Root/
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/thread_info.h> |
67 | #include <linux/rcupdate.h> |
68 | #include <linux/netdevice.h> |
69 | #include <linux/proc_fs.h> |
70 | #include <linux/seq_file.h> |
71 | #include <linux/mutex.h> |
72 | #include <linux/wanrouter.h> |
73 | #include <linux/if_bridge.h> |
74 | #include <linux/if_frad.h> |
75 | #include <linux/if_vlan.h> |
76 | #include <linux/init.h> |
77 | #include <linux/poll.h> |
78 | #include <linux/cache.h> |
79 | #include <linux/module.h> |
80 | #include <linux/highmem.h> |
81 | #include <linux/mount.h> |
82 | #include <linux/security.h> |
83 | #include <linux/syscalls.h> |
84 | #include <linux/compat.h> |
85 | #include <linux/kmod.h> |
86 | #include <linux/audit.h> |
87 | #include <linux/wireless.h> |
88 | #include <linux/nsproxy.h> |
89 | #include <linux/magic.h> |
90 | #include <linux/slab.h> |
91 | |
92 | #include <asm/uaccess.h> |
93 | #include <asm/unistd.h> |
94 | |
95 | #include <net/compat.h> |
96 | #include <net/wext.h> |
97 | #include <net/cls_cgroup.h> |
98 | |
99 | #include <net/sock.h> |
100 | #include <linux/netfilter.h> |
101 | |
102 | #include <linux/if_tun.h> |
103 | #include <linux/ipv6_route.h> |
104 | #include <linux/route.h> |
105 | #include <linux/sockios.h> |
106 | #include <linux/atalk.h> |
107 | |
108 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare); |
109 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, |
110 | unsigned long nr_segs, loff_t pos); |
111 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, |
112 | unsigned long nr_segs, loff_t pos); |
113 | static int sock_mmap(struct file *file, struct vm_area_struct *vma); |
114 | |
115 | static int sock_close(struct inode *inode, struct file *file); |
116 | static unsigned int sock_poll(struct file *file, |
117 | struct poll_table_struct *wait); |
118 | static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); |
119 | #ifdef CONFIG_COMPAT |
120 | static long compat_sock_ioctl(struct file *file, |
121 | unsigned int cmd, unsigned long arg); |
122 | #endif |
123 | static int sock_fasync(int fd, struct file *filp, int on); |
124 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
125 | int offset, size_t size, loff_t *ppos, int more); |
126 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, |
127 | struct pipe_inode_info *pipe, size_t len, |
128 | unsigned int flags); |
129 | |
130 | /* |
131 | * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear |
132 | * in the operation structures but are done directly via the socketcall() multiplexor. |
133 | */ |
134 | |
135 | static const struct file_operations socket_file_ops = { |
136 | .owner = THIS_MODULE, |
137 | .llseek = no_llseek, |
138 | .aio_read = sock_aio_read, |
139 | .aio_write = sock_aio_write, |
140 | .poll = sock_poll, |
141 | .unlocked_ioctl = sock_ioctl, |
142 | #ifdef CONFIG_COMPAT |
143 | .compat_ioctl = compat_sock_ioctl, |
144 | #endif |
145 | .mmap = sock_mmap, |
146 | .open = sock_no_open, /* special open code to disallow open via /proc */ |
147 | .release = sock_close, |
148 | .fasync = sock_fasync, |
149 | .sendpage = sock_sendpage, |
150 | .splice_write = generic_splice_sendpage, |
151 | .splice_read = sock_splice_read, |
152 | }; |
153 | |
154 | /* |
155 | * The protocol list. Each protocol is registered in here. |
156 | */ |
157 | |
158 | static DEFINE_SPINLOCK(net_family_lock); |
159 | static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; |
160 | |
161 | /* |
162 | * Statistics counters of the socket lists |
163 | */ |
164 | |
165 | static DEFINE_PER_CPU(int, sockets_in_use); |
166 | |
167 | /* |
168 | * Support routines. |
169 | * Move socket addresses back and forth across the kernel/user |
170 | * divide and look after the messy bits. |
171 | */ |
172 | |
173 | /** |
174 | * move_addr_to_kernel - copy a socket address into kernel space |
175 | * @uaddr: Address in user space |
176 | * @kaddr: Address in kernel space |
177 | * @ulen: Length in user space |
178 | * |
179 | * The address is copied into kernel space. If the provided address is |
180 | * too long an error code of -EINVAL is returned. If the copy gives |
181 | * invalid addresses -EFAULT is returned. On a success 0 is returned. |
182 | */ |
183 | |
184 | int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr) |
185 | { |
186 | if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) |
187 | return -EINVAL; |
188 | if (ulen == 0) |
189 | return 0; |
190 | if (copy_from_user(kaddr, uaddr, ulen)) |
191 | return -EFAULT; |
192 | return audit_sockaddr(ulen, kaddr); |
193 | } |
194 | |
195 | /** |
196 | * move_addr_to_user - copy an address to user space |
197 | * @kaddr: kernel space address |
198 | * @klen: length of address in kernel |
199 | * @uaddr: user space address |
200 | * @ulen: pointer to user length field |
201 | * |
202 | * The value pointed to by ulen on entry is the buffer length available. |
203 | * This is overwritten with the buffer space used. -EINVAL is returned |
204 | * if an overlong buffer is specified or a negative buffer size. -EFAULT |
205 | * is returned if either the buffer or the length field are not |
206 | * accessible. |
207 | * After copying the data up to the limit the user specifies, the true |
208 | * length of the data is written over the length limit the user |
209 | * specified. Zero is returned for a success. |
210 | */ |
211 | |
212 | static int move_addr_to_user(struct sockaddr *kaddr, int klen, |
213 | void __user *uaddr, int __user *ulen) |
214 | { |
215 | int err; |
216 | int len; |
217 | |
218 | err = get_user(len, ulen); |
219 | if (err) |
220 | return err; |
221 | if (len > klen) |
222 | len = klen; |
223 | if (len < 0 || len > sizeof(struct sockaddr_storage)) |
224 | return -EINVAL; |
225 | if (len) { |
226 | if (audit_sockaddr(klen, kaddr)) |
227 | return -ENOMEM; |
228 | if (copy_to_user(uaddr, kaddr, len)) |
229 | return -EFAULT; |
230 | } |
231 | /* |
232 | * "fromlen shall refer to the value before truncation.." |
233 | * 1003.1g |
234 | */ |
235 | return __put_user(klen, ulen); |
236 | } |
237 | |
238 | static struct kmem_cache *sock_inode_cachep __read_mostly; |
239 | |
240 | static struct inode *sock_alloc_inode(struct super_block *sb) |
241 | { |
242 | struct socket_alloc *ei; |
243 | struct socket_wq *wq; |
244 | |
245 | ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); |
246 | if (!ei) |
247 | return NULL; |
248 | wq = kmalloc(sizeof(*wq), GFP_KERNEL); |
249 | if (!wq) { |
250 | kmem_cache_free(sock_inode_cachep, ei); |
251 | return NULL; |
252 | } |
253 | init_waitqueue_head(&wq->wait); |
254 | wq->fasync_list = NULL; |
255 | RCU_INIT_POINTER(ei->socket.wq, wq); |
256 | |
257 | ei->socket.state = SS_UNCONNECTED; |
258 | ei->socket.flags = 0; |
259 | ei->socket.ops = NULL; |
260 | ei->socket.sk = NULL; |
261 | ei->socket.file = NULL; |
262 | |
263 | return &ei->vfs_inode; |
264 | } |
265 | |
266 | static void sock_destroy_inode(struct inode *inode) |
267 | { |
268 | struct socket_alloc *ei; |
269 | struct socket_wq *wq; |
270 | |
271 | ei = container_of(inode, struct socket_alloc, vfs_inode); |
272 | wq = rcu_dereference_protected(ei->socket.wq, 1); |
273 | kfree_rcu(wq, rcu); |
274 | kmem_cache_free(sock_inode_cachep, ei); |
275 | } |
276 | |
277 | static void init_once(void *foo) |
278 | { |
279 | struct socket_alloc *ei = (struct socket_alloc *)foo; |
280 | |
281 | inode_init_once(&ei->vfs_inode); |
282 | } |
283 | |
284 | static int init_inodecache(void) |
285 | { |
286 | sock_inode_cachep = kmem_cache_create("sock_inode_cache", |
287 | sizeof(struct socket_alloc), |
288 | 0, |
289 | (SLAB_HWCACHE_ALIGN | |
290 | SLAB_RECLAIM_ACCOUNT | |
291 | SLAB_MEM_SPREAD), |
292 | init_once); |
293 | if (sock_inode_cachep == NULL) |
294 | return -ENOMEM; |
295 | return 0; |
296 | } |
297 | |
298 | static const struct super_operations sockfs_ops = { |
299 | .alloc_inode = sock_alloc_inode, |
300 | .destroy_inode = sock_destroy_inode, |
301 | .statfs = simple_statfs, |
302 | }; |
303 | |
304 | /* |
305 | * sockfs_dname() is called from d_path(). |
306 | */ |
307 | static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) |
308 | { |
309 | return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", |
310 | dentry->d_inode->i_ino); |
311 | } |
312 | |
313 | static const struct dentry_operations sockfs_dentry_operations = { |
314 | .d_dname = sockfs_dname, |
315 | }; |
316 | |
317 | static struct dentry *sockfs_mount(struct file_system_type *fs_type, |
318 | int flags, const char *dev_name, void *data) |
319 | { |
320 | return mount_pseudo(fs_type, "socket:", &sockfs_ops, |
321 | &sockfs_dentry_operations, SOCKFS_MAGIC); |
322 | } |
323 | |
324 | static struct vfsmount *sock_mnt __read_mostly; |
325 | |
326 | static struct file_system_type sock_fs_type = { |
327 | .name = "sockfs", |
328 | .mount = sockfs_mount, |
329 | .kill_sb = kill_anon_super, |
330 | }; |
331 | |
332 | /* |
333 | * Obtains the first available file descriptor and sets it up for use. |
334 | * |
335 | * These functions create file structures and maps them to fd space |
336 | * of the current process. On success it returns file descriptor |
337 | * and file struct implicitly stored in sock->file. |
338 | * Note that another thread may close file descriptor before we return |
339 | * from this function. We use the fact that now we do not refer |
340 | * to socket after mapping. If one day we will need it, this |
341 | * function will increment ref. count on file by 1. |
342 | * |
343 | * In any case returned fd MAY BE not valid! |
344 | * This race condition is unavoidable |
345 | * with shared fd spaces, we cannot solve it inside kernel, |
346 | * but we take care of internal coherence yet. |
347 | */ |
348 | |
349 | static int sock_alloc_file(struct socket *sock, struct file **f, int flags) |
350 | { |
351 | struct qstr name = { .name = "" }; |
352 | struct path path; |
353 | struct file *file; |
354 | int fd; |
355 | |
356 | fd = get_unused_fd_flags(flags); |
357 | if (unlikely(fd < 0)) |
358 | return fd; |
359 | |
360 | path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name); |
361 | if (unlikely(!path.dentry)) { |
362 | put_unused_fd(fd); |
363 | return -ENOMEM; |
364 | } |
365 | path.mnt = mntget(sock_mnt); |
366 | |
367 | d_instantiate(path.dentry, SOCK_INODE(sock)); |
368 | SOCK_INODE(sock)->i_fop = &socket_file_ops; |
369 | |
370 | file = alloc_file(&path, FMODE_READ | FMODE_WRITE, |
371 | &socket_file_ops); |
372 | if (unlikely(!file)) { |
373 | /* drop dentry, keep inode */ |
374 | ihold(path.dentry->d_inode); |
375 | path_put(&path); |
376 | put_unused_fd(fd); |
377 | return -ENFILE; |
378 | } |
379 | |
380 | sock->file = file; |
381 | file->f_flags = O_RDWR | (flags & O_NONBLOCK); |
382 | file->f_pos = 0; |
383 | file->private_data = sock; |
384 | |
385 | *f = file; |
386 | return fd; |
387 | } |
388 | |
389 | int sock_map_fd(struct socket *sock, int flags) |
390 | { |
391 | struct file *newfile; |
392 | int fd = sock_alloc_file(sock, &newfile, flags); |
393 | |
394 | if (likely(fd >= 0)) |
395 | fd_install(fd, newfile); |
396 | |
397 | return fd; |
398 | } |
399 | EXPORT_SYMBOL(sock_map_fd); |
400 | |
401 | static struct socket *sock_from_file(struct file *file, int *err) |
402 | { |
403 | if (file->f_op == &socket_file_ops) |
404 | return file->private_data; /* set in sock_map_fd */ |
405 | |
406 | *err = -ENOTSOCK; |
407 | return NULL; |
408 | } |
409 | |
410 | /** |
411 | * sockfd_lookup - Go from a file number to its socket slot |
412 | * @fd: file handle |
413 | * @err: pointer to an error code return |
414 | * |
415 | * The file handle passed in is locked and the socket it is bound |
416 | * too is returned. If an error occurs the err pointer is overwritten |
417 | * with a negative errno code and NULL is returned. The function checks |
418 | * for both invalid handles and passing a handle which is not a socket. |
419 | * |
420 | * On a success the socket object pointer is returned. |
421 | */ |
422 | |
423 | struct socket *sockfd_lookup(int fd, int *err) |
424 | { |
425 | struct file *file; |
426 | struct socket *sock; |
427 | |
428 | file = fget(fd); |
429 | if (!file) { |
430 | *err = -EBADF; |
431 | return NULL; |
432 | } |
433 | |
434 | sock = sock_from_file(file, err); |
435 | if (!sock) |
436 | fput(file); |
437 | return sock; |
438 | } |
439 | EXPORT_SYMBOL(sockfd_lookup); |
440 | |
441 | static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) |
442 | { |
443 | struct file *file; |
444 | struct socket *sock; |
445 | |
446 | *err = -EBADF; |
447 | file = fget_light(fd, fput_needed); |
448 | if (file) { |
449 | sock = sock_from_file(file, err); |
450 | if (sock) |
451 | return sock; |
452 | fput_light(file, *fput_needed); |
453 | } |
454 | return NULL; |
455 | } |
456 | |
457 | /** |
458 | * sock_alloc - allocate a socket |
459 | * |
460 | * Allocate a new inode and socket object. The two are bound together |
461 | * and initialised. The socket is then returned. If we are out of inodes |
462 | * NULL is returned. |
463 | */ |
464 | |
