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> | |
80 | #include <linux/divert.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> | |
3ec3b2fb | 86 | #include <linux/audit.h> |
d86b5e0e | 87 | #include <linux/wireless.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 LT |
112 | static int sock_fasync(int fd, struct file *filp, int on); |
113 | static ssize_t sock_readv(struct file *file, const struct iovec *vector, | |
114 | unsigned long count, loff_t *ppos); | |
115 | static ssize_t sock_writev(struct file *file, const struct iovec *vector, | |
89bddce5 | 116 | unsigned long count, loff_t *ppos); |
1da177e4 LT |
117 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
118 | int offset, size_t size, loff_t *ppos, int more); | |
119 | ||
1da177e4 LT |
120 | /* |
121 | * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear | |
122 | * in the operation structures but are done directly via the socketcall() multiplexor. | |
123 | */ | |
124 | ||
125 | static struct file_operations socket_file_ops = { | |
126 | .owner = THIS_MODULE, | |
127 | .llseek = no_llseek, | |
128 | .aio_read = sock_aio_read, | |
129 | .aio_write = sock_aio_write, | |
130 | .poll = sock_poll, | |
131 | .unlocked_ioctl = sock_ioctl, | |
89bbfc95 SP |
132 | #ifdef CONFIG_COMPAT |
133 | .compat_ioctl = compat_sock_ioctl, | |
134 | #endif | |
1da177e4 LT |
135 | .mmap = sock_mmap, |
136 | .open = sock_no_open, /* special open code to disallow open via /proc */ | |
137 | .release = sock_close, | |
138 | .fasync = sock_fasync, | |
139 | .readv = sock_readv, | |
140 | .writev = sock_writev, | |
5274f052 JA |
141 | .sendpage = sock_sendpage, |
142 | .splice_write = generic_splice_sendpage, | |
1da177e4 LT |
143 | }; |
144 | ||
145 | /* | |
146 | * The protocol list. Each protocol is registered in here. | |
147 | */ | |
148 | ||
1da177e4 | 149 | static DEFINE_SPINLOCK(net_family_lock); |
f0fd27d4 | 150 | static const struct net_proto_family *net_families[NPROTO] __read_mostly; |
1da177e4 | 151 | |
1da177e4 LT |
152 | /* |
153 | * Statistics counters of the socket lists | |
154 | */ | |
155 | ||
156 | static DEFINE_PER_CPU(int, sockets_in_use) = 0; | |
157 | ||
158 | /* | |
89bddce5 SH |
159 | * Support routines. |
160 | * Move socket addresses back and forth across the kernel/user | |
161 | * divide and look after the messy bits. | |
1da177e4 LT |
162 | */ |
163 | ||
89bddce5 | 164 | #define MAX_SOCK_ADDR 128 /* 108 for Unix domain - |
1da177e4 LT |
165 | 16 for IP, 16 for IPX, |
166 | 24 for IPv6, | |
89bddce5 | 167 | about 80 for AX.25 |
1da177e4 LT |
168 | must be at least one bigger than |
169 | the AF_UNIX size (see net/unix/af_unix.c | |
89bddce5 | 170 | :unix_mkname()). |
1da177e4 | 171 | */ |
89bddce5 | 172 | |
1da177e4 LT |
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, void *kaddr) | |
185 | { | |
89bddce5 | 186 | if (ulen < 0 || ulen > MAX_SOCK_ADDR) |
1da177e4 | 187 | return -EINVAL; |
89bddce5 | 188 | if (ulen == 0) |
1da177e4 | 189 | return 0; |
89bddce5 | 190 | if (copy_from_user(kaddr, uaddr, ulen)) |
1da177e4 | 191 | return -EFAULT; |
3ec3b2fb | 192 | return audit_sockaddr(ulen, kaddr); |
1da177e4 LT |
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 | */ | |
89bddce5 SH |
211 | |
212 | int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, | |
213 | int __user *ulen) | |
1da177e4 LT |
214 | { |
215 | int err; | |
216 | int len; | |
217 | ||
89bddce5 SH |
218 | err = get_user(len, ulen); |
219 | if (err) | |
1da177e4 | 220 | return err; |
89bddce5 SH |
221 | if (len > klen) |
222 | len = klen; | |
223 | if (len < 0 || len > MAX_SOCK_ADDR) | |
1da177e4 | 224 | return -EINVAL; |
89bddce5 | 225 | if (len) { |
d6fe3945 SG |
226 | if (audit_sockaddr(klen, kaddr)) |
227 | return -ENOMEM; | |
89bddce5 | 228 | if (copy_to_user(uaddr, kaddr, len)) |
1da177e4 LT |
229 | return -EFAULT; |
230 | } | |
231 | /* | |
89bddce5 SH |
232 | * "fromlen shall refer to the value before truncation.." |
233 | * 1003.1g | |
1da177e4 LT |
234 | */ |
235 | return __put_user(klen, ulen); | |
236 | } | |
237 | ||
238 | #define SOCKFS_MAGIC 0x534F434B | |
239 | ||
89bddce5 | 240 | static kmem_cache_t *sock_inode_cachep __read_mostly; |
1da177e4 LT |
241 | |
242 | static struct inode *sock_alloc_inode(struct super_block *sb) | |
243 | { | |
244 | struct socket_alloc *ei; | |
89bddce5 SH |
245 | |
246 | ei = kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL); | |
1da177e4 LT |
247 | if (!ei) |
248 | return NULL; | |
249 | init_waitqueue_head(&ei->socket.wait); | |
89bddce5 | 250 | |
1da177e4 LT |
251 | ei->socket.fasync_list = NULL; |
252 | ei->socket.state = SS_UNCONNECTED; | |
253 | ei->socket.flags = 0; | |
254 | ei->socket.ops = NULL; | |
255 | ei->socket.sk = NULL; | |
256 | ei->socket.file = NULL; | |
1da177e4 LT |
257 | |
258 | return &ei->vfs_inode; | |
259 | } | |
260 | ||
261 | static void sock_destroy_inode(struct inode *inode) | |
262 | { | |
263 | kmem_cache_free(sock_inode_cachep, | |
264 | container_of(inode, struct socket_alloc, vfs_inode)); | |
265 | } | |
266 | ||
89bddce5 | 267 | static void init_once(void *foo, kmem_cache_t *cachep, unsigned long flags) |
1da177e4 | 268 | { |
89bddce5 | 269 | struct socket_alloc *ei = (struct socket_alloc *)foo; |
1da177e4 | 270 | |
89bddce5 SH |
271 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) |
272 | == SLAB_CTOR_CONSTRUCTOR) | |
1da177e4 LT |
273 | inode_init_once(&ei->vfs_inode); |
274 | } | |
89bddce5 | 275 | |
1da177e4 LT |
276 | static int init_inodecache(void) |
277 | { | |
278 | sock_inode_cachep = kmem_cache_create("sock_inode_cache", | |
89bddce5 SH |
279 | sizeof(struct socket_alloc), |
280 | 0, | |
281 | (SLAB_HWCACHE_ALIGN | | |
282 | SLAB_RECLAIM_ACCOUNT | | |
283 | SLAB_MEM_SPREAD), | |
284 | init_once, | |
285 | NULL); | |
1da177e4 LT |
286 | if (sock_inode_cachep == NULL) |
287 | return -ENOMEM; | |
288 | return 0; | |
289 | } | |
290 | ||
291 | static struct super_operations sockfs_ops = { | |
292 | .alloc_inode = sock_alloc_inode, | |
293 | .destroy_inode =sock_destroy_inode, | |
294 | .statfs = simple_statfs, | |
295 | }; | |
296 | ||
454e2398 | 297 | static int sockfs_get_sb(struct file_system_type *fs_type, |
89bddce5 SH |
298 | int flags, const char *dev_name, void *data, |
299 | struct vfsmount *mnt) | |
1da177e4 | 300 | { |
454e2398 DH |
301 | return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC, |
302 | mnt); | |
1da177e4 LT |
303 | } |
304 | ||
ba89966c | 305 | static struct vfsmount *sock_mnt __read_mostly; |
1da177e4 LT |
306 | |
307 | static struct file_system_type sock_fs_type = { | |
308 | .name = "sockfs", | |
309 | .get_sb = sockfs_get_sb, | |
310 | .kill_sb = kill_anon_super, | |
311 | }; | |
89bddce5 | 312 | |
1da177e4 LT |
313 | static int sockfs_delete_dentry(struct dentry *dentry) |
314 | { | |
315 | return 1; | |
316 | } | |
317 | static struct dentry_operations sockfs_dentry_operations = { | |
89bddce5 | 318 | .d_delete = sockfs_delete_dentry, |
1da177e4 LT |
319 | }; |
320 | ||
321 | /* | |
322 | * Obtains the first available file descriptor and sets it up for use. | |
323 | * | |
39d8c1b6 DM |
324 | * These functions create file structures and maps them to fd space |
325 | * of the current process. On success it returns file descriptor | |
1da177e4 LT |
326 | * and file struct implicitly stored in sock->file. |
327 | * Note that another thread may close file descriptor before we return | |
328 | * from this function. We use the fact that now we do not refer | |
329 | * to socket after mapping. If one day we will need it, this | |
330 | * function will increment ref. count on file by 1. | |
331 | * | |
332 | * In any case returned fd MAY BE not valid! | |
333 | * This race condition is unavoidable | |
334 | * with shared fd spaces, we cannot solve it inside kernel, | |
335 | * but we take care of internal coherence yet. | |
336 | */ | |
337 | ||
39d8c1b6 | 338 | static int sock_alloc_fd(struct file **filep) |
1da177e4 LT |
339 | { |
340 | int fd; | |
1da177e4 LT |
341 | |
342 | fd = get_unused_fd(); | |
39d8c1b6 | 343 | if (likely(fd >= 0)) { |
1da177e4 LT |
344 | struct file *file = get_empty_filp(); |
345 | ||
39d8c1b6 DM |
346 | *filep = file; |
347 | if (unlikely(!file)) { | |
1da177e4 | 348 | put_unused_fd(fd); |
39d8c1b6 | 349 | return -ENFILE; |
1da177e4 | 350 | } |
39d8c1b6 DM |
351 | } else |
352 | *filep = NULL; | |
353 | return fd; | |
354 | } | |
1da177e4 | 355 | |
39d8c1b6 DM |
356 | static int sock_attach_fd(struct socket *sock, struct file *file) |
357 | { | |
358 | struct qstr this; | |
359 | char name[32]; | |
360 | ||
361 | this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino); | |
362 | this.name = name; | |
363 | this.hash = SOCK_INODE(sock)->i_ino; | |
364 | ||
365 | file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this); | |
366 | if (unlikely(!file->f_dentry)) | |
367 | return -ENOMEM; | |
368 | ||
369 | file->f_dentry->d_op = &sockfs_dentry_operations; | |
370 | d_add(file->f_dentry, SOCK_INODE(sock)); | |
371 | file->f_vfsmnt = mntget(sock_mnt); | |
372 | file->f_mapping = file->f_dentry->d_inode->i_mapping; | |
373 | ||
374 | sock->file = file; | |
375 | file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops; | |
376 | file->f_mode = FMODE_READ | FMODE_WRITE; | |
377 | file->f_flags = O_RDWR; | |
378 | file->f_pos = 0; | |
379 | file->private_data = sock; | |
1da177e4 | 380 | |
39d8c1b6 DM |
381 | return 0; |
382 | } | |
383 | ||
384 | int sock_map_fd(struct socket *sock) | |
385 | { | |
386 | struct file *newfile; | |
387 | int fd = sock_alloc_fd(&newfile); | |
388 | ||
389 | if (likely(fd >= 0)) { | |
