]>
Commit | Line | Data |
---|---|---|
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 | ||
61 | #include <linux/mm.h> | |
62 | #include <linux/socket.h> | |
63 | #include <linux/file.h> | |
64 | #include <linux/net.h> | |
65 | #include <linux/interrupt.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 | ||
89 | #include <asm/uaccess.h> | |
90 | #include <asm/unistd.h> | |
91 | ||
92 | #include <net/compat.h> | |
93 | #include <net/wext.h> | |
94 | ||
95 | #include <net/sock.h> | |
96 | #include <linux/netfilter.h> | |
97 | ||
98 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare); | |
99 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, | |
100 | unsigned long nr_segs, loff_t pos); | |
101 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
102 | unsigned long nr_segs, loff_t pos); | |
103 | static int sock_mmap(struct file *file, struct vm_area_struct *vma); | |
104 | ||
105 | static int sock_close(struct inode *inode, struct file *file); | |
106 | static unsigned int sock_poll(struct file *file, | |
107 | struct poll_table_struct *wait); | |
108 | static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); | |
109 | #ifdef CONFIG_COMPAT | |
110 | static long compat_sock_ioctl(struct file *file, | |
111 | unsigned int cmd, unsigned long arg); | |
112 | #endif | |
113 | static int sock_fasync(int fd, struct file *filp, int on); | |
114 | static ssize_t sock_sendpage(struct file *file, struct page *page, | |
115 | int offset, size_t size, loff_t *ppos, int more); | |
116 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, | |
117 | struct pipe_inode_info *pipe, size_t len, | |
118 | unsigned int flags); | |
119 | ||
120 | /* | |
121 | * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear | |
122 | * in the operation structures but are done directly via the socketcall() multiplexor. | |
123 | */ | |
124 | ||
125 | static const struct file_operations socket_file_ops = { | |
126 | .owner = THIS_MODULE, | |
127 | .llseek = no_llseek, | |
128 | .aio_read = sock_aio_read, | |
129 | .aio_write = sock_aio_write, | |
130 | .poll = sock_poll, | |
131 | .unlocked_ioctl = sock_ioctl, | |
132 | #ifdef CONFIG_COMPAT | |
133 | .compat_ioctl = compat_sock_ioctl, | |
134 | #endif | |
135 | .mmap = sock_mmap, | |
136 | .open = sock_no_open, /* special open code to disallow open via /proc */ | |
137 | .release = sock_close, | |
138 | .fasync = sock_fasync, | |
139 | .sendpage = sock_sendpage, | |
140 | .splice_write = generic_splice_sendpage, | |
141 | .splice_read = sock_splice_read, | |
142 | }; | |
143 | ||
144 | /* | |
145 | * The protocol list. Each protocol is registered in here. | |
146 | */ | |
147 | ||
148 | static DEFINE_SPINLOCK(net_family_lock); | |
149 | static const struct net_proto_family *net_families[NPROTO] __read_mostly; | |
150 | ||
151 | /* | |
152 | * Statistics counters of the socket lists | |
153 | */ | |
154 | ||
155 | static DEFINE_PER_CPU(int, sockets_in_use) = 0; | |
156 | ||
157 | /* | |
158 | * Support routines. | |
159 | * Move socket addresses back and forth across the kernel/user | |
160 | * divide and look after the messy bits. | |
161 | */ | |
162 | ||
163 | #define MAX_SOCK_ADDR 128 /* 108 for Unix domain - | |
164 | 16 for IP, 16 for IPX, | |
165 | 24 for IPv6, | |
166 | about 80 for AX.25 | |
167 | must be at least one bigger than | |
168 | the AF_UNIX size (see net/unix/af_unix.c | |
169 | :unix_mkname()). | |
170 | */ | |
171 | ||
172 | /** | |
173 | * move_addr_to_kernel - copy a socket address into kernel space | |
174 | * @uaddr: Address in user space | |
175 | * @kaddr: Address in kernel space | |
176 | * @ulen: Length in user space | |
177 | * | |
178 | * The address is copied into kernel space. If the provided address is | |
179 | * too long an error code of -EINVAL is returned. If the copy gives | |
180 | * invalid addresses -EFAULT is returned. On a success 0 is returned. | |
181 | */ | |
182 | ||
183 | int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr) | |
184 | { | |
185 | if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) | |
186 | return -EINVAL; | |
187 | if (ulen == 0) | |
188 | return 0; | |
189 | if (copy_from_user(kaddr, uaddr, ulen)) | |
190 | return -EFAULT; | |
191 | return audit_sockaddr(ulen, kaddr); | |
192 | } | |
193 | ||
194 | /** | |
195 | * move_addr_to_user - copy an address to user space | |
196 | * @kaddr: kernel space address | |
197 | * @klen: length of address in kernel | |
198 | * @uaddr: user space address | |
199 | * @ulen: pointer to user length field | |
200 | * | |
201 | * The value pointed to by ulen on entry is the buffer length available. | |
202 | * This is overwritten with the buffer space used. -EINVAL is returned | |
203 | * if an overlong buffer is specified or a negative buffer size. -EFAULT | |
204 | * is returned if either the buffer or the length field are not | |
205 | * accessible. | |
206 | * After copying the data up to the limit the user specifies, the true | |
207 | * length of the data is written over the length limit the user | |
208 | * specified. Zero is returned for a success. | |
209 | */ | |
210 | ||
211 | int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr, | |
212 | int __user *ulen) | |
213 | { | |
214 | int err; | |
215 | int len; | |
216 | ||
217 | err = get_user(len, ulen); | |
218 | if (err) | |
219 | return err; | |
220 | if (len > klen) | |
221 | len = klen; | |
222 | if (len < 0 || len > sizeof(struct sockaddr_storage)) | |
223 | return -EINVAL; | |
224 | if (len) { | |
225 | if (audit_sockaddr(klen, kaddr)) | |
226 | return -ENOMEM; | |
227 | if (copy_to_user(uaddr, kaddr, len)) | |
228 | return -EFAULT; | |
229 | } | |
230 | /* | |
231 | * "fromlen shall refer to the value before truncation.." | |
232 | * 1003.1g | |
233 | */ | |
234 | return __put_user(klen, ulen); | |
235 | } | |
236 | ||
237 | #define SOCKFS_MAGIC 0x534F434B | |
238 | ||
239 | static struct kmem_cache *sock_inode_cachep __read_mostly; | |
240 | ||
241 | static struct inode *sock_alloc_inode(struct super_block *sb) | |
242 | { | |
243 | struct socket_alloc *ei; | |
244 | ||
245 | ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); | |
246 | if (!ei) | |
247 | return NULL; | |
248 | init_waitqueue_head(&ei->socket.wait); | |
249 | ||
250 | ei->socket.fasync_list = NULL; | |
251 | ei->socket.state = SS_UNCONNECTED; | |
252 | ei->socket.flags = 0; | |
253 | ei->socket.ops = NULL; | |
254 | ei->socket.sk = NULL; | |
255 | ei->socket.file = NULL; | |
256 | ||
257 | return &ei->vfs_inode; | |
258 | } | |
259 | ||
260 | static void sock_destroy_inode(struct inode *inode) | |
261 | { | |
262 | kmem_cache_free(sock_inode_cachep, | |
263 | container_of(inode, struct socket_alloc, vfs_inode)); | |
264 | } | |
265 | ||
266 | static void init_once(struct kmem_cache *cachep, void *foo) | |
267 | { | |
268 | struct socket_alloc *ei = (struct socket_alloc *)foo; | |
269 | ||
270 | inode_init_once(&ei->vfs_inode); | |
271 | } | |
272 | ||
273 | static int init_inodecache(void) | |
274 | { | |
275 | sock_inode_cachep = kmem_cache_create("sock_inode_cache", | |
276 | sizeof(struct socket_alloc), | |
277 | 0, | |
278 | (SLAB_HWCACHE_ALIGN | | |
279 | SLAB_RECLAIM_ACCOUNT | | |
280 | SLAB_MEM_SPREAD), | |
281 | init_once); | |
282 | if (sock_inode_cachep == NULL) | |
283 | return -ENOMEM; | |
284 | return 0; | |
285 | } | |
286 | ||
287 | static struct super_operations sockfs_ops = { | |
288 | .alloc_inode = sock_alloc_inode, | |
289 | .destroy_inode =sock_destroy_inode, | |
290 | .statfs = simple_statfs, | |
291 | }; | |
292 | ||
293 | static int sockfs_get_sb(struct file_system_type *fs_type, | |
294 | int flags, const char *dev_name, void *data, | |
295 | struct vfsmount *mnt) | |
296 | { | |
297 | return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC, | |
298 | mnt); | |
299 | } | |
300 | ||
301 | static struct vfsmount *sock_mnt __read_mostly; | |
302 | ||
303 | static struct file_system_type sock_fs_type = { | |
304 | .name = "sockfs", | |
305 | .get_sb = sockfs_get_sb, | |
306 | .kill_sb = kill_anon_super, | |
307 | }; | |
308 | ||
309 | static int sockfs_delete_dentry(struct dentry *dentry) | |
310 | { | |
311 | /* | |
312 | * At creation time, we pretended this dentry was hashed | |
313 | * (by clearing DCACHE_UNHASHED bit in d_flags) | |
314 | * At delete time, we restore the truth : not hashed. | |
315 | * (so that dput() can proceed correctly) | |
316 | */ | |
317 | dentry->d_flags |= DCACHE_UNHASHED; | |
318 | return 0; | |
319 | } | |
320 | ||
321 | /* | |
322 | * sockfs_dname() is called from d_path(). | |
323 | */ | |
324 | static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) | |
325 | { | |
326 | return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", | |
327 | dentry->d_inode->i_ino); | |
328 | } | |
329 | ||
330 | static struct dentry_operations sockfs_dentry_operations = { | |
331 | .d_delete = sockfs_delete_dentry, | |
332 | .d_dname = sockfs_dname, | |
333 | }; | |
334 | ||
335 | /* | |
336 | * Obtains the first available file descriptor and sets it up for use. | |
337 | * | |
338 | * These functions create file structures and maps them to fd space | |
339 | * of the current process. On success it returns file descriptor | |
340 | * and file struct implicitly stored in sock->file. | |
