nfssvc.c 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874
  1. /*
  2. * Central processing for nfsd.
  3. *
  4. * Authors: Olaf Kirch (okir@monad.swb.de)
  5. *
  6. * Copyright (C) 1995, 1996, 1997 Olaf Kirch <okir@monad.swb.de>
  7. */
  8. #include <linux/sched.h>
  9. #include <linux/freezer.h>
  10. #include <linux/module.h>
  11. #include <linux/fs_struct.h>
  12. #include <linux/swap.h>
  13. #include <linux/sunrpc/stats.h>
  14. #include <linux/sunrpc/svcsock.h>
  15. #include <linux/sunrpc/svc_xprt.h>
  16. #include <linux/lockd/bind.h>
  17. #include <linux/nfsacl.h>
  18. #include <linux/seq_file.h>
  19. #include <linux/inetdevice.h>
  20. #include <net/addrconf.h>
  21. #include <net/ipv6.h>
  22. #include <net/net_namespace.h>
  23. #include "nfsd.h"
  24. #include "cache.h"
  25. #include "vfs.h"
  26. #include "netns.h"
  27. #define NFSDDBG_FACILITY NFSDDBG_SVC
  28. extern struct svc_program nfsd_program;
  29. static int nfsd(void *vrqstp);
  30. /*
  31. * nfsd_mutex protects nn->nfsd_serv -- both the pointer itself and the members
  32. * of the svc_serv struct. In particular, ->sv_nrthreads but also to some
  33. * extent ->sv_temp_socks and ->sv_permsocks. It also protects nfsdstats.th_cnt
  34. *
  35. * If (out side the lock) nn->nfsd_serv is non-NULL, then it must point to a
  36. * properly initialised 'struct svc_serv' with ->sv_nrthreads > 0. That number
  37. * of nfsd threads must exist and each must listed in ->sp_all_threads in each
  38. * entry of ->sv_pools[].
  39. *
  40. * Transitions of the thread count between zero and non-zero are of particular
  41. * interest since the svc_serv needs to be created and initialized at that
  42. * point, or freed.
  43. *
  44. * Finally, the nfsd_mutex also protects some of the global variables that are
  45. * accessed when nfsd starts and that are settable via the write_* routines in
  46. * nfsctl.c. In particular:
  47. *
  48. * user_recovery_dirname
  49. * user_lease_time
  50. * nfsd_versions
  51. */
  52. DEFINE_MUTEX(nfsd_mutex);
  53. /*
  54. * nfsd_drc_lock protects nfsd_drc_max_pages and nfsd_drc_pages_used.
  55. * nfsd_drc_max_pages limits the total amount of memory available for
  56. * version 4.1 DRC caches.
  57. * nfsd_drc_pages_used tracks the current version 4.1 DRC memory usage.
  58. */
  59. spinlock_t nfsd_drc_lock;
  60. unsigned long nfsd_drc_max_mem;
  61. unsigned long nfsd_drc_mem_used;
  62. #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
  63. static struct svc_stat nfsd_acl_svcstats;
  64. static struct svc_version * nfsd_acl_version[] = {
  65. [2] = &nfsd_acl_version2,
  66. [3] = &nfsd_acl_version3,
  67. };
  68. #define NFSD_ACL_MINVERS 2
  69. #define NFSD_ACL_NRVERS ARRAY_SIZE(nfsd_acl_version)
  70. static struct svc_version *nfsd_acl_versions[NFSD_ACL_NRVERS];
  71. static struct svc_program nfsd_acl_program = {
  72. .pg_prog = NFS_ACL_PROGRAM,
  73. .pg_nvers = NFSD_ACL_NRVERS,
  74. .pg_vers = nfsd_acl_versions,
  75. .pg_name = "nfsacl",
  76. .pg_class = "nfsd",
  77. .pg_stats = &nfsd_acl_svcstats,
  78. .pg_authenticate = &svc_set_client,
  79. };
  80. static struct svc_stat nfsd_acl_svcstats = {
  81. .program = &nfsd_acl_program,
  82. };
  83. #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */
  84. static struct svc_version * nfsd_version[] = {
  85. [2] = &nfsd_version2,
  86. #if defined(CONFIG_NFSD_V3)
  87. [3] = &nfsd_version3,
