ip6_flowlabel.c 19 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818
  1. /*
  2. * ip6_flowlabel.c IPv6 flowlabel manager.
  3. *
  4. * This program is free software; you can redistribute it and/or
  5. * modify it under the terms of the GNU General Public License
  6. * as published by the Free Software Foundation; either version
  7. * 2 of the License, or (at your option) any later version.
  8. *
  9. * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10. */
  11. #include <linux/capability.h>
  12. #include <linux/errno.h>
  13. #include <linux/types.h>
  14. #include <linux/socket.h>
  15. #include <linux/net.h>
  16. #include <linux/netdevice.h>
  17. #include <linux/if_arp.h>
  18. #include <linux/in6.h>
  19. #include <linux/route.h>
  20. #include <linux/proc_fs.h>
  21. #include <linux/seq_file.h>
  22. #include <linux/slab.h>
  23. #include <linux/export.h>
  24. #include <linux/pid_namespace.h>
  25. #include <net/net_namespace.h>
  26. #include <net/sock.h>
  27. #include <net/ipv6.h>
  28. #include <net/ndisc.h>
  29. #include <net/protocol.h>
  30. #include <net/ip6_route.h>
  31. #include <net/addrconf.h>
  32. #include <net/rawv6.h>
  33. #include <net/icmp.h>
  34. #include <net/transp_v6.h>
  35. #include <asm/uaccess.h>
  36. #define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified
  37. in old IPv6 RFC. Well, it was reasonable value.
  38. */
  39. #define FL_MAX_LINGER 60 /* Maximal linger timeout */
  40. /* FL hash table */
  41. #define FL_MAX_PER_SOCK 32
  42. #define FL_MAX_SIZE 4096
  43. #define FL_HASH_MASK 255
  44. #define FL_HASH(l) (ntohl(l)&FL_HASH_MASK)
  45. static atomic_t fl_size = ATOMIC_INIT(0);
  46. static struct ip6_flowlabel *fl_ht[FL_HASH_MASK+1];
  47. static void ip6_fl_gc(unsigned long dummy);
  48. static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
  49. /* FL hash table lock: it protects only of GC */
  50. static DEFINE_RWLOCK(ip6_fl_lock);
  51. /* Big socket sock */
  52. static DEFINE_RWLOCK(ip6_sk_fl_lock);
  53. static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label)
  54. {
  55. struct ip6_flowlabel *fl;
  56. for (fl=fl_ht[FL_HASH(label)]; fl; fl = fl->next) {
  57. if (fl->label == label && net_eq(fl->fl_net, net))
  58. return fl;
  59. }
  60. return NULL;
  61. }
  62. static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label)
  63. {
  64. struct ip6_flowlabel *fl;
  65. read_lock_bh(&ip6_fl_lock);
  66. fl = __fl_lookup(net, label);
  67. if (fl)
  68. atomic_inc(&fl->users);
  69. read_unlock_bh(&ip6_fl_lock);
  70. return fl;
  71. }
  72. static void fl_free(struct ip6_flowlabel *fl)
  73. {
  74. if (fl) {
  75. if (fl->share == IPV6_FL_S_PROCESS)
  76. put_pid(fl->owner.pid);
  77. release_net(fl->fl_net);
  78. kfree(fl->opt);
  79. }
  80. kfree(fl);
  81. }
  82. static void fl_release(struct ip6_flowlabel *fl)
  83. {
  84. write_lock_bh(&ip6_fl_lock);
  85. fl->lastuse = jiffies;
  86. if (atomic_dec_and_test(&fl->users)) {
  87. unsigned long ttd = fl->lastuse + fl->linger;
  88. if (time_after(ttd, fl->expires))
  89. fl->expires = ttd;
  90. ttd = fl->expires;
  91. if (fl->opt && fl->share == IPV6_FL_S_EXCL) {
  92. struct ipv6_txoptions *opt = fl->opt;
  93. fl->opt = NULL;
  94. kfree(opt);
  95. }
  96. if (!timer_pending(&ip6_fl_gc_timer) ||
  97. time_after(ip6_fl_gc_timer.expires, ttd))
  98. mod_timer(&ip6_fl_gc_timer, ttd);
  99. }
  100. write_unlock_bh(&ip6_fl_lock);
  101. }
  102. static void ip6_fl_gc(unsigned long dummy)
  103. {
  104. int i;
  105. unsigned long now = jiffies;
  106. unsigned long sched = 0;
  107. write_lock(&ip6_fl_lock);