390 | int err = sock_attach_fd(sock, newfile); | |
391 | ||
392 | if (unlikely(err < 0)) { | |
393 | put_filp(newfile); | |
1da177e4 | 394 | put_unused_fd(fd); |
39d8c1b6 | 395 | return err; |
1da177e4 | 396 | } |
39d8c1b6 | 397 | fd_install(fd, newfile); |
1da177e4 | 398 | } |
1da177e4 LT |
399 | return fd; |
400 | } | |
401 | ||
6cb153ca BL |
402 | static struct socket *sock_from_file(struct file *file, int *err) |
403 | { | |
404 | struct inode *inode; | |
405 | struct socket *sock; | |
406 | ||
407 | if (file->f_op == &socket_file_ops) | |
408 | return file->private_data; /* set in sock_map_fd */ | |
409 | ||
410 | inode = file->f_dentry->d_inode; | |
411 | if (!S_ISSOCK(inode->i_mode)) { | |
412 | *err = -ENOTSOCK; | |
413 | return NULL; | |
414 | } | |
415 | ||
416 | sock = SOCKET_I(inode); | |
417 | if (sock->file != file) { | |
418 | printk(KERN_ERR "socki_lookup: socket file changed!\n"); | |
419 | sock->file = file; | |
420 | } | |
421 | return sock; | |
422 | } | |
423 | ||
1da177e4 LT |
424 | /** |
425 | * sockfd_lookup - Go from a file number to its socket slot | |
426 | * @fd: file handle | |
427 | * @err: pointer to an error code return | |
428 | * | |
429 | * The file handle passed in is locked and the socket it is bound | |
430 | * too is returned. If an error occurs the err pointer is overwritten | |
431 | * with a negative errno code and NULL is returned. The function checks | |
432 | * for both invalid handles and passing a handle which is not a socket. | |
433 | * | |
434 | * On a success the socket object pointer is returned. | |
435 | */ | |
436 | ||
437 | struct socket *sockfd_lookup(int fd, int *err) | |
438 | { | |
439 | struct file *file; | |
1da177e4 LT |
440 | struct socket *sock; |
441 | ||
89bddce5 SH |
442 | file = fget(fd); |
443 | if (!file) { | |
1da177e4 LT |
444 | *err = -EBADF; |
445 | return NULL; | |
446 | } | |
89bddce5 | 447 | |
6cb153ca BL |
448 | sock = sock_from_file(file, err); |
449 | if (!sock) | |
1da177e4 | 450 | fput(file); |
6cb153ca BL |
451 | return sock; |
452 | } | |
1da177e4 | 453 | |
6cb153ca BL |
454 | static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) |
455 | { | |
456 | struct file *file; | |
457 | struct socket *sock; | |
458 | ||
3672558c | 459 | *err = -EBADF; |
6cb153ca BL |
460 | file = fget_light(fd, fput_needed); |
461 | if (file) { | |
462 | sock = sock_from_file(file, err); | |
463 | if (sock) | |
464 | return sock; | |
465 | fput_light(file, *fput_needed); | |
1da177e4 | 466 | } |
6cb153ca | 467 | return NULL; |
1da177e4 LT |
468 | } |
469 | ||
470 | /** | |
471 | * sock_alloc - allocate a socket | |
89bddce5 | 472 | * |
1da177e4 LT |
473 | * Allocate a new inode and socket object. The two are bound together |
474 | * and initialised. The socket is then returned. If we are out of inodes | |
475 | * NULL is returned. | |
476 | */ | |
477 | ||
478 | static struct socket *sock_alloc(void) | |
479 | { | |
89bddce5 SH |
480 | struct inode *inode; |
481 | struct socket *sock; | |
1da177e4 LT |
482 | |
483 | inode = new_inode(sock_mnt->mnt_sb); | |
484 | if (!inode) | |
485 | return NULL; | |
486 | ||
487 | sock = SOCKET_I(inode); | |
488 | ||
89bddce5 | 489 | inode->i_mode = S_IFSOCK | S_IRWXUGO; |
1da177e4 LT |
490 | inode->i_uid = current->fsuid; |
491 | inode->i_gid = current->fsgid; | |
492 | ||
493 | get_cpu_var(sockets_in_use)++; | |
494 | put_cpu_var(sockets_in_use); | |
495 | return sock; | |
496 | } | |
497 | ||
498 | /* | |
499 | * In theory you can't get an open on this inode, but /proc provides | |
500 | * a back door. Remember to keep it shut otherwise you'll let the | |
501 | * creepy crawlies in. | |
502 | */ | |
89bddce5 | 503 | |
1da177e4 LT |
504 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare) |
505 | { | |
506 | return -ENXIO; | |
507 | } | |
508 | ||
4b6f5d20 | 509 | const struct file_operations bad_sock_fops = { |
1da177e4 LT |
510 | .owner = THIS_MODULE, |
511 | .open = sock_no_open, | |
512 | }; | |
513 | ||
514 | /** | |
515 | * sock_release - close a socket | |
516 | * @sock: socket to close | |
517 | * | |
518 | * The socket is released from the protocol stack if it has a release | |
519 | * callback, and the inode is then released if the socket is bound to | |
89bddce5 | 520 | * an inode not a file. |
1da177e4 | 521 | */ |
89bddce5 | 522 | |
1da177e4 LT |
523 | void sock_release(struct socket *sock) |
524 | { | |
525 | if (sock->ops) { | |
526 | struct module *owner = sock->ops->owner; | |
527 | ||
528 | sock->ops->release(sock); | |
529 | sock->ops = NULL; | |
530 | module_put(owner); | |
531 | } | |
532 | ||
533 | if (sock->fasync_list) | |
534 | printk(KERN_ERR "sock_release: fasync list not empty!\n"); | |
535 | ||
536 | get_cpu_var(sockets_in_use)--; | |
537 | put_cpu_var(sockets_in_use); | |
538 | if (!sock->file) { | |
539 | iput(SOCK_INODE(sock)); | |
540 | return; | |
541 | } | |
89bddce5 | 542 | sock->file = NULL; |
1da177e4 LT |
543 | } |
544 | ||
89bddce5 | 545 | static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
546 | struct msghdr *msg, size_t size) |
547 | { | |
548 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
549 | int err; | |
550 | ||
551 | si->sock = sock; | |
552 | si->scm = NULL; | |
553 | si->msg = msg; | |
554 | si->size = size; | |
555 | ||
556 | err = security_socket_sendmsg(sock, msg, size); | |
557 | if (err) | |
558 | return err; | |
559 | ||
560 | return sock->ops->sendmsg(iocb, sock, msg, size); | |
561 | } | |
562 | ||
563 | int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) | |
564 | { | |
565 | struct kiocb iocb; | |
566 | struct sock_iocb siocb; | |
567 | int ret; | |
568 | ||
569 | init_sync_kiocb(&iocb, NULL); | |
570 | iocb.private = &siocb; | |
571 | ret = __sock_sendmsg(&iocb, sock, msg, size); | |
572 | if (-EIOCBQUEUED == ret) | |
573 | ret = wait_on_sync_kiocb(&iocb); | |
574 | return ret; | |
575 | } | |
576 | ||
577 | int kernel_sendmsg(struct socket *sock, struct msghdr *msg, | |
578 | struct kvec *vec, size_t num, size_t size) | |
579 | { | |
580 | mm_segment_t oldfs = get_fs(); | |
581 | int result; | |
582 | ||
583 | set_fs(KERNEL_DS); | |
584 | /* | |
585 | * the following is safe, since for compiler definitions of kvec and | |
586 | * iovec are identical, yielding the same in-core layout and alignment | |
587 | */ | |
89bddce5 | 588 | msg->msg_iov = (struct iovec *)vec; |
1da177e4 LT |
589 | msg->msg_iovlen = num; |
590 | result = sock_sendmsg(sock, msg, size); | |
591 | set_fs(oldfs); | |
592 | return result; | |
593 | } | |
594 | ||
89bddce5 | 595 | static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
596 | struct msghdr *msg, size_t size, int flags) |
597 | { | |
598 | int err; | |
599 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
600 | ||
601 | si->sock = sock; | |
602 | si->scm = NULL; | |
603 | si->msg = msg; | |
604 | si->size = size; | |
605 | si->flags = flags; | |
606 | ||
607 | err = security_socket_recvmsg(sock, msg, size, flags); | |
608 | if (err) | |
609 | return err; | |
610 | ||
611 | return sock->ops->recvmsg(iocb, sock, msg, size, flags); | |
612 | } | |
613 | ||
89bddce5 | 614 | int sock_recvmsg(struct socket *sock, struct msghdr *msg, |
1da177e4 LT |
615 | size_t size, int flags) |
616 | { | |
617 | struct kiocb iocb; | |
618 | struct sock_iocb siocb; | |
619 | int ret; | |
620 | ||
89bddce5 | 621 | init_sync_kiocb(&iocb, NULL); |
1da177e4 LT |
622 | iocb.private = &siocb; |
623 | ret = __sock_recvmsg(&iocb, sock, msg, size, flags); | |
624 | if (-EIOCBQUEUED == ret) | |
625 | ret = wait_on_sync_kiocb(&iocb); | |
626 | return ret; | |
627 | } | |
628 | ||
89bddce5 SH |
629 | int kernel_recvmsg(struct socket *sock, struct msghdr *msg, |
630 | struct kvec *vec, size_t num, size_t size, int flags) | |
1da177e4 LT |
631 | { |
632 | mm_segment_t oldfs = get_fs(); | |
633 | int result; | |
634 | ||
635 | set_fs(KERNEL_DS); | |
636 | /* | |
637 | * the following is safe, since for compiler definitions of kvec and | |
638 | * iovec are identical, yielding the same in-core layout and alignment | |
639 | */ | |
89bddce5 | 640 | msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num; |
1da177e4 LT |
641 | result = sock_recvmsg(sock, msg, size, flags); |
642 | set_fs(oldfs); | |
643 | return result; | |
644 | } | |
645 | ||
646 | static void sock_aio_dtor(struct kiocb *iocb) | |
647 | { | |
648 | kfree(iocb->private); | |
649 | } | |
650 | ||
ce1d4d3e CH |
651 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
652 | int offset, size_t size, loff_t *ppos, int more) | |
1da177e4 | 653 | { |
1da177e4 LT |
654 | struct socket *sock; |
655 | int flags; | |
656 | ||
ce1d4d3e CH |
657 | sock = file->private_data; |
658 | ||
659 | flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT; | |
660 | if (more) | |
661 | flags |= MSG_MORE; | |
662 | ||
663 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
664 | } | |
1da177e4 | 665 | |
ce1d4d3e | 666 | static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb, |
89bddce5 | 667 | struct sock_iocb *siocb) |
ce1d4d3e CH |
668 | { |
669 | if (!is_sync_kiocb(iocb)) { | |
670 | siocb = kmalloc(sizeof(*siocb), GFP_KERNEL); | |
671 | if (!siocb) | |
672 | return NULL; | |
1da177e4 LT |
673 | iocb->ki_dtor = sock_aio_dtor; |
674 | } | |
1da177e4 | 675 | |
ce1d4d3e | 676 | siocb->kiocb = iocb; |
ce1d4d3e CH |
677 | iocb->private = siocb; |
678 | return siocb; | |
1da177e4 LT |
679 | } |
680 | ||
ce1d4d3e | 681 | static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb, |
027445c3 BP |
682 | struct file *file, const struct iovec *iov, |
683 | unsigned long nr_segs) | |
