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> | |
aaca0bdc | 66 | #include <linux/thread_info.h> |
55737fda | 67 | #include <linux/rcupdate.h> |
1da177e4 LT |
68 | #include <linux/netdevice.h> |
69 | #include <linux/proc_fs.h> | |
70 | #include <linux/seq_file.h> | |
4a3e2f71 | 71 | #include <linux/mutex.h> |
1da177e4 LT |
72 | #include <linux/wanrouter.h> |
73 | #include <linux/if_bridge.h> | |
20380731 ACM |
74 | #include <linux/if_frad.h> |
75 | #include <linux/if_vlan.h> | |
1da177e4 LT |
76 | #include <linux/init.h> |
77 | #include <linux/poll.h> | |
78 | #include <linux/cache.h> | |
79 | #include <linux/module.h> | |
80 | #include <linux/highmem.h> | |
1da177e4 LT |
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> |
1b8d7ae4 | 88 | #include <linux/nsproxy.h> |
1fd7317d | 89 | #include <linux/magic.h> |
1da177e4 LT |
90 | |
91 | #include <asm/uaccess.h> | |
92 | #include <asm/unistd.h> | |
93 | ||
94 | #include <net/compat.h> | |
87de87d5 | 95 | #include <net/wext.h> |
1da177e4 LT |
96 | |
97 | #include <net/sock.h> | |
98 | #include <linux/netfilter.h> | |
99 | ||
100 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare); | |
027445c3 BP |
101 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, |
102 | unsigned long nr_segs, loff_t pos); | |
103 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
104 | unsigned long nr_segs, loff_t pos); | |
89bddce5 | 105 | static int sock_mmap(struct file *file, struct vm_area_struct *vma); |
1da177e4 LT |
106 | |
107 | static int sock_close(struct inode *inode, struct file *file); | |
108 | static unsigned int sock_poll(struct file *file, | |
109 | struct poll_table_struct *wait); | |
89bddce5 | 110 | static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); |
89bbfc95 SP |
111 | #ifdef CONFIG_COMPAT |
112 | static long compat_sock_ioctl(struct file *file, | |
89bddce5 | 113 | unsigned int cmd, unsigned long arg); |
89bbfc95 | 114 | #endif |
1da177e4 | 115 | static int sock_fasync(int fd, struct file *filp, int on); |
1da177e4 LT |
116 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
117 | int offset, size_t size, loff_t *ppos, int more); | |
9c55e01c JA |
118 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, |
119 | struct pipe_inode_info *pipe, size_t len, | |
120 | unsigned int flags); | |
1da177e4 | 121 | |
1da177e4 LT |
122 | /* |
123 | * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear | |
124 | * in the operation structures but are done directly via the socketcall() multiplexor. | |
125 | */ | |
126 | ||
da7071d7 | 127 | static const struct file_operations socket_file_ops = { |
1da177e4 LT |
128 | .owner = THIS_MODULE, |
129 | .llseek = no_llseek, | |
130 | .aio_read = sock_aio_read, | |
131 | .aio_write = sock_aio_write, | |
132 | .poll = sock_poll, | |
133 | .unlocked_ioctl = sock_ioctl, | |
89bbfc95 SP |
134 | #ifdef CONFIG_COMPAT |
135 | .compat_ioctl = compat_sock_ioctl, | |
136 | #endif | |
1da177e4 LT |
137 | .mmap = sock_mmap, |
138 | .open = sock_no_open, /* special open code to disallow open via /proc */ | |
139 | .release = sock_close, | |
140 | .fasync = sock_fasync, | |
5274f052 JA |
141 | .sendpage = sock_sendpage, |
142 | .splice_write = generic_splice_sendpage, | |
