socket.c 74 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115
  1. /*
  2. * NET An implementation of the SOCKET network access protocol.
  3. *
  4. * Version: @(#)socket.c 1.1.93 18/02/95
  5. *
  6. * Authors: Orest Zborowski, <obz@Kodak.COM>
  7. * Ross Biro
  8. * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  9. *
  10. * Fixes:
  11. * Anonymous : NOTSOCK/BADF cleanup. Error fix in
  12. * shutdown()
  13. * Alan Cox : verify_area() fixes
  14. * Alan Cox : Removed DDI
  15. * Jonathan Kamens : SOCK_DGRAM reconnect bug
  16. * Alan Cox : Moved a load of checks to the very
  17. * top level.
  18. * Alan Cox : Move address structures to/from user
  19. * mode above the protocol layers.
  20. * Rob Janssen : Allow 0 length sends.
  21. * Alan Cox : Asynchronous I/O support (cribbed from the
  22. * tty drivers).
  23. * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
  24. * Jeff Uphoff : Made max number of sockets command-line
  25. * configurable.
  26. * Matti Aarnio : Made the number of sockets dynamic,
  27. * to be allocated when needed, and mr.
  28. * Uphoff's max is used as max to be
  29. * allowed to allocate.
  30. * Linus : Argh. removed all the socket allocation
  31. * altogether: it's in the inode now.
  32. * Alan Cox : Made sock_alloc()/sock_release() public
  33. * for NetROM and future kernel nfsd type
  34. * stuff.
  35. * Alan Cox : sendmsg/recvmsg basics.
  36. * Tom Dyas : Export net symbols.
  37. * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
  38. * Alan Cox : Added thread locking to sys_* calls
  39. * for sockets. May have errors at the
  40. * moment.
  41. * Kevin Buhr : Fixed the dumb errors in the above.
  42. * Andi Kleen : Some small cleanups, optimizations,
  43. * and fixed a copy_from_user() bug.
  44. * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
  45. * Tigran Aivazian : Made listen(2) backlog sanity checks
  46. * protocol-independent
  47. *
  48. *
  49. * This program is free software; you can redistribute it and/or
  50. * modify it under the terms of the GNU General Public License
  51. * as published by the Free Software Foundation; either version
  52. * 2 of the License, or (at your option) any later version.
  53. *
  54. *
  55. * This module is effectively the top level interface to the BSD socket
  56. * paradigm.
  57. *
  58. * Based upon Swansea University Computer Society NET3.039
  59. */
  60. #include <linux/mm.h>
  61. #include <linux/socket.h>
  62. #include <linux/file.h>
  63. #include <linux/net.h>
  64. #include <linux/interrupt.h>
  65. #include <linux/thread_info.h>
  66. #include <linux/rcupdate.h>
  67. #include <linux/netdevice.h>
  68. #include <linux/proc_fs.h>
  69. #include <linux/seq_file.h>
  70. #include <linux/mutex.h>
  71. #include <linux/wanrouter.h>
  72. #include <linux/if_bridge.h>
  73. #include <linux/if_frad.h>
  74. #include <linux/if_vlan.h>
  75. #include <linux/init.h>
  76. #include <linux/poll.h>
  77. #include <linux/cache.h>
  78. #include <linux/module.h>
  79. #include <linux/highmem.h>
  80. #include <linux/mount.h>
  81. #include <linux/security.h>
  82. #include <linux/syscalls.h>
  83. #include <linux/compat.h>
  84. #include <linux/kmod.h>
  85. #include <linux/audit.h>
  86. #include <linux/wireless.h>
  87. #include <linux/nsproxy.h>
  88. #include <linux/magic.h>
  89. #include <linux/slab.h>
  90. #include <asm/uaccess.h>
  91. #include <asm/unistd.h>
  92. #include <net/compat.h>
  93. #include <net/wext.h>
  94. #include <net/cls_cgroup.h>
  95. #include <net/sock.h>
  96. #include <linux/netfilter.h>
  97. #include <linux/if_tun.h>
  98. #include <linux/ipv6_route.h>
  99. #include <linux/route.h>
  100. #include <linux/sockios.h>
  101. #include <linux/atalk.h>
  102. static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
  103. static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
  104. unsigned long nr_segs, loff_t pos);
  105. static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
  106. unsigned long nr_segs, loff_t pos);
  107. static int sock_mmap(struct file *file, struct vm_area_struct *vma);
  108. static int sock_close(struct inode *inode, struct file *file);
  109. static unsigned int sock_poll(struct file *file,
  110. struct poll_table_struct *wait);
  111. static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
  112. #ifdef CONFIG_COMPAT
  113. static long compat_sock_ioctl(struct file *file,
  114. unsigned int cmd, unsigned long arg);
  115. #endif
  116. static int sock_fasync(int fd, struct file *filp, int on);
  117. static ssize_t sock_sendpage(struct file *file, struct page *page,
  118. int offset, size_t size, loff_t *ppos, int more);
  119. static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
  120. struct pipe_inode_info *pipe, size_t len,
  121. unsigned int flags);
  122. /*
  123. * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
  124. * in the operation structures but are done directly via the socketcall() multiplexor.
  125. */
  126. static const struct file_operations socket_file_ops = {
  127. .owner = THIS_MODULE,
  128. .llseek = no_llseek,
  129. .aio_read = sock_aio_read,
  130. .aio_write = sock_aio_write,
  131. .poll = sock_poll,
  132. .unlocked_ioctl = sock_ioctl,
  133. #ifdef CONFIG_COMPAT
  134. .compat_ioctl = compat_sock_ioctl,
  135. #endif
  136. .mmap = sock_mmap,
  137. .open = sock_no_open, /* special open code to disallow open via /proc */
  138. .release = sock_close,
  139. .fasync = sock_fasync,
  140. .sendpage = sock_sendpage,
  141. .splice_write = generic_splice_sendpage,
  142. .splice_read = sock_splice_read,
  143. };
  144. /*
  145. * The protocol list. Each protocol is registered in here.
  146. */
  147. static DEFINE_SPINLOCK(net_family_lock);
  148. static const struct net_proto_family *net_families[NPROTO] __read_mostly;
  149. /*
  150. * Statistics counters of the socket lists
  151. */
  152. static DEFINE_PER_CPU(int, sockets_in_use);
  153. /*
  154. * Support routines.
  155. * Move socket addresses back and forth across the kernel/user
  156. * divide and look after the messy bits.
  157. */
  158. /**
  159. * move_addr_to_kernel - copy a socket address into kernel space
  160. * @uaddr: Address in user space
  161. * @kaddr: Address in kernel space
  162. * @ulen: Length in user space
  163. *
  164. * The address is copied into kernel space. If the provided address is
  165. * too long an error code of -EINVAL is returned. If the copy gives
  166. * invalid addresses -EFAULT is returned. On a success 0 is returned.
  167. */
  168. int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
  169. {
  170. if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
  171. return -EINVAL;
  172. if (ulen == 0)
  173. return 0;
  174. if (copy_from_user(kaddr, uaddr, ulen))
  175. return -EFAULT;
  176. return audit_sockaddr(ulen, kaddr);
  177. }
  178. /**
  179. * move_addr_to_user - copy an address to user space
  180. * @kaddr: kernel space address
  181. * @klen: length of address in kernel
  182. * @uaddr: user space address
  183. * @ulen: pointer to user length field
  184. *
  185. * The value pointed to by ulen on entry is the buffer length available.
  186. * This is overwritten with the buffer space used. -EINVAL is returned
  187. * if an overlong buffer is specified or a negative buffer size. -EFAULT
  188. * is returned if either the buffer or the length field are not
  189. * accessible.
  190. * After copying the data up to the limit the user specifies, the true
  191. * length of the data is written over the length limit the user
  192. * specified. Zero is returned for a success.
  193. */
  194. int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
  195. int __user *ulen)
  196. {
  197. int err;
  198. int len;
  199. err = get_user(len, ulen);
  200. if (err)
  201. return err;
  202. if (len > klen)
  203. len = klen;
  204. if (len < 0 || len > sizeof(struct sockaddr_storage))
  205. return -EINVAL;
  206. if (len) {
  207. if (audit_sockaddr(klen, kaddr))
  208. return -ENOMEM;
  209. if (copy_to_user(uaddr, kaddr, len))
  210. return -EFAULT;
  211. }
  212. /*
  213. * "fromlen shall refer to the value before truncation.."
  214. * 1003.1g
  215. */
  216. return __put_user(klen, ulen);
  217. }
  218. static struct kmem_cache *sock_inode_cachep __read_mostly;
  219. static struct inode *sock_alloc_inode(struct super_block *sb)
  220. {
  221. struct socket_alloc *ei;
  222. ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
  223. if (!ei)
  224. return NULL;
  225. ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL);
  226. if (!ei->socket.wq) {
  227. kmem_cache_free(sock_inode_cachep, ei);
  228. return NULL;
  229. }
  230. init_waitqueue_head(&ei->socket.wq->wait);
  231. ei->socket.wq->fasync_list = NULL;
  232. ei->socket.state = SS_UNCONNECTED;
  233. ei->socket.flags = 0;
  234. ei->socket.ops = NULL;
  235. ei->socket.sk = NULL;
  236. ei->socket.file = NULL;
  237. return &ei->vfs_inode;
  238. }
  239. static void wq_free_rcu(struct rcu_head *head)
  240. {
  241. struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
  242. kfree(wq);
  243. }
  244. static void sock_destroy_inode(struct inode *inode)
  245. {
  246. struct socket_alloc *ei;
  247. ei = container_of(inode, struct socket_alloc, vfs_inode);
  248. call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
  249. kmem_cache_free(sock_inode_cachep, ei);
  250. }
  251. static void init_once(void *foo)
  252. {
  253. struct socket_alloc *ei = (struct socket_alloc *)foo;
  254. inode_init_once(&ei->vfs_inode);
  255. }
  256. static int init_inodecache(void)
  257. {
  258. sock_inode_cachep = kmem_cache_create("sock_inode_cache",
  259. sizeof(struct socket_alloc),
  260. 0,
  261. (SLAB_HWCACHE_ALIGN |
  262. SLAB_RECLAIM_ACCOUNT |
  263. SLAB_MEM_SPREAD),
  264. init_once);
  265. if (sock_inode_cachep == NULL)
  266. return -ENOMEM;
  267. return 0;
  268. }
  269. static const struct super_operations sockfs_ops = {
  270. .alloc_inode = sock_alloc_inode,
  271. .destroy_inode = sock_destroy_inode,
  272. .statfs = simple_statfs,
  273. };
  274. static int sockfs_get_sb(struct file_system_type *fs_type,
  275. int flags, const char *dev_name, void *data,
  276. struct vfsmount *mnt)
  277. {
  278. return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
  279. mnt);
  280. }
  281. static struct vfsmount *sock_mnt __read_mostly;
  282. static struct file_system_type sock_fs_type = {
  283. .name = "sockfs",
  284. .get_sb = sockfs_get_sb,
  285. .kill_sb = kill_anon_super,
  286. };
  287. /*
  288. * sockfs_dname() is called from d_path().
  289. */
  290. static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
  291. {
  292. return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
  293. dentry->d_inode->i_ino);
  294. }
  295. static const struct dentry_operations sockfs_dentry_operations = {
  296. .d_dname = sockfs_dname,
  297. };
  298. /*
  299. * Obtains the first available file descriptor and sets it up for use.
  300. *
  301. * These functions create file structures and maps them to fd space
  302. * of the current process. On success it returns file descriptor
  303. * and file struct implicitly stored in sock->file.
  304. * Note that another thread may close file descriptor before we return
  305. * from this function. We use the fact that now we do not refer
  306. * to socket after mapping. If one day we will need it, this
  307. * function will increment ref. count on file by 1.
  308. *
  309. * In any case returned fd MAY BE not valid!
