link.c 57 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131
  1. /*
  2. * net/tipc/link.c: TIPC link code
  3. *
  4. * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
  5. * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
  6. * All rights reserved.
  7. *
  8. * Redistribution and use in source and binary forms, with or without
  9. * modification, are permitted provided that the following conditions are met:
  10. *
  11. * 1. Redistributions of source code must retain the above copyright
  12. * notice, this list of conditions and the following disclaimer.
  13. * 2. Redistributions in binary form must reproduce the above copyright
  14. * notice, this list of conditions and the following disclaimer in the
  15. * documentation and/or other materials provided with the distribution.
  16. * 3. Neither the names of the copyright holders nor the names of its
  17. * contributors may be used to endorse or promote products derived from
  18. * this software without specific prior written permission.
  19. *
  20. * Alternatively, this software may be distributed under the terms of the
  21. * GNU General Public License ("GPL") version 2 as published by the Free
  22. * Software Foundation.
  23. *
  24. * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  25. * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26. * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27. * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  28. * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  29. * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  30. * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  31. * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  32. * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  33. * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  34. * POSSIBILITY OF SUCH DAMAGE.
  35. */
  36. #include "core.h"
  37. #include "subscr.h"
  38. #include "link.h"
  39. #include "bcast.h"
  40. #include "socket.h"
  41. #include "name_distr.h"
  42. #include "discover.h"
  43. #include "netlink.h"
  44. #include "monitor.h"
  45. #include <linux/pkt_sched.h>
  46. struct tipc_stats {
  47. u32 sent_pkts;
  48. u32 recv_pkts;
  49. u32 sent_states;
  50. u32 recv_states;
  51. u32 sent_probes;
  52. u32 recv_probes;
  53. u32 sent_nacks;
  54. u32 recv_nacks;
  55. u32 sent_acks;
  56. u32 sent_bundled;
  57. u32 sent_bundles;
  58. u32 recv_bundled;
  59. u32 recv_bundles;
  60. u32 retransmitted;
  61. u32 sent_fragmented;
  62. u32 sent_fragments;
  63. u32 recv_fragmented;
  64. u32 recv_fragments;
  65. u32 link_congs; /* # port sends blocked by congestion */
  66. u32 deferred_recv;
  67. u32 duplicates;
  68. u32 max_queue_sz; /* send queue size high water mark */
  69. u32 accu_queue_sz; /* used for send queue size profiling */
  70. u32 queue_sz_counts; /* used for send queue size profiling */
  71. u32 msg_length_counts; /* used for message length profiling */
  72. u32 msg_lengths_total; /* used for message length profiling */
  73. u32 msg_length_profile[7]; /* used for msg. length profiling */
  74. };
  75. /**
  76. * struct tipc_link - TIPC link data structure
  77. * @addr: network address of link's peer node
  78. * @name: link name character string
  79. * @media_addr: media address to use when sending messages over link
  80. * @timer: link timer
  81. * @net: pointer to namespace struct
  82. * @refcnt: reference counter for permanent references (owner node & timer)
  83. * @peer_session: link session # being used by peer end of link
  84. * @peer_bearer_id: bearer id used by link's peer endpoint
  85. * @bearer_id: local bearer id used by link
  86. * @tolerance: minimum link continuity loss needed to reset link [in ms]
  87. * @abort_limit: # of unacknowledged continuity probes needed to reset link
  88. * @state: current state of link FSM
  89. * @peer_caps: bitmap describing capabilities of peer node
  90. * @silent_intv_cnt: # of timer intervals without any reception from peer
  91. * @proto_msg: template for control messages generated by link
  92. * @pmsg: convenience pointer to "proto_msg" field
  93. * @priority: current link priority
  94. * @net_plane: current link network plane ('A' through 'H')
  95. * @mon_state: cookie with information needed by link monitor
  96. * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
  97. * @exp_msg_count: # of tunnelled messages expected during link changeover
  98. * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
  99. * @mtu: current maximum packet size for this link
  100. * @advertised_mtu: advertised own mtu when link is being established
  101. * @transmitq: queue for sent, non-acked messages
  102. * @backlogq: queue for messages waiting to be sent
  103. * @snt_nxt: next sequence number to use for outbound messages
  104. * @last_retransmitted: sequence number of most recently retransmitted message
  105. * @stale_count: # of identical retransmit requests made by peer
  106. * @ackers: # of peers that needs to ack each packet before it can be released
  107. * @acked: # last packet acked by a certain peer. Used for broadcast.
  108. * @rcv_nxt: next sequence number to expect for inbound messages
  109. * @deferred_queue: deferred queue saved OOS b'cast message received from node
  110. * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
  111. * @inputq: buffer queue for messages to be delivered upwards
  112. * @namedq: buffer queue for name table messages to be delivered upwards
  113. * @next_out: ptr to first unsent outbound message in queue
  114. * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
  115. * @long_msg_seq_no: next identifier to use for outbound fragmented messages
  116. * @reasm_buf: head of partially reassembled inbound message fragments
  117. * @bc_rcvr: marks that this is a broadcast receiver link
  118. * @stats: collects statistics regarding link activity
  119. */
  120. struct tipc_link {
  121. u32 addr;
  122. char name[TIPC_MAX_LINK_NAME];
  123. struct net *net;
  124. /* Management and link supervision data */
  125. u32 peer_session;
  126. u32 session;
  127. u32 peer_bearer_id;
  128. u32 bearer_id;
  129. u32 tolerance;
  130. u32 abort_limit;
  131. u32 state;
  132. u16 peer_caps;
  133. bool active;
  134. u32 silent_intv_cnt;
  135. char if_name[TIPC_MAX_IF_NAME];
  136. u32 priority;
  137. char net_plane;
  138. struct tipc_mon_state mon_state;
  139. u16 rst_cnt;
  140. /* Failover/synch */
  141. u16 drop_point;
  142. struct sk_buff *failover_reasm_skb;
  143. /* Max packet negotiation */
  144. u16 mtu;
  145. u16 advertised_mtu;
  146. /* Sending */
  147. struct sk_buff_head transmq;
  148. struct sk_buff_head backlogq;
  149. struct {
  150. u16 len;
  151. u16 limit;
  152. } backlog[5];
  153. u16 snd_nxt;
  154. u16 last_retransm;
  155. u16 window;
  156. u32 stale_count;
  157. /* Reception */
  158. u16 rcv_nxt;
  159. u32 rcv_unacked;
  160. struct sk_buff_head deferdq;
  161. struct sk_buff_head *inputq;
  162. struct sk_buff_head *namedq;
  163. /* Congestion handling */
  164. struct sk_buff_head wakeupq;
  165. /* Fragmentation/reassembly */
  166. struct sk_buff *reasm_buf;
  167. /* Broadcast */
  168. u16 ackers;
  169. u16 acked;
  170. struct tipc_link *bc_rcvlink;
  171. struct tipc_link *bc_sndlink;
  172. unsigned long prev_retr;
  173. u16 prev_from;
  174. u16 prev_to;
  175. u8 nack_state;
  176. bool bc_peer_is_up;
  177. /* Statistics */
  178. struct tipc_stats stats;
  179. };
  180. /*
  181. * Error message prefixes
  182. */
  183. static const char *link_co_err = "Link tunneling error, ";
  184. static const char *link_rst_msg = "Resetting link ";
  185. /* Send states for broadcast NACKs
  186. */
  187. enum {
  188. BC_NACK_SND_CONDITIONAL,
  189. BC_NACK_SND_UNCONDITIONAL,
  190. BC_NACK_SND_SUPPRESS,
  191. };
  192. #define TIPC_BC_RETR_LIMIT 10 /* [ms] */
  193. /*
  194. * Interval between NACKs when packets arrive out of order
  195. */
  196. #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
  197. /* Wildcard value for link session numbers. When it is known that
  198. * peer endpoint is down, any session number must be accepted.
