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
46 * @skb: frame to parse
48 int dccp_parse_options(struct sock
*sk
, struct dccp_request_sock
*dreq
,
51 struct dccp_sock
*dp
= dccp_sk(sk
);
52 const struct dccp_hdr
*dh
= dccp_hdr(skb
);
53 const u8 pkt_type
= DCCP_SKB_CB(skb
)->dccpd_type
;
54 unsigned char *options
= (unsigned char *)dh
+ dccp_hdr_len(skb
);
55 unsigned char *opt_ptr
= options
;
56 const unsigned char *opt_end
= (unsigned char *)dh
+
58 struct dccp_options_received
*opt_recv
= &dp
->dccps_options_received
;
59 unsigned char opt
, len
;
66 memset(opt_recv
, 0, sizeof(*opt_recv
));
69 while (opt_ptr
!= opt_end
) {
74 /* Check if this isn't a single byte option */
75 if (opt
> DCCPO_MAX_RESERVED
) {
76 if (opt_ptr
== opt_end
)
77 goto out_nonsensical_length
;
81 goto out_nonsensical_length
;
83 * Remove the type and len fields, leaving
90 if (opt_ptr
> opt_end
)
91 goto out_nonsensical_length
;
95 * CCID-specific options are ignored during connection setup, as
96 * negotiation may still be in progress (see RFC 4340, 10.3).
97 * The same applies to Ack Vectors, as these depend on the CCID.
99 if (dreq
!= NULL
&& (opt
>= DCCPO_MIN_RX_CCID_SPECIFIC
||
100 opt
== DCCPO_ACK_VECTOR_0
|| opt
== DCCPO_ACK_VECTOR_1
))
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 opt: NDP count=%llu\n", dccp_role(sk
),
118 (unsigned long long)opt_recv
->dccpor_ndp
);
120 case DCCPO_CHANGE_L
... DCCPO_CONFIRM_R
:
121 if (pkt_type
== DCCP_PKT_DATA
) /* RFC 4340, 6 */
124 goto out_invalid_option
;
125 rc
= dccp_feat_parse_options(sk
, dreq
, mandatory
, opt
,
126 *value
, value
+ 1, len
- 1);
128 goto out_featneg_failed
;
130 case DCCPO_TIMESTAMP
:
132 goto out_invalid_option
;
134 * RFC 4340 13.1: "The precise time corresponding to
135 * Timestamp Value zero is not specified". We use
136 * zero to indicate absence of a meaningful timestamp.
138 opt_val
= get_unaligned((__be32
*)value
);
139 if (unlikely(opt_val
== 0)) {
140 DCCP_WARN("Timestamp with zero value\n");
145 dreq
->dreq_timestamp_echo
= ntohl(opt_val
);
146 dreq
->dreq_timestamp_time
= dccp_timestamp();
148 opt_recv
->dccpor_timestamp
=
149 dp
->dccps_timestamp_echo
= ntohl(opt_val
);
150 dp
->dccps_timestamp_time
= dccp_timestamp();
152 dccp_pr_debug("%s rx opt: TIMESTAMP=%u, ackno=%llu\n",
153 dccp_role(sk
), ntohl(opt_val
),
155 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
156 /* schedule an Ack in case this sender is quiescent */
157 inet_csk_schedule_ack(sk
);
159 case DCCPO_TIMESTAMP_ECHO
:
160 if (len
!= 4 && len
!= 6 && len
!= 8)
161 goto out_invalid_option
;
163 opt_val
= get_unaligned((__be32
*)value
);
164 opt_recv
->dccpor_timestamp_echo
= ntohl(opt_val
);
166 dccp_pr_debug("%s rx opt: TIMESTAMP_ECHO=%u, len=%d, "
167 "ackno=%llu", dccp_role(sk
),
168 opt_recv
->dccpor_timestamp_echo
,
171 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
175 if (len
== 4) { /* no elapsed time included */
176 dccp_pr_debug_cat("\n");
180 if (len
== 6) { /* 2-byte elapsed time */
181 __be16 opt_val2
= get_unaligned((__be16
*)value
);
182 elapsed_time
= ntohs(opt_val2
);
183 } else { /* 4-byte elapsed time */
184 opt_val
= get_unaligned((__be32
*)value
);
185 elapsed_time
= ntohl(opt_val
);
188 dccp_pr_debug_cat(", ELAPSED_TIME=%u\n", elapsed_time
);
190 /* Give precedence to the biggest ELAPSED_TIME */
191 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
192 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
194 case DCCPO_ELAPSED_TIME
:
195 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 13.2 */
199 __be16 opt_val2
= get_unaligned((__be16
*)value
);
200 elapsed_time
= ntohs(opt_val2
);
201 } else if (len
== 4) {
202 opt_val
= get_unaligned((__be32
*)value
);
203 elapsed_time
= ntohl(opt_val
);
205 goto out_invalid_option
;
208 if (elapsed_time
> opt_recv
->dccpor_elapsed_time
)
209 opt_recv
->dccpor_elapsed_time
= elapsed_time
;
211 dccp_pr_debug("%s rx opt: ELAPSED_TIME=%d\n",
212 dccp_role(sk
), elapsed_time
);
214 case DCCPO_MIN_RX_CCID_SPECIFIC
... DCCPO_MAX_RX_CCID_SPECIFIC
:
215 if (ccid_hc_rx_parse_options(dp
->dccps_hc_rx_ccid
, sk
,
216 pkt_type
, opt
, value
, len
))
217 goto out_invalid_option
;
219 case DCCPO_ACK_VECTOR_0
:
220 case DCCPO_ACK_VECTOR_1
:
221 if (dccp_packet_without_ack(skb
)) /* RFC 4340, 11.4 */
224 * Ack vectors are processed by the TX CCID if it is
225 * interested. The RX CCID need not parse Ack Vectors,
226 * since it is only interested in clearing old state.
