vmci_transport.c 57 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145
  1. // SPDX-License-Identifier: GPL-2.0-only
  2. /*
  3. * VMware vSockets Driver
  4. *
  5. * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
  6. */
  7. #include <linux/types.h>
  8. #include <linux/bitops.h>
  9. #include <linux/cred.h>
  10. #include <linux/init.h>
  11. #include <linux/io.h>
  12. #include <linux/kernel.h>
  13. #include <linux/kmod.h>
  14. #include <linux/list.h>
  15. #include <linux/module.h>
  16. #include <linux/mutex.h>
  17. #include <linux/net.h>
  18. #include <linux/poll.h>
  19. #include <linux/skbuff.h>
  20. #include <linux/smp.h>
  21. #include <linux/socket.h>
  22. #include <linux/stddef.h>
  23. #include <linux/unistd.h>
  24. #include <linux/wait.h>
  25. #include <linux/workqueue.h>
  26. #include <net/sock.h>
  27. #include <net/af_vsock.h>
  28. #include "vmci_transport_notify.h"
  29. static int vmci_transport_recv_dgram_cb(void *data, struct vmci_datagram *dg);
  30. static int vmci_transport_recv_stream_cb(void *data, struct vmci_datagram *dg);
  31. static void vmci_transport_peer_detach_cb(u32 sub_id,
  32. const struct vmci_event_data *ed,
  33. void *client_data);
  34. static void vmci_transport_recv_pkt_work(struct work_struct *work);
  35. static void vmci_transport_cleanup(struct work_struct *work);
  36. static int vmci_transport_recv_listen(struct sock *sk,
  37. struct vmci_transport_packet *pkt);
  38. static int vmci_transport_recv_connecting_server(
  39. struct sock *sk,
  40. struct sock *pending,
  41. struct vmci_transport_packet *pkt);
  42. static int vmci_transport_recv_connecting_client(
  43. struct sock *sk,
  44. struct vmci_transport_packet *pkt);
  45. static int vmci_transport_recv_connecting_client_negotiate(
  46. struct sock *sk,
  47. struct vmci_transport_packet *pkt);
  48. static int vmci_transport_recv_connecting_client_invalid(
  49. struct sock *sk,
  50. struct vmci_transport_packet *pkt);
  51. static int vmci_transport_recv_connected(struct sock *sk,
  52. struct vmci_transport_packet *pkt);
  53. static bool vmci_transport_old_proto_override(bool *old_pkt_proto);
  54. static u16 vmci_transport_new_proto_supported_versions(void);
  55. static bool vmci_transport_proto_to_notify_struct(struct sock *sk, u16 *proto,
  56. bool old_pkt_proto);
  57. static bool vmci_check_transport(struct vsock_sock *vsk);
  58. struct vmci_transport_recv_pkt_info {
  59. struct work_struct work;
  60. struct sock *sk;
  61. struct vmci_transport_packet pkt;
  62. };
  63. static LIST_HEAD(vmci_transport_cleanup_list);
  64. static DEFINE_SPINLOCK(vmci_transport_cleanup_lock);
  65. static DECLARE_WORK(vmci_transport_cleanup_work, vmci_transport_cleanup);
  66. static struct vmci_handle vmci_transport_stream_handle = { VMCI_INVALID_ID,
  67. VMCI_INVALID_ID };
  68. static u32 vmci_transport_qp_resumed_sub_id = VMCI_INVALID_ID;
  69. static int PROTOCOL_OVERRIDE = -1;
  70. /* Helper function to convert from a VMCI error code to a VSock error code. */
  71. static s32 vmci_transport_error_to_vsock_error(s32 vmci_error)
  72. {
  73. switch (vmci_error) {
  74. case VMCI_ERROR_NO_MEM:
  75. return -ENOMEM;
  76. case VMCI_ERROR_DUPLICATE_ENTRY:
  77. case VMCI_ERROR_ALREADY_EXISTS:
  78. return -EADDRINUSE;
  79. case VMCI_ERROR_NO_ACCESS:
  80. return -EPERM;
  81. case VMCI_ERROR_NO_RESOURCES:
  82. return -ENOBUFS;
  83. case VMCI_ERROR_INVALID_RESOURCE:
  84. return -EHOSTUNREACH;
  85. case VMCI_ERROR_INVALID_ARGS:
  86. default:
  87. break;
  88. }
  89. return -EINVAL;
  90. }
  91. static u32 vmci_transport_peer_rid(u32 peer_cid)
  92. {
  93. if (VMADDR_CID_HYPERVISOR == peer_cid)
  94. return VMCI_TRANSPORT_HYPERVISOR_PACKET_RID;
  95. return VMCI_TRANSPORT_PACKET_RID;
  96. }
  97. static inline void
  98. vmci_transport_packet_init(struct vmci_transport_packet *pkt,
  99. struct sockaddr_vm *src,
  100. struct sockaddr_vm *dst,
  101. u8 type,
  102. u64 size,
  103. u64 mode,
  104. struct vmci_transport_waiting_info *wait,
  105. u16 proto,
  106. struct vmci_handle handle)
  107. {
  108. /* We register the stream control handler as an any cid handle so we
  109. * must always send from a source address of VMADDR_CID_ANY
  110. */
  111. pkt->dg.src = vmci_make_handle(VMADDR_CID_ANY,
  112. VMCI_TRANSPORT_PACKET_RID);
  113. pkt->dg.dst = vmci_make_handle(dst->svm_cid,
  114. vmci_transport_peer_rid(dst->svm_cid));
  115. pkt->dg.payload_size = sizeof(*pkt) - sizeof(pkt->dg);
  116. pkt->version = VMCI_TRANSPORT_PACKET_VERSION;
  117. pkt->type = type;
  118. pkt->src_port = src->svm_port;
  119. pkt->dst_port = dst->svm_port;
  120. memset(&pkt->proto, 0, sizeof(pkt->proto));
  121. memset(&pkt->_reserved2, 0, sizeof(pkt->_reserved2));
  122. switch (pkt->type) {
  123. case VMCI_TRANSPORT_PACKET_TYPE_INVALID:
  124. pkt->u.size = 0;
  125. break;
  126. case VMCI_TRANSPORT_PACKET_TYPE_REQUEST:
  127. case VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE:
  128. pkt->u.size = size;
  129. break;
  130. case VMCI_TRANSPORT_PACKET_TYPE_OFFER:
  131. case VMCI_TRANSPORT_PACKET_TYPE_ATTACH:
  132. pkt->u.handle = handle;
  133. break;
  134. case VMCI_TRANSPORT_PACKET_TYPE_WROTE:
  135. case VMCI_TRANSPORT_PACKET_TYPE_READ:
  136. case VMCI_TRANSPORT_PACKET_TYPE_RST:
  137. pkt->u.size = 0;
  138. break;
  139. case VMCI_TRANSPORT_PACKET_TYPE_SHUTDOWN:
  140. pkt->u.mode = mode;
  141. break;
  142. case VMCI_TRANSPORT_PACKET_TYPE_WAITING_READ:
  143. case VMCI_TRANSPORT_PACKET_TYPE_WAITING_WRITE:
  144. memcpy(&pkt->u.wait, wait, sizeof(pkt->u.wait));
  145. break;
  146. case VMCI_TRANSPORT_PACKET_TYPE_REQUEST2:
  147. case VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE2:
  148. pkt->u.size = size;
  149. pkt->proto = proto;
  150. break;
  151. }
  152. }
  153. static inline void
  154. vmci_transport_packet_get_addresses(struct vmci_transport_packet *pkt,
  155. struct sockaddr_vm *local,
  156. struct sockaddr_vm *remote)
  157. {
  158. vsock_addr_init(local, pkt->dg.dst.context, pkt->dst_port);
  159. vsock_addr_init(remote, pkt->dg.src.context, pkt->src_port);
  160. }
  161. static int
  162. __vmci_transport_send_control_pkt(struct vmci_transport_packet *pkt,
  163. struct sockaddr_vm *src,
  164. struct sockaddr_vm *dst,
  165. enum vmci_transport_packet_type type,
  166. u64 size,
  167. u64 mode,
  168. struct vmci_transport_waiting_info *wait,
  169. u16 proto,
  170. struct vmci_handle handle,
  171. bool convert_error)
  172. {
  173. int err;
  174. vmci_transport_packet_init(pkt, src, dst, type, size, mode, wait,
  175. proto, handle);
  176. err = vmci_datagram_send(&pkt->dg);
  177. if (convert_error && (err < 0))
  178. return vmci_transport_error_to_vsock_error(err);
  179. return err;
  180. }
  181. static int
  182. vmci_transport_reply_control_pkt_fast(struct vmci_transport_packet *pkt,
  183. enum vmci_transport_packet_type type,
  184. u64 size,
  185. u64 mode,
  186. struct vmci_transport_waiting_info *wait,
  187. struct vmci_handle handle)
  188. {
  189. struct vmci_transport_packet reply;
  190. struct sockaddr_vm src, dst;
  191. if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST) {
  192. return 0;
  193. } else {
  194. vmci_transport_packet_get_addresses(pkt, &src, &dst);
  195. return __vmci_transport_send_control_pkt(&reply, &src, &dst,
  196. type,
  197. size, mode, wait,
  198. VSOCK_PROTO_INVALID,
  199. handle, true);
  200. }
  201. }
  202. static int
  203. vmci_transport_send_control_pkt_bh(struct sockaddr_vm *src,
  204. struct sockaddr_vm *dst,
  205. enum vmci_transport_packet_type type,
  206. u64 size,
  207. u64 mode,
  208. struct vmci_transport_waiting_info *wait,
  209. struct vmci_handle handle)
  210. {
  211. /* Note that it is safe to use a single packet across all CPUs since
  212. * two tasklets of the same type are guaranteed to not ever run
  213. * simultaneously. If that ever changes, or VMCI stops using tasklets,
  214. * we can use per-cpu packets.