  199. */
  200. #define ANY_SESSION 0x10000
  201. /* Link FSM states:
  202. */
  203. enum {
  204. LINK_ESTABLISHED = 0xe,
  205. LINK_ESTABLISHING = 0xe << 4,
  206. LINK_RESET = 0x1 << 8,
  207. LINK_RESETTING = 0x2 << 12,
  208. LINK_PEER_RESET = 0xd << 16,
  209. LINK_FAILINGOVER = 0xf << 20,
  210. LINK_SYNCHING = 0xc << 24
  211. };
  212. /* Link FSM state checking routines
  213. */
  214. static int link_is_up(struct tipc_link *l)
  215. {
  216. return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
  217. }
  218. static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
  219. struct sk_buff_head *xmitq);
  220. static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
  221. u16 rcvgap, int tolerance, int priority,
  222. struct sk_buff_head *xmitq);
  223. static void link_print(struct tipc_link *l, const char *str);
  224. static int tipc_link_build_nack_msg(struct tipc_link *l,
  225. struct sk_buff_head *xmitq);
  226. static void tipc_link_build_bc_init_msg(struct tipc_link *l,
  227. struct sk_buff_head *xmitq);
  228. static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
  229. /*
  230. * Simple non-static link routines (i.e. referenced outside this file)
  231. */
  232. bool tipc_link_is_up(struct tipc_link *l)
  233. {
  234. return link_is_up(l);
  235. }
  236. bool tipc_link_peer_is_down(struct tipc_link *l)
  237. {
  238. return l->state == LINK_PEER_RESET;
  239. }
  240. bool tipc_link_is_reset(struct tipc_link *l)
  241. {
  242. return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
  243. }
  244. bool tipc_link_is_establishing(struct tipc_link *l)
  245. {
  246. return l->state == LINK_ESTABLISHING;
  247. }
  248. bool tipc_link_is_synching(struct tipc_link *l)
  249. {
  250. return l->state == LINK_SYNCHING;
  251. }
  252. bool tipc_link_is_failingover(struct tipc_link *l)
  253. {
  254. return l->state == LINK_FAILINGOVER;
  255. }
  256. bool tipc_link_is_blocked(struct tipc_link *l)
  257. {
  258. return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
  259. }
  260. static bool link_is_bc_sndlink(struct tipc_link *l)
  261. {
  262. return !l->bc_sndlink;
  263. }
  264. static bool link_is_bc_rcvlink(struct tipc_link *l)
  265. {
  266. return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
  267. }
  268. int tipc_link_is_active(struct tipc_link *l)
  269. {
  270. return l->active;
  271. }
  272. void tipc_link_set_active(struct tipc_link *l, bool active)
  273. {
  274. l->active = active;
  275. }
  276. u32 tipc_link_id(struct tipc_link *l)
  277. {
  278. return l->peer_bearer_id << 16 | l->bearer_id;
  279. }
  280. int tipc_link_window(struct tipc_link *l)
  281. {
  282. return l->window;
  283. }
  284. int tipc_link_prio(struct tipc_link *l)
  285. {
  286. return l->priority;
  287. }
  288. unsigned long tipc_link_tolerance(struct tipc_link *l)
  289. {
  290. return l->tolerance;
  291. }
  292. struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
  293. {
  294. return l->inputq;
  295. }
  296. char tipc_link_plane(struct tipc_link *l)
  297. {
  298. return l->net_plane;
  299. }
  300. void tipc_link_add_bc_peer(struct tipc_link *snd_l,
  301. struct tipc_link *uc_l,
  302. struct sk_buff_head *xmitq)
  303. {
  304. struct tipc_link *rcv_l = uc_l->bc_rcvlink;
  305. snd_l->ackers++;
  306. rcv_l->acked = snd_l->snd_nxt - 1;
  307. snd_l->state = LINK_ESTABLISHED;
  308. tipc_link_build_bc_init_msg(uc_l, xmitq);
  309. }
  310. void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
  311. struct tipc_link *rcv_l,
  312. struct sk_buff_head *xmitq)
  313. {
  314. u16 ack = snd_l->snd_nxt - 1;
  315. snd_l->ackers--;
  316. rcv_l->bc_peer_is_up = true;
  317. rcv_l->state = LINK_ESTABLISHED;
  318. tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
  319. tipc_link_reset(rcv_l);
  320. rcv_l->state = LINK_RESET;
  321. if (!snd_l->ackers) {
  322. tipc_link_reset(snd_l);
  323. snd_l->state = LINK_RESET;
  324. __skb_queue_purge(xmitq);
  325. }
  326. }
  327. int tipc_link_bc_peers(struct tipc_link *l)
  328. {
  329. return l->ackers;
  330. }
  331. u16 link_bc_rcv_gap(struct tipc_link *l)
  332. {
  333. struct sk_buff *skb = skb_peek(&l->deferdq);
  334. u16 gap = 0;
  335. if (more(l->snd_nxt, l->rcv_nxt))
  336. gap = l->snd_nxt - l->rcv_nxt;
  337. if (skb)
  338. gap = buf_seqno(skb) - l->rcv_nxt;
  339. return gap;
  340. }
  341. void tipc_link_set_mtu(struct tipc_link *l, int mtu)
  342. {
  343. l->mtu = mtu;
  344. }
  345. int tipc_link_mtu(struct tipc_link *l)
  346. {
  347. return l->mtu;
  348. }
  349. u16 tipc_link_rcv_nxt(struct tipc_link *l)
  350. {
  351. return l->rcv_nxt;
  352. }
  353. u16 tipc_link_acked(struct tipc_link *l)
  354. {
  355. return l->acked;
  356. }
  357. char *tipc_link_name(struct tipc_link *l)
  358. {
  359. return l->name;
  360. }
  361. /**
  362. * tipc_link_create - create a new link
  363. * @n: pointer to associated node
  364. * @if_name: associated interface name
  365. * @bearer_id: id (index) of associated bearer
  366. * @tolerance: link tolerance to be used by link
  367. * @net_plane: network plane (A,B,c..) this link belongs to
  368. * @mtu: mtu to be advertised by link
  369. * @priority: priority to be used by link
  370. * @window: send window to be used by link
  371. * @session: session to be used by link
  372. * @ownnode: identity of own node
  373. * @peer: node id of peer node
  374. * @peer_caps: bitmap describing peer node capabilities
  375. * @bc_sndlink: the namespace global link used for broadcast sending
  376. * @bc_rcvlink: the peer specific link used for broadcast reception
  377. * @inputq: queue to put messages ready for delivery
  378. * @namedq: queue to put binding table update messages ready for delivery
  379. * @link: return value, pointer to put the created link
  380. *
  381. * Returns true if link was created, otherwise false
  382. */
  383. bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
  384. int tolerance, char net_plane, u32 mtu, int priority,
  385. int window, u32 session, u32 ownnode, u32 peer,
  386. u16 peer_caps,
  387. struct tipc_link *bc_sndlink,
  388. struct tipc_link *bc_rcvlink,
  389. struct sk_buff_head *inputq,
  390. struct sk_buff_head *namedq,
  391. struct tipc_link **link)
  392. {
  393. struct tipc_link *l;
  394. l = kzalloc(sizeof(*l), GFP_ATOMIC);
  395. if (!l)
  396. return false;
  397. *link = l;
  398. l->session = session;
  399. /* Note: peer i/f name is completed by reset/activate message */
  400. sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
  401. tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
  402. if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
  403. strcpy(l->if_name, if_name);
  404. l->addr = peer;
  405. l->peer_caps = peer_caps;
  406. l->net = net;
  407. l->peer_session = ANY_SESSION;
  408. l->bearer_id = bearer_id;
  409. l->tolerance = tolerance;
  410. l->net_plane = net_plane;
  411. l->advertised_mtu = mtu;
  412. l->mtu = mtu;
  413. l->priority = priority;
  414. tipc_link_set_queue_limits(l, window);
  415. l->ackers = 1;
  416. l->bc_sndlink = bc_sndlink;
  417. l->bc_rcvlink = bc_rcvlink;
  418. l->inputq = inputq;
  419. l->namedq = namedq;
  420. l->state = LINK_RESETTING;
  421. __skb_queue_head_init(&l->transmq);
  422. __skb_queue_head_init(&l->backlogq);
  423. __skb_queue_head_init(&l->deferdq);
  424. skb_queue_head_init(&l->wakeupq);
  425. skb_queue_head_init(l->inputq);
  426. return true;
  427. }
  428. /**
  429. * tipc_link_bc_create - create new link to be used for broadcast
  430. * @n: pointer to associated node
  431. * @mtu: mtu to be used
  432. * @window: send window to be used
  433. * @inputq: queue to put messages ready for delivery
  434. * @namedq: queue to put binding table update messages ready for delivery
  435. * @link: return value, pointer to put the created link
  436. *
  437. * Returns true if link was created, otherwise false
  438. */
  439. bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
  440. int mtu, int window, u16 peer_caps,
  441. struct sk_buff_head *inputq,
  442. struct sk_buff_head *namedq,
  443. struct tipc_link *bc_sndlink,
  444. struct tipc_link **link)
  445. {
  446. struct tipc_link *l;
  447. if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
  448. 0, ownnode, peer, peer_caps, bc_sndlink,
  449. NULL, inputq, namedq, link))