9c55e01c | 143 | .splice_read = sock_splice_read, |
1da177e4 LT |
144 | }; |
145 | ||
146 | /* | |
147 | * The protocol list. Each protocol is registered in here. | |
148 | */ | |
149 | ||
1da177e4 | 150 | static DEFINE_SPINLOCK(net_family_lock); |
f0fd27d4 | 151 | static const struct net_proto_family *net_families[NPROTO] __read_mostly; |
1da177e4 | 152 | |
1da177e4 LT |
153 | /* |
154 | * Statistics counters of the socket lists | |
155 | */ | |
156 | ||
157 | static DEFINE_PER_CPU(int, sockets_in_use) = 0; | |
158 | ||
159 | /* | |
89bddce5 SH |
160 | * Support routines. |
161 | * Move socket addresses back and forth across the kernel/user | |
162 | * divide and look after the messy bits. | |
1da177e4 LT |
163 | */ |
164 | ||
89bddce5 | 165 | #define MAX_SOCK_ADDR 128 /* 108 for Unix domain - |
1da177e4 LT |
166 | 16 for IP, 16 for IPX, |
167 | 24 for IPv6, | |
89bddce5 | 168 | about 80 for AX.25 |
1da177e4 LT |
169 | must be at least one bigger than |
170 | the AF_UNIX size (see net/unix/af_unix.c | |
89bddce5 | 171 | :unix_mkname()). |
1da177e4 | 172 | */ |
89bddce5 | 173 | |
1da177e4 LT |
174 | /** |
175 | * move_addr_to_kernel - copy a socket address into kernel space | |
176 | * @uaddr: Address in user space | |
177 | * @kaddr: Address in kernel space | |
178 | * @ulen: Length in user space | |
179 | * | |
180 | * The address is copied into kernel space. If the provided address is | |
181 | * too long an error code of -EINVAL is returned. If the copy gives | |
182 | * invalid addresses -EFAULT is returned. On a success 0 is returned. | |
183 | */ | |
184 | ||
230b1839 | 185 | int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr) |
1da177e4 | 186 | { |
230b1839 | 187 | if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) |
1da177e4 | 188 | return -EINVAL; |
89bddce5 | 189 | if (ulen == 0) |
1da177e4 | 190 | return 0; |
89bddce5 | 191 | if (copy_from_user(kaddr, uaddr, ulen)) |
1da177e4 | 192 | return -EFAULT; |
3ec3b2fb | 193 | return audit_sockaddr(ulen, kaddr); |
1da177e4 LT |
194 | } |
195 | ||
196 | /** | |
197 | * move_addr_to_user - copy an address to user space | |
198 | * @kaddr: kernel space address | |
199 | * @klen: length of address in kernel | |
200 | * @uaddr: user space address | |
201 | * @ulen: pointer to user length field | |
202 | * | |
203 | * The value pointed to by ulen on entry is the buffer length available. | |
204 | * This is overwritten with the buffer space used. -EINVAL is returned | |
205 | * if an overlong buffer is specified or a negative buffer size. -EFAULT | |
206 | * is returned if either the buffer or the length field are not | |
207 | * accessible. | |
208 | * After copying the data up to the limit the user specifies, the true | |
209 | * length of the data is written over the length limit the user | |
210 | * specified. Zero is returned for a success. | |
211 | */ | |
89bddce5 | 212 | |
230b1839 | 213 | int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr, |
89bddce5 | 214 | int __user *ulen) |
1da177e4 LT |
215 | { |
216 | int err; | |
217 | int len; | |
218 | ||
89bddce5 SH |
219 | err = get_user(len, ulen); |
220 | if (err) | |
1da177e4 | 221 | return err; |
89bddce5 SH |
222 | if (len > klen) |
223 | len = klen; | |