341 | * Note that another thread may close file descriptor before we return | |
342 | * from this function. We use the fact that now we do not refer | |
343 | * to socket after mapping. If one day we will need it, this | |
344 | * function will increment ref. count on file by 1. | |
345 | * | |
346 | * In any case returned fd MAY BE not valid! | |
347 | * This race condition is unavoidable | |
348 | * with shared fd spaces, we cannot solve it inside kernel, | |
349 | * but we take care of internal coherence yet. | |
350 | */ | |
351 | ||
352 | static int sock_alloc_fd(struct file **filep, int flags) | |
353 | { | |
354 | int fd; | |
355 | ||
356 | fd = get_unused_fd_flags(flags); | |
357 | if (likely(fd >= 0)) { | |
358 | struct file *file = get_empty_filp(); | |
359 | ||
360 | *filep = file; | |
361 | if (unlikely(!file)) { | |
362 | put_unused_fd(fd); | |
363 | return -ENFILE; | |
364 | } | |
365 | } else | |
366 | *filep = NULL; | |
367 | return fd; | |
368 | } | |
369 | ||
370 | static int sock_attach_fd(struct socket *sock, struct file *file) | |
371 | { | |
372 | struct dentry *dentry; | |
373 | struct qstr name = { .name = "" }; | |
374 | ||
375 | dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name); | |
376 | if (unlikely(!dentry)) | |
377 | return -ENOMEM; | |
378 | ||
379 | dentry->d_op = &sockfs_dentry_operations; | |
380 | /* | |
381 | * We dont want to push this dentry into global dentry hash table. | |
382 | * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED | |
383 | * This permits a working /proc/$pid/fd/XXX on sockets | |
384 | */ | |
385 | dentry->d_flags &= ~DCACHE_UNHASHED; | |
386 | d_instantiate(dentry, SOCK_INODE(sock)); | |
387 | ||
388 | sock->file = file; | |
389 | init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE, | |
390 | &socket_file_ops); | |
391 | SOCK_INODE(sock)->i_fop = &socket_file_ops; | |
392 | file->f_flags = O_RDWR; | |
393 | file->f_pos = 0; | |
394 | file->private_data = sock; | |
395 | ||
396 | return 0; | |
397 | } | |
398 | ||
399 | int sock_map_fd(struct socket *sock, int flags) | |
400 | { | |
401 | struct file *newfile; | |
402 | int fd = sock_alloc_fd(&newfile, flags); | |
403 | ||
404 | if (likely(fd >= 0)) { | |
405 | int err = sock_attach_fd(sock, newfile); | |
406 | ||
407 | if (unlikely(err < 0)) { | |
408 | put_filp(newfile); | |
409 | put_unused_fd(fd); | |
410 | return err; | |
411 | } | |
412 | fd_install(fd, newfile); | |
413 | } | |
414 | return fd; | |
415 | } | |
416 | ||
417 | static struct socket *sock_from_file(struct file *file, int *err) | |
418 | { | |
419 | if (file->f_op == &socket_file_ops) | |
420 | return file->private_data; /* set in sock_map_fd */ | |
421 | ||
422 | *err = -ENOTSOCK; | |
423 | return NULL; | |
424 | } | |
425 | ||
426 | /** | |
427 | * sockfd_lookup - Go from a file number to its socket slot | |
428 | * @fd: file handle | |
429 | * @err: pointer to an error code return | |
430 | * | |
431 | * The file handle passed in is locked and the socket it is bound | |
432 | * too is returned. If an error occurs the err pointer is overwritten | |
433 | * with a negative errno code and NULL is returned. The function checks | |
434 | * for both invalid handles and passing a handle which is not a socket. | |
435 | * | |
436 | * On a success the socket object pointer is returned. | |
437 | */ | |
438 | ||
439 | struct socket *sockfd_lookup(int fd, int *err) | |
440 | { | |
441 | struct file *file; | |
442 | struct socket *sock; | |
443 | ||
444 | file = fget(fd); | |
445 | if (!file) { | |
446 | *err = -EBADF; | |
447 | return NULL; | |
448 | } | |
449 | ||
450 | sock = sock_from_file(file, err); | |
451 | if (!sock) | |
452 | fput(file); | |
453 | return sock; | |
454 | } | |
455 | ||
456 | static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) | |
457 | { | |
458 | struct file *file; | |
459 | struct socket *sock; | |
460 | ||
461 | *err = -EBADF; | |
462 | file = fget_light(fd, fput_needed); | |
463 | if (file) { | |
464 | sock = sock_from_file(file, err); | |
465 | if (sock) | |
466 | return sock; | |
467 | fput_light(file, *fput_needed); | |
468 | } | |
469 | return NULL; | |
470 | } | |
471 | ||
472 | /** | |
473 | * sock_alloc - allocate a socket | |
474 | * | |
475 | * Allocate a new inode and socket object. The two are bound together | |
476 | * and initialised. The socket is then returned. If we are out of inodes | |
477 | * NULL is returned. | |
478 | */ | |
479 | ||
480 | static struct socket *sock_alloc(void) | |
481 | { | |
482 | struct inode *inode; | |
483 | struct socket *sock; | |
484 | ||
485 | inode = new_inode(sock_mnt->mnt_sb); | |
486 | if (!inode) | |
487 | return NULL; | |
488 | ||
489 | sock = SOCKET_I(inode); | |
490 | ||
491 | inode->i_mode = S_IFSOCK | S_IRWXUGO; | |
492 | inode->i_uid = current->fsuid; | |
493 | inode->i_gid = current->fsgid; | |
494 | ||
495 | get_cpu_var(sockets_in_use)++; | |
496 | put_cpu_var(sockets_in_use); | |
497 | return sock; | |
498 | } | |
499 | ||
500 | /* | |
501 | * In theory you can't get an open on this inode, but /proc provides | |
502 | * a back door. Remember to keep it shut otherwise you'll let the | |
503 | * creepy crawlies in. | |
504 | */ | |
505 | ||
506 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare) | |
507 | { | |
508 | return -ENXIO; | |
509 | } | |
510 | ||
511 | const struct file_operations bad_sock_fops = { | |
512 | .owner = THIS_MODULE, | |
513 | .open = sock_no_open, | |
514 | }; | |
515 | ||
516 | /** | |
517 | * sock_release - close a socket | |
518 | * @sock: socket to close | |
519 | * | |
520 | * The socket is released from the protocol stack if it has a release | |
521 | * callback, and the inode is then released if the socket is bound to | |
522 | * an inode not a file. | |
523 | */ | |
524 | ||
525 | void sock_release(struct socket *sock) | |
526 | { | |
527 | if (sock->ops) { | |
528 | struct module *owner = sock->ops->owner; | |
529 | ||
530 | sock->ops->release(sock); | |
531 | sock->ops = NULL; | |
532 | module_put(owner); | |
533 | } | |
534 | ||
535 | if (sock->fasync_list) | |
536 | printk(KERN_ERR "sock_release: fasync list not empty!\n"); | |
537 | ||
538 | get_cpu_var(sockets_in_use)--; | |
539 | put_cpu_var(sockets_in_use); | |
540 | if (!sock->file) { | |
541 | iput(SOCK_INODE(sock)); | |
542 | return; | |
543 | } | |
544 | sock->file = NULL; | |
545 | } | |
546 | ||
547 | static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock, | |
548 | struct msghdr *msg, size_t size) | |
549 | { | |
550 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
551 | int err; | |
552 | ||
553 | si->sock = sock; | |
554 | si->scm = NULL; | |
555 | si->msg = msg; | |
556 | si->size = size; | |
557 | ||
558 | err = security_socket_sendmsg(sock, msg, size); | |
559 | if (err) | |
560 | return err; | |
561 | ||
562 | return sock->ops->sendmsg(iocb, sock, msg, size); | |
563 | } | |
564 | ||
565 | int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) | |
566 | { | |
567 | struct kiocb iocb; | |
568 | struct sock_iocb siocb; | |
569 | int ret; | |
570 | ||
571 | init_sync_kiocb(&iocb, NULL); | |
572 | iocb.private = &siocb; | |
573 | ret = __sock_sendmsg(&iocb, sock, msg, size); | |
574 | if (-EIOCBQUEUED == ret) | |
575 | ret = wait_on_sync_kiocb(&iocb); | |
576 | return ret; | |
577 | } | |
578 | ||
579 | int kernel_sendmsg(struct socket *sock, struct msghdr *msg, | |
580 | struct kvec *vec, size_t num, size_t size) | |
581 | { | |
582 | mm_segment_t oldfs = get_fs(); | |
583 | int result; | |
584 | ||
585 | set_fs(KERNEL_DS); | |
586 | /* | |
587 | * the following is safe, since for compiler definitions of kvec and | |
588 | * iovec are identical, yielding the same in-core layout and alignment | |
589 | */ | |
590 | msg->msg_iov = (struct iovec *)vec; | |
591 | msg->msg_iovlen = num; | |
592 | result = sock_sendmsg(sock, msg, size); | |
593 | set_fs(oldfs); | |
594 | return result; | |
595 | } | |
596 | ||
597 | /* | |
598 | * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) | |
599 | */ | |
600 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, | |
601 | struct sk_buff *skb) | |
602 | { | |
603 | ktime_t kt = skb->tstamp; | |
604 | ||
605 | if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { | |
606 | struct timeval tv; | |
607 | /* Race occurred between timestamp enabling and packet | |
608 | receiving. Fill in the current time for now. */ | |
609 | if (kt.tv64 == 0) | |
610 | kt = ktime_get_real(); | |
611 | skb->tstamp = kt; | |
612 | tv = ktime_to_timeval(kt); | |
613 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv); | |
614 | } else { | |
615 | struct timespec ts; | |
616 | /* Race occurred between timestamp enabling and packet | |
617 | receiving. Fill in the current time for now. */ | |
618 | if (kt.tv64 == 0) | |
619 | kt = ktime_get_real(); | |
620 | skb->tstamp = kt; | |
621 | ts = ktime_to_timespec(kt); | |
622 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts); | |
623 | } | |
624 | } | |
625 | ||
626 | EXPORT_SYMBOL_GPL(__sock_recv_timestamp); | |
627 | ||
628 | static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock, | |
629 | struct msghdr *msg, size_t size, int flags) | |
630 | { | |
631 | int err; | |
632 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
633 | ||
634 | si->sock = sock; | |