ce1d4d3e CH |
684 | { |
685 | struct socket *sock = file->private_data; | |
686 | size_t size = 0; | |
687 | int i; | |
1da177e4 | 688 | |
89bddce5 SH |
689 | for (i = 0; i < nr_segs; i++) |
690 | size += iov[i].iov_len; | |
1da177e4 | 691 | |
ce1d4d3e CH |
692 | msg->msg_name = NULL; |
693 | msg->msg_namelen = 0; | |
694 | msg->msg_control = NULL; | |
695 | msg->msg_controllen = 0; | |
89bddce5 | 696 | msg->msg_iov = (struct iovec *)iov; |
ce1d4d3e CH |
697 | msg->msg_iovlen = nr_segs; |
698 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
699 | ||
700 | return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags); | |
701 | } | |
702 | ||
703 | static ssize_t sock_readv(struct file *file, const struct iovec *iov, | |
704 | unsigned long nr_segs, loff_t *ppos) | |
1da177e4 | 705 | { |
ce1d4d3e CH |
706 | struct kiocb iocb; |
707 | struct sock_iocb siocb; | |
708 | struct msghdr msg; | |
709 | int ret; | |
710 | ||
89bddce5 | 711 | init_sync_kiocb(&iocb, NULL); |
ce1d4d3e CH |
712 | iocb.private = &siocb; |
713 | ||
027445c3 | 714 | ret = do_sock_read(&msg, &iocb, file, iov, nr_segs); |
ce1d4d3e CH |
715 | if (-EIOCBQUEUED == ret) |
716 | ret = wait_on_sync_kiocb(&iocb); | |
717 | return ret; | |
718 | } | |
719 | ||
027445c3 BP |
720 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, |
721 | unsigned long nr_segs, loff_t pos) | |
ce1d4d3e CH |
722 | { |
723 | struct sock_iocb siocb, *x; | |
724 | ||
1da177e4 LT |
725 | if (pos != 0) |
726 | return -ESPIPE; | |
027445c3 BP |
727 | |
728 | if (iocb->ki_left == 0) /* Match SYS5 behaviour */ | |
1da177e4 LT |
729 | return 0; |
730 | ||
027445c3 BP |
731 | |
732 | x = alloc_sock_iocb(iocb, &siocb); | |
ce1d4d3e CH |
733 | if (!x) |
734 | return -ENOMEM; | |
027445c3 | 735 | return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
1da177e4 LT |
736 | } |
737 | ||
ce1d4d3e | 738 | static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb, |
027445c3 BP |
739 | struct file *file, const struct iovec *iov, |
740 | unsigned long nr_segs) | |
1da177e4 | 741 | { |
ce1d4d3e CH |
742 | struct socket *sock = file->private_data; |
743 | size_t size = 0; | |
744 | int i; | |
1da177e4 | 745 | |
89bddce5 SH |
746 | for (i = 0; i < nr_segs; i++) |
747 | size += iov[i].iov_len; | |
1da177e4 | 748 | |
ce1d4d3e CH |
749 | msg->msg_name = NULL; |
750 | msg->msg_namelen = 0; | |
751 | msg->msg_control = NULL; | |
752 | msg->msg_controllen = 0; | |
89bddce5 | 753 | msg->msg_iov = (struct iovec *)iov; |
ce1d4d3e CH |
754 | msg->msg_iovlen = nr_segs; |
755 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
756 | if (sock->type == SOCK_SEQPACKET) | |
757 | msg->msg_flags |= MSG_EOR; | |
1da177e4 | 758 | |
ce1d4d3e | 759 | return __sock_sendmsg(iocb, sock, msg, size); |
1da177e4 LT |
760 | } |
761 | ||
ce1d4d3e CH |
762 | static ssize_t sock_writev(struct file *file, const struct iovec *iov, |
763 | unsigned long nr_segs, loff_t *ppos) | |
1da177e4 LT |
764 | { |
765 | struct msghdr msg; | |
ce1d4d3e CH |
766 | struct kiocb iocb; |
767 | struct sock_iocb siocb; | |
768 | int ret; | |
1da177e4 | 769 | |
ce1d4d3e CH |
770 | init_sync_kiocb(&iocb, NULL); |
771 | iocb.private = &siocb; | |
1da177e4 | 772 | |
027445c3 | 773 | ret = do_sock_write(&msg, &iocb, file, iov, nr_segs); |
ce1d4d3e CH |
774 | if (-EIOCBQUEUED == ret) |
775 | ret = wait_on_sync_kiocb(&iocb); | |
776 | return ret; | |
777 | } | |
1da177e4 | 778 | |
027445c3 BP |
779 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, |
780 | unsigned long nr_segs, loff_t pos) | |
ce1d4d3e CH |
781 | { |
782 | struct sock_iocb siocb, *x; | |
1da177e4 | 783 | |
ce1d4d3e CH |
784 | if (pos != 0) |
785 | return -ESPIPE; | |
027445c3 BP |
786 | |
787 | if (iocb->ki_left == 0) /* Match SYS5 behaviour */ | |
ce1d4d3e | 788 | return 0; |
1da177e4 | 789 | |
027445c3 | 790 | x = alloc_sock_iocb(iocb, &siocb); |
ce1d4d3e CH |
791 | if (!x) |
792 | return -ENOMEM; | |
1da177e4 | 793 | |
027445c3 | 794 | return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
1da177e4 LT |
795 | } |
796 | ||
1da177e4 LT |
797 | /* |
798 | * Atomic setting of ioctl hooks to avoid race | |
799 | * with module unload. | |
800 | */ | |
801 | ||
4a3e2f71 | 802 | static DEFINE_MUTEX(br_ioctl_mutex); |
89bddce5 | 803 | static int (*br_ioctl_hook) (unsigned int cmd, void __user *arg) = NULL; |
1da177e4 | 804 | |
89bddce5 | 805 | void brioctl_set(int (*hook) (unsigned int, void __user *)) |
1da177e4 | 806 | { |
4a3e2f71 | 807 | mutex_lock(&br_ioctl_mutex); |
1da177e4 | 808 | br_ioctl_hook = hook; |
4a3e2f71 | 809 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 | 810 | } |
89bddce5 | 811 | |
1da177e4 LT |
812 | EXPORT_SYMBOL(brioctl_set); |
813 | ||
4a3e2f71 | 814 | static DEFINE_MUTEX(vlan_ioctl_mutex); |
89bddce5 | 815 | static int (*vlan_ioctl_hook) (void __user *arg); |
1da177e4 | 816 | |
89bddce5 | 817 | void vlan_ioctl_set(int (*hook) (void __user *)) |
1da177e4 | 818 | { |
4a3e2f71 | 819 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 820 | vlan_ioctl_hook = hook; |
4a3e2f71 | 821 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 822 | } |
89bddce5 | 823 | |
1da177e4 LT |
824 | EXPORT_SYMBOL(vlan_ioctl_set); |
825 | ||
4a3e2f71 | 826 | static DEFINE_MUTEX(dlci_ioctl_mutex); |
89bddce5 | 827 | static int (*dlci_ioctl_hook) (unsigned int, void __user *); |
1da177e4 | 828 | |
89bddce5 | 829 | void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) |
1da177e4 | 830 | { |
4a3e2f71 | 831 | mutex_lock(&dlci_ioctl_mutex); |
1da177e4 | 832 | dlci_ioctl_hook = hook; |
4a3e2f71 | 833 | mutex_unlock(&dlci_ioctl_mutex); |
1da177e4 | 834 | } |
89bddce5 | 835 | |
1da177e4 LT |
836 | EXPORT_SYMBOL(dlci_ioctl_set); |
837 | ||
838 | /* | |
839 | * With an ioctl, arg may well be a user mode pointer, but we don't know | |
840 | * what to do with it - that's up to the protocol still. | |
841 | */ | |
842 | ||
843 | static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) | |
844 | { | |
845 | struct socket *sock; | |
846 | void __user *argp = (void __user *)arg; | |
847 | int pid, err; | |
848 | ||
b69aee04 | 849 | sock = file->private_data; |
1da177e4 LT |
850 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) { |
851 | err = dev_ioctl(cmd, argp); | |
852 | } else | |
d86b5e0e | 853 | #ifdef CONFIG_WIRELESS_EXT |
1da177e4 LT |
854 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { |
855 | err = dev_ioctl(cmd, argp); | |
856 | } else | |
89bddce5 SH |
857 | #endif /* CONFIG_WIRELESS_EXT */ |
858 | switch (cmd) { | |
1da177e4 LT |
859 | case FIOSETOWN: |
860 | case SIOCSPGRP: | |
861 | err = -EFAULT; | |
862 | if (get_user(pid, (int __user *)argp)) | |
863 | break; | |
864 | err = f_setown(sock->file, pid, 1); | |
865 | break; | |
866 | case FIOGETOWN: | |
867 | case SIOCGPGRP: | |
89bddce5 SH |
868 | err = put_user(sock->file->f_owner.pid, |
869 | (int __user *)argp); | |
1da177e4 LT |
870 | break; |
871 | case SIOCGIFBR: | |
872 | case SIOCSIFBR: | |
873 | case SIOCBRADDBR: | |
874 | case SIOCBRDELBR: | |
875 | err = -ENOPKG; | |
876 | if (!br_ioctl_hook) | |
877 | request_module("bridge"); | |
878 | ||
4a3e2f71 | 879 | mutex_lock(&br_ioctl_mutex); |
89bddce5 | 880 | if (br_ioctl_hook) |
1da177e4 | 881 | err = br_ioctl_hook(cmd, argp); |
4a3e2f71 | 882 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 LT |
883 | break; |
884 | case SIOCGIFVLAN: | |
885 | case SIOCSIFVLAN: | |
886 | err = -ENOPKG; | |
887 | if (!vlan_ioctl_hook) | |
888 | request_module("8021q"); | |
889 | ||
4a3e2f71 | 890 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 LT |
891 | if (vlan_ioctl_hook) |
892 | err = vlan_ioctl_hook(argp); | |
4a3e2f71 | 893 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 LT |
894 | break; |
895 | case SIOCGIFDIVERT: | |
896 | case SIOCSIFDIVERT: | |
89bddce5 | 897 | /* Convert this to call through a hook */ |
1da177e4 LT |
898 | err = divert_ioctl(cmd, argp); |
899 | break; | |
900 | case SIOCADDDLCI: | |
901 | case SIOCDELDLCI: | |
902 | err = -ENOPKG; | |
903 | if (!dlci_ioctl_hook) | |
904 | request_module("dlci"); | |
905 | ||
906 | if (dlci_ioctl_hook) { | |
4a3e2f71 | 907 | mutex_lock(&dlci_ioctl_mutex); |
1da177e4 | 908 | err = dlci_ioctl_hook(cmd, argp); |
4a3e2f71 | 909 | mutex_unlock(&dlci_ioctl_mutex); |
1da177e4 LT |
910 | } |
911 | break; | |
912 | default: | |
913 | err = sock->ops->ioctl(sock, cmd, arg); | |
b5e5fa5e CH |
914 | |
915 | /* | |
916 | * If this ioctl is unknown try to hand it down | |
917 | * to the NIC driver. | |
918 | */ | |
919 | if (err == -ENOIOCTLCMD) | |
920 | err = dev_ioctl(cmd, argp); | |
1da177e4 | 921 | break; |
89bddce5 | 922 | } |
1da177e4 LT |
923 | return err; |
924 | } | |
925 | ||
926 | int sock_create_lite(int family, int type, int protocol, struct socket **res) | |
927 | { | |
928 | int err; | |
929 | struct socket *sock = NULL; | |
89bddce5 | 930 | |
1da177e4 LT |
931 | err = security_socket_create(family, type, protocol, 1); |
932 | if (err) | |
933 | goto out; | |
934 | ||
935 | sock = sock_alloc(); | |
936 | if (!sock) { | |
937 | err = -ENOMEM; | |
938 | goto out; | |
939 | } | |
940 | ||
1da177e4 | 941 | sock->type = type; |
7420ed23 VY |
942 | err = security_socket_post_create(sock, family, type, protocol, 1); |
943 | if (err) | |
944 | goto out_release; | |
945 | ||
1da177e4 LT |
946 | out: |
947 | *res = sock; | |
948 | return err; | |
7420ed23 VY |
949 | out_release: |
950 | sock_release(sock); | |
951 | sock = NULL; | |
952 | goto out; | |
1da177e4 LT |
953 | } |
954 | ||
955 | /* No kernel lock held - perfect */ | |
89bddce5 | 956 | static unsigned int sock_poll(struct file *file, poll_table *wait) |
