]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
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> | |
02c30a84 | 7 | * Ross Biro |
1da177e4 LT |
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) | |
89bddce5 | 45 | * Tigran Aivazian : Made listen(2) backlog sanity checks |
1da177e4 LT |
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 | |
89bddce5 | 56 | * paradigm. |
1da177e4 LT |
57 | * |
58 | * Based upon Swansea University Computer Society NET3.039 | |
59 | */ | |
60 | ||
1da177e4 | 61 | #include <linux/mm.h> |
1da177e4 LT |
62 | #include <linux/socket.h> |
63 | #include <linux/file.h> | |
64 | #include <linux/net.h> | |
65 | #include <linux/interrupt.h> | |
55737fda | 66 | #include <linux/rcupdate.h> |
1da177e4 LT |
67 | #include <linux/netdevice.h> |
68 | #include <linux/proc_fs.h> | |
69 | #include <linux/seq_file.h> | |
4a3e2f71 | 70 | #include <linux/mutex.h> |
1da177e4 LT |
71 | #include <linux/wanrouter.h> |
72 | #include <linux/if_bridge.h> | |
20380731 ACM |
73 | #include <linux/if_frad.h> |
74 | #include <linux/if_vlan.h> | |
1da177e4 LT |
75 | #include <linux/init.h> |
76 | #include <linux/poll.h> | |
77 | #include <linux/cache.h> | |
78 | #include <linux/module.h> | |
79 | #include <linux/highmem.h> | |
1da177e4 LT |
80 | #include <linux/mount.h> |
81 | #include <linux/security.h> | |
82 | #include <linux/syscalls.h> | |
83 | #include <linux/compat.h> | |
84 | #include <linux/kmod.h> | |
3ec3b2fb | 85 | #include <linux/audit.h> |
d86b5e0e | 86 | #include <linux/wireless.h> |
1b8d7ae4 | 87 | #include <linux/nsproxy.h> |
1da177e4 LT |
88 | |
89 | #include <asm/uaccess.h> | |
90 | #include <asm/unistd.h> | |
91 | ||
92 | #include <net/compat.h> | |
93 | ||
94 | #include <net/sock.h> | |
95 | #include <linux/netfilter.h> | |
96 | ||
97 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare); | |
027445c3 BP |
98 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, |
99 | unsigned long nr_segs, loff_t pos); | |
100 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
101 | unsigned long nr_segs, loff_t pos); | |
89bddce5 | 102 | static int sock_mmap(struct file *file, struct vm_area_struct *vma); |
1da177e4 LT |
103 | |
104 | static int sock_close(struct inode *inode, struct file *file); | |
105 | static unsigned int sock_poll(struct file *file, | |
106 | struct poll_table_struct *wait); | |
89bddce5 | 107 | static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); |
89bbfc95 SP |
108 | #ifdef CONFIG_COMPAT |
109 | static long compat_sock_ioctl(struct file *file, | |
89bddce5 | 110 | unsigned int cmd, unsigned long arg); |
89bbfc95 | 111 | #endif |
1da177e4 | 112 | static int sock_fasync(int fd, struct file *filp, int on); |
1da177e4 LT |
113 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
114 | int offset, size_t size, loff_t *ppos, int more); | |
115 | ||
1da177e4 LT |
116 | /* |
117 | * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear | |
118 | * in the operation structures but are done directly via the socketcall() multiplexor. | |
119 | */ | |
120 | ||
da7071d7 | 121 | static const struct file_operations socket_file_ops = { |
1da177e4 LT |
122 | .owner = THIS_MODULE, |
123 | .llseek = no_llseek, | |
124 | .aio_read = sock_aio_read, | |
125 | .aio_write = sock_aio_write, | |
126 | .poll = sock_poll, | |
127 | .unlocked_ioctl = sock_ioctl, | |
89bbfc95 SP |
128 | #ifdef CONFIG_COMPAT |
129 | .compat_ioctl = compat_sock_ioctl, | |
130 | #endif | |
1da177e4 LT |
131 | .mmap = sock_mmap, |
132 | .open = sock_no_open, /* special open code to disallow open via /proc */ | |
133 | .release = sock_close, | |
134 | .fasync = sock_fasync, | |
5274f052 JA |
135 | .sendpage = sock_sendpage, |
136 | .splice_write = generic_splice_sendpage, | |
1da177e4 LT |
137 | }; |
138 | ||
139 | /* | |
140 | * The protocol list. Each protocol is registered in here. | |
141 | */ | |
142 | ||
1da177e4 | 143 | static DEFINE_SPINLOCK(net_family_lock); |
f0fd27d4 | 144 | static const struct net_proto_family *net_families[NPROTO] __read_mostly; |
1da177e4 | 145 | |
1da177e4 LT |
146 | /* |
147 | * Statistics counters of the socket lists | |
148 | */ | |
149 | ||
150 | static DEFINE_PER_CPU(int, sockets_in_use) = 0; | |
151 | ||
152 | /* | |
89bddce5 SH |
153 | * Support routines. |
154 | * Move socket addresses back and forth across the kernel/user | |
155 | * divide and look after the messy bits. | |
1da177e4 LT |
156 | */ |
157 | ||
89bddce5 | 158 | #define MAX_SOCK_ADDR 128 /* 108 for Unix domain - |
1da177e4 LT |
159 | 16 for IP, 16 for IPX, |
160 | 24 for IPv6, | |
89bddce5 | 161 | about 80 for AX.25 |
1da177e4 LT |
162 | must be at least one bigger than |
163 | the AF_UNIX size (see net/unix/af_unix.c | |
89bddce5 | 164 | :unix_mkname()). |
1da177e4 | 165 | */ |
89bddce5 | 166 | |
1da177e4 LT |
167 | /** |
168 | * move_addr_to_kernel - copy a socket address into kernel space | |
169 | * @uaddr: Address in user space | |
170 | * @kaddr: Address in kernel space | |
171 | * @ulen: Length in user space | |
172 | * | |
173 | * The address is copied into kernel space. If the provided address is | |
174 | * too long an error code of -EINVAL is returned. If the copy gives | |
175 | * invalid addresses -EFAULT is returned. On a success 0 is returned. | |
176 | */ | |
177 | ||
178 | int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr) | |
179 | { | |
89bddce5 | 180 | if (ulen < 0 || ulen > MAX_SOCK_ADDR) |
1da177e4 | 181 | return -EINVAL; |
89bddce5 | 182 | if (ulen == 0) |
1da177e4 | 183 | return 0; |
89bddce5 | 184 | if (copy_from_user(kaddr, uaddr, ulen)) |
1da177e4 | 185 | return -EFAULT; |
3ec3b2fb | 186 | return audit_sockaddr(ulen, kaddr); |
1da177e4 LT |
187 | } |
188 | ||
189 | /** | |
190 | * move_addr_to_user - copy an address to user space | |
191 | * @kaddr: kernel space address | |
192 | * @klen: length of address in kernel | |
193 | * @uaddr: user space address | |
194 | * @ulen: pointer to user length field | |
195 | * | |
196 | * The value pointed to by ulen on entry is the buffer length available. | |
197 | * This is overwritten with the buffer space used. -EINVAL is returned | |
198 | * if an overlong buffer is specified or a negative buffer size. -EFAULT | |
199 | * is returned if either the buffer or the length field are not | |
200 | * accessible. | |
201 | * After copying the data up to the limit the user specifies, the true | |
202 | * length of the data is written over the length limit the user | |
203 | * specified. Zero is returned for a success. | |
204 | */ | |
89bddce5 SH |
205 | |
206 | int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, | |
207 | int __user *ulen) | |
1da177e4 LT |
208 | { |
209 | int err; | |
210 | int len; | |
211 | ||
89bddce5 SH |
212 | err = get_user(len, ulen); |
213 | if (err) | |
1da177e4 | 214 | return err; |
89bddce5 SH |
215 | if (len > klen) |
216 | len = klen; | |
217 | if (len < 0 || len > MAX_SOCK_ADDR) | |
1da177e4 | 218 | return -EINVAL; |
89bddce5 | 219 | if (len) { |
d6fe3945 SG |
220 | if (audit_sockaddr(klen, kaddr)) |
221 | return -ENOMEM; | |
89bddce5 | 222 | if (copy_to_user(uaddr, kaddr, len)) |
1da177e4 LT |
223 | return -EFAULT; |
224 | } | |
225 | /* | |
89bddce5 SH |
226 | * "fromlen shall refer to the value before truncation.." |
227 | * 1003.1g | |
1da177e4 LT |
228 | */ |
229 | return __put_user(klen, ulen); | |
230 | } | |
231 | ||
232 | #define SOCKFS_MAGIC 0x534F434B | |
233 | ||
e18b890b | 234 | static struct kmem_cache *sock_inode_cachep __read_mostly; |
1da177e4 LT |
235 | |
236 | static struct inode *sock_alloc_inode(struct super_block *sb) | |
237 | { | |
238 | struct socket_alloc *ei; | |
89bddce5 | 239 | |
e94b1766 | 240 | ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); |
1da177e4 LT |
241 | if (!ei) |
242 | return NULL; | |
243 | init_waitqueue_head(&ei->socket.wait); | |
89bddce5 | 244 | |
1da177e4 LT |
245 | ei->socket.fasync_list = NULL; |
246 | ei->socket.state = SS_UNCONNECTED; | |
247 | ei->socket.flags = 0; | |
248 | ei->socket.ops = NULL; | |
249 | ei->socket.sk = NULL; | |
250 | ei->socket.file = NULL; | |
1da177e4 LT |
251 | |
252 | return &ei->vfs_inode; | |
253 | } | |
254 | ||
255 | static void sock_destroy_inode(struct inode *inode) | |
256 | { | |
257 | kmem_cache_free(sock_inode_cachep, | |
258 | container_of(inode, struct socket_alloc, vfs_inode)); | |
259 | } | |
260 | ||
e18b890b | 261 | static void init_once(void *foo, struct kmem_cache *cachep, unsigned long flags) |
1da177e4 | 262 | { |
89bddce5 | 263 | struct socket_alloc *ei = (struct socket_alloc *)foo; |
1da177e4 | 264 | |
a35afb83 | 265 | inode_init_once(&ei->vfs_inode); |
1da177e4 | 266 | } |
89bddce5 | 267 | |
1da177e4 LT |
268 | static int init_inodecache(void) |
269 | { | |
270 | sock_inode_cachep = kmem_cache_create("sock_inode_cache", | |
89bddce5 SH |
271 | sizeof(struct socket_alloc), |
272 | 0, | |
273 | (SLAB_HWCACHE_ALIGN | | |
274 | SLAB_RECLAIM_ACCOUNT | | |
275 | SLAB_MEM_SPREAD), | |
20c2df83 | 276 | init_once); |
1da177e4 LT |
277 | if (sock_inode_cachep == NULL) |
278 | return -ENOMEM; | |
279 | return 0; | |
280 | } | |
281 | ||
282 | static struct super_operations sockfs_ops = { | |
283 | .alloc_inode = sock_alloc_inode, | |
284 | .destroy_inode =sock_destroy_inode, | |
285 | .statfs = simple_statfs, | |
286 | }; | |
287 | ||
454e2398 | 288 | static int sockfs_get_sb(struct file_system_type *fs_type, |
89bddce5 SH |
289 | int flags, const char *dev_name, void *data, |
290 | struct vfsmount *mnt) | |
1da177e4 | 291 | { |
454e2398 DH |
292 | return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC, |
293 | mnt); | |
1da177e4 LT |
294 | } |
295 | ||
ba89966c | 296 | static struct vfsmount *sock_mnt __read_mostly; |
1da177e4 LT |
297 | |
298 | static struct file_system_type sock_fs_type = { | |
299 | .name = "sockfs", | |
300 | .get_sb = sockfs_get_sb, | |
301 | .kill_sb = kill_anon_super, | |
302 | }; | |
89bddce5 | 303 | |
1da177e4 LT |
304 | static int sockfs_delete_dentry(struct dentry *dentry) |
305 | { | |
304e61e6 ED |
306 | /* |
307 | * At creation time, we pretended this dentry was hashed | |
308 | * (by clearing DCACHE_UNHASHED bit in d_flags) | |
309 | * At delete time, we restore the truth : not hashed. | |
