ccid3.c 36 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286
  1. /*
  2. * net/dccp/ccids/ccid3.c
  3. *
  4. * Copyright (c) 2005 The University of Waikato, Hamilton, New Zealand.
  5. * Copyright (c) 2005-6 Ian McDonald <ian.mcdonald@jandi.co.nz>
  6. *
  7. * An implementation of the DCCP protocol
  8. *
  9. * This code has been developed by the University of Waikato WAND
  10. * research group. For further information please see http://www.wand.net.nz/
  11. *
  12. * This code also uses code from Lulea University, rereleased as GPL by its
  13. * authors:
  14. * Copyright (c) 2003 Nils-Erik Mattsson, Joacim Haggmark, Magnus Erixzon
  15. *
  16. * Changes to meet Linux coding standards, to make it meet latest ccid3 draft
  17. * and to make it work as a loadable module in the DCCP stack written by
  18. * Arnaldo Carvalho de Melo <acme@conectiva.com.br>.
  19. *
  20. * Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@conectiva.com.br>
  21. *
  22. * This program is free software; you can redistribute it and/or modify
  23. * it under the terms of the GNU General Public License as published by
  24. * the Free Software Foundation; either version 2 of the License, or
  25. * (at your option) any later version.
  26. *
  27. * This program is distributed in the hope that it will be useful,
  28. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  29. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  30. * GNU General Public License for more details.
  31. *
  32. * You should have received a copy of the GNU General Public License
  33. * along with this program; if not, write to the Free Software
  34. * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  35. */
  36. #include "../ccid.h"
  37. #include "../dccp.h"
  38. #include "lib/packet_history.h"
  39. #include "lib/loss_interval.h"
  40. #include "lib/tfrc.h"
  41. #include "ccid3.h"
  42. #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
  43. static int ccid3_debug;
  44. #define ccid3_pr_debug(format, a...) DCCP_PR_DEBUG(ccid3_debug, format, ##a)
  45. #else
  46. #define ccid3_pr_debug(format, a...)
  47. #endif
  48. static struct dccp_tx_hist *ccid3_tx_hist;
  49. static struct dccp_rx_hist *ccid3_rx_hist;
  50. static struct dccp_li_hist *ccid3_li_hist;
  51. /*
  52. * Transmitter Half-Connection Routines
  53. */
  54. #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
  55. static const char *ccid3_tx_state_name(enum ccid3_hc_tx_states state)
  56. {
  57. static char *ccid3_state_names[] = {
  58. [TFRC_SSTATE_NO_SENT] = "NO_SENT",
  59. [TFRC_SSTATE_NO_FBACK] = "NO_FBACK",
  60. [TFRC_SSTATE_FBACK] = "FBACK",
  61. [TFRC_SSTATE_TERM] = "TERM",
  62. };
  63. return ccid3_state_names[state];
  64. }
  65. #endif
  66. static void ccid3_hc_tx_set_state(struct sock *sk,
  67. enum ccid3_hc_tx_states state)
  68. {
  69. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  70. enum ccid3_hc_tx_states oldstate = hctx->ccid3hctx_state;
  71. ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
  72. dccp_role(sk), sk, ccid3_tx_state_name(oldstate),
  73. ccid3_tx_state_name(state));
  74. WARN_ON(state == oldstate);
  75. hctx->ccid3hctx_state = state;
  76. }
  77. /*
  78. * Compute the initial sending rate X_init according to RFC 3390:
  79. * w_init = min(4 * MSS, max(2 * MSS, 4380 bytes))
  80. * X_init = w_init / RTT
  81. * For consistency with other parts of the code, X_init is scaled by 2^6.
  82. */
  83. static inline u64 rfc3390_initial_rate(struct sock *sk)
  84. {
  85. const struct dccp_sock *dp = dccp_sk(sk);
  86. const __u32 w_init = min(4 * dp->dccps_mss_cache,
  87. max(2 * dp->dccps_mss_cache, 4380U));
  88. return scaled_div(w_init << 6, ccid3_hc_tx_sk(sk)->ccid3hctx_rtt);
  89. }
  90. /*
  91. * Recalculate t_ipi and delta (should be called whenever X changes)
  92. */
  93. static inline void ccid3_update_send_interval(struct ccid3_hc_tx_sock *hctx)
  94. {
  95. /* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */
  96. hctx->ccid3hctx_t_ipi = scaled_div32(((u64)hctx->ccid3hctx_s) << 6,
  97. hctx->ccid3hctx_x);
  98. /* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */
  99. hctx->ccid3hctx_delta = min_t(u32, hctx->ccid3hctx_t_ipi / 2,
  100. TFRC_OPSYS_HALF_TIME_GRAN);
  101. ccid3_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n",
  102. hctx->ccid3hctx_t_ipi, hctx->ccid3hctx_delta,
  103. hctx->ccid3hctx_s, (unsigned)(hctx->ccid3hctx_x >> 6));
  104. }
  105. /*
  106. * Update X by
  107. * If (p > 0)
  108. * X_calc = calcX(s, R, p);
  109. * X = max(min(X_calc, 2 * X_recv), s / t_mbi);
  110. * Else
  111. * If (now - tld >= R)
  112. * X = max(min(2 * X, 2 * X_recv), s / R);
  113. * tld = now;
  114. *
  115. * Note: X and X_recv are both stored in units of 64 * bytes/second, to support
  116. * fine-grained resolution of sending rates. This requires scaling by 2^6
  117. * throughout the code. Only X_calc is unscaled (in bytes/second).
  118. *
  119. */
  120. static void ccid3_hc_tx_update_x(struct sock *sk, struct timeval *now)
  121. {
  122. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  123. __u64 min_rate = 2 * hctx->ccid3hctx_x_recv;
  124. const __u64 old_x = hctx->ccid3hctx_x;
  125. /*
  126. * Handle IDLE periods: do not reduce below RFC3390 initial sending rate
  127. * when idling [RFC 4342, 5.1]. See also draft-ietf-dccp-rfc3448bis.
  128. * For consistency with X and X_recv, min_rate is also scaled by 2^6.