  450. return false;
  451. l = *link;
  452. strcpy(l->name, tipc_bclink_name);
  453. tipc_link_reset(l);
  454. l->state = LINK_RESET;
  455. l->ackers = 0;
  456. l->bc_rcvlink = l;
  457. /* Broadcast send link is always up */
  458. if (link_is_bc_sndlink(l))
  459. l->state = LINK_ESTABLISHED;
  460. /* Disable replicast if even a single peer doesn't support it */
  461. if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
  462. tipc_bcast_disable_rcast(net);
  463. return true;
  464. }
  465. /**
  466. * tipc_link_fsm_evt - link finite state machine
  467. * @l: pointer to link
  468. * @evt: state machine event to be processed
  469. */
  470. int tipc_link_fsm_evt(struct tipc_link *l, int evt)
  471. {
  472. int rc = 0;
  473. switch (l->state) {
  474. case LINK_RESETTING:
  475. switch (evt) {
  476. case LINK_PEER_RESET_EVT:
  477. l->state = LINK_PEER_RESET;
  478. break;
  479. case LINK_RESET_EVT:
  480. l->state = LINK_RESET;
  481. break;
  482. case LINK_FAILURE_EVT:
  483. case LINK_FAILOVER_BEGIN_EVT:
  484. case LINK_ESTABLISH_EVT:
  485. case LINK_FAILOVER_END_EVT:
  486. case LINK_SYNCH_BEGIN_EVT:
  487. case LINK_SYNCH_END_EVT:
  488. default:
  489. goto illegal_evt;
  490. }
  491. break;
  492. case LINK_RESET:
  493. switch (evt) {
  494. case LINK_PEER_RESET_EVT:
  495. l->state = LINK_ESTABLISHING;
  496. break;
  497. case LINK_FAILOVER_BEGIN_EVT:
  498. l->state = LINK_FAILINGOVER;
  499. case LINK_FAILURE_EVT:
  500. case LINK_RESET_EVT:
  501. case LINK_ESTABLISH_EVT:
  502. case LINK_FAILOVER_END_EVT:
  503. break;
  504. case LINK_SYNCH_BEGIN_EVT:
  505. case LINK_SYNCH_END_EVT:
  506. default:
  507. goto illegal_evt;
  508. }
  509. break;
  510. case LINK_PEER_RESET:
  511. switch (evt) {
  512. case LINK_RESET_EVT:
  513. l->state = LINK_ESTABLISHING;
  514. break;
  515. case LINK_PEER_RESET_EVT:
  516. case LINK_ESTABLISH_EVT:
  517. case LINK_FAILURE_EVT:
  518. break;
  519. case LINK_SYNCH_BEGIN_EVT:
  520. case LINK_SYNCH_END_EVT:
  521. case LINK_FAILOVER_BEGIN_EVT:
  522. case LINK_FAILOVER_END_EVT:
  523. default:
  524. goto illegal_evt;
  525. }
  526. break;
  527. case LINK_FAILINGOVER:
  528. switch (evt) {
  529. case LINK_FAILOVER_END_EVT:
  530. l->state = LINK_RESET;
  531. break;
  532. case LINK_PEER_RESET_EVT:
  533. case LINK_RESET_EVT:
  534. case LINK_ESTABLISH_EVT:
  535. case LINK_FAILURE_EVT:
  536. break;
  537. case LINK_FAILOVER_BEGIN_EVT:
  538. case LINK_SYNCH_BEGIN_EVT:
  539. case LINK_SYNCH_END_EVT:
  540. default:
  541. goto illegal_evt;
  542. }
  543. break;
  544. case LINK_ESTABLISHING:
  545. switch (evt) {
  546. case LINK_ESTABLISH_EVT:
  547. l->state = LINK_ESTABLISHED;
  548. break;
  549. case LINK_FAILOVER_BEGIN_EVT:
  550. l->state = LINK_FAILINGOVER;
  551. break;
  552. case LINK_RESET_EVT:
  553. l->state = LINK_RESET;
  554. break;
  555. case LINK_FAILURE_EVT:
  556. case LINK_PEER_RESET_EVT:
  557. case LINK_SYNCH_BEGIN_EVT:
  558. case LINK_FAILOVER_END_EVT:
  559. break;
  560. case LINK_SYNCH_END_EVT:
  561. default:
  562. goto illegal_evt;
  563. }
  564. break;
  565. case LINK_ESTABLISHED:
  566. switch (evt) {
  567. case LINK_PEER_RESET_EVT:
  568. l->state = LINK_PEER_RESET;
  569. rc |= TIPC_LINK_DOWN_EVT;
  570. break;
  571. case LINK_FAILURE_EVT:
  572. l->state = LINK_RESETTING;
  573. rc |= TIPC_LINK_DOWN_EVT;
  574. break;
  575. case LINK_RESET_EVT:
  576. l->state = LINK_RESET;
  577. break;
  578. case LINK_ESTABLISH_EVT:
  579. case LINK_SYNCH_END_EVT:
  580. break;
  581. case LINK_SYNCH_BEGIN_EVT:
  582. l->state = LINK_SYNCHING;
  583. break;
  584. case LINK_FAILOVER_BEGIN_EVT:
  585. case LINK_FAILOVER_END_EVT:
  586. default:
  587. goto illegal_evt;
  588. }
  589. break;
  590. case LINK_SYNCHING:
  591. switch (evt) {
  592. case LINK_PEER_RESET_EVT:
  593. l->state = LINK_PEER_RESET;
  594. rc |= TIPC_LINK_DOWN_EVT;
  595. break;
  596. case LINK_FAILURE_EVT:
  597. l->state = LINK_RESETTING;
  598. rc |= TIPC_LINK_DOWN_EVT;
  599. break;
  600. case LINK_RESET_EVT:
  601. l->state = LINK_RESET;
  602. break;
  603. case LINK_ESTABLISH_EVT:
  604. case LINK_SYNCH_BEGIN_EVT:
  605. break;
  606. case LINK_SYNCH_END_EVT:
  607. l->state = LINK_ESTABLISHED;
  608. break;
  609. case LINK_FAILOVER_BEGIN_EVT:
  610. case LINK_FAILOVER_END_EVT:
  611. default:
  612. goto illegal_evt;
  613. }
  614. break;
  615. default:
  616. pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
  617. }
  618. return rc;
  619. illegal_evt:
  620. pr_err("Illegal FSM event %x in state %x on link %s\n",
  621. evt, l->state, l->name);
  622. return rc;
  623. }
  624. /* link_profile_stats - update statistical profiling of traffic
  625. */
  626. static void link_profile_stats(struct tipc_link *l)
  627. {
  628. struct sk_buff *skb;
  629. struct tipc_msg *msg;
  630. int length;
  631. /* Update counters used in statistical profiling of send traffic */
  632. l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
  633. l->stats.queue_sz_counts++;
  634. skb = skb_peek(&l->transmq);
  635. if (!skb)
  636. return;
  637. msg = buf_msg(skb);
  638. length = msg_size(msg);
  639. if (msg_user(msg) == MSG_FRAGMENTER) {
  640. if (msg_type(msg) != FIRST_FRAGMENT)
  641. return;
  642. length = msg_size(msg_get_wrapped(msg));
  643. }
  644. l->stats.msg_lengths_total += length;
  645. l->stats.msg_length_counts++;
  646. if (length <= 64)
  647. l->stats.msg_length_profile[0]++;
  648. else if (length <= 256)
  649. l->stats.msg_length_profile[1]++;
  650. else if (length <= 1024)
  651. l->stats.msg_length_profile[2]++;
  652. else if (length <= 4096)
  653. l->stats.msg_length_profile[3]++;
  654. else if (length <= 16384)
  655. l->stats.msg_length_profile[4]++;
  656. else if (length <= 32768)
  657. l->stats.msg_length_profile[5]++;
  658. else
  659. l->stats.msg_length_profile[6]++;
  660. }
  661. /* tipc_link_timeout - perform periodic task as instructed from node timeout
  662. */
  663. int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
  664. {
  665. int mtyp = 0;
  666. int rc = 0;
  667. bool state = false;
  668. bool probe = false;
  669. bool setup = false;
  670. u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
  671. u16 bc_acked = l->bc_rcvlink->acked;
  672. struct tipc_mon_state *mstate = &l->mon_state;
  673. switch (l->state) {
  674. case LINK_ESTABLISHED:
  675. case LINK_SYNCHING:
  676. mtyp = STATE_MSG;
  677. link_profile_stats(l);
  678. tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
  679. if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
  680. return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
  681. state = bc_acked != bc_snt;
  682. state |= l->bc_rcvlink->rcv_unacked;
  683. state |= l->rcv_unacked;
  684. state |= !skb_queue_empty(&l->transmq);
  685. state |= !skb_queue_empty(&l->deferdq);
  686. probe = mstate->probing;
  687. probe |= l->silent_intv_cnt;
  688. if (probe || mstate->monitoring)
  689. l->silent_intv_cnt++;
  690. break;
  691. case LINK_RESET:
  692. setup = l->rst_cnt++ <= 4;
  693. setup |= !(l->rst_cnt % 16);
  694. mtyp = RESET_MSG;
  695. break;
  696. case LINK_ESTABLISHING:
  697. setup = true;
  698. mtyp = ACTIVATE_MSG;
  699. break;
  700. case LINK_PEER_RESET:
  701. case LINK_RESETTING:
  702. case LINK_FAILINGOVER:
  703. break;
  704. default:
  705. break;
  706. }
  707. if (state || probe || setup)
  708. tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, xmitq);
  709. return rc;
  710. }
  711. /**
  712. * link_schedule_user - schedule a message sender for wakeup after congestion
  713. * @l: congested link
  714. * @hdr: header of message that is being sent
  715. * Create pseudo msg to send back to user when congestion abates
  716. */
  717. static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
  718. {
  719. u32 dnode = tipc_own_addr(l->net);
  720. u32 dport = msg_origport(hdr);
  721. struct sk_buff *skb;
  722. /* Create and schedule wakeup pseudo message */
  723. skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
  724. dnode, l->addr, dport, 0, 0);
  725. if (!skb)
  726. return -ENOBUFS;
  727. msg_set_dest_droppable(buf_msg(skb), true);
  728. TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
  729. skb_queue_tail(&l->wakeupq, skb);
  730. l->stats.link_congs++;
  731. return -ELINKCONG;
  732. }
  733. /**