310 | * (so that dput() can proceed correctly) | |
311 | */ | |
312 | dentry->d_flags |= DCACHE_UNHASHED; | |
313 | return 0; | |
1da177e4 | 314 | } |
c23fbb6b ED |
315 | |
316 | /* | |
317 | * sockfs_dname() is called from d_path(). | |
318 | */ | |
319 | static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) | |
320 | { | |
321 | return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", | |
322 | dentry->d_inode->i_ino); | |
323 | } | |
324 | ||
1da177e4 | 325 | static struct dentry_operations sockfs_dentry_operations = { |
89bddce5 | 326 | .d_delete = sockfs_delete_dentry, |
c23fbb6b | 327 | .d_dname = sockfs_dname, |
1da177e4 LT |
328 | }; |
329 | ||
330 | /* | |
331 | * Obtains the first available file descriptor and sets it up for use. | |
332 | * | |
39d8c1b6 DM |
333 | * These functions create file structures and maps them to fd space |
334 | * of the current process. On success it returns file descriptor | |
1da177e4 LT |
335 | * and file struct implicitly stored in sock->file. |
336 | * Note that another thread may close file descriptor before we return | |
337 | * from this function. We use the fact that now we do not refer | |
338 | * to socket after mapping. If one day we will need it, this | |
339 | * function will increment ref. count on file by 1. | |
340 | * | |
341 | * In any case returned fd MAY BE not valid! | |
342 | * This race condition is unavoidable | |
343 | * with shared fd spaces, we cannot solve it inside kernel, | |
344 | * but we take care of internal coherence yet. | |
345 | */ | |
346 | ||
39d8c1b6 | 347 | static int sock_alloc_fd(struct file **filep) |
1da177e4 LT |
348 | { |
349 | int fd; | |
1da177e4 LT |
350 | |
351 | fd = get_unused_fd(); | |
39d8c1b6 | 352 | if (likely(fd >= 0)) { |
1da177e4 LT |
353 | struct file *file = get_empty_filp(); |
354 | ||
39d8c1b6 DM |
355 | *filep = file; |
356 | if (unlikely(!file)) { | |
1da177e4 | 357 | put_unused_fd(fd); |
39d8c1b6 | 358 | return -ENFILE; |
1da177e4 | 359 | } |
39d8c1b6 DM |
360 | } else |
361 | *filep = NULL; | |
362 | return fd; | |
363 | } | |
1da177e4 | 364 | |
39d8c1b6 DM |
365 | static int sock_attach_fd(struct socket *sock, struct file *file) |
366 | { | |
c23fbb6b | 367 | struct qstr name = { .name = "" }; |
39d8c1b6 | 368 | |
c23fbb6b | 369 | file->f_path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name); |
3126a42c | 370 | if (unlikely(!file->f_path.dentry)) |
39d8c1b6 DM |
371 | return -ENOMEM; |
372 | ||
3126a42c | 373 | file->f_path.dentry->d_op = &sockfs_dentry_operations; |
304e61e6 ED |
374 | /* |
375 | * We dont want to push this dentry into global dentry hash table. | |
376 | * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED | |
377 | * This permits a working /proc/$pid/fd/XXX on sockets | |
378 | */ | |
3126a42c JS |
379 | file->f_path.dentry->d_flags &= ~DCACHE_UNHASHED; |
380 | d_instantiate(file->f_path.dentry, SOCK_INODE(sock)); | |
381 | file->f_path.mnt = mntget(sock_mnt); | |
382 | file->f_mapping = file->f_path.dentry->d_inode->i_mapping; | |
39d8c1b6 DM |
383 | |
384 | sock->file = file; | |
385 | file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops; | |
386 | file->f_mode = FMODE_READ | FMODE_WRITE; | |
387 | file->f_flags = O_RDWR; | |
388 | file->f_pos = 0; | |
389 | file->private_data = sock; | |
1da177e4 | 390 | |
39d8c1b6 DM |
391 | return 0; |
392 | } | |
393 | ||
394 | int sock_map_fd(struct socket *sock) | |
395 | { | |
396 | struct file *newfile; | |
397 | int fd = sock_alloc_fd(&newfile); | |
398 | ||
399 | if (likely(fd >= 0)) { | |
400 | int err = sock_attach_fd(sock, newfile); | |
401 | ||
402 | if (unlikely(err < 0)) { | |
403 | put_filp(newfile); | |
1da177e4 | 404 | put_unused_fd(fd); |
39d8c1b6 | 405 | return err; |
1da177e4 | 406 | } |
39d8c1b6 | 407 | fd_install(fd, newfile); |
1da177e4 | 408 | } |
1da177e4 LT |
409 | return fd; |
410 | } | |
411 | ||
6cb153ca BL |
412 | static struct socket *sock_from_file(struct file *file, int *err) |
413 | { | |
6cb153ca BL |
414 | if (file->f_op == &socket_file_ops) |
415 | return file->private_data; /* set in sock_map_fd */ | |
416 | ||
23bb80d2 ED |
417 | *err = -ENOTSOCK; |
418 | return NULL; | |
6cb153ca BL |
419 | } |
420 | ||
1da177e4 LT |
421 | /** |
422 | * sockfd_lookup - Go from a file number to its socket slot | |
423 | * @fd: file handle | |
424 | * @err: pointer to an error code return | |
425 | * | |
426 | * The file handle passed in is locked and the socket it is bound | |
427 | * too is returned. If an error occurs the err pointer is overwritten | |
428 | * with a negative errno code and NULL is returned. The function checks | |
429 | * for both invalid handles and passing a handle which is not a socket. | |
430 | * | |
431 | * On a success the socket object pointer is returned. | |
432 | */ | |
433 | ||
434 | struct socket *sockfd_lookup(int fd, int *err) | |
435 | { | |
436 | struct file *file; | |
1da177e4 LT |
437 | struct socket *sock; |
438 | ||
89bddce5 SH |
439 | file = fget(fd); |
440 | if (!file) { | |
1da177e4 LT |
441 | *err = -EBADF; |
442 | return NULL; | |
443 | } | |
89bddce5 | 444 | |
6cb153ca BL |
445 | sock = sock_from_file(file, err); |
446 | if (!sock) | |
1da177e4 | 447 | fput(file); |
6cb153ca BL |
448 | return sock; |
449 | } | |
1da177e4 | 450 | |
6cb153ca BL |
451 | static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) |
452 | { | |
453 | struct file *file; | |
454 | struct socket *sock; | |
455 | ||
3672558c | 456 | *err = -EBADF; |
6cb153ca BL |
457 | file = fget_light(fd, fput_needed); |
458 | if (file) { | |
459 | sock = sock_from_file(file, err); | |
460 | if (sock) | |
461 | return sock; | |
462 | fput_light(file, *fput_needed); | |
1da177e4 | 463 | } |
6cb153ca | 464 | return NULL; |
1da177e4 LT |
465 | } |
466 | ||
467 | /** | |
468 | * sock_alloc - allocate a socket | |
89bddce5 | 469 | * |
1da177e4 LT |
470 | * Allocate a new inode and socket object. The two are bound together |
471 | * and initialised. The socket is then returned. If we are out of inodes | |
472 | * NULL is returned. | |
473 | */ | |
474 | ||
475 | static struct socket *sock_alloc(void) | |
476 | { | |
89bddce5 SH |
477 | struct inode *inode; |
478 | struct socket *sock; | |
1da177e4 LT |
479 | |
480 | inode = new_inode(sock_mnt->mnt_sb); | |
481 | if (!inode) | |
482 | return NULL; | |
483 | ||
484 | sock = SOCKET_I(inode); | |
485 | ||
89bddce5 | 486 | inode->i_mode = S_IFSOCK | S_IRWXUGO; |
1da177e4 LT |
487 | inode->i_uid = current->fsuid; |
488 | inode->i_gid = current->fsgid; | |
489 | ||
490 | get_cpu_var(sockets_in_use)++; | |
491 | put_cpu_var(sockets_in_use); | |
492 | return sock; | |
493 | } | |
494 | ||
495 | /* | |
496 | * In theory you can't get an open on this inode, but /proc provides | |
497 | * a back door. Remember to keep it shut otherwise you'll let the | |
498 | * creepy crawlies in. | |
499 | */ | |
89bddce5 | 500 | |
1da177e4 LT |
501 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare) |
502 | { | |
503 | return -ENXIO; | |
504 | } | |
505 | ||
4b6f5d20 | 506 | const struct file_operations bad_sock_fops = { |
1da177e4 LT |
507 | .owner = THIS_MODULE, |
508 | .open = sock_no_open, | |
509 | }; | |
510 | ||
511 | /** | |
512 | * sock_release - close a socket | |
513 | * @sock: socket to close | |
514 | * | |
515 | * The socket is released from the protocol stack if it has a release | |
516 | * callback, and the inode is then released if the socket is bound to | |
89bddce5 | 517 | * an inode not a file. |
1da177e4 | 518 | */ |
89bddce5 | 519 | |
1da177e4 LT |
520 | void sock_release(struct socket *sock) |
521 | { | |
522 | if (sock->ops) { | |
523 | struct module *owner = sock->ops->owner; | |
524 | ||
525 | sock->ops->release(sock); | |
526 | sock->ops = NULL; | |
527 | module_put(owner); | |
528 | } | |
529 | ||
530 | if (sock->fasync_list) | |
531 | printk(KERN_ERR "sock_release: fasync list not empty!\n"); | |
532 | ||
533 | get_cpu_var(sockets_in_use)--; | |
534 | put_cpu_var(sockets_in_use); | |
535 | if (!sock->file) { | |
536 | iput(SOCK_INODE(sock)); | |
537 | return; | |
538 | } | |
89bddce5 | 539 | sock->file = NULL; |
1da177e4 LT |
540 | } |
541 | ||
89bddce5 | 542 | static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
543 | struct msghdr *msg, size_t size) |
544 | { | |
545 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
546 | int err; | |
547 | ||
548 | si->sock = sock; | |
549 | si->scm = NULL; | |
550 | si->msg = msg; | |
551 | si->size = size; | |
552 | ||
553 | err = security_socket_sendmsg(sock, msg, size); | |
554 | if (err) | |
555 | return err; | |
556 | ||
557 | return sock->ops->sendmsg(iocb, sock, msg, size); | |
558 | } | |
559 | ||
560 | int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) | |
561 | { | |
562 | struct kiocb iocb; | |
563 | struct sock_iocb siocb; | |
564 | int ret; | |
565 | ||
566 | init_sync_kiocb(&iocb, NULL); | |
567 | iocb.private = &siocb; | |
568 | ret = __sock_sendmsg(&iocb, sock, msg, size); | |
569 | if (-EIOCBQUEUED == ret) | |
570 | ret = wait_on_sync_kiocb(&iocb); | |
571 | return ret; | |
572 | } | |
573 | ||
574 | int kernel_sendmsg(struct socket *sock, struct msghdr *msg, | |
575 | struct kvec *vec, size_t num, size_t size) | |
576 | { | |
577 | mm_segment_t oldfs = get_fs(); | |
578 | int result; | |
579 | ||
580 | set_fs(KERNEL_DS); | |
581 | /* | |
582 | * the following is safe, since for compiler definitions of kvec and | |
583 | * iovec are identical, yielding the same in-core layout and alignment | |
584 | */ | |
89bddce5 | 585 | msg->msg_iov = (struct iovec *)vec; |
1da177e4 LT |
586 | msg->msg_iovlen = num; |
587 | result = sock_sendmsg(sock, msg, size); | |
588 | set_fs(oldfs); | |
589 | return result; | |
590 | } | |
591 | ||
92f37fd2 ED |
592 | /* |
593 | * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) | |
594 | */ | |
595 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, | |
596 | struct sk_buff *skb) | |
597 | { | |
598 | ktime_t kt = skb->tstamp; | |
599 | ||
600 | if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { | |
601 | struct timeval tv; | |
602 | /* Race occurred between timestamp enabling and packet | |
603 | receiving. Fill in the current time for now. */ | |