  310. * This race condition is unavoidable
  311. * with shared fd spaces, we cannot solve it inside kernel,
  312. * but we take care of internal coherence yet.
  313. */
  314. static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
  315. {
  316. struct qstr name = { .name = "" };
  317. struct path path;
  318. struct file *file;
  319. int fd;
  320. fd = get_unused_fd_flags(flags);
  321. if (unlikely(fd < 0))
  322. return fd;
  323. path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
  324. if (unlikely(!path.dentry)) {
  325. put_unused_fd(fd);
  326. return -ENOMEM;
  327. }
  328. path.mnt = mntget(sock_mnt);
  329. path.dentry->d_op = &sockfs_dentry_operations;
  330. d_instantiate(path.dentry, SOCK_INODE(sock));
  331. SOCK_INODE(sock)->i_fop = &socket_file_ops;
  332. file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
  333. &socket_file_ops);
  334. if (unlikely(!file)) {
  335. /* drop dentry, keep inode */
  336. atomic_inc(&path.dentry->d_inode->i_count);
  337. path_put(&path);
  338. put_unused_fd(fd);
  339. return -ENFILE;
  340. }
  341. sock->file = file;
  342. file->f_flags = O_RDWR | (flags & O_NONBLOCK);
  343. file->f_pos = 0;
  344. file->private_data = sock;
  345. *f = file;
  346. return fd;
  347. }
  348. int sock_map_fd(struct socket *sock, int flags)
  349. {
  350. struct file *newfile;
  351. int fd = sock_alloc_file(sock, &newfile, flags);
  352. if (likely(fd >= 0))
  353. fd_install(fd, newfile);
  354. return fd;
  355. }
  356. EXPORT_SYMBOL(sock_map_fd);
  357. static struct socket *sock_from_file(struct file *file, int *err)
  358. {
  359. if (file->f_op == &socket_file_ops)
  360. return file->private_data; /* set in sock_map_fd */
  361. *err = -ENOTSOCK;
  362. return NULL;
  363. }
  364. /**
  365. * sockfd_lookup - Go from a file number to its socket slot
  366. * @fd: file handle
  367. * @err: pointer to an error code return
  368. *
  369. * The file handle passed in is locked and the socket it is bound
  370. * too is returned. If an error occurs the err pointer is overwritten
  371. * with a negative errno code and NULL is returned. The function checks
  372. * for both invalid handles and passing a handle which is not a socket.
  373. *
  374. * On a success the socket object pointer is returned.
  375. */
  376. struct socket *sockfd_lookup(int fd, int *err)
  377. {
  378. struct file *file;
  379. struct socket *sock;
  380. file = fget(fd);
  381. if (!file) {
  382. *err = -EBADF;
  383. return NULL;
  384. }
  385. sock = sock_from_file(file, err);
  386. if (!sock)
  387. fput(file);
  388. return sock;
  389. }
  390. EXPORT_SYMBOL(sockfd_lookup);
  391. static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
  392. {
  393. struct file *file;
  394. struct socket *sock;
  395. *err = -EBADF;
  396. file = fget_light(fd, fput_needed);
  397. if (file) {
  398. sock = sock_from_file(file, err);
  399. if (sock)
  400. return sock;
  401. fput_light(file, *fput_needed);
  402. }
  403. return NULL;
  404. }
  405. /**
  406. * sock_alloc - allocate a socket
  407. *
  408. * Allocate a new inode and socket object. The two are bound together
  409. * and initialised. The socket is then returned. If we are out of inodes
  410. * NULL is returned.
  411. */
  412. static struct socket *sock_alloc(void)
  413. {
  414. struct inode *inode;
  415. struct socket *sock;
  416. inode = new_inode(sock_mnt->mnt_sb);
  417. if (!inode)
  418. return NULL;
  419. sock = SOCKET_I(inode);
  420. kmemcheck_annotate_bitfield(sock, type);
  421. inode->i_mode = S_IFSOCK | S_IRWXUGO;
  422. inode->i_uid = current_fsuid();
  423. inode->i_gid = current_fsgid();
  424. percpu_add(sockets_in_use, 1);
  425. return sock;
  426. }
  427. /*
  428. * In theory you can't get an open on this inode, but /proc provides
  429. * a back door. Remember to keep it shut otherwise you'll let the
  430. * creepy crawlies in.
  431. */
  432. static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
  433. {
  434. return -ENXIO;
  435. }
  436. const struct file_operations bad_sock_fops = {
  437. .owner = THIS_MODULE,
  438. .open = sock_no_open,
  439. };
  440. /**
  441. * sock_release - close a socket
  442. * @sock: socket to close
  443. *
  444. * The socket is released from the protocol stack if it has a release
  445. * callback, and the inode is then released if the socket is bound to
  446. * an inode not a file.
  447. */
  448. void sock_release(struct socket *sock)
  449. {
  450. if (sock->ops) {
  451. struct module *owner = sock->ops->owner;
  452. sock->ops->release(sock);
  453. sock->ops = NULL;
  454. module_put(owner);
  455. }
  456. if (sock->wq->fasync_list)
  457. printk(KERN_ERR "sock_release: fasync list not empty!\n");
  458. percpu_sub(sockets_in_use, 1);
  459. if (!sock->file) {
  460. iput(SOCK_INODE(sock));
  461. return;
  462. }
  463. sock->file = NULL;
  464. }
  465. EXPORT_SYMBOL(sock_release);
  466. int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
  467. {
  468. *tx_flags = 0;
  469. if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
  470. *tx_flags |= SKBTX_HW_TSTAMP;
  471. if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
  472. *tx_flags |= SKBTX_SW_TSTAMP;
  473. return 0;
  474. }
  475. EXPORT_SYMBOL(sock_tx_timestamp);
  476. static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
  477. struct msghdr *msg, size_t size)
  478. {
  479. struct sock_iocb *si = kiocb_to_siocb(iocb);
  480. int err;
  481. sock_update_classid(sock->sk);
  482. si->sock = sock;
  483. si->scm = NULL;
  484. si->msg = msg;
  485. si->size = size;
  486. err = security_socket_sendmsg(sock, msg, size);
  487. if (err)
  488. return err;
  489. return sock->ops->sendmsg(iocb, sock, msg, size);
  490. }
  491. int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
  492. {
  493. struct kiocb iocb;
  494. struct sock_iocb siocb;
  495. int ret;
  496. init_sync_kiocb(&iocb, NULL);
  497. iocb.private = &siocb;
  498. ret = __sock_sendmsg(&iocb, sock, msg, size);
  499. if (-EIOCBQUEUED == ret)
  500. ret = wait_on_sync_kiocb(&iocb);
  501. return ret;
  502. }
  503. EXPORT_SYMBOL(sock_sendmsg);
  504. int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
  505. struct kvec *vec, size_t num, size_t size)
  506. {
  507. mm_segment_t oldfs = get_fs();
  508. int result;
  509. set_fs(KERNEL_DS);
  510. /*
  511. * the following is safe, since for compiler definitions of kvec and
  512. * iovec are identical, yielding the same in-core layout and alignment
  513. */
  514. msg->msg_iov = (struct iovec *)vec;
  515. msg->msg_iovlen = num;
  516. result = sock_sendmsg(sock, msg, size);
  517. set_fs(oldfs);
  518. return result;
  519. }
  520. EXPORT_SYMBOL(kernel_sendmsg);
  521. static int ktime2ts(ktime_t kt, struct timespec *ts)
  522. {
  523. if (kt.tv64) {
  524. *ts = ktime_to_timespec(kt);
  525. return 1;
  526. } else {
  527. return 0;
  528. }
  529. }
  530. /*
  531. * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
  532. */
  533. void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
  534. struct sk_buff *skb)
  535. {
  536. int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
  537. struct timespec ts[3];
  538. int empty = 1;
  539. struct skb_shared_hwtstamps *shhwtstamps =
  540. skb_hwtstamps(skb);
  541. /* Race occurred between timestamp enabling and packet
  542. receiving. Fill in the current time for now. */
  543. if (need_software_tstamp && skb->tstamp.tv64 == 0)
  544. __net_timestamp(skb);
  545. if (need_software_tstamp) {
  546. if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
  547. struct timeval tv;
  548. skb_get_timestamp(skb, &tv);
  549. put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
  550. sizeof(tv), &tv);
  551. } else {
  552. skb_get_timestampns(skb, &ts[0]);
  553. put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
  554. sizeof(ts[0]), &ts[0]);
  555. }
  556. }
  557. memset(ts, 0, sizeof(ts));
  558. if (skb->tstamp.tv64 &&
  559. sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
  560. skb_get_timestampns(skb, ts + 0);
  561. empty = 0;
  562. }
  563. if (shhwtstamps) {
  564. if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
  565. ktime2ts(shhwtstamps->syststamp, ts + 1))
  566. empty = 0;
  567. if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
  568. ktime2ts(shhwtstamps->hwtstamp, ts + 2))
  569. empty = 0;
  570. }
  571. if (!empty)
  572. put_cmsg(msg, SOL_SOCKET,
  573. SCM_TIMESTAMPING, sizeof(ts), &ts);
  574. }
  575. EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
  576. inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
  577. {
  578. if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
  579. put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
  580. sizeof(__u32), &skb->dropcount);
  581. }
  582. void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
  583. struct sk_buff *skb)
  584. {
  585. sock_recv_timestamp(msg, sk, skb);
  586. sock_recv_drops(msg, sk, skb);
  587. }
  588. EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
  589. static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
  590. struct msghdr *msg, size_t size, int flags)
  591. {
  592. struct sock_iocb *si = kiocb_to_siocb(iocb);
  593. sock_update_classid(sock->sk);
  594. si->sock = sock;
  595. si->scm = NULL;
  596. si->msg = msg;
  597. si->size = size;
  598. si->flags = flags;
  599. return sock->ops->recvmsg(iocb, sock, msg, size, flags);
  600. }
  601. static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
  602. struct msghdr *msg, size_t size, int flags)
  603. {
  604. int err = security_socket_recvmsg(sock, msg, size, flags);
  605. return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
  606. }
  607. int sock_recvmsg(struct socket *sock, struct msghdr *msg,
  608. size_t size, int flags)
  609. {
  610. struct kiocb iocb;
  611. struct sock_iocb siocb;
  612. int ret;
  613. init_sync_kiocb(&iocb, NULL);
  614. iocb.private = &siocb;
  615. ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
  616. if (-EIOCBQUEUED == ret)
  617. ret = wait_on_sync_kiocb(&iocb);
  618. return ret;
  619. }
  620. EXPORT_SYMBOL(sock_recvmsg);
  621. static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
  622. size_t size, int flags)
  623. {
  624. struct kiocb iocb;
  625. struct sock_iocb siocb;
  626. int ret;
  627. init_sync_kiocb(&iocb, NULL);
  628. iocb.private = &siocb;
  629. ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
  630. if (-EIOCBQUEUED == ret)
  631. ret = wait_on_sync_kiocb(&iocb);
  632. return ret;
  633. }
  634. int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
  635. struct kvec *vec, size_t num, size_t size, int flags)
  636. {
  637. mm_segment_t oldfs = get_fs();
  638. int result;
  639. set_fs(KERNEL_DS);
  640. /*
  641. * the following is safe, since for compiler definitions of kvec and
  642. * iovec are identical, yielding the same in-core layout and alignment
  643. */
  644. msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
  645. result = sock_recvmsg(sock, msg, size, flags);
  646. set_fs(oldfs);
  647. return result;
  648. }
  649. EXPORT_SYMBOL(kernel_recvmsg);
  650. static void sock_aio_dtor(struct kiocb *iocb)
  651. {
  652. kfree(iocb->private);
  653. }
  654. static ssize_t sock_sendpage(struct file *file, struct page *page,
  655. int offset, size_t size, loff_t *ppos, int more)
  656. {
  657. struct socket *sock;
  658. int flags;
  659. sock = file->private_data;
  660. flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
  661. if (more)
  662. flags |= MSG_MORE;
  663. return kernel_sendpage(sock, page, offset, size, flags);
  664. }
  665. static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
  666. struct pipe_inode_info *pipe, size_t len,
  667. unsigned int flags)
  668. {
  669. struct socket *sock = file->private_data;
  670. if (unlikely(!sock->ops->splice_read))
  671. return -EINVAL;
  672. sock_update_classid(sock->sk);
  673. return sock->ops->splice_read(sock, ppos, pipe, len, flags);
  674. }
  675. static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
  676. struct sock_iocb *siocb)
  677. {
  678. if (!is_sync_kiocb(iocb)) {
  679. siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
  680. if (!siocb)
  681. return NULL;
  682. iocb->ki_dtor = sock_aio_dtor;
  683. }
  684. siocb->kiocb = iocb;
  685. iocb->private = siocb;
  686. return siocb;
  687. }
  688. static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
  689. struct file *file, const struct iovec *iov,
  690. unsigned long nr_segs)
  691. {
  692. struct socket *sock = file->private_data;
  693. size_t size = 0;
  694. int i;
  695. for (i = 0; i < nr_segs; i++)
  696. size += iov[i].iov_len;
  697. msg->msg_name = NULL;
  698. msg->msg_namelen = 0;
  699. msg->msg_control = NULL;
  700. msg->msg_controllen = 0;
  701. msg->msg_iov = (struct iovec *)iov;
  702. msg->msg_iovlen = nr_segs;
  703. msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
  704. return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
  705. }
  706. static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
  707. unsigned long nr_segs, loff_t pos)
  708. {
  709. struct sock_iocb siocb, *x;
  710. if (pos != 0)
  711. return -ESPIPE;
  712. if (iocb->ki_left == 0) /* Match SYS5 behaviour */
  713. return 0;
  714. x = alloc_sock_iocb(iocb, &siocb);
  715. if (!x)
  716. return -ENOMEM;
  717. return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
  718. }
  719. static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
  720. struct file *file, const struct iovec *iov,
  721. unsigned long nr_segs)
  722. {
  723. struct socket *sock = file->private_data;
  724. size_t size = 0;
  725. int i;
  726. for (i = 0; i < nr_segs; i++)
  727. size += iov[i].iov_len;
  728. msg->msg_name = NULL;
  729. msg->msg_namelen = 0;
  730. msg->msg_control = NULL;
  731. msg->msg_controllen = 0;
  732. msg->msg_iov = (struct iovec *)iov;
  733. msg->msg_iovlen = nr_segs;
  734. msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
  735. if (sock->type == SOCK_SEQPACKET)
  736. msg->msg_flags |= MSG_EOR;
  737. return __sock_sendmsg(iocb, sock, msg, size);
  738. }
  739. static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
  740. unsigned long nr_segs, loff_t pos)
  741. {
  742. struct sock_iocb siocb, *x;
  743. if (pos != 0)
  744. return -ESPIPE;
  745. x = alloc_sock_iocb(iocb, &siocb);
  746. if (!x)
  747. return -ENOMEM;
  748. return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
  749. }
  750. /*
  751. * Atomic setting of ioctl hooks to avoid race
  752. * with module unload.