  734. * link_prepare_wakeup - prepare users for wakeup after congestion
  735. * @l: congested link
  736. * Wake up a number of waiting users, as permitted by available space
  737. * in the send queue
  738. */
  739. void link_prepare_wakeup(struct tipc_link *l)
  740. {
  741. struct sk_buff *skb, *tmp;
  742. int imp, i = 0;
  743. skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
  744. imp = TIPC_SKB_CB(skb)->chain_imp;
  745. if (l->backlog[imp].len < l->backlog[imp].limit) {
  746. skb_unlink(skb, &l->wakeupq);
  747. skb_queue_tail(l->inputq, skb);
  748. } else if (i++ > 10) {
  749. break;
  750. }
  751. }
  752. }
  753. void tipc_link_reset(struct tipc_link *l)
  754. {
  755. l->peer_session = ANY_SESSION;
  756. l->session++;
  757. l->mtu = l->advertised_mtu;
  758. __skb_queue_purge(&l->transmq);
  759. __skb_queue_purge(&l->deferdq);
  760. skb_queue_splice_init(&l->wakeupq, l->inputq);
  761. __skb_queue_purge(&l->backlogq);
  762. l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
  763. l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
  764. l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
  765. l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
  766. l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
  767. kfree_skb(l->reasm_buf);
  768. kfree_skb(l->failover_reasm_skb);
  769. l->reasm_buf = NULL;
  770. l->failover_reasm_skb = NULL;
  771. l->rcv_unacked = 0;
  772. l->snd_nxt = 1;
  773. l->rcv_nxt = 1;
  774. l->acked = 0;
  775. l->silent_intv_cnt = 0;
  776. l->rst_cnt = 0;
  777. l->stale_count = 0;
  778. l->bc_peer_is_up = false;
  779. memset(&l->mon_state, 0, sizeof(l->mon_state));
  780. tipc_link_reset_stats(l);
  781. }
  782. /**
  783. * tipc_link_xmit(): enqueue buffer list according to queue situation
  784. * @link: link to use
  785. * @list: chain of buffers containing message
  786. * @xmitq: returned list of packets to be sent by caller
  787. *
  788. * Consumes the buffer chain.
  789. * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
  790. * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
  791. */
  792. int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
  793. struct sk_buff_head *xmitq)
  794. {
  795. struct tipc_msg *hdr = buf_msg(skb_peek(list));
  796. unsigned int maxwin = l->window;
  797. int imp = msg_importance(hdr);
  798. unsigned int mtu = l->mtu;
  799. u16 ack = l->rcv_nxt - 1;
  800. u16 seqno = l->snd_nxt;
  801. u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
  802. struct sk_buff_head *transmq = &l->transmq;
  803. struct sk_buff_head *backlogq = &l->backlogq;
  804. struct sk_buff *skb, *_skb, *bskb;
  805. int pkt_cnt = skb_queue_len(list);
  806. int rc = 0;
  807. if (unlikely(msg_size(hdr) > mtu)) {
  808. skb_queue_purge(list);
  809. return -EMSGSIZE;
  810. }
  811. /* Allow oversubscription of one data msg per source at congestion */
  812. if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
  813. if (imp == TIPC_SYSTEM_IMPORTANCE) {
  814. pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
  815. return -ENOBUFS;
  816. }
  817. rc = link_schedule_user(l, hdr);
  818. }
  819. if (pkt_cnt > 1) {
  820. l->stats.sent_fragmented++;
  821. l->stats.sent_fragments += pkt_cnt;
  822. }
  823. /* Prepare each packet for sending, and add to relevant queue: */
  824. while (skb_queue_len(list)) {
  825. skb = skb_peek(list);
  826. hdr = buf_msg(skb);
  827. msg_set_seqno(hdr, seqno);
  828. msg_set_ack(hdr, ack);
  829. msg_set_bcast_ack(hdr, bc_ack);
  830. if (likely(skb_queue_len(transmq) < maxwin)) {
  831. _skb = skb_clone(skb, GFP_ATOMIC);
  832. if (!_skb) {
  833. skb_queue_purge(list);
  834. return -ENOBUFS;
  835. }
  836. __skb_dequeue(list);
  837. __skb_queue_tail(transmq, skb);
  838. __skb_queue_tail(xmitq, _skb);
  839. TIPC_SKB_CB(skb)->ackers = l->ackers;
  840. l->rcv_unacked = 0;
  841. l->stats.sent_pkts++;
  842. seqno++;
  843. continue;
  844. }
  845. if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
  846. kfree_skb(__skb_dequeue(list));
  847. l->stats.sent_bundled++;
  848. continue;
  849. }
  850. if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
  851. kfree_skb(__skb_dequeue(list));
  852. __skb_queue_tail(backlogq, bskb);
  853. l->backlog[msg_importance(buf_msg(bskb))].len++;
  854. l->stats.sent_bundled++;
  855. l->stats.sent_bundles++;
  856. continue;
  857. }
  858. l->backlog[imp].len += skb_queue_len(list);
  859. skb_queue_splice_tail_init(list, backlogq);
  860. }
  861. l->snd_nxt = seqno;
  862. return rc;
  863. }
  864. void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
  865. {
  866. struct sk_buff *skb, *_skb;
  867. struct tipc_msg *hdr;
  868. u16 seqno = l->snd_nxt;
  869. u16 ack = l->rcv_nxt - 1;
  870. u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
  871. while (skb_queue_len(&l->transmq) < l->window) {
  872. skb = skb_peek(&l->backlogq);
  873. if (!skb)
  874. break;
  875. _skb = skb_clone(skb, GFP_ATOMIC);
  876. if (!_skb)
  877. break;
  878. __skb_dequeue(&l->backlogq);
  879. hdr = buf_msg(skb);
  880. l->backlog[msg_importance(hdr)].len--;
  881. __skb_queue_tail(&l->transmq, skb);
  882. __skb_queue_tail(xmitq, _skb);
  883. TIPC_SKB_CB(skb)->ackers = l->ackers;
  884. msg_set_seqno(hdr, seqno);
  885. msg_set_ack(hdr, ack);
  886. msg_set_bcast_ack(hdr, bc_ack);
  887. l->rcv_unacked = 0;
  888. l->stats.sent_pkts++;
  889. seqno++;
  890. }
  891. l->snd_nxt = seqno;
  892. }
  893. static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
  894. {
  895. struct tipc_msg *hdr = buf_msg(skb);
  896. pr_warn("Retransmission failure on link <%s>\n", l->name);
  897. link_print(l, "Resetting link ");
  898. pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
  899. msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
  900. pr_info("sqno %u, prev: %x, src: %x\n",
  901. msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
  902. }
  903. int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
  904. struct sk_buff_head *xmitq)
  905. {
  906. struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
  907. struct tipc_msg *hdr;
  908. u16 ack = l->rcv_nxt - 1;
  909. u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
  910. if (!skb)
  911. return 0;
  912. /* Detect repeated retransmit failures on same packet */
  913. if (likely(l->last_retransm != buf_seqno(skb))) {
  914. l->last_retransm = buf_seqno(skb);
  915. l->stale_count = 1;
  916. } else if (++l->stale_count > 100) {
  917. link_retransmit_failure(l, skb);
  918. return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
  919. }
  920. /* Move forward to where retransmission should start */
  921. skb_queue_walk(&l->transmq, skb) {
  922. if (!less(buf_seqno(skb), from))
  923. break;
  924. }
  925. skb_queue_walk_from(&l->transmq, skb) {
  926. if (more(buf_seqno(skb), to))
  927. break;
  928. hdr = buf_msg(skb);
  929. _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
  930. if (!_skb)
  931. return 0;
  932. hdr = buf_msg(_skb);
  933. msg_set_ack(hdr, ack);
  934. msg_set_bcast_ack(hdr, bc_ack);
  935. _skb->priority = TC_PRIO_CONTROL;
  936. __skb_queue_tail(xmitq, _skb);
  937. l->stats.retransmitted++;
  938. }
  939. return 0;
  940. }
  941. /* tipc_data_input - deliver data and name distr msgs to upper layer
  942. *
  943. * Consumes buffer if message is of right type
  944. * Node lock must be held
  945. */
  946. static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
  947. struct sk_buff_head *inputq)
  948. {
  949. struct tipc_msg *hdr = buf_msg(skb);
  950. switch (msg_user(hdr)) {
  951. case TIPC_LOW_IMPORTANCE:
  952. case TIPC_MEDIUM_IMPORTANCE:
  953. case TIPC_HIGH_IMPORTANCE:
  954. case TIPC_CRITICAL_IMPORTANCE:
  955. if (unlikely(msg_type(hdr) == TIPC_MCAST_MSG)) {
  956. skb_queue_tail(l->bc_rcvlink->inputq, skb);
  957. return true;
  958. }
  959. case CONN_MANAGER:
  960. skb_queue_tail(inputq, skb);
  961. return true;
  962. case NAME_DISTRIBUTOR:
  963. l->bc_rcvlink->state = LINK_ESTABLISHED;
  964. skb_queue_tail(l->namedq, skb);
  965. return true;
  966. case MSG_BUNDLER:
  967. case TUNNEL_PROTOCOL:
  968. case MSG_FRAGMENTER:
  969. case BCAST_PROTOCOL:
  970. return false;
  971. default:
  972. pr_warn("Dropping received illegal msg type\n");
  973. kfree_skb(skb);
  974. return false;
  975. };
  976. }