  215. */
  216. static struct vmci_transport_packet pkt;
  217. return __vmci_transport_send_control_pkt(&pkt, src, dst, type,
  218. size, mode, wait,
  219. VSOCK_PROTO_INVALID, handle,
  220. false);
  221. }
  222. static int
  223. vmci_transport_alloc_send_control_pkt(struct sockaddr_vm *src,
  224. struct sockaddr_vm *dst,
  225. enum vmci_transport_packet_type type,
  226. u64 size,
  227. u64 mode,
  228. struct vmci_transport_waiting_info *wait,
  229. u16 proto,
  230. struct vmci_handle handle)
  231. {
  232. struct vmci_transport_packet *pkt;
  233. int err;
  234. pkt = kmalloc(sizeof(*pkt), GFP_KERNEL);
  235. if (!pkt)
  236. return -ENOMEM;
  237. err = __vmci_transport_send_control_pkt(pkt, src, dst, type, size,
  238. mode, wait, proto, handle,
  239. true);
  240. kfree(pkt);
  241. return err;
  242. }
  243. static int
  244. vmci_transport_send_control_pkt(struct sock *sk,
  245. enum vmci_transport_packet_type type,
  246. u64 size,
  247. u64 mode,
  248. struct vmci_transport_waiting_info *wait,
  249. u16 proto,
  250. struct vmci_handle handle)
  251. {
  252. struct vsock_sock *vsk;
  253. vsk = vsock_sk(sk);
  254. if (!vsock_addr_bound(&vsk->local_addr))
  255. return -EINVAL;
  256. if (!vsock_addr_bound(&vsk->remote_addr))
  257. return -EINVAL;
  258. return vmci_transport_alloc_send_control_pkt(&vsk->local_addr,
  259. &vsk->remote_addr,
  260. type, size, mode,
  261. wait, proto, handle);
  262. }
  263. static int vmci_transport_send_reset_bh(struct sockaddr_vm *dst,
  264. struct sockaddr_vm *src,
  265. struct vmci_transport_packet *pkt)
  266. {
  267. if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST)
  268. return 0;
  269. return vmci_transport_send_control_pkt_bh(
  270. dst, src,
  271. VMCI_TRANSPORT_PACKET_TYPE_RST, 0,
  272. 0, NULL, VMCI_INVALID_HANDLE);
  273. }
  274. static int vmci_transport_send_reset(struct sock *sk,
  275. struct vmci_transport_packet *pkt)
  276. {
  277. struct sockaddr_vm *dst_ptr;
  278. struct sockaddr_vm dst;
  279. struct vsock_sock *vsk;
  280. if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST)
  281. return 0;
  282. vsk = vsock_sk(sk);
  283. if (!vsock_addr_bound(&vsk->local_addr))
  284. return -EINVAL;
  285. if (vsock_addr_bound(&vsk->remote_addr)) {
  286. dst_ptr = &vsk->remote_addr;
  287. } else {
  288. vsock_addr_init(&dst, pkt->dg.src.context,
  289. pkt->src_port);
  290. dst_ptr = &dst;
  291. }
  292. return vmci_transport_alloc_send_control_pkt(&vsk->local_addr, dst_ptr,
  293. VMCI_TRANSPORT_PACKET_TYPE_RST,
  294. 0, 0, NULL, VSOCK_PROTO_INVALID,
  295. VMCI_INVALID_HANDLE);
  296. }
  297. static int vmci_transport_send_negotiate(struct sock *sk, size_t size)
  298. {
  299. return vmci_transport_send_control_pkt(
  300. sk,
  301. VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE,
  302. size, 0, NULL,
  303. VSOCK_PROTO_INVALID,
  304. VMCI_INVALID_HANDLE);
  305. }
  306. static int vmci_transport_send_negotiate2(struct sock *sk, size_t size,
  307. u16 version)
  308. {
  309. return vmci_transport_send_control_pkt(
  310. sk,
  311. VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE2,
  312. size, 0, NULL, version,
  313. VMCI_INVALID_HANDLE);
  314. }
  315. static int vmci_transport_send_qp_offer(struct sock *sk,
  316. struct vmci_handle handle)
  317. {
  318. return vmci_transport_send_control_pkt(
  319. sk, VMCI_TRANSPORT_PACKET_TYPE_OFFER, 0,
  320. 0, NULL,
  321. VSOCK_PROTO_INVALID, handle);
  322. }
  323. static int vmci_transport_send_attach(struct sock *sk,
  324. struct vmci_handle handle)
  325. {
  326. return vmci_transport_send_control_pkt(
  327. sk, VMCI_TRANSPORT_PACKET_TYPE_ATTACH,
  328. 0, 0, NULL, VSOCK_PROTO_INVALID,
  329. handle);
  330. }
  331. static int vmci_transport_reply_reset(struct vmci_transport_packet *pkt)
  332. {
  333. return vmci_transport_reply_control_pkt_fast(
  334. pkt,
  335. VMCI_TRANSPORT_PACKET_TYPE_RST,
  336. 0, 0, NULL,
  337. VMCI_INVALID_HANDLE);
  338. }
  339. static int vmci_transport_send_invalid_bh(struct sockaddr_vm *dst,
  340. struct sockaddr_vm *src)
  341. {
  342. return vmci_transport_send_control_pkt_bh(
  343. dst, src,
  344. VMCI_TRANSPORT_PACKET_TYPE_INVALID,
  345. 0, 0, NULL, VMCI_INVALID_HANDLE);
  346. }
  347. int vmci_transport_send_wrote_bh(struct sockaddr_vm *dst,
  348. struct sockaddr_vm *src)
  349. {
  350. return vmci_transport_send_control_pkt_bh(
  351. dst, src,
  352. VMCI_TRANSPORT_PACKET_TYPE_WROTE, 0,
  353. 0, NULL, VMCI_INVALID_HANDLE);
  354. }
  355. int vmci_transport_send_read_bh(struct sockaddr_vm *dst,
  356. struct sockaddr_vm *src)
  357. {
  358. return vmci_transport_send_control_pkt_bh(
  359. dst, src,
  360. VMCI_TRANSPORT_PACKET_TYPE_READ, 0,
  361. 0, NULL, VMCI_INVALID_HANDLE);
  362. }
  363. int vmci_transport_send_wrote(struct sock *sk)
  364. {
  365. return vmci_transport_send_control_pkt(
  366. sk, VMCI_TRANSPORT_PACKET_TYPE_WROTE, 0,
  367. 0, NULL, VSOCK_PROTO_INVALID,
  368. VMCI_INVALID_HANDLE);
  369. }
  370. int vmci_transport_send_read(struct sock *sk)
  371. {
  372. return vmci_transport_send_control_pkt(
  373. sk, VMCI_TRANSPORT_PACKET_TYPE_READ, 0,
  374. 0, NULL, VSOCK_PROTO_INVALID,
  375. VMCI_INVALID_HANDLE);
  376. }
  377. int vmci_transport_send_waiting_write(struct sock *sk,
  378. struct vmci_transport_waiting_info *wait)
  379. {
  380. return vmci_transport_send_control_pkt(
  381. sk, VMCI_TRANSPORT_PACKET_TYPE_WAITING_WRITE,
  382. 0, 0, wait, VSOCK_PROTO_INVALID,
  383. VMCI_INVALID_HANDLE);
  384. }
  385. int vmci_transport_send_waiting_read(struct sock *sk,
  386. struct vmci_transport_waiting_info *wait)
  387. {
  388. return vmci_transport_send_control_pkt(
  389. sk, VMCI_TRANSPORT_PACKET_TYPE_WAITING_READ,
  390. 0, 0, wait, VSOCK_PROTO_INVALID,
  391. VMCI_INVALID_HANDLE);
  392. }
  393. static int vmci_transport_shutdown(struct vsock_sock *vsk, int mode)
  394. {
  395. return vmci_transport_send_control_pkt(
  396. &vsk->sk,
  397. VMCI_TRANSPORT_PACKET_TYPE_SHUTDOWN,
  398. 0, mode, NULL,
  399. VSOCK_PROTO_INVALID,
  400. VMCI_INVALID_HANDLE);
  401. }
  402. static int vmci_transport_send_conn_request(struct sock *sk, size_t size)
  403. {
  404. return vmci_transport_send_control_pkt(sk,
  405. VMCI_TRANSPORT_PACKET_TYPE_REQUEST,
  406. size, 0, NULL,
  407. VSOCK_PROTO_INVALID,
  408. VMCI_INVALID_HANDLE);
  409. }
  410. static int vmci_transport_send_conn_request2(struct sock *sk, size_t size,
  411. u16 version)
  412. {
  413. return vmci_transport_send_control_pkt(
  414. sk, VMCI_TRANSPORT_PACKET_TYPE_REQUEST2,
  415. size, 0, NULL, version,
  416. VMCI_INVALID_HANDLE);
  417. }
  418. static struct sock *vmci_transport_get_pending(
  419. struct sock *listener,
  420. struct vmci_transport_packet *pkt)
  421. {
  422. struct vsock_sock *vlistener;
  423. struct vsock_sock *vpending;
  424. struct sock *pending;
  425. struct sockaddr_vm src;
  426. vsock_addr_init(&src, pkt->dg.src.context, pkt->src_port);
  427. vlistener = vsock_sk(listener);
  428. list_for_each_entry(vpending, &vlistener->pending_links,
  429. pending_links) {
  430. if (vsock_addr_equals_addr(&src, &vpending->remote_addr) &&
  431. pkt->dst_port == vpending->local_addr.svm_port) {
  432. pending = sk_vsock(vpending);
  433. sock_hold(pending);
  434. goto found;
  435. }
  436. }
  437. pending = NULL;
  438. found:
  439. return pending;
  440. }
  441. static void vmci_transport_release_pending(struct sock *pending)
  442. {
  443. sock_put(pending);
  444. }
  445. /* We allow two kinds of sockets to communicate with a restricted VM: 1)
  446. * trusted sockets 2) sockets from applications running as the same user as the
  447. * VM (this is only true for the host side and only when using hosted products)
  448. */
  449. static bool vmci_transport_is_trusted(struct vsock_sock *vsock, u32 peer_cid)
  450. {
  451. return vsock->trusted ||
  452. vmci_is_context_owner(peer_cid, vsock->owner->uid);
  453. }
  454. /* We allow sending datagrams to and receiving datagrams from a restricted VM
  455. * only if it is trusted as described in vmci_transport_is_trusted.