604 | if (kt.tv64 == 0) | |
605 | kt = ktime_get_real(); | |
606 | skb->tstamp = kt; | |
607 | tv = ktime_to_timeval(kt); | |
608 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv); | |
609 | } else { | |
610 | struct timespec ts; | |
611 | /* Race occurred between timestamp enabling and packet | |
612 | receiving. Fill in the current time for now. */ | |
613 | if (kt.tv64 == 0) | |
614 | kt = ktime_get_real(); | |
615 | skb->tstamp = kt; | |
616 | ts = ktime_to_timespec(kt); | |
617 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts); | |
618 | } | |
619 | } | |
620 | ||
7c81fd8b ACM |
621 | EXPORT_SYMBOL_GPL(__sock_recv_timestamp); |
622 | ||
89bddce5 | 623 | static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
624 | struct msghdr *msg, size_t size, int flags) |
625 | { | |
626 | int err; | |
627 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
628 | ||
629 | si->sock = sock; | |
630 | si->scm = NULL; | |
631 | si->msg = msg; | |
632 | si->size = size; | |
633 | si->flags = flags; | |
634 | ||
635 | err = security_socket_recvmsg(sock, msg, size, flags); | |
636 | if (err) | |
637 | return err; | |
638 | ||
639 | return sock->ops->recvmsg(iocb, sock, msg, size, flags); | |
640 | } | |
641 | ||
89bddce5 | 642 | int sock_recvmsg(struct socket *sock, struct msghdr *msg, |
1da177e4 LT |
643 | size_t size, int flags) |
644 | { | |
645 | struct kiocb iocb; | |
646 | struct sock_iocb siocb; | |
647 | int ret; | |
648 | ||
89bddce5 | 649 | init_sync_kiocb(&iocb, NULL); |
1da177e4 LT |
650 | iocb.private = &siocb; |
651 | ret = __sock_recvmsg(&iocb, sock, msg, size, flags); | |
652 | if (-EIOCBQUEUED == ret) | |
653 | ret = wait_on_sync_kiocb(&iocb); | |
654 | return ret; | |
655 | } | |
656 | ||
89bddce5 SH |
657 | int kernel_recvmsg(struct socket *sock, struct msghdr *msg, |
658 | struct kvec *vec, size_t num, size_t size, int flags) | |
1da177e4 LT |
659 | { |
660 | mm_segment_t oldfs = get_fs(); | |
661 | int result; | |
662 | ||
663 | set_fs(KERNEL_DS); | |
664 | /* | |
665 | * the following is safe, since for compiler definitions of kvec and | |
666 | * iovec are identical, yielding the same in-core layout and alignment | |
667 | */ | |
89bddce5 | 668 | msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num; |
1da177e4 LT |
669 | result = sock_recvmsg(sock, msg, size, flags); |
670 | set_fs(oldfs); | |
671 | return result; | |
672 | } | |
673 | ||
674 | static void sock_aio_dtor(struct kiocb *iocb) | |
675 | { | |
676 | kfree(iocb->private); | |
677 | } | |
678 | ||
ce1d4d3e CH |
679 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
680 | int offset, size_t size, loff_t *ppos, int more) | |
1da177e4 | 681 | { |
1da177e4 LT |
682 | struct socket *sock; |
683 | int flags; | |
684 | ||
ce1d4d3e CH |
685 | sock = file->private_data; |
686 | ||
687 | flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT; | |
688 | if (more) | |
689 | flags |= MSG_MORE; | |
690 | ||
691 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
692 | } | |
1da177e4 | 693 | |
ce1d4d3e | 694 | static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb, |
89bddce5 | 695 | struct sock_iocb *siocb) |
ce1d4d3e CH |
696 | { |
697 | if (!is_sync_kiocb(iocb)) { | |
698 | siocb = kmalloc(sizeof(*siocb), GFP_KERNEL); | |
699 | if (!siocb) | |
700 | return NULL; | |
1da177e4 LT |
701 | iocb->ki_dtor = sock_aio_dtor; |
702 | } | |
1da177e4 | 703 | |
ce1d4d3e | 704 | siocb->kiocb = iocb; |
ce1d4d3e CH |
705 | iocb->private = siocb; |
706 | return siocb; | |
1da177e4 LT |
707 | } |
708 | ||
ce1d4d3e | 709 | static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb, |
027445c3 BP |
710 | struct file *file, const struct iovec *iov, |
711 | unsigned long nr_segs) | |
ce1d4d3e CH |
712 | { |
713 | struct socket *sock = file->private_data; | |
714 | size_t size = 0; | |
715 | int i; | |
1da177e4 | 716 | |
89bddce5 SH |
717 | for (i = 0; i < nr_segs; i++) |
718 | size += iov[i].iov_len; | |
1da177e4 | 719 | |
ce1d4d3e CH |
720 | msg->msg_name = NULL; |
721 | msg->msg_namelen = 0; | |
722 | msg->msg_control = NULL; | |
723 | msg->msg_controllen = 0; | |
89bddce5 | 724 | msg->msg_iov = (struct iovec *)iov; |
ce1d4d3e CH |
725 | msg->msg_iovlen = nr_segs; |
726 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
727 | ||
728 | return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags); | |
729 | } | |
730 | ||
027445c3 BP |
731 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, |
732 | unsigned long nr_segs, loff_t pos) | |
ce1d4d3e CH |
733 | { |
734 | struct sock_iocb siocb, *x; | |
735 | ||
1da177e4 LT |
736 | if (pos != 0) |
737 | return -ESPIPE; | |
027445c3 BP |
738 | |
739 | if (iocb->ki_left == 0) /* Match SYS5 behaviour */ | |
1da177e4 LT |
740 | return 0; |
741 | ||
027445c3 BP |
742 | |
743 | x = alloc_sock_iocb(iocb, &siocb); | |
ce1d4d3e CH |
744 | if (!x) |
745 | return -ENOMEM; | |
027445c3 | 746 | return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
1da177e4 LT |
747 | } |
748 | ||
ce1d4d3e | 749 | static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb, |
027445c3 BP |
750 | struct file *file, const struct iovec *iov, |
751 | unsigned long nr_segs) | |
1da177e4 | 752 | { |
ce1d4d3e CH |
753 | struct socket *sock = file->private_data; |
754 | size_t size = 0; | |
755 | int i; | |
1da177e4 | 756 | |
89bddce5 SH |
757 | for (i = 0; i < nr_segs; i++) |
758 | size += iov[i].iov_len; | |
1da177e4 | 759 | |
ce1d4d3e CH |
760 | msg->msg_name = NULL; |
761 | msg->msg_namelen = 0; | |
762 | msg->msg_control = NULL; | |
763 | msg->msg_controllen = 0; | |
89bddce5 | 764 | msg->msg_iov = (struct iovec *)iov; |
ce1d4d3e CH |
765 | msg->msg_iovlen = nr_segs; |
766 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
767 | if (sock->type == SOCK_SEQPACKET) | |
768 | msg->msg_flags |= MSG_EOR; | |
1da177e4 | 769 | |
ce1d4d3e | 770 | return __sock_sendmsg(iocb, sock, msg, size); |
1da177e4 LT |
771 | } |
772 | ||
027445c3 BP |
773 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, |
774 | unsigned long nr_segs, loff_t pos) | |
ce1d4d3e CH |
775 | { |
776 | struct sock_iocb siocb, *x; | |
1da177e4 | 777 | |
ce1d4d3e CH |
778 | if (pos != 0) |
779 | return -ESPIPE; | |
027445c3 | 780 | |
027445c3 | 781 | x = alloc_sock_iocb(iocb, &siocb); |
ce1d4d3e CH |
782 | if (!x) |
783 | return -ENOMEM; | |
1da177e4 | 784 | |
027445c3 | 785 | return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
1da177e4 LT |
786 | } |
787 | ||
1da177e4 LT |
788 | /* |
789 | * Atomic setting of ioctl hooks to avoid race | |
790 | * with module unload. | |
791 | */ | |
792 | ||
4a3e2f71 | 793 | static DEFINE_MUTEX(br_ioctl_mutex); |
89bddce5 | 794 | static int (*br_ioctl_hook) (unsigned int cmd, void __user *arg) = NULL; |
1da177e4 | 795 | |
89bddce5 | 796 | void brioctl_set(int (*hook) (unsigned int, void __user *)) |
1da177e4 | 797 | { |
4a3e2f71 | 798 | mutex_lock(&br_ioctl_mutex); |
1da177e4 | 799 | br_ioctl_hook = hook; |
4a3e2f71 | 800 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 | 801 | } |
89bddce5 | 802 | |
1da177e4 LT |
803 | EXPORT_SYMBOL(brioctl_set); |
804 | ||
4a3e2f71 | 805 | static DEFINE_MUTEX(vlan_ioctl_mutex); |
89bddce5 | 806 | static int (*vlan_ioctl_hook) (void __user *arg); |
1da177e4 | 807 | |
89bddce5 | 808 | void vlan_ioctl_set(int (*hook) (void __user *)) |
1da177e4 | 809 | { |
4a3e2f71 | 810 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 811 | vlan_ioctl_hook = hook; |
4a3e2f71 | 812 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 813 | } |
89bddce5 | 814 | |
1da177e4 LT |
815 | EXPORT_SYMBOL(vlan_ioctl_set); |
816 | ||
4a3e2f71 | 817 | static DEFINE_MUTEX(dlci_ioctl_mutex); |
89bddce5 | 818 | static int (*dlci_ioctl_hook) (unsigned int, void __user *); |
1da177e4 | 819 | |
89bddce5 | 820 | void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) |
1da177e4 | 821 | { |
4a3e2f71 | 822 | mutex_lock(&dlci_ioctl_mutex); |
1da177e4 | 823 | dlci_ioctl_hook = hook; |
4a3e2f71 | 824 | mutex_unlock(&dlci_ioctl_mutex); |
1da177e4 | 825 | } |
89bddce5 | 826 | |
1da177e4 LT |
827 | EXPORT_SYMBOL(dlci_ioctl_set); |
828 | ||
829 | /* | |
830 | * With an ioctl, arg may well be a user mode pointer, but we don't know | |
831 | * what to do with it - that's up to the protocol still. | |
832 | */ | |
833 | ||
834 | static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) | |
835 | { | |
836 | struct socket *sock; | |
837 | void __user *argp = (void __user *)arg; | |
838 | int pid, err; | |
839 | ||
b69aee04 | 840 | sock = file->private_data; |
1da177e4 LT |
841 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) { |
842 | err = dev_ioctl(cmd, argp); | |
843 | } else | |
d86b5e0e | 844 | #ifdef CONFIG_WIRELESS_EXT |
1da177e4 LT |
845 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { |
846 | err = dev_ioctl(cmd, argp); | |
847 | } else | |
89bddce5 SH |
848 | #endif /* CONFIG_WIRELESS_EXT */ |
849 | switch (cmd) { | |
1da177e4 LT |
850 | case FIOSETOWN: |
851 | case SIOCSPGRP: | |
852 | err = -EFAULT; | |
853 | if (get_user(pid, (int __user *)argp)) | |
854 | break; | |
855 | err = f_setown(sock->file, pid, 1); | |
856 | break; | |
857 | case FIOGETOWN: | |
858 | case SIOCGPGRP: | |
609d7fa9 | 859 | err = put_user(f_getown(sock->file), |
89bddce5 | 860 | (int __user *)argp); |
1da177e4 LT |
861 | break; |
862 | case SIOCGIFBR: | |
863 | case SIOCSIFBR: | |
864 | case SIOCBRADDBR: | |
865 | case SIOCBRDELBR: | |
866 | err = -ENOPKG; | |
867 | if (!br_ioctl_hook) | |
868 | request_module("bridge"); | |
869 | ||
4a3e2f71 | 870 | mutex_lock(&br_ioctl_mutex); |
89bddce5 | 871 | if (br_ioctl_hook) |
1da177e4 | 872 | err = br_ioctl_hook(cmd, argp); |
4a3e2f71 | 873 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 LT |
874 | break; |
875 | case SIOCGIFVLAN: | |
876 | case SIOCSIFVLAN: | |
877 | err = -ENOPKG; | |
878 | if (!vlan_ioctl_hook) | |
879 | request_module("8021q"); | |
880 | ||
4a3e2f71 | 881 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 LT |
882 | if (vlan_ioctl_hook) |
883 | err = vlan_ioctl_hook(argp); | |
4a3e2f71 | 884 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 885 | break; |
1da177e4 LT |
886 | case SIOCADDDLCI: |
887 | case SIOCDELDLCI: | |
888 | err = -ENOPKG; | |