  108. for (i=0; i<=FL_HASH_MASK; i++) {
  109. struct ip6_flowlabel *fl, **flp;
  110. flp = &fl_ht[i];
  111. while ((fl=*flp) != NULL) {
  112. if (atomic_read(&fl->users) == 0) {
  113. unsigned long ttd = fl->lastuse + fl->linger;
  114. if (time_after(ttd, fl->expires))
  115. fl->expires = ttd;
  116. ttd = fl->expires;
  117. if (time_after_eq(now, ttd)) {
  118. *flp = fl->next;
  119. fl_free(fl);
  120. atomic_dec(&fl_size);
  121. continue;
  122. }
  123. if (!sched || time_before(ttd, sched))
  124. sched = ttd;
  125. }
  126. flp = &fl->next;
  127. }
  128. }
  129. if (!sched && atomic_read(&fl_size))
  130. sched = now + FL_MAX_LINGER;
  131. if (sched) {
  132. mod_timer(&ip6_fl_gc_timer, sched);
  133. }
  134. write_unlock(&ip6_fl_lock);
  135. }
  136. static void __net_exit ip6_fl_purge(struct net *net)
  137. {
  138. int i;
  139. write_lock(&ip6_fl_lock);
  140. for (i = 0; i <= FL_HASH_MASK; i++) {
  141. struct ip6_flowlabel *fl, **flp;
  142. flp = &fl_ht[i];
  143. while ((fl = *flp) != NULL) {
  144. if (net_eq(fl->fl_net, net) &&
  145. atomic_read(&fl->users) == 0) {
  146. *flp = fl->next;
  147. fl_free(fl);
  148. atomic_dec(&fl_size);
  149. continue;
  150. }
  151. flp = &fl->next;
  152. }
  153. }
  154. write_unlock(&ip6_fl_lock);
  155. }
  156. static struct ip6_flowlabel *fl_intern(struct net *net,
  157. struct ip6_flowlabel *fl, __be32 label)
  158. {
  159. struct ip6_flowlabel *lfl;
  160. fl->label = label & IPV6_FLOWLABEL_MASK;
  161. write_lock_bh(&ip6_fl_lock);
  162. if (label == 0) {
  163. for (;;) {
  164. fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK;
  165. if (fl->label) {
  166. lfl = __fl_lookup(net, fl->label);
  167. if (lfl == NULL)
  168. break;
  169. }
  170. }
  171. } else {
  172. /*
  173. * we dropper the ip6_fl_lock, so this entry could reappear
  174. * and we need to recheck with it.
  175. *
  176. * OTOH no need to search the active socket first, like it is
  177. * done in ipv6_flowlabel_opt - sock is locked, so new entry
  178. * with the same label can only appear on another sock
  179. */
  180. lfl = __fl_lookup(net, fl->label);
  181. if (lfl != NULL) {
  182. atomic_inc(&lfl->users);
  183. write_unlock_bh(&ip6_fl_lock);
  184. return lfl;
  185. }
  186. }
  187. fl->lastuse = jiffies;
  188. fl->next = fl_ht[FL_HASH(fl->label)];
  189. fl_ht[FL_HASH(fl->label)] = fl;
  190. atomic_inc(&fl_size);
  191. write_unlock_bh(&ip6_fl_lock);
  192. return NULL;
  193. }
  194. /* Socket flowlabel lists */
  195. struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label)
  196. {
  197. struct ipv6_fl_socklist *sfl;
  198. struct ipv6_pinfo *np = inet6_sk(sk);
  199. label &= IPV6_FLOWLABEL_MASK;
  200. read_lock_bh(&ip6_sk_fl_lock);
  201. for (sfl=np->ipv6_fl_list; sfl; sfl = sfl->next) {
  202. struct ip6_flowlabel *fl = sfl->fl;
  203. if (fl->label == label) {
  204. fl->lastuse = jiffies;
  205. atomic_inc(&fl->users);
  206. read_unlock_bh(&ip6_sk_fl_lock);
  207. return fl;
  208. }
  209. }
  210. read_unlock_bh(&ip6_sk_fl_lock);
  211. return NULL;
  212. }
  213. EXPORT_SYMBOL_GPL(fl6_sock_lookup);
  214. void fl6_free_socklist(struct sock *sk)
  215. {
  216. struct ipv6_pinfo *np = inet6_sk(sk);
  217. struct ipv6_fl_socklist *sfl;
  218. while ((sfl = np->ipv6_fl_list) != NULL) {
  219. np->ipv6_fl_list = sfl->next;
  220. fl_release(sfl->fl);
  221. kfree(sfl);
  222. }
  223. }
  224. /* Service routines */
  225. /*
  226. It is the only difficult place. flowlabel enforces equal headers
  227. before and including routing header, however user may supply options
  228. following rthdr.