  456. */
  457. static bool vmci_transport_allow_dgram(struct vsock_sock *vsock, u32 peer_cid)
  458. {
  459. if (VMADDR_CID_HYPERVISOR == peer_cid)
  460. return true;
  461. if (vsock->cached_peer != peer_cid) {
  462. vsock->cached_peer = peer_cid;
  463. if (!vmci_transport_is_trusted(vsock, peer_cid) &&
  464. (vmci_context_get_priv_flags(peer_cid) &
  465. VMCI_PRIVILEGE_FLAG_RESTRICTED)) {
  466. vsock->cached_peer_allow_dgram = false;
  467. } else {
  468. vsock->cached_peer_allow_dgram = true;
  469. }
  470. }
  471. return vsock->cached_peer_allow_dgram;
  472. }
  473. static int
  474. vmci_transport_queue_pair_alloc(struct vmci_qp **qpair,
  475. struct vmci_handle *handle,
  476. u64 produce_size,
  477. u64 consume_size,
  478. u32 peer, u32 flags, bool trusted)
  479. {
  480. int err = 0;
  481. if (trusted) {
  482. /* Try to allocate our queue pair as trusted. This will only
  483. * work if vsock is running in the host.
  484. */
  485. err = vmci_qpair_alloc(qpair, handle, produce_size,
  486. consume_size,
  487. peer, flags,
  488. VMCI_PRIVILEGE_FLAG_TRUSTED);
  489. if (err != VMCI_ERROR_NO_ACCESS)
  490. goto out;
  491. }
  492. err = vmci_qpair_alloc(qpair, handle, produce_size, consume_size,
  493. peer, flags, VMCI_NO_PRIVILEGE_FLAGS);
  494. out:
  495. if (err < 0) {
  496. pr_err_once("Could not attach to queue pair with %d\n", err);
  497. err = vmci_transport_error_to_vsock_error(err);
  498. }
  499. return err;
  500. }
  501. static int
  502. vmci_transport_datagram_create_hnd(u32 resource_id,
  503. u32 flags,
  504. vmci_datagram_recv_cb recv_cb,
  505. void *client_data,
  506. struct vmci_handle *out_handle)
  507. {
  508. int err = 0;
  509. /* Try to allocate our datagram handler as trusted. This will only work
  510. * if vsock is running in the host.
  511. */
  512. err = vmci_datagram_create_handle_priv(resource_id, flags,
  513. VMCI_PRIVILEGE_FLAG_TRUSTED,
  514. recv_cb,
  515. client_data, out_handle);
  516. if (err == VMCI_ERROR_NO_ACCESS)
  517. err = vmci_datagram_create_handle(resource_id, flags,
  518. recv_cb, client_data,
  519. out_handle);
  520. return err;
  521. }
  522. /* This is invoked as part of a tasklet that's scheduled when the VMCI
  523. * interrupt fires. This is run in bottom-half context and if it ever needs to
  524. * sleep it should defer that work to a work queue.
  525. */
  526. static int vmci_transport_recv_dgram_cb(void *data, struct vmci_datagram *dg)
  527. {
  528. struct sock *sk;
  529. size_t size;
  530. struct sk_buff *skb;
  531. struct vsock_sock *vsk;
  532. sk = (struct sock *)data;
  533. /* This handler is privileged when this module is running on the host.
  534. * We will get datagrams from all endpoints (even VMs that are in a
  535. * restricted context). If we get one from a restricted context then
  536. * the destination socket must be trusted.
  537. *
  538. * NOTE: We access the socket struct without holding the lock here.
  539. * This is ok because the field we are interested is never modified
  540. * outside of the create and destruct socket functions.
  541. */
  542. vsk = vsock_sk(sk);
  543. if (!vmci_transport_allow_dgram(vsk, dg->src.context))
  544. return VMCI_ERROR_NO_ACCESS;
  545. size = VMCI_DG_SIZE(dg);
  546. /* Attach the packet to the socket's receive queue as an sk_buff. */
  547. skb = alloc_skb(size, GFP_ATOMIC);
  548. if (!skb)
  549. return VMCI_ERROR_NO_MEM;
  550. /* sk_receive_skb() will do a sock_put(), so hold here. */
  551. sock_hold(sk);
  552. skb_put(skb, size);
  553. memcpy(skb->data, dg, size);
  554. sk_receive_skb(sk, skb, 0);
  555. return VMCI_SUCCESS;
  556. }
  557. static bool vmci_transport_stream_allow(u32 cid, u32 port)
  558. {
  559. static const u32 non_socket_contexts[] = {
  560. VMADDR_CID_LOCAL,
  561. };
  562. int i;
  563. BUILD_BUG_ON(sizeof(cid) != sizeof(*non_socket_contexts));
  564. for (i = 0; i < ARRAY_SIZE(non_socket_contexts); i++) {
  565. if (cid == non_socket_contexts[i])
  566. return false;
  567. }
  568. return true;
  569. }
  570. /* This is invoked as part of a tasklet that's scheduled when the VMCI
  571. * interrupt fires. This is run in bottom-half context but it defers most of
  572. * its work to the packet handling work queue.
  573. */
  574. static int vmci_transport_recv_stream_cb(void *data, struct vmci_datagram *dg)
  575. {
  576. struct sock *sk;
  577. struct sockaddr_vm dst;
  578. struct sockaddr_vm src;
  579. struct vmci_transport_packet *pkt;
  580. struct vsock_sock *vsk;
  581. bool bh_process_pkt;
  582. int err;
  583. sk = NULL;
  584. err = VMCI_SUCCESS;
  585. bh_process_pkt = false;
  586. /* Ignore incoming packets from contexts without sockets, or resources
  587. * that aren't vsock implementations.
  588. */
  589. if (!vmci_transport_stream_allow(dg->src.context, -1)
  590. || vmci_transport_peer_rid(dg->src.context) != dg->src.resource)
  591. return VMCI_ERROR_NO_ACCESS;
  592. if (VMCI_DG_SIZE(dg) < sizeof(*pkt))
  593. /* Drop datagrams that do not contain full VSock packets. */
  594. return VMCI_ERROR_INVALID_ARGS;
  595. pkt = (struct vmci_transport_packet *)dg;
  596. /* Find the socket that should handle this packet. First we look for a
  597. * connected socket and if there is none we look for a socket bound to
  598. * the destintation address.
  599. */
  600. vsock_addr_init(&src, pkt->dg.src.context, pkt->src_port);
  601. vsock_addr_init(&dst, pkt->dg.dst.context, pkt->dst_port);
  602. sk = vsock_find_connected_socket(&src, &dst);
  603. if (!sk) {
  604. sk = vsock_find_bound_socket(&dst);
  605. if (!sk) {
  606. /* We could not find a socket for this specified
  607. * address. If this packet is a RST, we just drop it.
  608. * If it is another packet, we send a RST. Note that
  609. * we do not send a RST reply to RSTs so that we do not
  610. * continually send RSTs between two endpoints.
  611. *
  612. * Note that since this is a reply, dst is src and src
  613. * is dst.
  614. */
  615. if (vmci_transport_send_reset_bh(&dst, &src, pkt) < 0)
  616. pr_err("unable to send reset\n");
  617. err = VMCI_ERROR_NOT_FOUND;
  618. goto out;
  619. }
  620. }
  621. /* If the received packet type is beyond all types known to this
  622. * implementation, reply with an invalid message. Hopefully this will
  623. * help when implementing backwards compatibility in the future.
  624. */
  625. if (pkt->type >= VMCI_TRANSPORT_PACKET_TYPE_MAX) {
  626. vmci_transport_send_invalid_bh(&dst, &src);
  627. err = VMCI_ERROR_INVALID_ARGS;
  628. goto out;
  629. }
  630. /* This handler is privileged when this module is running on the host.
  631. * We will get datagram connect requests from all endpoints (even VMs
  632. * that are in a restricted context). If we get one from a restricted
  633. * context then the destination socket must be trusted.
  634. *
  635. * NOTE: We access the socket struct without holding the lock here.
  636. * This is ok because the field we are interested is never modified
  637. * outside of the create and destruct socket functions.
  638. */
  639. vsk = vsock_sk(sk);
  640. if (!vmci_transport_allow_dgram(vsk, pkt->dg.src.context)) {
  641. err = VMCI_ERROR_NO_ACCESS;
  642. goto out;
  643. }
  644. /* We do most everything in a work queue, but let's fast path the
  645. * notification of reads and writes to help data transfer performance.
  646. * We can only do this if there is no process context code executing
  647. * for this socket since that may change the state.
  648. */
  649. bh_lock_sock(sk);
  650. if (!sock_owned_by_user(sk)) {
  651. /* The local context ID may be out of date, update it. */
  652. vsk->local_addr.svm_cid = dst.svm_cid;
  653. if (sk->sk_state == TCP_ESTABLISHED)
  654. vmci_trans(vsk)->notify_ops->handle_notify_pkt(
  655. sk, pkt, true, &dst, &src,
  656. &bh_process_pkt);
  657. }
  658. bh_unlock_sock(sk);
  659. if (!bh_process_pkt) {
  660. struct vmci_transport_recv_pkt_info *recv_pkt_info;
  661. recv_pkt_info = kmalloc(sizeof(*recv_pkt_info), GFP_ATOMIC);
  662. if (!recv_pkt_info) {
  663. if (vmci_transport_send_reset_bh(&dst, &src, pkt) < 0)
  664. pr_err("unable to send reset\n");
  665. err = VMCI_ERROR_NO_MEM;
  666. goto out;
  667. }
  668. recv_pkt_info->sk = sk;
  669. memcpy(&recv_pkt_info->pkt, pkt, sizeof(recv_pkt_info->pkt));
  670. INIT_WORK(&recv_pkt_info->work, vmci_transport_recv_pkt_work);
  671. schedule_work(&recv_pkt_info->work);
  672. /* Clear sk so that the reference count incremented by one of
  673. * the Find functions above is not decremented below. We need
  674. * that reference count for the packet handler we've scheduled
  675. * to run.
  676. */
  677. sk = NULL;
  678. }
  679. out:
  680. if (sk)
  681. sock_put(sk);
  682. return err;
  683. }
  684. static void vmci_transport_handle_detach(struct sock *sk)
  685. {
  686. struct vsock_sock *vsk;
  687. vsk = vsock_sk(sk);
  688. if (!vmci_handle_is_invalid(vmci_trans(vsk)->qp_handle)) {
  689. sock_set_flag(sk, SOCK_DONE);
  690. /* On a detach the peer will not be sending or receiving
  691. * anymore.