  129. */
  130. if (unlikely(hctx->ccid3hctx_idle)) {
  131. min_rate = rfc3390_initial_rate(sk);
  132. min_rate = max(min_rate, 2 * hctx->ccid3hctx_x_recv);
  133. }
  134. if (hctx->ccid3hctx_p > 0) {
  135. hctx->ccid3hctx_x = min(((__u64)hctx->ccid3hctx_x_calc) << 6,
  136. min_rate);
  137. hctx->ccid3hctx_x = max(hctx->ccid3hctx_x,
  138. (((__u64)hctx->ccid3hctx_s) << 6) /
  139. TFRC_T_MBI);
  140. } else if (timeval_delta(now, &hctx->ccid3hctx_t_ld) -
  141. (suseconds_t)hctx->ccid3hctx_rtt >= 0) {
  142. hctx->ccid3hctx_x =
  143. max(min(2 * hctx->ccid3hctx_x, min_rate),
  144. scaled_div(((__u64)hctx->ccid3hctx_s) << 6,
  145. hctx->ccid3hctx_rtt));
  146. hctx->ccid3hctx_t_ld = *now;
  147. }
  148. if (hctx->ccid3hctx_x != old_x) {
  149. ccid3_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, "
  150. "X_recv=%u\n", (unsigned)(old_x >> 6),
  151. (unsigned)(hctx->ccid3hctx_x >> 6),
  152. hctx->ccid3hctx_x_calc,
  153. (unsigned)(hctx->ccid3hctx_x_recv >> 6));
  154. ccid3_update_send_interval(hctx);
  155. }
  156. }
  157. /*
  158. * Track the mean packet size `s' (cf. RFC 4342, 5.3 and RFC 3448, 4.1)
  159. * @len: DCCP packet payload size in bytes
  160. */
  161. static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock *hctx, int len)
  162. {
  163. const u16 old_s = hctx->ccid3hctx_s;
  164. hctx->ccid3hctx_s = old_s == 0 ? len : (9 * old_s + len) / 10;
  165. if (hctx->ccid3hctx_s != old_s)
  166. ccid3_update_send_interval(hctx);
  167. }
  168. /*
  169. * Update Window Counter using the algorithm from [RFC 4342, 8.1].
  170. * The algorithm is not applicable if RTT < 4 microseconds.
  171. */
  172. static inline void ccid3_hc_tx_update_win_count(struct ccid3_hc_tx_sock *hctx,
  173. struct timeval *now)
  174. {
  175. suseconds_t delta;
  176. u32 quarter_rtts;
  177. if (unlikely(hctx->ccid3hctx_rtt < 4)) /* avoid divide-by-zero */
  178. return;
  179. delta = timeval_delta(now, &hctx->ccid3hctx_t_last_win_count);
  180. DCCP_BUG_ON(delta < 0);
  181. quarter_rtts = (u32)delta / (hctx->ccid3hctx_rtt / 4);
  182. if (quarter_rtts > 0) {
  183. hctx->ccid3hctx_t_last_win_count = *now;
  184. hctx->ccid3hctx_last_win_count += min_t(u32, quarter_rtts, 5);
  185. hctx->ccid3hctx_last_win_count &= 0xF; /* mod 16 */
  186. ccid3_pr_debug("now at %#X\n", hctx->ccid3hctx_last_win_count);
  187. }
  188. }
  189. static void ccid3_hc_tx_no_feedback_timer(unsigned long data)
  190. {
  191. struct sock *sk = (struct sock *)data;
  192. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  193. struct timeval now;
  194. unsigned long t_nfb = USEC_PER_SEC / 5;
  195. bh_lock_sock(sk);
  196. if (sock_owned_by_user(sk)) {
  197. /* Try again later. */
  198. /* XXX: set some sensible MIB */
  199. goto restart_timer;
  200. }
  201. ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk), sk,
  202. ccid3_tx_state_name(hctx->ccid3hctx_state));
  203. hctx->ccid3hctx_idle = 1;
  204. switch (hctx->ccid3hctx_state) {
  205. case TFRC_SSTATE_NO_FBACK:
  206. /* RFC 3448, 4.4: Halve send rate directly */
  207. hctx->ccid3hctx_x = max(hctx->ccid3hctx_x / 2,
  208. (((__u64)hctx->ccid3hctx_s) << 6) /
  209. TFRC_T_MBI);
  210. ccid3_pr_debug("%s(%p, state=%s), updated tx rate to %u "
  211. "bytes/s\n", dccp_role(sk), sk,
  212. ccid3_tx_state_name(hctx->ccid3hctx_state),
  213. (unsigned)(hctx->ccid3hctx_x >> 6));
  214. /* The value of R is still undefined and so we can not recompute
  215. * the timout value. Keep initial value as per [RFC 4342, 5]. */
  216. t_nfb = TFRC_INITIAL_TIMEOUT;
  217. ccid3_update_send_interval(hctx);
  218. break;
  219. case TFRC_SSTATE_FBACK:
  220. /*
  221. * Modify the cached value of X_recv [RFC 3448, 4.4]
  222. *
  223. * If (p == 0 || X_calc > 2 * X_recv)
  224. * X_recv = max(X_recv / 2, s / (2 * t_mbi));
  225. * Else
  226. * X_recv = X_calc / 4;
  227. *
  228. * Note that X_recv is scaled by 2^6 while X_calc is not
  229. */
  230. BUG_ON(hctx->ccid3hctx_p && !hctx->ccid3hctx_x_calc);
  231. if (hctx->ccid3hctx_p == 0 ||
  232. (hctx->ccid3hctx_x_calc > (hctx->ccid3hctx_x_recv >> 5))) {
  233. hctx->ccid3hctx_x_recv =
  234. max(hctx->ccid3hctx_x_recv / 2,
  235. (((__u64)hctx->ccid3hctx_s) << 6) /
  236. (2 * TFRC_T_MBI));
  237. if (hctx->ccid3hctx_p == 0)
  238. dccp_timestamp(sk, &now);
  239. } else {
  240. hctx->ccid3hctx_x_recv = hctx->ccid3hctx_x_calc;
  241. hctx->ccid3hctx_x_recv <<= 4;
  242. }
  243. /* Now recalculate X [RFC 3448, 4.3, step (4)] */
  244. ccid3_hc_tx_update_x(sk, &now);
  245. /*
  246. * Schedule no feedback timer to expire in
  247. * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
  248. * See comments in packet_recv() regarding the value of t_RTO.
  249. */
  250. t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
  251. break;
  252. case TFRC_SSTATE_NO_SENT:
  253. DCCP_BUG("%s(%p) - Illegal state NO_SENT", dccp_role(sk), sk);
  254. /* fall through */
  255. case TFRC_SSTATE_TERM:
  256. goto out;
  257. }
  258. restart_timer:
  259. sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
  260. jiffies + usecs_to_jiffies(t_nfb));
  261. out:
  262. bh_unlock_sock(sk);
  263. sock_put(sk);
  264. }
  265. /*
  266. * returns
  267. * > 0: delay (in msecs) that should pass before actually sending
  268. * = 0: can send immediately
  269. * < 0: error condition; do not send packet
  270. */
  271. static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
  272. {
  273. struct dccp_sock *dp = dccp_sk(sk);
  274. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  275. struct timeval now;
  276. suseconds_t delay;
  277. BUG_ON(hctx == NULL);
  278. /*
  279. * This function is called only for Data and DataAck packets. Sending
  280. * zero-sized Data(Ack)s is theoretically possible, but for congestion
  281. * control this case is pathological - ignore it.