889 | if (!dlci_ioctl_hook) | |
890 | request_module("dlci"); | |
891 | ||
892 | if (dlci_ioctl_hook) { | |
4a3e2f71 | 893 | mutex_lock(&dlci_ioctl_mutex); |
1da177e4 | 894 | err = dlci_ioctl_hook(cmd, argp); |
4a3e2f71 | 895 | mutex_unlock(&dlci_ioctl_mutex); |
1da177e4 LT |
896 | } |
897 | break; | |
898 | default: | |
899 | err = sock->ops->ioctl(sock, cmd, arg); | |
b5e5fa5e CH |
900 | |
901 | /* | |
902 | * If this ioctl is unknown try to hand it down | |
903 | * to the NIC driver. | |
904 | */ | |
905 | if (err == -ENOIOCTLCMD) | |
906 | err = dev_ioctl(cmd, argp); | |
1da177e4 | 907 | break; |
89bddce5 | 908 | } |
1da177e4 LT |
909 | return err; |
910 | } | |
911 | ||
912 | int sock_create_lite(int family, int type, int protocol, struct socket **res) | |
913 | { | |
914 | int err; | |
915 | struct socket *sock = NULL; | |
89bddce5 | 916 | |
1da177e4 LT |
917 | err = security_socket_create(family, type, protocol, 1); |
918 | if (err) | |
919 | goto out; | |
920 | ||
921 | sock = sock_alloc(); | |
922 | if (!sock) { | |
923 | err = -ENOMEM; | |
924 | goto out; | |
925 | } | |
926 | ||
1da177e4 | 927 | sock->type = type; |
7420ed23 VY |
928 | err = security_socket_post_create(sock, family, type, protocol, 1); |
929 | if (err) | |
930 | goto out_release; | |
931 | ||
1da177e4 LT |
932 | out: |
933 | *res = sock; | |
934 | return err; | |
7420ed23 VY |
935 | out_release: |
936 | sock_release(sock); | |
937 | sock = NULL; | |
938 | goto out; | |
1da177e4 LT |
939 | } |
940 | ||
941 | /* No kernel lock held - perfect */ | |
89bddce5 | 942 | static unsigned int sock_poll(struct file *file, poll_table *wait) |
1da177e4 LT |
943 | { |
944 | struct socket *sock; | |
945 | ||
946 | /* | |
89bddce5 | 947 | * We can't return errors to poll, so it's either yes or no. |
1da177e4 | 948 | */ |
b69aee04 | 949 | sock = file->private_data; |
1da177e4 LT |
950 | return sock->ops->poll(file, sock, wait); |
951 | } | |
952 | ||
89bddce5 | 953 | static int sock_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 954 | { |
b69aee04 | 955 | struct socket *sock = file->private_data; |
1da177e4 LT |
956 | |
957 | return sock->ops->mmap(file, sock, vma); | |
958 | } | |
959 | ||
20380731 | 960 | static int sock_close(struct inode *inode, struct file *filp) |
1da177e4 LT |
961 | { |
962 | /* | |
89bddce5 SH |
963 | * It was possible the inode is NULL we were |
964 | * closing an unfinished socket. | |
1da177e4 LT |
965 | */ |
966 | ||
89bddce5 | 967 | if (!inode) { |
1da177e4 LT |
968 | printk(KERN_DEBUG "sock_close: NULL inode\n"); |
969 | return 0; | |
970 | } | |
971 | sock_fasync(-1, filp, 0); | |
972 | sock_release(SOCKET_I(inode)); | |
973 | return 0; | |
974 | } | |
975 | ||
976 | /* | |
977 | * Update the socket async list | |
978 | * | |
979 | * Fasync_list locking strategy. | |
980 | * | |
981 | * 1. fasync_list is modified only under process context socket lock | |
982 | * i.e. under semaphore. | |
983 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) | |
984 | * or under socket lock. | |
985 | * 3. fasync_list can be used from softirq context, so that | |
986 | * modification under socket lock have to be enhanced with | |
987 | * write_lock_bh(&sk->sk_callback_lock). | |
988 | * --ANK (990710) | |
989 | */ | |
990 | ||
991 | static int sock_fasync(int fd, struct file *filp, int on) | |
992 | { | |
89bddce5 | 993 | struct fasync_struct *fa, *fna = NULL, **prev; |
1da177e4 LT |
994 | struct socket *sock; |
995 | struct sock *sk; | |
996 | ||
89bddce5 | 997 | if (on) { |
8b3a7005 | 998 | fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL); |
89bddce5 | 999 | if (fna == NULL) |
1da177e4 LT |
1000 | return -ENOMEM; |
1001 | } | |
1002 | ||
b69aee04 | 1003 | sock = filp->private_data; |
1da177e4 | 1004 | |
89bddce5 SH |
1005 | sk = sock->sk; |
1006 | if (sk == NULL) { | |
1da177e4 LT |
1007 | kfree(fna); |
1008 | return -EINVAL; | |
1009 | } | |
1010 | ||
1011 | lock_sock(sk); | |
1012 | ||
89bddce5 | 1013 | prev = &(sock->fasync_list); |
1da177e4 | 1014 | |
89bddce5 SH |
1015 | for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev) |
1016 | if (fa->fa_file == filp) | |
1da177e4 LT |
1017 | break; |
1018 | ||
89bddce5 SH |
1019 | if (on) { |
1020 | if (fa != NULL) { | |
1da177e4 | 1021 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1022 | fa->fa_fd = fd; |
1da177e4 LT |
1023 | write_unlock_bh(&sk->sk_callback_lock); |
1024 | ||
1025 | kfree(fna); | |
1026 | goto out; | |
1027 | } | |
89bddce5 SH |
1028 | fna->fa_file = filp; |
1029 | fna->fa_fd = fd; | |
1030 | fna->magic = FASYNC_MAGIC; | |
1031 | fna->fa_next = sock->fasync_list; | |
1da177e4 | 1032 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1033 | sock->fasync_list = fna; |
1da177e4 | 1034 | write_unlock_bh(&sk->sk_callback_lock); |
89bddce5 SH |
1035 | } else { |
1036 | if (fa != NULL) { | |
1da177e4 | 1037 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1038 | *prev = fa->fa_next; |
1da177e4 LT |
1039 | write_unlock_bh(&sk->sk_callback_lock); |
1040 | kfree(fa); | |
1041 | } | |
1042 | } | |
1043 | ||
1044 | out: | |
1045 | release_sock(sock->sk); | |
1046 | return 0; | |
1047 | } | |
1048 | ||
1049 | /* This function may be called only under socket lock or callback_lock */ | |
1050 | ||
1051 | int sock_wake_async(struct socket *sock, int how, int band) | |
1052 | { | |
1053 | if (!sock || !sock->fasync_list) | |
1054 | return -1; | |
89bddce5 | 1055 | switch (how) { |
1da177e4 | 1056 | case 1: |
89bddce5 | 1057 | |
1da177e4 LT |
1058 | if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags)) |
1059 | break; | |
1060 | goto call_kill; | |
1061 | case 2: | |
1062 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags)) | |
1063 | break; | |
1064 | /* fall through */ | |
1065 | case 0: | |
89bddce5 | 1066 | call_kill: |
1da177e4 LT |
1067 | __kill_fasync(sock->fasync_list, SIGIO, band); |
1068 | break; | |
1069 | case 3: | |
1070 | __kill_fasync(sock->fasync_list, SIGURG, band); | |
1071 | } | |
1072 | return 0; | |
1073 | } | |
1074 | ||
1b8d7ae4 | 1075 | static int __sock_create(struct net *net, int family, int type, int protocol, |
89bddce5 | 1076 | struct socket **res, int kern) |
1da177e4 LT |
1077 | { |
1078 | int err; | |
1079 | struct socket *sock; | |
55737fda | 1080 | const struct net_proto_family *pf; |
1da177e4 LT |
1081 | |
1082 | /* | |
89bddce5 | 1083 | * Check protocol is in range |
1da177e4 LT |
1084 | */ |
1085 | if (family < 0 || family >= NPROTO) | |
1086 | return -EAFNOSUPPORT; | |
1087 | if (type < 0 || type >= SOCK_MAX) | |
1088 | return -EINVAL; | |
1089 | ||
1090 | /* Compatibility. | |
1091 | ||
1092 | This uglymoron is moved from INET layer to here to avoid | |
1093 | deadlock in module load. | |
1094 | */ | |
1095 | if (family == PF_INET && type == SOCK_PACKET) { | |
89bddce5 | 1096 | static int warned; |
1da177e4 LT |
1097 | if (!warned) { |
1098 | warned = 1; | |
89bddce5 SH |
1099 | printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", |
1100 | current->comm); | |
1da177e4 LT |
1101 | } |
1102 | family = PF_PACKET; | |
1103 | } | |
1104 | ||
1105 | err = security_socket_create(family, type, protocol, kern); | |
1106 | if (err) | |
1107 | return err; | |
89bddce5 | 1108 | |
55737fda SH |
1109 | /* |
1110 | * Allocate the socket and allow the family to set things up. if | |
1111 | * the protocol is 0, the family is instructed to select an appropriate | |
1112 | * default. | |
1113 | */ | |
1114 | sock = sock_alloc(); | |
1115 | if (!sock) { | |
1116 | if (net_ratelimit()) | |
1117 | printk(KERN_WARNING "socket: no more sockets\n"); | |
1118 | return -ENFILE; /* Not exactly a match, but its the | |
1119 | closest posix thing */ | |
1120 | } | |
1121 | ||
1122 | sock->type = type; | |
1123 | ||
1da177e4 | 1124 | #if defined(CONFIG_KMOD) |
89bddce5 SH |
1125 | /* Attempt to load a protocol module if the find failed. |
1126 | * | |
1127 | * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user | |
1da177e4 LT |
1128 | * requested real, full-featured networking support upon configuration. |
1129 | * Otherwise module support will break! | |
1130 | */ | |
55737fda | 1131 | if (net_families[family] == NULL) |
89bddce5 | 1132 | request_module("net-pf-%d", family); |
1da177e4 LT |
1133 | #endif |
1134 | ||
55737fda SH |
1135 | rcu_read_lock(); |
1136 | pf = rcu_dereference(net_families[family]); | |
1137 | err = -EAFNOSUPPORT; | |
1138 | if (!pf) | |
1139 | goto out_release; | |
1da177e4 LT |
1140 | |
1141 | /* | |
1142 | * We will call the ->create function, that possibly is in a loadable | |
1143 | * module, so we have to bump that loadable module refcnt first. | |
1144 | */ | |
55737fda | 1145 | if (!try_module_get(pf->owner)) |
1da177e4 LT |
1146 | goto out_release; |
1147 | ||
55737fda SH |
1148 | /* Now protected by module ref count */ |
1149 | rcu_read_unlock(); | |
1150 | ||
1b8d7ae4 | 1151 | err = pf->create(net, sock, protocol); |
55737fda | 1152 | if (err < 0) |
1da177e4 | 1153 | goto out_module_put; |
a79af59e | 1154 | |
1da177e4 LT |
1155 | /* |
1156 | * Now to bump the refcnt of the [loadable] module that owns this | |
1157 | * socket at sock_release time we decrement its refcnt. | |
1158 | */ | |
55737fda SH |
1159 | if (!try_module_get(sock->ops->owner)) |
1160 | goto out_module_busy; | |
1161 | ||
1da177e4 LT |
1162 | /* |
1163 | * Now that we're done with the ->create function, the [loadable] | |
1164 | * module can have its refcnt decremented | |
1165 | */ | |
55737fda | 1166 | module_put(pf->owner); |
7420ed23 VY |
1167 | err = security_socket_post_create(sock, family, type, protocol, kern); |
1168 | if (err) | |
3b185525 | 1169 | goto out_sock_release; |
55737fda | 1170 | *res = sock; |
1da177e4 | 1171 | |
55737fda SH |
1172 | return 0; |
1173 | ||
1174 | out_module_busy: | |
1175 | err = -EAFNOSUPPORT; | |
1da177e4 | 1176 | out_module_put: |
55737fda SH |
1177 | sock->ops = NULL; |
1178 | module_put(pf->owner); | |
1179 | out_sock_release: | |
1da177e4 | 1180 | sock_release(sock); |
55737fda SH |
1181 | return err; |
1182 | ||
1183 | out_release: | |
1184 | rcu_read_unlock(); | |
1185 | goto out_sock_release; | |
1da177e4 LT |
1186 | } |
1187 | ||
1188 | int sock_create(int family, int type, int protocol, struct socket **res) | |