  753. */
  754. static DEFINE_MUTEX(br_ioctl_mutex);
  755. static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
  756. void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
  757. {
  758. mutex_lock(&br_ioctl_mutex);
  759. br_ioctl_hook = hook;
  760. mutex_unlock(&br_ioctl_mutex);
  761. }
  762. EXPORT_SYMBOL(brioctl_set);
  763. static DEFINE_MUTEX(vlan_ioctl_mutex);
  764. static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
  765. void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
  766. {
  767. mutex_lock(&vlan_ioctl_mutex);
  768. vlan_ioctl_hook = hook;
  769. mutex_unlock(&vlan_ioctl_mutex);
  770. }
  771. EXPORT_SYMBOL(vlan_ioctl_set);
  772. static DEFINE_MUTEX(dlci_ioctl_mutex);
  773. static int (*dlci_ioctl_hook) (unsigned int, void __user *);
  774. void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
  775. {
  776. mutex_lock(&dlci_ioctl_mutex);
  777. dlci_ioctl_hook = hook;
  778. mutex_unlock(&dlci_ioctl_mutex);
  779. }
  780. EXPORT_SYMBOL(dlci_ioctl_set);
  781. static long sock_do_ioctl(struct net *net, struct socket *sock,
  782. unsigned int cmd, unsigned long arg)
  783. {
  784. int err;
  785. void __user *argp = (void __user *)arg;
  786. err = sock->ops->ioctl(sock, cmd, arg);
  787. /*
  788. * If this ioctl is unknown try to hand it down
  789. * to the NIC driver.
  790. */
  791. if (err == -ENOIOCTLCMD)
  792. err = dev_ioctl(net, cmd, argp);
  793. return err;
  794. }
  795. /*
  796. * With an ioctl, arg may well be a user mode pointer, but we don't know
  797. * what to do with it - that's up to the protocol still.
  798. */
  799. static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
  800. {
  801. struct socket *sock;
  802. struct sock *sk;
  803. void __user *argp = (void __user *)arg;
  804. int pid, err;
  805. struct net *net;
  806. sock = file->private_data;
  807. sk = sock->sk;
  808. net = sock_net(sk);
  809. if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
  810. err = dev_ioctl(net, cmd, argp);
  811. } else
  812. #ifdef CONFIG_WEXT_CORE
  813. if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
  814. err = dev_ioctl(net, cmd, argp);
  815. } else
  816. #endif
  817. switch (cmd) {
  818. case FIOSETOWN:
  819. case SIOCSPGRP:
  820. err = -EFAULT;
  821. if (get_user(pid, (int __user *)argp))
  822. break;
  823. err = f_setown(sock->file, pid, 1);
  824. break;
  825. case FIOGETOWN:
  826. case SIOCGPGRP:
  827. err = put_user(f_getown(sock->file),
  828. (int __user *)argp);
  829. break;
  830. case SIOCGIFBR:
  831. case SIOCSIFBR:
  832. case SIOCBRADDBR:
  833. case SIOCBRDELBR:
  834. err = -ENOPKG;
  835. if (!br_ioctl_hook)
  836. request_module("bridge");
  837. mutex_lock(&br_ioctl_mutex);
  838. if (br_ioctl_hook)
  839. err = br_ioctl_hook(net, cmd, argp);
  840. mutex_unlock(&br_ioctl_mutex);
  841. break;
  842. case SIOCGIFVLAN:
  843. case SIOCSIFVLAN:
  844. err = -ENOPKG;
  845. if (!vlan_ioctl_hook)
  846. request_module("8021q");
  847. mutex_lock(&vlan_ioctl_mutex);
  848. if (vlan_ioctl_hook)
  849. err = vlan_ioctl_hook(net, argp);
  850. mutex_unlock(&vlan_ioctl_mutex);
  851. break;
  852. case SIOCADDDLCI:
  853. case SIOCDELDLCI:
  854. err = -ENOPKG;
  855. if (!dlci_ioctl_hook)
  856. request_module("dlci");
  857. mutex_lock(&dlci_ioctl_mutex);
  858. if (dlci_ioctl_hook)
  859. err = dlci_ioctl_hook(cmd, argp);
  860. mutex_unlock(&dlci_ioctl_mutex);
  861. break;
  862. default:
  863. err = sock_do_ioctl(net, sock, cmd, arg);
  864. break;
  865. }
  866. return err;
  867. }
  868. int sock_create_lite(int family, int type, int protocol, struct socket **res)
  869. {
  870. int err;
  871. struct socket *sock = NULL;
  872. err = security_socket_create(family, type, protocol, 1);
  873. if (err)
  874. goto out;
  875. sock = sock_alloc();
  876. if (!sock) {
  877. err = -ENOMEM;
  878. goto out;
  879. }
  880. sock->type = type;
  881. err = security_socket_post_create(sock, family, type, protocol, 1);
  882. if (err)
  883. goto out_release;
  884. out:
  885. *res = sock;
  886. return err;
  887. out_release:
  888. sock_release(sock);
  889. sock = NULL;
  890. goto out;
  891. }
  892. EXPORT_SYMBOL(sock_create_lite);
  893. /* No kernel lock held - perfect */
  894. static unsigned int sock_poll(struct file *file, poll_table *wait)
  895. {
  896. struct socket *sock;
  897. /*
  898. * We can't return errors to poll, so it's either yes or no.
  899. */
  900. sock = file->private_data;
  901. return sock->ops->poll(file, sock, wait);
  902. }
  903. static int sock_mmap(struct file *file, struct vm_area_struct *vma)
  904. {
  905. struct socket *sock = file->private_data;
  906. return sock->ops->mmap(file, sock, vma);
  907. }
  908. static int sock_close(struct inode *inode, struct file *filp)
  909. {
  910. /*
  911. * It was possible the inode is NULL we were
  912. * closing an unfinished socket.
  913. */
  914. if (!inode) {
  915. printk(KERN_DEBUG "sock_close: NULL inode\n");
  916. return 0;
  917. }
  918. sock_release(SOCKET_I(inode));
  919. return 0;
  920. }
  921. /*
  922. * Update the socket async list
  923. *
  924. * Fasync_list locking strategy.
  925. *
  926. * 1. fasync_list is modified only under process context socket lock
  927. * i.e. under semaphore.
  928. * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
  929. * or under socket lock
  930. */
  931. static int sock_fasync(int fd, struct file *filp, int on)
  932. {
  933. struct socket *sock = filp->private_data;
  934. struct sock *sk = sock->sk;
  935. if (sk == NULL)
  936. return -EINVAL;
  937. lock_sock(sk);
  938. fasync_helper(fd, filp, on, &sock->wq->fasync_list);
  939. if (!sock->wq->fasync_list)
  940. sock_reset_flag(sk, SOCK_FASYNC);
  941. else
  942. sock_set_flag(sk, SOCK_FASYNC);
  943. release_sock(sk);
  944. return 0;
  945. }
  946. /* This function may be called only under socket lock or callback_lock or rcu_lock */
  947. int sock_wake_async(struct socket *sock, int how, int band)
  948. {
  949. struct socket_wq *wq;
  950. if (!sock)
  951. return -1;
  952. rcu_read_lock();
  953. wq = rcu_dereference(sock->wq);
  954. if (!wq || !wq->fasync_list) {
  955. rcu_read_unlock();
  956. return -1;
  957. }
  958. switch (how) {
  959. case SOCK_WAKE_WAITD:
  960. if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
  961. break;
  962. goto call_kill;
  963. case SOCK_WAKE_SPACE:
  964. if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
  965. break;
  966. /* fall through */
  967. case SOCK_WAKE_IO:
  968. call_kill:
  969. kill_fasync(&wq->fasync_list, SIGIO, band);
  970. break;
  971. case SOCK_WAKE_URG:
  972. kill_fasync(&wq->fasync_list, SIGURG, band);
  973. }
  974. rcu_read_unlock();
  975. return 0;
  976. }
  977. EXPORT_SYMBOL(sock_wake_async);
  978. static int __sock_create(struct net *net, int family, int type, int protocol,
  979. struct socket **res, int kern)
  980. {
  981. int err;
  982. struct socket *sock;
  983. const struct net_proto_family *pf;
  984. /*
  985. * Check protocol is in range
  986. */
  987. if (family < 0 || family >= NPROTO)
  988. return -EAFNOSUPPORT;
  989. if (type < 0 || type >= SOCK_MAX)
  990. return -EINVAL;
  991. /* Compatibility.
  992. This uglymoron is moved from INET layer to here to avoid
  993. deadlock in module load.