  282. */
  283. if (unlikely(skb->len == 0))
  284. return -EBADMSG;
  285. dccp_timestamp(sk, &now);
  286. switch (hctx->ccid3hctx_state) {
  287. case TFRC_SSTATE_NO_SENT:
  288. sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
  289. (jiffies +
  290. usecs_to_jiffies(TFRC_INITIAL_TIMEOUT)));
  291. hctx->ccid3hctx_last_win_count = 0;
  292. hctx->ccid3hctx_t_last_win_count = now;
  293. /* Set t_0 for initial packet */
  294. hctx->ccid3hctx_t_nom = now;
  295. hctx->ccid3hctx_s = skb->len;
  296. /*
  297. * Use initial RTT sample when available: recommended by erratum
  298. * to RFC 4342. This implements the initialisation procedure of
  299. * draft rfc3448bis, section 4.2. Remember, X is scaled by 2^6.
  300. */
  301. if (dp->dccps_syn_rtt) {
  302. ccid3_pr_debug("SYN RTT = %uus\n", dp->dccps_syn_rtt);
  303. hctx->ccid3hctx_rtt = dp->dccps_syn_rtt;
  304. hctx->ccid3hctx_x = rfc3390_initial_rate(sk);
  305. hctx->ccid3hctx_t_ld = now;
  306. } else {
  307. /* Sender does not have RTT sample: X = MSS/second */
  308. hctx->ccid3hctx_x = dp->dccps_mss_cache;
  309. hctx->ccid3hctx_x <<= 6;
  310. }
  311. ccid3_update_send_interval(hctx);
  312. ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK);
  313. break;
  314. case TFRC_SSTATE_NO_FBACK:
  315. case TFRC_SSTATE_FBACK:
  316. delay = timeval_delta(&hctx->ccid3hctx_t_nom, &now);
  317. ccid3_pr_debug("delay=%ld\n", (long)delay);
  318. /*
  319. * Scheduling of packet transmissions [RFC 3448, 4.6]
  320. *
  321. * if (t_now > t_nom - delta)
  322. * // send the packet now
  323. * else
  324. * // send the packet in (t_nom - t_now) milliseconds.
  325. */
  326. if (delay - (suseconds_t)hctx->ccid3hctx_delta >= 0)
  327. return delay / 1000L;
  328. ccid3_hc_tx_update_win_count(hctx, &now);
  329. break;
  330. case TFRC_SSTATE_TERM:
  331. DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
  332. return -EINVAL;
  333. }
  334. /* prepare to send now (add options etc.) */
  335. dp->dccps_hc_tx_insert_options = 1;
  336. DCCP_SKB_CB(skb)->dccpd_ccval = hctx->ccid3hctx_last_win_count;
  337. hctx->ccid3hctx_idle = 0;
  338. /* set the nominal send time for the next following packet */
  339. timeval_add_usecs(&hctx->ccid3hctx_t_nom, hctx->ccid3hctx_t_ipi);
  340. return 0;
  341. }
  342. static void ccid3_hc_tx_packet_sent(struct sock *sk, int more,
  343. unsigned int len)
  344. {
  345. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  346. struct timeval now;
  347. struct dccp_tx_hist_entry *packet;
  348. BUG_ON(hctx == NULL);
  349. ccid3_hc_tx_update_s(hctx, len);
  350. packet = dccp_tx_hist_entry_new(ccid3_tx_hist, GFP_ATOMIC);
  351. if (unlikely(packet == NULL)) {
  352. DCCP_CRIT("packet history - out of memory!");
  353. return;
  354. }
  355. dccp_tx_hist_add_entry(&hctx->ccid3hctx_hist, packet);
  356. dccp_timestamp(sk, &now);
  357. packet->dccphtx_tstamp = now;
  358. packet->dccphtx_seqno = dccp_sk(sk)->dccps_gss;
  359. packet->dccphtx_rtt = hctx->ccid3hctx_rtt;
  360. packet->dccphtx_sent = 1;
  361. }
  362. static void ccid3_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
  363. {
  364. const struct dccp_sock *dp = dccp_sk(sk);
  365. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  366. struct ccid3_options_received *opt_recv;
  367. struct dccp_tx_hist_entry *packet;
  368. struct timeval now;
  369. unsigned long t_nfb;
  370. u32 pinv, r_sample;
  371. BUG_ON(hctx == NULL);
  372. /* we are only interested in ACKs */
  373. if (!(DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK ||
  374. DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_DATAACK))
  375. return;
  376. opt_recv = &hctx->ccid3hctx_options_received;
  377. switch (hctx->ccid3hctx_state) {
  378. case TFRC_SSTATE_NO_FBACK:
  379. case TFRC_SSTATE_FBACK:
  380. /* get packet from history to look up t_recvdata */
  381. packet = dccp_tx_hist_find_entry(&hctx->ccid3hctx_hist,
  382. DCCP_SKB_CB(skb)->dccpd_ack_seq);
  383. if (unlikely(packet == NULL)) {
  384. DCCP_WARN("%s(%p), seqno %llu(%s) doesn't exist "
  385. "in history!\n", dccp_role(sk), sk,
  386. (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq,
  387. dccp_packet_name(DCCP_SKB_CB(skb)->dccpd_type));
  388. return;
  389. }
  390. /* Update receive rate in units of 64 * bytes/second */
  391. hctx->ccid3hctx_x_recv = opt_recv->ccid3or_receive_rate;
  392. hctx->ccid3hctx_x_recv <<= 6;
  393. /* Update loss event rate */
  394. pinv = opt_recv->ccid3or_loss_event_rate;
  395. if (pinv == ~0U || pinv == 0) /* see RFC 4342, 8.5 */
  396. hctx->ccid3hctx_p = 0;
  397. else /* can not exceed 100% */
  398. hctx->ccid3hctx_p = 1000000 / pinv;
  399. dccp_timestamp(sk, &now);
  400. /*
  401. * Calculate new round trip sample as per [RFC 3448, 4.3] by
  402. * R_sample = (now - t_recvdata) - t_elapsed
  403. */
  404. r_sample = dccp_sample_rtt(sk, &now, &packet->dccphtx_tstamp);
  405. /*
  406. * Update RTT estimate by
  407. * If (No feedback recv)
  408. * R = R_sample;
  409. * Else
  410. * R = q * R + (1 - q) * R_sample;
  411. *
  412. * q is a constant, RFC 3448 recomments 0.9
  413. */
  414. if (hctx->ccid3hctx_state == TFRC_SSTATE_NO_FBACK) {
  415. /*
  416. * Larger Initial Windows [RFC 4342, sec. 5]
  417. */
  418. hctx->ccid3hctx_rtt = r_sample;
  419. hctx->ccid3hctx_x = rfc3390_initial_rate(sk);
  420. hctx->ccid3hctx_t_ld = now;
  421. ccid3_update_send_interval(hctx);
  422. ccid3_pr_debug("%s(%p), s=%u, MSS=%u, "
  423. "R_sample=%uus, X=%u\n", dccp_role(sk),
  424. sk, hctx->ccid3hctx_s,
  425. dp->dccps_mss_cache, r_sample,
  426. (unsigned)(hctx->ccid3hctx_x >> 6));
  427. ccid3_hc_tx_set_state(sk, TFRC_SSTATE_FBACK);
  428. } else {