  88. #endif
  89. #if defined(CONFIG_NFSD_V4)
  90. [4] = &nfsd_version4,
  91. #endif
  92. };
  93. #define NFSD_MINVERS 2
  94. #define NFSD_NRVERS ARRAY_SIZE(nfsd_version)
  95. static struct svc_version *nfsd_versions[NFSD_NRVERS];
  96. struct svc_program nfsd_program = {
  97. #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
  98. .pg_next = &nfsd_acl_program,
  99. #endif
  100. .pg_prog = NFS_PROGRAM, /* program number */
  101. .pg_nvers = NFSD_NRVERS, /* nr of entries in nfsd_version */
  102. .pg_vers = nfsd_versions, /* version table */
  103. .pg_name = "nfsd", /* program name */
  104. .pg_class = "nfsd", /* authentication class */
  105. .pg_stats = &nfsd_svcstats, /* version table */
  106. .pg_authenticate = &svc_set_client, /* export authentication */
  107. };
  108. static bool nfsd_supported_minorversions[NFSD_SUPPORTED_MINOR_VERSION + 1] = {
  109. [0] = 1,
  110. [1] = 1,
  111. [2] = 1,
  112. };
  113. int nfsd_vers(int vers, enum vers_op change)
  114. {
  115. if (vers < NFSD_MINVERS || vers >= NFSD_NRVERS)
  116. return 0;
  117. switch(change) {
  118. case NFSD_SET:
  119. nfsd_versions[vers] = nfsd_version[vers];
  120. #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
  121. if (vers < NFSD_ACL_NRVERS)
  122. nfsd_acl_versions[vers] = nfsd_acl_version[vers];
  123. #endif
  124. break;
  125. case NFSD_CLEAR:
  126. nfsd_versions[vers] = NULL;
  127. #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
  128. if (vers < NFSD_ACL_NRVERS)
  129. nfsd_acl_versions[vers] = NULL;
  130. #endif
  131. break;
  132. case NFSD_TEST:
  133. return nfsd_versions[vers] != NULL;
  134. case NFSD_AVAIL:
  135. return nfsd_version[vers] != NULL;
  136. }
  137. return 0;
  138. }
  139. int nfsd_minorversion(u32 minorversion, enum vers_op change)
  140. {
  141. if (minorversion > NFSD_SUPPORTED_MINOR_VERSION)
  142. return -1;
  143. switch(change) {
  144. case NFSD_SET:
  145. nfsd_supported_minorversions[minorversion] = true;
  146. break;
  147. case NFSD_CLEAR:
  148. nfsd_supported_minorversions[minorversion] = false;
  149. break;
  150. case NFSD_TEST:
  151. return nfsd_supported_minorversions[minorversion];
  152. case NFSD_AVAIL:
  153. return minorversion <= NFSD_SUPPORTED_MINOR_VERSION;
  154. }
  155. return 0;
  156. }
  157. /*
  158. * Maximum number of nfsd processes
  159. */
  160. #define NFSD_MAXSERVS 8192
  161. int nfsd_nrthreads(struct net *net)
  162. {
  163. int rv = 0;
  164. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  165. mutex_lock(&nfsd_mutex);
  166. if (nn->nfsd_serv)
  167. rv = nn->nfsd_serv->sv_nrthreads;
  168. mutex_unlock(&nfsd_mutex);
  169. return rv;
  170. }
  171. static int nfsd_init_socks(struct net *net)
  172. {
  173. int error;
  174. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  175. if (!list_empty(&nn->nfsd_serv->sv_permsocks))
  176. return 0;
  177. error = svc_create_xprt(nn->nfsd_serv, "udp", net, PF_INET, NFS_PORT,
  178. SVC_SOCK_DEFAULTS);
  179. if (error < 0)
  180. return error;
  181. error = svc_create_xprt(nn->nfsd_serv, "tcp", net, PF_INET, NFS_PORT,
  182. SVC_SOCK_DEFAULTS);
  183. if (error < 0)
  184. return error;
  185. return 0;
  186. }
  187. static int nfsd_users = 0;
  188. static int nfsd_startup_generic(int nrservs)
  189. {
  190. int ret;
  191. if (nfsd_users++)
  192. return 0;
  193. /*
  194. * Readahead param cache - will no-op if it already exists.