230b1839 | 224 | if (len < 0 || len > sizeof(struct sockaddr_storage)) |
1da177e4 | 225 | return -EINVAL; |
89bddce5 | 226 | if (len) { |
d6fe3945 SG |
227 | if (audit_sockaddr(klen, kaddr)) |
228 | return -ENOMEM; | |
89bddce5 | 229 | if (copy_to_user(uaddr, kaddr, len)) |
1da177e4 LT |
230 | return -EFAULT; |
231 | } | |
232 | /* | |
89bddce5 SH |
233 | * "fromlen shall refer to the value before truncation.." |
234 | * 1003.1g | |
1da177e4 LT |
235 | */ |
236 | return __put_user(klen, ulen); | |
237 | } | |
238 | ||
e18b890b | 239 | static struct kmem_cache *sock_inode_cachep __read_mostly; |
1da177e4 LT |
240 | |
241 | static struct inode *sock_alloc_inode(struct super_block *sb) | |
242 | { | |
243 | struct socket_alloc *ei; | |
89bddce5 | 244 | |
e94b1766 | 245 | ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); |
1da177e4 LT |
246 | if (!ei) |
247 | return NULL; | |
248 | init_waitqueue_head(&ei->socket.wait); | |
89bddce5 | 249 | |
1da177e4 LT |
250 | ei->socket.fasync_list = NULL; |
251 | ei->socket.state = SS_UNCONNECTED; | |
252 | ei->socket.flags = 0; | |
253 | ei->socket.ops = NULL; | |
254 | ei->socket.sk = NULL; | |
255 | ei->socket.file = NULL; | |
1da177e4 LT |
256 | |
257 | return &ei->vfs_inode; | |
258 | } | |
259 | ||
260 | static void sock_destroy_inode(struct inode *inode) | |
261 | { | |
262 | kmem_cache_free(sock_inode_cachep, | |
263 | container_of(inode, struct socket_alloc, vfs_inode)); | |
264 | } | |
265 | ||
51cc5068 | 266 | static void init_once(void *foo) |
1da177e4 | 267 | { |
89bddce5 | 268 | struct socket_alloc *ei = (struct socket_alloc *)foo; |
1da177e4 | 269 | |
a35afb83 | 270 | inode_init_once(&ei->vfs_inode); |
1da177e4 | 271 | } |
89bddce5 | 272 | |
1da177e4 LT |
273 | static int init_inodecache(void) |
274 | { | |
275 | sock_inode_cachep = kmem_cache_create("sock_inode_cache", | |
89bddce5 SH |
276 | sizeof(struct socket_alloc), |
277 | 0, | |
278 | (SLAB_HWCACHE_ALIGN | | |
279 | SLAB_RECLAIM_ACCOUNT | | |
280 | SLAB_MEM_SPREAD), | |
20c2df83 | 281 | init_once); |
1da177e4 LT |
282 | if (sock_inode_cachep == NULL) |
283 | return -ENOMEM; | |
284 | return 0; | |
285 | } | |
286 | ||
b87221de | 287 | static const struct super_operations sockfs_ops = { |
1da177e4 LT |
288 | .alloc_inode = sock_alloc_inode, |
289 | .destroy_inode =sock_destroy_inode, | |
290 | .statfs = simple_statfs, | |
291 | }; | |
292 | ||
454e2398 | 293 | static int sockfs_get_sb(struct file_system_type *fs_type, |
89bddce5 SH |
294 | int flags, const char *dev_name, void *data, |
295 | struct vfsmount *mnt) | |
1da177e4 | 296 | { |
454e2398 DH |
297 | return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC, |
298 | mnt); | |
1da177e4 LT |
299 | } |
300 | ||
ba89966c | 301 | static struct vfsmount *sock_mnt __read_mostly; |
1da177e4 LT |
302 | |
303 | static struct file_system_type sock_fs_type = { | |
304 | .name = "sockfs", | |
305 | .get_sb = sockfs_get_sb, | |
306 | .kill_sb = kill_anon_super, | |
307 | }; | |
89bddce5 | 308 | |
1da177e4 LT |
309 | static int sockfs_delete_dentry(struct dentry *dentry) |
310 | { | |
304e61e6 ED |
311 | /* |
312 | * At creation time, we pretended this dentry was hashed | |
313 | * (by clearing DCACHE_UNHASHED bit in d_flags) | |
314 | * At delete time, we restore the truth : not hashed. | |