635 | si->scm = NULL; | |
636 | si->msg = msg; | |
637 | si->size = size; | |
638 | si->flags = flags; | |
639 | ||
640 | err = security_socket_recvmsg(sock, msg, size, flags); | |
641 | if (err) | |
642 | return err; | |
643 | ||
644 | return sock->ops->recvmsg(iocb, sock, msg, size, flags); | |
645 | } | |
646 | ||
647 | int sock_recvmsg(struct socket *sock, struct msghdr *msg, | |
648 | size_t size, int flags) | |
649 | { | |
650 | struct kiocb iocb; | |
651 | struct sock_iocb siocb; | |
652 | int ret; | |
653 | ||
654 | init_sync_kiocb(&iocb, NULL); | |
655 | iocb.private = &siocb; | |
656 | ret = __sock_recvmsg(&iocb, sock, msg, size, flags); | |
657 | if (-EIOCBQUEUED == ret) | |
658 | ret = wait_on_sync_kiocb(&iocb); | |
659 | return ret; | |
660 | } | |
661 | ||
662 | int kernel_recvmsg(struct socket *sock, struct msghdr *msg, | |
663 | struct kvec *vec, size_t num, size_t size, int flags) | |
664 | { | |
665 | mm_segment_t oldfs = get_fs(); | |
666 | int result; | |
667 | ||
668 | set_fs(KERNEL_DS); | |
669 | /* | |
670 | * the following is safe, since for compiler definitions of kvec and | |
671 | * iovec are identical, yielding the same in-core layout and alignment | |
672 | */ | |
673 | msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num; | |
674 | result = sock_recvmsg(sock, msg, size, flags); | |
675 | set_fs(oldfs); | |
676 | return result; | |
677 | } | |
678 | ||
679 | static void sock_aio_dtor(struct kiocb *iocb) | |
680 | { | |
681 | kfree(iocb->private); | |
682 | } | |
683 | ||
684 | static ssize_t sock_sendpage(struct file *file, struct page *page, | |
685 | int offset, size_t size, loff_t *ppos, int more) | |
686 | { | |
687 | struct socket *sock; | |
688 | int flags; | |
689 | ||
690 | sock = file->private_data; | |
691 | ||
692 | flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT; | |
693 | if (more) | |
694 | flags |= MSG_MORE; | |
695 | ||
696 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
697 | } | |
698 | ||
699 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, | |
700 | struct pipe_inode_info *pipe, size_t len, | |
701 | unsigned int flags) | |
702 | { | |
703 | struct socket *sock = file->private_data; | |
704 | ||
705 | if (unlikely(!sock->ops->splice_read)) | |
706 | return -EINVAL; | |
707 | ||
708 | return sock->ops->splice_read(sock, ppos, pipe, len, flags); | |
709 | } | |
710 | ||
711 | static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb, | |
712 | struct sock_iocb *siocb) | |
713 | { | |
714 | if (!is_sync_kiocb(iocb)) { | |
715 | siocb = kmalloc(sizeof(*siocb), GFP_KERNEL); | |
716 | if (!siocb) | |
717 | return NULL; | |
718 | iocb->ki_dtor = sock_aio_dtor; | |
719 | } | |
720 | ||
721 | siocb->kiocb = iocb; | |
722 | iocb->private = siocb; | |
723 | return siocb; | |
724 | } | |
725 | ||
726 | static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb, | |
727 | struct file *file, const struct iovec *iov, | |
728 | unsigned long nr_segs) | |
729 | { | |
730 | struct socket *sock = file->private_data; | |
731 | size_t size = 0; | |
732 | int i; | |
733 | ||
734 | for (i = 0; i < nr_segs; i++) | |
735 | size += iov[i].iov_len; | |
736 | ||
737 | msg->msg_name = NULL; | |
738 | msg->msg_namelen = 0; | |
739 | msg->msg_control = NULL; | |
740 | msg->msg_controllen = 0; | |
741 | msg->msg_iov = (struct iovec *)iov; | |
742 | msg->msg_iovlen = nr_segs; | |
743 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
744 | ||
745 | return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags); | |
746 | } | |
747 | ||
748 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, | |
749 | unsigned long nr_segs, loff_t pos) | |
750 | { | |
751 | struct sock_iocb siocb, *x; | |
752 | ||
753 | if (pos != 0) | |
754 | return -ESPIPE; | |
755 | ||
756 | if (iocb->ki_left == 0) /* Match SYS5 behaviour */ | |
757 | return 0; | |
758 | ||
759 | ||
760 | x = alloc_sock_iocb(iocb, &siocb); | |
761 | if (!x) | |
762 | return -ENOMEM; | |
763 | return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); | |
764 | } | |
765 | ||
766 | static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb, | |
767 | struct file *file, const struct iovec *iov, | |
768 | unsigned long nr_segs) | |
769 | { | |
770 | struct socket *sock = file->private_data; | |
771 | size_t size = 0; | |
772 | int i; | |
773 | ||
774 | for (i = 0; i < nr_segs; i++) | |
775 | size += iov[i].iov_len; | |
776 | ||
777 | msg->msg_name = NULL; | |
778 | msg->msg_namelen = 0; | |
779 | msg->msg_control = NULL; | |
780 | msg->msg_controllen = 0; | |
781 | msg->msg_iov = (struct iovec *)iov; | |
782 | msg->msg_iovlen = nr_segs; | |
783 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
784 | if (sock->type == SOCK_SEQPACKET) | |
785 | msg->msg_flags |= MSG_EOR; | |
786 | ||
787 | return __sock_sendmsg(iocb, sock, msg, size); | |
788 | } | |
789 | ||
790 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
791 | unsigned long nr_segs, loff_t pos) | |
792 | { | |
793 | struct sock_iocb siocb, *x; | |
794 | ||
795 | if (pos != 0) | |
796 | return -ESPIPE; | |
797 | ||
798 | x = alloc_sock_iocb(iocb, &siocb); | |
799 | if (!x) | |
800 | return -ENOMEM; | |
801 | ||
802 | return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); | |
803 | } | |
804 | ||
805 | /* | |
806 | * Atomic setting of ioctl hooks to avoid race | |
807 | * with module unload. | |
808 | */ | |
809 | ||
810 | static DEFINE_MUTEX(br_ioctl_mutex); | |
811 | static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL; | |
812 | ||
813 | void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) | |
814 | { | |
815 | mutex_lock(&br_ioctl_mutex); | |
816 | br_ioctl_hook = hook; | |
817 | mutex_unlock(&br_ioctl_mutex); | |
818 | } | |
819 | ||
820 | EXPORT_SYMBOL(brioctl_set); | |
821 | ||
822 | static DEFINE_MUTEX(vlan_ioctl_mutex); | |
823 | static int (*vlan_ioctl_hook) (struct net *, void __user *arg); | |
824 | ||
825 | void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) | |
826 | { | |
827 | mutex_lock(&vlan_ioctl_mutex); | |
828 | vlan_ioctl_hook = hook; | |
829 | mutex_unlock(&vlan_ioctl_mutex); | |
830 | } | |
831 | ||
832 | EXPORT_SYMBOL(vlan_ioctl_set); | |
833 | ||
834 | static DEFINE_MUTEX(dlci_ioctl_mutex); | |
835 | static int (*dlci_ioctl_hook) (unsigned int, void __user *); | |
836 | ||
837 | void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) | |
838 | { | |
839 | mutex_lock(&dlci_ioctl_mutex); | |
840 | dlci_ioctl_hook = hook; | |
841 | mutex_unlock(&dlci_ioctl_mutex); | |
842 | } | |
843 | ||
844 | EXPORT_SYMBOL(dlci_ioctl_set); | |
845 | ||
846 | /* | |
847 | * With an ioctl, arg may well be a user mode pointer, but we don't know | |
848 | * what to do with it - that's up to the protocol still. | |
849 | */ | |
850 | ||
851 | static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) | |
852 | { | |
853 | struct socket *sock; | |
854 | struct sock *sk; | |
855 | void __user *argp = (void __user *)arg; | |
856 | int pid, err; | |
857 | struct net *net; | |
858 | ||
859 | sock = file->private_data; | |
860 | sk = sock->sk; | |
861 | net = sock_net(sk); | |
862 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) { | |
863 | err = dev_ioctl(net, cmd, argp); | |
864 | } else | |
865 | #ifdef CONFIG_WIRELESS_EXT | |
866 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { | |
867 | err = dev_ioctl(net, cmd, argp); | |
868 | } else | |
869 | #endif /* CONFIG_WIRELESS_EXT */ | |
870 | switch (cmd) { | |
871 | case FIOSETOWN: | |
872 | case SIOCSPGRP: | |
873 | err = -EFAULT; | |
874 | if (get_user(pid, (int __user *)argp)) | |
875 | break; | |
876 | err = f_setown(sock->file, pid, 1); | |
877 | break; | |
878 | case FIOGETOWN: | |
879 | case SIOCGPGRP: | |
880 | err = put_user(f_getown(sock->file), | |
881 | (int __user *)argp); | |
882 | break; | |
883 | case SIOCGIFBR: | |
884 | case SIOCSIFBR: | |
885 | case SIOCBRADDBR: | |
886 | case SIOCBRDELBR: | |
887 | err = -ENOPKG; | |
888 | if (!br_ioctl_hook) | |
889 | request_module("bridge"); | |
890 | ||
891 | mutex_lock(&br_ioctl_mutex); | |
892 | if (br_ioctl_hook) | |
893 | err = br_ioctl_hook(net, cmd, argp); | |
894 | mutex_unlock(&br_ioctl_mutex); | |
895 | break; | |
896 | case SIOCGIFVLAN: | |
897 | case SIOCSIFVLAN: | |
898 | err = -ENOPKG; | |
899 | if (!vlan_ioctl_hook) | |
900 | request_module("8021q"); | |
901 | ||
902 | mutex_lock(&vlan_ioctl_mutex); | |
903 | if (vlan_ioctl_hook) | |
904 | err = vlan_ioctl_hook(net, argp); | |
905 | mutex_unlock(&vlan_ioctl_mutex); | |
906 | break; | |
907 | case SIOCADDDLCI: | |
908 | case SIOCDELDLCI: | |
909 | err = -ENOPKG; | |
910 | if (!dlci_ioctl_hook) | |
911 | request_module("dlci"); | |
912 | ||
913 | mutex_lock(&dlci_ioctl_mutex); | |
914 | if (dlci_ioctl_hook) | |
915 | err = dlci_ioctl_hook(cmd, argp); | |
916 | mutex_unlock(&dlci_ioctl_mutex); | |
917 | break; | |
918 | default: | |
919 | err = sock->ops->ioctl(sock, cmd, arg); | |
920 | ||
921 | /* | |
922 | * If this ioctl is unknown try to hand it down | |
923 | * to the NIC driver. | |
924 | */ | |
925 | if (err == -ENOIOCTLCMD) | |
926 | err = dev_ioctl(net, cmd, argp); | |
927 | break; | |
928 | } | |
929 | return err; | |
930 | } | |
931 | ||
932 | int sock_create_lite(int family, int type, int protocol, struct socket **res) | |
933 | { | |
934 | int err; | |