  229. */
  230. struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space,
  231. struct ip6_flowlabel * fl,
  232. struct ipv6_txoptions * fopt)
  233. {
  234. struct ipv6_txoptions * fl_opt = fl->opt;
  235. if (fopt == NULL || fopt->opt_flen == 0)
  236. return fl_opt;
  237. if (fl_opt != NULL) {
  238. opt_space->hopopt = fl_opt->hopopt;
  239. opt_space->dst0opt = fl_opt->dst0opt;
  240. opt_space->srcrt = fl_opt->srcrt;
  241. opt_space->opt_nflen = fl_opt->opt_nflen;
  242. } else {
  243. if (fopt->opt_nflen == 0)
  244. return fopt;
  245. opt_space->hopopt = NULL;
  246. opt_space->dst0opt = NULL;
  247. opt_space->srcrt = NULL;
  248. opt_space->opt_nflen = 0;
  249. }
  250. opt_space->dst1opt = fopt->dst1opt;
  251. opt_space->opt_flen = fopt->opt_flen;
  252. return opt_space;
  253. }
  254. EXPORT_SYMBOL_GPL(fl6_merge_options);
  255. static unsigned long check_linger(unsigned long ttl)
  256. {
  257. if (ttl < FL_MIN_LINGER)
  258. return FL_MIN_LINGER*HZ;
  259. if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
  260. return 0;
  261. return ttl*HZ;
  262. }
  263. static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
  264. {
  265. linger = check_linger(linger);
  266. if (!linger)
  267. return -EPERM;
  268. expires = check_linger(expires);
  269. if (!expires)
  270. return -EPERM;
  271. fl->lastuse = jiffies;
  272. if (time_before(fl->linger, linger))
  273. fl->linger = linger;
  274. if (time_before(expires, fl->linger))
  275. expires = fl->linger;
  276. if (time_before(fl->expires, fl->lastuse + expires))
  277. fl->expires = fl->lastuse + expires;
  278. return 0;
  279. }
  280. static struct ip6_flowlabel *
  281. fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq,
  282. char __user *optval, int optlen, int *err_p)
  283. {
  284. struct ip6_flowlabel *fl = NULL;
  285. int olen;
  286. int addr_type;
  287. int err;
  288. olen = optlen - CMSG_ALIGN(sizeof(*freq));
  289. err = -EINVAL;
  290. if (olen > 64 * 1024)
  291. goto done;
  292. err = -ENOMEM;
  293. fl = kzalloc(sizeof(*fl), GFP_KERNEL);
  294. if (fl == NULL)
  295. goto done;
  296. if (olen > 0) {
  297. struct msghdr msg;
  298. struct flowi6 flowi6;
  299. int junk;
  300. err = -ENOMEM;
  301. fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
  302. if (fl->opt == NULL)
  303. goto done;
  304. memset(fl->opt, 0, sizeof(*fl->opt));
  305. fl->opt->tot_len = sizeof(*fl->opt) + olen;
  306. err = -EFAULT;
  307. if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
  308. goto done;
  309. msg.msg_controllen = olen;
  310. msg.msg_control = (void*)(fl->opt+1);
  311. memset(&flowi6, 0, sizeof(flowi6));
  312. err = datagram_send_ctl(net, sk, &msg, &flowi6, fl->opt, &junk,
  313. &junk, &junk);
  314. if (err)
  315. goto done;
  316. err = -EINVAL;
  317. if (fl->opt->opt_flen)
  318. goto done;
  319. if (fl->opt->opt_nflen == 0) {
  320. kfree(fl->opt);
  321. fl->opt = NULL;
  322. }
  323. }
  324. fl->fl_net = hold_net(net);
  325. fl->expires = jiffies;
  326. err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
  327. if (err)
  328. goto done;
  329. fl->share = freq->flr_share;
  330. addr_type = ipv6_addr_type(&freq->flr_dst);
  331. if ((addr_type & IPV6_ADDR_MAPPED) ||
  332. addr_type == IPV6_ADDR_ANY) {
  333. err = -EINVAL;
  334. goto done;
  335. }
  336. fl->dst = freq->flr_dst;
  337. atomic_set(&fl->users, 1);
  338. switch (fl->share) {
  339. case IPV6_FL_S_EXCL:
  340. case IPV6_FL_S_ANY:
  341. break;
  342. case IPV6_FL_S_PROCESS:
  343. fl->owner.pid = get_task_pid(current, PIDTYPE_PID);
  344. break;
  345. case IPV6_FL_S_USER:
  346. fl->owner.uid = current_euid();
  347. break;
  348. default:
  349. err = -EINVAL;
  350. goto done;
  351. }
  352. return fl;
  353. done:
  354. fl_free(fl);
  355. *err_p = err;
  356. return NULL;
  357. }
  358. static int mem_check(struct sock *sk)
  359. {
  360. struct ipv6_pinfo *np = inet6_sk(sk);
  361. struct ipv6_fl_socklist *sfl;
  362. int room = FL_MAX_SIZE - atomic_read(&fl_size);
  363. int count = 0;