315 | * (so that dput() can proceed correctly) | |
316 | */ | |
317 | dentry->d_flags |= DCACHE_UNHASHED; | |
318 | return 0; | |
1da177e4 | 319 | } |
c23fbb6b ED |
320 | |
321 | /* | |
322 | * sockfs_dname() is called from d_path(). | |
323 | */ | |
324 | static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) | |
325 | { | |
326 | return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", | |
327 | dentry->d_inode->i_ino); | |
328 | } | |
329 | ||
3ba13d17 | 330 | static const struct dentry_operations sockfs_dentry_operations = { |
89bddce5 | 331 | .d_delete = sockfs_delete_dentry, |
c23fbb6b | 332 | .d_dname = sockfs_dname, |
1da177e4 LT |
333 | }; |
334 | ||
335 | /* | |
336 | * Obtains the first available file descriptor and sets it up for use. | |
337 | * | |
39d8c1b6 DM |
338 | * These functions create file structures and maps them to fd space |
339 | * of the current process. On success it returns file descriptor | |
1da177e4 LT |
340 | * and file struct implicitly stored in sock->file. |
341 | * Note that another thread may close file descriptor before we return | |
342 | * from this function. We use the fact that now we do not refer | |
343 | * to socket after mapping. If one day we will need it, this | |
344 | * function will increment ref. count on file by 1. | |
345 | * | |
346 | * In any case returned fd MAY BE not valid! | |
347 | * This race condition is unavoidable | |
348 | * with shared fd spaces, we cannot solve it inside kernel, | |
349 | * but we take care of internal coherence yet. | |
350 | */ | |
351 | ||
a677a039 | 352 | static int sock_alloc_fd(struct file **filep, int flags) |
1da177e4 LT |
353 | { |
354 | int fd; | |
1da177e4 | 355 | |
a677a039 | 356 | fd = get_unused_fd_flags(flags); |
39d8c1b6 | 357 | if (likely(fd >= 0)) { |
1da177e4 LT |
358 | struct file *file = get_empty_filp(); |
359 | ||
39d8c1b6 DM |
360 | *filep = file; |
361 | if (unlikely(!file)) { | |
1da177e4 | 362 | put_unused_fd(fd); |
39d8c1b6 | 363 | return -ENFILE; |
1da177e4 | 364 | } |
39d8c1b6 DM |
365 | } else |
366 | *filep = NULL; | |
367 | return fd; | |
368 | } | |
1da177e4 | 369 | |
77d27200 | 370 | static int sock_attach_fd(struct socket *sock, struct file *file, int flags) |
39d8c1b6 | 371 | { |
ce8d2cdf | 372 | struct dentry *dentry; |
c23fbb6b | 373 | struct qstr name = { .name = "" }; |
39d8c1b6 | 374 | |
ce8d2cdf DH |
375 | dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name); |
376 | if (unlikely(!dentry)) | |
39d8c1b6 DM |
377 | return -ENOMEM; |
378 | ||
ce8d2cdf | 379 | dentry->d_op = &sockfs_dentry_operations; |
304e61e6 ED |
380 | /* |
381 | * We dont want to push this dentry into global dentry hash table. | |
382 | * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED | |
383 | * This permits a working /proc/$pid/fd/XXX on sockets | |
384 | */ | |
ce8d2cdf DH |
385 | dentry->d_flags &= ~DCACHE_UNHASHED; |
386 | d_instantiate(dentry, SOCK_INODE(sock)); | |
39d8c1b6 DM |
387 | |
388 | sock->file = file; | |
ce8d2cdf DH |
389 | init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE, |
390 | &socket_file_ops); | |
391 | SOCK_INODE(sock)->i_fop = &socket_file_ops; | |
77d27200 | 392 | file->f_flags = O_RDWR | (flags & O_NONBLOCK); |