465 | static struct socket *sock_alloc(void) |
466 | { |
467 | struct inode *inode; |
468 | struct socket *sock; |
469 | |
470 | inode = new_inode_pseudo(sock_mnt->mnt_sb); |
471 | if (!inode) |
472 | return NULL; |
473 | |
474 | sock = SOCKET_I(inode); |
475 | |
476 | kmemcheck_annotate_bitfield(sock, type); |
477 | inode->i_ino = get_next_ino(); |
478 | inode->i_mode = S_IFSOCK | S_IRWXUGO; |
479 | inode->i_uid = current_fsuid(); |
480 | inode->i_gid = current_fsgid(); |
481 | |
482 | percpu_add(sockets_in_use, 1); |
483 | return sock; |
484 | } |
485 | |
486 | /* |
487 | * In theory you can't get an open on this inode, but /proc provides |
488 | * a back door. Remember to keep it shut otherwise you'll let the |
489 | * creepy crawlies in. |
490 | */ |
491 | |
492 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare) |
493 | { |
494 | return -ENXIO; |
495 | } |
496 | |
497 | const struct file_operations bad_sock_fops = { |
498 | .owner = THIS_MODULE, |
499 | .open = sock_no_open, |
500 | .llseek = noop_llseek, |
501 | }; |
502 | |
503 | /** |
504 | * sock_release - close a socket |
505 | * @sock: socket to close |
506 | * |
507 | * The socket is released from the protocol stack if it has a release |
508 | * callback, and the inode is then released if the socket is bound to |
509 | * an inode not a file. |
510 | */ |
511 | |
512 | void sock_release(struct socket *sock) |
513 | { |
514 | if (sock->ops) { |
515 | struct module *owner = sock->ops->owner; |
516 | |
517 | sock->ops->release(sock); |
518 | sock->ops = NULL; |
519 | module_put(owner); |
520 | } |
521 | |
522 | if (rcu_dereference_protected(sock->wq, 1)->fasync_list) |
523 | printk(KERN_ERR "sock_release: fasync list not empty!\n"); |
524 | |
525 | percpu_sub(sockets_in_use, 1); |
526 | if (!sock->file) { |
527 | iput(SOCK_INODE(sock)); |
528 | return; |
529 | } |
530 | sock->file = NULL; |
531 | } |
532 | EXPORT_SYMBOL(sock_release); |
533 | |
534 | int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags) |
535 | { |
536 | *tx_flags = 0; |
537 | if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE)) |
538 | *tx_flags |= SKBTX_HW_TSTAMP; |
539 | if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE)) |
540 | *tx_flags |= SKBTX_SW_TSTAMP; |
541 | if (sock_flag(sk, SOCK_WIFI_STATUS)) |
542 | *tx_flags |= SKBTX_WIFI_STATUS; |
543 | return 0; |
544 | } |
545 | EXPORT_SYMBOL(sock_tx_timestamp); |
546 | |
547 | static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock, |
548 | struct msghdr *msg, size_t size) |
549 | { |
550 | struct sock_iocb *si = kiocb_to_siocb(iocb); |
551 | |
552 | sock_update_classid(sock->sk); |
553 | |
554 | sock_update_netprioidx(sock->sk); |
555 | |
556 | si->sock = sock; |
557 | si->scm = NULL; |
558 | si->msg = msg; |
559 | si->size = size; |
560 | |
561 | return sock->ops->sendmsg(iocb, sock, msg, size); |
562 | } |
563 | |
564 | static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock, |
565 | struct msghdr *msg, size_t size) |
566 | { |
567 | int err = security_socket_sendmsg(sock, msg, size); |
568 | |
569 | return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size); |
570 | } |
571 | |
572 | int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) |
573 | { |
574 | struct kiocb iocb; |
575 | struct sock_iocb siocb; |
576 | int ret; |
577 | |
578 | init_sync_kiocb(&iocb, NULL); |
579 | iocb.private = &siocb; |
580 | ret = __sock_sendmsg(&iocb, sock, msg, size); |
581 | if (-EIOCBQUEUED == ret) |
582 | ret = wait_on_sync_kiocb(&iocb); |
583 | return ret; |
584 | } |
585 | EXPORT_SYMBOL(sock_sendmsg); |
586 | |
587 | static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size) |
588 | { |
589 | struct kiocb iocb; |
590 | struct sock_iocb siocb; |
591 | int ret; |
592 | |
593 | init_sync_kiocb(&iocb, NULL); |
594 | iocb.private = &siocb; |
595 | ret = __sock_sendmsg_nosec(&iocb, sock, msg, size); |
596 | if (-EIOCBQUEUED == ret) |
597 | ret = wait_on_sync_kiocb(&iocb); |
598 | return ret; |
599 | } |
600 | |
601 | int kernel_sendmsg(struct socket *sock, struct msghdr *msg, |
602 | struct kvec *vec, size_t num, size_t size) |
603 | { |
604 | mm_segment_t oldfs = get_fs(); |
605 | int result; |
606 | |
607 | set_fs(KERNEL_DS); |
608 | /* |
609 | * the following is safe, since for compiler definitions of kvec and |
610 | * iovec are identical, yielding the same in-core layout and alignment |
611 | */ |
612 | msg->msg_iov = (struct iovec *)vec; |
613 | msg->msg_iovlen = num; |
614 | result = sock_sendmsg(sock, msg, size); |
615 | set_fs(oldfs); |
616 | return result; |
617 | } |
618 | EXPORT_SYMBOL(kernel_sendmsg); |
619 | |
620 | static int ktime2ts(ktime_t kt, struct timespec *ts) |
621 | { |
622 | if (kt.tv64) { |
623 | *ts = ktime_to_timespec(kt); |
624 | return 1; |
625 | } else { |
626 | return 0; |
627 | } |
628 | } |
629 | |
630 | /* |
631 | * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) |
632 | */ |
633 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
634 | struct sk_buff *skb) |
635 | { |
636 | int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); |
637 | struct timespec ts[3]; |
638 | int empty = 1; |
639 | struct skb_shared_hwtstamps *shhwtstamps = |
640 | skb_hwtstamps(skb); |
641 | |
642 | /* Race occurred between timestamp enabling and packet |
643 | receiving. Fill in the current time for now. */ |
644 | if (need_software_tstamp && skb->tstamp.tv64 == 0) |
645 | __net_timestamp(skb); |
646 | |
647 | if (need_software_tstamp) { |
648 | if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { |
649 | struct timeval tv; |
650 | skb_get_timestamp(skb, &tv); |
651 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, |
652 | sizeof(tv), &tv); |
653 | } else { |
654 | skb_get_timestampns(skb, &ts[0]); |
655 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, |
656 | sizeof(ts[0]), &ts[0]); |
657 | } |
658 | } |
659 | |
660 | |
661 | memset(ts, 0, sizeof(ts)); |
662 | if (skb->tstamp.tv64 && |
663 | sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) { |
664 | skb_get_timestampns(skb, ts + 0); |
665 | empty = 0; |
666 | } |
667 | if (shhwtstamps) { |
668 | if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) && |
669 | ktime2ts(shhwtstamps->syststamp, ts + 1)) |
670 | empty = 0; |
671 | if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) && |
672 | ktime2ts(shhwtstamps->hwtstamp, ts + 2)) |
673 | empty = 0; |
674 | } |
675 | if (!empty) |
676 | put_cmsg(msg, SOL_SOCKET, |
677 | SCM_TIMESTAMPING, sizeof(ts), &ts); |
678 | } |
679 | EXPORT_SYMBOL_GPL(__sock_recv_timestamp); |
680 | |
681 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, |
682 | struct sk_buff *skb) |
683 | { |
684 | int ack; |
685 | |
686 | if (!sock_flag(sk, SOCK_WIFI_STATUS)) |
687 | return; |
688 | if (!skb->wifi_acked_valid) |
689 | return; |
690 | |
691 | ack = skb->wifi_acked; |
692 | |
693 | put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); |
694 | } |
695 | EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); |
696 | |
697 | static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, |
698 | struct sk_buff *skb) |
699 | { |
700 | if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount) |
701 | put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, |
702 | sizeof(__u32), &skb->dropcount); |
703 | } |
704 | |
705 | void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
706 | struct sk_buff *skb) |
707 | { |
708 | sock_recv_timestamp(msg, sk, skb); |
709 | sock_recv_drops(msg, sk, skb); |
710 | } |
711 | EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops); |
712 | |
713 | static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock, |
714 | struct msghdr *msg, size_t size, int flags) |
715 | { |
716 | struct sock_iocb *si = kiocb_to_siocb(iocb); |
717 | |
718 | sock_update_classid(sock->sk); |
719 | |
720 | si->sock = sock; |
721 | si->scm = NULL; |
722 | si->msg = msg; |
723 | si->size = size; |
724 | si->flags = flags; |
725 | |
726 | return sock->ops->recvmsg(iocb, sock, msg, size, flags); |
727 | } |
728 | |
729 | static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock, |
730 | struct msghdr *msg, size_t size, int flags) |
731 | { |
732 | int err = security_socket_recvmsg(sock, msg, size, flags); |
733 | |
734 | return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags); |
735 | } |
736 | |
737 | int sock_recvmsg(struct socket *sock, struct msghdr *msg, |
738 | size_t size, int flags) |
739 | { |
740 | struct kiocb iocb; |
741 | struct sock_iocb siocb; |
742 | int ret; |
743 | |
744 | init_sync_kiocb(&iocb, NULL); |
745 | iocb.private = &siocb; |
746 | ret = __sock_recvmsg(&iocb, sock, msg, size, flags); |
747 | if (-EIOCBQUEUED == ret) |
748 | ret = wait_on_sync_kiocb(&iocb); |
749 | return ret; |
750 | } |
751 | EXPORT_SYMBOL(sock_recvmsg); |
752 | |
753 | static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, |
754 | size_t size, int flags) |
755 | { |
756 | struct kiocb iocb; |
757 | struct sock_iocb siocb; |
758 | int ret; |
759 | |
760 | init_sync_kiocb(&iocb, NULL); |
761 | iocb.private = &siocb; |
762 | ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags); |
763 | if (-EIOCBQUEUED == ret) |
764 | ret = wait_on_sync_kiocb(&iocb); |
765 | return ret; |
766 | } |
767 | |
768 | /** |
769 | * kernel_recvmsg - Receive a message from a socket (kernel space) |
770 | * @sock: The socket to receive the message from |
771 | * @msg: Received message |
772 | * @vec: Input s/g array for message data |
773 | * @num: Size of input s/g array |
774 | * @size: Number of bytes to read |
775 | * @flags: Message flags (MSG_DONTWAIT, etc...) |
776 | * |
777 | * On return the msg structure contains the scatter/gather array passed in the |
778 | * vec argument. The array is modified so that it consists of the unfilled |
779 | * portion of the original array. |
780 | * |
781 | * The returned value is the total number of bytes received, or an error. |
782 | */ |
783 | int kernel_recvmsg(struct socket *sock, struct msghdr *msg, |
784 | struct kvec *vec, size_t num, size_t size, int flags) |
785 | { |
786 | mm_segment_t oldfs = get_fs(); |
787 | int result; |
788 | |
789 | set_fs(KERNEL_DS); |
790 | /* |
791 | * the following is safe, since for compiler definitions of kvec and |
792 | * iovec are identical, yielding the same in-core layout and alignment |
793 | */ |
794 | msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num; |
795 | result = sock_recvmsg(sock, msg, size, flags); |
796 | set_fs(oldfs); |
797 | return result; |
798 | } |
799 | EXPORT_SYMBOL(kernel_recvmsg); |
800 | |
801 | static void sock_aio_dtor(struct kiocb *iocb) |
802 | { |
803 | kfree(iocb->private); |
804 | } |
805 | |
806 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
807 | int offset, size_t size, loff_t *ppos, int more) |
808 | { |
809 | struct socket *sock; |
810 | int flags; |
811 | |
812 | sock = file->private_data; |
813 | |
814 | flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT; |
815 | if (more) |
816 | flags |= MSG_MORE; |
817 | |
818 | return kernel_sendpage(sock, page, offset, size, flags); |
819 | } |
820 | |
821 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, |
822 | struct pipe_inode_info *pipe, size_t len, |
823 | unsigned int flags) |
824 | { |
825 | struct socket *sock = file->private_data; |
826 | |
827 | if (unlikely(!sock->ops->splice_read)) |
828 | return -EINVAL; |
829 | |
830 | sock_update_classid(sock->sk); |
831 | |
832 | return sock->ops->splice_read(sock, ppos, pipe, len, flags); |
833 | } |
834 | |
835 | static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb, |
836 | struct sock_iocb *siocb) |
837 | { |
838 | if (!is_sync_kiocb(iocb)) { |
839 | siocb = kmalloc(sizeof(*siocb), GFP_KERNEL); |
840 | if (!siocb) |
841 | return NULL; |
842 | iocb->ki_dtor = sock_aio_dtor; |
843 | } |
844 | |
845 | siocb->kiocb = iocb; |
846 | iocb->private = siocb; |
847 | return siocb; |
848 | } |
849 | |
850 | static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb, |
851 | struct file *file, const struct iovec *iov, |
852 | unsigned long nr_segs) |
853 | { |
854 | struct socket *sock = file->private_data; |
855 | size_t size = 0; |
856 | int i; |
857 | |
858 | for (i = 0; i < nr_segs; i++) |
859 | size += iov[i].iov_len; |
860 | |
861 | msg->msg_name = NULL; |
862 | msg->msg_namelen = 0; |
863 | msg->msg_control = NULL; |
864 | msg->msg_controllen = 0; |
865 | msg->msg_iov = (struct iovec *)iov; |
866 | msg->msg_iovlen = nr_segs; |
867 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; |
868 | |
869 | return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags); |
870 | } |
871 | |
872 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, |
873 | unsigned long nr_segs, loff_t pos) |
874 | { |
875 | struct sock_iocb siocb, *x; |
876 | |
877 | if (pos != 0) |
878 | return -ESPIPE; |
879 | |
880 | if (iocb->ki_left == 0) /* Match SYS5 behaviour */ |
881 | return 0; |
882 | |
883 | |
884 | x = alloc_sock_iocb(iocb, &siocb); |
885 | if (!x) |
886 | return -ENOMEM; |
887 | return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
888 | } |