935 | struct socket *sock = NULL; | |
936 | ||
937 | err = security_socket_create(family, type, protocol, 1); | |
938 | if (err) | |
939 | goto out; | |
940 | ||
941 | sock = sock_alloc(); | |
942 | if (!sock) { | |
943 | err = -ENOMEM; | |
944 | goto out; | |
945 | } | |
946 | ||
947 | sock->type = type; | |
948 | err = security_socket_post_create(sock, family, type, protocol, 1); | |
949 | if (err) | |
950 | goto out_release; | |
951 | ||
952 | out: | |
953 | *res = sock; | |
954 | return err; | |
955 | out_release: | |
956 | sock_release(sock); | |
957 | sock = NULL; | |
958 | goto out; | |
959 | } | |
960 | ||
961 | /* No kernel lock held - perfect */ | |
962 | static unsigned int sock_poll(struct file *file, poll_table *wait) | |
963 | { | |
964 | struct socket *sock; | |
965 | ||
966 | /* | |
967 | * We can't return errors to poll, so it's either yes or no. | |
968 | */ | |
969 | sock = file->private_data; | |
970 | return sock->ops->poll(file, sock, wait); | |
971 | } | |
972 | ||
973 | static int sock_mmap(struct file *file, struct vm_area_struct *vma) | |
974 | { | |
975 | struct socket *sock = file->private_data; | |
976 | ||
977 | return sock->ops->mmap(file, sock, vma); | |
978 | } | |
979 | ||
980 | static int sock_close(struct inode *inode, struct file *filp) | |
981 | { | |
982 | /* | |
983 | * It was possible the inode is NULL we were | |
984 | * closing an unfinished socket. | |
985 | */ | |
986 | ||
987 | if (!inode) { | |
988 | printk(KERN_DEBUG "sock_close: NULL inode\n"); | |
989 | return 0; | |
990 | } | |
991 | sock_fasync(-1, filp, 0); | |
992 | sock_release(SOCKET_I(inode)); | |
993 | return 0; | |
994 | } | |
995 | ||
996 | /* | |
997 | * Update the socket async list | |
998 | * | |
999 | * Fasync_list locking strategy. | |
1000 | * | |
1001 | * 1. fasync_list is modified only under process context socket lock | |
1002 | * i.e. under semaphore. | |
1003 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) | |
1004 | * or under socket lock. | |
1005 | * 3. fasync_list can be used from softirq context, so that | |
1006 | * modification under socket lock have to be enhanced with | |
1007 | * write_lock_bh(&sk->sk_callback_lock). | |
1008 | * --ANK (990710) | |
1009 | */ | |
1010 | ||
1011 | static int sock_fasync(int fd, struct file *filp, int on) | |
1012 | { | |
1013 | struct fasync_struct *fa, *fna = NULL, **prev; | |
1014 | struct socket *sock; | |
1015 | struct sock *sk; | |
1016 | ||
1017 | if (on) { | |
1018 | fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL); | |
1019 | if (fna == NULL) | |
1020 | return -ENOMEM; | |
1021 | } | |
1022 | ||
1023 | sock = filp->private_data; | |
1024 | ||
1025 | sk = sock->sk; | |
1026 | if (sk == NULL) { | |
1027 | kfree(fna); | |
1028 | return -EINVAL; | |
1029 | } | |
1030 | ||
1031 | lock_sock(sk); | |
1032 | ||
1033 | prev = &(sock->fasync_list); | |
1034 | ||
1035 | for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev) | |
1036 | if (fa->fa_file == filp) | |
1037 | break; | |
1038 | ||
1039 | if (on) { | |
1040 | if (fa != NULL) { | |
1041 | write_lock_bh(&sk->sk_callback_lock); | |
1042 | fa->fa_fd = fd; | |
1043 | write_unlock_bh(&sk->sk_callback_lock); | |
1044 | ||
1045 | kfree(fna); | |
1046 | goto out; | |
1047 | } | |
1048 | fna->fa_file = filp; | |
1049 | fna->fa_fd = fd; | |
1050 | fna->magic = FASYNC_MAGIC; | |
1051 | fna->fa_next = sock->fasync_list; | |
1052 | write_lock_bh(&sk->sk_callback_lock); | |
1053 | sock->fasync_list = fna; | |
1054 | write_unlock_bh(&sk->sk_callback_lock); | |
1055 | } else { | |
1056 | if (fa != NULL) { | |
1057 | write_lock_bh(&sk->sk_callback_lock); | |
1058 | *prev = fa->fa_next; | |
1059 | write_unlock_bh(&sk->sk_callback_lock); | |
1060 | kfree(fa); | |
1061 | } | |
1062 | } | |
1063 | ||
1064 | out: | |
1065 | release_sock(sock->sk); | |
1066 | return 0; | |
1067 | } | |
1068 | ||
1069 | /* This function may be called only under socket lock or callback_lock */ | |
1070 | ||
1071 | int sock_wake_async(struct socket *sock, int how, int band) | |
1072 | { | |
1073 | if (!sock || !sock->fasync_list) | |
1074 | return -1; | |
1075 | switch (how) { | |
1076 | case SOCK_WAKE_WAITD: | |
1077 | if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags)) | |
1078 | break; | |
1079 | goto call_kill; | |
1080 | case SOCK_WAKE_SPACE: | |
1081 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags)) | |
1082 | break; | |
1083 | /* fall through */ | |
1084 | case SOCK_WAKE_IO: | |
1085 | call_kill: | |
1086 | __kill_fasync(sock->fasync_list, SIGIO, band); | |
1087 | break; | |
1088 | case SOCK_WAKE_URG: | |
1089 | __kill_fasync(sock->fasync_list, SIGURG, band); | |
1090 | } | |
1091 | return 0; | |
1092 | } | |
1093 | ||
1094 | static int __sock_create(struct net *net, int family, int type, int protocol, | |
1095 | struct socket **res, int kern) | |
1096 | { | |
1097 | int err; | |
1098 | struct socket *sock; | |
1099 | const struct net_proto_family *pf; | |
1100 | ||
1101 | /* | |
1102 | * Check protocol is in range | |
1103 | */ | |
1104 | if (family < 0 || family >= NPROTO) | |
1105 | return -EAFNOSUPPORT; | |
1106 | if (type < 0 || type >= SOCK_MAX) | |
1107 | return -EINVAL; | |
1108 | ||
1109 | /* Compatibility. | |
1110 | ||
1111 | This uglymoron is moved from INET layer to here to avoid | |
1112 | deadlock in module load. | |
1113 | */ | |
1114 | if (family == PF_INET && type == SOCK_PACKET) { | |
1115 | static int warned; | |
1116 | if (!warned) { | |
1117 | warned = 1; | |
1118 | printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", | |
1119 | current->comm); | |
1120 | } | |
1121 | family = PF_PACKET; | |
1122 | } | |
1123 | ||
1124 | err = security_socket_create(family, type, protocol, kern); | |
1125 | if (err) | |
1126 | return err; | |
1127 | ||
1128 | /* | |
1129 | * Allocate the socket and allow the family to set things up. if | |
1130 | * the protocol is 0, the family is instructed to select an appropriate | |
1131 | * default. | |
1132 | */ | |
1133 | sock = sock_alloc(); | |
1134 | if (!sock) { | |
1135 | if (net_ratelimit()) | |
1136 | printk(KERN_WARNING "socket: no more sockets\n"); | |
1137 | return -ENFILE; /* Not exactly a match, but its the | |
1138 | closest posix thing */ | |
1139 | } | |
1140 | ||
1141 | sock->type = type; | |
1142 | ||
1143 | #if defined(CONFIG_KMOD) | |
1144 | /* Attempt to load a protocol module if the find failed. | |
1145 | * | |
1146 | * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user | |
1147 | * requested real, full-featured networking support upon configuration. | |
1148 | * Otherwise module support will break! | |
1149 | */ | |
1150 | if (net_families[family] == NULL) | |
1151 | request_module("net-pf-%d", family); | |
1152 | #endif | |
1153 | ||
1154 | rcu_read_lock(); | |
1155 | pf = rcu_dereference(net_families[family]); | |
1156 | err = -EAFNOSUPPORT; | |
1157 | if (!pf) | |
1158 | goto out_release; | |
1159 | ||
1160 | /* | |
1161 | * We will call the ->create function, that possibly is in a loadable | |
1162 | * module, so we have to bump that loadable module refcnt first. | |
1163 | */ | |
1164 | if (!try_module_get(pf->owner)) | |
1165 | goto out_release; | |
1166 | ||
1167 | /* Now protected by module ref count */ | |
1168 | rcu_read_unlock(); | |
1169 | ||
1170 | err = pf->create(net, sock, protocol); | |
1171 | if (err < 0) | |
1172 | goto out_module_put; | |
1173 | ||
1174 | /* | |
1175 | * Now to bump the refcnt of the [loadable] module that owns this | |
1176 | * socket at sock_release time we decrement its refcnt. | |
1177 | */ | |
1178 | if (!try_module_get(sock->ops->owner)) | |
1179 | goto out_module_busy; | |
1180 | ||
1181 | /* | |
1182 | * Now that we're done with the ->create function, the [loadable] | |
1183 | * module can have its refcnt decremented | |
1184 | */ | |
1185 | module_put(pf->owner); | |
1186 | err = security_socket_post_create(sock, family, type, protocol, kern); | |
1187 | if (err) | |
1188 | goto out_sock_release; | |
1189 | *res = sock; | |
1190 | ||
1191 | return 0; | |
1192 | ||
1193 | out_module_busy: | |
1194 | err = -EAFNOSUPPORT; | |
1195 | out_module_put: | |
1196 | sock->ops = NULL; | |
1197 | module_put(pf->owner); | |
1198 | out_sock_release: | |
1199 | sock_release(sock); | |
1200 | return err; | |
1201 | ||
1202 | out_release: | |
1203 | rcu_read_unlock(); | |
1204 | goto out_sock_release; | |
1205 | } | |
1206 | ||
1207 | int sock_create(int family, int type, int protocol, struct socket **res) | |
1208 | { | |
1209 | return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); | |
1210 | } | |
1211 | ||
1212 | int sock_create_kern(int family, int type, int protocol, struct socket **res) | |
1213 | { | |
1214 | return __sock_create(&init_net, family, type, protocol, res, 1); | |
1215 | } | |
1216 | ||
1217 | asmlinkage long sys_socket(int family, int type, int protocol) | |
1218 | { | |
1219 | int retval; | |
1220 | struct socket *sock; | |
1221 | int flags; | |
1222 | ||
1223 | flags = type & ~SOCK_TYPE_MASK; | |
1224 | if (flags & ~SOCK_CLOEXEC) | |
1225 | return -EINVAL; | |
1226 | type &= SOCK_TYPE_MASK; | |
1227 | ||
1228 | retval = sock_create(family, type, protocol, &sock); | |
1229 | if (retval < 0) | |
1230 | goto out; | |
1231 | ||
1232 | retval = sock_map_fd(sock, flags & O_CLOEXEC); | |