  195. * (Note therefore results will be suboptimal if number of
  196. * threads is modified after nfsd start.)
  197. */
  198. ret = nfsd_racache_init(2*nrservs);
  199. if (ret)
  200. goto dec_users;
  201. ret = nfs4_state_start();
  202. if (ret)
  203. goto out_racache;
  204. return 0;
  205. out_racache:
  206. nfsd_racache_shutdown();
  207. dec_users:
  208. nfsd_users--;
  209. return ret;
  210. }
  211. static void nfsd_shutdown_generic(void)
  212. {
  213. if (--nfsd_users)
  214. return;
  215. nfs4_state_shutdown();
  216. nfsd_racache_shutdown();
  217. }
  218. static bool nfsd_needs_lockd(void)
  219. {
  220. #if defined(CONFIG_NFSD_V3)
  221. return (nfsd_versions[2] != NULL) || (nfsd_versions[3] != NULL);
  222. #else
  223. return (nfsd_versions[2] != NULL);
  224. #endif
  225. }
  226. static int nfsd_startup_net(int nrservs, struct net *net)
  227. {
  228. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  229. int ret;
  230. if (nn->nfsd_net_up)
  231. return 0;
  232. ret = nfsd_startup_generic(nrservs);
  233. if (ret)
  234. return ret;
  235. ret = nfsd_init_socks(net);
  236. if (ret)
  237. goto out_socks;
  238. if (nfsd_needs_lockd() && !nn->lockd_up) {
  239. ret = lockd_up(net);
  240. if (ret)
  241. goto out_socks;
  242. nn->lockd_up = 1;
  243. }
  244. ret = nfs4_state_start_net(net);
  245. if (ret)
  246. goto out_lockd;
  247. nn->nfsd_net_up = true;
  248. return 0;
  249. out_lockd:
  250. if (nn->lockd_up) {
  251. lockd_down(net);
  252. nn->lockd_up = 0;
  253. }
  254. out_socks:
  255. nfsd_shutdown_generic();
  256. return ret;
  257. }
  258. static void nfsd_shutdown_net(struct net *net)
  259. {
  260. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  261. nfs4_state_shutdown_net(net);
  262. if (nn->lockd_up) {
  263. lockd_down(net);
  264. nn->lockd_up = 0;
  265. }
  266. nn->nfsd_net_up = false;
  267. nfsd_shutdown_generic();
  268. }
  269. static int nfsd_inetaddr_event(struct notifier_block *this, unsigned long event,
  270. void *ptr)
  271. {
  272. struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
  273. struct net_device *dev = ifa->ifa_dev->dev;
  274. struct net *net = dev_net(dev);
  275. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  276. struct sockaddr_in sin;
  277. if (event != NETDEV_DOWN)
  278. goto out;
  279. if (nn->nfsd_serv) {
  280. dprintk("nfsd_inetaddr_event: removed %pI4\n", &ifa->ifa_local);
  281. sin.sin_family = AF_INET;
  282. sin.sin_addr.s_addr = ifa->ifa_local;
  283. svc_age_temp_xprts_now(nn->nfsd_serv, (struct sockaddr *)&sin);
  284. }
  285. out:
  286. return NOTIFY_DONE;
  287. }
  288. static struct notifier_block nfsd_inetaddr_notifier = {
  289. .notifier_call = nfsd_inetaddr_event,
  290. };
  291. #if IS_ENABLED(CONFIG_IPV6)
  292. static int nfsd_inet6addr_event(struct notifier_block *this,
  293. unsigned long event, void *ptr)
  294. {