  364. if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
  365. return 0;
  366. for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next)
  367. count++;
  368. if (room <= 0 ||
  369. ((count >= FL_MAX_PER_SOCK ||
  370. (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) &&
  371. !capable(CAP_NET_ADMIN)))
  372. return -ENOBUFS;
  373. return 0;
  374. }
  375. static bool ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2)
  376. {
  377. if (h1 == h2)
  378. return false;
  379. if (h1 == NULL || h2 == NULL)
  380. return true;
  381. if (h1->hdrlen != h2->hdrlen)
  382. return true;
  383. return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1));
  384. }
  385. static bool ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2)
  386. {
  387. if (o1 == o2)
  388. return false;
  389. if (o1 == NULL || o2 == NULL)
  390. return true;
  391. if (o1->opt_nflen != o2->opt_nflen)
  392. return true;
  393. if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt))
  394. return true;
  395. if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt))
  396. return true;
  397. if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt))
  398. return true;
  399. return false;
  400. }
  401. static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
  402. struct ip6_flowlabel *fl)
  403. {
  404. write_lock_bh(&ip6_sk_fl_lock);
  405. sfl->fl = fl;
  406. sfl->next = np->ipv6_fl_list;
  407. np->ipv6_fl_list = sfl;
  408. write_unlock_bh(&ip6_sk_fl_lock);
  409. }
  410. int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
  411. {
  412. int uninitialized_var(err);
  413. struct net *net = sock_net(sk);
  414. struct ipv6_pinfo *np = inet6_sk(sk);
  415. struct in6_flowlabel_req freq;
  416. struct ipv6_fl_socklist *sfl1=NULL;
  417. struct ipv6_fl_socklist *sfl, **sflp;
  418. struct ip6_flowlabel *fl, *fl1 = NULL;
  419. if (optlen < sizeof(freq))
  420. return -EINVAL;
  421. if (copy_from_user(&freq, optval, sizeof(freq)))
  422. return -EFAULT;
  423. switch (freq.flr_action) {
  424. case IPV6_FL_A_PUT:
  425. write_lock_bh(&ip6_sk_fl_lock);
  426. for (sflp = &np->ipv6_fl_list; (sfl=*sflp)!=NULL; sflp = &sfl->next) {
  427. if (sfl->fl->label == freq.flr_label) {
  428. if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
  429. np->flow_label &= ~IPV6_FLOWLABEL_MASK;
  430. *sflp = sfl->next;
  431. write_unlock_bh(&ip6_sk_fl_lock);
  432. fl_release(sfl->fl);
  433. kfree(sfl);
  434. return 0;
  435. }
  436. }
  437. write_unlock_bh(&ip6_sk_fl_lock);
  438. return -ESRCH;
  439. case IPV6_FL_A_RENEW:
  440. read_lock_bh(&ip6_sk_fl_lock);
  441. for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
  442. if (sfl->fl->label == freq.flr_label) {
  443. err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
  444. read_unlock_bh(&ip6_sk_fl_lock);
  445. return err;
  446. }
  447. }
  448. read_unlock_bh(&ip6_sk_fl_lock);
  449. if (freq.flr_share == IPV6_FL_S_NONE &&
  450. ns_capable(net->user_ns, CAP_NET_ADMIN)) {
  451. fl = fl_lookup(net, freq.flr_label);
  452. if (fl) {
  453. err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
  454. fl_release(fl);
  455. return err;
  456. }
  457. }
  458. return -ESRCH;
  459. case IPV6_FL_A_GET:
  460. if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
  461. return -EINVAL;
  462. fl = fl_create(net, sk, &freq, optval, optlen, &err);
  463. if (fl == NULL)
  464. return err;
  465. sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
  466. if (freq.flr_label) {
  467. err = -EEXIST;
  468. read_lock_bh(&ip6_sk_fl_lock);
  469. for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
  470. if (sfl->fl->label == freq.flr_label) {
  471. if (freq.flr_flags&IPV6_FL_F_EXCL) {
  472. read_unlock_bh(&ip6_sk_fl_lock);
  473. goto done;
  474. }
  475. fl1 = sfl->fl;
  476. atomic_inc(&fl1->users);
  477. break;
  478. }
  479. }
  480. read_unlock_bh(&ip6_sk_fl_lock);
  481. if (fl1 == NULL)
  482. fl1 = fl_lookup(net, freq.flr_label);
  483. if (fl1) {
  484. recheck:
  485. err = -EEXIST;
  486. if (freq.flr_flags&IPV6_FL_F_EXCL)