889 | |
890 | static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb, |
891 | struct file *file, const struct iovec *iov, |
892 | unsigned long nr_segs) |
893 | { |
894 | struct socket *sock = file->private_data; |
895 | size_t size = 0; |
896 | int i; |
897 | |
898 | for (i = 0; i < nr_segs; i++) |
899 | size += iov[i].iov_len; |
900 | |
901 | msg->msg_name = NULL; |
902 | msg->msg_namelen = 0; |
903 | msg->msg_control = NULL; |
904 | msg->msg_controllen = 0; |
905 | msg->msg_iov = (struct iovec *)iov; |
906 | msg->msg_iovlen = nr_segs; |
907 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; |
908 | if (sock->type == SOCK_SEQPACKET) |
909 | msg->msg_flags |= MSG_EOR; |
910 | |
911 | return __sock_sendmsg(iocb, sock, msg, size); |
912 | } |
913 | |
914 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, |
915 | unsigned long nr_segs, loff_t pos) |
916 | { |
917 | struct sock_iocb siocb, *x; |
918 | |
919 | if (pos != 0) |
920 | return -ESPIPE; |
921 | |
922 | x = alloc_sock_iocb(iocb, &siocb); |
923 | if (!x) |
924 | return -ENOMEM; |
925 | |
926 | return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
927 | } |
928 | |
929 | /* |
930 | * Atomic setting of ioctl hooks to avoid race |
931 | * with module unload. |
932 | */ |
933 | |
934 | static DEFINE_MUTEX(br_ioctl_mutex); |
935 | static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg); |
936 | |
937 | void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) |
938 | { |
939 | mutex_lock(&br_ioctl_mutex); |
940 | br_ioctl_hook = hook; |
941 | mutex_unlock(&br_ioctl_mutex); |
942 | } |
943 | EXPORT_SYMBOL(brioctl_set); |
944 | |
945 | static DEFINE_MUTEX(vlan_ioctl_mutex); |
946 | static int (*vlan_ioctl_hook) (struct net *, void __user *arg); |
947 | |
948 | void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) |
949 | { |
950 | mutex_lock(&vlan_ioctl_mutex); |
951 | vlan_ioctl_hook = hook; |
952 | mutex_unlock(&vlan_ioctl_mutex); |
953 | } |
954 | EXPORT_SYMBOL(vlan_ioctl_set); |
955 | |
956 | static DEFINE_MUTEX(dlci_ioctl_mutex); |
957 | static int (*dlci_ioctl_hook) (unsigned int, void __user *); |
958 | |
959 | void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) |
960 | { |
961 | mutex_lock(&dlci_ioctl_mutex); |
962 | dlci_ioctl_hook = hook; |
963 | mutex_unlock(&dlci_ioctl_mutex); |
964 | } |
965 | EXPORT_SYMBOL(dlci_ioctl_set); |
966 | |
967 | static long sock_do_ioctl(struct net *net, struct socket *sock, |
968 | unsigned int cmd, unsigned long arg) |
969 | { |
970 | int err; |
971 | void __user *argp = (void __user *)arg; |
972 | |
973 | err = sock->ops->ioctl(sock, cmd, arg); |
974 | |
975 | /* |
976 | * If this ioctl is unknown try to hand it down |
977 | * to the NIC driver. |
978 | */ |
979 | if (err == -ENOIOCTLCMD) |
980 | err = dev_ioctl(net, cmd, argp); |
981 | |
982 | return err; |
983 | } |
984 | |
985 | /* |
986 | * With an ioctl, arg may well be a user mode pointer, but we don't know |
987 | * what to do with it - that's up to the protocol still. |
988 | */ |
989 | |
990 | static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) |
991 | { |
992 | struct socket *sock; |
993 | struct sock *sk; |
994 | void __user *argp = (void __user *)arg; |
995 | int pid, err; |
996 | struct net *net; |
997 | |
998 | sock = file->private_data; |
999 | sk = sock->sk; |
1000 | net = sock_net(sk); |
1001 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) { |
1002 | err = dev_ioctl(net, cmd, argp); |
1003 | } else |
1004 | #ifdef CONFIG_WEXT_CORE |
1005 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { |
1006 | err = dev_ioctl(net, cmd, argp); |
1007 | } else |
1008 | #endif |
1009 | switch (cmd) { |
1010 | case FIOSETOWN: |
1011 | case SIOCSPGRP: |
1012 | err = -EFAULT; |
1013 | if (get_user(pid, (int __user *)argp)) |
1014 | break; |
1015 | err = f_setown(sock->file, pid, 1); |
1016 | break; |
1017 | case FIOGETOWN: |
1018 | case SIOCGPGRP: |
1019 | err = put_user(f_getown(sock->file), |
1020 | (int __user *)argp); |
1021 | break; |
1022 | case SIOCGIFBR: |
1023 | case SIOCSIFBR: |
1024 | case SIOCBRADDBR: |
1025 | case SIOCBRDELBR: |
1026 | err = -ENOPKG; |
1027 | if (!br_ioctl_hook) |
1028 | request_module("bridge"); |
1029 | |
1030 | mutex_lock(&br_ioctl_mutex); |
1031 | if (br_ioctl_hook) |
1032 | err = br_ioctl_hook(net, cmd, argp); |
1033 | mutex_unlock(&br_ioctl_mutex); |
1034 | break; |
1035 | case SIOCGIFVLAN: |
1036 | case SIOCSIFVLAN: |
1037 | err = -ENOPKG; |
1038 | if (!vlan_ioctl_hook) |
1039 | request_module("8021q"); |
1040 | |
1041 | mutex_lock(&vlan_ioctl_mutex); |
1042 | if (vlan_ioctl_hook) |
1043 | err = vlan_ioctl_hook(net, argp); |
1044 | mutex_unlock(&vlan_ioctl_mutex); |
1045 | break; |
1046 | case SIOCADDDLCI: |
1047 | case SIOCDELDLCI: |
1048 | err = -ENOPKG; |
1049 | if (!dlci_ioctl_hook) |
1050 | request_module("dlci"); |
1051 | |
1052 | mutex_lock(&dlci_ioctl_mutex); |
1053 | if (dlci_ioctl_hook) |
1054 | err = dlci_ioctl_hook(cmd, argp); |
1055 | mutex_unlock(&dlci_ioctl_mutex); |
1056 | break; |
1057 | default: |
1058 | err = sock_do_ioctl(net, sock, cmd, arg); |
1059 | break; |
1060 | } |
1061 | return err; |
1062 | } |
1063 | |
1064 | int sock_create_lite(int family, int type, int protocol, struct socket **res) |
1065 | { |
1066 | int err; |
1067 | struct socket *sock = NULL; |
1068 | |
1069 | err = security_socket_create(family, type, protocol, 1); |
1070 | if (err) |
1071 | goto out; |
1072 | |
1073 | sock = sock_alloc(); |
1074 | if (!sock) { |
1075 | err = -ENOMEM; |
1076 | goto out; |
1077 | } |
1078 | |
1079 | sock->type = type; |
1080 | err = security_socket_post_create(sock, family, type, protocol, 1); |
1081 | if (err) |
1082 | goto out_release; |
1083 | |
1084 | out: |
1085 | *res = sock; |
1086 | return err; |
1087 | out_release: |
1088 | sock_release(sock); |
1089 | sock = NULL; |
1090 | goto out; |
1091 | } |
1092 | EXPORT_SYMBOL(sock_create_lite); |
1093 | |
1094 | /* No kernel lock held - perfect */ |
1095 | static unsigned int sock_poll(struct file *file, poll_table *wait) |
1096 | { |
1097 | struct socket *sock; |
1098 | |
1099 | /* |
1100 | * We can't return errors to poll, so it's either yes or no. |
1101 | */ |
1102 | sock = file->private_data; |
1103 | return sock->ops->poll(file, sock, wait); |
1104 | } |
1105 | |
1106 | static int sock_mmap(struct file *file, struct vm_area_struct *vma) |
1107 | { |
1108 | struct socket *sock = file->private_data; |
1109 | |
1110 | return sock->ops->mmap(file, sock, vma); |
1111 | } |
1112 | |
1113 | static int sock_close(struct inode *inode, struct file *filp) |
1114 | { |
1115 | /* |
1116 | * It was possible the inode is NULL we were |
1117 | * closing an unfinished socket. |
1118 | */ |
1119 | |
1120 | if (!inode) { |
1121 | printk(KERN_DEBUG "sock_close: NULL inode\n"); |
1122 | return 0; |
1123 | } |
1124 | sock_release(SOCKET_I(inode)); |
1125 | return 0; |
1126 | } |
1127 | |
1128 | /* |
1129 | * Update the socket async list |
1130 | * |
1131 | * Fasync_list locking strategy. |
1132 | * |
1133 | * 1. fasync_list is modified only under process context socket lock |
1134 | * i.e. under semaphore. |
1135 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) |
1136 | * or under socket lock |
1137 | */ |
1138 | |
1139 | static int sock_fasync(int fd, struct file *filp, int on) |
1140 | { |
1141 | struct socket *sock = filp->private_data; |
1142 | struct sock *sk = sock->sk; |
1143 | struct socket_wq *wq; |
1144 | |
1145 | if (sk == NULL) |
1146 | return -EINVAL; |
1147 | |
1148 | lock_sock(sk); |
1149 | wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk)); |
1150 | fasync_helper(fd, filp, on, &wq->fasync_list); |
1151 | |
1152 | if (!wq->fasync_list) |
1153 | sock_reset_flag(sk, SOCK_FASYNC); |
1154 | else |
1155 | sock_set_flag(sk, SOCK_FASYNC); |
1156 | |
1157 | release_sock(sk); |
1158 | return 0; |
1159 | } |
1160 | |
1161 | /* This function may be called only under socket lock or callback_lock or rcu_lock */ |
1162 | |
1163 | int sock_wake_async(struct socket *sock, int how, int band) |
1164 | { |
1165 | struct socket_wq *wq; |
1166 | |
1167 | if (!sock) |
1168 | return -1; |
1169 | rcu_read_lock(); |
1170 | wq = rcu_dereference(sock->wq); |
1171 | if (!wq || !wq->fasync_list) { |
1172 | rcu_read_unlock(); |
1173 | return -1; |
1174 | } |
1175 | switch (how) { |
1176 | case SOCK_WAKE_WAITD: |
1177 | if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags)) |
1178 | break; |
1179 | goto call_kill; |
1180 | case SOCK_WAKE_SPACE: |
1181 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags)) |
1182 | break; |
1183 | /* fall through */ |
1184 | case SOCK_WAKE_IO: |
1185 | call_kill: |
1186 | kill_fasync(&wq->fasync_list, SIGIO, band); |
1187 | break; |
1188 | case SOCK_WAKE_URG: |
1189 | kill_fasync(&wq->fasync_list, SIGURG, band); |
1190 | } |
1191 | rcu_read_unlock(); |
1192 | return 0; |
1193 | } |
1194 | EXPORT_SYMBOL(sock_wake_async); |
1195 | |
1196 | int __sock_create(struct net *net, int family, int type, int protocol, |
1197 | struct socket **res, int kern) |
1198 | { |
1199 | int err; |
1200 | struct socket *sock; |
1201 | const struct net_proto_family *pf; |
1202 | |
1203 | /* |
1204 | * Check protocol is in range |
1205 | */ |
1206 | if (family < 0 || family >= NPROTO) |
1207 | return -EAFNOSUPPORT; |
1208 | if (type < 0 || type >= SOCK_MAX) |
1209 | return -EINVAL; |
1210 | |
1211 | /* Compatibility. |
1212 | |
1213 | This uglymoron is moved from INET layer to here to avoid |
1214 | deadlock in module load. |
1215 | */ |
1216 | if (family == PF_INET && type == SOCK_PACKET) { |
1217 | static int warned; |
1218 | if (!warned) { |
1219 | warned = 1; |
1220 | printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", |
1221 | current->comm); |
1222 | } |
1223 | family = PF_PACKET; |
1224 | } |
1225 | |
1226 | err = security_socket_create(family, type, protocol, kern); |
1227 | if (err) |
1228 | return err; |
1229 | |
1230 | /* |
1231 | * Allocate the socket and allow the family to set things up. if |
1232 | * the protocol is 0, the family is instructed to select an appropriate |
1233 | * default. |
1234 | */ |
1235 | sock = sock_alloc(); |
1236 | if (!sock) { |
1237 | if (net_ratelimit()) |
1238 | printk(KERN_WARNING "socket: no more sockets\n"); |
1239 | return -ENFILE; /* Not exactly a match, but its the |
1240 | closest posix thing */ |
1241 | } |
1242 | |
1243 | sock->type = type; |
1244 | |
1245 | #ifdef CONFIG_MODULES |
1246 | /* Attempt to load a protocol module if the find failed. |
1247 | * |
1248 | * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user |
1249 | * requested real, full-featured networking support upon configuration. |
1250 | * Otherwise module support will break! |
1251 | */ |
1252 | if (rcu_access_pointer(net_families[family]) == NULL) |
1253 | request_module("net-pf-%d", family); |
1254 | #endif |
1255 | |
1256 | rcu_read_lock(); |
1257 | pf = rcu_dereference(net_families[family]); |
1258 | err = -EAFNOSUPPORT; |
1259 | if (!pf) |
1260 | goto out_release; |
1261 | |
1262 | /* |
1263 | * We will call the ->create function, that possibly is in a loadable |
1264 | * module, so we have to bump that loadable module refcnt first. |
1265 | */ |
1266 | if (!try_module_get(pf->owner)) |
1267 | goto out_release; |
1268 | |
1269 | /* Now protected by module ref count */ |
1270 | rcu_read_unlock(); |
1271 | |
1272 | err = pf->create(net, sock, protocol, kern); |
1273 | if (err < 0) |
1274 | goto out_module_put; |
1275 | |
1276 | /* |
1277 | * Now to bump the refcnt of the [loadable] module that owns this |
1278 | * socket at sock_release time we decrement its refcnt. |
1279 | */ |
1280 | if (!try_module_get(sock->ops->owner)) |
1281 | goto out_module_busy; |
1282 | |
1283 | /* |
1284 | * Now that we're done with the ->create function, the [loadable] |
1285 | * module can have its refcnt decremented |
1286 | */ |
1287 | module_put(pf->owner); |
1288 | err = security_socket_post_create(sock, family, type, protocol, kern); |
1289 | if (err) |
1290 | goto out_sock_release; |
1291 | *res = sock; |
1292 | |
1293 | return 0; |
1294 | |
1295 | out_module_busy: |
1296 | err = -EAFNOSUPPORT; |
1297 | out_module_put: |
1298 | sock->ops = NULL; |
1299 | module_put(pf->owner); |
1300 | out_sock_release: |
1301 | sock_release(sock); |
1302 | return err; |
1303 | |
1304 | out_release: |
1305 | rcu_read_unlock(); |
1306 | goto out_sock_release; |
1307 | } |
1308 | EXPORT_SYMBOL(__sock_create); |
1309 | |
1310 | int sock_create(int family, int type, int protocol, struct socket **res) |
1311 | { |
1312 | return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); |
1313 | } |
1314 | EXPORT_SYMBOL(sock_create); |
1315 | |
1316 | int sock_create_kern(int family, int type, int protocol, struct socket **res) |
1317 | { |
1318 | return __sock_create(&init_net, family, type, protocol, res, 1); |
1319 | } |
1320 | EXPORT_SYMBOL(sock_create_kern); |
1321 | |
1322 | SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) |
1323 | { |
1324 | int retval; |
1325 | struct socket *sock; |
1326 | int flags; |
1327 | |