  295. struct inet6_ifaddr *ifa = (struct inet6_ifaddr *)ptr;
  296. struct net_device *dev = ifa->idev->dev;
  297. struct net *net = dev_net(dev);
  298. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  299. struct sockaddr_in6 sin6;
  300. if (event != NETDEV_DOWN)
  301. goto out;
  302. if (nn->nfsd_serv) {
  303. dprintk("nfsd_inet6addr_event: removed %pI6\n", &ifa->addr);
  304. sin6.sin6_family = AF_INET6;
  305. sin6.sin6_addr = ifa->addr;
  306. svc_age_temp_xprts_now(nn->nfsd_serv, (struct sockaddr *)&sin6);
  307. }
  308. out:
  309. return NOTIFY_DONE;
  310. }
  311. static struct notifier_block nfsd_inet6addr_notifier = {
  312. .notifier_call = nfsd_inet6addr_event,
  313. };
  314. #endif
  315. /* Only used under nfsd_mutex, so this atomic may be overkill: */
  316. static atomic_t nfsd_notifier_refcount = ATOMIC_INIT(0);
  317. static void nfsd_last_thread(struct svc_serv *serv, struct net *net)
  318. {
  319. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  320. /* check if the notifier still has clients */
  321. if (atomic_dec_return(&nfsd_notifier_refcount) == 0) {
  322. unregister_inetaddr_notifier(&nfsd_inetaddr_notifier);
  323. #if IS_ENABLED(CONFIG_IPV6)
  324. unregister_inet6addr_notifier(&nfsd_inet6addr_notifier);
  325. #endif
  326. }
  327. /*
  328. * write_ports can create the server without actually starting
  329. * any threads--if we get shut down before any threads are
  330. * started, then nfsd_last_thread will be run before any of this
  331. * other initialization has been done except the rpcb information.
  332. */
  333. svc_rpcb_cleanup(serv, net);
  334. if (!nn->nfsd_net_up)
  335. return;
  336. nfsd_shutdown_net(net);
  337. printk(KERN_WARNING "nfsd: last server has exited, flushing export "
  338. "cache\n");
  339. nfsd_export_flush(net);
  340. }
  341. void nfsd_reset_versions(void)
  342. {
  343. int found_one = 0;
  344. int i;
  345. for (i = NFSD_MINVERS; i < NFSD_NRVERS; i++) {
  346. if (nfsd_program.pg_vers[i])
  347. found_one = 1;
  348. }
  349. if (!found_one) {
  350. for (i = NFSD_MINVERS; i < NFSD_NRVERS; i++)
  351. nfsd_program.pg_vers[i] = nfsd_version[i];
  352. #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
  353. for (i = NFSD_ACL_MINVERS; i < NFSD_ACL_NRVERS; i++)
  354. nfsd_acl_program.pg_vers[i] =
  355. nfsd_acl_version[i];
  356. #endif
  357. }
  358. }
  359. /*
  360. * Each session guarantees a negotiated per slot memory cache for replies
  361. * which in turn consumes memory beyond the v2/v3/v4.0 server. A dedicated
  362. * NFSv4.1 server might want to use more memory for a DRC than a machine
  363. * with mutiple services.
  364. *
  365. * Impose a hard limit on the number of pages for the DRC which varies
  366. * according to the machines free pages. This is of course only a default.
  367. *
  368. * For now this is a #defined shift which could be under admin control
  369. * in the future.
