ackvec
*av
= dp
->dccps_hc_rx_ackvec
;
399 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
400 const u16 buflen
= dccp_ackvec_buflen(av
);
401 /* Figure out how many options do we need to represent the ackvec */
402 const u8 nr_opts
= DIV_ROUND_UP(buflen
, DCCP_SINGLE_OPT_MAXLEN
);
403 u16 len
= buflen
+ 2 * nr_opts
;
405 const unsigned char *tail
, *from
;
408 if (dcb
->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
) {
409 DCCP_WARN("Lacking space for %u bytes on %s packet\n", len
,
410 dccp_packet_name(dcb
->dccpd_type
));
414 * Since Ack Vectors are variable-length, we can not always predict
415 * their size. To catch exception cases where the space is running out
416 * on the skb, a separate Sync is scheduled to carry the Ack Vector.
418 if (len
> DCCPAV_MIN_OPTLEN
&&
419 len
+ dcb
->dccpd_opt_len
+ skb
->len
> dp
->dccps_mss_cache
) {
420 DCCP_WARN("No space left for Ack Vector (%u) on skb (%u+%u), "
421 "MPS=%u ==> reduce payload size?\n", len
, skb
->len
,
422 dcb
->dccpd_opt_len
, dp
->dccps_mss_cache
);
423 dp
->dccps_sync_scheduled
= 1;
426 dcb
->dccpd_opt_len
+= len
;
428 to
= skb_push(skb
, len
);
430 from
= av
->av_buf
+ av
->av_buf_head
;
431 tail
= av
->av_buf
+ DCCPAV_MAX_ACKVEC_LEN
;
433 for (i
= 0; i
< nr_opts
; ++i
) {
436 if (len
> DCCP_SINGLE_OPT_MAXLEN
)
437 copylen
= DCCP_SINGLE_OPT_MAXLEN
;
440 * RFC 4340, 12.2: Encode the Nonce Echo for this Ack Vector via
441 * its type; ack_nonce is the sum of all individual buf_nonce's.
443 nonce
^= av
->av_buf_nonce
[i
];
445 *to
++ = DCCPO_ACK_VECTOR_0
+ av
->av_buf_nonce
[i
];
448 /* Check if buf_head wraps */
449 if (from
+ copylen
> tail
) {
450 const u16 tailsize
= tail
- from
;
452 memcpy(to
, from
, tailsize
);
459 memcpy(to
, from
, copylen
);
465 * Each sent Ack Vector is recorded in the list, as per A.2 of RFC 4340.
467 if (dccp_ackvec_update_records(av
, dcb
->dccpd_seq
, nonce
))
473 * dccp_insert_option_mandatory - Mandatory option (5.8.2)
474 * Note that since we are using skb_push, this function needs to be called
475 * _after_ inserting the option it is supposed to influence (stack order).
477 int dccp_insert_option_mandatory(struct sk_buff
*skb
)
479 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
>= DCCP_MAX_OPT_LEN
)
482 DCCP_SKB_CB(skb
)->dccpd_opt_len
++;
483 *(u8
*)skb_push(skb
, 1) = DCCPO_MANDATORY
;
488 * dccp_insert_fn_opt - Insert single Feature-Negotiation option into @skb
489 * @type: %DCCPO_CHANGE_L, %DCCPO_CHANGE_R, %DCCPO_CONFIRM_L, %DCCPO_CONFIRM_R
490 * @feat: one out of %dccp_feature_numbers
491 * @val: NN value or SP array (preferred element first) to copy
492 * @len: true length of @val in bytes (excluding first element repetition)
493 * @repeat_first: whether to copy the first element of @val twice
495 * The last argument is used to construct Confirm options, where the preferred
496 * value and the preference list appear separately (RFC 4340, 6.3.1). Preference
497 * lists are kept such that the preferred entry is always first, so we only need
498 * to copy twice, and avoid the overhead of cloning into a bigger array.
500 int dccp_insert_fn_opt(struct sk_buff
*skb
, u8 type
, u8 feat
,
501 u8
*val
, u8 len
, bool repeat_first
)
505 /* take the `Feature' field and possible repetition into account */
506 if (len
> (DCCP_SINGLE_OPT_MAXLEN
- 2)) {
507 DCCP_WARN("length %u for feature %u too large\n", len
, feat
);
511 if (unlikely(val
== NULL
|| len
== 0))
512 len
= repeat_first
= false;
513 tot_len
= 3 + repeat_first
+ len
;
515 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ tot_len
> DCCP_MAX_OPT_LEN
) {
516 DCCP_WARN("packet too small for feature %d option!\n", feat
);
519 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= tot_len
;
521 to
= skb_push(skb
, tot_len
);
529 memcpy(to
, val
, len
);
533 /* The length of all options needs to be a multiple of 4 (5.8) */
534 static void dccp_insert_option_padding(struct sk_buff
*skb
)
536 int padding
= DCCP_SKB_CB(skb
)->dccpd_opt_len
% 4;
539 padding
= 4 - padding
;
540 memset(skb_push(skb
, padding
), 0, padding
);
541 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= padding
;
545 int dccp_insert_options(struct sock
*sk
, struct sk_buff
*skb
)
547 struct dccp_sock
*dp
= dccp_sk(sk
);
549 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
551 if (dp
->dccps_send_ndp_count
&& dccp_insert_option_ndp(sk
, skb
))
554 if (DCCP_SKB_CB(skb
)->dccpd_type
!= DCCP_PKT_DATA
) {
556 /* Feature Negotiation */
557 if (dccp_feat_insert_opts(dp
, NULL
, skb
))
560 if (DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_REQUEST
) {
562 * Obtain RTT sample from Request/Response exchange.
563 * This is currently used for TFRC initialisation.
565 if (dccp_insert_option_timestamp(skb
))
568 } else if (dccp_ackvec_pending(sk
) &&
569 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 if (dp
->dccps_timestamp_echo
!= 0 &&
581 dccp_insert_option_timestamp_echo(dp
, NULL
, skb
))
584 dccp_insert_option_padding(skb
);
588 int dccp_insert_options_rsk(struct dccp_request_sock
*dreq
, struct sk_buff
*skb
)
590 DCCP_SKB_CB(skb
)->dccpd_opt_len
= 0;
592 if (dccp_feat_insert_opts(NULL
, dreq
, skb
))
595 /* Obtain RTT sample from Response/Ack exchange (used by TFRC). */
596 if (dccp_insert_option_timestamp(skb
))
599 if (dreq
->dreq_timestamp_echo
!= 0 &&
600 dccp_insert_option_timestamp_echo(NULL
, dreq
, skb
))
603 dccp_insert_option_padding(skb
);