1189 | { | |
1b8d7ae4 | 1190 | return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); |
1da177e4 LT |
1191 | } |
1192 | ||
1193 | int sock_create_kern(int family, int type, int protocol, struct socket **res) | |
1194 | { | |
1b8d7ae4 | 1195 | return __sock_create(&init_net, family, type, protocol, res, 1); |
1da177e4 LT |
1196 | } |
1197 | ||
1198 | asmlinkage long sys_socket(int family, int type, int protocol) | |
1199 | { | |
1200 | int retval; | |
1201 | struct socket *sock; | |
1202 | ||
1203 | retval = sock_create(family, type, protocol, &sock); | |
1204 | if (retval < 0) | |
1205 | goto out; | |
1206 | ||
1207 | retval = sock_map_fd(sock); | |
1208 | if (retval < 0) | |
1209 | goto out_release; | |
1210 | ||
1211 | out: | |
1212 | /* It may be already another descriptor 8) Not kernel problem. */ | |
1213 | return retval; | |
1214 | ||
1215 | out_release: | |
1216 | sock_release(sock); | |
1217 | return retval; | |
1218 | } | |
1219 | ||
1220 | /* | |
1221 | * Create a pair of connected sockets. | |
1222 | */ | |
1223 | ||
89bddce5 SH |
1224 | asmlinkage long sys_socketpair(int family, int type, int protocol, |
1225 | int __user *usockvec) | |
1da177e4 LT |
1226 | { |
1227 | struct socket *sock1, *sock2; | |
1228 | int fd1, fd2, err; | |
db349509 | 1229 | struct file *newfile1, *newfile2; |
1da177e4 LT |
1230 | |
1231 | /* | |
1232 | * Obtain the first socket and check if the underlying protocol | |
1233 | * supports the socketpair call. | |
1234 | */ | |
1235 | ||
1236 | err = sock_create(family, type, protocol, &sock1); | |
1237 | if (err < 0) | |
1238 | goto out; | |
1239 | ||
1240 | err = sock_create(family, type, protocol, &sock2); | |
1241 | if (err < 0) | |
1242 | goto out_release_1; | |
1243 | ||
1244 | err = sock1->ops->socketpair(sock1, sock2); | |
89bddce5 | 1245 | if (err < 0) |
1da177e4 LT |
1246 | goto out_release_both; |
1247 | ||
db349509 AV |
1248 | fd1 = sock_alloc_fd(&newfile1); |
1249 | if (unlikely(fd1 < 0)) | |
1250 | goto out_release_both; | |
1da177e4 | 1251 | |
db349509 AV |
1252 | fd2 = sock_alloc_fd(&newfile2); |
1253 | if (unlikely(fd2 < 0)) { | |
1254 | put_filp(newfile1); | |
1255 | put_unused_fd(fd1); | |
1da177e4 | 1256 | goto out_release_both; |
db349509 | 1257 | } |
1da177e4 | 1258 | |
db349509 AV |
1259 | err = sock_attach_fd(sock1, newfile1); |
1260 | if (unlikely(err < 0)) { | |
1261 | goto out_fd2; | |
1262 | } | |
1263 | ||
1264 | err = sock_attach_fd(sock2, newfile2); | |
1265 | if (unlikely(err < 0)) { | |
1266 | fput(newfile1); | |
1267 | goto out_fd1; | |
1268 | } | |
1269 | ||
1270 | err = audit_fd_pair(fd1, fd2); | |
1271 | if (err < 0) { | |
1272 | fput(newfile1); | |
1273 | fput(newfile2); | |
1274 | goto out_fd; | |
1275 | } | |
1da177e4 | 1276 | |
db349509 AV |
1277 | fd_install(fd1, newfile1); |
1278 | fd_install(fd2, newfile2); | |
1da177e4 LT |
1279 | /* fd1 and fd2 may be already another descriptors. |
1280 | * Not kernel problem. | |
1281 | */ | |
1282 | ||
89bddce5 | 1283 | err = put_user(fd1, &usockvec[0]); |
1da177e4 LT |
1284 | if (!err) |
1285 | err = put_user(fd2, &usockvec[1]); | |
1286 | if (!err) | |
1287 | return 0; | |
1288 | ||
1289 | sys_close(fd2); | |
1290 | sys_close(fd1); | |
1291 | return err; | |
1292 | ||
1da177e4 | 1293 | out_release_both: |
89bddce5 | 1294 | sock_release(sock2); |
1da177e4 | 1295 | out_release_1: |
89bddce5 | 1296 | sock_release(sock1); |
1da177e4 LT |
1297 | out: |
1298 | return err; | |
db349509 AV |
1299 | |
1300 | out_fd2: | |
1301 | put_filp(newfile1); | |
1302 | sock_release(sock1); | |
1303 | out_fd1: | |
1304 | put_filp(newfile2); | |
1305 | sock_release(sock2); | |
1306 | out_fd: | |
1307 | put_unused_fd(fd1); | |
1308 | put_unused_fd(fd2); | |
1309 | goto out; | |
1da177e4 LT |
1310 | } |
1311 | ||
1da177e4 LT |
1312 | /* |
1313 | * Bind a name to a socket. Nothing much to do here since it's | |
1314 | * the protocol's responsibility to handle the local address. | |
1315 | * | |
1316 | * We move the socket address to kernel space before we call | |
1317 | * the protocol layer (having also checked the address is ok). | |
1318 | */ | |
1319 | ||
1320 | asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) | |
1321 | { | |
1322 | struct socket *sock; | |
1323 | char address[MAX_SOCK_ADDR]; | |
6cb153ca | 1324 | int err, fput_needed; |
1da177e4 | 1325 | |
89bddce5 | 1326 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
e71a4783 | 1327 | if (sock) { |
89bddce5 SH |
1328 | err = move_addr_to_kernel(umyaddr, addrlen, address); |
1329 | if (err >= 0) { | |
1330 | err = security_socket_bind(sock, | |
1331 | (struct sockaddr *)address, | |
1332 | addrlen); | |
6cb153ca BL |
1333 | if (!err) |
1334 | err = sock->ops->bind(sock, | |
89bddce5 SH |
1335 | (struct sockaddr *) |
1336 | address, addrlen); | |
1da177e4 | 1337 | } |
6cb153ca | 1338 | fput_light(sock->file, fput_needed); |
89bddce5 | 1339 | } |
1da177e4 LT |
1340 | return err; |
1341 | } | |
1342 | ||
1da177e4 LT |
1343 | /* |
1344 | * Perform a listen. Basically, we allow the protocol to do anything | |
1345 | * necessary for a listen, and if that works, we mark the socket as | |
1346 | * ready for listening. | |
1347 | */ | |
1348 | ||
7a42c217 | 1349 | int sysctl_somaxconn __read_mostly = SOMAXCONN; |
1da177e4 LT |
1350 | |
1351 | asmlinkage long sys_listen(int fd, int backlog) | |
1352 | { | |
1353 | struct socket *sock; | |
6cb153ca | 1354 | int err, fput_needed; |
89bddce5 SH |
1355 | |
1356 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1357 | if (sock) { | |
1358 | if ((unsigned)backlog > sysctl_somaxconn) | |
1da177e4 LT |
1359 | backlog = sysctl_somaxconn; |
1360 | ||
1361 | err = security_socket_listen(sock, backlog); | |
6cb153ca BL |
1362 | if (!err) |
1363 | err = sock->ops->listen(sock, backlog); | |
1da177e4 | 1364 | |
6cb153ca | 1365 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1366 | } |
1367 | return err; | |
1368 | } | |
1369 | ||
1da177e4 LT |
1370 | /* |
1371 | * For accept, we attempt to create a new socket, set up the link | |
1372 | * with the client, wake up the client, then return the new | |
1373 | * connected fd. We collect the address of the connector in kernel | |
1374 | * space and move it to user at the very end. This is unclean because | |
1375 | * we open the socket then return an error. | |
1376 | * | |
1377 | * 1003.1g adds the ability to recvmsg() to query connection pending | |
1378 | * status to recvmsg. We need to add that support in a way thats | |
1379 | * clean when we restucture accept also. | |
1380 | */ | |
1381 | ||
89bddce5 SH |
1382 | asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, |
1383 | int __user *upeer_addrlen) | |
1da177e4 LT |
1384 | { |
1385 | struct socket *sock, *newsock; | |
39d8c1b6 | 1386 | struct file *newfile; |
6cb153ca | 1387 | int err, len, newfd, fput_needed; |
1da177e4 LT |
1388 | char address[MAX_SOCK_ADDR]; |
1389 | ||
6cb153ca | 1390 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1391 | if (!sock) |
1392 | goto out; | |
1393 | ||
1394 | err = -ENFILE; | |
89bddce5 | 1395 | if (!(newsock = sock_alloc())) |
1da177e4 LT |
1396 | goto out_put; |
1397 | ||
1398 | newsock->type = sock->type; | |
1399 | newsock->ops = sock->ops; | |
1400 | ||
1da177e4 LT |
1401 | /* |
1402 | * We don't need try_module_get here, as the listening socket (sock) | |
1403 | * has the protocol module (sock->ops->owner) held. | |
1404 | */ | |
1405 | __module_get(newsock->ops->owner); | |
1406 | ||
39d8c1b6 DM |
1407 | newfd = sock_alloc_fd(&newfile); |
1408 | if (unlikely(newfd < 0)) { | |
1409 | err = newfd; | |
9a1875e6 DM |
1410 | sock_release(newsock); |
1411 | goto out_put; | |
39d8c1b6 DM |
1412 | } |
1413 | ||
1414 | err = sock_attach_fd(newsock, newfile); | |
1415 | if (err < 0) | |
79f4f642 | 1416 | goto out_fd_simple; |
39d8c1b6 | 1417 | |
a79af59e FF |
1418 | err = security_socket_accept(sock, newsock); |
1419 | if (err) | |
39d8c1b6 | 1420 | goto out_fd; |
a79af59e | 1421 | |
1da177e4 LT |
1422 | err = sock->ops->accept(sock, newsock, sock->file->f_flags); |
1423 | if (err < 0) | |
39d8c1b6 | 1424 | goto out_fd; |
1da177e4 LT |
1425 | |
1426 | if (upeer_sockaddr) { | |
89bddce5 SH |
1427 | if (newsock->ops->getname(newsock, (struct sockaddr *)address, |
1428 | &len, 2) < 0) { | |
1da177e4 | 1429 | err = -ECONNABORTED; |
39d8c1b6 | 1430 | goto out_fd; |
1da177e4 | 1431 | } |
89bddce5 SH |
1432 | err = move_addr_to_user(address, len, upeer_sockaddr, |
1433 | upeer_addrlen); | |
1da177e4 | 1434 | if (err < 0) |
39d8c1b6 | 1435 | goto out_fd; |
1da177e4 LT |
1436 | } |
1437 | ||
1438 | /* File flags are not inherited via accept() unlike another OSes. */ | |
1439 | ||
39d8c1b6 DM |
1440 | fd_install(newfd, newfile); |
1441 | err = newfd; | |
1da177e4 LT |
1442 | |
1443 | security_socket_post_accept(sock, newsock); | |
1444 | ||
1445 | out_put: | |
6cb153ca | 1446 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1447 | out: |
1448 | return err; | |
79f4f642 AD |
1449 | out_fd_simple: |
1450 | sock_release(newsock); | |
1451 | put_filp(newfile); | |
1452 | put_unused_fd(newfd); | |
1453 | goto out_put; | |
39d8c1b6 | 1454 | out_fd: |
9606a216 | 1455 | fput(newfile); |
39d8c1b6 | 1456 | put_unused_fd(newfd); |
1da177e4 LT |
1457 | goto out_put; |
1458 | } | |
1459 | ||
1da177e4 LT |
1460 | /* |
1461 | * Attempt to connect to a socket with the server address. The address | |
1462 | * is in user space so we verify it is OK and move it to kernel space. | |
1463 | * | |
1464 | * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to | |
1465 | * break bindings | |
1466 | * | |
1467 | * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and | |
1468 | * other SEQPACKET protocols that take time to connect() as it doesn't | |
1469 | * include the -EINPROGRESS status for such sockets. | |
1470 | */ | |
1471 | ||
89bddce5 SH |
1472 | asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, |
1473 | int addrlen) | |
1da177e4 LT |
1474 | { |
1475 | struct socket *sock; | |
1476 | char address[MAX_SOCK_ADDR]; | |
6cb153ca | 1477 | int err, fput_needed; |
1da177e4 | 1478 | |
6cb153ca | 1479 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1480 | if (!sock) |