1da177e4 LT |
957 | { |
958 | struct socket *sock; | |
959 | ||
960 | /* | |
89bddce5 | 961 | * We can't return errors to poll, so it's either yes or no. |
1da177e4 | 962 | */ |
b69aee04 | 963 | sock = file->private_data; |
1da177e4 LT |
964 | return sock->ops->poll(file, sock, wait); |
965 | } | |
966 | ||
89bddce5 | 967 | static int sock_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 968 | { |
b69aee04 | 969 | struct socket *sock = file->private_data; |
1da177e4 LT |
970 | |
971 | return sock->ops->mmap(file, sock, vma); | |
972 | } | |
973 | ||
20380731 | 974 | static int sock_close(struct inode *inode, struct file *filp) |
1da177e4 LT |
975 | { |
976 | /* | |
89bddce5 SH |
977 | * It was possible the inode is NULL we were |
978 | * closing an unfinished socket. | |
1da177e4 LT |
979 | */ |
980 | ||
89bddce5 | 981 | if (!inode) { |
1da177e4 LT |
982 | printk(KERN_DEBUG "sock_close: NULL inode\n"); |
983 | return 0; | |
984 | } | |
985 | sock_fasync(-1, filp, 0); | |
986 | sock_release(SOCKET_I(inode)); | |
987 | return 0; | |
988 | } | |
989 | ||
990 | /* | |
991 | * Update the socket async list | |
992 | * | |
993 | * Fasync_list locking strategy. | |
994 | * | |
995 | * 1. fasync_list is modified only under process context socket lock | |
996 | * i.e. under semaphore. | |
997 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) | |
998 | * or under socket lock. | |
999 | * 3. fasync_list can be used from softirq context, so that | |
1000 | * modification under socket lock have to be enhanced with | |
1001 | * write_lock_bh(&sk->sk_callback_lock). | |
1002 | * --ANK (990710) | |
1003 | */ | |
1004 | ||
1005 | static int sock_fasync(int fd, struct file *filp, int on) | |
1006 | { | |
89bddce5 | 1007 | struct fasync_struct *fa, *fna = NULL, **prev; |
1da177e4 LT |
1008 | struct socket *sock; |
1009 | struct sock *sk; | |
1010 | ||
89bddce5 | 1011 | if (on) { |
8b3a7005 | 1012 | fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL); |
89bddce5 | 1013 | if (fna == NULL) |
1da177e4 LT |
1014 | return -ENOMEM; |
1015 | } | |
1016 | ||
b69aee04 | 1017 | sock = filp->private_data; |
1da177e4 | 1018 | |
89bddce5 SH |
1019 | sk = sock->sk; |
1020 | if (sk == NULL) { | |
1da177e4 LT |
1021 | kfree(fna); |
1022 | return -EINVAL; | |
1023 | } | |
1024 | ||
1025 | lock_sock(sk); | |
1026 | ||
89bddce5 | 1027 | prev = &(sock->fasync_list); |
1da177e4 | 1028 | |
89bddce5 SH |
1029 | for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev) |
1030 | if (fa->fa_file == filp) | |
1da177e4 LT |
1031 | break; |
1032 | ||
89bddce5 SH |
1033 | if (on) { |
1034 | if (fa != NULL) { | |
1da177e4 | 1035 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1036 | fa->fa_fd = fd; |
1da177e4 LT |
1037 | write_unlock_bh(&sk->sk_callback_lock); |
1038 | ||
1039 | kfree(fna); | |
1040 | goto out; | |
1041 | } | |
89bddce5 SH |
1042 | fna->fa_file = filp; |
1043 | fna->fa_fd = fd; | |
1044 | fna->magic = FASYNC_MAGIC; | |
1045 | fna->fa_next = sock->fasync_list; | |
1da177e4 | 1046 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1047 | sock->fasync_list = fna; |
1da177e4 | 1048 | write_unlock_bh(&sk->sk_callback_lock); |
89bddce5 SH |
1049 | } else { |
1050 | if (fa != NULL) { | |
1da177e4 | 1051 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1052 | *prev = fa->fa_next; |
1da177e4 LT |
1053 | write_unlock_bh(&sk->sk_callback_lock); |
1054 | kfree(fa); | |
1055 | } | |
1056 | } | |
1057 | ||
1058 | out: | |
1059 | release_sock(sock->sk); | |
1060 | return 0; | |
1061 | } | |
1062 | ||
1063 | /* This function may be called only under socket lock or callback_lock */ | |
1064 | ||
1065 | int sock_wake_async(struct socket *sock, int how, int band) | |
1066 | { | |
1067 | if (!sock || !sock->fasync_list) | |
1068 | return -1; | |
89bddce5 | 1069 | switch (how) { |
1da177e4 | 1070 | case 1: |
89bddce5 | 1071 | |
1da177e4 LT |
1072 | if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags)) |
1073 | break; | |
1074 | goto call_kill; | |
1075 | case 2: | |
1076 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags)) | |
1077 | break; | |
1078 | /* fall through */ | |
1079 | case 0: | |
89bddce5 | 1080 | call_kill: |
1da177e4 LT |
1081 | __kill_fasync(sock->fasync_list, SIGIO, band); |
1082 | break; | |
1083 | case 3: | |
1084 | __kill_fasync(sock->fasync_list, SIGURG, band); | |
1085 | } | |
1086 | return 0; | |
1087 | } | |
1088 | ||
89bddce5 SH |
1089 | static int __sock_create(int family, int type, int protocol, |
1090 | struct socket **res, int kern) | |
1da177e4 LT |
1091 | { |
1092 | int err; | |
1093 | struct socket *sock; | |
55737fda | 1094 | const struct net_proto_family *pf; |
1da177e4 LT |
1095 | |
1096 | /* | |
89bddce5 | 1097 | * Check protocol is in range |
1da177e4 LT |
1098 | */ |
1099 | if (family < 0 || family >= NPROTO) | |
1100 | return -EAFNOSUPPORT; | |
1101 | if (type < 0 || type >= SOCK_MAX) | |
1102 | return -EINVAL; | |
1103 | ||
1104 | /* Compatibility. | |
1105 | ||
1106 | This uglymoron is moved from INET layer to here to avoid | |
1107 | deadlock in module load. | |
1108 | */ | |
1109 | if (family == PF_INET && type == SOCK_PACKET) { | |
89bddce5 | 1110 | static int warned; |
1da177e4 LT |
1111 | if (!warned) { |
1112 | warned = 1; | |
89bddce5 SH |
1113 | printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", |
1114 | current->comm); | |
1da177e4 LT |
1115 | } |
1116 | family = PF_PACKET; | |
1117 | } | |
1118 | ||
1119 | err = security_socket_create(family, type, protocol, kern); | |
1120 | if (err) | |
1121 | return err; | |
89bddce5 | 1122 | |
55737fda SH |
1123 | /* |
1124 | * Allocate the socket and allow the family to set things up. if | |
1125 | * the protocol is 0, the family is instructed to select an appropriate | |
1126 | * default. | |
1127 | */ | |
1128 | sock = sock_alloc(); | |
1129 | if (!sock) { | |
1130 | if (net_ratelimit()) | |
1131 | printk(KERN_WARNING "socket: no more sockets\n"); | |
1132 | return -ENFILE; /* Not exactly a match, but its the | |
1133 | closest posix thing */ | |
1134 | } | |
1135 | ||
1136 | sock->type = type; | |
1137 | ||
1da177e4 | 1138 | #if defined(CONFIG_KMOD) |
89bddce5 SH |
1139 | /* Attempt to load a protocol module if the find failed. |
1140 | * | |
1141 | * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user | |
1da177e4 LT |
1142 | * requested real, full-featured networking support upon configuration. |
1143 | * Otherwise module support will break! | |
1144 | */ | |
55737fda | 1145 | if (net_families[family] == NULL) |
89bddce5 | 1146 | request_module("net-pf-%d", family); |
1da177e4 LT |
1147 | #endif |
1148 | ||
55737fda SH |
1149 | rcu_read_lock(); |
1150 | pf = rcu_dereference(net_families[family]); | |
1151 | err = -EAFNOSUPPORT; | |
1152 | if (!pf) | |
1153 | goto out_release; | |
1da177e4 LT |
1154 | |
1155 | /* | |
1156 | * We will call the ->create function, that possibly is in a loadable | |
1157 | * module, so we have to bump that loadable module refcnt first. | |
1158 | */ | |
55737fda | 1159 | if (!try_module_get(pf->owner)) |
1da177e4 LT |
1160 | goto out_release; |
1161 | ||
55737fda SH |
1162 | /* Now protected by module ref count */ |
1163 | rcu_read_unlock(); | |
1164 | ||
1165 | err = pf->create(sock, protocol); | |
1166 | if (err < 0) | |
1da177e4 | 1167 | goto out_module_put; |
a79af59e | 1168 | |
1da177e4 LT |
1169 | /* |
1170 | * Now to bump the refcnt of the [loadable] module that owns this | |
1171 | * socket at sock_release time we decrement its refcnt. | |
1172 | */ | |
55737fda SH |
1173 | if (!try_module_get(sock->ops->owner)) |
1174 | goto out_module_busy; | |
1175 | ||
1da177e4 LT |
1176 | /* |
1177 | * Now that we're done with the ->create function, the [loadable] | |
1178 | * module can have its refcnt decremented | |
1179 | */ | |
55737fda | 1180 | module_put(pf->owner); |
7420ed23 VY |
1181 | err = security_socket_post_create(sock, family, type, protocol, kern); |
1182 | if (err) | |
1183 | goto out_release; | |
55737fda | 1184 | *res = sock; |
1da177e4 | 1185 | |
55737fda SH |
1186 | return 0; |
1187 | ||
1188 | out_module_busy: | |
1189 | err = -EAFNOSUPPORT; | |
1da177e4 | 1190 | out_module_put: |
55737fda SH |
1191 | sock->ops = NULL; |
1192 | module_put(pf->owner); | |
1193 | out_sock_release: | |
1da177e4 | 1194 | sock_release(sock); |
55737fda SH |
1195 | return err; |
1196 | ||
1197 | out_release: | |
1198 | rcu_read_unlock(); | |
1199 | goto out_sock_release; | |
1da177e4 LT |
1200 | } |
1201 | ||
1202 | int sock_create(int family, int type, int protocol, struct socket **res) | |
1203 | { | |
1204 | return __sock_create(family, type, protocol, res, 0); | |
1205 | } | |
1206 | ||
1207 | int sock_create_kern(int family, int type, int protocol, struct socket **res) | |
1208 | { | |
1209 | return __sock_create(family, type, protocol, res, 1); | |
1210 | } | |
1211 | ||
1212 | asmlinkage long sys_socket(int family, int type, int protocol) | |
1213 | { | |
1214 | int retval; | |
1215 | struct socket *sock; | |
1216 | ||
1217 | retval = sock_create(family, type, protocol, &sock); | |
1218 | if (retval < 0) | |
1219 | goto out; | |
1220 | ||
1221 | retval = sock_map_fd(sock); | |
1222 | if (retval < 0) | |
1223 | goto out_release; | |
1224 | ||
1225 | out: | |
1226 | /* It may be already another descriptor 8) Not kernel problem. */ | |
1227 | return retval; | |
1228 | ||
1229 | out_release: | |
1230 | sock_release(sock); | |
1231 | return retval; | |
1232 | } | |
1233 | ||
1234 | /* | |
1235 | * Create a pair of connected sockets. | |
1236 | */ | |
1237 | ||
89bddce5 SH |
