aufs: ioctl, from aufs2.2-3.0
[zen-stable.git] / net / dccp / options.c
blob68fa6b7a3e016bbd7d9a3bf173040b39bc86eb20
1 /*
2 * net/dccp/options.c
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>
21 #include "ackvec.h"
22 #include "ccid.h"
23 #include "dccp.h"
24 #include "feat.h"
26 u64 dccp_decode_value_var(const u8 *bf, const u8 len)
28 u64 value = 0;
30 if (len >= DCCP_OPTVAL_MAXLEN)
31 value += ((u64)*bf++) << 40;
32 if (len > 4)
33 value += ((u64)*bf++) << 32;
34 if (len > 3)
35 value += ((u64)*bf++) << 24;
36 if (len > 2)
37 value += ((u64)*bf++) << 16;
38 if (len > 1)
39 value += ((u64)*bf++) << 8;
40 if (len > 0)
41 value += *bf;
43 return value;
46 /**
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,
52 struct sk_buff *skb)
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 +
60 (dh->dccph_doff * 4);
61 struct dccp_options_received *opt_recv = &dp->dccps_options_received;
62 unsigned char opt, len;
63 unsigned char *uninitialized_var(value);
64 u32 elapsed_time;
65 __be32 opt_val;
66 int rc;
67 int mandatory = 0;
69 memset(opt_recv, 0, sizeof(*opt_recv));
71 opt = len = 0;
72 while (opt_ptr != opt_end) {
73 opt = *opt_ptr++;
74 len = 0;
75 value = NULL;
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;
82 len = *opt_ptr++;
83 if (len < 2)
84 goto out_nonsensical_length;
86 * Remove the type and len fields, leaving
87 * just the value size
89 len -= 2;
90 value = opt_ptr;
91 opt_ptr += len;
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))
104 goto ignore_option;
106 switch (opt) {
107 case DCCPO_PADDING:
108 break;
109 case DCCPO_MANDATORY:
110 if (mandatory)
111 goto out_invalid_option;
112 if (pkt_type != DCCP_PKT_DATA)
113 mandatory = 1;
114 break;
115 case DCCPO_NDP_COUNT:
116 if (len > 6)
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);
122 break;
123 case DCCPO_CHANGE_L ... DCCPO_CONFIRM_R:
124 if (pkt_type == DCCP_PKT_DATA) /* RFC 4340, 6 */
125 break;
126 if (len == 0)
127 goto out_invalid_option;
128 rc = dccp_feat_parse_options(sk, dreq, mandatory, opt,
129 *value, value + 1, len - 1);
130 if (rc)
131 goto out_featneg_failed;
132 break;
133 case DCCPO_TIMESTAMP:
134 if (len != 4)
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");
144 break;
147 if (dreq != NULL) {
148 dreq->dreq_timestamp_echo = ntohl(opt_val);
149 dreq->dreq_timestamp_time = dccp_timestamp();
150 } else {
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),
157 (unsigned long long)
158 DCCP_SKB_CB(skb)->dccpd_ack_seq);
159 /* schedule an Ack in case this sender is quiescent */
160 inet_csk_schedule_ack(sk);
161 break;
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,
172 len + 2,
173 (unsigned long long)
174 DCCP_SKB_CB(skb)->dccpd_ack_seq);
176 value += 4;
178 if (len == 4) { /* no elapsed time included */
179 dccp_pr_debug_cat("\n");
180 break;
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;
196 break;
197 case DCCPO_ELAPSED_TIME:
198 if (dccp_packet_without_ack(skb)) /* RFC 4340, 13.2 */
199 break;
201 if (len == 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);
207 } else {
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);
216 break;
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;
221 break;
222 case DCCPO_ACK_VECTOR_0:
223 case DCCPO_ACK_VECTOR_1:
224 if (dccp_packet_without_ack(skb)) /* RFC 4340, 11.4 */
225 break;
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.
230 * Fall through.