1233 | if (retval < 0) | |
1234 | goto out_release; | |
1235 | ||
1236 | out: | |
1237 | /* It may be already another descriptor 8) Not kernel problem. */ | |
1238 | return retval; | |
1239 | ||
1240 | out_release: | |
1241 | sock_release(sock); | |
1242 | return retval; | |
1243 | } | |
1244 | ||
1245 | /* | |
1246 | * Create a pair of connected sockets. | |
1247 | */ | |
1248 | ||
1249 | asmlinkage long sys_socketpair(int family, int type, int protocol, | |
1250 | int __user *usockvec) | |
1251 | { | |
1252 | struct socket *sock1, *sock2; | |
1253 | int fd1, fd2, err; | |
1254 | struct file *newfile1, *newfile2; | |
1255 | int flags; | |
1256 | ||
1257 | flags = type & ~SOCK_TYPE_MASK; | |
1258 | if (flags & ~SOCK_CLOEXEC) | |
1259 | return -EINVAL; | |
1260 | type &= SOCK_TYPE_MASK; | |
1261 | ||
1262 | /* | |
1263 | * Obtain the first socket and check if the underlying protocol | |
1264 | * supports the socketpair call. | |
1265 | */ | |
1266 | ||
1267 | err = sock_create(family, type, protocol, &sock1); | |
1268 | if (err < 0) | |
1269 | goto out; | |
1270 | ||
1271 | err = sock_create(family, type, protocol, &sock2); | |
1272 | if (err < 0) | |
1273 | goto out_release_1; | |
1274 | ||
1275 | err = sock1->ops->socketpair(sock1, sock2); | |
1276 | if (err < 0) | |
1277 | goto out_release_both; | |
1278 | ||
1279 | fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC); | |
1280 | if (unlikely(fd1 < 0)) { | |
1281 | err = fd1; | |
1282 | goto out_release_both; | |
1283 | } | |
1284 | ||
1285 | fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC); | |
1286 | if (unlikely(fd2 < 0)) { | |
1287 | err = fd2; | |
1288 | put_filp(newfile1); | |
1289 | put_unused_fd(fd1); | |
1290 | goto out_release_both; | |
1291 | } | |
1292 | ||
1293 | err = sock_attach_fd(sock1, newfile1); | |
1294 | if (unlikely(err < 0)) { | |
1295 | goto out_fd2; | |
1296 | } | |
1297 | ||
1298 | err = sock_attach_fd(sock2, newfile2); | |
1299 | if (unlikely(err < 0)) { | |
1300 | fput(newfile1); | |
1301 | goto out_fd1; | |
1302 | } | |
1303 | ||
1304 | err = audit_fd_pair(fd1, fd2); | |
1305 | if (err < 0) { | |
1306 | fput(newfile1); | |
1307 | fput(newfile2); | |
1308 | goto out_fd; | |
1309 | } | |
1310 | ||
1311 | fd_install(fd1, newfile1); | |
1312 | fd_install(fd2, newfile2); | |
1313 | /* fd1 and fd2 may be already another descriptors. | |
1314 | * Not kernel problem. | |
1315 | */ | |
1316 | ||
1317 | err = put_user(fd1, &usockvec[0]); | |
1318 | if (!err) | |
1319 | err = put_user(fd2, &usockvec[1]); | |
1320 | if (!err) | |
1321 | return 0; | |
1322 | ||
1323 | sys_close(fd2); | |
1324 | sys_close(fd1); | |
1325 | return err; | |
1326 | ||
1327 | out_release_both: | |
1328 | sock_release(sock2); | |
1329 | out_release_1: | |
1330 | sock_release(sock1); | |
1331 | out: | |
1332 | return err; | |
1333 | ||
1334 | out_fd2: | |
1335 | put_filp(newfile1); | |
1336 | sock_release(sock1); | |
1337 | out_fd1: | |
1338 | put_filp(newfile2); | |
1339 | sock_release(sock2); | |
1340 | out_fd: | |
1341 | put_unused_fd(fd1); | |
1342 | put_unused_fd(fd2); | |
1343 | goto out; | |
1344 | } | |
1345 | ||
1346 | /* | |
1347 | * Bind a name to a socket. Nothing much to do here since it's | |
1348 | * the protocol's responsibility to handle the local address. | |
1349 | * | |
1350 | * We move the socket address to kernel space before we call | |
1351 | * the protocol layer (having also checked the address is ok). | |
1352 | */ | |
1353 | ||
1354 | asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) | |
1355 | { | |
1356 | struct socket *sock; | |
1357 | struct sockaddr_storage address; | |
1358 | int err, fput_needed; | |
1359 | ||
1360 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1361 | if (sock) { | |
1362 | err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address); | |
1363 | if (err >= 0) { | |
1364 | err = security_socket_bind(sock, | |
1365 | (struct sockaddr *)&address, | |
1366 | addrlen); | |
1367 | if (!err) | |
1368 | err = sock->ops->bind(sock, | |
1369 | (struct sockaddr *) | |
1370 | &address, addrlen); | |
1371 | } | |
1372 | fput_light(sock->file, fput_needed); | |
1373 | } | |
1374 | return err; | |
1375 | } | |
1376 | ||
1377 | /* | |
1378 | * Perform a listen. Basically, we allow the protocol to do anything | |
1379 | * necessary for a listen, and if that works, we mark the socket as | |
1380 | * ready for listening. | |
1381 | */ | |
1382 | ||
1383 | asmlinkage long sys_listen(int fd, int backlog) | |
1384 | { | |
1385 | struct socket *sock; | |
1386 | int err, fput_needed; | |
1387 | int somaxconn; | |
1388 | ||
1389 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1390 | if (sock) { | |
1391 | somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; | |
1392 | if ((unsigned)backlog > somaxconn) | |
1393 | backlog = somaxconn; | |
1394 | ||
1395 | err = security_socket_listen(sock, backlog); | |
1396 | if (!err) | |
1397 | err = sock->ops->listen(sock, backlog); | |
1398 | ||
1399 | fput_light(sock->file, fput_needed); | |
1400 | } | |
1401 | return err; | |
1402 | } | |
1403 | ||
1404 | /* | |
1405 | * For accept, we attempt to create a new socket, set up the link | |
1406 | * with the client, wake up the client, then return the new | |
1407 | * connected fd. We collect the address of the connector in kernel | |
1408 | * space and move it to user at the very end. This is unclean because | |
1409 | * we open the socket then return an error. | |
1410 | * | |
1411 | * 1003.1g adds the ability to recvmsg() to query connection pending | |
1412 | * status to recvmsg. We need to add that support in a way thats | |
1413 | * clean when we restucture accept also. | |
1414 | */ | |
1415 | ||
1416 | asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, | |
1417 | int __user *upeer_addrlen) | |
1418 | { | |
1419 | struct socket *sock, *newsock; | |
1420 | struct file *newfile; | |
1421 | int err, len, newfd, fput_needed; | |
1422 | struct sockaddr_storage address; | |
1423 | ||
1424 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1425 | if (!sock) | |
1426 | goto out; | |
1427 | ||
1428 | err = -ENFILE; | |
1429 | if (!(newsock = sock_alloc())) | |
1430 | goto out_put; | |
1431 | ||
1432 | newsock->type = sock->type; | |
1433 | newsock->ops = sock->ops; | |
1434 | ||
1435 | /* | |
1436 | * We don't need try_module_get here, as the listening socket (sock) | |
1437 | * has the protocol module (sock->ops->owner) held. | |
1438 | */ | |
1439 | __module_get(newsock->ops->owner); | |
1440 | ||
1441 | newfd = sock_alloc_fd(&newfile, 0); | |
1442 | if (unlikely(newfd < 0)) { | |
1443 | err = newfd; | |
1444 | sock_release(newsock); | |
1445 | goto out_put; | |
1446 | } | |
1447 | ||
1448 | err = sock_attach_fd(newsock, newfile); | |
1449 | if (err < 0) | |
1450 | goto out_fd_simple; | |
1451 | ||
1452 | err = security_socket_accept(sock, newsock); | |
1453 | if (err) | |
1454 | goto out_fd; | |
1455 | ||
1456 | err = sock->ops->accept(sock, newsock, sock->file->f_flags); | |
1457 | if (err < 0) | |
1458 | goto out_fd; | |
1459 | ||
1460 | if (upeer_sockaddr) { | |
1461 | if (newsock->ops->getname(newsock, (struct sockaddr *)&address, | |
1462 | &len, 2) < 0) { | |
1463 | err = -ECONNABORTED; | |
1464 | goto out_fd; | |
1465 | } | |
1466 | err = move_addr_to_user((struct sockaddr *)&address, | |
1467 | len, upeer_sockaddr, upeer_addrlen); | |
1468 | if (err < 0) | |
1469 | goto out_fd; | |
1470 | } | |
1471 | ||
1472 | /* File flags are not inherited via accept() unlike another OSes. */ | |
1473 | ||
1474 | fd_install(newfd, newfile); | |
1475 | err = newfd; | |
1476 | ||
1477 | security_socket_post_accept(sock, newsock); | |
1478 | ||
1479 | out_put: | |
1480 | fput_light(sock->file, fput_needed); | |
1481 | out: | |
1482 | return err; | |
1483 | out_fd_simple: | |
1484 | sock_release(newsock); | |
1485 | put_filp(newfile); | |
1486 | put_unused_fd(newfd); | |
1487 | goto out_put; | |
1488 | out_fd: | |
1489 | fput(newfile); | |
1490 | put_unused_fd(newfd); | |
1491 | goto out_put; | |
1492 | } | |
1493 | ||
1494 | /* | |
1495 | * Attempt to connect to a socket with the server address. The address | |
1496 | * is in user space so we verify it is OK and move it to kernel space. | |
1497 | * | |
1498 | * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to | |
1499 | * break bindings | |
1500 | * | |
1501 | * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and | |
1502 | * other SEQPACKET protocols that take time to connect() as it doesn't | |
1503 | * include the -EINPROGRESS status for such sockets. | |
1504 | */ | |
1505 | ||
1506 | asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, | |
1507 | int addrlen) | |
1508 | { | |
1509 | struct socket *sock; | |
1510 | struct sockaddr_storage address; | |
1511 | int err, fput_needed; | |
1512 | ||
1513 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1514 | if (!sock) | |
1515 | goto out; | |
1516 | err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address); | |
1517 | if (err < 0) | |
1518 | goto out_put; | |
1519 | ||
1520 | err = | |
1521 | security_socket_connect(sock, (struct sockaddr *)&address, addrlen); | |
1522 | if (err) | |
1523 | goto out_put; | |
1524 | ||
1525 | err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, | |
1526 | sock->file->f_flags); | |