229 case DCCPO_MIN_TX_CCID_SPECIFIC
... DCCPO_MAX_TX_CCID_SPECIFIC
:
230 if (ccid_hc_tx_parse_options(dp
->dccps_hc_tx_ccid
, sk
,
231 pkt_type
, opt
, value
, len
))
232 goto out_invalid_option
;
235 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
236 "implemented, ignoring", sk
, opt
, len
);
240 if (opt
!= DCCPO_MANDATORY
)
244 /* mandatory was the last byte in option list -> reset connection */
246 goto out_invalid_option
;
248 out_nonsensical_length
:
249 /* RFC 4340, 5.8: ignore option and all remaining option space */
253 DCCP_INC_STATS(DCCP_MIB_INVALIDOPT
);
254 rc
= DCCP_RESET_CODE_OPTION_ERROR
;
256 DCCP_WARN("DCCP(%p): Option %d (len=%d) error=%u\n", sk
, opt
, len
, rc
);
257 DCCP_SKB_CB(skb
)->dccpd_reset_code
= rc
;
258 DCCP_SKB_CB(skb
)->dccpd_reset_data
[0] = opt
;
259 DCCP_SKB_CB(skb
)->dccpd_reset_data
[1] = len
> 0 ? value
[0] : 0;
260 DCCP_SKB_CB(skb
)->dccpd_reset_data
[2] = len
> 1 ? value
[1] : 0;
264 EXPORT_SYMBOL_GPL(dccp_parse_options
);
266 void dccp_encode_value_var(const u64 value
, u8
*to
, const u8 len
)
268 if (len
>= DCCP_OPTVAL_MAXLEN
)
269 *to
++ = (value
& 0xFF0000000000ull
) >> 40;
271 *to
++ = (value
& 0xFF00000000ull
) >> 32;
273 *to
++ = (value
& 0xFF000000) >> 24;
275 *to
++ = (value
& 0xFF0000) >> 16;
277 *to
++ = (value
& 0xFF00) >> 8;
279 *to
++ = (value
& 0xFF);
282 static inline u8
dccp_ndp_len(const u64 ndp
)
284 if (likely(ndp
<= 0xFF))
286 return likely(ndp
<= USHRT_MAX
) ? 2 : (ndp
<= UINT_MAX
? 4 : 6);
289 int dccp_insert_option(struct sk_buff
*skb
, const unsigned char option
,
290 const void *value
, const unsigned char len
)
294 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
+ 2 > DCCP_MAX_OPT_LEN
)
297 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
+ 2;
299 to
= skb_push(skb
, len
+ 2);
303 memcpy(to
, value
, len
);
307 EXPORT_SYMBOL_GPL(dccp_insert_option
);
309 static int dccp_insert_option_ndp(struct sock
*sk
, struct sk_buff
*skb
)
311 struct dccp_sock
*dp
= dccp_sk(sk
);
312 u64 ndp
= dp
->dccps_ndp_count
;
314 if (dccp_non_data_packet(skb
))
315 ++dp
->dccps_ndp_count
;
317 dp
->dccps_ndp_count
= 0;
321 const int ndp_len
= dccp_ndp_len(ndp
);
322 const int len
= ndp_len
+ 2;
324 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
327 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
329 ptr
= skb_push(skb
, len
);
330 *ptr
++ = DCCPO_NDP_COUNT
;
332 dccp_encode_value_var(ndp
, ptr
, ndp_len
);
338 static inline int dccp_elapsed_time_len(const u32 elapsed_time
)
340 return elapsed_time
== 0 ? 0 : elapsed_time
<= 0xFFFF ? 2 : 4;
343 static int dccp_insert_option_timestamp(struct sk_buff
*skb
)
345 __be32 now
= htonl(dccp_timestamp());
346 /* yes this will overflow but that is the point as we want a
347 * 10 usec 32 bit timer which mean it wraps every 11.9 hours */
349 return dccp_insert_option(skb
, DCCPO_TIMESTAMP
, &now
, sizeof(now
));
352 static int dccp_insert_option_timestamp_echo(struct dccp_sock
*dp
,
353 struct dccp_request_sock
*dreq
,
358 u32 elapsed_time
, elapsed_time_len
, len
;
361 elapsed_time
= dccp_timestamp() - dreq
->dreq_timestamp_time
;
362 tstamp_echo
= htonl(dreq
->dreq_timestamp_echo
);
363 dreq
->dreq_timestamp_echo
= 0;
365 elapsed_time
= dccp_timestamp() - dp
->dccps_timestamp_time
;
366 tstamp_echo
= htonl(dp