1328 | /* Check the SOCK_* constants for consistency. */ |
1329 | BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); |
1330 | BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); |
1331 | BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); |
1332 | BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); |
1333 | |
1334 | flags = type & ~SOCK_TYPE_MASK; |
1335 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
1336 | return -EINVAL; |
1337 | type &= SOCK_TYPE_MASK; |
1338 | |
1339 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1340 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; |
1341 | |
1342 | retval = sock_create(family, type, protocol, &sock); |
1343 | if (retval < 0) |
1344 | goto out; |
1345 | |
1346 | retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); |
1347 | if (retval < 0) |
1348 | goto out_release; |
1349 | |
1350 | out: |
1351 | /* It may be already another descriptor 8) Not kernel problem. */ |
1352 | return retval; |
1353 | |
1354 | out_release: |
1355 | sock_release(sock); |
1356 | return retval; |
1357 | } |
1358 | |
1359 | /* |
1360 | * Create a pair of connected sockets. |
1361 | */ |
1362 | |
1363 | SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, |
1364 | int __user *, usockvec) |
1365 | { |
1366 | struct socket *sock1, *sock2; |
1367 | int fd1, fd2, err; |
1368 | struct file *newfile1, *newfile2; |
1369 | int flags; |
1370 | |
1371 | flags = type & ~SOCK_TYPE_MASK; |
1372 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
1373 | return -EINVAL; |
1374 | type &= SOCK_TYPE_MASK; |
1375 | |
1376 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1377 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; |
1378 | |
1379 | /* |
1380 | * Obtain the first socket and check if the underlying protocol |
1381 | * supports the socketpair call. |
1382 | */ |
1383 | |
1384 | err = sock_create(family, type, protocol, &sock1); |
1385 | if (err < 0) |
1386 | goto out; |
1387 | |
1388 | err = sock_create(family, type, protocol, &sock2); |
1389 | if (err < 0) |
1390 | goto out_release_1; |
1391 | |
1392 | err = sock1->ops->socketpair(sock1, sock2); |
1393 | if (err < 0) |
1394 | goto out_release_both; |
1395 | |
1396 | fd1 = sock_alloc_file(sock1, &newfile1, flags); |
1397 | if (unlikely(fd1 < 0)) { |
1398 | err = fd1; |
1399 | goto out_release_both; |
1400 | } |
1401 | |
1402 | fd2 = sock_alloc_file(sock2, &newfile2, flags); |
1403 | if (unlikely(fd2 < 0)) { |
1404 | err = fd2; |
1405 | fput(newfile1); |
1406 | put_unused_fd(fd1); |
1407 | sock_release(sock2); |
1408 | goto out; |
1409 | } |
1410 | |
1411 | audit_fd_pair(fd1, fd2); |
1412 | fd_install(fd1, newfile1); |
1413 | fd_install(fd2, newfile2); |
1414 | /* fd1 and fd2 may be already another descriptors. |
1415 | * Not kernel problem. |
1416 | */ |
1417 | |
1418 | err = put_user(fd1, &usockvec[0]); |
1419 | if (!err) |
1420 | err = put_user(fd2, &usockvec[1]); |
1421 | if (!err) |
1422 | return 0; |
1423 | |
1424 | sys_close(fd2); |
1425 | sys_close(fd1); |
1426 | return err; |
1427 | |
1428 | out_release_both: |
1429 | sock_release(sock2); |
1430 | out_release_1: |
1431 | sock_release(sock1); |
1432 | out: |
1433 | return err; |
1434 | } |
1435 | |
1436 | /* |
1437 | * Bind a name to a socket. Nothing much to do here since it's |
1438 | * the protocol's responsibility to handle the local address. |
1439 | * |
1440 | * We move the socket address to kernel space before we call |
1441 | * the protocol layer (having also checked the address is ok). |
1442 | */ |
1443 | |
1444 | SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) |
1445 | { |
1446 | struct socket *sock; |
1447 | struct sockaddr_storage address; |
1448 | int err, fput_needed; |
1449 | |
1450 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1451 | if (sock) { |
1452 | err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address); |
1453 | if (err >= 0) { |
1454 | err = security_socket_bind(sock, |
1455 | (struct sockaddr *)&address, |
1456 | addrlen); |
1457 | if (!err) |
1458 | err = sock->ops->bind(sock, |
1459 | (struct sockaddr *) |
1460 | &address, addrlen); |
1461 | } |
1462 | fput_light(sock->file, fput_needed); |
1463 | } |
1464 | return err; |
1465 | } |
1466 | |
1467 | /* |
1468 | * Perform a listen. Basically, we allow the protocol to do anything |
1469 | * necessary for a listen, and if that works, we mark the socket as |
1470 | * ready for listening. |
1471 | */ |
1472 | |
1473 | SYSCALL_DEFINE2(listen, int, fd, int, backlog) |
1474 | { |
1475 | struct socket *sock; |
1476 | int err, fput_needed; |
1477 | int somaxconn; |
1478 | |
1479 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1480 | if (sock) { |
1481 | somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; |
1482 | if ((unsigned)backlog > somaxconn) |
1483 | backlog = somaxconn; |
1484 | |
1485 | err = security_socket_listen(sock, backlog); |
1486 | if (!err) |
1487 | err = sock->ops->listen(sock, backlog); |
1488 | |
1489 | fput_light(sock->file, fput_needed); |
1490 | } |
1491 | return err; |
1492 | } |
1493 | |
1494 | /* |
1495 | * For accept, we attempt to create a new socket, set up the link |
1496 | * with the client, wake up the client, then return the new |
1497 | * connected fd. We collect the address of the connector in kernel |
1498 | * space and move it to user at the very end. This is unclean because |
1499 | * we open the socket then return an error. |
1500 | * |
1501 | * 1003.1g adds the ability to recvmsg() to query connection pending |
1502 | * status to recvmsg. We need to add that support in a way thats |
1503 | * clean when we restucture accept also. |
1504 | */ |
1505 | |
1506 | SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, |
1507 | int __user *, upeer_addrlen, int, flags) |
1508 | { |
1509 | struct socket *sock, *newsock; |
1510 | struct file *newfile; |
1511 | int err, len, newfd, fput_needed; |
1512 | struct sockaddr_storage address; |
1513 | |
1514 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
1515 | return -EINVAL; |
1516 | |
1517 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1518 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; |
1519 | |
1520 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1521 | if (!sock) |
1522 | goto out; |
1523 | |
1524 | err = -ENFILE; |
1525 | newsock = sock_alloc(); |
1526 | if (!newsock) |
1527 | goto out_put; |
1528 | |
1529 | newsock->type = sock->type; |
1530 | newsock->ops = sock->ops; |
1531 | |
1532 | /* |
1533 | * We don't need try_module_get here, as the listening socket (sock) |
1534 | * has the protocol module (sock->ops->owner) held. |
1535 | */ |
1536 | __module_get(newsock->ops->owner); |
1537 | |
1538 | newfd = sock_alloc_file(newsock, &newfile, flags); |
1539 | if (unlikely(newfd < 0)) { |
1540 | err = newfd; |
1541 | sock_release(newsock); |
1542 | goto out_put; |
1543 | } |
1544 | |
1545 | err = security_socket_accept(sock, newsock); |
1546 | if (err) |
1547 | goto out_fd; |
1548 | |
1549 | err = sock->ops->accept(sock, newsock, sock->file->f_flags); |
1550 | if (err < 0) |
1551 | goto out_fd; |
1552 | |
1553 | if (upeer_sockaddr) { |
1554 | if (newsock->ops->getname(newsock, (struct sockaddr *)&address, |
1555 | &len, 2) < 0) { |
1556 | err = -ECONNABORTED; |
1557 | goto out_fd; |
1558 | } |
1559 | err = move_addr_to_user((struct sockaddr *)&address, |
1560 | len, upeer_sockaddr, upeer_addrlen); |
1561 | if (err < 0) |
1562 | goto out_fd; |
1563 | } |
1564 | |
1565 | /* File flags are not inherited via accept() unlike another OSes. */ |
1566 | |
1567 | fd_install(newfd, newfile); |
1568 | err = newfd; |
1569 | |
1570 | out_put: |
1571 | fput_light(sock->file, fput_needed); |
1572 | out: |
1573 | return err; |
1574 | out_fd: |
1575 | fput(newfile); |
1576 | put_unused_fd(newfd); |
1577 | goto out_put; |
1578 | } |
1579 | |
1580 | SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, |
1581 | int __user *, upeer_addrlen) |
1582 | { |
1583 | return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); |
1584 | } |
1585 | |
1586 | /* |
1587 | * Attempt to connect to a socket with the server address. The address |
1588 | * is in user space so we verify it is OK and move it to kernel space. |
1589 | * |
1590 | * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to |
1591 | * break bindings |
1592 | * |
1593 | * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and |
1594 | * other SEQPACKET protocols that take time to connect() as it doesn't |
1595 | * include the -EINPROGRESS status for such sockets. |
1596 | */ |
1597 | |
1598 | SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, |
1599 | int, addrlen) |
1600 | { |
1601 | struct socket *sock; |
1602 | struct sockaddr_storage address; |
1603 | int err, fput_needed; |
1604 | |
1605 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1606 | if (!sock) |
1607 | goto out; |
1608 | err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address); |
1609 | if (err < 0) |
1610 | goto out_put; |
1611 | |
1612 | err = |
1613 | security_socket_connect(sock, (struct sockaddr *)&address, addrlen); |
1614 | if (err) |
1615 | goto out_put; |
1616 | |
1617 | err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, |
1618 | sock->file->f_flags); |
1619 | out_put: |
1620 | fput_light(sock->file, fput_needed); |
1621 | out: |
1622 | return err; |
1623 | } |
1624 | |
1625 | /* |
1626 | * Get the local address ('name') of a socket object. Move the obtained |
1627 | * name to user space. |
1628 | */ |
1629 | |
1630 | SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, |
1631 | int __user *, usockaddr_len) |
1632 | { |
1633 | struct socket *sock; |
1634 | struct sockaddr_storage address; |
1635 | int len, err, fput_needed; |
1636 | |
1637 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1638 | if (!sock) |
1639 | goto out; |
1640 | |
1641 | err = security_socket_getsockname(sock); |
1642 | if (err) |
1643 | goto out_put; |
1644 | |
1645 | err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0); |
1646 | if (err) |
1647 | goto out_put; |
1648 | err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len); |
1649 | |
1650 | out_put: |
1651 | fput_light(sock->file, fput_needed); |
1652 | out: |
1653 | return err; |
1654 | } |
1655 | |
1656 | /* |
1657 | * Get the remote address ('name') of a socket object. Move the obtained |
1658 | * name to user space. |
1659 | */ |
1660 | |
1661 | SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, |
1662 | int __user *, usockaddr_len) |
1663 | { |
1664 | struct socket *sock; |
1665 | struct sockaddr_storage address; |
1666 | int len, err, fput_needed; |
1667 | |
1668 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1669 | if (sock != NULL) { |
1670 | err = security_socket_getpeername(sock); |
1671 | if (err) { |
1672 | fput_light(sock->file, fput_needed); |
1673 | return err; |
1674 | } |
1675 | |
1676 | err = |
1677 | sock->ops->getname(sock, (struct sockaddr *)&address, &len, |
1678 | 1); |
1679 | if (!err) |
1680 | err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, |
1681 | usockaddr_len); |
1682 | fput_light(sock->file, fput_needed); |
1683 | } |
1684 | return err; |
1685 | } |
1686 | |
1687 | /* |
1688 | * Send a datagram to a given address. We move the address into kernel |
1689 | * space and check the user space data area is readable before invoking |
1690 | * the protocol. |
1691 | */ |
1692 | |
1693 | SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, |
1694 | unsigned, flags, struct sockaddr __user *, addr, |
1695 | int, addr_len) |
1696 | { |
1697 | struct socket *sock; |
1698 | struct sockaddr_storage address; |
1699 | int err; |
1700 | struct msghdr msg; |
1701 | struct iovec iov; |
1702 | int fput_needed; |
1703 | |
1704 | if (len > INT_MAX) |
1705 | len = INT_MAX; |
1706 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1707 | if (!sock) |
1708 | goto out; |
1709 | |
1710 | iov.iov_base = buff; |
1711 | iov.iov_len = len; |
1712 | msg.msg_name = NULL; |
1713 | msg.msg_iov = &iov; |
1714 | msg.msg_iovlen = 1; |
1715 | msg.msg_control = NULL; |
1716 | msg.msg_controllen = 0; |
1717 | msg.msg_namelen = 0; |
1718 | if (addr) { |
1719 | err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address); |
1720 | if (err < 0) |
1721 | goto out_put; |
1722 | msg.msg_name = (struct sockaddr *)&address; |
1723 | msg.msg_namelen = addr_len; |
1724 | } |
1725 | if (sock->file->f_flags & O_NONBLOCK) |
1726 | flags |= MSG_DONTWAIT; |
1727 | msg.msg_flags = flags; |
1728 | err = sock_sendmsg(sock, &msg, len); |
1729 | |
1730 | out_put: |
1731 | fput_light(sock->file, fput_needed); |
1732 | out: |
1733 | return err; |
1734 | } |
1735 | |
1736 | /* |
1737 | * Send a datagram down a socket. |
1738 | */ |
1739 | |
1740 | SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, |
1741 | unsigned, flags) |
1742 | { |
1743 | return sys_sendto(fd, buff, len, flags, NULL, 0); |
1744 | } |
1745 | |
1746 | /* |
1747 | * Receive a frame from the socket and optionally record the address of the |
1748 | * sender. We verify the buffers are writable and if needed move the |