  370. */
  371. static void set_max_drc(void)
  372. {
  373. #define NFSD_DRC_SIZE_SHIFT 10
  374. nfsd_drc_max_mem = (nr_free_buffer_pages()
  375. >> NFSD_DRC_SIZE_SHIFT) * PAGE_SIZE;
  376. nfsd_drc_mem_used = 0;
  377. spin_lock_init(&nfsd_drc_lock);
  378. dprintk("%s nfsd_drc_max_mem %lu \n", __func__, nfsd_drc_max_mem);
  379. }
  380. static int nfsd_get_default_max_blksize(void)
  381. {
  382. struct sysinfo i;
  383. unsigned long long target;
  384. unsigned long ret;
  385. si_meminfo(&i);
  386. target = (i.totalram - i.totalhigh) << PAGE_SHIFT;
  387. /*
  388. * Aim for 1/4096 of memory per thread This gives 1MB on 4Gig
  389. * machines, but only uses 32K on 128M machines. Bottom out at
  390. * 8K on 32M and smaller. Of course, this is only a default.
  391. */
  392. target >>= 12;
  393. ret = NFSSVC_MAXBLKSIZE;
  394. while (ret > target && ret >= 8*1024*2)
  395. ret /= 2;
  396. return ret;
  397. }
  398. static struct svc_serv_ops nfsd_thread_sv_ops = {
  399. .svo_shutdown = nfsd_last_thread,
  400. .svo_function = nfsd,
  401. .svo_enqueue_xprt = svc_xprt_do_enqueue,
  402. .svo_setup = svc_set_num_threads,
  403. .svo_module = THIS_MODULE,
  404. };
  405. int nfsd_create_serv(struct net *net)
  406. {
  407. int error;
  408. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  409. WARN_ON(!mutex_is_locked(&nfsd_mutex));
  410. if (nn->nfsd_serv) {
  411. svc_get(nn->nfsd_serv);
  412. return 0;
  413. }
  414. if (nfsd_max_blksize == 0)
  415. nfsd_max_blksize = nfsd_get_default_max_blksize();
  416. nfsd_reset_versions();
  417. nn->nfsd_serv = svc_create_pooled(&nfsd_program, nfsd_max_blksize,
  418. &nfsd_thread_sv_ops);
  419. if (nn->nfsd_serv == NULL)
  420. return -ENOMEM;
  421. nn->nfsd_serv->sv_maxconn = nn->max_connections;
  422. error = svc_bind(nn->nfsd_serv, net);
  423. if (error < 0) {
  424. svc_destroy(nn->nfsd_serv);
  425. return error;
  426. }
  427. set_max_drc();
  428. /* check if the notifier is already set */
  429. if (atomic_inc_return(&nfsd_notifier_refcount) == 1) {
  430. register_inetaddr_notifier(&nfsd_inetaddr_notifier);
  431. #if IS_ENABLED(CONFIG_IPV6)
  432. register_inet6addr_notifier(&nfsd_inet6addr_notifier);
  433. #endif
  434. }
  435. do_gettimeofday(&nn->nfssvc_boot); /* record boot time */
  436. return 0;
  437. }
  438. int nfsd_nrpools(struct net *net)
  439. {
  440. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  441. if (nn->nfsd_serv == NULL)
  442. return 0;
  443. else
  444. return nn->nfsd_serv->sv_nrpools;
  445. }
  446. int nfsd_get_nrthreads(int n, int *nthreads, struct net *net)
  447. {
  448. int i = 0;
  449. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  450. if (nn->nfsd_serv != NULL) {
  451. for (i = 0; i < nn->nfsd_serv->sv_nrpools && i < n; i++)
  452. nthreads[i] = nn->nfsd_serv->sv_pools[i].sp_nrthreads;
  453. }
  454. return 0;
  455. }
  456. void nfsd_destroy(struct net *net)
  457. {
  458. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  459. int destroy = (nn->nfsd_serv->sv_nrthreads == 1);
  460. if (destroy)
  461. svc_shutdown_net(nn->nfsd_serv, net);
  462. svc_destroy(nn->nfsd_serv);
  463. if (destroy)
  464. nn->nfsd_serv = NULL;
  465. }
  466. int nfsd_set_nrthreads(int n, int *nthreads, struct net *net)
  467. {
  468. int i = 0;
  469. int tot = 0;
  470. int err = 0;
  471. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  472. WARN_ON(!mutex_is_locked(&nfsd_mutex));
  473. if (nn->nfsd_serv == NULL || n <= 0)
  474. return 0;
  475. if (n > nn->nfsd_serv->sv_nrpools)
  476. n = nn->nfsd_serv->sv_nrpools;
  477. /* enforce a global maximum number of threads */
  478. tot = 0;
  479. for (i = 0; i < n; i++) {
  480. nthreads[i] = min(nthreads[i], NFSD_MAXSERVS);
  481. tot += nthreads[i];
  482. }
  483. if (tot > NFSD_MAXSERVS) {
  484. /* total too large: scale down requested numbers */
  485. for (i = 0; i < n && tot > 0; i++) {
  486. int new = nthreads[i] * NFSD_MAXSERVS / tot;
  487. tot -= (nthreads[i] - new);
  488. nthreads[i] = new;
  489. }
  490. for (i = 0; i < n && tot > 0; i++) {
  491. nthreads[i]--;
  492. tot--;
  493. }
  494. }
  495. /*
  496. * There must always be a thread in pool 0; the admin
  497. * can't shut down NFS completely using pool_threads.