  692. */
  693. vsk->peer_shutdown = SHUTDOWN_MASK;
  694. /* We should not be sending anymore since the peer won't be
  695. * there to receive, but we can still receive if there is data
  696. * left in our consume queue. If the local endpoint is a host,
  697. * we can't call vsock_stream_has_data, since that may block,
  698. * but a host endpoint can't read data once the VM has
  699. * detached, so there is no available data in that case.
  700. */
  701. if (vsk->local_addr.svm_cid == VMADDR_CID_HOST ||
  702. vsock_stream_has_data(vsk) <= 0) {
  703. if (sk->sk_state == TCP_SYN_SENT) {
  704. /* The peer may detach from a queue pair while
  705. * we are still in the connecting state, i.e.,
  706. * if the peer VM is killed after attaching to
  707. * a queue pair, but before we complete the
  708. * handshake. In that case, we treat the detach
  709. * event like a reset.
  710. */
  711. sk->sk_state = TCP_CLOSE;
  712. sk->sk_err = ECONNRESET;
  713. sk->sk_error_report(sk);
  714. return;
  715. }
  716. sk->sk_state = TCP_CLOSE;
  717. }
  718. sk->sk_state_change(sk);
  719. }
  720. }
  721. static void vmci_transport_peer_detach_cb(u32 sub_id,
  722. const struct vmci_event_data *e_data,
  723. void *client_data)
  724. {
  725. struct vmci_transport *trans = client_data;
  726. const struct vmci_event_payload_qp *e_payload;
  727. e_payload = vmci_event_data_const_payload(e_data);
  728. /* XXX This is lame, we should provide a way to lookup sockets by
  729. * qp_handle.
  730. */
  731. if (vmci_handle_is_invalid(e_payload->handle) ||
  732. !vmci_handle_is_equal(trans->qp_handle, e_payload->handle))
  733. return;
  734. /* We don't ask for delayed CBs when we subscribe to this event (we
  735. * pass 0 as flags to vmci_event_subscribe()). VMCI makes no
  736. * guarantees in that case about what context we might be running in,
  737. * so it could be BH or process, blockable or non-blockable. So we
  738. * need to account for all possible contexts here.
  739. */
  740. spin_lock_bh(&trans->lock);
  741. if (!trans->sk)
  742. goto out;
  743. /* Apart from here, trans->lock is only grabbed as part of sk destruct,
  744. * where trans->sk isn't locked.
  745. */
  746. bh_lock_sock(trans->sk);
  747. vmci_transport_handle_detach(trans->sk);
  748. bh_unlock_sock(trans->sk);
  749. out:
  750. spin_unlock_bh(&trans->lock);
  751. }
  752. static void vmci_transport_qp_resumed_cb(u32 sub_id,
  753. const struct vmci_event_data *e_data,
  754. void *client_data)
  755. {
  756. vsock_for_each_connected_socket(vmci_transport_handle_detach);
  757. }
  758. static void vmci_transport_recv_pkt_work(struct work_struct *work)
  759. {
  760. struct vmci_transport_recv_pkt_info *recv_pkt_info;
  761. struct vmci_transport_packet *pkt;
  762. struct sock *sk;
  763. recv_pkt_info =
  764. container_of(work, struct vmci_transport_recv_pkt_info, work);
  765. sk = recv_pkt_info->sk;
  766. pkt = &recv_pkt_info->pkt;
  767. lock_sock(sk);
  768. /* The local context ID may be out of date. */
  769. vsock_sk(sk)->local_addr.svm_cid = pkt->dg.dst.context;
  770. switch (sk->sk_state) {
  771. case TCP_LISTEN:
  772. vmci_transport_recv_listen(sk, pkt);
  773. break;
  774. case TCP_SYN_SENT:
  775. /* Processing of pending connections for servers goes through
  776. * the listening socket, so see vmci_transport_recv_listen()
  777. * for that path.
  778. */
  779. vmci_transport_recv_connecting_client(sk, pkt);
  780. break;
  781. case TCP_ESTABLISHED:
  782. vmci_transport_recv_connected(sk, pkt);
  783. break;
  784. default:
  785. /* Because this function does not run in the same context as
  786. * vmci_transport_recv_stream_cb it is possible that the
  787. * socket has closed. We need to let the other side know or it
  788. * could be sitting in a connect and hang forever. Send a
  789. * reset to prevent that.
  790. */
  791. vmci_transport_send_reset(sk, pkt);
  792. break;
  793. }
  794. release_sock(sk);
  795. kfree(recv_pkt_info);
  796. /* Release reference obtained in the stream callback when we fetched
  797. * this socket out of the bound or connected list.
  798. */
  799. sock_put(sk);
  800. }
  801. static int vmci_transport_recv_listen(struct sock *sk,
  802. struct vmci_transport_packet *pkt)
  803. {
  804. struct sock *pending;
  805. struct vsock_sock *vpending;
  806. int err;
  807. u64 qp_size;
  808. bool old_request = false;
  809. bool old_pkt_proto = false;
  810. err = 0;
  811. /* Because we are in the listen state, we could be receiving a packet
  812. * for ourself or any previous connection requests that we received.
  813. * If it's the latter, we try to find a socket in our list of pending
  814. * connections and, if we do, call the appropriate handler for the
  815. * state that that socket is in. Otherwise we try to service the
  816. * connection request.
  817. */
  818. pending = vmci_transport_get_pending(sk, pkt);
  819. if (pending) {
  820. lock_sock(pending);
  821. /* The local context ID may be out of date. */
  822. vsock_sk(pending)->local_addr.svm_cid = pkt->dg.dst.context;
  823. switch (pending->sk_state) {
  824. case TCP_SYN_SENT:
  825. err = vmci_transport_recv_connecting_server(sk,
  826. pending,
  827. pkt);
  828. break;
  829. default:
  830. vmci_transport_send_reset(pending, pkt);
  831. err = -EINVAL;
  832. }
  833. if (err < 0)
  834. vsock_remove_pending(sk, pending);
  835. release_sock(pending);
  836. vmci_transport_release_pending(pending);
  837. return err;
  838. }
  839. /* The listen state only accepts connection requests. Reply with a
  840. * reset unless we received a reset.
  841. */
  842. if (!(pkt->type == VMCI_TRANSPORT_PACKET_TYPE_REQUEST ||
  843. pkt->type == VMCI_TRANSPORT_PACKET_TYPE_REQUEST2)) {
  844. vmci_transport_reply_reset(pkt);
  845. return -EINVAL;
  846. }
  847. if (pkt->u.size == 0) {
  848. vmci_transport_reply_reset(pkt);
  849. return -EINVAL;
  850. }
  851. /* If this socket can't accommodate this connection request, we send a
  852. * reset. Otherwise we create and initialize a child socket and reply
  853. * with a connection negotiation.
  854. */
  855. if (sk->sk_ack_backlog >= sk->sk_max_ack_backlog) {
  856. vmci_transport_reply_reset(pkt);
  857. return -ECONNREFUSED;
  858. }
  859. pending = vsock_create_connected(sk);
  860. if (!pending) {
  861. vmci_transport_send_reset(sk, pkt);
  862. return -ENOMEM;
  863. }
  864. vpending = vsock_sk(pending);
  865. vsock_addr_init(&vpending->local_addr, pkt->dg.dst.context,
  866. pkt->dst_port);
  867. vsock_addr_init(&vpending->remote_addr, pkt->dg.src.context,
  868. pkt->src_port);
  869. err = vsock_assign_transport(vpending, vsock_sk(sk));
  870. /* Transport assigned (looking at remote_addr) must be the same
  871. * where we received the request.
  872. */
  873. if (err || !vmci_check_transport(vpending)) {
  874. vmci_transport_send_reset(sk, pkt);
  875. sock_put(pending);
  876. return err;
  877. }
  878. /* If the proposed size fits within our min/max, accept it. Otherwise
  879. * propose our own size.
  880. */
  881. if (pkt->u.size >= vpending->buffer_min_size &&
  882. pkt->u.size <= vpending->buffer_max_size) {
  883. qp_size = pkt->u.size;
  884. } else {
  885. qp_size = vpending->buffer_size;
  886. }
  887. /* Figure out if we are using old or new requests based on the
  888. * overrides pkt types sent by our peer.
  889. */
  890. if (vmci_transport_old_proto_override(&old_pkt_proto)) {
  891. old_request = old_pkt_proto;
  892. } else {
  893. if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_REQUEST)
  894. old_request = true;
  895. else if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_REQUEST2)
  896. old_request = false;
  897. }
  898. if (old_request) {
  899. /* Handle a REQUEST (or override) */
  900. u16 version = VSOCK_PROTO_INVALID;
  901. if (vmci_transport_proto_to_notify_struct(
  902. pending, &version, true))
  903. err = vmci_transport_send_negotiate(pending, qp_size);
  904. else
  905. err = -EINVAL;
  906. } else {
  907. /* Handle a REQUEST2 (or override) */
  908. int proto_int = pkt->proto;
  909. int pos;
  910. u16 active_proto_version = 0;
  911. /* The list of possible protocols is the intersection of all
  912. * protocols the client supports ... plus all the protocols we
  913. * support.
  914. */
  915. proto_int &= vmci_transport_new_proto_supported_versions();
  916. /* We choose the highest possible protocol version and use that
  917. * one.
  918. */
  919. pos = fls(proto_int);
  920. if (pos) {
  921. active_proto_version = (1 << (pos - 1));
  922. if (vmci_transport_proto_to_notify_struct(
  923. pending, &active_proto_version, false))
  924. err = vmci_transport_send_negotiate2(pending,
  925. qp_size,
  926. active_proto_version);
  927. else
  928. err = -EINVAL;
  929. } else {
  930. err = -EINVAL;
  931. }
  932. }
  933. if (err < 0) {
  934. vmci_transport_send_reset(sk, pkt);
  935. sock_put(pending);
  936. err = vmci_transport_error_to_vsock_error(err);
  937. goto out;
  938. }
  939. vsock_add_pending(sk, pending);
  940. sk_acceptq_added(sk);
  941. pending->sk_state = TCP_SYN_SENT;
  942. vmci_trans(vpending)->produce_size =
  943. vmci_trans(vpending)->consume_size = qp_size;
  944. vpending->buffer_size = qp_size;
  945. vmci_trans(vpending)->notify_ops->process_request(pending);
  946. /* We might never receive another message for this socket and it's not
  947. * connected to any process, so we have to ensure it gets cleaned up
  948. * ourself. Our delayed work function will take care of that. Note
  949. * that we do not ever cancel this function since we have few
  950. * guarantees about its state when calling cancel_delayed_work().