  429. hctx->ccid3hctx_rtt = (9 * hctx->ccid3hctx_rtt +
  430. r_sample) / 10;
  431. /* Update sending rate (step 4 of [RFC 3448, 4.3]) */
  432. if (hctx->ccid3hctx_p > 0)
  433. hctx->ccid3hctx_x_calc =
  434. tfrc_calc_x(hctx->ccid3hctx_s,
  435. hctx->ccid3hctx_rtt,
  436. hctx->ccid3hctx_p);
  437. ccid3_hc_tx_update_x(sk, &now);
  438. ccid3_pr_debug("%s(%p), RTT=%uus (sample=%uus), s=%u, "
  439. "p=%u, X_calc=%u, X_recv=%u, X=%u\n",
  440. dccp_role(sk),
  441. sk, hctx->ccid3hctx_rtt, r_sample,
  442. hctx->ccid3hctx_s, hctx->ccid3hctx_p,
  443. hctx->ccid3hctx_x_calc,
  444. (unsigned)(hctx->ccid3hctx_x_recv >> 6),
  445. (unsigned)(hctx->ccid3hctx_x >> 6));
  446. }
  447. /* unschedule no feedback timer */
  448. sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
  449. /* remove all packets older than the one acked from history */
  450. dccp_tx_hist_purge_older(ccid3_tx_hist,
  451. &hctx->ccid3hctx_hist, packet);
  452. /*
  453. * As we have calculated new ipi, delta, t_nom it is possible
  454. * that we now can send a packet, so wake up dccp_wait_for_ccid
  455. */
  456. sk->sk_write_space(sk);
  457. /*
  458. * Update timeout interval for the nofeedback timer.
  459. * We use a configuration option to increase the lower bound.
  460. * This can help avoid triggering the nofeedback timer too
  461. * often ('spinning') on LANs with small RTTs.
  462. */
  463. hctx->ccid3hctx_t_rto = max_t(u32, 4 * hctx->ccid3hctx_rtt,
  464. CONFIG_IP_DCCP_CCID3_RTO *
  465. (USEC_PER_SEC/1000));
  466. /*
  467. * Schedule no feedback timer to expire in
  468. * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
  469. */
  470. t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
  471. ccid3_pr_debug("%s(%p), Scheduled no feedback timer to "
  472. "expire in %lu jiffies (%luus)\n",
  473. dccp_role(sk),
  474. sk, usecs_to_jiffies(t_nfb), t_nfb);
  475. sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
  476. jiffies + usecs_to_jiffies(t_nfb));
  477. /* set idle flag */
  478. hctx->ccid3hctx_idle = 1;
  479. break;
  480. case TFRC_SSTATE_NO_SENT: /* fall through */
  481. case TFRC_SSTATE_TERM: /* ignore feedback when closing */
  482. break;
  483. }
  484. }
  485. static int ccid3_hc_tx_parse_options(struct sock *sk, unsigned char option,
  486. unsigned char len, u16 idx,
  487. unsigned char *value)
  488. {
  489. int rc = 0;
  490. const struct dccp_sock *dp = dccp_sk(sk);
  491. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  492. struct ccid3_options_received *opt_recv;
  493. BUG_ON(hctx == NULL);
  494. opt_recv = &hctx->ccid3hctx_options_received;
  495. if (opt_recv->ccid3or_seqno != dp->dccps_gsr) {
  496. opt_recv->ccid3or_seqno = dp->dccps_gsr;
  497. opt_recv->ccid3or_loss_event_rate = ~0;
  498. opt_recv->ccid3or_loss_intervals_idx = 0;
  499. opt_recv->ccid3or_loss_intervals_len = 0;
  500. opt_recv->ccid3or_receive_rate = 0;
  501. }
  502. switch (option) {
  503. case TFRC_OPT_LOSS_EVENT_RATE:
  504. if (unlikely(len != 4)) {
  505. DCCP_WARN("%s(%p), invalid len %d "
  506. "for TFRC_OPT_LOSS_EVENT_RATE\n",
  507. dccp_role(sk), sk, len);
  508. rc = -EINVAL;
  509. } else {
  510. opt_recv->ccid3or_loss_event_rate =
  511. ntohl(*(__be32 *)value);
  512. ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
  513. dccp_role(sk), sk,
  514. opt_recv->ccid3or_loss_event_rate);
  515. }
  516. break;
  517. case TFRC_OPT_LOSS_INTERVALS:
  518. opt_recv->ccid3or_loss_intervals_idx = idx;
  519. opt_recv->ccid3or_loss_intervals_len = len;
  520. ccid3_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n",
  521. dccp_role(sk), sk,
  522. opt_recv->ccid3or_loss_intervals_idx,
  523. opt_recv->ccid3or_loss_intervals_len);
  524. break;
  525. case TFRC_OPT_RECEIVE_RATE:
  526. if (unlikely(len != 4)) {
  527. DCCP_WARN("%s(%p), invalid len %d "
  528. "for TFRC_OPT_RECEIVE_RATE\n",
  529. dccp_role(sk), sk, len);
  530. rc = -EINVAL;
  531. } else {
  532. opt_recv->ccid3or_receive_rate =
  533. ntohl(*(__be32 *)value);
  534. ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n",
  535. dccp_role(sk), sk,
  536. opt_recv->ccid3or_receive_rate);
  537. }
  538. break;
  539. }
  540. return rc;
  541. }
  542. static int ccid3_hc_tx_init(struct ccid *ccid, struct sock *sk)
  543. {
  544. struct ccid3_hc_tx_sock *hctx = ccid_priv(ccid);
  545. hctx->ccid3hctx_s = 0;
  546. hctx->ccid3hctx_rtt = 0;
  547. hctx->ccid3hctx_state = TFRC_SSTATE_NO_SENT;
  548. INIT_LIST_HEAD(&hctx->ccid3hctx_hist);
  549. hctx->ccid3hctx_no_feedback_timer.function =
  550. ccid3_hc_tx_no_feedback_timer;
  551. hctx->ccid3hctx_no_feedback_timer.data = (unsigned long)sk;
  552. init_timer(&hctx->ccid3hctx_no_feedback_timer);
  553. return 0;
  554. }
  555. static void ccid3_hc_tx_exit(struct sock *sk)
  556. {
  557. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  558. BUG_ON(hctx == NULL);
  559. ccid3_hc_tx_set_state(sk, TFRC_SSTATE_TERM);
  560. sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
  561. /* Empty packet history */
  562. dccp_tx_hist_purge(ccid3_tx_hist, &hctx->ccid3hctx_hist);
  563. }
  564. static void ccid3_hc_tx_get_info(struct sock *sk, struct tcp_info *info)
  565. {
  566. const struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  567. /* Listen socks doesn't have a private CCID block */
  568. if (sk->sk_state == DCCP_LISTEN)
  569. return;
  570. BUG_ON(hctx == NULL);
  571. info->tcpi_rto = hctx->ccid3hctx_t_rto;
  572. info->tcpi_rtt = hctx->ccid3hctx_rtt;
  573. }
  574. static int ccid3_hc_tx_getsockopt(struct sock *sk, const int optname, int len,
  575. u32 __user *optval, int __user *optlen)
  576. {
  577. const struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  578. const void *val;