  498. */
  499. if (nthreads[0] == 0)
  500. nthreads[0] = 1;
  501. /* apply the new numbers */
  502. svc_get(nn->nfsd_serv);
  503. for (i = 0; i < n; i++) {
  504. err = nn->nfsd_serv->sv_ops->svo_setup(nn->nfsd_serv,
  505. &nn->nfsd_serv->sv_pools[i], nthreads[i]);
  506. if (err)
  507. break;
  508. }
  509. nfsd_destroy(net);
  510. return err;
  511. }
  512. /*
  513. * Adjust the number of threads and return the new number of threads.
  514. * This is also the function that starts the server if necessary, if
  515. * this is the first time nrservs is nonzero.
  516. */
  517. int
  518. nfsd_svc(int nrservs, struct net *net)
  519. {
  520. int error;
  521. bool nfsd_up_before;
  522. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  523. mutex_lock(&nfsd_mutex);
  524. dprintk("nfsd: creating service\n");
  525. nrservs = max(nrservs, 0);
  526. nrservs = min(nrservs, NFSD_MAXSERVS);
  527. error = 0;
  528. if (nrservs == 0 && nn->nfsd_serv == NULL)
  529. goto out;
  530. error = nfsd_create_serv(net);
  531. if (error)
  532. goto out;
  533. nfsd_up_before = nn->nfsd_net_up;
  534. error = nfsd_startup_net(nrservs, net);
  535. if (error)
  536. goto out_destroy;
  537. error = nn->nfsd_serv->sv_ops->svo_setup(nn->nfsd_serv,
  538. NULL, nrservs);
  539. if (error)
  540. goto out_shutdown;
  541. /* We are holding a reference to nn->nfsd_serv which
  542. * we don't want to count in the return value,
  543. * so subtract 1
  544. */
  545. error = nn->nfsd_serv->sv_nrthreads - 1;
  546. out_shutdown:
  547. if (error < 0 && !nfsd_up_before)
  548. nfsd_shutdown_net(net);
  549. out_destroy:
  550. nfsd_destroy(net); /* Release server */
  551. out:
  552. mutex_unlock(&nfsd_mutex);
  553. return error;
  554. }
  555. /*
  556. * This is the NFS server kernel thread
  557. */
  558. static int
  559. nfsd(void *vrqstp)
  560. {
  561. struct svc_rqst *rqstp = (struct svc_rqst *) vrqstp;
  562. struct svc_xprt *perm_sock = list_entry(rqstp->rq_server->sv_permsocks.next, typeof(struct svc_xprt), xpt_list);
  563. struct net *net = perm_sock->xpt_net;
  564. struct nfsd_net *nn = net_generic(net, nfsd_net_id);
  565. int err;
  566. /* Lock module and set up kernel thread */
  567. mutex_lock(&nfsd_mutex);
  568. /* At this point, the thread shares current->fs
  569. * with the init process. We need to create files with a
  570. * umask of 0 instead of init's umask. */
  571. if (unshare_fs_struct() < 0) {
  572. printk("Unable to start nfsd thread: out of memory\n");
  573. goto out;
  574. }
  575. current->fs->umask = 0;
  576. /*
  577. * thread is spawned with all signals set to SIG_IGN, re-enable
  578. * the ones that will bring down the thread
  579. */
  580. allow_signal(SIGKILL);
  581. allow_signal(SIGHUP);
  582. allow_signal(SIGINT);
  583. allow_signal(SIGQUIT);
  584. nfsdstats.th_cnt++;
  585. mutex_unlock(&nfsd_mutex);
  586. set_freezable();
  587. /*
  588. * The main request loop
  589. */
  590. for (;;) {
  591. /* Update sv_maxconn if it has changed */
  592. rqstp->rq_server->sv_maxconn = nn->max_connections;
  593. /*
  594. * Find a socket with data available and call its
  595. * recvfrom routine.