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;
236 break;
237 default:
238 DCCP_CRIT("DCCP(%p): option %d(len=%d) not "
239 "implemented, ignoring", sk, opt, len);
240 break;
242 ignore_option:
243 if (opt != DCCPO_MANDATORY)
244 mandatory = 0;
247 /* mandatory was the last byte in option list -> reset connection */
248 if (mandatory)
249 goto out_invalid_option;
251 out_nonsensical_length:
252 /* RFC 4340, 5.8: ignore option and all remaining option space */
253 return 0;
255 out_invalid_option:
256 DCCP_INC_STATS_BH(DCCP_MIB_INVALIDOPT);
257 rc = DCCP_RESET_CODE_OPTION_ERROR;
258 out_featneg_failed:
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;
264 return -1;
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;
273 if (len > 4)
274 *to++ = (value & 0xFF00000000ull) >> 32;
275 if (len > 3)
276 *to++ = (value & 0xFF000000) >> 24;
277 if (len > 2)
278 *to++ = (value & 0xFF0000) >> 16;
279 if (len > 1)
280 *to++ = (value & 0xFF00) >> 8;
281 if (len > 0)
282 *to++ = (value & 0xFF);
285 static inline u8 dccp_ndp_len(const u64 ndp)
287 if (likely(ndp <= 0xFF))
288 return 1;
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)
295 unsigned char *to;
297 if (DCCP_SKB_CB(skb)->dccpd_opt_len + len + 2 > DCCP_MAX_OPT_LEN)
298 return -1;
300 DCCP_SKB_CB(skb)->dccpd_opt_len += len + 2;
302 to = skb_push(skb, len + 2);
303 *to++ = option;
304 *to++ = len + 2;
306 memcpy(to, value, len);
307 return 0;
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;
319 else
320 dp->dccps_ndp_count = 0;
322 if (ndp > 0) {
323 unsigned char *ptr;
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)
328 return -1;
330 DCCP_SKB_CB(skb)->dccpd_opt_len += len;
332 ptr = skb_push(skb, len);
333 *ptr++ = DCCPO_NDP_COUNT;
334 *ptr++ = len;
335 dccp_encode_value_var(ndp, ptr, ndp_len);
338 return 0;
341 static inline int dccp_elapsed_time_len(const u32 elapsed_time)
343 return elapsed_time == 0 ? 0 : elapsed_time <= 0xFFFF ? 2 : 4;
346 /* FIXME: This function is currently not used anywhere */
347 int dccp_insert_option_elapsed_time(struct sk_buff *skb, u32 elapsed_time)
349 const int elapsed_time_len = dccp_elapsed_time_len(elapsed_time);
350 const int len = 2 + elapsed_time_len;
351 unsigned char *to;
353 if (elapsed_time_len == 0)
354 return 0;
356 if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
357 return -1;
359 DCCP_SKB_CB(skb)->dccpd_opt_len += len;
361 to = skb_push(skb, len);
362 *to++ = DCCPO_ELAPSED_TIME;
363 *to++ = len;
365 if (elapsed_time_len == 2) {
366 const __be16 var16 = htons((u16)elapsed_time);
367 memcpy(to, &var16, 2);
368 } else {
369 const __be32 var32 = htonl(elapsed_time);
370 memcpy(to, &var32, 4);
373 return 0;
376 EXPORT_SYMBOL_GPL(dccp_insert_option_elapsed_time);
378 static int dccp_insert_option_timestamp(struct sk_buff *skb)
380 __be32 now = htonl(dccp_timestamp());
381 /* yes this will overflow but that is the point as we want a
382 * 10 usec 32 bit timer which mean it wraps every 11.9 hours */
384 return dccp_insert_option(skb, DCCPO_TIMESTAMP, &now, sizeof(now));
387 static int dccp_insert_option_timestamp_echo(struct dccp_sock *dp,
388 struct dccp_request_sock *dreq,
389 struct sk_buff *skb)
391 __be32 tstamp_echo;
392 unsigned char *to;
393 u32 elapsed_time, elapsed_time_len, len;
395 if (dreq != NULL) {
396 elapsed_time = dccp_timestamp() - dreq->dreq_timestamp_time;