1749 | * sender address from kernel to user space. |
1750 | */ |
1751 | |
1752 | SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, |
1753 | unsigned, flags, struct sockaddr __user *, addr, |
1754 | int __user *, addr_len) |
1755 | { |
1756 | struct socket *sock; |
1757 | struct iovec iov; |
1758 | struct msghdr msg; |
1759 | struct sockaddr_storage address; |
1760 | int err, err2; |
1761 | int fput_needed; |
1762 | |
1763 | if (size > INT_MAX) |
1764 | size = INT_MAX; |
1765 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1766 | if (!sock) |
1767 | goto out; |
1768 | |
1769 | msg.msg_control = NULL; |
1770 | msg.msg_controllen = 0; |
1771 | msg.msg_iovlen = 1; |
1772 | msg.msg_iov = &iov; |
1773 | iov.iov_len = size; |
1774 | iov.iov_base = ubuf; |
1775 | msg.msg_name = (struct sockaddr *)&address; |
1776 | msg.msg_namelen = sizeof(address); |
1777 | if (sock->file->f_flags & O_NONBLOCK) |
1778 | flags |= MSG_DONTWAIT; |
1779 | err = sock_recvmsg(sock, &msg, size, flags); |
1780 | |
1781 | if (err >= 0 && addr != NULL) { |
1782 | err2 = move_addr_to_user((struct sockaddr *)&address, |
1783 | msg.msg_namelen, addr, addr_len); |
1784 | if (err2 < 0) |
1785 | err = err2; |
1786 | } |
1787 | |
1788 | fput_light(sock->file, fput_needed); |
1789 | out: |
1790 | return err; |
1791 | } |
1792 | |
1793 | /* |
1794 | * Receive a datagram from a socket. |
1795 | */ |
1796 | |
1797 | asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size, |
1798 | unsigned flags) |
1799 | { |
1800 | return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); |
1801 | } |
1802 | |
1803 | /* |
1804 | * Set a socket option. Because we don't know the option lengths we have |
1805 | * to pass the user mode parameter for the protocols to sort out. |
1806 | */ |
1807 | |
1808 | SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, |
1809 | char __user *, optval, int, optlen) |
1810 | { |
1811 | int err, fput_needed; |
1812 | struct socket *sock; |
1813 | |
1814 | if (optlen < 0) |
1815 | return -EINVAL; |
1816 | |
1817 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1818 | if (sock != NULL) { |
1819 | err = security_socket_setsockopt(sock, level, optname); |
1820 | if (err) |
1821 | goto out_put; |
1822 | |
1823 | if (level == SOL_SOCKET) |
1824 | err = |
1825 | sock_setsockopt(sock, level, optname, optval, |
1826 | optlen); |
1827 | else |
1828 | err = |
1829 | sock->ops->setsockopt(sock, level, optname, optval, |
1830 | optlen); |
1831 | out_put: |
1832 | fput_light(sock->file, fput_needed); |
1833 | } |
1834 | return err; |
1835 | } |
1836 | |
1837 | /* |
1838 | * Get a socket option. Because we don't know the option lengths we have |
1839 | * to pass a user mode parameter for the protocols to sort out. |
1840 | */ |
1841 | |
1842 | SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, |
1843 | char __user *, optval, int __user *, optlen) |
1844 | { |
1845 | int err, fput_needed; |
1846 | struct socket *sock; |
1847 | |
1848 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1849 | if (sock != NULL) { |
1850 | err = security_socket_getsockopt(sock, level, optname); |
1851 | if (err) |
1852 | goto out_put; |
1853 | |
1854 | if (level == SOL_SOCKET) |
1855 | err = |
1856 | sock_getsockopt(sock, level, optname, optval, |
1857 | optlen); |
1858 | else |
1859 | err = |
1860 | sock->ops->getsockopt(sock, level, optname, optval, |
1861 | optlen); |
1862 | out_put: |
1863 | fput_light(sock->file, fput_needed); |
1864 | } |
1865 | return err; |
1866 | } |
1867 | |
1868 | /* |
1869 | * Shutdown a socket. |
1870 | */ |
1871 | |
1872 | SYSCALL_DEFINE2(shutdown, int, fd, int, how) |
1873 | { |
1874 | int err, fput_needed; |
1875 | struct socket *sock; |
1876 | |
1877 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1878 | if (sock != NULL) { |
1879 | err = security_socket_shutdown(sock, how); |
1880 | if (!err) |
1881 | err = sock->ops->shutdown(sock, how); |
1882 | fput_light(sock->file, fput_needed); |
1883 | } |
1884 | return err; |
1885 | } |
1886 | |
1887 | /* A couple of helpful macros for getting the address of the 32/64 bit |
1888 | * fields which are the same type (int / unsigned) on our platforms. |
1889 | */ |
1890 | #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) |
1891 | #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) |
1892 | #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) |
1893 | |
1894 | struct used_address { |
1895 | struct sockaddr_storage name; |
1896 | unsigned int name_len; |
1897 | }; |
1898 | |
1899 | static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg, |
1900 | struct msghdr *msg_sys, unsigned flags, |
1901 | struct used_address *used_address) |
1902 | { |
1903 | struct compat_msghdr __user *msg_compat = |
1904 | (struct compat_msghdr __user *)msg; |
1905 | struct sockaddr_storage address; |
1906 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; |
1907 | unsigned char ctl[sizeof(struct cmsghdr) + 20] |
1908 | __attribute__ ((aligned(sizeof(__kernel_size_t)))); |
1909 | /* 20 is size of ipv6_pktinfo */ |
1910 | unsigned char *ctl_buf = ctl; |
1911 | int err, ctl_len, iov_size, total_len; |
1912 | |
1913 | err = -EFAULT; |
1914 | if (MSG_CMSG_COMPAT & flags) { |
1915 | if (get_compat_msghdr(msg_sys, msg_compat)) |
1916 | return -EFAULT; |
1917 | } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr))) |
1918 | return -EFAULT; |
1919 | |
1920 | /* do not move before msg_sys is valid */ |
1921 | err = -EMSGSIZE; |
1922 | if (msg_sys->msg_iovlen > UIO_MAXIOV) |
1923 | goto out; |
1924 | |
1925 | /* Check whether to allocate the iovec area */ |
1926 | err = -ENOMEM; |
1927 | iov_size = msg_sys->msg_iovlen * sizeof(struct iovec); |
1928 | if (msg_sys->msg_iovlen > UIO_FASTIOV) { |
1929 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); |
1930 | if (!iov) |
1931 | goto out; |
1932 | } |
1933 | |
1934 | /* This will also move the address data into kernel space */ |
1935 | if (MSG_CMSG_COMPAT & flags) { |
1936 | err = verify_compat_iovec(msg_sys, iov, |
1937 | (struct sockaddr *)&address, |
1938 | VERIFY_READ); |
1939 | } else |
1940 | err = verify_iovec(msg_sys, iov, |
1941 | (struct sockaddr *)&address, |
1942 | VERIFY_READ); |
1943 | if (err < 0) |
1944 | goto out_freeiov; |
1945 | total_len = err; |
1946 | |
1947 | err = -ENOBUFS; |
1948 | |
1949 | if (msg_sys->msg_controllen > INT_MAX) |
1950 | goto out_freeiov; |
1951 | ctl_len = msg_sys->msg_controllen; |
1952 | if ((MSG_CMSG_COMPAT & flags) && ctl_len) { |
1953 | err = |
1954 | cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, |
1955 | sizeof(ctl)); |
1956 | if (err) |
1957 | goto out_freeiov; |
1958 | ctl_buf = msg_sys->msg_control; |
1959 | ctl_len = msg_sys->msg_controllen; |
1960 | } else if (ctl_len) { |
1961 | if (ctl_len > sizeof(ctl)) { |
1962 | ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); |
1963 | if (ctl_buf == NULL) |
1964 | goto out_freeiov; |
1965 | } |
1966 | err = -EFAULT; |
1967 | /* |
1968 | * Careful! Before this, msg_sys->msg_control contains a user pointer. |
1969 | * Afterwards, it will be a kernel pointer. Thus the compiler-assisted |
1970 | * checking falls down on this. |
1971 | */ |
1972 | if (copy_from_user(ctl_buf, |
1973 | (void __user __force *)msg_sys->msg_control, |
1974 | ctl_len)) |
1975 | goto out_freectl; |
1976 | msg_sys->msg_control = ctl_buf; |
1977 | } |
1978 | msg_sys->msg_flags = flags; |
1979 | |
1980 | if (sock->file->f_flags & O_NONBLOCK) |
1981 | msg_sys->msg_flags |= MSG_DONTWAIT; |
1982 | /* |
1983 | * If this is sendmmsg() and current destination address is same as |
1984 | * previously succeeded address, omit asking LSM's decision. |
1985 | * used_address->name_len is initialized to UINT_MAX so that the first |
1986 | * destination address never matches. |
1987 | */ |
1988 | if (used_address && msg_sys->msg_name && |
1989 | used_address->name_len == msg_sys->msg_namelen && |
1990 | !memcmp(&used_address->name, msg_sys->msg_name, |
1991 | used_address->name_len)) { |
1992 | err = sock_sendmsg_nosec(sock, msg_sys, total_len); |
1993 | goto out_freectl; |
1994 | } |
1995 | err = sock_sendmsg(sock, msg_sys, total_len); |
1996 | /* |
1997 | * If this is sendmmsg() and sending to current destination address was |
1998 | * successful, remember it. |
1999 | */ |
2000 | if (used_address && err >= 0) { |
2001 | used_address->name_len = msg_sys->msg_namelen; |
2002 | if (msg_sys->msg_name) |
2003 | memcpy(&used_address->name, msg_sys->msg_name, |
2004 | used_address->name_len); |
2005 | } |
2006 | |
2007 | out_freectl: |
2008 | if (ctl_buf != ctl) |
2009 | sock_kfree_s(sock->sk, ctl_buf, ctl_len); |
2010 | out_freeiov: |
2011 | if (iov != iovstack) |
2012 | sock_kfree_s(sock->sk, iov, iov_size); |
2013 | out: |
2014 | return err; |
2015 | } |
2016 | |
2017 | /* |
2018 | * BSD sendmsg interface |
2019 | */ |
2020 | |
2021 | SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags) |
2022 | { |
2023 | int fput_needed, err; |
2024 | struct msghdr msg_sys; |
2025 | struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2026 | |
2027 | if (!sock) |
2028 | goto out; |
2029 | |
2030 | err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL); |
2031 | |
2032 | fput_light(sock->file, fput_needed); |
2033 | out: |
2034 | return err; |
2035 | } |
2036 | |
2037 | /* |
2038 | * Linux sendmmsg interface |
2039 | */ |
2040 | |
2041 | int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, |
2042 | unsigned int flags) |
2043 | { |
2044 | int fput_needed, err, datagrams; |
2045 | struct socket *sock; |
2046 | struct mmsghdr __user *entry; |
2047 | struct compat_mmsghdr __user *compat_entry; |
2048 | struct msghdr msg_sys; |
2049 | struct used_address used_address; |
2050 | |
2051 | if (vlen > UIO_MAXIOV) |
2052 | vlen = UIO_MAXIOV; |
2053 | |
2054 | datagrams = 0; |
2055 | |
2056 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2057 | if (!sock) |
2058 | return err; |
2059 | |
2060 | used_address.name_len = UINT_MAX; |
2061 | entry = mmsg; |
2062 | compat_entry = (struct compat_mmsghdr __user *)mmsg; |
2063 | err = 0; |
2064 | |
2065 | while (datagrams < vlen) { |
2066 | if (MSG_CMSG_COMPAT & flags) { |
2067 | err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry, |
2068 | &msg_sys, flags, &used_address); |
2069 | if (err < 0) |
2070 | break; |
2071 | err = __put_user(err, &compat_entry->msg_len); |
2072 | ++compat_entry; |
2073 | } else { |
2074 | err = __sys_sendmsg(sock, (struct msghdr __user *)entry, |
2075 | &msg_sys, flags, &used_address); |
2076 | if (err < 0) |
2077 | break; |
2078 | err = put_user(err, &entry->msg_len); |
2079 | ++entry; |
2080 | } |
2081 | |
2082 | if (err) |
2083 | break; |
2084 | ++datagrams; |
2085 | } |
2086 | |
2087 | fput_light(sock->file, fput_needed); |
2088 | |
2089 | /* We only return an error if no datagrams were able to be sent */ |
2090 | if (datagrams != 0) |
2091 | return datagrams; |
2092 | |
2093 | return err; |
2094 | } |
2095 | |
2096 | SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, |
2097 | unsigned int, vlen, unsigned int, flags) |
2098 | { |
2099 | return __sys_sendmmsg(fd, mmsg, vlen, flags); |
2100 | } |
2101 | |
2102 | static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg, |
2103 | struct msghdr *msg_sys, unsigned flags, int nosec) |
2104 | { |
2105 | struct compat_msghdr __user *msg_compat = |
2106 | (struct compat_msghdr __user *)msg; |
2107 | struct iovec iovstack[UIO_FASTIOV]; |
2108 | struct iovec *iov = iovstack; |
2109 | unsigned long cmsg_ptr; |
2110 | int err, iov_size, total_len, len; |
2111 | |
2112 | /* kernel mode address */ |
2113 | struct sockaddr_storage addr; |
2114 | |
2115 | /* user mode address pointers */ |
2116 | struct sockaddr __user *uaddr; |
2117 | int __user *uaddr_len; |
2118 | |
2119 | if (MSG_CMSG_COMPAT & flags) { |
2120 | if (get_compat_msghdr(msg_sys, msg_compat)) |
2121 | return -EFAULT; |
2122 | } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr))) |
2123 | return -EFAULT; |
2124 | |
2125 | err = -EMSGSIZE; |
2126 | if (msg_sys->msg_iovlen > UIO_MAXIOV) |
2127 | goto out; |
2128 | |
2129 | /* Check whether to allocate the iovec area */ |
2130 | err = -ENOMEM; |
2131 | iov_size = msg_sys->msg_iovlen * sizeof(struct iovec); |
2132 | if (msg_sys->msg_iovlen > UIO_FASTIOV) { |
2133 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); |
2134 | if (!iov) |
2135 | goto out; |
2136 | } |
2137 | |
2138 | /* |
2139 | * Save the user-mode address (verify_iovec will change the |
2140 | * kernel msghdr to use the kernel address space) |
2141 | */ |
2142 | |
2143 | uaddr = (__force void __user *)msg_sys->msg_name; |
2144 | uaddr_len = COMPAT_NAMELEN(msg); |
2145 | if (MSG_CMSG_COMPAT & flags) { |
2146 | err = verify_compat_iovec(msg_sys, iov, |
2147 | (struct sockaddr *)&addr, |
2148 | VERIFY_WRITE); |
2149 | } else |
2150 | err = verify_iovec(msg_sys, iov, |
2151 | (struct sockaddr *)&addr, |
2152 | VERIFY_WRITE); |
2153 | if (err < 0) |
2154 | goto out_freeiov; |
2155 | total_len = err; |
2156 | |
2157 | cmsg_ptr = (unsigned long)msg_sys->msg_control; |