  579. /* Listen socks doesn't have a private CCID block */
  580. if (sk->sk_state == DCCP_LISTEN)
  581. return -EINVAL;
  582. switch (optname) {
  583. case DCCP_SOCKOPT_CCID_TX_INFO:
  584. if (len < sizeof(hctx->ccid3hctx_tfrc))
  585. return -EINVAL;
  586. len = sizeof(hctx->ccid3hctx_tfrc);
  587. val = &hctx->ccid3hctx_tfrc;
  588. break;
  589. default:
  590. return -ENOPROTOOPT;
  591. }
  592. if (put_user(len, optlen) || copy_to_user(optval, val, len))
  593. return -EFAULT;
  594. return 0;
  595. }
  596. /*
  597. * Receiver Half-Connection Routines
  598. */
  599. #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
  600. static const char *ccid3_rx_state_name(enum ccid3_hc_rx_states state)
  601. {
  602. static char *ccid3_rx_state_names[] = {
  603. [TFRC_RSTATE_NO_DATA] = "NO_DATA",
  604. [TFRC_RSTATE_DATA] = "DATA",
  605. [TFRC_RSTATE_TERM] = "TERM",
  606. };
  607. return ccid3_rx_state_names[state];
  608. }
  609. #endif
  610. static void ccid3_hc_rx_set_state(struct sock *sk,
  611. enum ccid3_hc_rx_states state)
  612. {
  613. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  614. enum ccid3_hc_rx_states oldstate = hcrx->ccid3hcrx_state;
  615. ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
  616. dccp_role(sk), sk, ccid3_rx_state_name(oldstate),
  617. ccid3_rx_state_name(state));
  618. WARN_ON(state == oldstate);
  619. hcrx->ccid3hcrx_state = state;
  620. }
  621. static inline void ccid3_hc_rx_update_s(struct ccid3_hc_rx_sock *hcrx, int len)
  622. {
  623. if (unlikely(len == 0)) /* don't update on empty packets (e.g. ACKs) */
  624. ccid3_pr_debug("Packet payload length is 0 - not updating\n");
  625. else
  626. hcrx->ccid3hcrx_s = hcrx->ccid3hcrx_s == 0 ? len :
  627. (9 * hcrx->ccid3hcrx_s + len) / 10;
  628. }
  629. static void ccid3_hc_rx_send_feedback(struct sock *sk)
  630. {
  631. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  632. struct dccp_sock *dp = dccp_sk(sk);
  633. struct dccp_rx_hist_entry *packet;
  634. struct timeval now;
  635. suseconds_t delta;
  636. ccid3_pr_debug("%s(%p) - entry \n", dccp_role(sk), sk);
  637. dccp_timestamp(sk, &now);
  638. switch (hcrx->ccid3hcrx_state) {
  639. case TFRC_RSTATE_NO_DATA:
  640. hcrx->ccid3hcrx_x_recv = 0;
  641. break;
  642. case TFRC_RSTATE_DATA:
  643. delta = timeval_delta(&now,
  644. &hcrx->ccid3hcrx_tstamp_last_feedback);
  645. DCCP_BUG_ON(delta < 0);
  646. hcrx->ccid3hcrx_x_recv =
  647. scaled_div32(hcrx->ccid3hcrx_bytes_recv, delta);
  648. break;
  649. case TFRC_RSTATE_TERM:
  650. DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
  651. return;
  652. }
  653. packet = dccp_rx_hist_find_data_packet(&hcrx->ccid3hcrx_hist);
  654. if (unlikely(packet == NULL)) {
  655. DCCP_WARN("%s(%p), no data packet in history!\n",
  656. dccp_role(sk), sk);
  657. return;
  658. }
  659. hcrx->ccid3hcrx_tstamp_last_feedback = now;
  660. hcrx->ccid3hcrx_ccval_last_counter = packet->dccphrx_ccval;
  661. hcrx->ccid3hcrx_bytes_recv = 0;
  662. /* Elapsed time information [RFC 4340, 13.2] in units of 10 * usecs */
  663. delta = timeval_delta(&now, &packet->dccphrx_tstamp);
  664. DCCP_BUG_ON(delta < 0);
  665. hcrx->ccid3hcrx_elapsed_time = delta / 10;
  666. if (hcrx->ccid3hcrx_p == 0)
  667. hcrx->ccid3hcrx_pinv = ~0U; /* see RFC 4342, 8.5 */
  668. else if (hcrx->ccid3hcrx_p > 1000000) {
  669. DCCP_WARN("p (%u) > 100%%\n", hcrx->ccid3hcrx_p);
  670. hcrx->ccid3hcrx_pinv = 1; /* use 100% in this case */
  671. } else
  672. hcrx->ccid3hcrx_pinv = 1000000 / hcrx->ccid3hcrx_p;
  673. dp->dccps_hc_rx_insert_options = 1;
  674. dccp_send_ack(sk);
  675. }
  676. static int ccid3_hc_rx_insert_options(struct sock *sk, struct sk_buff *skb)
  677. {
  678. const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  679. __be32 x_recv, pinv;
  680. BUG_ON(hcrx == NULL);
  681. if (!(sk->sk_state == DCCP_OPEN || sk->sk_state == DCCP_PARTOPEN))
  682. return 0;
  683. DCCP_SKB_CB(skb)->dccpd_ccval = hcrx->ccid3hcrx_ccval_last_counter;
  684. if (dccp_packet_without_ack(skb))
  685. return 0;
  686. x_recv = htonl(hcrx->ccid3hcrx_x_recv);
  687. pinv = htonl(hcrx->ccid3hcrx_pinv);
  688. if ((hcrx->ccid3hcrx_elapsed_time != 0 &&
  689. dccp_insert_option_elapsed_time(sk, skb,
  690. hcrx->ccid3hcrx_elapsed_time)) ||
  691. dccp_insert_option_timestamp(sk, skb) ||
  692. dccp_insert_option(sk, skb, TFRC_OPT_LOSS_EVENT_RATE,
  693. &pinv, sizeof(pinv)) ||
  694. dccp_insert_option(sk, skb, TFRC_OPT_RECEIVE_RATE,
  695. &x_recv, sizeof(x_recv)))
  696. return -1;
  697. return 0;
  698. }
  699. /* calculate first loss interval
  700. *
  701. * returns estimated loss interval in usecs */
  702. static u32 ccid3_hc_rx_calc_first_li(struct sock *sk)
  703. {
  704. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  705. struct dccp_rx_hist_entry *entry, *next, *tail = NULL;
  706. u32 x_recv, p;
  707. suseconds_t rtt, delta;
  708. struct timeval tstamp = { 0, };
  709. int interval = 0;
  710. int win_count = 0;
  711. int step = 0;
  712. u64 fval;
  713. list_for_each_entry_safe(entry, next, &hcrx->ccid3hcrx_hist,
  714. dccphrx_node) {
  715. if (dccp_rx_hist_entry_data_packet(entry)) {
  716. tail = entry;
  717. switch (step) {
  718. case 0:
  719. tstamp = entry->dccphrx_tstamp;
  720. win_count = entry->dccphrx_ccval;
  721. step = 1;
  722. break;
  723. case 1:
  724. interval = win_count - entry->dccphrx_ccval;
  725. if (interval < 0)
  726. interval += TFRC_WIN_COUNT_LIMIT;
  727. if (interval > 4)
  728. goto found;
  729. break;
  730. }
  731. }
  732. }
  733. if (unlikely(step == 0)) {
  734. DCCP_WARN("%s(%p), packet history has no data packets!\n",
  735. dccp_role(sk), sk);
  736. return ~0;