39d8c1b6 DM |
393 | file->f_pos = 0; |
394 | file->private_data = sock; | |
1da177e4 | 395 | |
39d8c1b6 DM |
396 | return 0; |
397 | } | |
398 | ||
a677a039 | 399 | int sock_map_fd(struct socket *sock, int flags) |
39d8c1b6 DM |
400 | { |
401 | struct file *newfile; | |
a677a039 | 402 | int fd = sock_alloc_fd(&newfile, flags); |
39d8c1b6 DM |
403 | |
404 | if (likely(fd >= 0)) { | |
77d27200 | 405 | int err = sock_attach_fd(sock, newfile, flags); |
39d8c1b6 DM |
406 | |
407 | if (unlikely(err < 0)) { | |
408 | put_filp(newfile); | |
1da177e4 | 409 | put_unused_fd(fd); |
39d8c1b6 | 410 | return err; |
1da177e4 | 411 | } |
39d8c1b6 | 412 | fd_install(fd, newfile); |
1da177e4 | 413 | } |
1da177e4 LT |
414 | return fd; |
415 | } | |
416 | ||
6cb153ca BL |
417 | static struct socket *sock_from_file(struct file *file, int *err) |
418 | { | |
6cb153ca BL |
419 | if (file->f_op == &socket_file_ops) |
420 | return file->private_data; /* set in sock_map_fd */ | |
421 | ||
23bb80d2 ED |
422 | *err = -ENOTSOCK; |
423 | return NULL; | |
6cb153ca BL |
424 | } |
425 | ||
1da177e4 LT |
426 | /** |
427 | * sockfd_lookup - Go from a file number to its socket slot | |
428 | * @fd: file handle | |
429 | * @err: pointer to an error code return | |
430 | * | |
431 | * The file handle passed in is locked and the socket it is bound | |
432 | * too is returned. If an error occurs the err pointer is overwritten | |
433 | * with a negative errno code and NULL is returned. The function checks | |
434 | * for both invalid handles and passing a handle which is not a socket. | |
435 | * | |
436 | * On a success the socket object pointer is returned. | |
437 | */ | |
438 | ||
439 | struct socket *sockfd_lookup(int fd, int *err) | |
440 | { | |
441 | struct file *file; | |
1da177e4 LT |
442 | struct socket *sock; |
443 | ||
89bddce5 SH |
444 | file = fget(fd); |
445 | if (!file) { | |
1da177e4 LT |
446 | *err = -EBADF; |
447 | return NULL; | |
448 | } | |
89bddce5 | 449 | |
6cb153ca BL |
450 | sock = sock_from_file(file, err); |
451 | if (!sock) | |
1da177e4 | 452 | fput(file); |
6cb153ca BL |
453 | return sock; |
454 | } | |
1da177e4 | 455 | |
6cb153ca BL |
456 | static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) |
457 | { | |
458 | struct file *file; | |
459 | struct socket *sock; | |
460 | ||
3672558c | 461 | *err = -EBADF; |
6cb153ca BL |
462 | file = fget_light(fd, fput_needed); |
463 | if (file) { | |
464 | sock = sock_from_file(file, err); | |
465 | if (sock) | |
466 | return sock; | |
467 | fput_light(file, *fput_needed); | |
1da177e4 | 468 | } |
6cb153ca | 469 | return NULL; |
1da177e4 LT |
470 | } |
471 | ||
472 | /** | |
473 | * sock_alloc - allocate a socket | |
89bddce5 | 474 | * |
1da177e4 LT |
475 | * Allocate a new inode and socket object. The two are bound together |
476 | * and initialised. The socket is then returned. If we are out of inodes | |
477 | * NULL is returned. | |
478 | */ | |
479 | ||
480 | static struct socket *sock_alloc(void) | |
481 | { | |
89bddce5 SH |
482 | struct inode *inode; |
483 | struct socket *sock; | |
1da177e4 LT |
484 | |
485 | inode = new_inode(sock_mnt->mnt_sb); | |
486 | if (!inode) | |
487 | return NULL; | |
488 | ||
489 | sock = SOCKET_I(inode); | |
490 | ||