  994. */
  995. if (family == PF_INET && type == SOCK_PACKET) {
  996. static int warned;
  997. if (!warned) {
  998. warned = 1;
  999. printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
  1000. current->comm);
  1001. }
  1002. family = PF_PACKET;
  1003. }
  1004. err = security_socket_create(family, type, protocol, kern);
  1005. if (err)
  1006. return err;
  1007. /*
  1008. * Allocate the socket and allow the family to set things up. if
  1009. * the protocol is 0, the family is instructed to select an appropriate
  1010. * default.
  1011. */
  1012. sock = sock_alloc();
  1013. if (!sock) {
  1014. if (net_ratelimit())
  1015. printk(KERN_WARNING "socket: no more sockets\n");
  1016. return -ENFILE; /* Not exactly a match, but its the
  1017. closest posix thing */
  1018. }
  1019. sock->type = type;
  1020. #ifdef CONFIG_MODULES
  1021. /* Attempt to load a protocol module if the find failed.
  1022. *
  1023. * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
  1024. * requested real, full-featured networking support upon configuration.
  1025. * Otherwise module support will break!
  1026. */
  1027. if (net_families[family] == NULL)
  1028. request_module("net-pf-%d", family);
  1029. #endif
  1030. rcu_read_lock();
  1031. pf = rcu_dereference(net_families[family]);
  1032. err = -EAFNOSUPPORT;
  1033. if (!pf)
  1034. goto out_release;
  1035. /*
  1036. * We will call the ->create function, that possibly is in a loadable
  1037. * module, so we have to bump that loadable module refcnt first.
  1038. */
  1039. if (!try_module_get(pf->owner))
  1040. goto out_release;
  1041. /* Now protected by module ref count */
  1042. rcu_read_unlock();
  1043. err = pf->create(net, sock, protocol, kern);
  1044. if (err < 0)
  1045. goto out_module_put;
  1046. /*
  1047. * Now to bump the refcnt of the [loadable] module that owns this
  1048. * socket at sock_release time we decrement its refcnt.
  1049. */
  1050. if (!try_module_get(sock->ops->owner))
  1051. goto out_module_busy;
  1052. /*
  1053. * Now that we're done with the ->create function, the [loadable]
  1054. * module can have its refcnt decremented
  1055. */
  1056. module_put(pf->owner);
  1057. err = security_socket_post_create(sock, family, type, protocol, kern);
  1058. if (err)
  1059. goto out_sock_release;
  1060. *res = sock;
  1061. return 0;
  1062. out_module_busy:
  1063. err = -EAFNOSUPPORT;
  1064. out_module_put:
  1065. sock->ops = NULL;
  1066. module_put(pf->owner);
  1067. out_sock_release:
  1068. sock_release(sock);
  1069. return err;
  1070. out_release:
  1071. rcu_read_unlock();
  1072. goto out_sock_release;
  1073. }
  1074. int sock_create(int family, int type, int protocol, struct socket **res)
  1075. {
  1076. return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
  1077. }
  1078. EXPORT_SYMBOL(sock_create);
  1079. int sock_create_kern(int family, int type, int protocol, struct socket **res)
  1080. {
  1081. return __sock_create(&init_net, family, type, protocol, res, 1);
  1082. }
  1083. EXPORT_SYMBOL(sock_create_kern);
  1084. SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
  1085. {
  1086. int retval;
  1087. struct socket *sock;
  1088. int flags;
  1089. /* Check the SOCK_* constants for consistency. */
  1090. BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
  1091. BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
  1092. BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
  1093. BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
  1094. flags = type & ~SOCK_TYPE_MASK;
  1095. if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
  1096. return -EINVAL;
  1097. type &= SOCK_TYPE_MASK;
  1098. if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
  1099. flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
  1100. retval = sock_create(family, type, protocol, &sock);
  1101. if (retval < 0)
  1102. goto out;
  1103. retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
  1104. if (retval < 0)
  1105. goto out_release;
  1106. out:
  1107. /* It may be already another descriptor 8) Not kernel problem. */
  1108. return retval;
  1109. out_release:
  1110. sock_release(sock);
  1111. return retval;
  1112. }
  1113. /*
  1114. * Create a pair of connected sockets.
  1115. */
  1116. SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
  1117. int __user *, usockvec)
  1118. {
  1119. struct socket *sock1, *sock2;
  1120. int fd1, fd2, err;
  1121. struct file *newfile1, *newfile2;
  1122. int flags;
  1123. flags = type & ~SOCK_TYPE_MASK;
  1124. if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
  1125. return -EINVAL;
  1126. type &= SOCK_TYPE_MASK;
  1127. if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
  1128. flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
  1129. /*
  1130. * Obtain the first socket and check if the underlying protocol
  1131. * supports the socketpair call.
  1132. */
  1133. err = sock_create(family, type, protocol, &sock1);
  1134. if (err < 0)
  1135. goto out;
  1136. err = sock_create(family, type, protocol, &sock2);
  1137. if (err < 0)
  1138. goto out_release_1;
  1139. err = sock1->ops->socketpair(sock1, sock2);
  1140. if (err < 0)
  1141. goto out_release_both;
  1142. fd1 = sock_alloc_file(sock1, &newfile1, flags);
  1143. if (unlikely(fd1 < 0)) {
  1144. err = fd1;
  1145. goto out_release_both;
  1146. }
  1147. fd2 = sock_alloc_file(sock2, &newfile2, flags);
  1148. if (unlikely(fd2 < 0)) {
  1149. err = fd2;
  1150. fput(newfile1);
  1151. put_unused_fd(fd1);
  1152. sock_release(sock2);
  1153. goto out;
  1154. }
  1155. audit_fd_pair(fd1, fd2);
  1156. fd_install(fd1, newfile1);
  1157. fd_install(fd2, newfile2);
  1158. /* fd1 and fd2 may be already another descriptors.
  1159. * Not kernel problem.
  1160. */
  1161. err = put_user(fd1, &usockvec[0]);
  1162. if (!err)
  1163. err = put_user(fd2, &usockvec[1]);
  1164. if (!err)
  1165. return 0;
  1166. sys_close(fd2);
  1167. sys_close(fd1);
  1168. return err;
  1169. out_release_both:
  1170. sock_release(sock2);
  1171. out_release_1:
  1172. sock_release(sock1);
  1173. out:
  1174. return err;
  1175. }
  1176. /*
  1177. * Bind a name to a socket. Nothing much to do here since it's
  1178. * the protocol's responsibility to handle the local address.
  1179. *
  1180. * We move the socket address to kernel space before we call
  1181. * the protocol layer (having also checked the address is ok).
  1182. */
  1183. SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
  1184. {
  1185. struct socket *sock;
  1186. struct sockaddr_storage address;
  1187. int err, fput_needed;
  1188. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1189. if (sock) {
  1190. err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
  1191. if (err >= 0) {
  1192. err = security_socket_bind(sock,
  1193. (struct sockaddr *)&address,
  1194. addrlen);
  1195. if (!err)
  1196. err = sock->ops->bind(sock,
  1197. (struct sockaddr *)
  1198. &address, addrlen);
  1199. }
  1200. fput_light(sock->file, fput_needed);
  1201. }
  1202. return err;
  1203. }
  1204. /*
  1205. * Perform a listen. Basically, we allow the protocol to do anything
  1206. * necessary for a listen, and if that works, we mark the socket as
  1207. * ready for listening.
  1208. */
  1209. SYSCALL_DEFINE2(listen, int, fd, int, backlog)
  1210. {
  1211. struct socket *sock;
  1212. int err, fput_needed;
  1213. int somaxconn;
  1214. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1215. if (sock) {
  1216. somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
  1217. if ((unsigned)backlog > somaxconn)
  1218. backlog = somaxconn;
  1219. err = security_socket_listen(sock, backlog);
  1220. if (!err)
  1221. err = sock->ops->listen(sock, backlog);
  1222. fput_light(sock->file, fput_needed);
  1223. }
  1224. return err;
  1225. }
  1226. /*
  1227. * For accept, we attempt to create a new socket, set up the link
  1228. * with the client, wake up the client, then return the new
  1229. * connected fd. We collect the address of the connector in kernel
  1230. * space and move it to user at the very end. This is unclean because
  1231. * we open the socket then return an error.
  1232. *
  1233. * 1003.1g adds the ability to recvmsg() to query connection pending
  1234. * status to recvmsg. We need to add that support in a way thats
  1235. * clean when we restucture accept also.
  1236. */
  1237. SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
  1238. int __user *, upeer_addrlen, int, flags)
  1239. {
  1240. struct socket *sock, *newsock;
  1241. struct file *newfile;
  1242. int err, len, newfd, fput_needed;
  1243. struct sockaddr_storage address;
  1244. if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
  1245. return -EINVAL;
  1246. if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
  1247. flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
  1248. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1249. if (!sock)
  1250. goto out;
  1251. err = -ENFILE;
  1252. newsock = sock_alloc();
  1253. if (!newsock)
  1254. goto out_put;
  1255. newsock->type = sock->type;
  1256. newsock->ops = sock->ops;
  1257. /*
  1258. * We don't need try_module_get here, as the listening socket (sock)
  1259. * has the protocol module (sock->ops->owner) held.
  1260. */
  1261. __module_get(newsock->ops->owner);
  1262. newfd = sock_alloc_file(newsock, &newfile, flags);
  1263. if (unlikely(newfd < 0)) {
  1264. err = newfd;
  1265. sock_release(newsock);
  1266. goto out_put;
  1267. }
  1268. err = security_socket_accept(sock, newsock);
  1269. if (err)
  1270. goto out_fd;
  1271. err = sock->ops->accept(sock, newsock, sock->file->f_flags);
  1272. if (err < 0)
  1273. goto out_fd;
  1274. if (upeer_sockaddr) {
  1275. if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
  1276. &len, 2) < 0) {
  1277. err = -ECONNABORTED;
  1278. goto out_fd;
  1279. }
  1280. err = move_addr_to_user((struct sockaddr *)&address,
  1281. len, upeer_sockaddr, upeer_addrlen);
  1282. if (err < 0)
  1283. goto out_fd;
  1284. }
  1285. /* File flags are not inherited via accept() unlike another OSes. */
  1286. fd_install(newfd, newfile);
  1287. err = newfd;
  1288. out_put:
  1289. fput_light(sock->file, fput_needed);
  1290. out:
  1291. return err;
  1292. out_fd:
  1293. fput(newfile);
  1294. put_unused_fd(newfd);
  1295. goto out_put;
  1296. }
  1297. SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
  1298. int __user *, upeer_addrlen)
  1299. {
  1300. return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
  1301. }
  1302. /*
  1303. * Attempt to connect to a socket with the server address. The address
  1304. * is in user space so we verify it is OK and move it to kernel space.
  1305. *
  1306. * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
  1307. * break bindings
  1308. *
  1309. * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
  1310. * other SEQPACKET protocols that take time to connect() as it doesn't
  1311. * include the -EINPROGRESS status for such sockets.
  1312. */
  1313. SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
  1314. int, addrlen)
  1315. {
  1316. struct socket *sock;
  1317. struct sockaddr_storage address;
  1318. int err, fput_needed;
  1319. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1320. if (!sock)
  1321. goto out;
  1322. err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
  1323. if (err < 0)
  1324. goto out_put;
  1325. err =
  1326. security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
  1327. if (err)
  1328. goto out_put;
  1329. err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
  1330. sock->file->f_flags);
  1331. out_put:
  1332. fput_light(sock->file, fput_needed);
  1333. out:
  1334. return err;
  1335. }
  1336. /*
  1337. * Get the local address ('name') of a socket object. Move the obtained
  1338. * name to user space.