  737. }
  738. if (unlikely(interval == 0)) {
  739. DCCP_WARN("%s(%p), Could not find a win_count interval > 0."
  740. "Defaulting to 1\n", dccp_role(sk), sk);
  741. interval = 1;
  742. }
  743. found:
  744. if (!tail) {
  745. DCCP_CRIT("tail is null\n");
  746. return ~0;
  747. }
  748. delta = timeval_delta(&tstamp, &tail->dccphrx_tstamp);
  749. DCCP_BUG_ON(delta < 0);
  750. rtt = delta * 4 / interval;
  751. ccid3_pr_debug("%s(%p), approximated RTT to %dus\n",
  752. dccp_role(sk), sk, (int)rtt);
  753. /*
  754. * Determine the length of the first loss interval via inverse lookup.
  755. * Assume that X_recv can be computed by the throughput equation
  756. * s
  757. * X_recv = --------
  758. * R * fval
  759. * Find some p such that f(p) = fval; return 1/p [RFC 3448, 6.3.1].
  760. */
  761. if (rtt == 0) { /* would result in divide-by-zero */
  762. DCCP_WARN("RTT==0\n");
  763. return ~0;
  764. }
  765. dccp_timestamp(sk, &tstamp);
  766. delta = timeval_delta(&tstamp, &hcrx->ccid3hcrx_tstamp_last_feedback);
  767. DCCP_BUG_ON(delta <= 0);
  768. x_recv = scaled_div32(hcrx->ccid3hcrx_bytes_recv, delta);
  769. if (x_recv == 0) { /* would also trigger divide-by-zero */
  770. DCCP_WARN("X_recv==0\n");
  771. if ((x_recv = hcrx->ccid3hcrx_x_recv) == 0) {
  772. DCCP_BUG("stored value of X_recv is zero");
  773. return ~0;
  774. }
  775. }
  776. fval = scaled_div(hcrx->ccid3hcrx_s, rtt);
  777. fval = scaled_div32(fval, x_recv);
  778. p = tfrc_calc_x_reverse_lookup(fval);
  779. ccid3_pr_debug("%s(%p), receive rate=%u bytes/s, implied "
  780. "loss rate=%u\n", dccp_role(sk), sk, x_recv, p);
  781. if (p == 0)
  782. return ~0;
  783. else
  784. return 1000000 / p;
  785. }
  786. static void ccid3_hc_rx_update_li(struct sock *sk, u64 seq_loss, u8 win_loss)
  787. {
  788. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  789. struct dccp_li_hist_entry *head;
  790. u64 seq_temp;
  791. if (list_empty(&hcrx->ccid3hcrx_li_hist)) {
  792. if (!dccp_li_hist_interval_new(ccid3_li_hist,
  793. &hcrx->ccid3hcrx_li_hist, seq_loss, win_loss))
  794. return;
  795. head = list_entry(hcrx->ccid3hcrx_li_hist.next,
  796. struct dccp_li_hist_entry, dccplih_node);
  797. head->dccplih_interval = ccid3_hc_rx_calc_first_li(sk);
  798. } else {
  799. struct dccp_li_hist_entry *entry;
  800. struct list_head *tail;
  801. head = list_entry(hcrx->ccid3hcrx_li_hist.next,
  802. struct dccp_li_hist_entry, dccplih_node);
  803. /* FIXME win count check removed as was wrong */
  804. /* should make this check with receive history */
  805. /* and compare there as per section 10.2 of RFC4342 */
  806. /* new loss event detected */
  807. /* calculate last interval length */
  808. seq_temp = dccp_delta_seqno(head->dccplih_seqno, seq_loss);
  809. entry = dccp_li_hist_entry_new(ccid3_li_hist, GFP_ATOMIC);
  810. if (entry == NULL) {
  811. DCCP_BUG("out of memory - can not allocate entry");
  812. return;
  813. }
  814. list_add(&entry->dccplih_node, &hcrx->ccid3hcrx_li_hist);
  815. tail = hcrx->ccid3hcrx_li_hist.prev;
  816. list_del(tail);
  817. kmem_cache_free(ccid3_li_hist->dccplih_slab, tail);
  818. /* Create the newest interval */
  819. entry->dccplih_seqno = seq_loss;
  820. entry->dccplih_interval = seq_temp;
  821. entry->dccplih_win_count = win_loss;
  822. }
  823. }
  824. static int ccid3_hc_rx_detect_loss(struct sock *sk,
  825. struct dccp_rx_hist_entry *packet)
  826. {
  827. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  828. struct dccp_rx_hist_entry *rx_hist =
  829. dccp_rx_hist_head(&hcrx->ccid3hcrx_hist);
  830. u64 seqno = packet->dccphrx_seqno;
  831. u64 tmp_seqno;
  832. int loss = 0;
  833. u8 ccval;
  834. tmp_seqno = hcrx->ccid3hcrx_seqno_nonloss;
  835. if (!rx_hist ||
  836. follows48(packet->dccphrx_seqno, hcrx->ccid3hcrx_seqno_nonloss)) {
  837. hcrx->ccid3hcrx_seqno_nonloss = seqno;
  838. hcrx->ccid3hcrx_ccval_nonloss = packet->dccphrx_ccval;
  839. goto detect_out;
  840. }
  841. while (dccp_delta_seqno(hcrx->ccid3hcrx_seqno_nonloss, seqno)
  842. > TFRC_RECV_NUM_LATE_LOSS) {
  843. loss = 1;
  844. ccid3_hc_rx_update_li(sk, hcrx->ccid3hcrx_seqno_nonloss,
  845. hcrx->ccid3hcrx_ccval_nonloss);
  846. tmp_seqno = hcrx->ccid3hcrx_seqno_nonloss;
  847. dccp_inc_seqno(&tmp_seqno);
  848. hcrx->ccid3hcrx_seqno_nonloss = tmp_seqno;
  849. dccp_inc_seqno(&tmp_seqno);
  850. while (dccp_rx_hist_find_entry(&hcrx->ccid3hcrx_hist,
  851. tmp_seqno, &ccval)) {
  852. hcrx->ccid3hcrx_seqno_nonloss = tmp_seqno;