  487. goto release;
  488. err = -EPERM;
  489. if (fl1->share == IPV6_FL_S_EXCL ||
  490. fl1->share != fl->share ||
  491. ((fl1->share == IPV6_FL_S_PROCESS) &&
  492. (fl1->owner.pid == fl->owner.pid)) ||
  493. ((fl1->share == IPV6_FL_S_USER) &&
  494. uid_eq(fl1->owner.uid, fl->owner.uid)))
  495. goto release;
  496. err = -EINVAL;
  497. if (!ipv6_addr_equal(&fl1->dst, &fl->dst) ||
  498. ipv6_opt_cmp(fl1->opt, fl->opt))
  499. goto release;
  500. err = -ENOMEM;
  501. if (sfl1 == NULL)
  502. goto release;
  503. if (fl->linger > fl1->linger)
  504. fl1->linger = fl->linger;
  505. if ((long)(fl->expires - fl1->expires) > 0)
  506. fl1->expires = fl->expires;
  507. fl_link(np, sfl1, fl1);
  508. fl_free(fl);
  509. return 0;
  510. release:
  511. fl_release(fl1);
  512. goto done;
  513. }
  514. }
  515. err = -ENOENT;
  516. if (!(freq.flr_flags&IPV6_FL_F_CREATE))
  517. goto done;
  518. err = -ENOMEM;
  519. if (sfl1 == NULL || (err = mem_check(sk)) != 0)
  520. goto done;
  521. fl1 = fl_intern(net, fl, freq.flr_label);
  522. if (fl1 != NULL)
  523. goto recheck;
  524. if (!freq.flr_label) {
  525. if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
  526. &fl->label, sizeof(fl->label))) {
  527. /* Intentionally ignore fault. */
  528. }
  529. }
  530. fl_link(np, sfl1, fl);
  531. return 0;
  532. default:
  533. return -EINVAL;
  534. }
  535. done:
  536. fl_free(fl);
  537. kfree(sfl1);
  538. return err;
  539. }
  540. #ifdef CONFIG_PROC_FS
  541. struct ip6fl_iter_state {
  542. struct seq_net_private p;
  543. struct pid_namespace *pid_ns;
  544. int bucket;
  545. };
  546. #define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private)
  547. static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
  548. {
  549. struct ip6_flowlabel *fl = NULL;
  550. struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
  551. struct net *net = seq_file_net(seq);
  552. for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
  553. fl = fl_ht[state->bucket];
  554. while (fl && !net_eq(fl->fl_net, net))
  555. fl = fl->next;
  556. if (fl)
  557. break;
  558. }
  559. return fl;
  560. }
  561. static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
  562. {
  563. struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
  564. struct net *net = seq_file_net(seq);
  565. fl = fl->next;
  566. try_again:
  567. while (fl && !net_eq(fl->fl_net, net))
  568. fl = fl->next;
  569. while (!fl) {
  570. if (++state->bucket <= FL_HASH_MASK) {
  571. fl = fl_ht[state->bucket];
  572. goto try_again;
  573. } else
  574. break;
  575. }
  576. return fl;
  577. }
  578. static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
  579. {
  580. struct ip6_flowlabel *fl = ip6fl_get_first(seq);
  581. if (fl)
  582. while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
  583. --pos;
  584. return pos ? NULL : fl;
  585. }
  586. static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
  587. __acquires(ip6_fl_lock)
  588. {
  589. read_lock_bh(&ip6_fl_lock);
  590. return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
  591. }
  592. static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
  593. {
  594. struct ip6_flowlabel *fl;
  595. if (v == SEQ_START_TOKEN)
  596. fl = ip6fl_get_first(seq);
  597. else
  598. fl = ip6fl_get_next(seq, v);
  599. ++*pos;
  600. return fl;
  601. }
  602. static void ip6fl_seq_stop(struct seq_file *seq, void *v)
  603. __releases(ip6_fl_lock)
  604. {
  605. read_unlock_bh(&ip6_fl_lock);
  606. }
  607. static int ip6fl_seq_show(struct seq_file *seq, void *v)
  608. {
  609. struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
  610. if (v == SEQ_START_TOKEN)
  611. seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n",
  612. "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt");
  613. else {
  614. struct ip6_flowlabel *fl = v;
  615. seq_printf(seq,
  616. "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
  617. (unsigned int)ntohl(fl->label),
  618. fl->share,
  619. ((fl->share == IPV6_FL_S_PROCESS) ?