29a020d3 | 491 | kmemcheck_annotate_bitfield(sock, type); |
89bddce5 | 492 | inode->i_mode = S_IFSOCK | S_IRWXUGO; |
8192b0c4 DH |
493 | inode->i_uid = current_fsuid(); |
494 | inode->i_gid = current_fsgid(); | |
1da177e4 | 495 | |
4e69489a | 496 | percpu_add(sockets_in_use, 1); |
1da177e4 LT |
497 | return sock; |
498 | } | |
499 | ||
500 | /* | |
501 | * In theory you can't get an open on this inode, but /proc provides | |
502 | * a back door. Remember to keep it shut otherwise you'll let the | |
503 | * creepy crawlies in. | |
504 | */ | |
89bddce5 | 505 | |
1da177e4 LT |
506 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare) |
507 | { | |
508 | return -ENXIO; | |
509 | } | |
510 | ||
4b6f5d20 | 511 | const struct file_operations bad_sock_fops = { |
1da177e4 LT |
512 | .owner = THIS_MODULE, |
513 | .open = sock_no_open, | |
514 | }; | |
515 | ||
516 | /** | |
517 | * sock_release - close a socket | |
518 | * @sock: socket to close | |
519 | * | |
520 | * The socket is released from the protocol stack if it has a release | |
521 | * callback, and the inode is then released if the socket is bound to | |
89bddce5 | 522 | * an inode not a file. |
1da177e4 | 523 | */ |
89bddce5 | 524 | |
1da177e4 LT |
525 | void sock_release(struct socket *sock) |
526 | { | |
527 | if (sock->ops) { | |
528 | struct module *owner = sock->ops->owner; | |
529 | ||
530 | sock->ops->release(sock); | |
531 | sock->ops = NULL; | |
532 | module_put(owner); | |
533 | } | |
534 | ||
535 | if (sock->fasync_list) | |
536 | printk(KERN_ERR "sock_release: fasync list not empty!\n"); | |
537 | ||
4e69489a | 538 | percpu_sub(sockets_in_use, 1); |
1da177e4 LT |
539 | if (!sock->file) { |
540 | iput(SOCK_INODE(sock)); | |
541 | return; | |
542 | } | |
89bddce5 | 543 | sock->file = NULL; |
1da177e4 LT |
544 | } |
545 | ||
20d49473 PO |
546 | int sock_tx_timestamp(struct msghdr *msg, struct sock *sk, |
547 | union skb_shared_tx *shtx) | |
548 | { | |
549 | shtx->flags = 0; | |
550 | if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE)) | |
551 | shtx->hardware = 1; | |
552 | if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE)) | |
553 | shtx->software = 1; | |
554 | return 0; | |
555 | } | |
556 | EXPORT_SYMBOL(sock_tx_timestamp); | |
557 | ||
89bddce5 | 558 | static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
559 | struct msghdr *msg, size_t size) |
560 | { | |
561 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
562 | int err; | |
563 | ||
564 | si->sock = sock; | |
565 | si->scm = NULL; | |
566 | si->msg = msg; | |
567 | si->size = size; | |
568 | ||
569 | err = security_socket_sendmsg(sock, msg, size); | |
570 | if (err) | |
571 | return err; | |
572 | ||
573 | return sock->ops->sendmsg(iocb, sock, msg, size); | |
574 | } | |
575 | ||
576 | int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) | |
577 | { | |
578 | struct kiocb iocb; | |
579 | struct sock_iocb siocb; | |
580 | int ret; | |
581 | ||
582 | init_sync_kiocb(&iocb, NULL); | |
583 | iocb.private = &siocb; | |
584 | ret = __sock_sendmsg(&iocb, sock, msg, size); | |
585 | if (-EIOCBQUEUED == ret) | |
586 | ret = wait_on_sync_kiocb(&iocb); | |
587 | return ret; | |
588 | } | |
589 | ||
590 | int kernel_sendmsg(struct socket *sock, struct msghdr *msg, | |