  1339. */
  1340. SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
  1341. int __user *, usockaddr_len)
  1342. {
  1343. struct socket *sock;
  1344. struct sockaddr_storage address;
  1345. int len, err, fput_needed;
  1346. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1347. if (!sock)
  1348. goto out;
  1349. err = security_socket_getsockname(sock);
  1350. if (err)
  1351. goto out_put;
  1352. err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
  1353. if (err)
  1354. goto out_put;
  1355. err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
  1356. out_put:
  1357. fput_light(sock->file, fput_needed);
  1358. out:
  1359. return err;
  1360. }
  1361. /*
  1362. * Get the remote address ('name') of a socket object. Move the obtained
  1363. * name to user space.
  1364. */
  1365. SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
  1366. int __user *, usockaddr_len)
  1367. {
  1368. struct socket *sock;
  1369. struct sockaddr_storage address;
  1370. int len, err, fput_needed;
  1371. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1372. if (sock != NULL) {
  1373. err = security_socket_getpeername(sock);
  1374. if (err) {
  1375. fput_light(sock->file, fput_needed);
  1376. return err;
  1377. }
  1378. err =
  1379. sock->ops->getname(sock, (struct sockaddr *)&address, &len,
  1380. 1);
  1381. if (!err)
  1382. err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
  1383. usockaddr_len);
  1384. fput_light(sock->file, fput_needed);
  1385. }
  1386. return err;
  1387. }
  1388. /*
  1389. * Send a datagram to a given address. We move the address into kernel
  1390. * space and check the user space data area is readable before invoking
  1391. * the protocol.
  1392. */
  1393. SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
  1394. unsigned, flags, struct sockaddr __user *, addr,
  1395. int, addr_len)
  1396. {
  1397. struct socket *sock;
  1398. struct sockaddr_storage address;
  1399. int err;
  1400. struct msghdr msg;
  1401. struct iovec iov;
  1402. int fput_needed;
  1403. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1404. if (!sock)
  1405. goto out;
  1406. iov.iov_base = buff;
  1407. iov.iov_len = len;
  1408. msg.msg_name = NULL;
  1409. msg.msg_iov = &iov;
  1410. msg.msg_iovlen = 1;
  1411. msg.msg_control = NULL;
  1412. msg.msg_controllen = 0;
  1413. msg.msg_namelen = 0;
  1414. if (addr) {
  1415. err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
  1416. if (err < 0)
  1417. goto out_put;
  1418. msg.msg_name = (struct sockaddr *)&address;
  1419. msg.msg_namelen = addr_len;
  1420. }
  1421. if (sock->file->f_flags & O_NONBLOCK)
  1422. flags |= MSG_DONTWAIT;
  1423. msg.msg_flags = flags;
  1424. err = sock_sendmsg(sock, &msg, len);
  1425. out_put:
  1426. fput_light(sock->file, fput_needed);
  1427. out:
  1428. return err;
  1429. }
  1430. /*
  1431. * Send a datagram down a socket.
  1432. */
  1433. SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
  1434. unsigned, flags)
  1435. {
  1436. return sys_sendto(fd, buff, len, flags, NULL, 0);
  1437. }
  1438. /*
  1439. * Receive a frame from the socket and optionally record the address of the
  1440. * sender. We verify the buffers are writable and if needed move the
  1441. * sender address from kernel to user space.
  1442. */
  1443. SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
  1444. unsigned, flags, struct sockaddr __user *, addr,
  1445. int __user *, addr_len)
  1446. {
  1447. struct socket *sock;
  1448. struct iovec iov;
  1449. struct msghdr msg;
  1450. struct sockaddr_storage address;
  1451. int err, err2;
  1452. int fput_needed;
  1453. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1454. if (!sock)
  1455. goto out;
  1456. msg.msg_control = NULL;
  1457. msg.msg_controllen = 0;
  1458. msg.msg_iovlen = 1;
  1459. msg.msg_iov = &iov;
  1460. iov.iov_len = size;
  1461. iov.iov_base = ubuf;
  1462. msg.msg_name = (struct sockaddr *)&address;
  1463. msg.msg_namelen = sizeof(address);
  1464. if (sock->file->f_flags & O_NONBLOCK)
  1465. flags |= MSG_DONTWAIT;
  1466. err = sock_recvmsg(sock, &msg, size, flags);
  1467. if (err >= 0 && addr != NULL) {
  1468. err2 = move_addr_to_user((struct sockaddr *)&address,
  1469. msg.msg_namelen, addr, addr_len);
  1470. if (err2 < 0)
  1471. err = err2;
  1472. }
  1473. fput_light(sock->file, fput_needed);
  1474. out:
  1475. return err;
  1476. }
  1477. /*
  1478. * Receive a datagram from a socket.
  1479. */
  1480. asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
  1481. unsigned flags)
  1482. {
  1483. return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
  1484. }
  1485. /*
  1486. * Set a socket option. Because we don't know the option lengths we have
  1487. * to pass the user mode parameter for the protocols to sort out.
  1488. */
  1489. SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
  1490. char __user *, optval, int, optlen)
  1491. {
  1492. int err, fput_needed;
  1493. struct socket *sock;
  1494. if (optlen < 0)
  1495. return -EINVAL;
  1496. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1497. if (sock != NULL) {
  1498. err = security_socket_setsockopt(sock, level, optname);
  1499. if (err)
  1500. goto out_put;
  1501. if (level == SOL_SOCKET)
  1502. err =
  1503. sock_setsockopt(sock, level, optname, optval,
  1504. optlen);
  1505. else
  1506. err =
  1507. sock->ops->setsockopt(sock, level, optname, optval,
  1508. optlen);
  1509. out_put:
  1510. fput_light(sock->file, fput_needed);
  1511. }
  1512. return err;
  1513. }
  1514. /*
  1515. * Get a socket option. Because we don't know the option lengths we have
  1516. * to pass a user mode parameter for the protocols to sort out.
  1517. */
  1518. SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
  1519. char __user *, optval, int __user *, optlen)
  1520. {
  1521. int err, fput_needed;
  1522. struct socket *sock;
  1523. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1524. if (sock != NULL) {
  1525. err = security_socket_getsockopt(sock, level, optname);
  1526. if (err)
  1527. goto out_put;
  1528. if (level == SOL_SOCKET)
  1529. err =
  1530. sock_getsockopt(sock, level, optname, optval,
  1531. optlen);
  1532. else
  1533. err =
  1534. sock->ops->getsockopt(sock, level, optname, optval,
  1535. optlen);
  1536. out_put:
  1537. fput_light(sock->file, fput_needed);
  1538. }
  1539. return err;
  1540. }
  1541. /*
  1542. * Shutdown a socket.
  1543. */
  1544. SYSCALL_DEFINE2(shutdown, int, fd, int, how)
  1545. {
  1546. int err, fput_needed;
  1547. struct socket *sock;
  1548. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1549. if (sock != NULL) {
  1550. err = security_socket_shutdown(sock, how);
  1551. if (!err)
  1552. err = sock->ops->shutdown(sock, how);
  1553. fput_light(sock->file, fput_needed);
  1554. }
  1555. return err;
  1556. }
  1557. /* A couple of helpful macros for getting the address of the 32/64 bit
  1558. * fields which are the same type (int / unsigned) on our platforms.
  1559. */
  1560. #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
  1561. #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
  1562. #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
  1563. /*
  1564. * BSD sendmsg interface
  1565. */
  1566. SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
  1567. {
  1568. struct compat_msghdr __user *msg_compat =
  1569. (struct compat_msghdr __user *)msg;
  1570. struct socket *sock;
  1571. struct sockaddr_storage address;
  1572. struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
  1573. unsigned char ctl[sizeof(struct cmsghdr) + 20]
  1574. __attribute__ ((aligned(sizeof(__kernel_size_t))));
  1575. /* 20 is size of ipv6_pktinfo */
  1576. unsigned char *ctl_buf = ctl;
  1577. struct msghdr msg_sys;
  1578. int err, ctl_len, iov_size, total_len;
  1579. int fput_needed;
  1580. err = -EFAULT;
  1581. if (MSG_CMSG_COMPAT & flags) {
  1582. if (get_compat_msghdr(&msg_sys, msg_compat))
  1583. return -EFAULT;
  1584. } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
  1585. return -EFAULT;
  1586. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1587. if (!sock)
  1588. goto out;
  1589. /* do not move before msg_sys is valid */
  1590. err = -EMSGSIZE;
  1591. if (msg_sys.msg_iovlen > UIO_MAXIOV)
  1592. goto out_put;
  1593. /* Check whether to allocate the iovec area */
  1594. err = -ENOMEM;
  1595. iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
  1596. if (msg_sys.msg_iovlen > UIO_FASTIOV) {
  1597. iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
  1598. if (!iov)
  1599. goto out_put;
  1600. }
  1601. /* This will also move the address data into kernel space */
  1602. if (MSG_CMSG_COMPAT & flags) {
  1603. err = verify_compat_iovec(&msg_sys, iov,
  1604. (struct sockaddr *)&address,
  1605. VERIFY_READ);
  1606. } else
  1607. err = verify_iovec(&msg_sys, iov,
  1608. (struct sockaddr *)&address,
  1609. VERIFY_READ);
  1610. if (err < 0)
  1611. goto out_freeiov;
  1612. total_len = err;
  1613. err = -ENOBUFS;
  1614. if (msg_sys.msg_controllen > INT_MAX)
  1615. goto out_freeiov;
  1616. ctl_len = msg_sys.msg_controllen;
  1617. if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
  1618. err =
  1619. cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
  1620. sizeof(ctl));
  1621. if (err)
  1622. goto out_freeiov;
  1623. ctl_buf = msg_sys.msg_control;
  1624. ctl_len = msg_sys.msg_controllen;
  1625. } else if (ctl_len) {
  1626. if (ctl_len > sizeof(ctl)) {
  1627. ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
  1628. if (ctl_buf == NULL)
  1629. goto out_freeiov;
  1630. }
  1631. err = -EFAULT;
  1632. /*
  1633. * Careful! Before this, msg_sys.msg_control contains a user pointer.
  1634. * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
  1635. * checking falls down on this.