397 tstamp_echo = htonl(dreq->dreq_timestamp_echo);
398 dreq->dreq_timestamp_echo = 0;
399 } else {
400 elapsed_time = dccp_timestamp() - dp->dccps_timestamp_time;
401 tstamp_echo = htonl(dp->dccps_timestamp_echo);
402 dp->dccps_timestamp_echo = 0;
405 elapsed_time_len = dccp_elapsed_time_len(elapsed_time);
406 len = 6 + elapsed_time_len;
408 if (DCCP_SKB_CB(skb)->dccpd_opt_len + len > DCCP_MAX_OPT_LEN)
409 return -1;
411 DCCP_SKB_CB(skb)->dccpd_opt_len += len;
413 to = skb_push(skb, len);
414 *to++ = DCCPO_TIMESTAMP_ECHO;
415 *to++ = len;
417 memcpy(to, &tstamp_echo, 4);
418 to += 4;
420 if (elapsed_time_len == 2) {
421 const __be16 var16 = htons((u16)elapsed_time);
422 memcpy(to, &var16, 2);
423 } else if (elapsed_time_len == 4) {
424 const __be32 var32 = htonl(elapsed_time);
425 memcpy(to, &var32, 4);
428 return 0;
431 static int dccp_insert_option_ackvec(struct sock *sk, struct sk_buff *skb)
433 struct dccp_sock *dp = dccp_sk(sk);
434 struct dccp_ackvec *av = dp->dccps_hc_rx_ackvec;
435 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
436 const u16 buflen = dccp_ackvec_buflen(av);
437 /* Figure out how many options do we need to represent the ackvec */
438 const u8 nr_opts = DIV_ROUND_UP(buflen, DCCP_SINGLE_OPT_MAXLEN);
439 u16 len = buflen + 2 * nr_opts;
440 u8 i, nonce = 0;
441 const unsigned char *tail, *from;
442 unsigned char *to;
444 if (dcb->dccpd_opt_len + len > DCCP_MAX_OPT_LEN) {
445 DCCP_WARN("Lacking space for %u bytes on %s packet\n", len,
446 dccp_packet_name(dcb->dccpd_type));
447 return -1;
450 * Since Ack Vectors are variable-length, we can not always predict
451 * their size. To catch exception cases where the space is running out
452 * on the skb, a separate Sync is scheduled to carry the Ack Vector.
454 if (len > DCCPAV_MIN_OPTLEN &&
455 len + dcb->dccpd_opt_len + skb->len > dp->dccps_mss_cache) {
456 DCCP_WARN("No space left for Ack Vector (%u) on skb (%u+%u), "
457 "MPS=%u ==> reduce payload size?\n", len, skb->len,
458 dcb->dccpd_opt_len, dp->dccps_mss_cache);
459 dp->dccps_sync_scheduled = 1;
460 return 0;
462 dcb->dccpd_opt_len += len;
464 to = skb_push(skb, len);
465 len = buflen;
466 from = av->av_buf + av->av_buf_head;
467 tail = av->av_buf + DCCPAV_MAX_ACKVEC_LEN;
469 for (i = 0; i < nr_opts; ++i) {
470 int copylen = len;
472 if (len > DCCP_SINGLE_OPT_MAXLEN)
473 copylen = DCCP_SINGLE_OPT_MAXLEN;
476 * RFC 4340, 12.2: Encode the Nonce Echo for this Ack Vector via
477 * its type; ack_nonce is the sum of all individual buf_nonce's.
479 nonce ^= av->av_buf_nonce[i];
481 *to++ = DCCPO_ACK_VECTOR_0 + av->av_buf_nonce[i];
482 *to++ = copylen + 2;
484 /* Check if buf_head wraps */
485 if (from + copylen > tail) {
486 const u16 tailsize = tail - from;
488 memcpy(to, from, tailsize);
489 to += tailsize;
490 len -= tailsize;
491 copylen -= tailsize;
492 from = av->av_buf;
495 memcpy(to, from, copylen);
496 from += copylen;
497 to += copylen;
498 len -= copylen;
501 * Each sent Ack Vector is recorded in the list, as per A.2 of RFC 4340.
503 if (dccp_ackvec_update_records(av, dcb->dccpd_seq, nonce))
504 return -ENOBUFS;
505 return 0;
509 * dccp_insert_option_mandatory - Mandatory option (5.8.2)
510 * Note that since we are using skb_push, this function needs to be called
511 * _after_ inserting the option it is supposed to influence (stack order).