->dccps_timestamp_echo
);
367 dp
->dccps_timestamp_echo
= 0;
370 elapsed_time_len
= dccp_elapsed_time_len(elapsed_time
);
371 len
= 6 + elapsed_time_len
;
373 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
)
376 DCCP_SKB_CB(skb
)->dccpd_opt_len
+= len
;
378 to
= skb_push(skb
, len
);
379 *to
++ = DCCPO_TIMESTAMP_ECHO
;
382 memcpy(to
, &tstamp_echo
, 4);
385 if (elapsed_time_len
== 2) {
386 const __be16 var16
= htons((u16
)elapsed_time
);
387 memcpy(to
, &var16
, 2);
388 } else if (elapsed_time_len
== 4) {
389 const __be32 var32
= htonl(elapsed_time
);
390 memcpy(to
, &var32
, 4);
396 static int dccp_insert_option_ackvec(struct sock
*sk
, struct sk_buff
*skb
)
398 struct dccp_sock
*dp
= dccp_sk(sk
);
399 struct dccp_ackvec
*av
= dp
->dccps_hc_rx_ackvec
;
400 struct dccp_skb_cb
*dcb
= DCCP_SKB_CB(skb
);
401 const u16 buflen
= dccp_ackvec_buflen(av
);
402 /* Figure out how many options do we need to represent the ackvec */
403 const u8 nr_opts
= DIV_ROUND_UP(buflen
, DCCP_SINGLE_OPT_MAXLEN
);
404 u16 len
= buflen
+ 2 * nr_opts
;
406 const unsigned char *tail
, *from
;
409 if (dcb
->dccpd_opt_len
+ len
> DCCP_MAX_OPT_LEN
) {
410 DCCP_WARN("Lacking space for %u bytes on %s packet\n", len
,
411 dccp_packet_name(dcb
->dccpd_type
));
415 * Since Ack Vectors are variable-length, we can not always predict
416 * their size. To catch exception cases where the space is running out
417 * on the skb, a separate Sync is scheduled to carry the Ack Vector.
419 if (len
> DCCPAV_MIN_OPTLEN
&&
420 len
+ dcb
->dccpd_opt_len
+ skb
->len
> dp
->dccps_mss_cache
) {
421 DCCP_WARN("No space left for Ack Vector (%u) on skb (%u+%u), "
422 "MPS=%u ==> reduce payload size?\n", len
, skb
->len
,
423 dcb
->dccpd_opt_len
, dp
->dccps_mss_cache
);
424 dp
->dccps_sync_scheduled
= 1;
427 dcb
->dccpd_opt_len
+= len
;
429 to
= skb_push(skb
, len
);
431 from
= av
->av_buf
+ av
->av_buf_head
;
432 tail
= av
->av_buf
+ DCCPAV_MAX_ACKVEC_LEN
;
434 for (i
= 0; i
< nr_opts
; ++i
) {
437 if (len
> DCCP_SINGLE_OPT_MAXLEN
)
438 copylen
= DCCP_SINGLE_OPT_MAXLEN
;
441 * RFC 4340, 12.2: Encode the Nonce Echo for this Ack Vector via
442 * its type; ack_nonce is the sum of all individual buf_nonce's.
444 nonce
^= av
->av_buf_nonce
[i
];
446 *to
++ = DCCPO_ACK_VECTOR_0
+ av
->av_buf_nonce
[i
];
449 /* Check if buf_head wraps */
450 if (from
+ copylen
> tail
) {
451 const u16 tailsize
= tail
- from
;
453 memcpy(to
, from
, tailsize
);
460 memcpy(to
, from
, copylen
);
466 * Each sent Ack Vector is recorded in the list, as per A.2 of RFC 4340.
468 if (dccp_ackvec_update_records(av
, dcb
->dccpd_seq
, nonce
))
474 * dccp_insert_option_mandatory - Mandatory option (5.8.2)
475 * @skb: frame into which to insert option
477 * Note that since we are using skb_push, this function needs to be called
478 * _after_ inserting the option it is supposed to influence (stack order).
480 int dccp_insert_option_mandatory(struct sk_buff
*skb
)
482 if (DCCP_SKB_CB(skb
)->dccpd_opt_len
>= DCCP_MAX_OPT_LEN
)
485 DCCP_SKB_CB(skb
)->dccpd_opt_len
++;
486 *(u8
*)skb_push(skb
, 1) = DCCPO_MANDATORY
;
491 * dccp_insert_fn_opt - Insert single Feature-Negotiation option into @skb
492 * @skb: frame to insert feature negotiation option into
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
);