  977. /* tipc_link_input - process packet that has passed link protocol check
  978. *
  979. * Consumes buffer
  980. */
  981. static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
  982. struct sk_buff_head *inputq)
  983. {
  984. struct tipc_msg *hdr = buf_msg(skb);
  985. struct sk_buff **reasm_skb = &l->reasm_buf;
  986. struct sk_buff *iskb;
  987. struct sk_buff_head tmpq;
  988. int usr = msg_user(hdr);
  989. int rc = 0;
  990. int pos = 0;
  991. int ipos = 0;
  992. if (unlikely(usr == TUNNEL_PROTOCOL)) {
  993. if (msg_type(hdr) == SYNCH_MSG) {
  994. __skb_queue_purge(&l->deferdq);
  995. goto drop;
  996. }
  997. if (!tipc_msg_extract(skb, &iskb, &ipos))
  998. return rc;
  999. kfree_skb(skb);
  1000. skb = iskb;
  1001. hdr = buf_msg(skb);
  1002. if (less(msg_seqno(hdr), l->drop_point))
  1003. goto drop;
  1004. if (tipc_data_input(l, skb, inputq))
  1005. return rc;
  1006. usr = msg_user(hdr);
  1007. reasm_skb = &l->failover_reasm_skb;
  1008. }
  1009. if (usr == MSG_BUNDLER) {
  1010. skb_queue_head_init(&tmpq);
  1011. l->stats.recv_bundles++;
  1012. l->stats.recv_bundled += msg_msgcnt(hdr);
  1013. while (tipc_msg_extract(skb, &iskb, &pos))
  1014. tipc_data_input(l, iskb, &tmpq);
  1015. tipc_skb_queue_splice_tail(&tmpq, inputq);
  1016. return 0;
  1017. } else if (usr == MSG_FRAGMENTER) {
  1018. l->stats.recv_fragments++;
  1019. if (tipc_buf_append(reasm_skb, &skb)) {
  1020. l->stats.recv_fragmented++;
  1021. tipc_data_input(l, skb, inputq);
  1022. } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
  1023. pr_warn_ratelimited("Unable to build fragment list\n");
  1024. return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
  1025. }
  1026. return 0;
  1027. } else if (usr == BCAST_PROTOCOL) {
  1028. tipc_bcast_lock(l->net);
  1029. tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
  1030. tipc_bcast_unlock(l->net);
  1031. }
  1032. drop:
  1033. kfree_skb(skb);
  1034. return 0;
  1035. }
  1036. static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
  1037. {
  1038. bool released = false;
  1039. struct sk_buff *skb, *tmp;
  1040. skb_queue_walk_safe(&l->transmq, skb, tmp) {
  1041. if (more(buf_seqno(skb), acked))
  1042. break;
  1043. __skb_unlink(skb, &l->transmq);
  1044. kfree_skb(skb);
  1045. released = true;
  1046. }
  1047. return released;
  1048. }
  1049. /* tipc_link_build_state_msg: prepare link state message for transmission
  1050. *
  1051. * Note that sending of broadcast ack is coordinated among nodes, to reduce
  1052. * risk of ack storms towards the sender
  1053. */
  1054. int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
  1055. {
  1056. if (!l)
  1057. return 0;
  1058. /* Broadcast ACK must be sent via a unicast link => defer to caller */
  1059. if (link_is_bc_rcvlink(l)) {
  1060. if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
  1061. return 0;
  1062. l->rcv_unacked = 0;
  1063. /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
  1064. l->snd_nxt = l->rcv_nxt;
  1065. return TIPC_LINK_SND_STATE;
  1066. }
  1067. /* Unicast ACK */
  1068. l->rcv_unacked = 0;
  1069. l->stats.sent_acks++;
  1070. tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
  1071. return 0;
  1072. }
  1073. /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
  1074. */
  1075. void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
  1076. {
  1077. int mtyp = RESET_MSG;
  1078. struct sk_buff *skb;
  1079. if (l->state == LINK_ESTABLISHING)
  1080. mtyp = ACTIVATE_MSG;
  1081. tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
  1082. /* Inform peer that this endpoint is going down if applicable */
  1083. skb = skb_peek_tail(xmitq);
  1084. if (skb && (l->state == LINK_RESET))
  1085. msg_set_peer_stopping(buf_msg(skb), 1);
  1086. }
  1087. /* tipc_link_build_nack_msg: prepare link nack message for transmission
  1088. * Note that sending of broadcast NACK is coordinated among nodes, to
  1089. * reduce the risk of NACK storms towards the sender
  1090. */
  1091. static int tipc_link_build_nack_msg(struct tipc_link *l,
  1092. struct sk_buff_head *xmitq)
  1093. {
  1094. u32 def_cnt = ++l->stats.deferred_recv;
  1095. int match1, match2;
  1096. if (link_is_bc_rcvlink(l)) {
  1097. match1 = def_cnt & 0xf;
  1098. match2 = tipc_own_addr(l->net) & 0xf;
  1099. if (match1 == match2)
  1100. return TIPC_LINK_SND_STATE;
  1101. return 0;
  1102. }
  1103. if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
  1104. tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
  1105. return 0;
  1106. }
  1107. /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
  1108. * @l: the link that should handle the message
  1109. * @skb: TIPC packet
  1110. * @xmitq: queue to place packets to be sent after this call
  1111. */
  1112. int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
  1113. struct sk_buff_head *xmitq)
  1114. {
  1115. struct sk_buff_head *defq = &l->deferdq;
  1116. struct tipc_msg *hdr;
  1117. u16 seqno, rcv_nxt, win_lim;
  1118. int rc = 0;
  1119. do {
  1120. hdr = buf_msg(skb);
  1121. seqno = msg_seqno(hdr);
  1122. rcv_nxt = l->rcv_nxt;
  1123. win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
  1124. /* Verify and update link state */
  1125. if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
  1126. return tipc_link_proto_rcv(l, skb, xmitq);
  1127. if (unlikely(!link_is_up(l))) {
  1128. if (l->state == LINK_ESTABLISHING)
  1129. rc = TIPC_LINK_UP_EVT;
  1130. goto drop;
  1131. }
  1132. /* Don't send probe at next timeout expiration */
  1133. l->silent_intv_cnt = 0;
  1134. /* Drop if outside receive window */
  1135. if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
  1136. l->stats.duplicates++;
  1137. goto drop;
  1138. }
  1139. /* Forward queues and wake up waiting users */
  1140. if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
  1141. tipc_link_advance_backlog(l, xmitq);
  1142. if (unlikely(!skb_queue_empty(&l->wakeupq)))
  1143. link_prepare_wakeup(l);
  1144. }
  1145. /* Defer delivery if sequence gap */
  1146. if (unlikely(seqno != rcv_nxt)) {
  1147. __tipc_skb_queue_sorted(defq, seqno, skb);
  1148. rc |= tipc_link_build_nack_msg(l, xmitq);
  1149. break;
  1150. }
  1151. /* Deliver packet */
  1152. l->rcv_nxt++;
  1153. l->stats.recv_pkts++;
  1154. if (!tipc_data_input(l, skb, l->inputq))
  1155. rc |= tipc_link_input(l, skb, l->inputq);
  1156. if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
  1157. rc |= tipc_link_build_state_msg(l, xmitq);
  1158. if (unlikely(rc & ~TIPC_LINK_SND_STATE))
  1159. break;
  1160. } while ((skb = __skb_dequeue(defq)));
  1161. return rc;
  1162. drop:
  1163. kfree_skb(skb);
  1164. return rc;
  1165. }
  1166. static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
  1167. u16 rcvgap, int tolerance, int priority,
  1168. struct sk_buff_head *xmitq)
  1169. {
  1170. struct tipc_link *bcl = l->bc_rcvlink;
  1171. struct sk_buff *skb;
  1172. struct tipc_msg *hdr;
  1173. struct sk_buff_head *dfq = &l->deferdq;
  1174. bool node_up = link_is_up(bcl);
  1175. struct tipc_mon_state *mstate = &l->mon_state;
  1176. int dlen = 0;
  1177. void *data;
  1178. /* Don't send protocol message during reset or link failover */
  1179. if (tipc_link_is_blocked(l))
  1180. return;
  1181. if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
  1182. return;
  1183. if (!skb_queue_empty(dfq))
  1184. rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
  1185. skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
  1186. tipc_max_domain_size, l->addr,
  1187. tipc_own_addr(l->net), 0, 0, 0);
  1188. if (!skb)
  1189. return;
  1190. hdr = buf_msg(skb);
  1191. data = msg_data(hdr);
  1192. msg_set_session(hdr, l->session);
  1193. msg_set_bearer_id(hdr, l->bearer_id);
  1194. msg_set_net_plane(hdr, l->net_plane);
  1195. msg_set_next_sent(hdr, l->snd_nxt);
  1196. msg_set_ack(hdr, l->rcv_nxt - 1);
  1197. msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
  1198. msg_set_bc_ack_invalid(hdr, !node_up);
  1199. msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
  1200. msg_set_link_tolerance(hdr, tolerance);
  1201. msg_set_linkprio(hdr, priority);
  1202. msg_set_redundant_link(hdr, node_up);
  1203. msg_set_seq_gap(hdr, 0);
  1204. msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
  1205. if (mtyp == STATE_MSG) {
  1206. msg_set_seq_gap(hdr, rcvgap);