1527 | out_put: | |
1528 | fput_light(sock->file, fput_needed); | |
1529 | out: | |
1530 | return err; | |
1531 | } | |
1532 | ||
1533 | /* | |
1534 | * Get the local address ('name') of a socket object. Move the obtained | |
1535 | * name to user space. | |
1536 | */ | |
1537 | ||
1538 | asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, | |
1539 | int __user *usockaddr_len) | |
1540 | { | |
1541 | struct socket *sock; | |
1542 | struct sockaddr_storage address; | |
1543 | int len, err, fput_needed; | |
1544 | ||
1545 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1546 | if (!sock) | |
1547 | goto out; | |
1548 | ||
1549 | err = security_socket_getsockname(sock); | |
1550 | if (err) | |
1551 | goto out_put; | |
1552 | ||
1553 | err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0); | |
1554 | if (err) | |
1555 | goto out_put; | |
1556 | err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len); | |
1557 | ||
1558 | out_put: | |
1559 | fput_light(sock->file, fput_needed); | |
1560 | out: | |
1561 | return err; | |
1562 | } | |
1563 | ||
1564 | /* | |
1565 | * Get the remote address ('name') of a socket object. Move the obtained | |
1566 | * name to user space. | |
1567 | */ | |
1568 | ||
1569 | asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, | |
1570 | int __user *usockaddr_len) | |
1571 | { | |
1572 | struct socket *sock; | |
1573 | struct sockaddr_storage address; | |
1574 | int len, err, fput_needed; | |
1575 | ||
1576 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1577 | if (sock != NULL) { | |
1578 | err = security_socket_getpeername(sock); | |
1579 | if (err) { | |
1580 | fput_light(sock->file, fput_needed); | |
1581 | return err; | |
1582 | } | |
1583 | ||
1584 | err = | |
1585 | sock->ops->getname(sock, (struct sockaddr *)&address, &len, | |
1586 | 1); | |
1587 | if (!err) | |
1588 | err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, | |
1589 | usockaddr_len); | |
1590 | fput_light(sock->file, fput_needed); | |
1591 | } | |
1592 | return err; | |
1593 | } | |
1594 | ||
1595 | /* | |
1596 | * Send a datagram to a given address. We move the address into kernel | |
1597 | * space and check the user space data area is readable before invoking | |
1598 | * the protocol. | |
1599 | */ | |
1600 | ||
1601 | asmlinkage long sys_sendto(int fd, void __user *buff, size_t len, | |
1602 | unsigned flags, struct sockaddr __user *addr, | |
1603 | int addr_len) | |
1604 | { | |
1605 | struct socket *sock; | |
1606 | struct sockaddr_storage address; | |
1607 | int err; | |
1608 | struct msghdr msg; | |
1609 | struct iovec iov; | |
1610 | int fput_needed; | |
1611 | ||
1612 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1613 | if (!sock) | |
1614 | goto out; | |
1615 | ||
1616 | iov.iov_base = buff; | |
1617 | iov.iov_len = len; | |
1618 | msg.msg_name = NULL; | |
1619 | msg.msg_iov = &iov; | |
1620 | msg.msg_iovlen = 1; | |
1621 | msg.msg_control = NULL; | |
1622 | msg.msg_controllen = 0; | |
1623 | msg.msg_namelen = 0; | |
1624 | if (addr) { | |
1625 | err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address); | |
1626 | if (err < 0) | |
1627 | goto out_put; | |
1628 | msg.msg_name = (struct sockaddr *)&address; | |
1629 | msg.msg_namelen = addr_len; | |
1630 | } | |
1631 | if (sock->file->f_flags & O_NONBLOCK) | |
1632 | flags |= MSG_DONTWAIT; | |
1633 | msg.msg_flags = flags; | |
1634 | err = sock_sendmsg(sock, &msg, len); | |
1635 | ||
1636 | out_put: | |
1637 | fput_light(sock->file, fput_needed); | |
1638 | out: | |
1639 | return err; | |
1640 | } | |
1641 | ||
1642 | /* | |
1643 | * Send a datagram down a socket. | |
1644 | */ | |
1645 | ||
1646 | asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags) | |
1647 | { | |
1648 | return sys_sendto(fd, buff, len, flags, NULL, 0); | |
1649 | } | |
1650 | ||
1651 | /* | |
1652 | * Receive a frame from the socket and optionally record the address of the | |
1653 | * sender. We verify the buffers are writable and if needed move the | |
1654 | * sender address from kernel to user space. | |
1655 | */ | |
1656 | ||
1657 | asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size, | |
1658 | unsigned flags, struct sockaddr __user *addr, | |
1659 | int __user *addr_len) | |
1660 | { | |
1661 | struct socket *sock; | |
1662 | struct iovec iov; | |
1663 | struct msghdr msg; | |
1664 | struct sockaddr_storage address; | |
1665 | int err, err2; | |
1666 | int fput_needed; | |
1667 | ||
1668 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1669 | if (!sock) | |
1670 | goto out; | |
1671 | ||
1672 | msg.msg_control = NULL; | |
1673 | msg.msg_controllen = 0; | |
1674 | msg.msg_iovlen = 1; | |
1675 | msg.msg_iov = &iov; | |
1676 | iov.iov_len = size; | |
1677 | iov.iov_base = ubuf; | |
1678 | msg.msg_name = (struct sockaddr *)&address; | |
1679 | msg.msg_namelen = sizeof(address); | |
1680 | if (sock->file->f_flags & O_NONBLOCK) | |
1681 | flags |= MSG_DONTWAIT; | |
1682 | err = sock_recvmsg(sock, &msg, size, flags); | |
1683 | ||
1684 | if (err >= 0 && addr != NULL) { | |
1685 | err2 = move_addr_to_user((struct sockaddr *)&address, | |
1686 | msg.msg_namelen, addr, addr_len); | |
1687 | if (err2 < 0) | |
1688 | err = err2; | |
1689 | } | |
1690 | ||
1691 | fput_light(sock->file, fput_needed); | |
1692 | out: | |
1693 | return err; | |
1694 | } | |
1695 | ||
1696 | /* | |
1697 | * Receive a datagram from a socket. | |
1698 | */ | |
1699 | ||
1700 | asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size, | |
1701 | unsigned flags) | |
1702 | { | |
1703 | return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); | |
1704 | } | |
1705 | ||
1706 | /* | |
1707 | * Set a socket option. Because we don't know the option lengths we have | |
1708 | * to pass the user mode parameter for the protocols to sort out. | |
1709 | */ | |
1710 | ||
1711 | asmlinkage long sys_setsockopt(int fd, int level, int optname, | |
1712 | char __user *optval, int optlen) | |
1713 | { | |
1714 | int err, fput_needed; | |
1715 | struct socket *sock; | |
1716 | ||
1717 | if (optlen < 0) | |
1718 | return -EINVAL; | |
1719 | ||
1720 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1721 | if (sock != NULL) { | |
1722 | err = security_socket_setsockopt(sock, level, optname); | |
1723 | if (err) | |
1724 | goto out_put; | |
1725 | ||
1726 | if (level == SOL_SOCKET) | |
1727 | err = | |
1728 | sock_setsockopt(sock, level, optname, optval, | |
1729 | optlen); | |
1730 | else | |
1731 | err = | |
1732 | sock->ops->setsockopt(sock, level, optname, optval, | |
1733 | optlen); | |
1734 | out_put: | |
1735 | fput_light(sock->file, fput_needed); | |
1736 | } | |
1737 | return err; | |
1738 | } | |
1739 | ||
1740 | /* | |
1741 | * Get a socket option. Because we don't know the option lengths we have | |
1742 | * to pass a user mode parameter for the protocols to sort out. | |
1743 | */ | |
1744 | ||
1745 | asmlinkage long sys_getsockopt(int fd, int level, int optname, | |
1746 | char __user *optval, int __user *optlen) | |
1747 | { | |
1748 | int err, fput_needed; | |
1749 | struct socket *sock; | |
1750 | ||
1751 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1752 | if (sock != NULL) { | |
1753 | err = security_socket_getsockopt(sock, level, optname); | |
1754 | if (err) | |
1755 | goto out_put; | |
1756 | ||
1757 | if (level == SOL_SOCKET) | |
1758 | err = | |
1759 | sock_getsockopt(sock, level, optname, optval, | |
1760 | optlen); | |
1761 | else | |
1762 | err = | |
1763 | sock->ops->getsockopt(sock, level, optname, optval, | |
1764 | optlen); | |
1765 | out_put: | |
1766 | fput_light(sock->file, fput_needed); | |
1767 | } | |
1768 | return err; | |
1769 | } | |
1770 | ||
1771 | /* | |
1772 | * Shutdown a socket. | |
1773 | */ | |
1774 | ||
1775 | asmlinkage long sys_shutdown(int fd, int how) | |
1776 | { | |
1777 | int err, fput_needed; | |
1778 | struct socket *sock; | |
1779 | ||
1780 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1781 | if (sock != NULL) { | |
1782 | err = security_socket_shutdown(sock, how); | |
1783 | if (!err) | |
1784 | err = sock->ops->shutdown(sock, how); | |
1785 | fput_light(sock->file, fput_needed); | |
1786 | } | |
1787 | return err; | |
1788 | } | |
1789 | ||
1790 | /* A couple of helpful macros for getting the address of the 32/64 bit | |
1791 | * fields which are the same type (int / unsigned) on our platforms. | |
1792 | */ | |
1793 | #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) | |
1794 | #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) | |
1795 | #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) | |
1796 | ||
1797 | /* | |
1798 | * BSD sendmsg interface | |
1799 | */ | |
1800 | ||
1801 | asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags) | |
1802 | { | |
1803 | struct compat_msghdr __user *msg_compat = | |
1804 | (struct compat_msghdr __user *)msg; | |
1805 | struct socket *sock; | |
1806 | struct sockaddr_storage address; | |
1807 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; | |
1808 | unsigned char ctl[sizeof(struct cmsghdr) + 20] | |
1809 | __attribute__ ((aligned(sizeof(__kernel_size_t)))); | |
1810 | /* 20 is size of ipv6_pktinfo */ | |
1811 | unsigned char *ctl_buf = ctl; | |
1812 | struct msghdr msg_sys; | |
1813 | int err, ctl_len, iov_size, total_len; | |