1481 | goto out; | |
1482 | err = move_addr_to_kernel(uservaddr, addrlen, address); | |
1483 | if (err < 0) | |
1484 | goto out_put; | |
1485 | ||
89bddce5 SH |
1486 | err = |
1487 | security_socket_connect(sock, (struct sockaddr *)address, addrlen); | |
1da177e4 LT |
1488 | if (err) |
1489 | goto out_put; | |
1490 | ||
89bddce5 | 1491 | err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen, |
1da177e4 LT |
1492 | sock->file->f_flags); |
1493 | out_put: | |
6cb153ca | 1494 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1495 | out: |
1496 | return err; | |
1497 | } | |
1498 | ||
1499 | /* | |
1500 | * Get the local address ('name') of a socket object. Move the obtained | |
1501 | * name to user space. | |
1502 | */ | |
1503 | ||
89bddce5 SH |
1504 | asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, |
1505 | int __user *usockaddr_len) | |
1da177e4 LT |
1506 | { |
1507 | struct socket *sock; | |
1508 | char address[MAX_SOCK_ADDR]; | |
6cb153ca | 1509 | int len, err, fput_needed; |
89bddce5 | 1510 | |
6cb153ca | 1511 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1512 | if (!sock) |
1513 | goto out; | |
1514 | ||
1515 | err = security_socket_getsockname(sock); | |
1516 | if (err) | |
1517 | goto out_put; | |
1518 | ||
1519 | err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0); | |
1520 | if (err) | |
1521 | goto out_put; | |
1522 | err = move_addr_to_user(address, len, usockaddr, usockaddr_len); | |
1523 | ||
1524 | out_put: | |
6cb153ca | 1525 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1526 | out: |
1527 | return err; | |
1528 | } | |
1529 | ||
1530 | /* | |
1531 | * Get the remote address ('name') of a socket object. Move the obtained | |
1532 | * name to user space. | |
1533 | */ | |
1534 | ||
89bddce5 SH |
1535 | asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, |
1536 | int __user *usockaddr_len) | |
1da177e4 LT |
1537 | { |
1538 | struct socket *sock; | |
1539 | char address[MAX_SOCK_ADDR]; | |
6cb153ca | 1540 | int len, err, fput_needed; |
1da177e4 | 1541 | |
89bddce5 SH |
1542 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1543 | if (sock != NULL) { | |
1da177e4 LT |
1544 | err = security_socket_getpeername(sock); |
1545 | if (err) { | |
6cb153ca | 1546 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1547 | return err; |
1548 | } | |
1549 | ||
89bddce5 SH |
1550 | err = |
1551 | sock->ops->getname(sock, (struct sockaddr *)address, &len, | |
1552 | 1); | |
1da177e4 | 1553 | if (!err) |
89bddce5 SH |
1554 | err = move_addr_to_user(address, len, usockaddr, |
1555 | usockaddr_len); | |
6cb153ca | 1556 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1557 | } |
1558 | return err; | |
1559 | } | |
1560 | ||
1561 | /* | |
1562 | * Send a datagram to a given address. We move the address into kernel | |
1563 | * space and check the user space data area is readable before invoking | |
1564 | * the protocol. | |
1565 | */ | |
1566 | ||
89bddce5 SH |
1567 | asmlinkage long sys_sendto(int fd, void __user *buff, size_t len, |
1568 | unsigned flags, struct sockaddr __user *addr, | |
1569 | int addr_len) | |
1da177e4 LT |
1570 | { |
1571 | struct socket *sock; | |
1572 | char address[MAX_SOCK_ADDR]; | |
1573 | int err; | |
1574 | struct msghdr msg; | |
1575 | struct iovec iov; | |
6cb153ca BL |
1576 | int fput_needed; |
1577 | struct file *sock_file; | |
1578 | ||
1579 | sock_file = fget_light(fd, &fput_needed); | |
4387ff75 | 1580 | err = -EBADF; |
6cb153ca | 1581 | if (!sock_file) |
4387ff75 | 1582 | goto out; |
6cb153ca BL |
1583 | |
1584 | sock = sock_from_file(sock_file, &err); | |
1da177e4 | 1585 | if (!sock) |
6cb153ca | 1586 | goto out_put; |
89bddce5 SH |
1587 | iov.iov_base = buff; |
1588 | iov.iov_len = len; | |
1589 | msg.msg_name = NULL; | |
1590 | msg.msg_iov = &iov; | |
1591 | msg.msg_iovlen = 1; | |
1592 | msg.msg_control = NULL; | |
1593 | msg.msg_controllen = 0; | |
1594 | msg.msg_namelen = 0; | |
6cb153ca | 1595 | if (addr) { |
1da177e4 LT |
1596 | err = move_addr_to_kernel(addr, addr_len, address); |
1597 | if (err < 0) | |
1598 | goto out_put; | |
89bddce5 SH |
1599 | msg.msg_name = address; |
1600 | msg.msg_namelen = addr_len; | |
1da177e4 LT |
1601 | } |
1602 | if (sock->file->f_flags & O_NONBLOCK) | |
1603 | flags |= MSG_DONTWAIT; | |
1604 | msg.msg_flags = flags; | |
1605 | err = sock_sendmsg(sock, &msg, len); | |
1606 | ||
89bddce5 | 1607 | out_put: |
6cb153ca | 1608 | fput_light(sock_file, fput_needed); |
4387ff75 | 1609 | out: |
1da177e4 LT |
1610 | return err; |
1611 | } | |
1612 | ||
1613 | /* | |
89bddce5 | 1614 | * Send a datagram down a socket. |
1da177e4 LT |
1615 | */ |
1616 | ||
89bddce5 | 1617 | asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags) |
1da177e4 LT |
1618 | { |
1619 | return sys_sendto(fd, buff, len, flags, NULL, 0); | |
1620 | } | |
1621 | ||
1622 | /* | |
89bddce5 | 1623 | * Receive a frame from the socket and optionally record the address of the |
1da177e4 LT |
1624 | * sender. We verify the buffers are writable and if needed move the |
1625 | * sender address from kernel to user space. | |
1626 | */ | |
1627 | ||
89bddce5 SH |
1628 | asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size, |
1629 | unsigned flags, struct sockaddr __user *addr, | |
1630 | int __user *addr_len) | |
1da177e4 LT |
1631 | { |
1632 | struct socket *sock; | |
1633 | struct iovec iov; | |
1634 | struct msghdr msg; | |
1635 | char address[MAX_SOCK_ADDR]; | |
89bddce5 | 1636 | int err, err2; |
6cb153ca BL |
1637 | struct file *sock_file; |
1638 | int fput_needed; | |
1639 | ||
1640 | sock_file = fget_light(fd, &fput_needed); | |
4387ff75 | 1641 | err = -EBADF; |
6cb153ca | 1642 | if (!sock_file) |
4387ff75 | 1643 | goto out; |
1da177e4 | 1644 | |
6cb153ca | 1645 | sock = sock_from_file(sock_file, &err); |
1da177e4 | 1646 | if (!sock) |
4387ff75 | 1647 | goto out_put; |
1da177e4 | 1648 | |
89bddce5 SH |
1649 | msg.msg_control = NULL; |
1650 | msg.msg_controllen = 0; | |
1651 | msg.msg_iovlen = 1; | |
1652 | msg.msg_iov = &iov; | |
1653 | iov.iov_len = size; | |
1654 | iov.iov_base = ubuf; | |
1655 | msg.msg_name = address; | |
1656 | msg.msg_namelen = MAX_SOCK_ADDR; | |
1da177e4 LT |
1657 | if (sock->file->f_flags & O_NONBLOCK) |
1658 | flags |= MSG_DONTWAIT; | |
89bddce5 | 1659 | err = sock_recvmsg(sock, &msg, size, flags); |
1da177e4 | 1660 | |
89bddce5 SH |
1661 | if (err >= 0 && addr != NULL) { |
1662 | err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len); | |
1663 | if (err2 < 0) | |
1664 | err = err2; | |
1da177e4 | 1665 | } |
4387ff75 | 1666 | out_put: |
6cb153ca | 1667 | fput_light(sock_file, fput_needed); |
4387ff75 | 1668 | out: |
1da177e4 LT |
1669 | return err; |
1670 | } | |
1671 | ||
1672 | /* | |
89bddce5 | 1673 | * Receive a datagram from a socket. |
1da177e4 LT |
1674 | */ |
1675 | ||
89bddce5 SH |
1676 | asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size, |
1677 | unsigned flags) | |
1da177e4 LT |
1678 | { |
1679 | return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); | |
1680 | } | |
1681 | ||
1682 | /* | |
1683 | * Set a socket option. Because we don't know the option lengths we have | |
1684 | * to pass the user mode parameter for the protocols to sort out. | |
1685 | */ | |
1686 | ||
89bddce5 SH |
1687 | asmlinkage long sys_setsockopt(int fd, int level, int optname, |
1688 | char __user *optval, int optlen) | |
1da177e4 | 1689 | { |
6cb153ca | 1690 | int err, fput_needed; |
1da177e4 LT |
1691 | struct socket *sock; |
1692 | ||
1693 | if (optlen < 0) | |
1694 | return -EINVAL; | |
89bddce5 SH |
1695 | |
1696 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1697 | if (sock != NULL) { | |
1698 | err = security_socket_setsockopt(sock, level, optname); | |
6cb153ca BL |
1699 | if (err) |
1700 | goto out_put; | |
1da177e4 LT |
1701 | |
1702 | if (level == SOL_SOCKET) | |
89bddce5 SH |
1703 | err = |
1704 | sock_setsockopt(sock, level, optname, optval, | |
1705 | optlen); | |
1da177e4 | 1706 | else |
89bddce5 SH |
1707 | err = |
1708 | sock->ops->setsockopt(sock, level, optname, optval, | |
1709 | optlen); | |
6cb153ca BL |
1710 | out_put: |
1711 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1712 | } |
1713 | return err; | |
1714 | } | |
1715 | ||
1716 | /* | |
1717 | * Get a socket option. Because we don't know the option lengths we have | |
1718 | * to pass a user mode parameter for the protocols to sort out. | |
1719 | */ | |
1720 | ||
89bddce5 SH |
1721 | asmlinkage long sys_getsockopt(int fd, int level, int optname, |
1722 | char __user *optval, int __user *optlen) | |
1da177e4 | 1723 | { |
6cb153ca | 1724 | int err, fput_needed; |
1da177e4 LT |
1725 | struct socket *sock; |
1726 | ||
89bddce5 SH |
1727 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1728 | if (sock != NULL) { | |
6cb153ca BL |
1729 | err = security_socket_getsockopt(sock, level, optname); |
1730 | if (err) | |
1731 | goto out_put; | |
1da177e4 LT |
1732 | |
1733 | if (level == SOL_SOCKET) | |
89bddce5 SH |
1734 | err = |
1735 | sock_getsockopt(sock, level, optname, optval, | |
1736 | optlen); | |
1da177e4 | 1737 | else |
89bddce5 SH |
1738 | err = |
1739 | sock->ops->getsockopt(sock, level, optname, optval, | |
1740 | optlen); | |
6cb153ca BL |
1741 | out_put: |
1742 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1743 | } |
1744 | return err; | |
1745 | } | |
1746 | ||
1da177e4 LT |
1747 | /* |
1748 | * Shutdown a socket. | |
1749 | */ | |
1750 | ||
1751 | asmlinkage long sys_shutdown(int fd, int how) | |
1752 | { | |
6cb153ca | 1753 | int err, fput_needed; |
1da177e4 LT |
1754 | struct socket *sock; |
1755 | ||
89bddce5 SH |
1756 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1757 | if (sock != NULL) { | |
1da177e4 | 1758 | err = security_socket_shutdown(sock, how); |
6cb153ca BL |
1759 | if (!err) |
1760 | err = sock->ops->shutdown(sock, how); | |
1761 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1762 | } |
1763 | return err; | |
1764 | } | |
1765 | ||
89bddce5 | 1766 | /* A couple of helpful macros for getting the address of the 32/64 bit |
1da177e4 LT |
1767 | * fields which are the same type (int / unsigned) on our platforms. |
1768 | */ | |
1769 | #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) | |