2158 | msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); |
2159 | |
2160 | if (sock->file->f_flags & O_NONBLOCK) |
2161 | flags |= MSG_DONTWAIT; |
2162 | err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, |
2163 | total_len, flags); |
2164 | if (err < 0) |
2165 | goto out_freeiov; |
2166 | len = err; |
2167 | |
2168 | if (uaddr != NULL) { |
2169 | err = move_addr_to_user((struct sockaddr *)&addr, |
2170 | msg_sys->msg_namelen, uaddr, |
2171 | uaddr_len); |
2172 | if (err < 0) |
2173 | goto out_freeiov; |
2174 | } |
2175 | err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), |
2176 | COMPAT_FLAGS(msg)); |
2177 | if (err) |
2178 | goto out_freeiov; |
2179 | if (MSG_CMSG_COMPAT & flags) |
2180 | err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, |
2181 | &msg_compat->msg_controllen); |
2182 | else |
2183 | err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, |
2184 | &msg->msg_controllen); |
2185 | if (err) |
2186 | goto out_freeiov; |
2187 | err = len; |
2188 | |
2189 | out_freeiov: |
2190 | if (iov != iovstack) |
2191 | sock_kfree_s(sock->sk, iov, iov_size); |
2192 | out: |
2193 | return err; |
2194 | } |
2195 | |
2196 | /* |
2197 | * BSD recvmsg interface |
2198 | */ |
2199 | |
2200 | SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg, |
2201 | unsigned int, flags) |
2202 | { |
2203 | int fput_needed, err; |
2204 | struct msghdr msg_sys; |
2205 | struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2206 | |
2207 | if (!sock) |
2208 | goto out; |
2209 | |
2210 | err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0); |
2211 | |
2212 | fput_light(sock->file, fput_needed); |
2213 | out: |
2214 | return err; |
2215 | } |
2216 | |
2217 | /* |
2218 | * Linux recvmmsg interface |
2219 | */ |
2220 | |
2221 | int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, |
2222 | unsigned int flags, struct timespec *timeout) |
2223 | { |
2224 | int fput_needed, err, datagrams; |
2225 | struct socket *sock; |
2226 | struct mmsghdr __user *entry; |
2227 | struct compat_mmsghdr __user *compat_entry; |
2228 | struct msghdr msg_sys; |
2229 | struct timespec end_time; |
2230 | |
2231 | if (timeout && |
2232 | poll_select_set_timeout(&end_time, timeout->tv_sec, |
2233 | timeout->tv_nsec)) |
2234 | return -EINVAL; |
2235 | |
2236 | datagrams = 0; |
2237 | |
2238 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2239 | if (!sock) |
2240 | return err; |
2241 | |
2242 | err = sock_error(sock->sk); |
2243 | if (err) |
2244 | goto out_put; |
2245 | |
2246 | entry = mmsg; |
2247 | compat_entry = (struct compat_mmsghdr __user *)mmsg; |
2248 | |
2249 | while (datagrams < vlen) { |
2250 | /* |
2251 | * No need to ask LSM for more than the first datagram. |
2252 | */ |
2253 | if (MSG_CMSG_COMPAT & flags) { |
2254 | err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry, |
2255 | &msg_sys, flags & ~MSG_WAITFORONE, |
2256 | datagrams); |
2257 | if (err < 0) |
2258 | break; |
2259 | err = __put_user(err, &compat_entry->msg_len); |
2260 | ++compat_entry; |
2261 | } else { |
2262 | err = __sys_recvmsg(sock, (struct msghdr __user *)entry, |
2263 | &msg_sys, flags & ~MSG_WAITFORONE, |
2264 | datagrams); |
2265 | if (err < 0) |
2266 | break; |
2267 | err = put_user(err, &entry->msg_len); |
2268 | ++entry; |
2269 | } |
2270 | |
2271 | if (err) |
2272 | break; |
2273 | ++datagrams; |
2274 | |
2275 | /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ |
2276 | if (flags & MSG_WAITFORONE) |
2277 | flags |= MSG_DONTWAIT; |
2278 | |
2279 | if (timeout) { |
2280 | ktime_get_ts(timeout); |
2281 | *timeout = timespec_sub(end_time, *timeout); |
2282 | if (timeout->tv_sec < 0) { |
2283 | timeout->tv_sec = timeout->tv_nsec = 0; |
2284 | break; |
2285 | } |
2286 | |
2287 | /* Timeout, return less than vlen datagrams */ |
2288 | if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) |
2289 | break; |
2290 | } |
2291 | |
2292 | /* Out of band data, return right away */ |
2293 | if (msg_sys.msg_flags & MSG_OOB) |
2294 | break; |
2295 | } |
2296 | |
2297 | out_put: |
2298 | fput_light(sock->file, fput_needed); |
2299 | |
2300 | if (err == 0) |
2301 | return datagrams; |
2302 | |
2303 | if (datagrams != 0) { |
2304 | /* |
2305 | * We may return less entries than requested (vlen) if the |
2306 | * sock is non block and there aren't enough datagrams... |
2307 | */ |
2308 | if (err != -EAGAIN) { |
2309 | /* |
2310 | * ... or if recvmsg returns an error after we |
2311 | * received some datagrams, where we record the |
2312 | * error to return on the next call or if the |
2313 | * app asks about it using getsockopt(SO_ERROR). |
2314 | */ |
2315 | sock->sk->sk_err = -err; |
2316 | } |
2317 | |
2318 | return datagrams; |
2319 | } |
2320 | |
2321 | return err; |
2322 | } |
2323 | |
2324 | SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, |
2325 | unsigned int, vlen, unsigned int, flags, |
2326 | struct timespec __user *, timeout) |
2327 | { |
2328 | int datagrams; |
2329 | struct timespec timeout_sys; |
2330 | |
2331 | if (!timeout) |
2332 | return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL); |
2333 | |
2334 | if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys))) |
2335 | return -EFAULT; |
2336 | |
2337 | datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); |
2338 | |
2339 | if (datagrams > 0 && |
2340 | copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys))) |
2341 | datagrams = -EFAULT; |
2342 | |
2343 | return datagrams; |
2344 | } |
2345 | |
2346 | #ifdef __ARCH_WANT_SYS_SOCKETCALL |
2347 | /* Argument list sizes for sys_socketcall */ |
2348 | #define AL(x) ((x) * sizeof(unsigned long)) |
2349 | static const unsigned char nargs[21] = { |
2350 | AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), |
2351 | AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), |
2352 | AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), |
2353 | AL(4), AL(5), AL(4) |
2354 | }; |
2355 | |
2356 | #undef AL |
2357 | |
2358 | /* |
2359 | * System call vectors. |
2360 | * |
2361 | * Argument checking cleaned up. Saved 20% in size. |
2362 | * This function doesn't need to set the kernel lock because |
2363 | * it is set by the callees. |
2364 | */ |
2365 | |
2366 | SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) |
2367 | { |
2368 | unsigned long a[6]; |
2369 | unsigned long a0, a1; |
2370 | int err; |
2371 | unsigned int len; |
2372 | |
2373 | if (call < 1 || call > SYS_SENDMMSG) |
2374 | return -EINVAL; |
2375 | |
2376 | len = nargs[call]; |
2377 | if (len > sizeof(a)) |
2378 | return -EINVAL; |
2379 | |
2380 | /* copy_from_user should be SMP safe. */ |
2381 | if (copy_from_user(a, args, len)) |
2382 | return -EFAULT; |
2383 | |
2384 | audit_socketcall(nargs[call] / sizeof(unsigned long), a); |
2385 | |
2386 | a0 = a[0]; |
2387 | a1 = a[1]; |
2388 | |
2389 | switch (call) { |
2390 | case SYS_SOCKET: |
2391 | err = sys_socket(a0, a1, a[2]); |
2392 | break; |
2393 | case SYS_BIND: |
2394 | err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]); |
2395 | break; |
2396 | case SYS_CONNECT: |
2397 | err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]); |
2398 | break; |
2399 | case SYS_LISTEN: |
2400 | err = sys_listen(a0, a1); |
2401 | break; |
2402 | case SYS_ACCEPT: |
2403 | err = sys_accept4(a0, (struct sockaddr __user *)a1, |
2404 | (int __user *)a[2], 0); |
2405 | break; |
2406 | case SYS_GETSOCKNAME: |
2407 | err = |
2408 | sys_getsockname(a0, (struct sockaddr __user *)a1, |
2409 | (int __user *)a[2]); |
2410 | break; |
2411 | case SYS_GETPEERNAME: |
2412 | err = |
2413 | sys_getpeername(a0, (struct sockaddr __user *)a1, |
2414 | (int __user *)a[2]); |
2415 | break; |
2416 | case SYS_SOCKETPAIR: |
2417 | err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]); |
2418 | break; |
2419 | case SYS_SEND: |
2420 | err = sys_send(a0, (void __user *)a1, a[2], a[3]); |
2421 | break; |
2422 | case SYS_SENDTO: |
2423 | err = sys_sendto(a0, (void __user *)a1, a[2], a[3], |
2424 | (struct sockaddr __user *)a[4], a[5]); |
2425 | break; |
2426 | case SYS_RECV: |
2427 | err = sys_recv(a0, (void __user *)a1, a[2], a[3]); |
2428 | break; |
2429 | case SYS_RECVFROM: |
2430 | err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3], |
2431 | (struct sockaddr __user *)a[4], |
2432 | (int __user *)a[5]); |
2433 | break; |
2434 | case SYS_SHUTDOWN: |
2435 | err = sys_shutdown(a0, a1); |
2436 | break; |
2437 | case SYS_SETSOCKOPT: |
2438 | err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]); |
2439 | break; |
2440 | case SYS_GETSOCKOPT: |
2441 | err = |
2442 | sys_getsockopt(a0, a1, a[2], (char __user *)a[3], |
2443 | (int __user *)a[4]); |
2444 | break; |
2445 | case SYS_SENDMSG: |
2446 | err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]); |
2447 | break; |
2448 | case SYS_SENDMMSG: |
2449 | err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]); |
2450 | break; |
2451 | case SYS_RECVMSG: |
2452 | err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]); |
2453 | break; |
2454 | case SYS_RECVMMSG: |
2455 | err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3], |
2456 | (struct timespec __user *)a[4]); |
2457 | break; |
2458 | case SYS_ACCEPT4: |
2459 | err = sys_accept4(a0, (struct sockaddr __user *)a1, |
2460 | (int __user *)a[2], a[3]); |
2461 | break; |
2462 | default: |
2463 | err = -EINVAL; |
2464 | break; |
2465 | } |
2466 | return err; |
2467 | } |
2468 | |
2469 | #endif /* __ARCH_WANT_SYS_SOCKETCALL */ |
2470 | |
2471 | /** |
2472 | * sock_register - add a socket protocol handler |
2473 | * @ops: description of protocol |
2474 | * |
2475 | * This function is called by a protocol handler that wants to |
2476 | * advertise its address family, and have it linked into the |
2477 | * socket interface. The value ops->family coresponds to the |
2478 | * socket system call protocol family. |
2479 | */ |
2480 | int sock_register(const struct net_proto_family *ops) |
2481 | { |
2482 | int err; |
2483 | |
2484 | if (ops->family >= NPROTO) { |
2485 | printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, |
2486 | NPROTO); |
2487 | return -ENOBUFS; |
2488 | } |
2489 | |
2490 | spin_lock(&net_family_lock); |
2491 | if (rcu_dereference_protected(net_families[ops->family], |
2492 | lockdep_is_held(&net_family_lock))) |
2493 | err = -EEXIST; |
2494 | else { |
2495 | rcu_assign_pointer(net_families[ops->family], ops); |
2496 | err = 0; |
2497 | } |
2498 | spin_unlock(&net_family_lock); |
2499 | |
2500 | printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family); |
2501 | return err; |
2502 | } |
2503 | EXPORT_SYMBOL(sock_register); |
2504 | |
2505 | /** |
2506 | * sock_unregister - remove a protocol handler |
2507 | * @family: protocol family to remove |
2508 | * |
2509 | * This function is called by a protocol handler that wants to |
2510 | * remove its address family, and have it unlinked from the |
2511 | * new socket creation. |
2512 | * |
2513 | * If protocol handler is a module, then it can use module reference |
2514 | * counts to protect against new references. If protocol handler is not |
2515 | * a module then it needs to provide its own protection in |
2516 | * the ops->create routine. |
2517 | */ |
2518 | void sock_unregister(int family) |
2519 | { |
2520 | BUG_ON(family < 0 || family >= NPROTO); |
2521 | |
2522 | spin_lock(&net_family_lock); |
2523 | RCU_INIT_POINTER(net_families[family], NULL); |
2524 | spin_unlock(&net_family_lock); |
2525 | |
2526 | synchronize_rcu(); |
2527 | |
2528 | printk(KERN_INFO "NET: Unregistered protocol family %d\n", family); |
2529 | } |
2530 | EXPORT_SYMBOL(sock_unregister); |
2531 | |
2532 | static int __init sock_init(void) |
2533 | { |
2534 | int err; |
2535 | |
2536 | /* |
2537 | * Initialize sock SLAB cache. |
2538 | */ |
2539 | |
2540 | sk_init(); |
2541 | |
2542 | /* |
2543 | * Initialize skbuff SLAB cache |
2544 | */ |
2545 | skb_init(); |
2546 | |
2547 | /* |
2548 | * Initialize the protocols module. |
2549 | */ |
2550 | |
2551 | init_inodecache(); |
2552 | |
2553 | err = register_filesystem(&sock_fs_type); |
2554 | if (err) |
2555 | goto out_fs; |
2556 | sock_mnt = kern_mount(&sock_fs_type); |
2557 | if (IS_ERR(sock_mnt)) { |
2558 | err = PTR_ERR(sock_mnt); |
2559 | goto out_mount; |
2560 | } |
2561 | |
2562 | /* The real protocol initialization is performed in later initcalls. |
2563 | */ |
2564 | |
2565 | #ifdef CONFIG_NETFILTER |
2566 | netfilter_init(); |
2567 | #endif |
2568 | |
2569 | #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING |
2570 | skb_timestamping_init(); |
2571 | #endif |
2572 | |
2573 | out: |
2574 | return err; |
2575 | |
2576 | out_mount: |
2577 | unregister_filesystem(&sock_fs_type); |
2578 | out_fs: |
2579 | goto out; |
2580 | } |
2581 | |
2582 | core_initcall(sock_init); /* early initcall */ |
2583 | |
2584 | #ifdef CONFIG_PROC_FS |
2585 | void socket_seq_show(struct seq_file *seq) |
2586 | { |
2587 | int cpu; |
2588 | int counter = 0; |
2589 | |
2590 | for_each_possible_cpu(cpu) |
2591 | counter += per_cpu(sockets_in_use, cpu); |
2592 | |
2593 | /* It can be negative, by the way. 8) */ |
2594 | if (counter < 0) |