1238 | asmlinkage long sys_socketpair(int family, int type, int protocol, |
1239 | int __user *usockvec) | |
1da177e4 LT |
1240 | { |
1241 | struct socket *sock1, *sock2; | |
1242 | int fd1, fd2, err; | |
1243 | ||
1244 | /* | |
1245 | * Obtain the first socket and check if the underlying protocol | |
1246 | * supports the socketpair call. | |
1247 | */ | |
1248 | ||
1249 | err = sock_create(family, type, protocol, &sock1); | |
1250 | if (err < 0) | |
1251 | goto out; | |
1252 | ||
1253 | err = sock_create(family, type, protocol, &sock2); | |
1254 | if (err < 0) | |
1255 | goto out_release_1; | |
1256 | ||
1257 | err = sock1->ops->socketpair(sock1, sock2); | |
89bddce5 | 1258 | if (err < 0) |
1da177e4 LT |
1259 | goto out_release_both; |
1260 | ||
1261 | fd1 = fd2 = -1; | |
1262 | ||
1263 | err = sock_map_fd(sock1); | |
1264 | if (err < 0) | |
1265 | goto out_release_both; | |
1266 | fd1 = err; | |
1267 | ||
1268 | err = sock_map_fd(sock2); | |
1269 | if (err < 0) | |
1270 | goto out_close_1; | |
1271 | fd2 = err; | |
1272 | ||
1273 | /* fd1 and fd2 may be already another descriptors. | |
1274 | * Not kernel problem. | |
1275 | */ | |
1276 | ||
89bddce5 | 1277 | err = put_user(fd1, &usockvec[0]); |
1da177e4 LT |
1278 | if (!err) |
1279 | err = put_user(fd2, &usockvec[1]); | |
1280 | if (!err) | |
1281 | return 0; | |
1282 | ||
1283 | sys_close(fd2); | |
1284 | sys_close(fd1); | |
1285 | return err; | |
1286 | ||
1287 | out_close_1: | |
89bddce5 | 1288 | sock_release(sock2); |
1da177e4 LT |
1289 | sys_close(fd1); |
1290 | return err; | |
1291 | ||
1292 | out_release_both: | |
89bddce5 | 1293 | sock_release(sock2); |
1da177e4 | 1294 | out_release_1: |
89bddce5 | 1295 | sock_release(sock1); |
1da177e4 LT |
1296 | out: |
1297 | return err; | |
1298 | } | |
1299 | ||
1da177e4 LT |
1300 | /* |
1301 | * Bind a name to a socket. Nothing much to do here since it's | |
1302 | * the protocol's responsibility to handle the local address. | |
1303 | * | |
1304 | * We move the socket address to kernel space before we call | |
1305 | * the protocol layer (having also checked the address is ok). | |
1306 | */ | |
1307 | ||
1308 | asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) | |
1309 | { | |
1310 | struct socket *sock; | |
1311 | char address[MAX_SOCK_ADDR]; | |
6cb153ca | 1312 | int err, fput_needed; |
1da177e4 | 1313 | |
89bddce5 SH |
1314 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1315 | if(sock) { | |
1316 | err = move_addr_to_kernel(umyaddr, addrlen, address); | |
1317 | if (err >= 0) { | |
1318 | err = security_socket_bind(sock, | |
1319 | (struct sockaddr *)address, | |
1320 | addrlen); | |
6cb153ca BL |
1321 | if (!err) |
1322 | err = sock->ops->bind(sock, | |
89bddce5 SH |
1323 | (struct sockaddr *) |
1324 | address, addrlen); | |
1da177e4 | 1325 | } |
6cb153ca | 1326 | fput_light(sock->file, fput_needed); |
89bddce5 | 1327 | } |
1da177e4 LT |
1328 | return err; |
1329 | } | |
1330 | ||
1da177e4 LT |
1331 | /* |
1332 | * Perform a listen. Basically, we allow the protocol to do anything | |
1333 | * necessary for a listen, and if that works, we mark the socket as | |
1334 | * ready for listening. | |
1335 | */ | |
1336 | ||
7a42c217 | 1337 | int sysctl_somaxconn __read_mostly = SOMAXCONN; |
1da177e4 LT |
1338 | |
1339 | asmlinkage long sys_listen(int fd, int backlog) | |
1340 | { | |
1341 | struct socket *sock; | |
6cb153ca | 1342 | int err, fput_needed; |
89bddce5 SH |
1343 | |
1344 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1345 | if (sock) { | |
1346 | if ((unsigned)backlog > sysctl_somaxconn) | |
1da177e4 LT |
1347 | backlog = sysctl_somaxconn; |
1348 | ||
1349 | err = security_socket_listen(sock, backlog); | |
6cb153ca BL |
1350 | if (!err) |
1351 | err = sock->ops->listen(sock, backlog); | |
1da177e4 | 1352 | |
6cb153ca | 1353 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1354 | } |
1355 | return err; | |
1356 | } | |
1357 | ||
1da177e4 LT |
1358 | /* |
1359 | * For accept, we attempt to create a new socket, set up the link | |
1360 | * with the client, wake up the client, then return the new | |
1361 | * connected fd. We collect the address of the connector in kernel | |
1362 | * space and move it to user at the very end. This is unclean because | |
1363 | * we open the socket then return an error. | |
1364 | * | |
1365 | * 1003.1g adds the ability to recvmsg() to query connection pending | |
1366 | * status to recvmsg. We need to add that support in a way thats | |
1367 | * clean when we restucture accept also. | |
1368 | */ | |
1369 | ||
89bddce5 SH |
1370 | asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, |
1371 | int __user *upeer_addrlen) | |
1da177e4 LT |
1372 | { |
1373 | struct socket *sock, *newsock; | |
39d8c1b6 | 1374 | struct file *newfile; |
6cb153ca | 1375 | int err, len, newfd, fput_needed; |
1da177e4 LT |
1376 | char address[MAX_SOCK_ADDR]; |
1377 | ||
6cb153ca | 1378 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1379 | if (!sock) |
1380 | goto out; | |
1381 | ||
1382 | err = -ENFILE; | |
89bddce5 | 1383 | if (!(newsock = sock_alloc())) |
1da177e4 LT |
1384 | goto out_put; |
1385 | ||
1386 | newsock->type = sock->type; | |
1387 | newsock->ops = sock->ops; | |
1388 | ||
1da177e4 LT |
1389 | /* |
1390 | * We don't need try_module_get here, as the listening socket (sock) | |
1391 | * has the protocol module (sock->ops->owner) held. | |
1392 | */ | |
1393 | __module_get(newsock->ops->owner); | |
1394 | ||
39d8c1b6 DM |
1395 | newfd = sock_alloc_fd(&newfile); |
1396 | if (unlikely(newfd < 0)) { | |
1397 | err = newfd; | |
9a1875e6 DM |
1398 | sock_release(newsock); |
1399 | goto out_put; | |
39d8c1b6 DM |
1400 | } |
1401 | ||
1402 | err = sock_attach_fd(newsock, newfile); | |
1403 | if (err < 0) | |
1404 | goto out_fd; | |
1405 | ||
a79af59e FF |
1406 | err = security_socket_accept(sock, newsock); |
1407 | if (err) | |
39d8c1b6 | 1408 | goto out_fd; |
a79af59e | 1409 | |
1da177e4 LT |
1410 | err = sock->ops->accept(sock, newsock, sock->file->f_flags); |
1411 | if (err < 0) | |
39d8c1b6 | 1412 | goto out_fd; |
1da177e4 LT |
1413 | |
1414 | if (upeer_sockaddr) { | |
89bddce5 SH |
1415 | if (newsock->ops->getname(newsock, (struct sockaddr *)address, |
1416 | &len, 2) < 0) { | |
1da177e4 | 1417 | err = -ECONNABORTED; |
39d8c1b6 | 1418 | goto out_fd; |
1da177e4 | 1419 | } |
89bddce5 SH |
1420 | err = move_addr_to_user(address, len, upeer_sockaddr, |
1421 | upeer_addrlen); | |
1da177e4 | 1422 | if (err < 0) |
39d8c1b6 | 1423 | goto out_fd; |
1da177e4 LT |
1424 | } |
1425 | ||
1426 | /* File flags are not inherited via accept() unlike another OSes. */ | |
1427 | ||
39d8c1b6 DM |
1428 | fd_install(newfd, newfile); |
1429 | err = newfd; | |
1da177e4 LT |
1430 | |
1431 | security_socket_post_accept(sock, newsock); | |
1432 | ||
1433 | out_put: | |
6cb153ca | 1434 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1435 | out: |
1436 | return err; | |
39d8c1b6 | 1437 | out_fd: |
9606a216 | 1438 | fput(newfile); |
39d8c1b6 | 1439 | put_unused_fd(newfd); |
1da177e4 LT |
1440 | goto out_put; |
1441 | } | |
1442 | ||
1da177e4 LT |
1443 | /* |
1444 | * Attempt to connect to a socket with the server address. The address | |
1445 | * is in user space so we verify it is OK and move it to kernel space. | |
1446 | * | |
1447 | * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to | |
1448 | * break bindings | |
1449 | * | |
1450 | * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and | |
1451 | * other SEQPACKET protocols that take time to connect() as it doesn't | |
1452 | * include the -EINPROGRESS status for such sockets. | |
1453 | */ | |
1454 | ||
89bddce5 SH |
1455 | asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, |
1456 | int addrlen) | |
1da177e4 LT |
1457 | { |
1458 | struct socket *sock; | |
1459 | char address[MAX_SOCK_ADDR]; | |
6cb153ca | 1460 | int err, fput_needed; |
1da177e4 | 1461 | |
6cb153ca | 1462 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1463 | if (!sock) |
1464 | goto out; | |
1465 | err = move_addr_to_kernel(uservaddr, addrlen, address); | |
1466 | if (err < 0) | |
1467 | goto out_put; | |
1468 | ||
89bddce5 SH |
1469 | err = |
1470 | security_socket_connect(sock, (struct sockaddr *)address, addrlen); | |
1da177e4 LT |
1471 | if (err) |
1472 | goto out_put; | |
1473 | ||
89bddce5 | 1474 | err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen, |
1da177e4 LT |
1475 | sock->file->f_flags); |
1476 | out_put: | |
6cb153ca | 1477 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1478 | out: |
1479 | return err; | |
1480 | } | |
1481 | ||
1482 | /* | |
1483 | * Get the local address ('name') of a socket object. Move the obtained | |
1484 | * name to user space. | |
1485 | */ | |
1486 | ||
89bddce5 SH |
1487 | asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, |
1488 | int __user *usockaddr_len) | |
1da177e4 LT |
1489 | { |
1490 | struct socket *sock; | |
1491 | char address[MAX_SOCK_ADDR]; | |
6cb153ca | 1492 | int len, err, fput_needed; |
89bddce5 | 1493 | |
6cb153ca | 1494 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1495 | if (!sock) |
1496 | goto out; | |
1497 | ||
1498 | err = security_socket_getsockname(sock); | |
1499 | if (err) | |
1500 | goto out_put; | |
1501 | ||
1502 | err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0); | |
1503 | if (err) | |
1504 | goto out_put; | |
1505 | err = move_addr_to_user(address, len, usockaddr, usockaddr_len); | |
1506 | ||
1507 | out_put: | |
6cb153ca | 1508 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1509 | out: |
1510 | return err; | |
1511 | } | |
1512 | ||
1513 | /* | |