1770 | #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) | |
1771 | #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) | |
1772 | ||
1da177e4 LT |
1773 | /* |
1774 | * BSD sendmsg interface | |
1775 | */ | |
1776 | ||
1777 | asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags) | |
1778 | { | |
89bddce5 SH |
1779 | struct compat_msghdr __user *msg_compat = |
1780 | (struct compat_msghdr __user *)msg; | |
1da177e4 LT |
1781 | struct socket *sock; |
1782 | char address[MAX_SOCK_ADDR]; | |
1783 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; | |
b9d717a7 | 1784 | unsigned char ctl[sizeof(struct cmsghdr) + 20] |
89bddce5 SH |
1785 | __attribute__ ((aligned(sizeof(__kernel_size_t)))); |
1786 | /* 20 is size of ipv6_pktinfo */ | |
1da177e4 LT |
1787 | unsigned char *ctl_buf = ctl; |
1788 | struct msghdr msg_sys; | |
1789 | int err, ctl_len, iov_size, total_len; | |
6cb153ca | 1790 | int fput_needed; |
89bddce5 | 1791 | |
1da177e4 LT |
1792 | err = -EFAULT; |
1793 | if (MSG_CMSG_COMPAT & flags) { | |
1794 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1795 | return -EFAULT; | |
89bddce5 SH |
1796 | } |
1797 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1da177e4 LT |
1798 | return -EFAULT; |
1799 | ||
6cb153ca | 1800 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
89bddce5 | 1801 | if (!sock) |
1da177e4 LT |
1802 | goto out; |
1803 | ||
1804 | /* do not move before msg_sys is valid */ | |
1805 | err = -EMSGSIZE; | |
1806 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
1807 | goto out_put; | |
1808 | ||
89bddce5 | 1809 | /* Check whether to allocate the iovec area */ |
1da177e4 LT |
1810 | err = -ENOMEM; |
1811 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
1812 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
1813 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
1814 | if (!iov) | |
1815 | goto out_put; | |
1816 | } | |
1817 | ||
1818 | /* This will also move the address data into kernel space */ | |
1819 | if (MSG_CMSG_COMPAT & flags) { | |
1820 | err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ); | |
1821 | } else | |
1822 | err = verify_iovec(&msg_sys, iov, address, VERIFY_READ); | |
89bddce5 | 1823 | if (err < 0) |
1da177e4 LT |
1824 | goto out_freeiov; |
1825 | total_len = err; | |
1826 | ||
1827 | err = -ENOBUFS; | |
1828 | ||
1829 | if (msg_sys.msg_controllen > INT_MAX) | |
1830 | goto out_freeiov; | |
89bddce5 | 1831 | ctl_len = msg_sys.msg_controllen; |
1da177e4 | 1832 | if ((MSG_CMSG_COMPAT & flags) && ctl_len) { |
89bddce5 SH |
1833 | err = |
1834 | cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, | |
1835 | sizeof(ctl)); | |
1da177e4 LT |
1836 | if (err) |
1837 | goto out_freeiov; | |
1838 | ctl_buf = msg_sys.msg_control; | |
8920e8f9 | 1839 | ctl_len = msg_sys.msg_controllen; |
1da177e4 | 1840 | } else if (ctl_len) { |
89bddce5 | 1841 | if (ctl_len > sizeof(ctl)) { |
1da177e4 | 1842 | ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); |
89bddce5 | 1843 | if (ctl_buf == NULL) |
1da177e4 LT |
1844 | goto out_freeiov; |
1845 | } | |
1846 | err = -EFAULT; | |
1847 | /* | |
1848 | * Careful! Before this, msg_sys.msg_control contains a user pointer. | |
1849 | * Afterwards, it will be a kernel pointer. Thus the compiler-assisted | |
1850 | * checking falls down on this. | |
1851 | */ | |
89bddce5 SH |
1852 | if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control, |
1853 | ctl_len)) | |
1da177e4 LT |
1854 | goto out_freectl; |
1855 | msg_sys.msg_control = ctl_buf; | |
1856 | } | |
1857 | msg_sys.msg_flags = flags; | |
1858 | ||
1859 | if (sock->file->f_flags & O_NONBLOCK) | |
1860 | msg_sys.msg_flags |= MSG_DONTWAIT; | |
1861 | err = sock_sendmsg(sock, &msg_sys, total_len); | |
1862 | ||
1863 | out_freectl: | |
89bddce5 | 1864 | if (ctl_buf != ctl) |
1da177e4 LT |
1865 | sock_kfree_s(sock->sk, ctl_buf, ctl_len); |
1866 | out_freeiov: | |
1867 | if (iov != iovstack) | |
1868 | sock_kfree_s(sock->sk, iov, iov_size); | |
1869 | out_put: | |
6cb153ca | 1870 | fput_light(sock->file, fput_needed); |
89bddce5 | 1871 | out: |
1da177e4 LT |
1872 | return err; |
1873 | } | |
1874 | ||
1875 | /* | |
1876 | * BSD recvmsg interface | |
1877 | */ | |
1878 | ||
89bddce5 SH |
1879 | asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, |
1880 | unsigned int flags) | |
1da177e4 | 1881 | { |
89bddce5 SH |
1882 | struct compat_msghdr __user *msg_compat = |
1883 | (struct compat_msghdr __user *)msg; | |
1da177e4 LT |
1884 | struct socket *sock; |
1885 | struct iovec iovstack[UIO_FASTIOV]; | |
89bddce5 | 1886 | struct iovec *iov = iovstack; |
1da177e4 LT |
1887 | struct msghdr msg_sys; |
1888 | unsigned long cmsg_ptr; | |
1889 | int err, iov_size, total_len, len; | |
6cb153ca | 1890 | int fput_needed; |
1da177e4 LT |
1891 | |
1892 | /* kernel mode address */ | |
1893 | char addr[MAX_SOCK_ADDR]; | |
1894 | ||
1895 | /* user mode address pointers */ | |
1896 | struct sockaddr __user *uaddr; | |
1897 | int __user *uaddr_len; | |
89bddce5 | 1898 | |
1da177e4 LT |
1899 | if (MSG_CMSG_COMPAT & flags) { |
1900 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1901 | return -EFAULT; | |
89bddce5 SH |
1902 | } |
1903 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1904 | return -EFAULT; | |
1da177e4 | 1905 | |
6cb153ca | 1906 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1907 | if (!sock) |
1908 | goto out; | |
1909 | ||
1910 | err = -EMSGSIZE; | |
1911 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
1912 | goto out_put; | |
89bddce5 SH |
1913 | |
1914 | /* Check whether to allocate the iovec area */ | |
1da177e4 LT |
1915 | err = -ENOMEM; |
1916 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
1917 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
1918 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
1919 | if (!iov) | |
1920 | goto out_put; | |
1921 | } | |
1922 | ||
1923 | /* | |
89bddce5 SH |
1924 | * Save the user-mode address (verify_iovec will change the |
1925 | * kernel msghdr to use the kernel address space) | |
1da177e4 | 1926 | */ |
89bddce5 SH |
1927 | |
1928 | uaddr = (void __user *)msg_sys.msg_name; | |
1da177e4 LT |
1929 | uaddr_len = COMPAT_NAMELEN(msg); |
1930 | if (MSG_CMSG_COMPAT & flags) { | |
1931 | err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE); | |
1932 | } else | |
1933 | err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE); | |
1934 | if (err < 0) | |
1935 | goto out_freeiov; | |
89bddce5 | 1936 | total_len = err; |
1da177e4 LT |
1937 | |
1938 | cmsg_ptr = (unsigned long)msg_sys.msg_control; | |
4a19542e | 1939 | msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); |
89bddce5 | 1940 | |
1da177e4 LT |
1941 | if (sock->file->f_flags & O_NONBLOCK) |
1942 | flags |= MSG_DONTWAIT; | |
1943 | err = sock_recvmsg(sock, &msg_sys, total_len, flags); | |
1944 | if (err < 0) | |
1945 | goto out_freeiov; | |
1946 | len = err; | |
1947 | ||
1948 | if (uaddr != NULL) { | |
89bddce5 SH |
1949 | err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, |
1950 | uaddr_len); | |
1da177e4 LT |
1951 | if (err < 0) |
1952 | goto out_freeiov; | |
1953 | } | |
37f7f421 DM |
1954 | err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT), |
1955 | COMPAT_FLAGS(msg)); | |
1da177e4 LT |
1956 | if (err) |
1957 | goto out_freeiov; | |
1958 | if (MSG_CMSG_COMPAT & flags) | |
89bddce5 | 1959 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, |
1da177e4 LT |
1960 | &msg_compat->msg_controllen); |
1961 | else | |
89bddce5 | 1962 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, |
1da177e4 LT |
1963 | &msg->msg_controllen); |
1964 | if (err) | |
1965 | goto out_freeiov; | |
1966 | err = len; | |
1967 | ||
1968 | out_freeiov: | |
1969 | if (iov != iovstack) | |
1970 | sock_kfree_s(sock->sk, iov, iov_size); | |
1971 | out_put: | |
6cb153ca | 1972 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1973 | out: |
1974 | return err; | |
1975 | } | |
1976 | ||
1977 | #ifdef __ARCH_WANT_SYS_SOCKETCALL | |
1978 | ||
1979 | /* Argument list sizes for sys_socketcall */ | |
1980 | #define AL(x) ((x) * sizeof(unsigned long)) | |
89bddce5 SH |
1981 | static const unsigned char nargs[18]={ |
1982 | AL(0),AL(3),AL(3),AL(3),AL(2),AL(3), | |
1983 | AL(3),AL(3),AL(4),AL(4),AL(4),AL(6), | |
1984 | AL(6),AL(2),AL(5),AL(5),AL(3),AL(3) | |
1985 | }; | |
1986 | ||
1da177e4 LT |
1987 | #undef AL |
1988 | ||
1989 | /* | |
89bddce5 | 1990 | * System call vectors. |
1da177e4 LT |
1991 | * |
1992 | * Argument checking cleaned up. Saved 20% in size. | |
1993 | * This function doesn't need to set the kernel lock because | |
89bddce5 | 1994 | * it is set by the callees. |
1da177e4 LT |
1995 | */ |
1996 | ||
1997 | asmlinkage long sys_socketcall(int call, unsigned long __user *args) | |
1998 | { | |
1999 | unsigned long a[6]; | |
89bddce5 | 2000 | unsigned long a0, a1; |
1da177e4 LT |
2001 | int err; |
2002 | ||
89bddce5 | 2003 | if (call < 1 || call > SYS_RECVMSG) |
1da177e4 LT |
2004 | return -EINVAL; |
2005 | ||
2006 | /* copy_from_user should be SMP safe. */ | |
2007 | if (copy_from_user(a, args, nargs[call])) | |
2008 | return -EFAULT; | |
3ec3b2fb | 2009 | |
89bddce5 | 2010 | err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); |
3ec3b2fb DW |
2011 | if (err) |
2012 | return err; | |
2013 | ||
89bddce5 SH |
2014 | a0 = a[0]; |
2015 | a1 = a[1]; | |
2016 | ||
2017 | switch (call) { | |
2018 | case SYS_SOCKET: | |
2019 | err = sys_socket(a0, a1, a[2]); | |
2020 | break; | |
2021 | case SYS_BIND: | |
2022 | err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]); | |
2023 | break; | |
2024 | case SYS_CONNECT: | |
2025 | err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]); | |
2026 | break; | |
2027 | case SYS_LISTEN: | |
2028 | err = sys_listen(a0, a1); | |
2029 | break; | |
2030 | case SYS_ACCEPT: | |
2031 | err = | |
2032 | sys_accept(a0, (struct sockaddr __user *)a1, | |
2033 | (int __user *)a[2]); | |
2034 | break; | |
2035 | case SYS_GETSOCKNAME: | |
2036 | err = | |
2037 | sys_getsockname(a0, (struct sockaddr __user *)a1, | |
2038 | (int __user *)a[2]); | |
2039 | break; | |
2040 | case SYS_GETPEERNAME: | |
2041 | err = | |
2042 | sys_getpeername(a0, (struct sockaddr __user *)a1, | |