  951. * Instead we hold a reference on the socket for that function and make
  952. * it capable of handling cases where it needs to do nothing but
  953. * release that reference.
  954. */
  955. vpending->listener = sk;
  956. sock_hold(sk);
  957. sock_hold(pending);
  958. schedule_delayed_work(&vpending->pending_work, HZ);
  959. out:
  960. return err;
  961. }
  962. static int
  963. vmci_transport_recv_connecting_server(struct sock *listener,
  964. struct sock *pending,
  965. struct vmci_transport_packet *pkt)
  966. {
  967. struct vsock_sock *vpending;
  968. struct vmci_handle handle;
  969. struct vmci_qp *qpair;
  970. bool is_local;
  971. u32 flags;
  972. u32 detach_sub_id;
  973. int err;
  974. int skerr;
  975. vpending = vsock_sk(pending);
  976. detach_sub_id = VMCI_INVALID_ID;
  977. switch (pkt->type) {
  978. case VMCI_TRANSPORT_PACKET_TYPE_OFFER:
  979. if (vmci_handle_is_invalid(pkt->u.handle)) {
  980. vmci_transport_send_reset(pending, pkt);
  981. skerr = EPROTO;
  982. err = -EINVAL;
  983. goto destroy;
  984. }
  985. break;
  986. default:
  987. /* Close and cleanup the connection. */
  988. vmci_transport_send_reset(pending, pkt);
  989. skerr = EPROTO;
  990. err = pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST ? 0 : -EINVAL;
  991. goto destroy;
  992. }
  993. /* In order to complete the connection we need to attach to the offered
  994. * queue pair and send an attach notification. We also subscribe to the
  995. * detach event so we know when our peer goes away, and we do that
  996. * before attaching so we don't miss an event. If all this succeeds,
  997. * we update our state and wakeup anything waiting in accept() for a
  998. * connection.
  999. */
  1000. /* We don't care about attach since we ensure the other side has
  1001. * attached by specifying the ATTACH_ONLY flag below.
  1002. */
  1003. err = vmci_event_subscribe(VMCI_EVENT_QP_PEER_DETACH,
  1004. vmci_transport_peer_detach_cb,
  1005. vmci_trans(vpending), &detach_sub_id);
  1006. if (err < VMCI_SUCCESS) {
  1007. vmci_transport_send_reset(pending, pkt);
  1008. err = vmci_transport_error_to_vsock_error(err);
  1009. skerr = -err;
  1010. goto destroy;
  1011. }
  1012. vmci_trans(vpending)->detach_sub_id = detach_sub_id;
  1013. /* Now attach to the queue pair the client created. */
  1014. handle = pkt->u.handle;
  1015. /* vpending->local_addr always has a context id so we do not need to
  1016. * worry about VMADDR_CID_ANY in this case.
  1017. */
  1018. is_local =
  1019. vpending->remote_addr.svm_cid == vpending->local_addr.svm_cid;
  1020. flags = VMCI_QPFLAG_ATTACH_ONLY;
  1021. flags |= is_local ? VMCI_QPFLAG_LOCAL : 0;
  1022. err = vmci_transport_queue_pair_alloc(
  1023. &qpair,
  1024. &handle,
  1025. vmci_trans(vpending)->produce_size,
  1026. vmci_trans(vpending)->consume_size,
  1027. pkt->dg.src.context,
  1028. flags,
  1029. vmci_transport_is_trusted(
  1030. vpending,
  1031. vpending->remote_addr.svm_cid));
  1032. if (err < 0) {
  1033. vmci_transport_send_reset(pending, pkt);
  1034. skerr = -err;
  1035. goto destroy;
  1036. }
  1037. vmci_trans(vpending)->qp_handle = handle;
  1038. vmci_trans(vpending)->qpair = qpair;
  1039. /* When we send the attach message, we must be ready to handle incoming
  1040. * control messages on the newly connected socket. So we move the
  1041. * pending socket to the connected state before sending the attach
  1042. * message. Otherwise, an incoming packet triggered by the attach being
  1043. * received by the peer may be processed concurrently with what happens
  1044. * below after sending the attach message, and that incoming packet
  1045. * will find the listening socket instead of the (currently) pending
  1046. * socket. Note that enqueueing the socket increments the reference
  1047. * count, so even if a reset comes before the connection is accepted,
  1048. * the socket will be valid until it is removed from the queue.
  1049. *
  1050. * If we fail sending the attach below, we remove the socket from the
  1051. * connected list and move the socket to TCP_CLOSE before
  1052. * releasing the lock, so a pending slow path processing of an incoming
  1053. * packet will not see the socket in the connected state in that case.
  1054. */
  1055. pending->sk_state = TCP_ESTABLISHED;
  1056. vsock_insert_connected(vpending);
  1057. /* Notify our peer of our attach. */
  1058. err = vmci_transport_send_attach(pending, handle);
  1059. if (err < 0) {
  1060. vsock_remove_connected(vpending);
  1061. pr_err("Could not send attach\n");
  1062. vmci_transport_send_reset(pending, pkt);
  1063. err = vmci_transport_error_to_vsock_error(err);
  1064. skerr = -err;
  1065. goto destroy;
  1066. }
  1067. /* We have a connection. Move the now connected socket from the
  1068. * listener's pending list to the accept queue so callers of accept()
  1069. * can find it.
  1070. */
  1071. vsock_remove_pending(listener, pending);
  1072. vsock_enqueue_accept(listener, pending);
  1073. /* Callers of accept() will be be waiting on the listening socket, not
  1074. * the pending socket.
  1075. */
  1076. listener->sk_data_ready(listener);
  1077. return 0;
  1078. destroy:
  1079. pending->sk_err = skerr;
  1080. pending->sk_state = TCP_CLOSE;
  1081. /* As long as we drop our reference, all necessary cleanup will handle
  1082. * when the cleanup function drops its reference and our destruct
  1083. * implementation is called. Note that since the listen handler will
  1084. * remove pending from the pending list upon our failure, the cleanup
  1085. * function won't drop the additional reference, which is why we do it
  1086. * here.
  1087. */
  1088. sock_put(pending);
  1089. return err;
  1090. }
  1091. static int
  1092. vmci_transport_recv_connecting_client(struct sock *sk,
  1093. struct vmci_transport_packet *pkt)
  1094. {
  1095. struct vsock_sock *vsk;
  1096. int err;
  1097. int skerr;
  1098. vsk = vsock_sk(sk);
  1099. switch (pkt->type) {
  1100. case VMCI_TRANSPORT_PACKET_TYPE_ATTACH:
  1101. if (vmci_handle_is_invalid(pkt->u.handle) ||
  1102. !vmci_handle_is_equal(pkt->u.handle,
  1103. vmci_trans(vsk)->qp_handle)) {
  1104. skerr = EPROTO;
  1105. err = -EINVAL;
  1106. goto destroy;
  1107. }
  1108. /* Signify the socket is connected and wakeup the waiter in
  1109. * connect(). Also place the socket in the connected table for
  1110. * accounting (it can already be found since it's in the bound
  1111. * table).
  1112. */
  1113. sk->sk_state = TCP_ESTABLISHED;
  1114. sk->sk_socket->state = SS_CONNECTED;
  1115. vsock_insert_connected(vsk);
  1116. sk->sk_state_change(sk);
  1117. break;
  1118. case VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE:
  1119. case VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE2:
  1120. if (pkt->u.size == 0
  1121. || pkt->dg.src.context != vsk->remote_addr.svm_cid
  1122. || pkt->src_port != vsk->remote_addr.svm_port
  1123. || !vmci_handle_is_invalid(vmci_trans(vsk)->qp_handle)
  1124. || vmci_trans(vsk)->qpair
  1125. || vmci_trans(vsk)->produce_size != 0
  1126. || vmci_trans(vsk)->consume_size != 0
  1127. || vmci_trans(vsk)->detach_sub_id != VMCI_INVALID_ID) {
  1128. skerr = EPROTO;
  1129. err = -EINVAL;
  1130. goto destroy;
  1131. }
  1132. err = vmci_transport_recv_connecting_client_negotiate(sk, pkt);
  1133. if (err) {
  1134. skerr = -err;
  1135. goto destroy;
  1136. }
  1137. break;
  1138. case VMCI_TRANSPORT_PACKET_TYPE_INVALID:
  1139. err = vmci_transport_recv_connecting_client_invalid(sk, pkt);
  1140. if (err) {
  1141. skerr = -err;
  1142. goto destroy;
  1143. }
  1144. break;
  1145. case VMCI_TRANSPORT_PACKET_TYPE_RST:
  1146. /* Older versions of the linux code (WS 6.5 / ESX 4.0) used to
  1147. * continue processing here after they sent an INVALID packet.
  1148. * This meant that we got a RST after the INVALID. We ignore a
  1149. * RST after an INVALID. The common code doesn't send the RST
  1150. * ... so we can hang if an old version of the common code
  1151. * fails between getting a REQUEST and sending an OFFER back.
  1152. * Not much we can do about it... except hope that it doesn't
  1153. * happen.
  1154. */
  1155. if (vsk->ignore_connecting_rst) {
  1156. vsk->ignore_connecting_rst = false;
  1157. } else {
  1158. skerr = ECONNRESET;
  1159. err = 0;
  1160. goto destroy;
  1161. }
  1162. break;
  1163. default:
  1164. /* Close and cleanup the connection. */
  1165. skerr = EPROTO;
  1166. err = -EINVAL;
  1167. goto destroy;
  1168. }
  1169. return 0;
  1170. destroy:
  1171. vmci_transport_send_reset(sk, pkt);
  1172. sk->sk_state = TCP_CLOSE;
  1173. sk->sk_err = skerr;
  1174. sk->sk_error_report(sk);
  1175. return err;
  1176. }
  1177. static int vmci_transport_recv_connecting_client_negotiate(
  1178. struct sock *sk,
  1179. struct vmci_transport_packet *pkt)
  1180. {
  1181. int err;
  1182. struct vsock_sock *vsk;
  1183. struct vmci_handle handle;
  1184. struct vmci_qp *qpair;
  1185. u32 detach_sub_id;
  1186. bool is_local;
  1187. u32 flags;
  1188. bool old_proto = true;
  1189. bool old_pkt_proto;
  1190. u16 version;
  1191. vsk = vsock_sk(sk);
  1192. handle = VMCI_INVALID_HANDLE;
  1193. detach_sub_id = VMCI_INVALID_ID;
  1194. /* If we have gotten here then we should be past the point where old
  1195. * linux vsock could have sent the bogus rst.