  1636. */
  1637. if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
  1638. ctl_len))
  1639. goto out_freectl;
  1640. msg_sys.msg_control = ctl_buf;
  1641. }
  1642. msg_sys.msg_flags = flags;
  1643. if (sock->file->f_flags & O_NONBLOCK)
  1644. msg_sys.msg_flags |= MSG_DONTWAIT;
  1645. err = sock_sendmsg(sock, &msg_sys, total_len);
  1646. out_freectl:
  1647. if (ctl_buf != ctl)
  1648. sock_kfree_s(sock->sk, ctl_buf, ctl_len);
  1649. out_freeiov:
  1650. if (iov != iovstack)
  1651. sock_kfree_s(sock->sk, iov, iov_size);
  1652. out_put:
  1653. fput_light(sock->file, fput_needed);
  1654. out:
  1655. return err;
  1656. }
  1657. static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
  1658. struct msghdr *msg_sys, unsigned flags, int nosec)
  1659. {
  1660. struct compat_msghdr __user *msg_compat =
  1661. (struct compat_msghdr __user *)msg;
  1662. struct iovec iovstack[UIO_FASTIOV];
  1663. struct iovec *iov = iovstack;
  1664. unsigned long cmsg_ptr;
  1665. int err, iov_size, total_len, len;
  1666. /* kernel mode address */
  1667. struct sockaddr_storage addr;
  1668. /* user mode address pointers */
  1669. struct sockaddr __user *uaddr;
  1670. int __user *uaddr_len;
  1671. if (MSG_CMSG_COMPAT & flags) {
  1672. if (get_compat_msghdr(msg_sys, msg_compat))
  1673. return -EFAULT;
  1674. } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
  1675. return -EFAULT;
  1676. err = -EMSGSIZE;
  1677. if (msg_sys->msg_iovlen > UIO_MAXIOV)
  1678. goto out;
  1679. /* Check whether to allocate the iovec area */
  1680. err = -ENOMEM;
  1681. iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
  1682. if (msg_sys->msg_iovlen > UIO_FASTIOV) {
  1683. iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
  1684. if (!iov)
  1685. goto out;
  1686. }
  1687. /*
  1688. * Save the user-mode address (verify_iovec will change the
  1689. * kernel msghdr to use the kernel address space)
  1690. */
  1691. uaddr = (__force void __user *)msg_sys->msg_name;
  1692. uaddr_len = COMPAT_NAMELEN(msg);
  1693. if (MSG_CMSG_COMPAT & flags) {
  1694. err = verify_compat_iovec(msg_sys, iov,
  1695. (struct sockaddr *)&addr,
  1696. VERIFY_WRITE);
  1697. } else
  1698. err = verify_iovec(msg_sys, iov,
  1699. (struct sockaddr *)&addr,
  1700. VERIFY_WRITE);
  1701. if (err < 0)
  1702. goto out_freeiov;
  1703. total_len = err;
  1704. cmsg_ptr = (unsigned long)msg_sys->msg_control;
  1705. msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
  1706. if (sock->file->f_flags & O_NONBLOCK)
  1707. flags |= MSG_DONTWAIT;
  1708. err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
  1709. total_len, flags);
  1710. if (err < 0)
  1711. goto out_freeiov;
  1712. len = err;
  1713. if (uaddr != NULL) {
  1714. err = move_addr_to_user((struct sockaddr *)&addr,
  1715. msg_sys->msg_namelen, uaddr,
  1716. uaddr_len);
  1717. if (err < 0)
  1718. goto out_freeiov;
  1719. }
  1720. err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
  1721. COMPAT_FLAGS(msg));
  1722. if (err)
  1723. goto out_freeiov;
  1724. if (MSG_CMSG_COMPAT & flags)
  1725. err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
  1726. &msg_compat->msg_controllen);
  1727. else
  1728. err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
  1729. &msg->msg_controllen);
  1730. if (err)
  1731. goto out_freeiov;
  1732. err = len;
  1733. out_freeiov:
  1734. if (iov != iovstack)
  1735. sock_kfree_s(sock->sk, iov, iov_size);
  1736. out:
  1737. return err;
  1738. }
  1739. /*
  1740. * BSD recvmsg interface
  1741. */
  1742. SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
  1743. unsigned int, flags)
  1744. {
  1745. int fput_needed, err;
  1746. struct msghdr msg_sys;
  1747. struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1748. if (!sock)
  1749. goto out;
  1750. err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
  1751. fput_light(sock->file, fput_needed);
  1752. out:
  1753. return err;
  1754. }
  1755. /*
  1756. * Linux recvmmsg interface
  1757. */
  1758. int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
  1759. unsigned int flags, struct timespec *timeout)
  1760. {
  1761. int fput_needed, err, datagrams;
  1762. struct socket *sock;
  1763. struct mmsghdr __user *entry;
  1764. struct compat_mmsghdr __user *compat_entry;
  1765. struct msghdr msg_sys;
  1766. struct timespec end_time;
  1767. if (timeout &&
  1768. poll_select_set_timeout(&end_time, timeout->tv_sec,
  1769. timeout->tv_nsec))
  1770. return -EINVAL;
  1771. datagrams = 0;
  1772. sock = sockfd_lookup_light(fd, &err, &fput_needed);
  1773. if (!sock)
  1774. return err;
  1775. err = sock_error(sock->sk);
  1776. if (err)
  1777. goto out_put;
  1778. entry = mmsg;
  1779. compat_entry = (struct compat_mmsghdr __user *)mmsg;
  1780. while (datagrams < vlen) {
  1781. /*
  1782. * No need to ask LSM for more than the first datagram.
  1783. */
  1784. if (MSG_CMSG_COMPAT & flags) {
  1785. err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
  1786. &msg_sys, flags, datagrams);
  1787. if (err < 0)
  1788. break;
  1789. err = __put_user(err, &compat_entry->msg_len);
  1790. ++compat_entry;
  1791. } else {
  1792. err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
  1793. &msg_sys, flags, datagrams);
  1794. if (err < 0)
  1795. break;
  1796. err = put_user(err, &entry->msg_len);
  1797. ++entry;
  1798. }
  1799. if (err)
  1800. break;
  1801. ++datagrams;
  1802. /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
  1803. if (flags & MSG_WAITFORONE)
  1804. flags |= MSG_DONTWAIT;
  1805. if (timeout) {
  1806. ktime_get_ts(timeout);
  1807. *timeout = timespec_sub(end_time, *timeout);
  1808. if (timeout->tv_sec < 0) {
  1809. timeout->tv_sec = timeout->tv_nsec = 0;
  1810. break;
  1811. }
  1812. /* Timeout, return less than vlen datagrams */
  1813. if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
  1814. break;
  1815. }
  1816. /* Out of band data, return right away */
  1817. if (msg_sys.msg_flags & MSG_OOB)
  1818. break;
  1819. }
  1820. out_put:
  1821. fput_light(sock->file, fput_needed);
  1822. if (err == 0)
  1823. return datagrams;
  1824. if (datagrams != 0) {
  1825. /*
  1826. * We may return less entries than requested (vlen) if the
  1827. * sock is non block and there aren't enough datagrams...
  1828. */
  1829. if (err != -EAGAIN) {
  1830. /*
  1831. * ... or if recvmsg returns an error after we
  1832. * received some datagrams, where we record the
  1833. * error to return on the next call or if the
  1834. * app asks about it using getsockopt(SO_ERROR).
  1835. */
  1836. sock->sk->sk_err = -err;
  1837. }
  1838. return datagrams;
  1839. }
  1840. return err;
  1841. }
  1842. SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
  1843. unsigned int, vlen, unsigned int, flags,
  1844. struct timespec __user *, timeout)
  1845. {
  1846. int datagrams;
  1847. struct timespec timeout_sys;
  1848. if (!timeout)
  1849. return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
  1850. if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
  1851. return -EFAULT;
  1852. datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
  1853. if (datagrams > 0 &&
  1854. copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
  1855. datagrams = -EFAULT;
  1856. return datagrams;
  1857. }
  1858. #ifdef __ARCH_WANT_SYS_SOCKETCALL
  1859. /* Argument list sizes for sys_socketcall */
  1860. #define AL(x) ((x) * sizeof(unsigned long))
  1861. static const unsigned char nargs[20] = {
  1862. AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
  1863. AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
  1864. AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
  1865. AL(4), AL(5)
  1866. };
  1867. #undef AL
  1868. /*
  1869. * System call vectors.
  1870. *
  1871. * Argument checking cleaned up. Saved 20% in size.
  1872. * This function doesn't need to set the kernel lock because
  1873. * it is set by the callees.
  1874. */
  1875. SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
  1876. {
  1877. unsigned long a[6];
  1878. unsigned long a0, a1;
  1879. int err;
  1880. unsigned int len;
  1881. if (call < 1 || call > SYS_RECVMMSG)
  1882. return -EINVAL;
  1883. len = nargs[call];
  1884. if (len > sizeof(a))
  1885. return -EINVAL;
  1886. /* copy_from_user should be SMP safe. */
  1887. if (copy_from_user(a, args, len))
  1888. return -EFAULT;
  1889. audit_socketcall(nargs[call] / sizeof(unsigned long), a);
  1890. a0 = a[0];
  1891. a1 = a[1];
  1892. switch (call) {
  1893. case SYS_SOCKET:
  1894. err = sys_socket(a0, a1, a[2]);
  1895. break;
  1896. case SYS_BIND:
  1897. err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
  1898. break;
  1899. case SYS_CONNECT:
  1900. err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
  1901. break;
  1902. case SYS_LISTEN:
  1903. err = sys_listen(a0, a1);
  1904. break;
  1905. case SYS_ACCEPT:
  1906. err = sys_accept4(a0, (struct sockaddr __user *)a1,
  1907. (int __user *)a[2], 0);
  1908. break;
  1909. case SYS_GETSOCKNAME:
  1910. err =
  1911. sys_getsockname(a0, (struct sockaddr __user *)a1,
  1912. (int __user *)a[2]);
  1913. break;
  1914. case SYS_GETPEERNAME:
  1915. err =
  1916. sys_getpeername(a0, (struct sockaddr __user *)a1,
  1917. (int __user *)a[2]);
  1918. break;
  1919. case SYS_SOCKETPAIR:
  1920. err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
  1921. break;
  1922. case SYS_SEND:
  1923. err = sys_send(a0, (void __user *)a1, a[2], a[3]);
  1924. break;
  1925. case SYS_SENDTO:
  1926. err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
  1927. (struct sockaddr __user *)a[4], a[5]);
  1928. break;
  1929. case SYS_RECV:
  1930. err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
  1931. break;
  1932. case SYS_RECVFROM:
  1933. err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
  1934. (struct sockaddr __user *)a[4],
  1935. (int __user *)a[5]);
  1936. break;
  1937. case SYS_SHUTDOWN:
  1938. err = sys_shutdown(a0, a1);
  1939. break;
  1940. case SYS_SETSOCKOPT:
  1941. err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
  1942. break;
  1943. case SYS_GETSOCKOPT:
  1944. err =
  1945. sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
  1946. (int __user *)a[4]);
  1947. break;
  1948. case SYS_SENDMSG:
  1949. err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
  1950. break;
  1951. case SYS_RECVMSG:
  1952. err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
  1953. break;
  1954. case SYS_RECVMMSG:
  1955. err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
  1956. (struct timespec __user *)a[4]);
  1957. break;
  1958. case SYS_ACCEPT4:
  1959. err = sys_accept4(a0, (struct sockaddr __user *)a1,
  1960. (int __user *)a[2], a[3]);
  1961. break;
  1962. default:
  1963. err = -EINVAL;
  1964. break;
  1965. }
  1966. return err;
  1967. }
  1968. #endif /* __ARCH_WANT_SYS_SOCKETCALL */
  1969. /**
  1970. * sock_register - add a socket protocol handler
  1971. * @ops: description of protocol
  1972. *
  1973. * This function is called by a protocol handler that wants to
  1974. * advertise its address family, and have it linked into the
  1975. * socket interface. The value ops->family coresponds to the
  1976. * socket system call protocol family.
  1977. */
  1978. int sock_register(const struct net_proto_family *ops)
  1979. {
  1980. int err;
  1981. if (ops->family >= NPROTO) {
  1982. printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
  1983. NPROTO);
  1984. return -ENOBUFS;
  1985. }
  1986. spin_lock(&net_family_lock);
  1987. if (net_families[ops->family])
  1988. err = -EEXIST;
  1989. else {
  1990. net_families[ops->family] = ops;
  1991. err = 0;
  1992. }
  1993. spin_unlock(&net_family_lock);
  1994. printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
  1995. return err;
  1996. }
  1997. EXPORT_SYMBOL(sock_register);
  1998. /**
  1999. * sock_unregister - remove a protocol handler
  2000. * @family: protocol family to remove
  2001. *
  2002. * This function is called by a protocol handler that wants to
  2003. * remove its address family, and have it unlinked from the
  2004. * new socket creation.
  2005. *
  2006. * If protocol handler is a module, then it can use module reference
  2007. * counts to protect against new references. If protocol handler is not
  2008. * a module then it needs to provide its own protection in
  2009. * the ops->create routine.
  2010. */
  2011. void sock_unregister(int family)
  2012. {
  2013. BUG_ON(family < 0 || family >= NPROTO);
  2014. spin_lock(&net_family_lock);
  2015. net_families[family] = NULL;
  2016. spin_unlock(&net_family_lock);
  2017. synchronize_rcu();
  2018. printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
  2019. }
  2020. EXPORT_SYMBOL(sock_unregister);
  2021. static int __init sock_init(void)
  2022. {
  2023. /*
  2024. * Initialize sock SLAB cache.
  2025. */
  2026. sk_init();
  2027. /*
  2028. * Initialize skbuff SLAB cache
  2029. */
  2030. skb_init();
  2031. /*
  2032. * Initialize the protocols module.