1514 | * Get the remote address ('name') of a socket object. Move the obtained | |
1515 | * name to user space. | |
1516 | */ | |
1517 | ||
89bddce5 SH |
1518 | asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, |
1519 | int __user *usockaddr_len) | |
1da177e4 LT |
1520 | { |
1521 | struct socket *sock; | |
1522 | char address[MAX_SOCK_ADDR]; | |
6cb153ca | 1523 | int len, err, fput_needed; |
1da177e4 | 1524 | |
89bddce5 SH |
1525 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1526 | if (sock != NULL) { | |
1da177e4 LT |
1527 | err = security_socket_getpeername(sock); |
1528 | if (err) { | |
6cb153ca | 1529 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1530 | return err; |
1531 | } | |
1532 | ||
89bddce5 SH |
1533 | err = |
1534 | sock->ops->getname(sock, (struct sockaddr *)address, &len, | |
1535 | 1); | |
1da177e4 | 1536 | if (!err) |
89bddce5 SH |
1537 | err = move_addr_to_user(address, len, usockaddr, |
1538 | usockaddr_len); | |
6cb153ca | 1539 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1540 | } |
1541 | return err; | |
1542 | } | |
1543 | ||
1544 | /* | |
1545 | * Send a datagram to a given address. We move the address into kernel | |
1546 | * space and check the user space data area is readable before invoking | |
1547 | * the protocol. | |
1548 | */ | |
1549 | ||
89bddce5 SH |
1550 | asmlinkage long sys_sendto(int fd, void __user *buff, size_t len, |
1551 | unsigned flags, struct sockaddr __user *addr, | |
1552 | int addr_len) | |
1da177e4 LT |
1553 | { |
1554 | struct socket *sock; | |
1555 | char address[MAX_SOCK_ADDR]; | |
1556 | int err; | |
1557 | struct msghdr msg; | |
1558 | struct iovec iov; | |
6cb153ca BL |
1559 | int fput_needed; |
1560 | struct file *sock_file; | |
1561 | ||
1562 | sock_file = fget_light(fd, &fput_needed); | |
1563 | if (!sock_file) | |
1564 | return -EBADF; | |
1565 | ||
1566 | sock = sock_from_file(sock_file, &err); | |
1da177e4 | 1567 | if (!sock) |
6cb153ca | 1568 | goto out_put; |
89bddce5 SH |
1569 | iov.iov_base = buff; |
1570 | iov.iov_len = len; | |
1571 | msg.msg_name = NULL; | |
1572 | msg.msg_iov = &iov; | |
1573 | msg.msg_iovlen = 1; | |
1574 | msg.msg_control = NULL; | |
1575 | msg.msg_controllen = 0; | |
1576 | msg.msg_namelen = 0; | |
6cb153ca | 1577 | if (addr) { |
1da177e4 LT |
1578 | err = move_addr_to_kernel(addr, addr_len, address); |
1579 | if (err < 0) | |
1580 | goto out_put; | |
89bddce5 SH |
1581 | msg.msg_name = address; |
1582 | msg.msg_namelen = addr_len; | |
1da177e4 LT |
1583 | } |
1584 | if (sock->file->f_flags & O_NONBLOCK) | |
1585 | flags |= MSG_DONTWAIT; | |
1586 | msg.msg_flags = flags; | |
1587 | err = sock_sendmsg(sock, &msg, len); | |
1588 | ||
89bddce5 | 1589 | out_put: |
6cb153ca | 1590 | fput_light(sock_file, fput_needed); |
1da177e4 LT |
1591 | return err; |
1592 | } | |
1593 | ||
1594 | /* | |
89bddce5 | 1595 | * Send a datagram down a socket. |
1da177e4 LT |
1596 | */ |
1597 | ||
89bddce5 | 1598 | asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags) |
1da177e4 LT |
1599 | { |
1600 | return sys_sendto(fd, buff, len, flags, NULL, 0); | |
1601 | } | |
1602 | ||
1603 | /* | |
89bddce5 | 1604 | * Receive a frame from the socket and optionally record the address of the |
1da177e4 LT |
1605 | * sender. We verify the buffers are writable and if needed move the |
1606 | * sender address from kernel to user space. | |
1607 | */ | |
1608 | ||
89bddce5 SH |
1609 | asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size, |
1610 | unsigned flags, struct sockaddr __user *addr, | |
1611 | int __user *addr_len) | |
1da177e4 LT |
1612 | { |
1613 | struct socket *sock; | |
1614 | struct iovec iov; | |
1615 | struct msghdr msg; | |
1616 | char address[MAX_SOCK_ADDR]; | |
89bddce5 | 1617 | int err, err2; |
6cb153ca BL |
1618 | struct file *sock_file; |
1619 | int fput_needed; | |
1620 | ||
1621 | sock_file = fget_light(fd, &fput_needed); | |
1622 | if (!sock_file) | |
1623 | return -EBADF; | |
1da177e4 | 1624 | |
6cb153ca | 1625 | sock = sock_from_file(sock_file, &err); |
1da177e4 LT |
1626 | if (!sock) |
1627 | goto out; | |
1628 | ||
89bddce5 SH |
1629 | msg.msg_control = NULL; |
1630 | msg.msg_controllen = 0; | |
1631 | msg.msg_iovlen = 1; | |
1632 | msg.msg_iov = &iov; | |
1633 | iov.iov_len = size; | |
1634 | iov.iov_base = ubuf; | |
1635 | msg.msg_name = address; | |
1636 | msg.msg_namelen = MAX_SOCK_ADDR; | |
1da177e4 LT |
1637 | if (sock->file->f_flags & O_NONBLOCK) |
1638 | flags |= MSG_DONTWAIT; | |
89bddce5 | 1639 | err = sock_recvmsg(sock, &msg, size, flags); |
1da177e4 | 1640 | |
89bddce5 SH |
1641 | if (err >= 0 && addr != NULL) { |
1642 | err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len); | |
1643 | if (err2 < 0) | |
1644 | err = err2; | |
1da177e4 | 1645 | } |
1da177e4 | 1646 | out: |
6cb153ca | 1647 | fput_light(sock_file, fput_needed); |
1da177e4 LT |
1648 | return err; |
1649 | } | |
1650 | ||
1651 | /* | |
89bddce5 | 1652 | * Receive a datagram from a socket. |
1da177e4 LT |
1653 | */ |
1654 | ||
89bddce5 SH |
1655 | asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size, |
1656 | unsigned flags) | |
1da177e4 LT |
1657 | { |
1658 | return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); | |
1659 | } | |
1660 | ||
1661 | /* | |
1662 | * Set a socket option. Because we don't know the option lengths we have | |
1663 | * to pass the user mode parameter for the protocols to sort out. | |
1664 | */ | |
1665 | ||
89bddce5 SH |
1666 | asmlinkage long sys_setsockopt(int fd, int level, int optname, |
1667 | char __user *optval, int optlen) | |
1da177e4 | 1668 | { |
6cb153ca | 1669 | int err, fput_needed; |
1da177e4 LT |
1670 | struct socket *sock; |
1671 | ||
1672 | if (optlen < 0) | |
1673 | return -EINVAL; | |
89bddce5 SH |
1674 | |
1675 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1676 | if (sock != NULL) { | |
1677 | err = security_socket_setsockopt(sock, level, optname); | |
6cb153ca BL |
1678 | if (err) |
1679 | goto out_put; | |
1da177e4 LT |
1680 | |
1681 | if (level == SOL_SOCKET) | |
89bddce5 SH |
1682 | err = |
1683 | sock_setsockopt(sock, level, optname, optval, | |
1684 | optlen); | |
1da177e4 | 1685 | else |
89bddce5 SH |
1686 | err = |
1687 | sock->ops->setsockopt(sock, level, optname, optval, | |
1688 | optlen); | |
6cb153ca BL |
1689 | out_put: |
1690 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1691 | } |
1692 | return err; | |
1693 | } | |
1694 | ||
1695 | /* | |
1696 | * Get a socket option. Because we don't know the option lengths we have | |
1697 | * to pass a user mode parameter for the protocols to sort out. | |
1698 | */ | |
1699 | ||
89bddce5 SH |
1700 | asmlinkage long sys_getsockopt(int fd, int level, int optname, |
1701 | char __user *optval, int __user *optlen) | |
1da177e4 | 1702 | { |
6cb153ca | 1703 | int err, fput_needed; |
1da177e4 LT |
1704 | struct socket *sock; |
1705 | ||
89bddce5 SH |
1706 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1707 | if (sock != NULL) { | |
6cb153ca BL |
1708 | err = security_socket_getsockopt(sock, level, optname); |
1709 | if (err) | |
1710 | goto out_put; | |
1da177e4 LT |
1711 | |
1712 | if (level == SOL_SOCKET) | |
89bddce5 SH |
1713 | err = |
1714 | sock_getsockopt(sock, level, optname, optval, | |
1715 | optlen); | |
1da177e4 | 1716 | else |
89bddce5 SH |
1717 | err = |
1718 | sock->ops->getsockopt(sock, level, optname, optval, | |
1719 | optlen); | |
6cb153ca BL |
1720 | out_put: |
1721 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1722 | } |
1723 | return err; | |
1724 | } | |
1725 | ||
1da177e4 LT |
1726 | /* |
1727 | * Shutdown a socket. | |
1728 | */ | |
1729 | ||
1730 | asmlinkage long sys_shutdown(int fd, int how) | |
1731 | { | |
6cb153ca | 1732 | int err, fput_needed; |
1da177e4 LT |
1733 | struct socket *sock; |
1734 | ||
89bddce5 SH |
1735 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1736 | if (sock != NULL) { | |
1da177e4 | 1737 | err = security_socket_shutdown(sock, how); |
6cb153ca BL |
1738 | if (!err) |
1739 | err = sock->ops->shutdown(sock, how); | |
1740 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1741 | } |
1742 | return err; | |
1743 | } | |
1744 | ||
89bddce5 | 1745 | /* A couple of helpful macros for getting the address of the 32/64 bit |
1da177e4 LT |
1746 | * fields which are the same type (int / unsigned) on our platforms. |
1747 | */ | |
1748 | #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) | |
1749 | #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) | |
1750 | #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) | |
1751 | ||
1da177e4 LT |
1752 | /* |
1753 | * BSD sendmsg interface | |
1754 | */ | |
1755 | ||
1756 | asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags) | |
1757 | { | |
89bddce5 SH |
1758 | struct compat_msghdr __user *msg_compat = |
1759 | (struct compat_msghdr __user *)msg; | |
1da177e4 LT |
1760 | struct socket *sock; |
1761 | char address[MAX_SOCK_ADDR]; | |
1762 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; | |
b9d717a7 | 1763 | unsigned char ctl[sizeof(struct cmsghdr) + 20] |
89bddce5 SH |
1764 | __attribute__ ((aligned(sizeof(__kernel_size_t)))); |
1765 | /* 20 is size of ipv6_pktinfo */ | |
1da177e4 LT |
1766 | unsigned char *ctl_buf = ctl; |
1767 | struct msghdr msg_sys; | |
1768 | int err, ctl_len, iov_size, total_len; | |
6cb153ca | 1769 | int fput_needed; |
89bddce5 | 1770 | |
1da177e4 LT |
1771 | err = -EFAULT; |
1772 | if (MSG_CMSG_COMPAT & flags) { | |
1773 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1774 | return -EFAULT; | |
89bddce5 SH |
1775 | } |
1776 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1da177e4 LT |