  1196. */
  1197. vsk->sent_request = false;
  1198. vsk->ignore_connecting_rst = false;
  1199. /* Verify that we're OK with the proposed queue pair size */
  1200. if (pkt->u.size < vsk->buffer_min_size ||
  1201. pkt->u.size > vsk->buffer_max_size) {
  1202. err = -EINVAL;
  1203. goto destroy;
  1204. }
  1205. /* At this point we know the CID the peer is using to talk to us. */
  1206. if (vsk->local_addr.svm_cid == VMADDR_CID_ANY)
  1207. vsk->local_addr.svm_cid = pkt->dg.dst.context;
  1208. /* Setup the notify ops to be the highest supported version that both
  1209. * the server and the client support.
  1210. */
  1211. if (vmci_transport_old_proto_override(&old_pkt_proto)) {
  1212. old_proto = old_pkt_proto;
  1213. } else {
  1214. if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE)
  1215. old_proto = true;
  1216. else if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE2)
  1217. old_proto = false;
  1218. }
  1219. if (old_proto)
  1220. version = VSOCK_PROTO_INVALID;
  1221. else
  1222. version = pkt->proto;
  1223. if (!vmci_transport_proto_to_notify_struct(sk, &version, old_proto)) {
  1224. err = -EINVAL;
  1225. goto destroy;
  1226. }
  1227. /* Subscribe to detach events first.
  1228. *
  1229. * XXX We attach once for each queue pair created for now so it is easy
  1230. * to find the socket (it's provided), but later we should only
  1231. * subscribe once and add a way to lookup sockets by queue pair handle.
  1232. */
  1233. err = vmci_event_subscribe(VMCI_EVENT_QP_PEER_DETACH,
  1234. vmci_transport_peer_detach_cb,
  1235. vmci_trans(vsk), &detach_sub_id);
  1236. if (err < VMCI_SUCCESS) {
  1237. err = vmci_transport_error_to_vsock_error(err);
  1238. goto destroy;
  1239. }
  1240. /* Make VMCI select the handle for us. */
  1241. handle = VMCI_INVALID_HANDLE;
  1242. is_local = vsk->remote_addr.svm_cid == vsk->local_addr.svm_cid;
  1243. flags = is_local ? VMCI_QPFLAG_LOCAL : 0;
  1244. err = vmci_transport_queue_pair_alloc(&qpair,
  1245. &handle,
  1246. pkt->u.size,
  1247. pkt->u.size,
  1248. vsk->remote_addr.svm_cid,
  1249. flags,
  1250. vmci_transport_is_trusted(
  1251. vsk,
  1252. vsk->
  1253. remote_addr.svm_cid));
  1254. if (err < 0)
  1255. goto destroy;
  1256. err = vmci_transport_send_qp_offer(sk, handle);
  1257. if (err < 0) {
  1258. err = vmci_transport_error_to_vsock_error(err);
  1259. goto destroy;
  1260. }
  1261. vmci_trans(vsk)->qp_handle = handle;
  1262. vmci_trans(vsk)->qpair = qpair;
  1263. vmci_trans(vsk)->produce_size = vmci_trans(vsk)->consume_size =
  1264. pkt->u.size;
  1265. vmci_trans(vsk)->detach_sub_id = detach_sub_id;
  1266. vmci_trans(vsk)->notify_ops->process_negotiate(sk);
  1267. return 0;
  1268. destroy:
  1269. if (detach_sub_id != VMCI_INVALID_ID)
  1270. vmci_event_unsubscribe(detach_sub_id);
  1271. if (!vmci_handle_is_invalid(handle))
  1272. vmci_qpair_detach(&qpair);
  1273. return err;
  1274. }
  1275. static int
  1276. vmci_transport_recv_connecting_client_invalid(struct sock *sk,
  1277. struct vmci_transport_packet *pkt)
  1278. {
  1279. int err = 0;
  1280. struct vsock_sock *vsk = vsock_sk(sk);
  1281. if (vsk->sent_request) {
  1282. vsk->sent_request = false;
  1283. vsk->ignore_connecting_rst = true;
  1284. err = vmci_transport_send_conn_request(sk, vsk->buffer_size);
  1285. if (err < 0)
  1286. err = vmci_transport_error_to_vsock_error(err);
  1287. else
  1288. err = 0;
  1289. }
  1290. return err;
  1291. }
  1292. static int vmci_transport_recv_connected(struct sock *sk,
  1293. struct vmci_transport_packet *pkt)
  1294. {
  1295. struct vsock_sock *vsk;
  1296. bool pkt_processed = false;
  1297. /* In cases where we are closing the connection, it's sufficient to
  1298. * mark the state change (and maybe error) and wake up any waiting
  1299. * threads. Since this is a connected socket, it's owned by a user
  1300. * process and will be cleaned up when the failure is passed back on
  1301. * the current or next system call. Our system call implementations
  1302. * must therefore check for error and state changes on entry and when
  1303. * being awoken.
  1304. */
  1305. switch (pkt->type) {
  1306. case VMCI_TRANSPORT_PACKET_TYPE_SHUTDOWN:
  1307. if (pkt->u.mode) {
  1308. vsk = vsock_sk(sk);
  1309. vsk->peer_shutdown |= pkt->u.mode;
  1310. sk->sk_state_change(sk);
  1311. }
  1312. break;
  1313. case VMCI_TRANSPORT_PACKET_TYPE_RST:
  1314. vsk = vsock_sk(sk);
  1315. /* It is possible that we sent our peer a message (e.g a
  1316. * WAITING_READ) right before we got notified that the peer had
  1317. * detached. If that happens then we can get a RST pkt back
  1318. * from our peer even though there is data available for us to
  1319. * read. In that case, don't shutdown the socket completely but
  1320. * instead allow the local client to finish reading data off
  1321. * the queuepair. Always treat a RST pkt in connected mode like
  1322. * a clean shutdown.
  1323. */
  1324. sock_set_flag(sk, SOCK_DONE);
  1325. vsk->peer_shutdown = SHUTDOWN_MASK;
  1326. if (vsock_stream_has_data(vsk) <= 0)
  1327. sk->sk_state = TCP_CLOSING;
  1328. sk->sk_state_change(sk);
  1329. break;
  1330. default:
  1331. vsk = vsock_sk(sk);
  1332. vmci_trans(vsk)->notify_ops->handle_notify_pkt(
  1333. sk, pkt, false, NULL, NULL,
  1334. &pkt_processed);
  1335. if (!pkt_processed)
  1336. return -EINVAL;
  1337. break;
  1338. }
  1339. return 0;
  1340. }
  1341. static int vmci_transport_socket_init(struct vsock_sock *vsk,
  1342. struct vsock_sock *psk)
  1343. {
  1344. vsk->trans = kmalloc(sizeof(struct vmci_transport), GFP_KERNEL);
  1345. if (!vsk->trans)
  1346. return -ENOMEM;
  1347. vmci_trans(vsk)->dg_handle = VMCI_INVALID_HANDLE;
  1348. vmci_trans(vsk)->qp_handle = VMCI_INVALID_HANDLE;
  1349. vmci_trans(vsk)->qpair = NULL;
  1350. vmci_trans(vsk)->produce_size = vmci_trans(vsk)->consume_size = 0;
  1351. vmci_trans(vsk)->detach_sub_id = VMCI_INVALID_ID;
  1352. vmci_trans(vsk)->notify_ops = NULL;
  1353. INIT_LIST_HEAD(&vmci_trans(vsk)->elem);
  1354. vmci_trans(vsk)->sk = &vsk->sk;
  1355. spin_lock_init(&vmci_trans(vsk)->lock);
  1356. return 0;
  1357. }
  1358. static void vmci_transport_free_resources(struct list_head *transport_list)
  1359. {
  1360. while (!list_empty(transport_list)) {
  1361. struct vmci_transport *transport =
  1362. list_first_entry(transport_list, struct vmci_transport,
  1363. elem);
  1364. list_del(&transport->elem);
  1365. if (transport->detach_sub_id != VMCI_INVALID_ID) {
  1366. vmci_event_unsubscribe(transport->detach_sub_id);
  1367. transport->detach_sub_id = VMCI_INVALID_ID;
  1368. }
  1369. if (!vmci_handle_is_invalid(transport->qp_handle)) {
  1370. vmci_qpair_detach(&transport->qpair);
  1371. transport->qp_handle = VMCI_INVALID_HANDLE;
  1372. transport->produce_size = 0;
  1373. transport->consume_size = 0;
  1374. }
  1375. kfree(transport);
  1376. }
  1377. }
  1378. static void vmci_transport_cleanup(struct work_struct *work)
  1379. {
  1380. LIST_HEAD(pending);
  1381. spin_lock_bh(&vmci_transport_cleanup_lock);
  1382. list_replace_init(&vmci_transport_cleanup_list, &pending);
  1383. spin_unlock_bh(&vmci_transport_cleanup_lock);
  1384. vmci_transport_free_resources(&pending);
  1385. }
  1386. static void vmci_transport_destruct(struct vsock_sock *vsk)
  1387. {
  1388. /* transport can be NULL if we hit a failure at init() time */
  1389. if (!vmci_trans(vsk))
  1390. return;
  1391. /* Ensure that the detach callback doesn't use the sk/vsk
  1392. * we are about to destruct.