2595 | counter = 0; |
2596 | |
2597 | seq_printf(seq, "sockets: used %d\n", counter); |
2598 | } |
2599 | #endif /* CONFIG_PROC_FS */ |
2600 | |
2601 | #ifdef CONFIG_COMPAT |
2602 | static int do_siocgstamp(struct net *net, struct socket *sock, |
2603 | unsigned int cmd, struct compat_timeval __user *up) |
2604 | { |
2605 | mm_segment_t old_fs = get_fs(); |
2606 | struct timeval ktv; |
2607 | int err; |
2608 | |
2609 | set_fs(KERNEL_DS); |
2610 | err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv); |
2611 | set_fs(old_fs); |
2612 | if (!err) { |
2613 | err = put_user(ktv.tv_sec, &up->tv_sec); |
2614 | err |= __put_user(ktv.tv_usec, &up->tv_usec); |
2615 | } |
2616 | return err; |
2617 | } |
2618 | |
2619 | static int do_siocgstampns(struct net *net, struct socket *sock, |
2620 | unsigned int cmd, struct compat_timespec __user *up) |
2621 | { |
2622 | mm_segment_t old_fs = get_fs(); |
2623 | struct timespec kts; |
2624 | int err; |
2625 | |
2626 | set_fs(KERNEL_DS); |
2627 | err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts); |
2628 | set_fs(old_fs); |
2629 | if (!err) { |
2630 | err = put_user(kts.tv_sec, &up->tv_sec); |
2631 | err |= __put_user(kts.tv_nsec, &up->tv_nsec); |
2632 | } |
2633 | return err; |
2634 | } |
2635 | |
2636 | static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32) |
2637 | { |
2638 | struct ifreq __user *uifr; |
2639 | int err; |
2640 | |
2641 | uifr = compat_alloc_user_space(sizeof(struct ifreq)); |
2642 | if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) |
2643 | return -EFAULT; |
2644 | |
2645 | err = dev_ioctl(net, SIOCGIFNAME, uifr); |
2646 | if (err) |
2647 | return err; |
2648 | |
2649 | if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq))) |
2650 | return -EFAULT; |
2651 | |
2652 | return 0; |
2653 | } |
2654 | |
2655 | static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32) |
2656 | { |
2657 | struct compat_ifconf ifc32; |
2658 | struct ifconf ifc; |
2659 | struct ifconf __user *uifc; |
2660 | struct compat_ifreq __user *ifr32; |
2661 | struct ifreq __user *ifr; |
2662 | unsigned int i, j; |
2663 | int err; |
2664 | |
2665 | if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf))) |
2666 | return -EFAULT; |
2667 | |
2668 | if (ifc32.ifcbuf == 0) { |
2669 | ifc32.ifc_len = 0; |
2670 | ifc.ifc_len = 0; |
2671 | ifc.ifc_req = NULL; |
2672 | uifc = compat_alloc_user_space(sizeof(struct ifconf)); |
2673 | } else { |
2674 | size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) * |
2675 | sizeof(struct ifreq); |
2676 | uifc = compat_alloc_user_space(sizeof(struct ifconf) + len); |
2677 | ifc.ifc_len = len; |
2678 | ifr = ifc.ifc_req = (void __user *)(uifc + 1); |
2679 | ifr32 = compat_ptr(ifc32.ifcbuf); |
2680 | for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) { |
2681 | if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq))) |
2682 | return -EFAULT; |
2683 | ifr++; |
2684 | ifr32++; |
2685 | } |
2686 | } |
2687 | if (copy_to_user(uifc, &ifc, sizeof(struct ifconf))) |
2688 | return -EFAULT; |
2689 | |
2690 | err = dev_ioctl(net, SIOCGIFCONF, uifc); |
2691 | if (err) |
2692 | return err; |
2693 | |
2694 | if (copy_from_user(&ifc, uifc, sizeof(struct ifconf))) |
2695 | return -EFAULT; |
2696 | |
2697 | ifr = ifc.ifc_req; |
2698 | ifr32 = compat_ptr(ifc32.ifcbuf); |
2699 | for (i = 0, j = 0; |
2700 | i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len; |
2701 | i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) { |
2702 | if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq))) |
2703 | return -EFAULT; |
2704 | ifr32++; |
2705 | ifr++; |
2706 | } |
2707 | |
2708 | if (ifc32.ifcbuf == 0) { |
2709 | /* Translate from 64-bit structure multiple to |
2710 | * a 32-bit one. |
2711 | */ |
2712 | i = ifc.ifc_len; |
2713 | i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq)); |
2714 | ifc32.ifc_len = i; |
2715 | } else { |
2716 | ifc32.ifc_len = i; |
2717 | } |
2718 | if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf))) |
2719 | return -EFAULT; |
2720 | |
2721 | return 0; |
2722 | } |
2723 | |
2724 | static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32) |
2725 | { |
2726 | struct compat_ethtool_rxnfc __user *compat_rxnfc; |
2727 | bool convert_in = false, convert_out = false; |
2728 | size_t buf_size = ALIGN(sizeof(struct ifreq), 8); |
2729 | struct ethtool_rxnfc __user *rxnfc; |
2730 | struct ifreq __user *ifr; |
2731 | u32 rule_cnt = 0, actual_rule_cnt; |
2732 | u32 ethcmd; |
2733 | u32 data; |
2734 | int ret; |
2735 | |
2736 | if (get_user(data, &ifr32->ifr_ifru.ifru_data)) |
2737 | return -EFAULT; |
2738 | |
2739 | compat_rxnfc = compat_ptr(data); |
2740 | |
2741 | if (get_user(ethcmd, &compat_rxnfc->cmd)) |
2742 | return -EFAULT; |
2743 | |
2744 | /* Most ethtool structures are defined without padding. |
2745 | * Unfortunately struct ethtool_rxnfc is an exception. |
2746 | */ |
2747 | switch (ethcmd) { |
2748 | default: |
2749 | break; |
2750 | case ETHTOOL_GRXCLSRLALL: |
2751 | /* Buffer size is variable */ |
2752 | if (get_user(rule_cnt, &compat_rxnfc->rule_cnt)) |
2753 | return -EFAULT; |
2754 | if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32)) |
2755 | return -ENOMEM; |
2756 | buf_size += rule_cnt * sizeof(u32); |
2757 | /* fall through */ |
2758 | case ETHTOOL_GRXRINGS: |
2759 | case ETHTOOL_GRXCLSRLCNT: |
2760 | case ETHTOOL_GRXCLSRULE: |
2761 | case ETHTOOL_SRXCLSRLINS: |
2762 | convert_out = true; |
2763 | /* fall through */ |
2764 | case ETHTOOL_SRXCLSRLDEL: |
2765 | buf_size += sizeof(struct ethtool_rxnfc); |
2766 | convert_in = true; |
2767 | break; |
2768 | } |
2769 | |
2770 | ifr = compat_alloc_user_space(buf_size); |
2771 | rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8); |
2772 | |
2773 | if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ)) |
2774 | return -EFAULT; |
2775 | |
2776 | if (put_user(convert_in ? rxnfc : compat_ptr(data), |
2777 | &ifr->ifr_ifru.ifru_data)) |
2778 | return -EFAULT; |
2779 | |
2780 | if (convert_in) { |
2781 | /* We expect there to be holes between fs.m_ext and |
2782 | * fs.ring_cookie and at the end of fs, but nowhere else. |
2783 | */ |
2784 | BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) + |
2785 | sizeof(compat_rxnfc->fs.m_ext) != |
2786 | offsetof(struct ethtool_rxnfc, fs.m_ext) + |
2787 | sizeof(rxnfc->fs.m_ext)); |
2788 | BUILD_BUG_ON( |
2789 | offsetof(struct compat_ethtool_rxnfc, fs.location) - |
2790 | offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) != |
2791 | offsetof(struct ethtool_rxnfc, fs.location) - |
2792 | offsetof(struct ethtool_rxnfc, fs.ring_cookie)); |
2793 | |
2794 | if (copy_in_user(rxnfc, compat_rxnfc, |
2795 | (void *)(&rxnfc->fs.m_ext + 1) - |
2796 | (void *)rxnfc) || |
2797 | copy_in_user(&rxnfc->fs.ring_cookie, |
2798 | &compat_rxnfc->fs.ring_cookie, |
2799 | (void *)(&rxnfc->fs.location + 1) - |
2800 | (void *)&rxnfc->fs.ring_cookie) || |
2801 | copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt, |
2802 | sizeof(rxnfc->rule_cnt))) |
2803 | return -EFAULT; |
2804 | } |
2805 | |
2806 | ret = dev_ioctl(net, SIOCETHTOOL, ifr); |
2807 | if (ret) |
2808 | return ret; |
2809 | |
2810 | if (convert_out) { |
2811 | if (copy_in_user(compat_rxnfc, rxnfc, |
2812 | (const void *)(&rxnfc->fs.m_ext + 1) - |
2813 | (const void *)rxnfc) || |
2814 | copy_in_user(&compat_rxnfc->fs.ring_cookie, |
2815 | &rxnfc->fs.ring_cookie, |
2816 | (const void *)(&rxnfc->fs.location + 1) - |
2817 | (const void *)&rxnfc->fs.ring_cookie) || |
2818 | copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt, |
2819 | sizeof(rxnfc->rule_cnt))) |
2820 | return -EFAULT; |
2821 | |
2822 | if (ethcmd == ETHTOOL_GRXCLSRLALL) { |
2823 | /* As an optimisation, we only copy the actual |
2824 | * number of rules that the underlying |
2825 | * function returned. Since Mallory might |
2826 | * change the rule count in user memory, we |
2827 | * check that it is less than the rule count |
2828 | * originally given (as the user buffer size), |
2829 | * which has been range-checked. |
2830 | */ |
2831 | if (get_user(actual_rule_cnt, &rxnfc->rule_cnt)) |
2832 | return -EFAULT; |
2833 | if (actual_rule_cnt < rule_cnt) |
2834 | rule_cnt = actual_rule_cnt; |
2835 | if (copy_in_user(&compat_rxnfc->rule_locs[0], |
2836 | &rxnfc->rule_locs[0], |
2837 | rule_cnt * sizeof(u32))) |
2838 | return -EFAULT; |
2839 | } |
2840 | } |
2841 | |
2842 | return 0; |
2843 | } |
2844 | |
2845 | static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) |
2846 | { |
2847 | void __user *uptr; |
2848 | compat_uptr_t uptr32; |
2849 | struct ifreq __user *uifr; |
2850 | |
2851 | uifr = compat_alloc_user_space(sizeof(*uifr)); |
2852 | if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) |
2853 | return -EFAULT; |
2854 | |
2855 | if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) |
2856 | return -EFAULT; |
2857 | |
2858 | uptr = compat_ptr(uptr32); |
2859 | |
2860 | if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc)) |
2861 | return -EFAULT; |
2862 | |
2863 | return dev_ioctl(net, SIOCWANDEV, uifr); |
2864 | } |
2865 | |
2866 | static int bond_ioctl(struct net *net, unsigned int cmd, |
2867 | struct compat_ifreq __user *ifr32) |
2868 | { |
2869 | struct ifreq kifr; |
2870 | struct ifreq __user *uifr; |
2871 | mm_segment_t old_fs; |
2872 | int err; |
2873 | u32 data; |
2874 | void __user *datap; |
2875 | |
2876 | switch (cmd) { |
2877 | case SIOCBONDENSLAVE: |
2878 | case SIOCBONDRELEASE: |
2879 | case SIOCBONDSETHWADDR: |
2880 | case SIOCBONDCHANGEACTIVE: |
2881 | if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq))) |
2882 | return -EFAULT; |
2883 | |
2884 | old_fs = get_fs(); |
2885 | set_fs(KERNEL_DS); |
2886 | err = dev_ioctl(net, cmd, |
2887 | (struct ifreq __user __force *) &kifr); |
2888 | set_fs(old_fs); |
2889 | |
2890 | return err; |
2891 | case SIOCBONDSLAVEINFOQUERY: |
2892 | case SIOCBONDINFOQUERY: |
2893 | uifr = compat_alloc_user_space(sizeof(*uifr)); |
2894 | if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ)) |
2895 | return -EFAULT; |
2896 | |
2897 | if (get_user(data, &ifr32->ifr_ifru.ifru_data)) |
2898 | return -EFAULT; |
2899 | |
2900 | datap = compat_ptr(data); |
2901 | if (put_user(datap, &uifr->ifr_ifru.ifru_data)) |
2902 | return -EFAULT; |
2903 | |
2904 | return dev_ioctl(net, cmd, uifr); |
2905 | default: |
2906 | return -ENOIOCTLCMD; |
2907 | } |
2908 | } |
2909 | |
2910 | static int siocdevprivate_ioctl(struct net *net, unsigned int cmd, |
2911 | struct compat_ifreq __user *u_ifreq32) |
2912 | { |
2913 | struct ifreq __user *u_ifreq64; |
2914 | char tmp_buf[IFNAMSIZ]; |
2915 | void __user *data64; |
2916 | u32 data32; |
2917 | |
2918 | if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]), |
2919 | IFNAMSIZ)) |
2920 | return -EFAULT; |
2921 | if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data)) |
2922 | return -EFAULT; |
2923 | data64 = compat_ptr(data32); |
2924 | |
2925 | u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64)); |
2926 | |
2927 | /* Don't check these user accesses, just let that get trapped |
2928 | * in the ioctl handler instead. |
2929 | */ |
2930 | if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0], |
2931 | IFNAMSIZ)) |
2932 | return -EFAULT; |
2933 | if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data)) |
2934 | return -EFAULT; |
2935 | |
2936 | return dev_ioctl(net, cmd, u_ifreq64); |
2937 | } |
2938 | |
2939 | static int dev_ifsioc(struct net *net, struct socket *sock, |
2940 | unsigned int cmd, struct compat_ifreq __user *uifr32) |
2941 | { |
2942 | struct ifreq __user *uifr; |
2943 | int err; |
2944 | |
2945 | uifr = compat_alloc_user_space(sizeof(*uifr)); |
2946 | if (copy_in_user(uifr, uifr32, sizeof(*uifr32))) |
2947 | return -EFAULT; |
2948 | |
2949 | err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr); |
2950 | |
2951 | if (!err) { |
2952 | switch (cmd) { |
2953 | case SIOCGIFFLAGS: |
2954 | case SIOCGIFMETRIC: |
2955 | case SIOCGIFMTU: |
2956 | case SIOCGIFMEM: |
2957 | case SIOCGIFHWADDR: |
2958 | case SIOCGIFINDEX: |
2959 | case SIOCGIFADDR: |
2960 | case SIOCGIFBRDADDR: |
2961 | case SIOCGIFDSTADDR: |
2962 | case SIOCGIFNETMASK: |
2963 | case SIOCGIFPFLAGS: |
2964 | case SIOCGIFTXQLEN: |
2965 | case SIOCGMIIPHY: |
2966 | case SIOCGMIIREG: |
2967 | if (copy_in_user(uifr32, uifr, sizeof(*uifr32))) |
2968 | err = -EFAULT; |
2969 | break; |
2970 | } |
2971 | } |
2972 | return err; |
2973 | } |
2974 | |
2975 | static int compat_sioc_ifmap(struct net *net, unsigned int cmd, |
2976 | struct compat_ifreq __user *uifr32) |
2977 | { |
2978 | struct ifreq ifr; |
2979 | struct compat_ifmap __user *uifmap32; |
2980 | mm_segment_t old_fs; |
2981 | int err; |
2982 | |
2983 | uifmap32 = &uifr32->ifr_ifru.ifru_map; |
2984 | err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name)); |
2985 | err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); |
2986 | err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); |
2987 | err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); |
2988 | err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq); |