1777 | return -EFAULT; |
1778 | ||
6cb153ca | 1779 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
89bddce5 | 1780 | if (!sock) |
1da177e4 LT |
1781 | goto out; |
1782 | ||
1783 | /* do not move before msg_sys is valid */ | |
1784 | err = -EMSGSIZE; | |
1785 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
1786 | goto out_put; | |
1787 | ||
89bddce5 | 1788 | /* Check whether to allocate the iovec area */ |
1da177e4 LT |
1789 | err = -ENOMEM; |
1790 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
1791 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
1792 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
1793 | if (!iov) | |
1794 | goto out_put; | |
1795 | } | |
1796 | ||
1797 | /* This will also move the address data into kernel space */ | |
1798 | if (MSG_CMSG_COMPAT & flags) { | |
1799 | err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ); | |
1800 | } else | |
1801 | err = verify_iovec(&msg_sys, iov, address, VERIFY_READ); | |
89bddce5 | 1802 | if (err < 0) |
1da177e4 LT |
1803 | goto out_freeiov; |
1804 | total_len = err; | |
1805 | ||
1806 | err = -ENOBUFS; | |
1807 | ||
1808 | if (msg_sys.msg_controllen > INT_MAX) | |
1809 | goto out_freeiov; | |
89bddce5 | 1810 | ctl_len = msg_sys.msg_controllen; |
1da177e4 | 1811 | if ((MSG_CMSG_COMPAT & flags) && ctl_len) { |
89bddce5 SH |
1812 | err = |
1813 | cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, | |
1814 | sizeof(ctl)); | |
1da177e4 LT |
1815 | if (err) |
1816 | goto out_freeiov; | |
1817 | ctl_buf = msg_sys.msg_control; | |
8920e8f9 | 1818 | ctl_len = msg_sys.msg_controllen; |
1da177e4 | 1819 | } else if (ctl_len) { |
89bddce5 | 1820 | if (ctl_len > sizeof(ctl)) { |
1da177e4 | 1821 | ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); |
89bddce5 | 1822 | if (ctl_buf == NULL) |
1da177e4 LT |
1823 | goto out_freeiov; |
1824 | } | |
1825 | err = -EFAULT; | |
1826 | /* | |
1827 | * Careful! Before this, msg_sys.msg_control contains a user pointer. | |
1828 | * Afterwards, it will be a kernel pointer. Thus the compiler-assisted | |
1829 | * checking falls down on this. | |
1830 | */ | |
89bddce5 SH |
1831 | if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control, |
1832 | ctl_len)) | |
1da177e4 LT |
1833 | goto out_freectl; |
1834 | msg_sys.msg_control = ctl_buf; | |
1835 | } | |
1836 | msg_sys.msg_flags = flags; | |
1837 | ||
1838 | if (sock->file->f_flags & O_NONBLOCK) | |
1839 | msg_sys.msg_flags |= MSG_DONTWAIT; | |
1840 | err = sock_sendmsg(sock, &msg_sys, total_len); | |
1841 | ||
1842 | out_freectl: | |
89bddce5 | 1843 | if (ctl_buf != ctl) |
1da177e4 LT |
1844 | sock_kfree_s(sock->sk, ctl_buf, ctl_len); |
1845 | out_freeiov: | |
1846 | if (iov != iovstack) | |
1847 | sock_kfree_s(sock->sk, iov, iov_size); | |
1848 | out_put: | |
6cb153ca | 1849 | fput_light(sock->file, fput_needed); |
89bddce5 | 1850 | out: |
1da177e4 LT |
1851 | return err; |
1852 | } | |
1853 | ||
1854 | /* | |
1855 | * BSD recvmsg interface | |
1856 | */ | |
1857 | ||
89bddce5 SH |
1858 | asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, |
1859 | unsigned int flags) | |
1da177e4 | 1860 | { |
89bddce5 SH |
1861 | struct compat_msghdr __user *msg_compat = |
1862 | (struct compat_msghdr __user *)msg; | |
1da177e4 LT |
1863 | struct socket *sock; |
1864 | struct iovec iovstack[UIO_FASTIOV]; | |
89bddce5 | 1865 | struct iovec *iov = iovstack; |
1da177e4 LT |
1866 | struct msghdr msg_sys; |
1867 | unsigned long cmsg_ptr; | |
1868 | int err, iov_size, total_len, len; | |
6cb153ca | 1869 | int fput_needed; |
1da177e4 LT |
1870 | |
1871 | /* kernel mode address */ | |
1872 | char addr[MAX_SOCK_ADDR]; | |
1873 | ||
1874 | /* user mode address pointers */ | |
1875 | struct sockaddr __user *uaddr; | |
1876 | int __user *uaddr_len; | |
89bddce5 | 1877 | |
1da177e4 LT |
1878 | if (MSG_CMSG_COMPAT & flags) { |
1879 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1880 | return -EFAULT; | |
89bddce5 SH |
1881 | } |
1882 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1883 | return -EFAULT; | |
1da177e4 | 1884 | |
6cb153ca | 1885 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1886 | if (!sock) |
1887 | goto out; | |
1888 | ||
1889 | err = -EMSGSIZE; | |
1890 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
1891 | goto out_put; | |
89bddce5 SH |
1892 | |
1893 | /* Check whether to allocate the iovec area */ | |
1da177e4 LT |
1894 | err = -ENOMEM; |
1895 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
1896 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
1897 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
1898 | if (!iov) | |
1899 | goto out_put; | |
1900 | } | |
1901 | ||
1902 | /* | |
89bddce5 SH |
1903 | * Save the user-mode address (verify_iovec will change the |
1904 | * kernel msghdr to use the kernel address space) | |
1da177e4 | 1905 | */ |
89bddce5 SH |
1906 | |
1907 | uaddr = (void __user *)msg_sys.msg_name; | |
1da177e4 LT |
1908 | uaddr_len = COMPAT_NAMELEN(msg); |
1909 | if (MSG_CMSG_COMPAT & flags) { | |
1910 | err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE); | |
1911 | } else | |
1912 | err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE); | |
1913 | if (err < 0) | |
1914 | goto out_freeiov; | |
89bddce5 | 1915 | total_len = err; |
1da177e4 LT |
1916 | |
1917 | cmsg_ptr = (unsigned long)msg_sys.msg_control; | |
1918 | msg_sys.msg_flags = 0; | |
1919 | if (MSG_CMSG_COMPAT & flags) | |
1920 | msg_sys.msg_flags = MSG_CMSG_COMPAT; | |
89bddce5 | 1921 | |
1da177e4 LT |
1922 | if (sock->file->f_flags & O_NONBLOCK) |
1923 | flags |= MSG_DONTWAIT; | |
1924 | err = sock_recvmsg(sock, &msg_sys, total_len, flags); | |
1925 | if (err < 0) | |
1926 | goto out_freeiov; | |
1927 | len = err; | |
1928 | ||
1929 | if (uaddr != NULL) { | |
89bddce5 SH |
1930 | err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, |
1931 | uaddr_len); | |
1da177e4 LT |
1932 | if (err < 0) |
1933 | goto out_freeiov; | |
1934 | } | |
37f7f421 DM |
1935 | err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT), |
1936 | COMPAT_FLAGS(msg)); | |
1da177e4 LT |
1937 | if (err) |
1938 | goto out_freeiov; | |
1939 | if (MSG_CMSG_COMPAT & flags) | |
89bddce5 | 1940 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, |
1da177e4 LT |
1941 | &msg_compat->msg_controllen); |
1942 | else | |
89bddce5 | 1943 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, |
1da177e4 LT |
1944 | &msg->msg_controllen); |
1945 | if (err) | |
1946 | goto out_freeiov; | |
1947 | err = len; | |
1948 | ||
1949 | out_freeiov: | |
1950 | if (iov != iovstack) | |
1951 | sock_kfree_s(sock->sk, iov, iov_size); | |
1952 | out_put: | |
6cb153ca | 1953 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1954 | out: |
1955 | return err; | |
1956 | } | |
1957 | ||
1958 | #ifdef __ARCH_WANT_SYS_SOCKETCALL | |
1959 | ||
1960 | /* Argument list sizes for sys_socketcall */ | |
1961 | #define AL(x) ((x) * sizeof(unsigned long)) | |
89bddce5 SH |
1962 | static const unsigned char nargs[18]={ |
1963 | AL(0),AL(3),AL(3),AL(3),AL(2),AL(3), | |
1964 | AL(3),AL(3),AL(4),AL(4),AL(4),AL(6), | |
1965 | AL(6),AL(2),AL(5),AL(5),AL(3),AL(3) | |
1966 | }; | |
1967 | ||
1da177e4 LT |
1968 | #undef AL |
1969 | ||
1970 | /* | |
89bddce5 | 1971 | * System call vectors. |
1da177e4 LT |
1972 | * |
1973 | * Argument checking cleaned up. Saved 20% in size. | |
1974 | * This function doesn't need to set the kernel lock because | |
89bddce5 | 1975 | * it is set by the callees. |
1da177e4 LT |
1976 | */ |
1977 | ||
1978 | asmlinkage long sys_socketcall(int call, unsigned long __user *args) | |
1979 | { | |
1980 | unsigned long a[6]; | |
89bddce5 | 1981 | unsigned long a0, a1; |
1da177e4 LT |
1982 | int err; |
1983 | ||
89bddce5 | 1984 | if (call < 1 || call > SYS_RECVMSG) |
1da177e4 LT |
1985 | return -EINVAL; |
1986 | ||
1987 | /* copy_from_user should be SMP safe. */ | |
1988 | if (copy_from_user(a, args, nargs[call])) | |
1989 | return -EFAULT; | |
3ec3b2fb | 1990 | |
89bddce5 | 1991 | err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); |
3ec3b2fb DW |
1992 | if (err) |
1993 | return err; | |
1994 | ||
89bddce5 SH |
1995 | a0 = a[0]; |
1996 | a1 = a[1]; | |
1997 | ||
1998 | switch (call) { | |
1999 | case SYS_SOCKET: | |
2000 | err = sys_socket(a0, a1, a[2]); | |
2001 | break; | |
2002 | case SYS_BIND: | |
2003 | err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]); | |
2004 | break; | |
2005 | case SYS_CONNECT: | |
2006 | err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]); | |
2007 | break; | |
2008 | case SYS_LISTEN: | |
2009 | err = sys_listen(a0, a1); | |
2010 | break; | |
2011 | case SYS_ACCEPT: | |
2012 | err = | |
2013 | sys_accept(a0, (struct sockaddr __user *)a1, | |
2014 | (int __user *)a[2]); | |
2015 | break; | |
2016 | case SYS_GETSOCKNAME: | |
2017 | err = | |
2018 | sys_getsockname(a0, (struct sockaddr __user *)a1, | |
2019 | (int __user *)a[2]); | |
2020 | break; | |
2021 | case SYS_GETPEERNAME: | |
2022 | err = | |
2023 | sys_getpeername(a0, (struct sockaddr __user *)a1, | |
2024 | (int __user *)a[2]); | |
2025 | break; | |
2026 | case SYS_SOCKETPAIR: | |
2027 | err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]); | |
2028 | break; | |
2029 | case SYS_SEND: | |
2030 | err = sys_send(a0, (void __user *)a1, a[2], a[3]); | |
2031 | break; | |
2032 | case SYS_SENDTO: | |
2033 | err = sys_sendto(a0, (void __user *)a1, a[2], a[3], | |
2034 | (struct sockaddr __user *)a[4], a[5]); | |
2035 | break; | |
2036 | case SYS_RECV: | |
2037 | err = sys_recv(a0, (void __user *)a1, a[2], a[3]); | |
2038 | break; | |
2039 | case SYS_RECVFROM: | |