591 | struct kvec *vec, size_t num, size_t size) | |
592 | { | |
593 | mm_segment_t oldfs = get_fs(); | |
594 | int result; | |
595 | ||
596 | set_fs(KERNEL_DS); | |
597 | /* | |
598 | * the following is safe, since for compiler definitions of kvec and | |
599 | * iovec are identical, yielding the same in-core layout and alignment | |
600 | */ | |
89bddce5 | 601 | msg->msg_iov = (struct iovec *)vec; |
1da177e4 LT |
602 | msg->msg_iovlen = num; |
603 | result = sock_sendmsg(sock, msg, size); | |
604 | set_fs(oldfs); | |
605 | return result; | |
606 | } | |
607 | ||
20d49473 PO |
608 | static int ktime2ts(ktime_t kt, struct timespec *ts) |
609 | { | |
610 | if (kt.tv64) { | |
611 | *ts = ktime_to_timespec(kt); | |
612 | return 1; | |
613 | } else { | |
614 | return 0; | |
615 | } | |
616 | } | |
617 | ||
92f37fd2 ED |
618 | /* |
619 | * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) | |
620 | */ | |
621 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, | |
622 | struct sk_buff *skb) | |
623 | { | |
20d49473 PO |
624 | int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); |
625 | struct timespec ts[3]; | |
626 | int empty = 1; | |
627 | struct skb_shared_hwtstamps *shhwtstamps = | |
628 | skb_hwtstamps(skb); | |
629 | ||
630 | /* Race occurred between timestamp enabling and packet | |
631 | receiving. Fill in the current time for now. */ | |
632 | if (need_software_tstamp && skb->tstamp.tv64 == 0) | |
633 | __net_timestamp(skb); | |
634 | ||
635 | if (need_software_tstamp) { | |
636 | if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { | |
637 | struct timeval tv; | |
638 | skb_get_timestamp(skb, &tv); | |
639 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, | |
640 | sizeof(tv), &tv); | |
641 | } else { | |
642 | struct timespec ts; | |
643 | skb_get_timestampns(skb, &ts); | |
644 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, | |
645 | sizeof(ts), &ts); | |
646 | } | |
647 | } | |
648 | ||
649 | ||
650 | memset(ts, 0, sizeof(ts)); | |
651 | if (skb->tstamp.tv64 && | |
652 | sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) { | |
653 | skb_get_timestampns(skb, ts + 0); | |
654 | empty = 0; | |
655 | } | |
656 | if (shhwtstamps) { | |
657 | if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) && | |
658 | ktime2ts(shhwtstamps->syststamp, ts + 1)) | |
659 | empty = 0; | |
660 | if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) && | |
661 | ktime2ts(shhwtstamps->hwtstamp, ts + 2)) | |
662 | empty = 0; | |
92f37fd2 | 663 | } |
20d49473 PO |
664 | if (!empty) |
665 | put_cmsg(msg, SOL_SOCKET, | |
666 | SCM_TIMESTAMPING, sizeof(ts), &ts); | |
92f37fd2 ED |
667 | } |
668 | ||
7c81fd8b ACM |
669 | EXPORT_SYMBOL_GPL(__sock_recv_timestamp); |
670 | ||
89bddce5 | 671 | static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
672 | struct msghdr *msg, size_t size, int flags) |
673 | { | |
674 | int err; | |
675 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
676 | ||
677 | si->sock = sock; | |
678 | si->scm = NULL; | |
679 | si->msg = msg; | |
680 | si->size = size; | |
681 | si->flags = flags; | |
682 | ||
683 | err = security_socket_recvmsg(sock, msg, size, flags); | |
684 | if (err) | |
685 | return err; | |
686 | ||