  2033. */
  2034. init_inodecache();
  2035. register_filesystem(&sock_fs_type);
  2036. sock_mnt = kern_mount(&sock_fs_type);
  2037. /* The real protocol initialization is performed in later initcalls.
  2038. */
  2039. #ifdef CONFIG_NETFILTER
  2040. netfilter_init();
  2041. #endif
  2042. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2043. skb_timestamping_init();
  2044. #endif
  2045. return 0;
  2046. }
  2047. core_initcall(sock_init); /* early initcall */
  2048. #ifdef CONFIG_PROC_FS
  2049. void socket_seq_show(struct seq_file *seq)
  2050. {
  2051. int cpu;
  2052. int counter = 0;
  2053. for_each_possible_cpu(cpu)
  2054. counter += per_cpu(sockets_in_use, cpu);
  2055. /* It can be negative, by the way. 8) */
  2056. if (counter < 0)
  2057. counter = 0;
  2058. seq_printf(seq, "sockets: used %d\n", counter);
  2059. }
  2060. #endif /* CONFIG_PROC_FS */
  2061. #ifdef CONFIG_COMPAT
  2062. static int do_siocgstamp(struct net *net, struct socket *sock,
  2063. unsigned int cmd, struct compat_timeval __user *up)
  2064. {
  2065. mm_segment_t old_fs = get_fs();
  2066. struct timeval ktv;
  2067. int err;
  2068. set_fs(KERNEL_DS);
  2069. err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
  2070. set_fs(old_fs);
  2071. if (!err) {
  2072. err = put_user(ktv.tv_sec, &up->tv_sec);
  2073. err |= __put_user(ktv.tv_usec, &up->tv_usec);
  2074. }
  2075. return err;
  2076. }
  2077. static int do_siocgstampns(struct net *net, struct socket *sock,
  2078. unsigned int cmd, struct compat_timespec __user *up)
  2079. {
  2080. mm_segment_t old_fs = get_fs();
  2081. struct timespec kts;
  2082. int err;
  2083. set_fs(KERNEL_DS);
  2084. err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
  2085. set_fs(old_fs);
  2086. if (!err) {
  2087. err = put_user(kts.tv_sec, &up->tv_sec);
  2088. err |= __put_user(kts.tv_nsec, &up->tv_nsec);
  2089. }
  2090. return err;
  2091. }
  2092. static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
  2093. {
  2094. struct ifreq __user *uifr;
  2095. int err;
  2096. uifr = compat_alloc_user_space(sizeof(struct ifreq));
  2097. if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
  2098. return -EFAULT;
  2099. err = dev_ioctl(net, SIOCGIFNAME, uifr);
  2100. if (err)
  2101. return err;
  2102. if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
  2103. return -EFAULT;
  2104. return 0;
  2105. }
  2106. static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
  2107. {
  2108. struct compat_ifconf ifc32;
  2109. struct ifconf ifc;
  2110. struct ifconf __user *uifc;
  2111. struct compat_ifreq __user *ifr32;
  2112. struct ifreq __user *ifr;
  2113. unsigned int i, j;
  2114. int err;
  2115. if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
  2116. return -EFAULT;
  2117. if (ifc32.ifcbuf == 0) {
  2118. ifc32.ifc_len = 0;
  2119. ifc.ifc_len = 0;
  2120. ifc.ifc_req = NULL;
  2121. uifc = compat_alloc_user_space(sizeof(struct ifconf));
  2122. } else {
  2123. size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
  2124. sizeof(struct ifreq);
  2125. uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
  2126. ifc.ifc_len = len;
  2127. ifr = ifc.ifc_req = (void __user *)(uifc + 1);
  2128. ifr32 = compat_ptr(ifc32.ifcbuf);
  2129. for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
  2130. if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
  2131. return -EFAULT;
  2132. ifr++;
  2133. ifr32++;
  2134. }
  2135. }
  2136. if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
  2137. return -EFAULT;
  2138. err = dev_ioctl(net, SIOCGIFCONF, uifc);
  2139. if (err)
  2140. return err;
  2141. if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
  2142. return -EFAULT;
  2143. ifr = ifc.ifc_req;
  2144. ifr32 = compat_ptr(ifc32.ifcbuf);
  2145. for (i = 0, j = 0;
  2146. i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
  2147. i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
  2148. if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
  2149. return -EFAULT;
  2150. ifr32++;
  2151. ifr++;
  2152. }
  2153. if (ifc32.ifcbuf == 0) {
  2154. /* Translate from 64-bit structure multiple to
  2155. * a 32-bit one.
  2156. */
  2157. i = ifc.ifc_len;
  2158. i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
  2159. ifc32.ifc_len = i;
  2160. } else {
  2161. ifc32.ifc_len = i;
  2162. }
  2163. if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
  2164. return -EFAULT;
  2165. return 0;
  2166. }
  2167. static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
  2168. {
  2169. struct ifreq __user *ifr;
  2170. u32 data;
  2171. void __user *datap;
  2172. ifr = compat_alloc_user_space(sizeof(*ifr));
  2173. if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
  2174. return -EFAULT;
  2175. if (get_user(data, &ifr32->ifr_ifru.ifru_data))
  2176. return -EFAULT;
  2177. datap = compat_ptr(data);
  2178. if (put_user(datap, &ifr->ifr_ifru.ifru_data))
  2179. return -EFAULT;
  2180. return dev_ioctl(net, SIOCETHTOOL, ifr);
  2181. }
  2182. static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
  2183. {
  2184. void __user *uptr;
  2185. compat_uptr_t uptr32;
  2186. struct ifreq __user *uifr;
  2187. uifr = compat_alloc_user_space(sizeof(*uifr));
  2188. if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
  2189. return -EFAULT;
  2190. if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
  2191. return -EFAULT;
  2192. uptr = compat_ptr(uptr32);
  2193. if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
  2194. return -EFAULT;
  2195. return dev_ioctl(net, SIOCWANDEV, uifr);
  2196. }
  2197. static int bond_ioctl(struct net *net, unsigned int cmd,
  2198. struct compat_ifreq __user *ifr32)
  2199. {
  2200. struct ifreq kifr;
  2201. struct ifreq __user *uifr;
  2202. mm_segment_t old_fs;
  2203. int err;
  2204. u32 data;
  2205. void __user *datap;
  2206. switch (cmd) {
  2207. case SIOCBONDENSLAVE:
  2208. case SIOCBONDRELEASE:
  2209. case SIOCBONDSETHWADDR:
  2210. case SIOCBONDCHANGEACTIVE:
  2211. if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
  2212. return -EFAULT;
  2213. old_fs = get_fs();
  2214. set_fs(KERNEL_DS);
  2215. err = dev_ioctl(net, cmd, &kifr);
  2216. set_fs(old_fs);
  2217. return err;
  2218. case SIOCBONDSLAVEINFOQUERY:
  2219. case SIOCBONDINFOQUERY:
  2220. uifr = compat_alloc_user_space(sizeof(*uifr));
  2221. if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
  2222. return -EFAULT;
  2223. if (get_user(data, &ifr32->ifr_ifru.ifru_data))
  2224. return -EFAULT;
  2225. datap = compat_ptr(data);
  2226. if (put_user(datap, &uifr->ifr_ifru.ifru_data))
  2227. return -EFAULT;
  2228. return dev_ioctl(net, cmd, uifr);
  2229. default:
  2230. return -EINVAL;
  2231. }
  2232. }
  2233. static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
  2234. struct compat_ifreq __user *u_ifreq32)
  2235. {
  2236. struct ifreq __user *u_ifreq64;
  2237. char tmp_buf[IFNAMSIZ];
  2238. void __user *data64;
  2239. u32 data32;
  2240. if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
  2241. IFNAMSIZ))
  2242. return -EFAULT;
  2243. if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
  2244. return -EFAULT;
  2245. data64 = compat_ptr(data32);
  2246. u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
  2247. /* Don't check these user accesses, just let that get trapped
  2248. * in the ioctl handler instead.
  2249. */
  2250. if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
  2251. IFNAMSIZ))
  2252. return -EFAULT;
  2253. if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
  2254. return -EFAULT;
  2255. return dev_ioctl(net, cmd, u_ifreq64);
  2256. }
  2257. static int dev_ifsioc(struct net *net, struct socket *sock,
  2258. unsigned int cmd, struct compat_ifreq __user *uifr32)
  2259. {
  2260. struct ifreq __user *uifr;
  2261. int err;
  2262. uifr = compat_alloc_user_space(sizeof(*uifr));
  2263. if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
  2264. return -EFAULT;
  2265. err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
  2266. if (!err) {
  2267. switch (cmd) {
  2268. case SIOCGIFFLAGS:
  2269. case SIOCGIFMETRIC:
  2270. case SIOCGIFMTU:
  2271. case SIOCGIFMEM:
  2272. case SIOCGIFHWADDR:
  2273. case SIOCGIFINDEX:
  2274. case SIOCGIFADDR:
  2275. case SIOCGIFBRDADDR:
  2276. case SIOCGIFDSTADDR:
  2277. case SIOCGIFNETMASK:
  2278. case SIOCGIFPFLAGS:
  2279. case SIOCGIFTXQLEN:
  2280. case SIOCGMIIPHY:
  2281. case SIOCGMIIREG:
  2282. if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
  2283. err = -EFAULT;
  2284. break;
  2285. }
  2286. }
  2287. return err;
  2288. }
  2289. static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
  2290. struct compat_ifreq __user *uifr32)
  2291. {
  2292. struct ifreq ifr;
  2293. struct compat_ifmap __user *uifmap32;
  2294. mm_segment_t old_fs;
  2295. int err;
  2296. uifmap32 = &uifr32->ifr_ifru.ifru_map;
  2297. err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
  2298. err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
  2299. err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
  2300. err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
  2301. err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
  2302. err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
  2303. err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
  2304. if (err)
  2305. return -EFAULT;
  2306. old_fs = get_fs();
  2307. set_fs(KERNEL_DS);
  2308. err = dev_ioctl(net, cmd, (void __user *)&ifr);
  2309. set_fs(old_fs);
  2310. if (cmd == SIOCGIFMAP && !err) {
  2311. err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
  2312. err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
  2313. err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
  2314. err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
  2315. err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
  2316. err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
  2317. err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
  2318. if (err)
  2319. err = -EFAULT;
  2320. }
  2321. return err;
  2322. }
  2323. static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
  2324. {
  2325. void __user *uptr;
  2326. compat_uptr_t uptr32;
  2327. struct ifreq __user *uifr;
  2328. uifr = compat_alloc_user_space(sizeof(*uifr));
  2329. if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
  2330. return -EFAULT;
  2331. if (get_user(uptr32, &uifr32->ifr_data))
  2332. return -EFAULT;
  2333. uptr = compat_ptr(uptr32);
  2334. if (put_user(uptr, &uifr->ifr_data))
  2335. return -EFAULT;
  2336. return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