  1393. */
  1394. spin_lock_bh(&vmci_trans(vsk)->lock);
  1395. vmci_trans(vsk)->sk = NULL;
  1396. spin_unlock_bh(&vmci_trans(vsk)->lock);
  1397. if (vmci_trans(vsk)->notify_ops)
  1398. vmci_trans(vsk)->notify_ops->socket_destruct(vsk);
  1399. spin_lock_bh(&vmci_transport_cleanup_lock);
  1400. list_add(&vmci_trans(vsk)->elem, &vmci_transport_cleanup_list);
  1401. spin_unlock_bh(&vmci_transport_cleanup_lock);
  1402. schedule_work(&vmci_transport_cleanup_work);
  1403. vsk->trans = NULL;
  1404. }
  1405. static void vmci_transport_release(struct vsock_sock *vsk)
  1406. {
  1407. vsock_remove_sock(vsk);
  1408. if (!vmci_handle_is_invalid(vmci_trans(vsk)->dg_handle)) {
  1409. vmci_datagram_destroy_handle(vmci_trans(vsk)->dg_handle);
  1410. vmci_trans(vsk)->dg_handle = VMCI_INVALID_HANDLE;
  1411. }
  1412. }
  1413. static int vmci_transport_dgram_bind(struct vsock_sock *vsk,
  1414. struct sockaddr_vm *addr)
  1415. {
  1416. u32 port;
  1417. u32 flags;
  1418. int err;
  1419. /* VMCI will select a resource ID for us if we provide
  1420. * VMCI_INVALID_ID.
  1421. */
  1422. port = addr->svm_port == VMADDR_PORT_ANY ?
  1423. VMCI_INVALID_ID : addr->svm_port;
  1424. if (port <= LAST_RESERVED_PORT && !capable(CAP_NET_BIND_SERVICE))
  1425. return -EACCES;
  1426. flags = addr->svm_cid == VMADDR_CID_ANY ?
  1427. VMCI_FLAG_ANYCID_DG_HND : 0;
  1428. err = vmci_transport_datagram_create_hnd(port, flags,
  1429. vmci_transport_recv_dgram_cb,
  1430. &vsk->sk,
  1431. &vmci_trans(vsk)->dg_handle);
  1432. if (err < VMCI_SUCCESS)
  1433. return vmci_transport_error_to_vsock_error(err);
  1434. vsock_addr_init(&vsk->local_addr, addr->svm_cid,
  1435. vmci_trans(vsk)->dg_handle.resource);
  1436. return 0;
  1437. }
  1438. static int vmci_transport_dgram_enqueue(
  1439. struct vsock_sock *vsk,
  1440. struct sockaddr_vm *remote_addr,
  1441. struct msghdr *msg,
  1442. size_t len)
  1443. {
  1444. int err;
  1445. struct vmci_datagram *dg;
  1446. if (len > VMCI_MAX_DG_PAYLOAD_SIZE)
  1447. return -EMSGSIZE;
  1448. if (!vmci_transport_allow_dgram(vsk, remote_addr->svm_cid))
  1449. return -EPERM;
  1450. /* Allocate a buffer for the user's message and our packet header. */
  1451. dg = kmalloc(len + sizeof(*dg), GFP_KERNEL);
  1452. if (!dg)
  1453. return -ENOMEM;
  1454. memcpy_from_msg(VMCI_DG_PAYLOAD(dg), msg, len);
  1455. dg->dst = vmci_make_handle(remote_addr->svm_cid,
  1456. remote_addr->svm_port);
  1457. dg->src = vmci_make_handle(vsk->local_addr.svm_cid,
  1458. vsk->local_addr.svm_port);
  1459. dg->payload_size = len;
  1460. err = vmci_datagram_send(dg);
  1461. kfree(dg);
  1462. if (err < 0)
  1463. return vmci_transport_error_to_vsock_error(err);
  1464. return err - sizeof(*dg);
  1465. }
  1466. static int vmci_transport_dgram_dequeue(struct vsock_sock *vsk,
  1467. struct msghdr *msg, size_t len,
  1468. int flags)
  1469. {
  1470. int err;
  1471. int noblock;
  1472. struct vmci_datagram *dg;
  1473. size_t payload_len;
  1474. struct sk_buff *skb;
  1475. noblock = flags & MSG_DONTWAIT;
  1476. if (flags & MSG_OOB || flags & MSG_ERRQUEUE)
  1477. return -EOPNOTSUPP;
  1478. /* Retrieve the head sk_buff from the socket's receive queue. */
  1479. err = 0;
  1480. skb = skb_recv_datagram(&vsk->sk, flags, noblock, &err);
  1481. if (!skb)
  1482. return err;
  1483. dg = (struct vmci_datagram *)skb->data;
  1484. if (!dg)
  1485. /* err is 0, meaning we read zero bytes. */
  1486. goto out;
  1487. payload_len = dg->payload_size;
  1488. /* Ensure the sk_buff matches the payload size claimed in the packet. */
  1489. if (payload_len != skb->len - sizeof(*dg)) {
  1490. err = -EINVAL;
  1491. goto out;
  1492. }
  1493. if (payload_len > len) {
  1494. payload_len = len;
  1495. msg->msg_flags |= MSG_TRUNC;
  1496. }
  1497. /* Place the datagram payload in the user's iovec. */
  1498. err = skb_copy_datagram_msg(skb, sizeof(*dg), msg, payload_len);
  1499. if (err)
  1500. goto out;
  1501. if (msg->msg_name) {
  1502. /* Provide the address of the sender. */
  1503. DECLARE_SOCKADDR(struct sockaddr_vm *, vm_addr, msg->msg_name);
  1504. vsock_addr_init(vm_addr, dg->src.context, dg->src.resource);
  1505. msg->msg_namelen = sizeof(*vm_addr);
  1506. }
  1507. err = payload_len;
  1508. out:
  1509. skb_free_datagram(&vsk->sk, skb);
  1510. return err;
  1511. }
  1512. static bool vmci_transport_dgram_allow(u32 cid, u32 port)
  1513. {
  1514. if (cid == VMADDR_CID_HYPERVISOR) {
  1515. /* Registrations of PBRPC Servers do not modify VMX/Hypervisor
  1516. * state and are allowed.
  1517. */
  1518. return port == VMCI_UNITY_PBRPC_REGISTER;
  1519. }
  1520. return true;
  1521. }
  1522. static int vmci_transport_connect(struct vsock_sock *vsk)
  1523. {
  1524. int err;
  1525. bool old_pkt_proto = false;
  1526. struct sock *sk = &vsk->sk;
  1527. if (vmci_transport_old_proto_override(&old_pkt_proto) &&
  1528. old_pkt_proto) {
  1529. err = vmci_transport_send_conn_request(sk, vsk->buffer_size);
  1530. if (err < 0) {
  1531. sk->sk_state = TCP_CLOSE;
  1532. return err;
  1533. }
  1534. } else {
  1535. int supported_proto_versions =
  1536. vmci_transport_new_proto_supported_versions();
  1537. err = vmci_transport_send_conn_request2(sk, vsk->buffer_size,
  1538. supported_proto_versions);
  1539. if (err < 0) {
  1540. sk->sk_state = TCP_CLOSE;
  1541. return err;
  1542. }
  1543. vsk->sent_request = true;
  1544. }
  1545. return err;
  1546. }
  1547. static ssize_t vmci_transport_stream_dequeue(
  1548. struct vsock_sock *vsk,
  1549. struct msghdr *msg,
  1550. size_t len,
  1551. int flags)
  1552. {
  1553. if (flags & MSG_PEEK)
  1554. return vmci_qpair_peekv(vmci_trans(vsk)->qpair, msg, len, 0);
  1555. else
  1556. return vmci_qpair_dequev(vmci_trans(vsk)->qpair, msg, len, 0);
  1557. }
  1558. static ssize_t vmci_transport_stream_enqueue(
  1559. struct vsock_sock *vsk,
  1560. struct msghdr *msg,
  1561. size_t len)
  1562. {
  1563. return vmci_qpair_enquev(vmci_trans(vsk)->qpair, msg, len, 0);
  1564. }
  1565. static s64 vmci_transport_stream_has_data(struct vsock_sock *vsk)
  1566. {
  1567. return vmci_qpair_consume_buf_ready(vmci_trans(vsk)->qpair);
  1568. }
  1569. static s64 vmci_transport_stream_has_space(struct vsock_sock *vsk)
  1570. {
  1571. return vmci_qpair_produce_free_space(vmci_trans(vsk)->qpair);
  1572. }
  1573. static u64 vmci_transport_stream_rcvhiwat(struct vsock_sock *vsk)
  1574. {
  1575. return vmci_trans(vsk)->consume_size;
  1576. }
  1577. static bool vmci_transport_stream_is_active(struct vsock_sock *vsk)
  1578. {
  1579. return !vmci_handle_is_invalid(vmci_trans(vsk)->qp_handle);
  1580. }
  1581. static int vmci_transport_notify_poll_in(
  1582. struct vsock_sock *vsk,
  1583. size_t target,
  1584. bool *data_ready_now)
  1585. {
  1586. return vmci_trans(vsk)->notify_ops->poll_in(
  1587. &vsk->sk, target, data_ready_now);
  1588. }
  1589. static int vmci_transport_notify_poll_out(
  1590. struct vsock_sock *vsk,
  1591. size_t target,
  1592. bool *space_available_now)
  1593. {
  1594. return vmci_trans(vsk)->notify_ops->poll_out(
  1595. &vsk->sk, target, space_available_now);
  1596. }
  1597. static int vmci_transport_notify_recv_init(
  1598. struct vsock_sock *vsk,
  1599. size_t target,
  1600. struct vsock_transport_recv_notify_data *data)
  1601. {
  1602. return vmci_trans(vsk)->notify_ops->recv_init(
  1603. &vsk->sk, target,
  1604. (struct vmci_transport_recv_notify_data *)data);
  1605. }
  1606. static int vmci_transport_notify_recv_pre_block(
  1607. struct vsock_sock *vsk,
  1608. size_t target,
  1609. struct vsock_transport_recv_notify_data *data)
  1610. {
  1611. return vmci_trans(vsk)->notify_ops->recv_pre_block(
  1612. &vsk->sk, target,
  1613. (struct vmci_transport_recv_notify_data *)data);
  1614. }
  1615. static int vmci_transport_notify_recv_pre_dequeue(
  1616. struct vsock_sock *vsk,
  1617. size_t target,
  1618. struct vsock_transport_recv_notify_data *data)
  1619. {
  1620. return vmci_trans(vsk)->notify_ops->recv_pre_dequeue(