1814 | int fput_needed; | |
1815 | ||
1816 | err = -EFAULT; | |
1817 | if (MSG_CMSG_COMPAT & flags) { | |
1818 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1819 | return -EFAULT; | |
1820 | } | |
1821 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1822 | return -EFAULT; | |
1823 | ||
1824 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1825 | if (!sock) | |
1826 | goto out; | |
1827 | ||
1828 | /* do not move before msg_sys is valid */ | |
1829 | err = -EMSGSIZE; | |
1830 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
1831 | goto out_put; | |
1832 | ||
1833 | /* Check whether to allocate the iovec area */ | |
1834 | err = -ENOMEM; | |
1835 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
1836 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
1837 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
1838 | if (!iov) | |
1839 | goto out_put; | |
1840 | } | |
1841 | ||
1842 | /* This will also move the address data into kernel space */ | |
1843 | if (MSG_CMSG_COMPAT & flags) { | |
1844 | err = verify_compat_iovec(&msg_sys, iov, | |
1845 | (struct sockaddr *)&address, | |
1846 | VERIFY_READ); | |
1847 | } else | |
1848 | err = verify_iovec(&msg_sys, iov, | |
1849 | (struct sockaddr *)&address, | |
1850 | VERIFY_READ); | |
1851 | if (err < 0) | |
1852 | goto out_freeiov; | |
1853 | total_len = err; | |
1854 | ||
1855 | err = -ENOBUFS; | |
1856 | ||
1857 | if (msg_sys.msg_controllen > INT_MAX) | |
1858 | goto out_freeiov; | |
1859 | ctl_len = msg_sys.msg_controllen; | |
1860 | if ((MSG_CMSG_COMPAT & flags) && ctl_len) { | |
1861 | err = | |
1862 | cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, | |
1863 | sizeof(ctl)); | |
1864 | if (err) | |
1865 | goto out_freeiov; | |
1866 | ctl_buf = msg_sys.msg_control; | |
1867 | ctl_len = msg_sys.msg_controllen; | |
1868 | } else if (ctl_len) { | |
1869 | if (ctl_len > sizeof(ctl)) { | |
1870 | ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); | |
1871 | if (ctl_buf == NULL) | |
1872 | goto out_freeiov; | |
1873 | } | |
1874 | err = -EFAULT; | |
1875 | /* | |
1876 | * Careful! Before this, msg_sys.msg_control contains a user pointer. | |
1877 | * Afterwards, it will be a kernel pointer. Thus the compiler-assisted | |
1878 | * checking falls down on this. | |
1879 | */ | |
1880 | if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control, | |
1881 | ctl_len)) | |
1882 | goto out_freectl; | |
1883 | msg_sys.msg_control = ctl_buf; | |
1884 | } | |
1885 | msg_sys.msg_flags = flags; | |
1886 | ||
1887 | if (sock->file->f_flags & O_NONBLOCK) | |
1888 | msg_sys.msg_flags |= MSG_DONTWAIT; | |
1889 | err = sock_sendmsg(sock, &msg_sys, total_len); | |
1890 | ||
1891 | out_freectl: | |
1892 | if (ctl_buf != ctl) | |
1893 | sock_kfree_s(sock->sk, ctl_buf, ctl_len); | |
1894 | out_freeiov: | |
1895 | if (iov != iovstack) | |
1896 | sock_kfree_s(sock->sk, iov, iov_size); | |
1897 | out_put: | |
1898 | fput_light(sock->file, fput_needed); | |
1899 | out: | |
1900 | return err; | |
1901 | } | |
1902 | ||
1903 | /* | |
1904 | * BSD recvmsg interface | |
1905 | */ | |
1906 | ||
1907 | asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, | |
1908 | unsigned int flags) | |
1909 | { | |
1910 | struct compat_msghdr __user *msg_compat = | |
1911 | (struct compat_msghdr __user *)msg; | |
1912 | struct socket *sock; | |
1913 | struct iovec iovstack[UIO_FASTIOV]; | |
1914 | struct iovec *iov = iovstack; | |
1915 | struct msghdr msg_sys; | |
1916 | unsigned long cmsg_ptr; | |
1917 | int err, iov_size, total_len, len; | |
1918 | int fput_needed; | |
1919 | ||
1920 | /* kernel mode address */ | |
1921 | struct sockaddr_storage addr; | |
1922 | ||
1923 | /* user mode address pointers */ | |
1924 | struct sockaddr __user *uaddr; | |
1925 | int __user *uaddr_len; | |
1926 | ||
1927 | if (MSG_CMSG_COMPAT & flags) { | |
1928 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1929 | return -EFAULT; | |
1930 | } | |
1931 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1932 | return -EFAULT; | |
1933 | ||
1934 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1935 | if (!sock) | |
1936 | goto out; | |
1937 | ||
1938 | err = -EMSGSIZE; | |
1939 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
1940 | goto out_put; | |
1941 | ||
1942 | /* Check whether to allocate the iovec area */ | |
1943 | err = -ENOMEM; | |
1944 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
1945 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
1946 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
1947 | if (!iov) | |
1948 | goto out_put; | |
1949 | } | |
1950 | ||
1951 | /* | |
1952 | * Save the user-mode address (verify_iovec will change the | |
1953 | * kernel msghdr to use the kernel address space) | |
1954 | */ | |
1955 | ||
1956 | uaddr = (__force void __user *)msg_sys.msg_name; | |
1957 | uaddr_len = COMPAT_NAMELEN(msg); | |
1958 | if (MSG_CMSG_COMPAT & flags) { | |
1959 | err = verify_compat_iovec(&msg_sys, iov, | |
1960 | (struct sockaddr *)&addr, | |
1961 | VERIFY_WRITE); | |
1962 | } else | |
1963 | err = verify_iovec(&msg_sys, iov, | |
1964 | (struct sockaddr *)&addr, | |
1965 | VERIFY_WRITE); | |
1966 | if (err < 0) | |
1967 | goto out_freeiov; | |
1968 | total_len = err; | |
1969 | ||
1970 | cmsg_ptr = (unsigned long)msg_sys.msg_control; | |
1971 | msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); | |
1972 | ||
1973 | if (sock->file->f_flags & O_NONBLOCK) | |
1974 | flags |= MSG_DONTWAIT; | |
1975 | err = sock_recvmsg(sock, &msg_sys, total_len, flags); | |
1976 | if (err < 0) | |
1977 | goto out_freeiov; | |
1978 | len = err; | |
1979 | ||
1980 | if (uaddr != NULL) { | |
1981 | err = move_addr_to_user((struct sockaddr *)&addr, | |
1982 | msg_sys.msg_namelen, uaddr, | |
1983 | uaddr_len); | |
1984 | if (err < 0) | |
1985 | goto out_freeiov; | |
1986 | } | |
1987 | err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT), | |
1988 | COMPAT_FLAGS(msg)); | |
1989 | if (err) | |
1990 | goto out_freeiov; | |
1991 | if (MSG_CMSG_COMPAT & flags) | |
1992 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, | |
1993 | &msg_compat->msg_controllen); | |
1994 | else | |
1995 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, | |
1996 | &msg->msg_controllen); | |
1997 | if (err) | |
1998 | goto out_freeiov; | |
1999 | err = len; | |
2000 | ||
2001 | out_freeiov: | |
2002 | if (iov != iovstack) | |
2003 | sock_kfree_s(sock->sk, iov, iov_size); | |
2004 | out_put: | |
2005 | fput_light(sock->file, fput_needed); | |
2006 | out: | |
2007 | return err; | |
2008 | } | |
2009 | ||
2010 | #ifdef __ARCH_WANT_SYS_SOCKETCALL | |
2011 | ||
2012 | /* Argument list sizes for sys_socketcall */ | |
2013 | #define AL(x) ((x) * sizeof(unsigned long)) | |
2014 | static const unsigned char nargs[18]={ | |
2015 | AL(0),AL(3),AL(3),AL(3),AL(2),AL(3), | |
2016 | AL(3),AL(3),AL(4),AL(4),AL(4),AL(6), | |
2017 | AL(6),AL(2),AL(5),AL(5),AL(3),AL(3) | |
2018 | }; | |
2019 | ||
2020 | #undef AL | |
2021 | ||
2022 | /* | |
2023 | * System call vectors. | |
2024 | * | |
2025 | * Argument checking cleaned up. Saved 20% in size. | |
2026 | * This function doesn't need to set the kernel lock because | |
2027 | * it is set by the callees. | |
2028 | */ | |
2029 | ||
2030 | asmlinkage long sys_socketcall(int call, unsigned long __user *args) | |
2031 | { | |
2032 | unsigned long a[6]; | |
2033 | unsigned long a0, a1; | |
2034 | int err; | |
2035 | ||
2036 | if (call < 1 || call > SYS_RECVMSG) | |
2037 | return -EINVAL; | |
2038 | ||
2039 | /* copy_from_user should be SMP safe. */ | |
2040 | if (copy_from_user(a, args, nargs[call])) | |
2041 | return -EFAULT; | |
2042 | ||
2043 | err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); | |
2044 | if (err) | |
2045 | return err; | |
2046 | ||
2047 | a0 = a[0]; | |
2048 | a1 = a[1]; | |
2049 | ||
2050 | switch (call) { | |
2051 | case SYS_SOCKET: | |
2052 | err = sys_socket(a0, a1, a[2]); | |
2053 | break; | |
2054 | case SYS_BIND: | |
2055 | err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]); | |
2056 | break; | |
2057 | case SYS_CONNECT: | |
2058 | err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]); | |
2059 | break; | |
2060 | case SYS_LISTEN: | |
2061 | err = sys_listen(a0, a1); | |
2062 | break; | |
2063 | case SYS_ACCEPT: | |
2064 | err = | |
2065 | sys_accept(a0, (struct sockaddr __user *)a1, | |
2066 | (int __user *)a[2]); | |
2067 | break; | |
2068 | case SYS_GETSOCKNAME: | |
2069 | err = | |
2070 | sys_getsockname(a0, (struct sockaddr __user *)a1, | |
2071 | (int __user *)a[2]); | |
2072 | break; | |
2073 | case SYS_GETPEERNAME: | |
2074 | err = | |
2075 | sys_getpeername(a0, (struct sockaddr __user *)a1, | |
2076 | (int __user *)a[2]); | |
2077 | break; | |
2078 | case SYS_SOCKETPAIR: | |
2079 | err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]); | |
2080 | break; | |
2081 | case SYS_SEND: | |
2082 | err = sys_send(a0, (void __user *)a1, a[2], a[3]); | |
2083 | break; | |
2084 | case SYS_SENDTO: | |
2085 | err = sys_sendto(a0, (void __user *)a1, a[2], a[3], | |
2086 | (struct sockaddr __user *)a[4], a[5]); | |
2087 | break; | |
2088 | case SYS_RECV: | |
2089 | err = sys_recv(a0, (void __user *)a1, a[2], a[3]); | |
2090 | break; | |
2091 | case SYS_RECVFROM: | |
2092 | err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3], | |