  1207. msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
  1208. msg_set_probe(hdr, probe);
  1209. tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
  1210. msg_set_size(hdr, INT_H_SIZE + dlen);
  1211. skb_trim(skb, INT_H_SIZE + dlen);
  1212. l->stats.sent_states++;
  1213. l->rcv_unacked = 0;
  1214. } else {
  1215. /* RESET_MSG or ACTIVATE_MSG */
  1216. msg_set_max_pkt(hdr, l->advertised_mtu);
  1217. strcpy(data, l->if_name);
  1218. msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
  1219. skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
  1220. }
  1221. if (probe)
  1222. l->stats.sent_probes++;
  1223. if (rcvgap)
  1224. l->stats.sent_nacks++;
  1225. skb->priority = TC_PRIO_CONTROL;
  1226. __skb_queue_tail(xmitq, skb);
  1227. }
  1228. /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
  1229. * with contents of the link's transmit and backlog queues.
  1230. */
  1231. void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
  1232. int mtyp, struct sk_buff_head *xmitq)
  1233. {
  1234. struct sk_buff *skb, *tnlskb;
  1235. struct tipc_msg *hdr, tnlhdr;
  1236. struct sk_buff_head *queue = &l->transmq;
  1237. struct sk_buff_head tmpxq, tnlq;
  1238. u16 pktlen, pktcnt, seqno = l->snd_nxt;
  1239. if (!tnl)
  1240. return;
  1241. skb_queue_head_init(&tnlq);
  1242. skb_queue_head_init(&tmpxq);
  1243. /* At least one packet required for safe algorithm => add dummy */
  1244. skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
  1245. BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
  1246. 0, 0, TIPC_ERR_NO_PORT);
  1247. if (!skb) {
  1248. pr_warn("%sunable to create tunnel packet\n", link_co_err);
  1249. return;
  1250. }
  1251. skb_queue_tail(&tnlq, skb);
  1252. tipc_link_xmit(l, &tnlq, &tmpxq);
  1253. __skb_queue_purge(&tmpxq);
  1254. /* Initialize reusable tunnel packet header */
  1255. tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
  1256. mtyp, INT_H_SIZE, l->addr);
  1257. pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
  1258. msg_set_msgcnt(&tnlhdr, pktcnt);
  1259. msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
  1260. tnl:
  1261. /* Wrap each packet into a tunnel packet */
  1262. skb_queue_walk(queue, skb) {
  1263. hdr = buf_msg(skb);
  1264. if (queue == &l->backlogq)
  1265. msg_set_seqno(hdr, seqno++);
  1266. pktlen = msg_size(hdr);
  1267. msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
  1268. tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
  1269. if (!tnlskb) {
  1270. pr_warn("%sunable to send packet\n", link_co_err);
  1271. return;
  1272. }
  1273. skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
  1274. skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
  1275. __skb_queue_tail(&tnlq, tnlskb);
  1276. }
  1277. if (queue != &l->backlogq) {
  1278. queue = &l->backlogq;
  1279. goto tnl;
  1280. }
  1281. tipc_link_xmit(tnl, &tnlq, xmitq);
  1282. if (mtyp == FAILOVER_MSG) {
  1283. tnl->drop_point = l->rcv_nxt;
  1284. tnl->failover_reasm_skb = l->reasm_buf;
  1285. l->reasm_buf = NULL;
  1286. }
  1287. }
  1288. /* tipc_link_proto_rcv(): receive link level protocol message :
  1289. * Note that network plane id propagates through the network, and may
  1290. * change at any time. The node with lowest numerical id determines
  1291. * network plane
  1292. */
  1293. static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
  1294. struct sk_buff_head *xmitq)
  1295. {
  1296. struct tipc_msg *hdr = buf_msg(skb);
  1297. u16 rcvgap = 0;
  1298. u16 ack = msg_ack(hdr);
  1299. u16 gap = msg_seq_gap(hdr);
  1300. u16 peers_snd_nxt = msg_next_sent(hdr);
  1301. u16 peers_tol = msg_link_tolerance(hdr);
  1302. u16 peers_prio = msg_linkprio(hdr);
  1303. u16 rcv_nxt = l->rcv_nxt;
  1304. u16 dlen = msg_data_sz(hdr);
  1305. int mtyp = msg_type(hdr);
  1306. void *data;
  1307. char *if_name;
  1308. int rc = 0;
  1309. if (tipc_link_is_blocked(l) || !xmitq)
  1310. goto exit;
  1311. if (tipc_own_addr(l->net) > msg_prevnode(hdr))
  1312. l->net_plane = msg_net_plane(hdr);
  1313. skb_linearize(skb);
  1314. hdr = buf_msg(skb);
  1315. data = msg_data(hdr);
  1316. switch (mtyp) {
  1317. case RESET_MSG:
  1318. /* Ignore duplicate RESET with old session number */
  1319. if ((less_eq(msg_session(hdr), l->peer_session)) &&
  1320. (l->peer_session != ANY_SESSION))
  1321. break;
  1322. /* fall thru' */
  1323. case ACTIVATE_MSG:
  1324. /* Complete own link name with peer's interface name */
  1325. if_name = strrchr(l->name, ':') + 1;
  1326. if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
  1327. break;
  1328. if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
  1329. break;
  1330. strncpy(if_name, data, TIPC_MAX_IF_NAME);
  1331. /* Update own tolerance if peer indicates a non-zero value */
  1332. if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
  1333. l->tolerance = peers_tol;
  1334. /* Update own priority if peer's priority is higher */
  1335. if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
  1336. l->priority = peers_prio;
  1337. /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
  1338. if (msg_peer_stopping(hdr))
  1339. rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
  1340. else if ((mtyp == RESET_MSG) || !link_is_up(l))
  1341. rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
  1342. /* ACTIVATE_MSG takes up link if it was already locally reset */
  1343. if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
  1344. rc = TIPC_LINK_UP_EVT;
  1345. l->peer_session = msg_session(hdr);
  1346. l->peer_bearer_id = msg_bearer_id(hdr);
  1347. if (l->mtu > msg_max_pkt(hdr))
  1348. l->mtu = msg_max_pkt(hdr);
  1349. break;
  1350. case STATE_MSG:
  1351. /* Update own tolerance if peer indicates a non-zero value */
  1352. if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
  1353. l->tolerance = peers_tol;
  1354. /* Update own prio if peer indicates a different value */
  1355. if ((peers_prio != l->priority) &&
  1356. in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
  1357. l->priority = peers_prio;
  1358. rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
  1359. }
  1360. l->silent_intv_cnt = 0;
  1361. l->stats.recv_states++;
  1362. if (msg_probe(hdr))
  1363. l->stats.recv_probes++;
  1364. if (!link_is_up(l)) {
  1365. if (l->state == LINK_ESTABLISHING)
  1366. rc = TIPC_LINK_UP_EVT;
  1367. break;
  1368. }
  1369. tipc_mon_rcv(l->net, data, dlen, l->addr,
  1370. &l->mon_state, l->bearer_id);
  1371. /* Send NACK if peer has sent pkts we haven't received yet */
  1372. if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
  1373. rcvgap = peers_snd_nxt - l->rcv_nxt;
  1374. if (rcvgap || (msg_probe(hdr)))
  1375. tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
  1376. 0, 0, xmitq);
  1377. tipc_link_release_pkts(l, ack);
  1378. /* If NACK, retransmit will now start at right position */
  1379. if (gap) {
  1380. rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
  1381. l->stats.recv_nacks++;
  1382. }
  1383. tipc_link_advance_backlog(l, xmitq);
  1384. if (unlikely(!skb_queue_empty(&l->wakeupq)))
  1385. link_prepare_wakeup(l);
  1386. }
  1387. exit:
  1388. kfree_skb(skb);
  1389. return rc;
  1390. }
  1391. /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
  1392. */
  1393. static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
  1394. u16 peers_snd_nxt,
  1395. struct sk_buff_head *xmitq)
  1396. {
  1397. struct sk_buff *skb;
  1398. struct tipc_msg *hdr;
  1399. struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
  1400. u16 ack = l->rcv_nxt - 1;
  1401. u16 gap_to = peers_snd_nxt - 1;
  1402. skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
  1403. 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
  1404. if (!skb)
  1405. return false;
  1406. hdr = buf_msg(skb);
  1407. msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
  1408. msg_set_bcast_ack(hdr, ack);
  1409. msg_set_bcgap_after(hdr, ack);
  1410. if (dfrd_skb)
  1411. gap_to = buf_seqno(dfrd_skb) - 1;
  1412. msg_set_bcgap_to(hdr, gap_to);
  1413. msg_set_non_seq(hdr, bcast);
  1414. __skb_queue_tail(xmitq, skb);
  1415. return true;
  1416. }
  1417. /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
  1418. *
  1419. * Give a newly added peer node the sequence number where it should
  1420. * start receiving and acking broadcast packets.