  596. */
  597. while ((err = svc_recv(rqstp, 60*60*HZ)) == -EAGAIN)
  598. ;
  599. if (err == -EINTR)
  600. break;
  601. validate_process_creds();
  602. svc_process(rqstp);
  603. validate_process_creds();
  604. }
  605. /* Clear signals before calling svc_exit_thread() */
  606. flush_signals(current);
  607. mutex_lock(&nfsd_mutex);
  608. nfsdstats.th_cnt --;
  609. out:
  610. rqstp->rq_server = NULL;
  611. /* Release the thread */
  612. svc_exit_thread(rqstp);
  613. nfsd_destroy(net);
  614. /* Release module */
  615. mutex_unlock(&nfsd_mutex);
  616. module_put_and_exit(0);
  617. return 0;
  618. }
  619. static __be32 map_new_errors(u32 vers, __be32 nfserr)
  620. {
  621. if (nfserr == nfserr_jukebox && vers == 2)
  622. return nfserr_dropit;
  623. if (nfserr == nfserr_wrongsec && vers < 4)
  624. return nfserr_acces;
  625. return nfserr;
  626. }
  627. /*
  628. * A write procedure can have a large argument, and a read procedure can
  629. * have a large reply, but no NFSv2 or NFSv3 procedure has argument and
  630. * reply that can both be larger than a page. The xdr code has taken
  631. * advantage of this assumption to be a sloppy about bounds checking in
  632. * some cases. Pending a rewrite of the NFSv2/v3 xdr code to fix that
  633. * problem, we enforce these assumptions here:
  634. */
  635. static bool nfs_request_too_big(struct svc_rqst *rqstp,
  636. struct svc_procedure *proc)
  637. {
  638. /*
  639. * The ACL code has more careful bounds-checking and is not
  640. * susceptible to this problem:
  641. */
  642. if (rqstp->rq_prog != NFS_PROGRAM)
  643. return false;
  644. /*
  645. * Ditto NFSv4 (which can in theory have argument and reply both
  646. * more than a page):
  647. */
  648. if (rqstp->rq_vers >= 4)
  649. return false;
  650. /* The reply will be small, we're OK: */
  651. if (proc->pc_xdrressize > 0 &&
  652. proc->pc_xdrressize < XDR_QUADLEN(PAGE_SIZE))
  653. return false;
  654. return rqstp->rq_arg.len > PAGE_SIZE;
  655. }
  656. int
  657. nfsd_dispatch(struct svc_rqst *rqstp, __be32 *statp)
  658. {
  659. struct svc_procedure *proc;
  660. kxdrproc_t xdr;
  661. __be32 nfserr;
  662. __be32 *nfserrp;
  663. dprintk("nfsd_dispatch: vers %d proc %d\n",
  664. rqstp->rq_vers, rqstp->rq_proc);
  665. proc = rqstp->rq_procinfo;
  666. if (nfs_request_too_big(rqstp, proc)) {
  667. dprintk("nfsd: NFSv%d argument too large\n", rqstp->rq_vers);
  668. *statp = rpc_garbage_args;
  669. return 1;
  670. }
  671. /*
  672. * Give the xdr decoder a chance to change this if it wants
  673. * (necessary in the NFSv4.0 compound case)
  674. */
  675. rqstp->rq_cachetype = proc->pc_cachetype;
  676. /* Decode arguments */