2093 | (struct sockaddr __user *)a[4], | |
2094 | (int __user *)a[5]); | |
2095 | break; | |
2096 | case SYS_SHUTDOWN: | |
2097 | err = sys_shutdown(a0, a1); | |
2098 | break; | |
2099 | case SYS_SETSOCKOPT: | |
2100 | err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]); | |
2101 | break; | |
2102 | case SYS_GETSOCKOPT: | |
2103 | err = | |
2104 | sys_getsockopt(a0, a1, a[2], (char __user *)a[3], | |
2105 | (int __user *)a[4]); | |
2106 | break; | |
2107 | case SYS_SENDMSG: | |
2108 | err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2109 | break; | |
2110 | case SYS_RECVMSG: | |
2111 | err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2112 | break; | |
2113 | default: | |
2114 | err = -EINVAL; | |
2115 | break; | |
2116 | } | |
2117 | return err; | |
2118 | } | |
2119 | ||
2120 | #endif /* __ARCH_WANT_SYS_SOCKETCALL */ | |
2121 | ||
2122 | /** | |
2123 | * sock_register - add a socket protocol handler | |
2124 | * @ops: description of protocol | |
2125 | * | |
2126 | * This function is called by a protocol handler that wants to | |
2127 | * advertise its address family, and have it linked into the | |
2128 | * socket interface. The value ops->family coresponds to the | |
2129 | * socket system call protocol family. | |
2130 | */ | |
2131 | int sock_register(const struct net_proto_family *ops) | |
2132 | { | |
2133 | int err; | |
2134 | ||
2135 | if (ops->family >= NPROTO) { | |
2136 | printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, | |
2137 | NPROTO); | |
2138 | return -ENOBUFS; | |
2139 | } | |
2140 | ||
2141 | spin_lock(&net_family_lock); | |
2142 | if (net_families[ops->family]) | |
2143 | err = -EEXIST; | |
2144 | else { | |
2145 | net_families[ops->family] = ops; | |
2146 | err = 0; | |
2147 | } | |
2148 | spin_unlock(&net_family_lock); | |
2149 | ||
2150 | printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family); | |
2151 | return err; | |
2152 | } | |
2153 | ||
2154 | /** | |
2155 | * sock_unregister - remove a protocol handler | |
2156 | * @family: protocol family to remove | |
2157 | * | |
2158 | * This function is called by a protocol handler that wants to | |
2159 | * remove its address family, and have it unlinked from the | |
2160 | * new socket creation. | |
2161 | * | |
2162 | * If protocol handler is a module, then it can use module reference | |
2163 | * counts to protect against new references. If protocol handler is not | |
2164 | * a module then it needs to provide its own protection in | |
2165 | * the ops->create routine. | |
2166 | */ | |
2167 | void sock_unregister(int family) | |
2168 | { | |
2169 | BUG_ON(family < 0 || family >= NPROTO); | |
2170 | ||
2171 | spin_lock(&net_family_lock); | |
2172 | net_families[family] = NULL; | |
2173 | spin_unlock(&net_family_lock); | |
2174 | ||
2175 | synchronize_rcu(); | |
2176 | ||
2177 | printk(KERN_INFO "NET: Unregistered protocol family %d\n", family); | |
2178 | } | |
2179 | ||
2180 | static int __init sock_init(void) | |
2181 | { | |
2182 | /* | |
2183 | * Initialize sock SLAB cache. | |
2184 | */ | |
2185 | ||
2186 | sk_init(); | |
2187 | ||
2188 | /* | |
2189 | * Initialize skbuff SLAB cache | |
2190 | */ | |
2191 | skb_init(); | |
2192 | ||
2193 | /* | |
2194 | * Initialize the protocols module. | |
2195 | */ | |
2196 | ||
2197 | init_inodecache(); | |
2198 | register_filesystem(&sock_fs_type); | |
2199 | sock_mnt = kern_mount(&sock_fs_type); | |
2200 | ||
2201 | /* The real protocol initialization is performed in later initcalls. | |
2202 | */ | |
2203 | ||
2204 | #ifdef CONFIG_NETFILTER | |
2205 | netfilter_init(); | |
2206 | #endif | |
2207 | ||
2208 | return 0; | |
2209 | } | |
2210 | ||
2211 | core_initcall(sock_init); /* early initcall */ | |
2212 | ||
2213 | #ifdef CONFIG_PROC_FS | |
2214 | void socket_seq_show(struct seq_file *seq) | |
2215 | { | |
2216 | int cpu; | |
2217 | int counter = 0; | |
2218 | ||
2219 | for_each_possible_cpu(cpu) | |
2220 | counter += per_cpu(sockets_in_use, cpu); | |
2221 | ||
2222 | /* It can be negative, by the way. 8) */ | |
2223 | if (counter < 0) | |
2224 | counter = 0; | |
2225 | ||
2226 | seq_printf(seq, "sockets: used %d\n", counter); | |
2227 | } | |
2228 | #endif /* CONFIG_PROC_FS */ | |
2229 | ||
2230 | #ifdef CONFIG_COMPAT | |
2231 | static long compat_sock_ioctl(struct file *file, unsigned cmd, | |
2232 | unsigned long arg) | |
2233 | { | |
2234 | struct socket *sock = file->private_data; | |
2235 | int ret = -ENOIOCTLCMD; | |
2236 | struct sock *sk; | |
2237 | struct net *net; | |
2238 | ||
2239 | sk = sock->sk; | |
2240 | net = sock_net(sk); | |
2241 | ||
2242 | if (sock->ops->compat_ioctl) | |
2243 | ret = sock->ops->compat_ioctl(sock, cmd, arg); | |
2244 | ||
2245 | if (ret == -ENOIOCTLCMD && | |
2246 | (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) | |
2247 | ret = compat_wext_handle_ioctl(net, cmd, arg); | |
2248 | ||
2249 | return ret; | |
2250 | } | |
2251 | #endif | |
2252 | ||
2253 | int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) | |
2254 | { | |
2255 | return sock->ops->bind(sock, addr, addrlen); | |
2256 | } | |
2257 | ||
2258 | int kernel_listen(struct socket *sock, int backlog) | |
2259 | { | |
2260 | return sock->ops->listen(sock, backlog); | |
2261 | } | |
2262 | ||
2263 | int kernel_accept(struct socket *sock, struct socket **newsock, int flags) | |
2264 | { | |
2265 | struct sock *sk = sock->sk; | |
2266 | int err; | |
2267 | ||
2268 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
2269 | newsock); | |
2270 | if (err < 0) | |
2271 | goto done; | |
2272 | ||
2273 | err = sock->ops->accept(sock, *newsock, flags); | |
2274 | if (err < 0) { | |
2275 | sock_release(*newsock); | |
2276 | *newsock = NULL; | |
2277 | goto done; | |
2278 | } | |
2279 | ||
2280 | (*newsock)->ops = sock->ops; | |
2281 | ||
2282 | done: | |
2283 | return err; | |
2284 | } | |
2285 | ||
2286 | int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, | |
2287 | int flags) | |
2288 | { | |
2289 | return sock->ops->connect(sock, addr, addrlen, flags); | |
2290 | } | |
2291 | ||
2292 | int kernel_getsockname(struct socket *sock, struct sockaddr *addr, | |
2293 | int *addrlen) | |
2294 | { | |
2295 | return sock->ops->getname(sock, addr, addrlen, 0); | |
2296 | } | |
2297 | ||
2298 | int kernel_getpeername(struct socket *sock, struct sockaddr *addr, | |
2299 | int *addrlen) | |
2300 | { | |
2301 | return sock->ops->getname(sock, addr, addrlen, 1); | |
2302 | } | |
2303 | ||
2304 | int kernel_getsockopt(struct socket *sock, int level, int optname, | |
2305 | char *optval, int *optlen) | |
2306 | { | |
2307 | mm_segment_t oldfs = get_fs(); | |
2308 | int err; | |
2309 | ||
2310 | set_fs(KERNEL_DS); | |
2311 | if (level == SOL_SOCKET) | |
2312 | err = sock_getsockopt(sock, level, optname, optval, optlen); | |
2313 | else | |
2314 | err = sock->ops->getsockopt(sock, level, optname, optval, | |
2315 | optlen); | |
2316 | set_fs(oldfs); | |
2317 | return err; | |
2318 | } | |
2319 | ||
2320 | int kernel_setsockopt(struct socket *sock, int level, int optname, | |
2321 | char *optval, int optlen) | |
2322 | { | |
2323 | mm_segment_t oldfs = get_fs(); | |
2324 | int err; | |
2325 | ||
2326 | set_fs(KERNEL_DS); | |
2327 | if (level == SOL_SOCKET) | |
2328 | err = sock_setsockopt(sock, level, optname, optval, optlen); | |
2329 | else | |
2330 | err = sock->ops->setsockopt(sock, level, optname, optval, | |
2331 | optlen); | |
2332 | set_fs(oldfs); | |
2333 | return err; | |
2334 | } | |
2335 | ||
2336 | int kernel_sendpage(struct socket *sock, struct page *page, int offset, | |
2337 | size_t size, int flags) | |
2338 | { | |
2339 | if (sock->ops->sendpage) | |
2340 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
2341 | ||
2342 | return sock_no_sendpage(sock, page, offset, size, flags); | |
2343 | } | |
2344 | ||
2345 | int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg) | |
2346 | { | |
2347 | mm_segment_t oldfs = get_fs(); | |
2348 | int err; | |
2349 | ||
2350 | set_fs(KERNEL_DS); | |
2351 | err = sock->ops->ioctl(sock, cmd, arg); | |
2352 | set_fs(oldfs); | |
2353 | ||
2354 | return err; | |
2355 | } | |
2356 | ||
2357 | int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) | |
2358 | { | |
2359 | return sock->ops->shutdown(sock, how); | |
2360 | } | |
2361 | ||
2362 | EXPORT_SYMBOL(sock_create); | |
2363 | EXPORT_SYMBOL(sock_create_kern); | |
2364 | EXPORT_SYMBOL(sock_create_lite); | |
2365 | EXPORT_SYMBOL(sock_map_fd); | |
2366 | EXPORT_SYMBOL(sock_recvmsg); | |
2367 | EXPORT_SYMBOL(sock_register); | |
2368 | EXPORT_SYMBOL(sock_release); | |
2369 | EXPORT_SYMBOL(sock_sendmsg); | |
2370 | EXPORT_SYMBOL(sock_unregister); | |
2371 | EXPORT_SYMBOL(sock_wake_async); | |
2372 | EXPORT_SYMBOL(sockfd_lookup); | |
2373 | EXPORT_SYMBOL(kernel_sendmsg); | |
2374 | EXPORT_SYMBOL(kernel_recvmsg); | |
2375 | EXPORT_SYMBOL(kernel_bind); | |
2376 | EXPORT_SYMBOL(kernel_listen); | |
2377 | EXPORT_SYMBOL(kernel_accept); | |
2378 | EXPORT_SYMBOL(kernel_connect); | |
2379 | EXPORT_SYMBOL(kernel_getsockname); | |
2380 | EXPORT_SYMBOL(kernel_getpeername); | |
2381 | EXPORT_SYMBOL(kernel_getsockopt); | |
2382 | EXPORT_SYMBOL(kernel_setsockopt); | |
2383 | EXPORT_SYMBOL(kernel_sendpage); | |
2384 | EXPORT_SYMBOL(kernel_sock_ioctl); | |
2385 | EXPORT_SYMBOL(kernel_sock_shutdown); |