  1421. */
  1422. static void tipc_link_build_bc_init_msg(struct tipc_link *l,
  1423. struct sk_buff_head *xmitq)
  1424. {
  1425. struct sk_buff_head list;
  1426. __skb_queue_head_init(&list);
  1427. if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
  1428. return;
  1429. msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
  1430. tipc_link_xmit(l, &list, xmitq);
  1431. }
  1432. /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
  1433. */
  1434. void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
  1435. {
  1436. int mtyp = msg_type(hdr);
  1437. u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
  1438. if (link_is_up(l))
  1439. return;
  1440. if (msg_user(hdr) == BCAST_PROTOCOL) {
  1441. l->rcv_nxt = peers_snd_nxt;
  1442. l->state = LINK_ESTABLISHED;
  1443. return;
  1444. }
  1445. if (l->peer_caps & TIPC_BCAST_SYNCH)
  1446. return;
  1447. if (msg_peer_node_is_up(hdr))
  1448. return;
  1449. /* Compatibility: accept older, less safe initial synch data */
  1450. if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
  1451. l->rcv_nxt = peers_snd_nxt;
  1452. }
  1453. /* link_bc_retr eval()- check if the indicated range can be retransmitted now
  1454. * - Adjust permitted range if there is overlap with previous retransmission
  1455. */
  1456. static bool link_bc_retr_eval(struct tipc_link *l, u16 *from, u16 *to)
  1457. {
  1458. unsigned long elapsed = jiffies_to_msecs(jiffies - l->prev_retr);
  1459. if (less(*to, *from))
  1460. return false;
  1461. /* New retransmission request */
  1462. if ((elapsed > TIPC_BC_RETR_LIMIT) ||
  1463. less(*to, l->prev_from) || more(*from, l->prev_to)) {
  1464. l->prev_from = *from;
  1465. l->prev_to = *to;
  1466. l->prev_retr = jiffies;
  1467. return true;
  1468. }
  1469. /* Inside range of previous retransmit */
  1470. if (!less(*from, l->prev_from) && !more(*to, l->prev_to))
  1471. return false;
  1472. /* Fully or partially outside previous range => exclude overlap */
  1473. if (less(*from, l->prev_from)) {
  1474. *to = l->prev_from - 1;
  1475. l->prev_from = *from;
  1476. }
  1477. if (more(*to, l->prev_to)) {
  1478. *from = l->prev_to + 1;
  1479. l->prev_to = *to;
  1480. }
  1481. l->prev_retr = jiffies;
  1482. return true;
  1483. }
  1484. /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
  1485. */
  1486. int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
  1487. struct sk_buff_head *xmitq)
  1488. {
  1489. struct tipc_link *snd_l = l->bc_sndlink;
  1490. u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
  1491. u16 from = msg_bcast_ack(hdr) + 1;
  1492. u16 to = from + msg_bc_gap(hdr) - 1;
  1493. int rc = 0;
  1494. if (!link_is_up(l))
  1495. return rc;
  1496. if (!msg_peer_node_is_up(hdr))
  1497. return rc;
  1498. /* Open when peer ackowledges our bcast init msg (pkt #1) */
  1499. if (msg_ack(hdr))
  1500. l->bc_peer_is_up = true;
  1501. if (!l->bc_peer_is_up)
  1502. return rc;
  1503. l->stats.recv_nacks++;
  1504. /* Ignore if peers_snd_nxt goes beyond receive window */
  1505. if (more(peers_snd_nxt, l->rcv_nxt + l->window))
  1506. return rc;
  1507. if (link_bc_retr_eval(snd_l, &from, &to))
  1508. rc = tipc_link_retrans(snd_l, from, to, xmitq);
  1509. l->snd_nxt = peers_snd_nxt;
  1510. if (link_bc_rcv_gap(l))
  1511. rc |= TIPC_LINK_SND_STATE;
  1512. /* Return now if sender supports nack via STATE messages */
  1513. if (l->peer_caps & TIPC_BCAST_STATE_NACK)
  1514. return rc;
  1515. /* Otherwise, be backwards compatible */
  1516. if (!more(peers_snd_nxt, l->rcv_nxt)) {
  1517. l->nack_state = BC_NACK_SND_CONDITIONAL;
  1518. return 0;
  1519. }
  1520. /* Don't NACK if one was recently sent or peeked */
  1521. if (l->nack_state == BC_NACK_SND_SUPPRESS) {
  1522. l->nack_state = BC_NACK_SND_UNCONDITIONAL;
  1523. return 0;
  1524. }
  1525. /* Conditionally delay NACK sending until next synch rcv */
  1526. if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
  1527. l->nack_state = BC_NACK_SND_UNCONDITIONAL;
  1528. if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
  1529. return 0;
  1530. }
  1531. /* Send NACK now but suppress next one */
  1532. tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
  1533. l->nack_state = BC_NACK_SND_SUPPRESS;
  1534. return 0;
  1535. }
  1536. void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
  1537. struct sk_buff_head *xmitq)
  1538. {
  1539. struct sk_buff *skb, *tmp;
  1540. struct tipc_link *snd_l = l->bc_sndlink;
  1541. if (!link_is_up(l) || !l->bc_peer_is_up)
  1542. return;
  1543. if (!more(acked, l->acked))
  1544. return;
  1545. /* Skip over packets peer has already acked */
  1546. skb_queue_walk(&snd_l->transmq, skb) {
  1547. if (more(buf_seqno(skb), l->acked))
  1548. break;
  1549. }
  1550. /* Update/release the packets peer is acking now */
  1551. skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
  1552. if (more(buf_seqno(skb), acked))
  1553. break;
  1554. if (!--TIPC_SKB_CB(skb)->ackers) {
  1555. __skb_unlink(skb, &snd_l->transmq);
  1556. kfree_skb(skb);
  1557. }
  1558. }
  1559. l->acked = acked;
  1560. tipc_link_advance_backlog(snd_l, xmitq);
  1561. if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
  1562. link_prepare_wakeup(snd_l);
  1563. }
  1564. /* tipc_link_bc_nack_rcv(): receive broadcast nack message
  1565. * This function is here for backwards compatibility, since
  1566. * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
  1567. */
  1568. int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
  1569. struct sk_buff_head *xmitq)
  1570. {
  1571. struct tipc_msg *hdr = buf_msg(skb);
  1572. u32 dnode = msg_destnode(hdr);
  1573. int mtyp = msg_type(hdr);
  1574. u16 acked = msg_bcast_ack(hdr);
  1575. u16 from = acked + 1;
  1576. u16 to = msg_bcgap_to(hdr);
  1577. u16 peers_snd_nxt = to + 1;
  1578. int rc = 0;
  1579. kfree_skb(skb);
  1580. if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
  1581. return 0;
  1582. if (mtyp != STATE_MSG)
  1583. return 0;
  1584. if (dnode == tipc_own_addr(l->net)) {
  1585. tipc_link_bc_ack_rcv(l, acked, xmitq);
  1586. rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
  1587. l->stats.recv_nacks++;
  1588. return rc;
  1589. }
  1590. /* Msg for other node => suppress own NACK at next sync if applicable */
  1591. if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
  1592. l->nack_state = BC_NACK_SND_SUPPRESS;
  1593. return 0;
  1594. }
  1595. void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
  1596. {
  1597. int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
  1598. l->window = win;
  1599. l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
  1600. l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
  1601. l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
  1602. l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
  1603. l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
  1604. }
  1605. /**
  1606. * link_reset_stats - reset link statistics
  1607. * @l: pointer to link
  1608. */
  1609. void tipc_link_reset_stats(struct tipc_link *l)
  1610. {
  1611. memset(&l->stats, 0, sizeof(l->stats));
  1612. }
  1613. static void link_print(struct tipc_link *l, const char *str)
  1614. {
  1615. struct sk_buff *hskb = skb_peek(&l->transmq);
  1616. u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
  1617. u16 tail = l->snd_nxt - 1;
  1618. pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
  1619. pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
  1620. skb_queue_len(&l->transmq), head, tail,
  1621. skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
  1622. }
  1623. /* Parse and validate nested (link) properties valid for media, bearer and link
  1624. */
  1625. int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
  1626. {
  1627. int err;
  1628. err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
  1629. tipc_nl_prop_policy, NULL);
  1630. if (err)
  1631. return err;
  1632. if (props[TIPC_NLA_PROP_PRIO]) {
  1633. u32 prio;
  1634. prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
  1635. if (prio > TIPC_MAX_LINK_PRI)
  1636. return -EINVAL;
  1637. }
  1638. if (props[TIPC_NLA_PROP_TOL]) {
  1639. u32 tol;
  1640. tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
  1641. if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
  1642. return -EINVAL;
  1643. }
  1644. if (props[TIPC_NLA_PROP_WIN]) {
  1645. u32 win;
  1646. win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
  1647. if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
  1648. return -EINVAL;
  1649. }
  1650. return 0;
  1651. }
  1652. static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
  1653. {
  1654. int i;
  1655. struct nlattr *stats;
  1656. struct nla_map {
  1657. u32 key;
  1658. u32 val;
  1659. };
  1660. struct nla_map map[] = {
  1661. {TIPC_NLA_STATS_RX_INFO, 0},
  1662. {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