  620. pid_nr_ns(fl->owner.pid, state->pid_ns) :
  621. ((fl->share == IPV6_FL_S_USER) ?
  622. from_kuid_munged(seq_user_ns(seq), fl->owner.uid) :
  623. 0)),
  624. atomic_read(&fl->users),
  625. fl->linger/HZ,
  626. (long)(fl->expires - jiffies)/HZ,
  627. &fl->dst,
  628. fl->opt ? fl->opt->opt_nflen : 0);
  629. }
  630. return 0;
  631. }
  632. static const struct seq_operations ip6fl_seq_ops = {
  633. .start = ip6fl_seq_start,
  634. .next = ip6fl_seq_next,
  635. .stop = ip6fl_seq_stop,
  636. .show = ip6fl_seq_show,
  637. };
  638. static int ip6fl_seq_open(struct inode *inode, struct file *file)
  639. {
  640. struct seq_file *seq;
  641. struct ip6fl_iter_state *state;
  642. int err;
  643. err = seq_open_net(inode, file, &ip6fl_seq_ops,
  644. sizeof(struct ip6fl_iter_state));
  645. if (!err) {
  646. seq = file->private_data;
  647. state = ip6fl_seq_private(seq);
  648. rcu_read_lock();
  649. state->pid_ns = get_pid_ns(task_active_pid_ns(current));
  650. rcu_read_unlock();
  651. }
  652. return err;
  653. }
  654. static int ip6fl_seq_release(struct inode *inode, struct file *file)
  655. {
  656. struct seq_file *seq = file->private_data;
  657. struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
  658. put_pid_ns(state->pid_ns);
  659. return seq_release_net(inode, file);
  660. }
  661. static const struct file_operations ip6fl_seq_fops = {
  662. .owner = THIS_MODULE,
  663. .open = ip6fl_seq_open,
  664. .read = seq_read,
  665. .llseek = seq_lseek,
  666. .release = ip6fl_seq_release,
  667. };
  668. static int __net_init ip6_flowlabel_proc_init(struct net *net)
  669. {
  670. if (!proc_net_fops_create(net, "ip6_flowlabel",
  671. S_IRUGO, &ip6fl_seq_fops))
  672. return -ENOMEM;
  673. return 0;
  674. }
  675. static void __net_exit ip6_flowlabel_proc_fini(struct net *net)
  676. {
  677. proc_net_remove(net, "ip6_flowlabel");
  678. }
  679. #else
  680. static inline int ip6_flowlabel_proc_init(struct net *net)
  681. {
  682. return 0;
  683. }
  684. static inline void ip6_flowlabel_proc_fini(struct net *net)
  685. {
  686. }
  687. #endif
  688. static void __net_exit ip6_flowlabel_net_exit(struct net *net)
  689. {
  690. ip6_fl_purge(net);
  691. ip6_flowlabel_proc_fini(net);
  692. }
  693. static struct pernet_operations ip6_flowlabel_net_ops = {
  694. .init = ip6_flowlabel_proc_init,
  695. .exit = ip6_flowlabel_net_exit,
  696. };
  697. int ip6_flowlabel_init(void)
  698. {
  699. return register_pernet_subsys(&ip6_flowlabel_net_ops);
  700. }
  701. void ip6_flowlabel_cleanup(void)
  702. {
  703. del_timer(&ip6_fl_gc_timer);
  704. unregister_pernet_subsys(&ip6_flowlabel_net_ops);
  705. }