2043 | (int __user *)a[2]); | |
2044 | break; | |
2045 | case SYS_SOCKETPAIR: | |
2046 | err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]); | |
2047 | break; | |
2048 | case SYS_SEND: | |
2049 | err = sys_send(a0, (void __user *)a1, a[2], a[3]); | |
2050 | break; | |
2051 | case SYS_SENDTO: | |
2052 | err = sys_sendto(a0, (void __user *)a1, a[2], a[3], | |
2053 | (struct sockaddr __user *)a[4], a[5]); | |
2054 | break; | |
2055 | case SYS_RECV: | |
2056 | err = sys_recv(a0, (void __user *)a1, a[2], a[3]); | |
2057 | break; | |
2058 | case SYS_RECVFROM: | |
2059 | err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3], | |
2060 | (struct sockaddr __user *)a[4], | |
2061 | (int __user *)a[5]); | |
2062 | break; | |
2063 | case SYS_SHUTDOWN: | |
2064 | err = sys_shutdown(a0, a1); | |
2065 | break; | |
2066 | case SYS_SETSOCKOPT: | |
2067 | err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]); | |
2068 | break; | |
2069 | case SYS_GETSOCKOPT: | |
2070 | err = | |
2071 | sys_getsockopt(a0, a1, a[2], (char __user *)a[3], | |
2072 | (int __user *)a[4]); | |
2073 | break; | |
2074 | case SYS_SENDMSG: | |
2075 | err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2076 | break; | |
2077 | case SYS_RECVMSG: | |
2078 | err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2079 | break; | |
2080 | default: | |
2081 | err = -EINVAL; | |
2082 | break; | |
1da177e4 LT |
2083 | } |
2084 | return err; | |
2085 | } | |
2086 | ||
89bddce5 | 2087 | #endif /* __ARCH_WANT_SYS_SOCKETCALL */ |
1da177e4 | 2088 | |
55737fda SH |
2089 | /** |
2090 | * sock_register - add a socket protocol handler | |
2091 | * @ops: description of protocol | |
2092 | * | |
1da177e4 LT |
2093 | * This function is called by a protocol handler that wants to |
2094 | * advertise its address family, and have it linked into the | |
55737fda SH |
2095 | * socket interface. The value ops->family coresponds to the |
2096 | * socket system call protocol family. | |
1da177e4 | 2097 | */ |
f0fd27d4 | 2098 | int sock_register(const struct net_proto_family *ops) |
1da177e4 LT |
2099 | { |
2100 | int err; | |
2101 | ||
2102 | if (ops->family >= NPROTO) { | |
89bddce5 SH |
2103 | printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, |
2104 | NPROTO); | |
1da177e4 LT |
2105 | return -ENOBUFS; |
2106 | } | |
55737fda SH |
2107 | |
2108 | spin_lock(&net_family_lock); | |
2109 | if (net_families[ops->family]) | |
2110 | err = -EEXIST; | |
2111 | else { | |
89bddce5 | 2112 | net_families[ops->family] = ops; |
1da177e4 LT |
2113 | err = 0; |
2114 | } | |
55737fda SH |
2115 | spin_unlock(&net_family_lock); |
2116 | ||
89bddce5 | 2117 | printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family); |
1da177e4 LT |
2118 | return err; |
2119 | } | |
2120 | ||
55737fda SH |
2121 | /** |
2122 | * sock_unregister - remove a protocol handler | |
2123 | * @family: protocol family to remove | |
2124 | * | |
1da177e4 LT |
2125 | * This function is called by a protocol handler that wants to |
2126 | * remove its address family, and have it unlinked from the | |
55737fda SH |
2127 | * new socket creation. |
2128 | * | |
2129 | * If protocol handler is a module, then it can use module reference | |
2130 | * counts to protect against new references. If protocol handler is not | |
2131 | * a module then it needs to provide its own protection in | |
2132 | * the ops->create routine. | |
1da177e4 | 2133 | */ |
f0fd27d4 | 2134 | void sock_unregister(int family) |
1da177e4 | 2135 | { |
f0fd27d4 | 2136 | BUG_ON(family < 0 || family >= NPROTO); |
1da177e4 | 2137 | |
55737fda | 2138 | spin_lock(&net_family_lock); |
89bddce5 | 2139 | net_families[family] = NULL; |
55737fda SH |
2140 | spin_unlock(&net_family_lock); |
2141 | ||
2142 | synchronize_rcu(); | |
2143 | ||
89bddce5 | 2144 | printk(KERN_INFO "NET: Unregistered protocol family %d\n", family); |
1da177e4 LT |
2145 | } |
2146 | ||
77d76ea3 | 2147 | static int __init sock_init(void) |
1da177e4 LT |
2148 | { |
2149 | /* | |
89bddce5 | 2150 | * Initialize sock SLAB cache. |
1da177e4 | 2151 | */ |
89bddce5 | 2152 | |
1da177e4 LT |
2153 | sk_init(); |
2154 | ||
1da177e4 | 2155 | /* |
89bddce5 | 2156 | * Initialize skbuff SLAB cache |
1da177e4 LT |
2157 | */ |
2158 | skb_init(); | |
1da177e4 LT |
2159 | |
2160 | /* | |
89bddce5 | 2161 | * Initialize the protocols module. |
1da177e4 LT |
2162 | */ |
2163 | ||
2164 | init_inodecache(); | |
2165 | register_filesystem(&sock_fs_type); | |
2166 | sock_mnt = kern_mount(&sock_fs_type); | |
77d76ea3 AK |
2167 | |
2168 | /* The real protocol initialization is performed in later initcalls. | |
1da177e4 LT |
2169 | */ |
2170 | ||
2171 | #ifdef CONFIG_NETFILTER | |
2172 | netfilter_init(); | |
2173 | #endif | |
cbeb321a DM |
2174 | |
2175 | return 0; | |
1da177e4 LT |
2176 | } |
2177 | ||
77d76ea3 AK |
2178 | core_initcall(sock_init); /* early initcall */ |
2179 | ||
1da177e4 LT |
2180 | #ifdef CONFIG_PROC_FS |
2181 | void socket_seq_show(struct seq_file *seq) | |
2182 | { | |
2183 | int cpu; | |
2184 | int counter = 0; | |
2185 | ||
6f912042 | 2186 | for_each_possible_cpu(cpu) |
89bddce5 | 2187 | counter += per_cpu(sockets_in_use, cpu); |
1da177e4 LT |
2188 | |
2189 | /* It can be negative, by the way. 8) */ | |
2190 | if (counter < 0) | |
2191 | counter = 0; | |
2192 | ||
2193 | seq_printf(seq, "sockets: used %d\n", counter); | |
2194 | } | |
89bddce5 | 2195 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 2196 | |
89bbfc95 SP |
2197 | #ifdef CONFIG_COMPAT |
2198 | static long compat_sock_ioctl(struct file *file, unsigned cmd, | |
89bddce5 | 2199 | unsigned long arg) |
89bbfc95 SP |
2200 | { |
2201 | struct socket *sock = file->private_data; | |
2202 | int ret = -ENOIOCTLCMD; | |
2203 | ||
2204 | if (sock->ops->compat_ioctl) | |
2205 | ret = sock->ops->compat_ioctl(sock, cmd, arg); | |
2206 | ||
2207 | return ret; | |
2208 | } | |
2209 | #endif | |
2210 | ||
ac5a488e SS |
2211 | int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) |
2212 | { | |
2213 | return sock->ops->bind(sock, addr, addrlen); | |
2214 | } | |
2215 | ||
2216 | int kernel_listen(struct socket *sock, int backlog) | |
2217 | { | |
2218 | return sock->ops->listen(sock, backlog); | |
2219 | } | |
2220 | ||
2221 | int kernel_accept(struct socket *sock, struct socket **newsock, int flags) | |
2222 | { | |
2223 | struct sock *sk = sock->sk; | |
2224 | int err; | |
2225 | ||
2226 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
2227 | newsock); | |
2228 | if (err < 0) | |
2229 | goto done; | |
2230 | ||
2231 | err = sock->ops->accept(sock, *newsock, flags); | |
2232 | if (err < 0) { | |
2233 | sock_release(*newsock); | |
2234 | goto done; | |
2235 | } | |
2236 | ||
2237 | (*newsock)->ops = sock->ops; | |
2238 | ||
2239 | done: | |
2240 | return err; | |
2241 | } | |
2242 | ||
2243 | int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, | |
4768fbcb | 2244 | int flags) |
ac5a488e SS |
2245 | { |
2246 | return sock->ops->connect(sock, addr, addrlen, flags); | |
2247 | } | |
2248 | ||
2249 | int kernel_getsockname(struct socket *sock, struct sockaddr *addr, | |
2250 | int *addrlen) | |
2251 | { | |
2252 | return sock->ops->getname(sock, addr, addrlen, 0); | |
2253 | } | |
2254 | ||
2255 | int kernel_getpeername(struct socket *sock, struct sockaddr *addr, | |
2256 | int *addrlen) | |
2257 | { | |
2258 | return sock->ops->getname(sock, addr, addrlen, 1); | |
2259 | } | |
2260 | ||
2261 | int kernel_getsockopt(struct socket *sock, int level, int optname, | |
2262 | char *optval, int *optlen) | |
2263 | { | |
2264 | mm_segment_t oldfs = get_fs(); | |
2265 | int err; | |
2266 | ||
2267 | set_fs(KERNEL_DS); | |
2268 | if (level == SOL_SOCKET) | |
2269 | err = sock_getsockopt(sock, level, optname, optval, optlen); | |
2270 | else | |
2271 | err = sock->ops->getsockopt(sock, level, optname, optval, | |
2272 | optlen); | |
2273 | set_fs(oldfs); | |
2274 | return err; | |
2275 | } | |
2276 | ||
2277 | int kernel_setsockopt(struct socket *sock, int level, int optname, | |
2278 | char *optval, int optlen) | |
2279 | { | |
2280 | mm_segment_t oldfs = get_fs(); | |
2281 | int err; | |
2282 | ||
2283 | set_fs(KERNEL_DS); | |
2284 | if (level == SOL_SOCKET) | |
2285 | err = sock_setsockopt(sock, level, optname, optval, optlen); | |
2286 | else | |
2287 | err = sock->ops->setsockopt(sock, level, optname, optval, | |
2288 | optlen); | |
2289 | set_fs(oldfs); | |
2290 | return err; | |
2291 | } | |
2292 | ||
2293 | int kernel_sendpage(struct socket *sock, struct page *page, int offset, | |
2294 | size_t size, int flags) | |
2295 | { | |
2296 | if (sock->ops->sendpage) | |
2297 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
2298 | ||
2299 | return sock_no_sendpage(sock, page, offset, size, flags); | |
2300 | } | |
2301 | ||
2302 | int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg) | |
2303 | { | |
2304 | mm_segment_t oldfs = get_fs(); | |
2305 | int err; | |
2306 | ||
2307 | set_fs(KERNEL_DS); | |
2308 | err = sock->ops->ioctl(sock, cmd, arg); | |
2309 | set_fs(oldfs); | |
2310 | ||
2311 | return err; | |
2312 | } | |
2313 | ||
1da177e4 LT |
2314 | /* ABI emulation layers need these two */ |
2315 | EXPORT_SYMBOL(move_addr_to_kernel); | |
2316 | EXPORT_SYMBOL(move_addr_to_user); | |
2317 | EXPORT_SYMBOL(sock_create); | |
2318 | EXPORT_SYMBOL(sock_create_kern); | |
2319 | EXPORT_SYMBOL(sock_create_lite); | |
2320 | EXPORT_SYMBOL(sock_map_fd); | |
2321 | EXPORT_SYMBOL(sock_recvmsg); | |
2322 | EXPORT_SYMBOL(sock_register); | |
2323 | EXPORT_SYMBOL(sock_release); | |
2324 | EXPORT_SYMBOL(sock_sendmsg); | |
2325 | EXPORT_SYMBOL(sock_unregister); | |
2326 | EXPORT_SYMBOL(sock_wake_async); | |
2327 | EXPORT_SYMBOL(sockfd_lookup); | |
2328 | EXPORT_SYMBOL(kernel_sendmsg); | |
2329 | EXPORT_SYMBOL(kernel_recvmsg); | |
ac5a488e SS |
2330 | EXPORT_SYMBOL(kernel_bind); |
2331 | EXPORT_SYMBOL(kernel_listen); | |
2332 | EXPORT_SYMBOL(kernel_accept); | |
2333 | EXPORT_SYMBOL(kernel_connect); | |
2334 | EXPORT_SYMBOL(kernel_getsockname); | |
2335 | EXPORT_SYMBOL(kernel_getpeername); | |
2336 | EXPORT_SYMBOL(kernel_getsockopt); | |
2337 | EXPORT_SYMBOL(kernel_setsockopt); | |
2338 | EXPORT_SYMBOL(kernel_sendpage); | |
2339 | EXPORT_SYMBOL(kernel_sock_ioctl); |