2989 | err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma); |
2990 | err |= __get_user(ifr.ifr_map.port, &uifmap32->port); |
2991 | if (err) |
2992 | return -EFAULT; |
2993 | |
2994 | old_fs = get_fs(); |
2995 | set_fs(KERNEL_DS); |
2996 | err = dev_ioctl(net, cmd, (void __user __force *)&ifr); |
2997 | set_fs(old_fs); |
2998 | |
2999 | if (cmd == SIOCGIFMAP && !err) { |
3000 | err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name)); |
3001 | err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); |
3002 | err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); |
3003 | err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); |
3004 | err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq); |
3005 | err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma); |
3006 | err |= __put_user(ifr.ifr_map.port, &uifmap32->port); |
3007 | if (err) |
3008 | err = -EFAULT; |
3009 | } |
3010 | return err; |
3011 | } |
3012 | |
3013 | static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32) |
3014 | { |
3015 | void __user *uptr; |
3016 | compat_uptr_t uptr32; |
3017 | struct ifreq __user *uifr; |
3018 | |
3019 | uifr = compat_alloc_user_space(sizeof(*uifr)); |
3020 | if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) |
3021 | return -EFAULT; |
3022 | |
3023 | if (get_user(uptr32, &uifr32->ifr_data)) |
3024 | return -EFAULT; |
3025 | |
3026 | uptr = compat_ptr(uptr32); |
3027 | |
3028 | if (put_user(uptr, &uifr->ifr_data)) |
3029 | return -EFAULT; |
3030 | |
3031 | return dev_ioctl(net, SIOCSHWTSTAMP, uifr); |
3032 | } |
3033 | |
3034 | struct rtentry32 { |
3035 | u32 rt_pad1; |
3036 | struct sockaddr rt_dst; /* target address */ |
3037 | struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */ |
3038 | struct sockaddr rt_genmask; /* target network mask (IP) */ |
3039 | unsigned short rt_flags; |
3040 | short rt_pad2; |
3041 | u32 rt_pad3; |
3042 | unsigned char rt_tos; |
3043 | unsigned char rt_class; |
3044 | short rt_pad4; |
3045 | short rt_metric; /* +1 for binary compatibility! */ |
3046 | /* char * */ u32 rt_dev; /* forcing the device at add */ |
3047 | u32 rt_mtu; /* per route MTU/Window */ |
3048 | u32 rt_window; /* Window clamping */ |
3049 | unsigned short rt_irtt; /* Initial RTT */ |
3050 | }; |
3051 | |
3052 | struct in6_rtmsg32 { |
3053 | struct in6_addr rtmsg_dst; |
3054 | struct in6_addr rtmsg_src; |
3055 | struct in6_addr rtmsg_gateway; |
3056 | u32 rtmsg_type; |
3057 | u16 rtmsg_dst_len; |
3058 | u16 rtmsg_src_len; |
3059 | u32 rtmsg_metric; |
3060 | u32 rtmsg_info; |
3061 | u32 rtmsg_flags; |
3062 | s32 rtmsg_ifindex; |
3063 | }; |
3064 | |
3065 | static int routing_ioctl(struct net *net, struct socket *sock, |
3066 | unsigned int cmd, void __user *argp) |
3067 | { |
3068 | int ret; |
3069 | void *r = NULL; |
3070 | struct in6_rtmsg r6; |
3071 | struct rtentry r4; |
3072 | char devname[16]; |
3073 | u32 rtdev; |
3074 | mm_segment_t old_fs = get_fs(); |
3075 | |
3076 | if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */ |
3077 | struct in6_rtmsg32 __user *ur6 = argp; |
3078 | ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst), |
3079 | 3 * sizeof(struct in6_addr)); |
3080 | ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type)); |
3081 | ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len)); |
3082 | ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len)); |
3083 | ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric)); |
3084 | ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info)); |
3085 | ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags)); |
3086 | ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex)); |
3087 | |
3088 | r = (void *) &r6; |
3089 | } else { /* ipv4 */ |
3090 | struct rtentry32 __user *ur4 = argp; |
3091 | ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst), |
3092 | 3 * sizeof(struct sockaddr)); |
3093 | ret |= __get_user(r4.rt_flags, &(ur4->rt_flags)); |
3094 | ret |= __get_user(r4.rt_metric, &(ur4->rt_metric)); |
3095 | ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu)); |
3096 | ret |= __get_user(r4.rt_window, &(ur4->rt_window)); |
3097 | ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt)); |
3098 | ret |= __get_user(rtdev, &(ur4->rt_dev)); |
3099 | if (rtdev) { |
3100 | ret |= copy_from_user(devname, compat_ptr(rtdev), 15); |
3101 | r4.rt_dev = (char __user __force *)devname; |
3102 | devname[15] = 0; |
3103 | } else |
3104 | r4.rt_dev = NULL; |
3105 | |
3106 | r = (void *) &r4; |
3107 | } |
3108 | |
3109 | if (ret) { |
3110 | ret = -EFAULT; |
3111 | goto out; |
3112 | } |
3113 | |
3114 | set_fs(KERNEL_DS); |
3115 | ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r); |
3116 | set_fs(old_fs); |
3117 | |
3118 | out: |
3119 | return ret; |
3120 | } |
3121 | |
3122 | /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE |
3123 | * for some operations; this forces use of the newer bridge-utils that |
3124 | * use compatible ioctls |
3125 | */ |
3126 | static int old_bridge_ioctl(compat_ulong_t __user *argp) |
3127 | { |
3128 | compat_ulong_t tmp; |
3129 | |
3130 | if (get_user(tmp, argp)) |
3131 | return -EFAULT; |
3132 | if (tmp == BRCTL_GET_VERSION) |
3133 | return BRCTL_VERSION + 1; |
3134 | return -EINVAL; |
3135 | } |
3136 | |
3137 | static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, |
3138 | unsigned int cmd, unsigned long arg) |
3139 | { |
3140 | void __user *argp = compat_ptr(arg); |
3141 | struct sock *sk = sock->sk; |
3142 | struct net *net = sock_net(sk); |
3143 | |
3144 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) |
3145 | return siocdevprivate_ioctl(net, cmd, argp); |
3146 | |
3147 | switch (cmd) { |
3148 | case SIOCSIFBR: |
3149 | case SIOCGIFBR: |
3150 | return old_bridge_ioctl(argp); |
3151 | case SIOCGIFNAME: |
3152 | return dev_ifname32(net, argp); |
3153 | case SIOCGIFCONF: |
3154 | return dev_ifconf(net, argp); |
3155 | case SIOCETHTOOL: |
3156 | return ethtool_ioctl(net, argp); |
3157 | case SIOCWANDEV: |
3158 | return compat_siocwandev(net, argp); |
3159 | case SIOCGIFMAP: |
3160 | case SIOCSIFMAP: |
3161 | return compat_sioc_ifmap(net, cmd, argp); |
3162 | case SIOCBONDENSLAVE: |
3163 | case SIOCBONDRELEASE: |
3164 | case SIOCBONDSETHWADDR: |
3165 | case SIOCBONDSLAVEINFOQUERY: |
3166 | case SIOCBONDINFOQUERY: |
3167 | case SIOCBONDCHANGEACTIVE: |
3168 | return bond_ioctl(net, cmd, argp); |
3169 | case SIOCADDRT: |
3170 | case SIOCDELRT: |
3171 | return routing_ioctl(net, sock, cmd, argp); |
3172 | case SIOCGSTAMP: |
3173 | return do_siocgstamp(net, sock, cmd, argp); |
3174 | case SIOCGSTAMPNS: |
3175 | return do_siocgstampns(net, sock, cmd, argp); |
3176 | case SIOCSHWTSTAMP: |
3177 | return compat_siocshwtstamp(net, argp); |
3178 | |
3179 | case FIOSETOWN: |
3180 | case SIOCSPGRP: |
3181 | case FIOGETOWN: |
3182 | case SIOCGPGRP: |
3183 | case SIOCBRADDBR: |
3184 | case SIOCBRDELBR: |
3185 | case SIOCGIFVLAN: |
3186 | case SIOCSIFVLAN: |
3187 | case SIOCADDDLCI: |
3188 | case SIOCDELDLCI: |
3189 | return sock_ioctl(file, cmd, arg); |
3190 | |
3191 | case SIOCGIFFLAGS: |
3192 | case SIOCSIFFLAGS: |
3193 | case SIOCGIFMETRIC: |
3194 | case SIOCSIFMETRIC: |
3195 | case SIOCGIFMTU: |
3196 | case SIOCSIFMTU: |
3197 | case SIOCGIFMEM: |
3198 | case SIOCSIFMEM: |
3199 | case SIOCGIFHWADDR: |
3200 | case SIOCSIFHWADDR: |
3201 | case SIOCADDMULTI: |
3202 | case SIOCDELMULTI: |
3203 | case SIOCGIFINDEX: |
3204 | case SIOCGIFADDR: |
3205 | case SIOCSIFADDR: |
3206 | case SIOCSIFHWBROADCAST: |
3207 | case SIOCDIFADDR: |
3208 | case SIOCGIFBRDADDR: |
3209 | case SIOCSIFBRDADDR: |
3210 | case SIOCGIFDSTADDR: |
3211 | case SIOCSIFDSTADDR: |
3212 | case SIOCGIFNETMASK: |
3213 | case SIOCSIFNETMASK: |
3214 | case SIOCSIFPFLAGS: |
3215 | case SIOCGIFPFLAGS: |
3216 | case SIOCGIFTXQLEN: |
3217 | case SIOCSIFTXQLEN: |
3218 | case SIOCBRADDIF: |
3219 | case SIOCBRDELIF: |
3220 | case SIOCSIFNAME: |
3221 | case SIOCGMIIPHY: |
3222 | case SIOCGMIIREG: |
3223 | case SIOCSMIIREG: |
3224 | return dev_ifsioc(net, sock, cmd, argp); |
3225 | |
3226 | case SIOCSARP: |
3227 | case SIOCGARP: |
3228 | case SIOCDARP: |
3229 | case SIOCATMARK: |
3230 | return sock_do_ioctl(net, sock, cmd, arg); |
3231 | } |
3232 | |
3233 | return -ENOIOCTLCMD; |
3234 | } |
3235 | |
3236 | static long compat_sock_ioctl(struct file *file, unsigned cmd, |
3237 | unsigned long arg) |
3238 | { |
3239 | struct socket *sock = file->private_data; |
3240 | int ret = -ENOIOCTLCMD; |
3241 | struct sock *sk; |
3242 | struct net *net; |
3243 | |
3244 | sk = sock->sk; |
3245 | net = sock_net(sk); |
3246 | |
3247 | if (sock->ops->compat_ioctl) |
3248 | ret = sock->ops->compat_ioctl(sock, cmd, arg); |
3249 | |
3250 | if (ret == -ENOIOCTLCMD && |
3251 | (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) |
3252 | ret = compat_wext_handle_ioctl(net, cmd, arg); |
3253 | |
3254 | if (ret == -ENOIOCTLCMD) |
3255 | ret = compat_sock_ioctl_trans(file, sock, cmd, arg); |
3256 | |
3257 | return ret; |
3258 | } |
3259 | #endif |
3260 | |
3261 | int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) |
3262 | { |
3263 | return sock->ops->bind(sock, addr, addrlen); |
3264 | } |
3265 | EXPORT_SYMBOL(kernel_bind); |
3266 | |
3267 | int kernel_listen(struct socket *sock, int backlog) |
3268 | { |
3269 | return sock->ops->listen(sock, backlog); |
3270 | } |
3271 | EXPORT_SYMBOL(kernel_listen); |
3272 | |
3273 | int kernel_accept(struct socket *sock, struct socket **newsock, int flags) |
3274 | { |
3275 | struct sock *sk = sock->sk; |
3276 | int err; |
3277 | |
3278 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, |
3279 | newsock); |
3280 | if (err < 0) |
3281 | goto done; |
3282 | |
3283 | err = sock->ops->accept(sock, *newsock, flags); |
3284 | if (err < 0) { |
3285 | sock_release(*newsock); |
3286 | *newsock = NULL; |
3287 | goto done; |
3288 | } |
3289 | |
3290 | (*newsock)->ops = sock->ops; |
3291 | __module_get((*newsock)->ops->owner); |
3292 | |
3293 | done: |
3294 | return err; |
3295 | } |
3296 | EXPORT_SYMBOL(kernel_accept); |
3297 | |
3298 | int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, |
3299 | int flags) |
3300 | { |
3301 | return sock->ops->connect(sock, addr, addrlen, flags); |
3302 | } |
3303 | EXPORT_SYMBOL(kernel_connect); |
3304 | |
3305 | int kernel_getsockname(struct socket *sock, struct sockaddr *addr, |
3306 | int *addrlen) |
3307 | { |
3308 | return sock->ops->getname(sock, addr, addrlen, 0); |
3309 | } |
3310 | EXPORT_SYMBOL(kernel_getsockname); |
3311 | |
3312 | int kernel_getpeername(struct socket *sock, struct sockaddr *addr, |
3313 | int *addrlen) |
3314 | { |
3315 | return sock->ops->getname(sock, addr, addrlen, 1); |
3316 | } |
3317 | EXPORT_SYMBOL(kernel_getpeername); |
3318 | |
3319 | int kernel_getsockopt(struct socket *sock, int level, int optname, |
3320 | char *optval, int *optlen) |
3321 | { |
3322 | mm_segment_t oldfs = get_fs(); |
3323 | char __user *uoptval; |
3324 | int __user *uoptlen; |
3325 | int err; |
3326 | |
3327 | uoptval = (char __user __force *) optval; |
3328 | uoptlen = (int __user __force *) optlen; |
3329 | |
3330 | set_fs(KERNEL_DS); |
3331 | if (level == SOL_SOCKET) |
3332 | err = sock_getsockopt(sock, level, optname, uoptval, uoptlen); |
3333 | else |
3334 | err = sock->ops->getsockopt(sock, level, optname, uoptval, |
3335 | uoptlen); |
3336 | set_fs(oldfs); |
3337 | return err; |
3338 | } |
3339 | EXPORT_SYMBOL(kernel_getsockopt); |
3340 | |
3341 | int kernel_setsockopt(struct socket *sock, int level, int optname, |
3342 | char *optval, unsigned int optlen) |
3343 | { |
3344 | mm_segment_t oldfs = get_fs(); |
3345 | char __user *uoptval; |
3346 | int err; |
3347 | |
3348 | uoptval = (char __user __force *) optval; |
3349 | |
3350 | set_fs(KERNEL_DS); |
3351 | if (level == SOL_SOCKET) |
3352 | err = sock_setsockopt(sock, level, optname, uoptval, optlen); |
3353 | else |
3354 | err = sock->ops->setsockopt(sock, level, optname, uoptval, |
3355 | optlen); |
3356 | set_fs(oldfs); |
3357 | return err; |
3358 | } |
3359 | EXPORT_SYMBOL(kernel_setsockopt); |
3360 | |
3361 | int kernel_sendpage(struct socket *sock, struct page *page, int offset, |
3362 | size_t size, int flags) |
3363 | { |
3364 | sock_update_classid(sock->sk); |
3365 | |
3366 | if (sock->ops->sendpage) |
3367 | return sock->ops->sendpage(sock, page, offset, size, flags); |
3368 | |
3369 | return sock_no_sendpage(sock, page, offset, size, flags); |
3370 | } |
3371 | EXPORT_SYMBOL(kernel_sendpage); |
3372 | |
3373 | int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg) |
3374 | { |
3375 | mm_segment_t oldfs = get_fs(); |
3376 | int err; |
3377 | |
3378 | set_fs(KERNEL_DS); |
3379 | err = sock->ops->ioctl(sock, cmd, arg); |
3380 | set_fs(oldfs); |
3381 | |
3382 | return err; |
3383 | } |
3384 | EXPORT_SYMBOL(kernel_sock_ioctl); |
3385 | |
3386 | int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) |
3387 | { |
3388 | return sock->ops->shutdown(sock, how); |
3389 | } |
3390 | EXPORT_SYMBOL(kernel_sock_shutdown); |
3391 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9