  2337. }
  2338. struct rtentry32 {
  2339. u32 rt_pad1;
  2340. struct sockaddr rt_dst; /* target address */
  2341. struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
  2342. struct sockaddr rt_genmask; /* target network mask (IP) */
  2343. unsigned short rt_flags;
  2344. short rt_pad2;
  2345. u32 rt_pad3;
  2346. unsigned char rt_tos;
  2347. unsigned char rt_class;
  2348. short rt_pad4;
  2349. short rt_metric; /* +1 for binary compatibility! */
  2350. /* char * */ u32 rt_dev; /* forcing the device at add */
  2351. u32 rt_mtu; /* per route MTU/Window */
  2352. u32 rt_window; /* Window clamping */
  2353. unsigned short rt_irtt; /* Initial RTT */
  2354. };
  2355. struct in6_rtmsg32 {
  2356. struct in6_addr rtmsg_dst;
  2357. struct in6_addr rtmsg_src;
  2358. struct in6_addr rtmsg_gateway;
  2359. u32 rtmsg_type;
  2360. u16 rtmsg_dst_len;
  2361. u16 rtmsg_src_len;
  2362. u32 rtmsg_metric;
  2363. u32 rtmsg_info;
  2364. u32 rtmsg_flags;
  2365. s32 rtmsg_ifindex;
  2366. };
  2367. static int routing_ioctl(struct net *net, struct socket *sock,
  2368. unsigned int cmd, void __user *argp)
  2369. {
  2370. int ret;
  2371. void *r = NULL;
  2372. struct in6_rtmsg r6;
  2373. struct rtentry r4;
  2374. char devname[16];
  2375. u32 rtdev;
  2376. mm_segment_t old_fs = get_fs();
  2377. if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
  2378. struct in6_rtmsg32 __user *ur6 = argp;
  2379. ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
  2380. 3 * sizeof(struct in6_addr));
  2381. ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
  2382. ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
  2383. ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
  2384. ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
  2385. ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
  2386. ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
  2387. ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
  2388. r = (void *) &r6;
  2389. } else { /* ipv4 */
  2390. struct rtentry32 __user *ur4 = argp;
  2391. ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
  2392. 3 * sizeof(struct sockaddr));
  2393. ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
  2394. ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
  2395. ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
  2396. ret |= __get_user(r4.rt_window, &(ur4->rt_window));
  2397. ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
  2398. ret |= __get_user(rtdev, &(ur4->rt_dev));
  2399. if (rtdev) {
  2400. ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
  2401. r4.rt_dev = devname; devname[15] = 0;
  2402. } else
  2403. r4.rt_dev = NULL;
  2404. r = (void *) &r4;
  2405. }
  2406. if (ret) {
  2407. ret = -EFAULT;
  2408. goto out;
  2409. }
  2410. set_fs(KERNEL_DS);
  2411. ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
  2412. set_fs(old_fs);
  2413. out:
  2414. return ret;
  2415. }
  2416. /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
  2417. * for some operations; this forces use of the newer bridge-utils that
  2418. * use compatiable ioctls
  2419. */
  2420. static int old_bridge_ioctl(compat_ulong_t __user *argp)
  2421. {
  2422. compat_ulong_t tmp;
  2423. if (get_user(tmp, argp))
  2424. return -EFAULT;
  2425. if (tmp == BRCTL_GET_VERSION)
  2426. return BRCTL_VERSION + 1;
  2427. return -EINVAL;
  2428. }
  2429. static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
  2430. unsigned int cmd, unsigned long arg)
  2431. {
  2432. void __user *argp = compat_ptr(arg);
  2433. struct sock *sk = sock->sk;
  2434. struct net *net = sock_net(sk);
  2435. if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
  2436. return siocdevprivate_ioctl(net, cmd, argp);
  2437. switch (cmd) {
  2438. case SIOCSIFBR:
  2439. case SIOCGIFBR:
  2440. return old_bridge_ioctl(argp);
  2441. case SIOCGIFNAME:
  2442. return dev_ifname32(net, argp);
  2443. case SIOCGIFCONF:
  2444. return dev_ifconf(net, argp);
  2445. case SIOCETHTOOL:
  2446. return ethtool_ioctl(net, argp);
  2447. case SIOCWANDEV:
  2448. return compat_siocwandev(net, argp);
  2449. case SIOCGIFMAP:
  2450. case SIOCSIFMAP:
  2451. return compat_sioc_ifmap(net, cmd, argp);
  2452. case SIOCBONDENSLAVE:
  2453. case SIOCBONDRELEASE:
  2454. case SIOCBONDSETHWADDR:
  2455. case SIOCBONDSLAVEINFOQUERY:
  2456. case SIOCBONDINFOQUERY:
  2457. case SIOCBONDCHANGEACTIVE:
  2458. return bond_ioctl(net, cmd, argp);
  2459. case SIOCADDRT:
  2460. case SIOCDELRT:
  2461. return routing_ioctl(net, sock, cmd, argp);
  2462. case SIOCGSTAMP:
  2463. return do_siocgstamp(net, sock, cmd, argp);
  2464. case SIOCGSTAMPNS:
  2465. return do_siocgstampns(net, sock, cmd, argp);
  2466. case SIOCSHWTSTAMP:
  2467. return compat_siocshwtstamp(net, argp);
  2468. case FIOSETOWN:
  2469. case SIOCSPGRP:
  2470. case FIOGETOWN:
  2471. case SIOCGPGRP:
  2472. case SIOCBRADDBR:
  2473. case SIOCBRDELBR:
  2474. case SIOCGIFVLAN:
  2475. case SIOCSIFVLAN:
  2476. case SIOCADDDLCI:
  2477. case SIOCDELDLCI:
  2478. return sock_ioctl(file, cmd, arg);
  2479. case SIOCGIFFLAGS:
  2480. case SIOCSIFFLAGS:
  2481. case SIOCGIFMETRIC:
  2482. case SIOCSIFMETRIC:
  2483. case SIOCGIFMTU:
  2484. case SIOCSIFMTU:
  2485. case SIOCGIFMEM:
  2486. case SIOCSIFMEM:
  2487. case SIOCGIFHWADDR:
  2488. case SIOCSIFHWADDR:
  2489. case SIOCADDMULTI:
  2490. case SIOCDELMULTI:
  2491. case SIOCGIFINDEX:
  2492. case SIOCGIFADDR:
  2493. case SIOCSIFADDR:
  2494. case SIOCSIFHWBROADCAST:
  2495. case SIOCDIFADDR:
  2496. case SIOCGIFBRDADDR:
  2497. case SIOCSIFBRDADDR:
  2498. case SIOCGIFDSTADDR:
  2499. case SIOCSIFDSTADDR:
  2500. case SIOCGIFNETMASK:
  2501. case SIOCSIFNETMASK:
  2502. case SIOCSIFPFLAGS:
  2503. case SIOCGIFPFLAGS:
  2504. case SIOCGIFTXQLEN:
  2505. case SIOCSIFTXQLEN:
  2506. case SIOCBRADDIF:
  2507. case SIOCBRDELIF:
  2508. case SIOCSIFNAME:
  2509. case SIOCGMIIPHY:
  2510. case SIOCGMIIREG:
  2511. case SIOCSMIIREG:
  2512. return dev_ifsioc(net, sock, cmd, argp);
  2513. case SIOCSARP:
  2514. case SIOCGARP:
  2515. case SIOCDARP:
  2516. case SIOCATMARK:
  2517. return sock_do_ioctl(net, sock, cmd, arg);
  2518. }
  2519. /* Prevent warning from compat_sys_ioctl, these always
  2520. * result in -EINVAL in the native case anyway. */
  2521. switch (cmd) {
  2522. case SIOCRTMSG:
  2523. case SIOCGIFCOUNT:
  2524. case SIOCSRARP:
  2525. case SIOCGRARP:
  2526. case SIOCDRARP:
  2527. case SIOCSIFLINK:
  2528. case SIOCGIFSLAVE:
  2529. case SIOCSIFSLAVE:
  2530. return -EINVAL;
  2531. }
  2532. return -ENOIOCTLCMD;
  2533. }
  2534. static long compat_sock_ioctl(struct file *file, unsigned cmd,
  2535. unsigned long arg)
  2536. {
  2537. struct socket *sock = file->private_data;
  2538. int ret = -ENOIOCTLCMD;
  2539. struct sock *sk;
  2540. struct net *net;
  2541. sk = sock->sk;
  2542. net = sock_net(sk);
  2543. if (sock->ops->compat_ioctl)
  2544. ret = sock->ops->compat_ioctl(sock, cmd, arg);
  2545. if (ret == -ENOIOCTLCMD &&
  2546. (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
  2547. ret = compat_wext_handle_ioctl(net, cmd, arg);
  2548. if (ret == -ENOIOCTLCMD)
  2549. ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
  2550. return ret;
  2551. }
  2552. #endif
  2553. int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
  2554. {
  2555. return sock->ops->bind(sock, addr, addrlen);
  2556. }
  2557. EXPORT_SYMBOL(kernel_bind);
  2558. int kernel_listen(struct socket *sock, int backlog)
  2559. {
  2560. return sock->ops->listen(sock, backlog);
  2561. }
  2562. EXPORT_SYMBOL(kernel_listen);
  2563. int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
  2564. {
  2565. struct sock *sk = sock->sk;
  2566. int err;
  2567. err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
  2568. newsock);
  2569. if (err < 0)
  2570. goto done;
  2571. err = sock->ops->accept(sock, *newsock, flags);
  2572. if (err < 0) {
  2573. sock_release(*newsock);
  2574. *newsock = NULL;
  2575. goto done;
  2576. }
  2577. (*newsock)->ops = sock->ops;
  2578. __module_get((*newsock)->ops->owner);
  2579. done:
  2580. return err;
  2581. }
  2582. EXPORT_SYMBOL(kernel_accept);
  2583. int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
  2584. int flags)
  2585. {
  2586. return sock->ops->connect(sock, addr, addrlen, flags);
  2587. }
  2588. EXPORT_SYMBOL(kernel_connect);
  2589. int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
  2590. int *addrlen)
  2591. {
  2592. return sock->ops->getname(sock, addr, addrlen, 0);
  2593. }
  2594. EXPORT_SYMBOL(kernel_getsockname);
  2595. int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
  2596. int *addrlen)
  2597. {
  2598. return sock->ops->getname(sock, addr, addrlen, 1);
  2599. }
  2600. EXPORT_SYMBOL(kernel_getpeername);
  2601. int kernel_getsockopt(struct socket *sock, int level, int optname,
  2602. char *optval, int *optlen)
  2603. {
  2604. mm_segment_t oldfs = get_fs();
  2605. int err;
  2606. set_fs(KERNEL_DS);
  2607. if (level == SOL_SOCKET)
  2608. err = sock_getsockopt(sock, level, optname, optval, optlen);
  2609. else
  2610. err = sock->ops->getsockopt(sock, level, optname, optval,
  2611. optlen);
  2612. set_fs(oldfs);
  2613. return err;
  2614. }
  2615. EXPORT_SYMBOL(kernel_getsockopt);
  2616. int kernel_setsockopt(struct socket *sock, int level, int optname,
  2617. char *optval, unsigned int optlen)
  2618. {
  2619. mm_segment_t oldfs = get_fs();
  2620. int err;
  2621. set_fs(KERNEL_DS);
  2622. if (level == SOL_SOCKET)
  2623. err = sock_setsockopt(sock, level, optname, optval, optlen);
  2624. else
  2625. err = sock->ops->setsockopt(sock, level, optname, optval,
  2626. optlen);
  2627. set_fs(oldfs);
  2628. return err;
  2629. }
  2630. EXPORT_SYMBOL(kernel_setsockopt);
  2631. int kernel_sendpage(struct socket *sock, struct page *page, int offset,
  2632. size_t size, int flags)
  2633. {
  2634. sock_update_classid(sock->sk);
  2635. if (sock->ops->sendpage)
  2636. return sock->ops->sendpage(sock, page, offset, size, flags);
  2637. return sock_no_sendpage(sock, page, offset, size, flags);
  2638. }
  2639. EXPORT_SYMBOL(kernel_sendpage);
  2640. int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
  2641. {
  2642. mm_segment_t oldfs = get_fs();
  2643. int err;
  2644. set_fs(KERNEL_DS);
  2645. err = sock->ops->ioctl(sock, cmd, arg);
  2646. set_fs(oldfs);
  2647. return err;
  2648. }
  2649. EXPORT_SYMBOL(kernel_sock_ioctl);
  2650. int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
  2651. {
  2652. return sock->ops->shutdown(sock, how);
  2653. }
  2654. EXPORT_SYMBOL(kernel_sock_shutdown);