2040 | err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3], | |
2041 | (struct sockaddr __user *)a[4], | |
2042 | (int __user *)a[5]); | |
2043 | break; | |
2044 | case SYS_SHUTDOWN: | |
2045 | err = sys_shutdown(a0, a1); | |
2046 | break; | |
2047 | case SYS_SETSOCKOPT: | |
2048 | err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]); | |
2049 | break; | |
2050 | case SYS_GETSOCKOPT: | |
2051 | err = | |
2052 | sys_getsockopt(a0, a1, a[2], (char __user *)a[3], | |
2053 | (int __user *)a[4]); | |
2054 | break; | |
2055 | case SYS_SENDMSG: | |
2056 | err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2057 | break; | |
2058 | case SYS_RECVMSG: | |
2059 | err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2060 | break; | |
2061 | default: | |
2062 | err = -EINVAL; | |
2063 | break; | |
1da177e4 LT |
2064 | } |
2065 | return err; | |
2066 | } | |
2067 | ||
89bddce5 | 2068 | #endif /* __ARCH_WANT_SYS_SOCKETCALL */ |
1da177e4 | 2069 | |
55737fda SH |
2070 | /** |
2071 | * sock_register - add a socket protocol handler | |
2072 | * @ops: description of protocol | |
2073 | * | |
1da177e4 LT |
2074 | * This function is called by a protocol handler that wants to |
2075 | * advertise its address family, and have it linked into the | |
55737fda SH |
2076 | * socket interface. The value ops->family coresponds to the |
2077 | * socket system call protocol family. | |
1da177e4 | 2078 | */ |
f0fd27d4 | 2079 | int sock_register(const struct net_proto_family *ops) |
1da177e4 LT |
2080 | { |
2081 | int err; | |
2082 | ||
2083 | if (ops->family >= NPROTO) { | |
89bddce5 SH |
2084 | printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, |
2085 | NPROTO); | |
1da177e4 LT |
2086 | return -ENOBUFS; |
2087 | } | |
55737fda SH |
2088 | |
2089 | spin_lock(&net_family_lock); | |
2090 | if (net_families[ops->family]) | |
2091 | err = -EEXIST; | |
2092 | else { | |
89bddce5 | 2093 | net_families[ops->family] = ops; |
1da177e4 LT |
2094 | err = 0; |
2095 | } | |
55737fda SH |
2096 | spin_unlock(&net_family_lock); |
2097 | ||
89bddce5 | 2098 | printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family); |
1da177e4 LT |
2099 | return err; |
2100 | } | |
2101 | ||
55737fda SH |
2102 | /** |
2103 | * sock_unregister - remove a protocol handler | |
2104 | * @family: protocol family to remove | |
2105 | * | |
1da177e4 LT |
2106 | * This function is called by a protocol handler that wants to |
2107 | * remove its address family, and have it unlinked from the | |
55737fda SH |
2108 | * new socket creation. |
2109 | * | |
2110 | * If protocol handler is a module, then it can use module reference | |
2111 | * counts to protect against new references. If protocol handler is not | |
2112 | * a module then it needs to provide its own protection in | |
2113 | * the ops->create routine. | |
1da177e4 | 2114 | */ |
f0fd27d4 | 2115 | void sock_unregister(int family) |
1da177e4 | 2116 | { |
f0fd27d4 | 2117 | BUG_ON(family < 0 || family >= NPROTO); |
1da177e4 | 2118 | |
55737fda | 2119 | spin_lock(&net_family_lock); |
89bddce5 | 2120 | net_families[family] = NULL; |
55737fda SH |
2121 | spin_unlock(&net_family_lock); |
2122 | ||
2123 | synchronize_rcu(); | |
2124 | ||
89bddce5 | 2125 | printk(KERN_INFO "NET: Unregistered protocol family %d\n", family); |
1da177e4 LT |
2126 | } |
2127 | ||
77d76ea3 | 2128 | static int __init sock_init(void) |
1da177e4 LT |
2129 | { |
2130 | /* | |
89bddce5 | 2131 | * Initialize sock SLAB cache. |
1da177e4 | 2132 | */ |
89bddce5 | 2133 | |
1da177e4 LT |
2134 | sk_init(); |
2135 | ||
1da177e4 | 2136 | /* |
89bddce5 | 2137 | * Initialize skbuff SLAB cache |
1da177e4 LT |
2138 | */ |
2139 | skb_init(); | |
1da177e4 LT |
2140 | |
2141 | /* | |
89bddce5 | 2142 | * Initialize the protocols module. |
1da177e4 LT |
2143 | */ |
2144 | ||
2145 | init_inodecache(); | |
2146 | register_filesystem(&sock_fs_type); | |
2147 | sock_mnt = kern_mount(&sock_fs_type); | |
77d76ea3 AK |
2148 | |
2149 | /* The real protocol initialization is performed in later initcalls. | |
1da177e4 LT |
2150 | */ |
2151 | ||
2152 | #ifdef CONFIG_NETFILTER | |
2153 | netfilter_init(); | |
2154 | #endif | |
cbeb321a DM |
2155 | |
2156 | return 0; | |
1da177e4 LT |
2157 | } |
2158 | ||
77d76ea3 AK |
2159 | core_initcall(sock_init); /* early initcall */ |
2160 | ||
1da177e4 LT |
2161 | #ifdef CONFIG_PROC_FS |
2162 | void socket_seq_show(struct seq_file *seq) | |
2163 | { | |
2164 | int cpu; | |
2165 | int counter = 0; | |
2166 | ||
6f912042 | 2167 | for_each_possible_cpu(cpu) |
89bddce5 | 2168 | counter += per_cpu(sockets_in_use, cpu); |
1da177e4 LT |
2169 | |
2170 | /* It can be negative, by the way. 8) */ | |
2171 | if (counter < 0) | |
2172 | counter = 0; | |
2173 | ||
2174 | seq_printf(seq, "sockets: used %d\n", counter); | |
2175 | } | |
89bddce5 | 2176 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 2177 | |
89bbfc95 SP |
2178 | #ifdef CONFIG_COMPAT |
2179 | static long compat_sock_ioctl(struct file *file, unsigned cmd, | |
89bddce5 | 2180 | unsigned long arg) |
89bbfc95 SP |
2181 | { |
2182 | struct socket *sock = file->private_data; | |
2183 | int ret = -ENOIOCTLCMD; | |
2184 | ||
2185 | if (sock->ops->compat_ioctl) | |
2186 | ret = sock->ops->compat_ioctl(sock, cmd, arg); | |
2187 | ||
2188 | return ret; | |
2189 | } | |
2190 | #endif | |
2191 | ||
ac5a488e SS |
2192 | int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) |
2193 | { | |
2194 | return sock->ops->bind(sock, addr, addrlen); | |
2195 | } | |
2196 | ||
2197 | int kernel_listen(struct socket *sock, int backlog) | |
2198 | { | |
2199 | return sock->ops->listen(sock, backlog); | |
2200 | } | |
2201 | ||
2202 | int kernel_accept(struct socket *sock, struct socket **newsock, int flags) | |
2203 | { | |
2204 | struct sock *sk = sock->sk; | |
2205 | int err; | |
2206 | ||
2207 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
2208 | newsock); | |
2209 | if (err < 0) | |
2210 | goto done; | |
2211 | ||
2212 | err = sock->ops->accept(sock, *newsock, flags); | |
2213 | if (err < 0) { | |
2214 | sock_release(*newsock); | |
2215 | goto done; | |
2216 | } | |
2217 | ||
2218 | (*newsock)->ops = sock->ops; | |
2219 | ||
2220 | done: | |
2221 | return err; | |
2222 | } | |
2223 | ||
2224 | int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, | |
2225 | int flags) | |
2226 | { | |
2227 | return sock->ops->connect(sock, addr, addrlen, flags); | |
2228 | } | |
2229 | ||
2230 | int kernel_getsockname(struct socket *sock, struct sockaddr *addr, | |
2231 | int *addrlen) | |
2232 | { | |
2233 | return sock->ops->getname(sock, addr, addrlen, 0); | |
2234 | } | |
2235 | ||
2236 | int kernel_getpeername(struct socket *sock, struct sockaddr *addr, | |
2237 | int *addrlen) | |
2238 | { | |
2239 | return sock->ops->getname(sock, addr, addrlen, 1); | |
2240 | } | |
2241 | ||
2242 | int kernel_getsockopt(struct socket *sock, int level, int optname, | |
2243 | char *optval, int *optlen) | |
2244 | { | |
2245 | mm_segment_t oldfs = get_fs(); | |
2246 | int err; | |
2247 | ||
2248 | set_fs(KERNEL_DS); | |
2249 | if (level == SOL_SOCKET) | |
2250 | err = sock_getsockopt(sock, level, optname, optval, optlen); | |
2251 | else | |
2252 | err = sock->ops->getsockopt(sock, level, optname, optval, | |
2253 | optlen); | |
2254 | set_fs(oldfs); | |
2255 | return err; | |
2256 | } | |
2257 | ||
2258 | int kernel_setsockopt(struct socket *sock, int level, int optname, | |
2259 | char *optval, int optlen) | |
2260 | { | |
2261 | mm_segment_t oldfs = get_fs(); | |
2262 | int err; | |
2263 | ||
2264 | set_fs(KERNEL_DS); | |
2265 | if (level == SOL_SOCKET) | |
2266 | err = sock_setsockopt(sock, level, optname, optval, optlen); | |
2267 | else | |
2268 | err = sock->ops->setsockopt(sock, level, optname, optval, | |
2269 | optlen); | |
2270 | set_fs(oldfs); | |
2271 | return err; | |
2272 | } | |
2273 | ||
2274 | int kernel_sendpage(struct socket *sock, struct page *page, int offset, | |
2275 | size_t size, int flags) | |
2276 | { | |
2277 | if (sock->ops->sendpage) | |
2278 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
2279 | ||
2280 | return sock_no_sendpage(sock, page, offset, size, flags); | |
2281 | } | |
2282 | ||
2283 | int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg) | |
2284 | { | |
2285 | mm_segment_t oldfs = get_fs(); | |
2286 | int err; | |
2287 | ||
2288 | set_fs(KERNEL_DS); | |
2289 | err = sock->ops->ioctl(sock, cmd, arg); | |
2290 | set_fs(oldfs); | |
2291 | ||
2292 | return err; | |
2293 | } | |
2294 | ||
1da177e4 LT |
2295 | /* ABI emulation layers need these two */ |
2296 | EXPORT_SYMBOL(move_addr_to_kernel); | |
2297 | EXPORT_SYMBOL(move_addr_to_user); | |
2298 | EXPORT_SYMBOL(sock_create); | |
2299 | EXPORT_SYMBOL(sock_create_kern); | |
2300 | EXPORT_SYMBOL(sock_create_lite); | |
2301 | EXPORT_SYMBOL(sock_map_fd); | |
2302 | EXPORT_SYMBOL(sock_recvmsg); | |
2303 | EXPORT_SYMBOL(sock_register); | |
2304 | EXPORT_SYMBOL(sock_release); | |
2305 | EXPORT_SYMBOL(sock_sendmsg); | |
2306 | EXPORT_SYMBOL(sock_unregister); | |
2307 | EXPORT_SYMBOL(sock_wake_async); | |
2308 | EXPORT_SYMBOL(sockfd_lookup); | |
2309 | EXPORT_SYMBOL(kernel_sendmsg); | |
2310 | EXPORT_SYMBOL(kernel_recvmsg); | |
ac5a488e SS |
2311 | EXPORT_SYMBOL(kernel_bind); |
2312 | EXPORT_SYMBOL(kernel_listen); | |
2313 | EXPORT_SYMBOL(kernel_accept); | |
2314 | EXPORT_SYMBOL(kernel_connect); | |
2315 | EXPORT_SYMBOL(kernel_getsockname); | |
2316 | EXPORT_SYMBOL(kernel_getpeername); | |
2317 | EXPORT_SYMBOL(kernel_getsockopt); | |
2318 | EXPORT_SYMBOL(kernel_setsockopt); | |
2319 | EXPORT_SYMBOL(kernel_sendpage); | |
2320 | EXPORT_SYMBOL(kernel_sock_ioctl); |