  853. hcrx->ccid3hcrx_ccval_nonloss = ccval;
  854. dccp_inc_seqno(&tmp_seqno);
  855. }
  856. }
  857. /* FIXME - this code could be simplified with above while */
  858. /* but works at moment */
  859. if (follows48(packet->dccphrx_seqno, hcrx->ccid3hcrx_seqno_nonloss)) {
  860. hcrx->ccid3hcrx_seqno_nonloss = seqno;
  861. hcrx->ccid3hcrx_ccval_nonloss = packet->dccphrx_ccval;
  862. }
  863. detect_out:
  864. dccp_rx_hist_add_packet(ccid3_rx_hist, &hcrx->ccid3hcrx_hist,
  865. &hcrx->ccid3hcrx_li_hist, packet,
  866. hcrx->ccid3hcrx_seqno_nonloss);
  867. return loss;
  868. }
  869. static void ccid3_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
  870. {
  871. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  872. const struct dccp_options_received *opt_recv;
  873. struct dccp_rx_hist_entry *packet;
  874. struct timeval now;
  875. u32 p_prev, r_sample, rtt_prev;
  876. int loss, payload_size;
  877. BUG_ON(hcrx == NULL);
  878. opt_recv = &dccp_sk(sk)->dccps_options_received;
  879. switch (DCCP_SKB_CB(skb)->dccpd_type) {
  880. case DCCP_PKT_ACK:
  881. if (hcrx->ccid3hcrx_state == TFRC_RSTATE_NO_DATA)
  882. return;
  883. case DCCP_PKT_DATAACK:
  884. if (opt_recv->dccpor_timestamp_echo == 0)
  885. break;
  886. rtt_prev = hcrx->ccid3hcrx_rtt;
  887. dccp_timestamp(sk, &now);
  888. r_sample = dccp_sample_rtt(sk, &now, NULL);
  889. if (hcrx->ccid3hcrx_state == TFRC_RSTATE_NO_DATA)
  890. hcrx->ccid3hcrx_rtt = r_sample;
  891. else
  892. hcrx->ccid3hcrx_rtt = (hcrx->ccid3hcrx_rtt * 9) / 10 +
  893. r_sample / 10;
  894. if (rtt_prev != hcrx->ccid3hcrx_rtt)
  895. ccid3_pr_debug("%s(%p), New RTT=%uus, elapsed time=%u\n",
  896. dccp_role(sk), sk, hcrx->ccid3hcrx_rtt,
  897. opt_recv->dccpor_elapsed_time);
  898. break;
  899. case DCCP_PKT_DATA:
  900. break;
  901. default: /* We're not interested in other packet types, move along */
  902. return;
  903. }
  904. packet = dccp_rx_hist_entry_new(ccid3_rx_hist, sk, opt_recv->dccpor_ndp,
  905. skb, GFP_ATOMIC);
  906. if (unlikely(packet == NULL)) {
  907. DCCP_WARN("%s(%p), Not enough mem to add rx packet "
  908. "to history, consider it lost!\n", dccp_role(sk), sk);
  909. return;
  910. }
  911. loss = ccid3_hc_rx_detect_loss(sk, packet);
  912. if (DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK)
  913. return;
  914. payload_size = skb->len - dccp_hdr(skb)->dccph_doff * 4;
  915. ccid3_hc_rx_update_s(hcrx, payload_size);
  916. switch (hcrx->ccid3hcrx_state) {
  917. case TFRC_RSTATE_NO_DATA:
  918. ccid3_pr_debug("%s(%p, state=%s), skb=%p, sending initial "
  919. "feedback\n", dccp_role(sk), sk,
  920. dccp_state_name(sk->sk_state), skb);
  921. ccid3_hc_rx_send_feedback(sk);
  922. ccid3_hc_rx_set_state(sk, TFRC_RSTATE_DATA);
  923. return;
  924. case TFRC_RSTATE_DATA:
  925. hcrx->ccid3hcrx_bytes_recv += payload_size;
  926. if (loss)
  927. break;
  928. dccp_timestamp(sk, &now);
  929. if ((timeval_delta(&now, &hcrx->ccid3hcrx_tstamp_last_ack) -
  930. (suseconds_t)hcrx->ccid3hcrx_rtt) >= 0) {
  931. hcrx->ccid3hcrx_tstamp_last_ack = now;
  932. ccid3_hc_rx_send_feedback(sk);
  933. }
  934. return;
  935. case TFRC_RSTATE_TERM:
  936. DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
  937. return;
  938. }
  939. /* Dealing with packet loss */
  940. ccid3_pr_debug("%s(%p, state=%s), data loss! Reacting...\n",
  941. dccp_role(sk), sk, dccp_state_name(sk->sk_state));
  942. p_prev = hcrx->ccid3hcrx_p;
  943. /* Calculate loss event rate */
  944. if (!list_empty(&hcrx->ccid3hcrx_li_hist)) {
  945. u32 i_mean = dccp_li_hist_calc_i_mean(&hcrx->ccid3hcrx_li_hist);
  946. /* Scaling up by 1000000 as fixed decimal */
  947. if (i_mean != 0)
  948. hcrx->ccid3hcrx_p = 1000000 / i_mean;
  949. } else
  950. DCCP_BUG("empty loss history");
  951. if (hcrx->ccid3hcrx_p > p_prev) {
  952. ccid3_hc_rx_send_feedback(sk);
  953. return;
  954. }
  955. }
  956. static int ccid3_hc_rx_init(struct ccid *ccid, struct sock *sk)
  957. {
  958. struct ccid3_hc_rx_sock *hcrx = ccid_priv(ccid);
  959. ccid3_pr_debug("entry\n");
  960. hcrx->ccid3hcrx_state = TFRC_RSTATE_NO_DATA;
  961. INIT_LIST_HEAD(&hcrx->ccid3hcrx_hist);
  962. INIT_LIST_HEAD(&hcrx->ccid3hcrx_li_hist);
  963. dccp_timestamp(sk, &hcrx->ccid3hcrx_tstamp_last_ack);
  964. hcrx->ccid3hcrx_tstamp_last_feedback = hcrx->ccid3hcrx_tstamp_last_ack;
  965. hcrx->ccid3hcrx_s = 0;
  966. hcrx->ccid3hcrx_rtt = 0;
  967. return 0;
  968. }
  969. static void ccid3_hc_rx_exit(struct sock *sk)