  677. xdr = proc->pc_decode;
  678. if (xdr && !xdr(rqstp, (__be32*)rqstp->rq_arg.head[0].iov_base,
  679. rqstp->rq_argp)) {
  680. dprintk("nfsd: failed to decode arguments!\n");
  681. *statp = rpc_garbage_args;
  682. return 1;
  683. }
  684. /* Check whether we have this call in the cache. */
  685. switch (nfsd_cache_lookup(rqstp)) {
  686. case RC_DROPIT:
  687. return 0;
  688. case RC_REPLY:
  689. return 1;
  690. case RC_DOIT:;
  691. /* do it */
  692. }
  693. /* need to grab the location to store the status, as
  694. * nfsv4 does some encoding while processing
  695. */
  696. nfserrp = rqstp->rq_res.head[0].iov_base
  697. + rqstp->rq_res.head[0].iov_len;
  698. rqstp->rq_res.head[0].iov_len += sizeof(__be32);
  699. /* Now call the procedure handler, and encode NFS status. */
  700. nfserr = proc->pc_func(rqstp, rqstp->rq_argp, rqstp->rq_resp);
  701. nfserr = map_new_errors(rqstp->rq_vers, nfserr);
  702. if (nfserr == nfserr_dropit || test_bit(RQ_DROPME, &rqstp->rq_flags)) {
  703. dprintk("nfsd: Dropping request; may be revisited later\n");
  704. nfsd_cache_update(rqstp, RC_NOCACHE, NULL);
  705. return 0;
  706. }
  707. if (rqstp->rq_proc != 0)
  708. *nfserrp++ = nfserr;
  709. /* Encode result.
  710. * For NFSv2, additional info is never returned in case of an error.
  711. */
  712. if (!(nfserr && rqstp->rq_vers == 2)) {
  713. xdr = proc->pc_encode;
  714. if (xdr && !xdr(rqstp, nfserrp,
  715. rqstp->rq_resp)) {
  716. /* Failed to encode result. Release cache entry */
  717. dprintk("nfsd: failed to encode result!\n");
  718. nfsd_cache_update(rqstp, RC_NOCACHE, NULL);
  719. *statp = rpc_system_err;
  720. return 1;
  721. }
  722. }
  723. /* Store reply in cache. */
  724. nfsd_cache_update(rqstp, rqstp->rq_cachetype, statp + 1);
  725. return 1;
  726. }
  727. int nfsd_pool_stats_open(struct inode *inode, struct file *file)
  728. {
  729. int ret;
  730. struct nfsd_net *nn = net_generic(inode->i_sb->s_fs_info, nfsd_net_id);
  731. mutex_lock(&nfsd_mutex);
  732. if (nn->nfsd_serv == NULL) {
  733. mutex_unlock(&nfsd_mutex);
  734. return -ENODEV;
  735. }
  736. /* bump up the psudo refcount while traversing */
  737. svc_get(nn->nfsd_serv);
  738. ret = svc_pool_stats_open(nn->nfsd_serv, file);
  739. mutex_unlock(&nfsd_mutex);
  740. return ret;
  741. }
  742. int nfsd_pool_stats_release(struct inode *inode, struct file *file)
  743. {
  744. int ret = seq_release(inode, file);
  745. struct net *net = inode->i_sb->s_fs_info;
  746. mutex_lock(&nfsd_mutex);
  747. /* this function really, really should have been called svc_put() */
  748. nfsd_destroy(net);
  749. mutex_unlock(&nfsd_mutex);
  750. return ret;
  751. }