  1663. {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
  1664. {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
  1665. {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
  1666. {TIPC_NLA_STATS_TX_INFO, 0},
  1667. {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
  1668. {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
  1669. {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
  1670. {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
  1671. {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
  1672. s->msg_length_counts : 1},
  1673. {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
  1674. {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
  1675. {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
  1676. {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
  1677. {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
  1678. {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
  1679. {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
  1680. {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
  1681. {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
  1682. {TIPC_NLA_STATS_RX_STATES, s->recv_states},
  1683. {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
  1684. {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
  1685. {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
  1686. {TIPC_NLA_STATS_TX_STATES, s->sent_states},
  1687. {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
  1688. {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
  1689. {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
  1690. {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
  1691. {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
  1692. {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
  1693. {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
  1694. {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
  1695. (s->accu_queue_sz / s->queue_sz_counts) : 0}
  1696. };
  1697. stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
  1698. if (!stats)
  1699. return -EMSGSIZE;
  1700. for (i = 0; i < ARRAY_SIZE(map); i++)
  1701. if (nla_put_u32(skb, map[i].key, map[i].val))
  1702. goto msg_full;
  1703. nla_nest_end(skb, stats);
  1704. return 0;
  1705. msg_full:
  1706. nla_nest_cancel(skb, stats);
  1707. return -EMSGSIZE;
  1708. }
  1709. /* Caller should hold appropriate locks to protect the link */
  1710. int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
  1711. struct tipc_link *link, int nlflags)
  1712. {
  1713. int err;
  1714. void *hdr;
  1715. struct nlattr *attrs;
  1716. struct nlattr *prop;
  1717. struct tipc_net *tn = net_generic(net, tipc_net_id);
  1718. hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
  1719. nlflags, TIPC_NL_LINK_GET);
  1720. if (!hdr)
  1721. return -EMSGSIZE;
  1722. attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
  1723. if (!attrs)
  1724. goto msg_full;
  1725. if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
  1726. goto attr_msg_full;
  1727. if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
  1728. tipc_cluster_mask(tn->own_addr)))
  1729. goto attr_msg_full;
  1730. if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
  1731. goto attr_msg_full;
  1732. if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
  1733. goto attr_msg_full;
  1734. if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
  1735. goto attr_msg_full;
  1736. if (tipc_link_is_up(link))
  1737. if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
  1738. goto attr_msg_full;
  1739. if (link->active)
  1740. if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
  1741. goto attr_msg_full;
  1742. prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
  1743. if (!prop)
  1744. goto attr_msg_full;
  1745. if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
  1746. goto prop_msg_full;
  1747. if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
  1748. goto prop_msg_full;
  1749. if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
  1750. link->window))
  1751. goto prop_msg_full;
  1752. if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
  1753. goto prop_msg_full;
  1754. nla_nest_end(msg->skb, prop);
  1755. err = __tipc_nl_add_stats(msg->skb, &link->stats);
  1756. if (err)
  1757. goto attr_msg_full;
  1758. nla_nest_end(msg->skb, attrs);
  1759. genlmsg_end(msg->skb, hdr);
  1760. return 0;
  1761. prop_msg_full:
  1762. nla_nest_cancel(msg->skb, prop);
  1763. attr_msg_full:
  1764. nla_nest_cancel(msg->skb, attrs);
  1765. msg_full:
  1766. genlmsg_cancel(msg->skb, hdr);
  1767. return -EMSGSIZE;
  1768. }
  1769. static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
  1770. struct tipc_stats *stats)
  1771. {
  1772. int i;
  1773. struct nlattr *nest;
  1774. struct nla_map {
  1775. __u32 key;
  1776. __u32 val;
  1777. };
  1778. struct nla_map map[] = {
  1779. {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
  1780. {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
  1781. {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
  1782. {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
  1783. {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
  1784. {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
  1785. {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
  1786. {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
  1787. {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
  1788. {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
  1789. {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
  1790. {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
  1791. {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
  1792. {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
  1793. {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
  1794. {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
  1795. {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
  1796. {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
  1797. {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
  1798. (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
  1799. };
  1800. nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
  1801. if (!nest)
  1802. return -EMSGSIZE;
  1803. for (i = 0; i < ARRAY_SIZE(map); i++)
  1804. if (nla_put_u32(skb, map[i].key, map[i].val))
  1805. goto msg_full;
  1806. nla_nest_end(skb, nest);
  1807. return 0;
  1808. msg_full:
  1809. nla_nest_cancel(skb, nest);
  1810. return -EMSGSIZE;
  1811. }
  1812. int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
  1813. {
  1814. int err;
  1815. void *hdr;
  1816. struct nlattr *attrs;
  1817. struct nlattr *prop;
  1818. struct tipc_net *tn = net_generic(net, tipc_net_id);
  1819. struct tipc_link *bcl = tn->bcl;
  1820. if (!bcl)
  1821. return 0;
  1822. tipc_bcast_lock(net);
  1823. hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
  1824. NLM_F_MULTI, TIPC_NL_LINK_GET);
  1825. if (!hdr) {
  1826. tipc_bcast_unlock(net);
  1827. return -EMSGSIZE;
  1828. }
  1829. attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
  1830. if (!attrs)
  1831. goto msg_full;
  1832. /* The broadcast link is always up */
  1833. if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
  1834. goto attr_msg_full;
  1835. if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
  1836. goto attr_msg_full;
  1837. if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
  1838. goto attr_msg_full;
  1839. if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
  1840. goto attr_msg_full;
  1841. if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
  1842. goto attr_msg_full;
  1843. prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
  1844. if (!prop)
  1845. goto attr_msg_full;
  1846. if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
  1847. goto prop_msg_full;
  1848. nla_nest_end(msg->skb, prop);
  1849. err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
  1850. if (err)
  1851. goto attr_msg_full;
  1852. tipc_bcast_unlock(net);
  1853. nla_nest_end(msg->skb, attrs);
  1854. genlmsg_end(msg->skb, hdr);
  1855. return 0;
  1856. prop_msg_full:
  1857. nla_nest_cancel(msg->skb, prop);
  1858. attr_msg_full:
  1859. nla_nest_cancel(msg->skb, attrs);
  1860. msg_full:
  1861. tipc_bcast_unlock(net);
  1862. genlmsg_cancel(msg->skb, hdr);
  1863. return -EMSGSIZE;
  1864. }
  1865. void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
  1866. struct sk_buff_head *xmitq)
  1867. {
  1868. l->tolerance = tol;
  1869. tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, tol, 0, xmitq);
  1870. }
  1871. void tipc_link_set_prio(struct tipc_link *l, u32 prio,
  1872. struct sk_buff_head *xmitq)
  1873. {
  1874. l->priority = prio;
  1875. tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, prio, xmitq);
  1876. }
  1877. void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
  1878. {
  1879. l->abort_limit = limit;
  1880. }