  970. {
  971. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  972. BUG_ON(hcrx == NULL);
  973. ccid3_hc_rx_set_state(sk, TFRC_RSTATE_TERM);
  974. /* Empty packet history */
  975. dccp_rx_hist_purge(ccid3_rx_hist, &hcrx->ccid3hcrx_hist);
  976. /* Empty loss interval history */
  977. dccp_li_hist_purge(ccid3_li_hist, &hcrx->ccid3hcrx_li_hist);
  978. }
  979. static void ccid3_hc_rx_get_info(struct sock *sk, struct tcp_info *info)
  980. {
  981. const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  982. /* Listen socks doesn't have a private CCID block */
  983. if (sk->sk_state == DCCP_LISTEN)
  984. return;
  985. BUG_ON(hcrx == NULL);
  986. info->tcpi_ca_state = hcrx->ccid3hcrx_state;
  987. info->tcpi_options |= TCPI_OPT_TIMESTAMPS;
  988. info->tcpi_rcv_rtt = hcrx->ccid3hcrx_rtt;
  989. }
  990. static int ccid3_hc_rx_getsockopt(struct sock *sk, const int optname, int len,
  991. u32 __user *optval, int __user *optlen)
  992. {
  993. const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  994. const void *val;
  995. /* Listen socks doesn't have a private CCID block */
  996. if (sk->sk_state == DCCP_LISTEN)
  997. return -EINVAL;
  998. switch (optname) {
  999. case DCCP_SOCKOPT_CCID_RX_INFO:
  1000. if (len < sizeof(hcrx->ccid3hcrx_tfrc))
  1001. return -EINVAL;
  1002. len = sizeof(hcrx->ccid3hcrx_tfrc);
  1003. val = &hcrx->ccid3hcrx_tfrc;
  1004. break;
  1005. default:
  1006. return -ENOPROTOOPT;
  1007. }
  1008. if (put_user(len, optlen) || copy_to_user(optval, val, len))
  1009. return -EFAULT;
  1010. return 0;
  1011. }
  1012. static struct ccid_operations ccid3 = {
  1013. .ccid_id = DCCPC_CCID3,
  1014. .ccid_name = "ccid3",
  1015. .ccid_owner = THIS_MODULE,
  1016. .ccid_hc_tx_obj_size = sizeof(struct ccid3_hc_tx_sock),
  1017. .ccid_hc_tx_init = ccid3_hc_tx_init,
  1018. .ccid_hc_tx_exit = ccid3_hc_tx_exit,
  1019. .ccid_hc_tx_send_packet = ccid3_hc_tx_send_packet,
  1020. .ccid_hc_tx_packet_sent = ccid3_hc_tx_packet_sent,
  1021. .ccid_hc_tx_packet_recv = ccid3_hc_tx_packet_recv,
  1022. .ccid_hc_tx_parse_options = ccid3_hc_tx_parse_options,
  1023. .ccid_hc_rx_obj_size = sizeof(struct ccid3_hc_rx_sock),
  1024. .ccid_hc_rx_init = ccid3_hc_rx_init,
  1025. .ccid_hc_rx_exit = ccid3_hc_rx_exit,
  1026. .ccid_hc_rx_insert_options = ccid3_hc_rx_insert_options,
  1027. .ccid_hc_rx_packet_recv = ccid3_hc_rx_packet_recv,
  1028. .ccid_hc_rx_get_info = ccid3_hc_rx_get_info,
  1029. .ccid_hc_tx_get_info = ccid3_hc_tx_get_info,
  1030. .ccid_hc_rx_getsockopt = ccid3_hc_rx_getsockopt,
  1031. .ccid_hc_tx_getsockopt = ccid3_hc_tx_getsockopt,
  1032. };
  1033. #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
  1034. module_param(ccid3_debug, int, 0444);
  1035. MODULE_PARM_DESC(ccid3_debug, "Enable debug messages");
  1036. #endif
  1037. static __init int ccid3_module_init(void)
  1038. {
  1039. int rc = -ENOBUFS;
  1040. ccid3_rx_hist = dccp_rx_hist_new("ccid3");
  1041. if (ccid3_rx_hist == NULL)
  1042. goto out;
  1043. ccid3_tx_hist = dccp_tx_hist_new("ccid3");
  1044. if (ccid3_tx_hist == NULL)
  1045. goto out_free_rx;
  1046. ccid3_li_hist = dccp_li_hist_new("ccid3");
  1047. if (ccid3_li_hist == NULL)
  1048. goto out_free_tx;
  1049. rc = ccid_register(&ccid3);
  1050. if (rc != 0)
  1051. goto out_free_loss_interval_history;
  1052. out:
  1053. return rc;
  1054. out_free_loss_interval_history:
  1055. dccp_li_hist_delete(ccid3_li_hist);
  1056. ccid3_li_hist = NULL;
  1057. out_free_tx:
  1058. dccp_tx_hist_delete(ccid3_tx_hist);
  1059. ccid3_tx_hist = NULL;
  1060. out_free_rx:
  1061. dccp_rx_hist_delete(ccid3_rx_hist);
  1062. ccid3_rx_hist = NULL;
  1063. goto out;
  1064. }
  1065. module_init(ccid3_module_init);
  1066. static __exit void ccid3_module_exit(void)
  1067. {
  1068. ccid_unregister(&ccid3);
  1069. if (ccid3_tx_hist != NULL) {
  1070. dccp_tx_hist_delete(ccid3_tx_hist);
  1071. ccid3_tx_hist = NULL;
  1072. }
  1073. if (ccid3_rx_hist != NULL) {
  1074. dccp_rx_hist_delete(ccid3_rx_hist);
  1075. ccid3_rx_hist = NULL;
  1076. }
  1077. if (ccid3_li_hist != NULL) {
  1078. dccp_li_hist_delete(ccid3_li_hist);
  1079. ccid3_li_hist = NULL;
  1080. }
  1081. }
  1082. module_exit(ccid3_module_exit);
  1083. MODULE_AUTHOR("Ian McDonald <ian.mcdonald@jandi.co.nz>, "
  1084. "Arnaldo Carvalho de Melo <acme@ghostprotocols.net>");
  1085. MODULE_DESCRIPTION("DCCP TFRC CCID3 CCID");
  1086. MODULE_LICENSE("GPL");
  1087. MODULE_ALIAS("net-dccp-ccid-3");