  1621. &vsk->sk, target,
  1622. (struct vmci_transport_recv_notify_data *)data);
  1623. }
  1624. static int vmci_transport_notify_recv_post_dequeue(
  1625. struct vsock_sock *vsk,
  1626. size_t target,
  1627. ssize_t copied,
  1628. bool data_read,
  1629. struct vsock_transport_recv_notify_data *data)
  1630. {
  1631. return vmci_trans(vsk)->notify_ops->recv_post_dequeue(
  1632. &vsk->sk, target, copied, data_read,
  1633. (struct vmci_transport_recv_notify_data *)data);
  1634. }
  1635. static int vmci_transport_notify_send_init(
  1636. struct vsock_sock *vsk,
  1637. struct vsock_transport_send_notify_data *data)
  1638. {
  1639. return vmci_trans(vsk)->notify_ops->send_init(
  1640. &vsk->sk,
  1641. (struct vmci_transport_send_notify_data *)data);
  1642. }
  1643. static int vmci_transport_notify_send_pre_block(
  1644. struct vsock_sock *vsk,
  1645. struct vsock_transport_send_notify_data *data)
  1646. {
  1647. return vmci_trans(vsk)->notify_ops->send_pre_block(
  1648. &vsk->sk,
  1649. (struct vmci_transport_send_notify_data *)data);
  1650. }
  1651. static int vmci_transport_notify_send_pre_enqueue(
  1652. struct vsock_sock *vsk,
  1653. struct vsock_transport_send_notify_data *data)
  1654. {
  1655. return vmci_trans(vsk)->notify_ops->send_pre_enqueue(
  1656. &vsk->sk,
  1657. (struct vmci_transport_send_notify_data *)data);
  1658. }
  1659. static int vmci_transport_notify_send_post_enqueue(
  1660. struct vsock_sock *vsk,
  1661. ssize_t written,
  1662. struct vsock_transport_send_notify_data *data)
  1663. {
  1664. return vmci_trans(vsk)->notify_ops->send_post_enqueue(
  1665. &vsk->sk, written,
  1666. (struct vmci_transport_send_notify_data *)data);
  1667. }
  1668. static bool vmci_transport_old_proto_override(bool *old_pkt_proto)
  1669. {
  1670. if (PROTOCOL_OVERRIDE != -1) {
  1671. if (PROTOCOL_OVERRIDE == 0)
  1672. *old_pkt_proto = true;
  1673. else
  1674. *old_pkt_proto = false;
  1675. pr_info("Proto override in use\n");
  1676. return true;
  1677. }
  1678. return false;
  1679. }
  1680. static bool vmci_transport_proto_to_notify_struct(struct sock *sk,
  1681. u16 *proto,
  1682. bool old_pkt_proto)
  1683. {
  1684. struct vsock_sock *vsk = vsock_sk(sk);
  1685. if (old_pkt_proto) {
  1686. if (*proto != VSOCK_PROTO_INVALID) {
  1687. pr_err("Can't set both an old and new protocol\n");
  1688. return false;
  1689. }
  1690. vmci_trans(vsk)->notify_ops = &vmci_transport_notify_pkt_ops;
  1691. goto exit;
  1692. }
  1693. switch (*proto) {
  1694. case VSOCK_PROTO_PKT_ON_NOTIFY:
  1695. vmci_trans(vsk)->notify_ops =
  1696. &vmci_transport_notify_pkt_q_state_ops;
  1697. break;
  1698. default:
  1699. pr_err("Unknown notify protocol version\n");
  1700. return false;
  1701. }
  1702. exit:
  1703. vmci_trans(vsk)->notify_ops->socket_init(sk);
  1704. return true;
  1705. }
  1706. static u16 vmci_transport_new_proto_supported_versions(void)
  1707. {
  1708. if (PROTOCOL_OVERRIDE != -1)
  1709. return PROTOCOL_OVERRIDE;
  1710. return VSOCK_PROTO_ALL_SUPPORTED;
  1711. }
  1712. static u32 vmci_transport_get_local_cid(void)
  1713. {
  1714. return vmci_get_context_id();
  1715. }
  1716. static struct vsock_transport vmci_transport = {
  1717. .module = THIS_MODULE,
  1718. .init = vmci_transport_socket_init,
  1719. .destruct = vmci_transport_destruct,
  1720. .release = vmci_transport_release,
  1721. .connect = vmci_transport_connect,
  1722. .dgram_bind = vmci_transport_dgram_bind,
  1723. .dgram_dequeue = vmci_transport_dgram_dequeue,
  1724. .dgram_enqueue = vmci_transport_dgram_enqueue,
  1725. .dgram_allow = vmci_transport_dgram_allow,
  1726. .stream_dequeue = vmci_transport_stream_dequeue,
  1727. .stream_enqueue = vmci_transport_stream_enqueue,
  1728. .stream_has_data = vmci_transport_stream_has_data,
  1729. .stream_has_space = vmci_transport_stream_has_space,
  1730. .stream_rcvhiwat = vmci_transport_stream_rcvhiwat,
  1731. .stream_is_active = vmci_transport_stream_is_active,
  1732. .stream_allow = vmci_transport_stream_allow,
  1733. .notify_poll_in = vmci_transport_notify_poll_in,
  1734. .notify_poll_out = vmci_transport_notify_poll_out,
  1735. .notify_recv_init = vmci_transport_notify_recv_init,
  1736. .notify_recv_pre_block = vmci_transport_notify_recv_pre_block,
  1737. .notify_recv_pre_dequeue = vmci_transport_notify_recv_pre_dequeue,
  1738. .notify_recv_post_dequeue = vmci_transport_notify_recv_post_dequeue,
  1739. .notify_send_init = vmci_transport_notify_send_init,
  1740. .notify_send_pre_block = vmci_transport_notify_send_pre_block,
  1741. .notify_send_pre_enqueue = vmci_transport_notify_send_pre_enqueue,
  1742. .notify_send_post_enqueue = vmci_transport_notify_send_post_enqueue,
  1743. .shutdown = vmci_transport_shutdown,
  1744. .get_local_cid = vmci_transport_get_local_cid,
  1745. };
  1746. static bool vmci_check_transport(struct vsock_sock *vsk)
  1747. {
  1748. return vsk->transport == &vmci_transport;
  1749. }
  1750. static void vmci_vsock_transport_cb(bool is_host)
  1751. {
  1752. int features;
  1753. if (is_host)
  1754. features = VSOCK_TRANSPORT_F_H2G;
  1755. else
  1756. features = VSOCK_TRANSPORT_F_G2H;
  1757. vsock_core_register(&vmci_transport, features);
  1758. }
  1759. static int __init vmci_transport_init(void)
  1760. {
  1761. int err;
  1762. /* Create the datagram handle that we will use to send and receive all
  1763. * VSocket control messages for this context.
  1764. */
  1765. err = vmci_transport_datagram_create_hnd(VMCI_TRANSPORT_PACKET_RID,
  1766. VMCI_FLAG_ANYCID_DG_HND,
  1767. vmci_transport_recv_stream_cb,
  1768. NULL,
  1769. &vmci_transport_stream_handle);
  1770. if (err < VMCI_SUCCESS) {
  1771. pr_err("Unable to create datagram handle. (%d)\n", err);
  1772. return vmci_transport_error_to_vsock_error(err);
  1773. }
  1774. err = vmci_event_subscribe(VMCI_EVENT_QP_RESUMED,
  1775. vmci_transport_qp_resumed_cb,
  1776. NULL, &vmci_transport_qp_resumed_sub_id);
  1777. if (err < VMCI_SUCCESS) {
  1778. pr_err("Unable to subscribe to resumed event. (%d)\n", err);
  1779. err = vmci_transport_error_to_vsock_error(err);
  1780. vmci_transport_qp_resumed_sub_id = VMCI_INVALID_ID;
  1781. goto err_destroy_stream_handle;
  1782. }
  1783. /* Register only with dgram feature, other features (H2G, G2H) will be
  1784. * registered when the first host or guest becomes active.
  1785. */
  1786. err = vsock_core_register(&vmci_transport, VSOCK_TRANSPORT_F_DGRAM);
  1787. if (err < 0)
  1788. goto err_unsubscribe;
  1789. err = vmci_register_vsock_callback(vmci_vsock_transport_cb);
  1790. if (err < 0)
  1791. goto err_unregister;
  1792. return 0;
  1793. err_unregister:
  1794. vsock_core_unregister(&vmci_transport);
  1795. err_unsubscribe:
  1796. vmci_event_unsubscribe(vmci_transport_qp_resumed_sub_id);
  1797. err_destroy_stream_handle:
  1798. vmci_datagram_destroy_handle(vmci_transport_stream_handle);
  1799. return err;
  1800. }
  1801. module_init(vmci_transport_init);
  1802. static void __exit vmci_transport_exit(void)
  1803. {
  1804. cancel_work_sync(&vmci_transport_cleanup_work);
  1805. vmci_transport_free_resources(&vmci_transport_cleanup_list);
  1806. if (!vmci_handle_is_invalid(vmci_transport_stream_handle)) {
  1807. if (vmci_datagram_destroy_handle(
  1808. vmci_transport_stream_handle) != VMCI_SUCCESS)
  1809. pr_err("Couldn't destroy datagram handle\n");
  1810. vmci_transport_stream_handle = VMCI_INVALID_HANDLE;
  1811. }
  1812. if (vmci_transport_qp_resumed_sub_id != VMCI_INVALID_ID) {
  1813. vmci_event_unsubscribe(vmci_transport_qp_resumed_sub_id);
  1814. vmci_transport_qp_resumed_sub_id = VMCI_INVALID_ID;
  1815. }
  1816. vmci_register_vsock_callback(NULL);
  1817. vsock_core_unregister(&vmci_transport);
  1818. }
  1819. module_exit(vmci_transport_exit);
  1820. MODULE_AUTHOR("VMware, Inc.");
  1821. MODULE_DESCRIPTION("VMCI transport for Virtual Sockets");
  1822. MODULE_VERSION("1.0.5.0-k");
  1823. MODULE_LICENSE("GPL v2");
  1824. MODULE_ALIAS("vmware_vsock");
  1825. MODULE_ALIAS_NETPROTO(PF_VSOCK);