513 int dccp_insert_option_mandatory(struct sk_buff *skb)
515 if (DCCP_SKB_CB(skb)->dccpd_opt_len >= DCCP_MAX_OPT_LEN)
516 return -1;
518 DCCP_SKB_CB(skb)->dccpd_opt_len++;
519 *skb_push(skb, 1) = DCCPO_MANDATORY;
520 return 0;
524 * dccp_insert_fn_opt - Insert single Feature-Negotiation option into @skb
525 * @type: %DCCPO_CHANGE_L, %DCCPO_CHANGE_R, %DCCPO_CONFIRM_L, %DCCPO_CONFIRM_R
526 * @feat: one out of %dccp_feature_numbers
527 * @val: NN value or SP array (preferred element first) to copy
528 * @len: true length of @val in bytes (excluding first element repetition)
529 * @repeat_first: whether to copy the first element of @val twice
530 * The last argument is used to construct Confirm options, where the preferred
531 * value and the preference list appear separately (RFC 4340, 6.3.1). Preference
532 * lists are kept such that the preferred entry is always first, so we only need
533 * to copy twice, and avoid the overhead of cloning into a bigger array.
535 int dccp_insert_fn_opt(struct sk_buff *skb, u8 type, u8 feat,
536 u8 *val, u8 len, bool repeat_first)
538 u8 tot_len, *to;
540 /* take the `Feature' field and possible repetition into account */
541 if (len > (DCCP_SINGLE_OPT_MAXLEN - 2)) {
542 DCCP_WARN("length %u for feature %u too large\n", len, feat);
543 return -1;
546 if (unlikely(val == NULL || len == 0))
547 len = repeat_first = false;
548 tot_len = 3 + repeat_first + len;
550 if (DCCP_SKB_CB(skb)->dccpd_opt_len + tot_len > DCCP_MAX_OPT_LEN) {
551 DCCP_WARN("packet too small for feature %d option!\n", feat);
552 return -1;
554 DCCP_SKB_CB(skb)->dccpd_opt_len += tot_len;
556 to = skb_push(skb, tot_len);
557 *to++ = type;
558 *to++ = tot_len;
559 *to++ = feat;
561 if (repeat_first)
562 *to++ = *val;
563 if (len)
564 memcpy(to, val, len);
565 return 0;
568 /* The length of all options needs to be a multiple of 4 (5.8) */
569 static void dccp_insert_option_padding(struct sk_buff *skb)
571 int padding = DCCP_SKB_CB(skb)->dccpd_opt_len % 4;
573 if (padding != 0) {
574 padding = 4 - padding;
575 memset(skb_push(skb, padding), 0, padding);
576 DCCP_SKB_CB(skb)->dccpd_opt_len += padding;
580 int dccp_insert_options(struct sock *sk, struct sk_buff *skb)
582 struct dccp_sock *dp = dccp_sk(sk);
584 DCCP_SKB_CB(skb)->dccpd_opt_len = 0;
586 if (dp->dccps_send_ndp_count && dccp_insert_option_ndp(sk, skb))
587 return -1;
589 if (DCCP_SKB_CB(skb)->dccpd_type != DCCP_PKT_DATA) {
591 /* Feature Negotiation */
592 if (dccp_feat_insert_opts(dp, NULL, skb))
593 return -1;
595 if (DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_REQUEST) {
597 * Obtain RTT sample from Request/Response exchange.
598 * This is currently used for TFRC initialisation.
600 if (dccp_insert_option_timestamp(skb))
601 return -1;
603 } else if (dccp_ackvec_pending(sk) &&
604 dccp_insert_option_ackvec(sk, skb)) {
605 return -1;
609 if (dp->dccps_hc_rx_insert_options) {
610 if (ccid_hc_rx_insert_options(dp->dccps_hc_rx_ccid, sk, skb))
611 return -1;
612 dp->dccps_hc_rx_insert_options = 0;
615 if (dp->dccps_timestamp_echo != 0 &&
616 dccp_insert_option_timestamp_echo(dp, NULL, skb))
617 return -1;
619 dccp_insert_option_padding(skb);
620 return 0;
623 int dccp_insert_options_rsk(struct dccp_request_sock *dreq, struct sk_buff *skb)
625 DCCP_SKB_CB(skb)->dccpd_opt_len = 0;
627 if (dccp_feat_insert_opts(NULL, dreq, skb))
628 return -1;
630 /* Obtain RTT sample from Response/Ack exchange (used by TFRC). */
631 if (dccp_insert_option_timestamp(skb))
632 return -1;
634 if (dreq->dreq_timestamp_echo != 0 &&
635 dccp_insert_option_timestamp_echo(NULL, dreq, skb))
636 return -1;
638 dccp_insert_option_padding(skb);
639 return 0;