687 | return sock->ops->recvmsg(iocb, sock, msg, size, flags); | |
688 | } | |
689 | ||
89bddce5 | 690 | int sock_recvmsg(struct socket *sock, struct msghdr *msg, |
1da177e4 LT |
691 | size_t size, int flags) |
692 | { | |
693 | struct kiocb iocb; | |
694 | struct sock_iocb siocb; | |
695 | int ret; | |
696 | ||
89bddce5 | 697 | init_sync_kiocb(&iocb, NULL); |
1da177e4 LT |
698 | iocb.private = &siocb; |
699 | ret = __sock_recvmsg(&iocb, sock, msg, size, flags); | |
700 | if (-EIOCBQUEUED == ret) | |
701 | ret = wait_on_sync_kiocb(&iocb); | |
702 | return ret; | |
703 | } | |
704 | ||
89bddce5 SH |
705 | int kernel_recvmsg(struct socket *sock, struct msghdr *msg, |
706 | struct kvec *vec, size_t num, size_t size, int flags) | |
1da177e4 LT |
707 | { |
708 | mm_segment_t oldfs = get_fs(); | |
709 | int result; | |
710 | ||
711 | set_fs(KERNEL_DS); | |
712 | /* | |
713 | * the following is safe, since for compiler definitions of kvec and | |
714 | * iovec are identical, yielding the same in-core layout and alignment | |
715 | */ | |
89bddce5 | 716 | msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num; |
1da177e4 LT |
717 | result = sock_recvmsg(sock, msg, size, flags); |
718 | set_fs(oldfs); | |
719 | return result; | |
720 | } | |
721 | ||
722 | static void sock_aio_dtor(struct kiocb *iocb) | |
723 | { | |
724 | kfree(iocb->private); | |
725 | } | |
726 | ||
ce1d4d3e CH |
727 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
728 | int offset, size_t size, loff_t *ppos, int more) | |
1da177e4 | 729 | { |
1da177e4 LT |
730 | struct socket *sock; |
731 | int flags; | |
732 | ||
ce1d4d3e CH |
733 | sock = file->private_data; |
734 | ||
735 | flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT; | |
736 | if (more) | |
737 | flags |= MSG_MORE; | |
738 | ||
e6949583 | 739 | return kernel_sendpage(sock, page, offset, size, flags); |
ce1d4d3e | 740 | } |
1da177e4 | 741 | |
9c55e01c JA |
742 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, |
743 | struct pipe_inode_info *pipe, size_t len, | |
744 | unsigned int flags) | |
745 | { | |
746 | struct socket *sock = file->private_data; | |
747 | ||
997b37da RDC |
748 | if (unlikely(!sock->ops->splice_read)) |
749 | return -EINVAL; | |
750 | ||
9c55e01c JA |
751 | return sock->ops->splice_read(sock, ppos, pipe, len, flags); |
752 | } | |
753 | ||
ce1d4d3e | 754 | static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb, |
89bddce5 | 755 | struct sock_iocb *siocb) |
ce1d4d3e CH |
756 | { |
757 | if (!is_sync_kiocb(iocb)) { | |
758 | siocb = kmalloc(sizeof(*siocb), GFP_KERNEL); | |
759 | if (!siocb) | |
760 | return NULL; | |
1da177e4 LT |
761 | iocb->ki_dtor = sock_aio_dtor; |
762 | } | |
1da177e4 | 763 | |
ce1d4d3e | 764 | siocb->kiocb = iocb; |
ce1d4d3e CH |
765 | iocb->private = siocb; |
766 | return siocb; | |
1da177e4 LT |
767 | } |
768 | ||
ce1d4d3e | 769 | static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb, |
027445c3 BP |
770 | struct file *file, const struct iovec *iov, |
771 | unsigned long nr_segs) | |
ce1d4d3e CH |
772 | { |
773 | struct socket *sock = file->private_data; | |
774 | size_t size = 0; | |
775 | int i; | |
1da177e4 | 776 | |
89bddce5 SH |
777 | for (i = 0; i < nr_segs; i++) |
778 | size += iov[i].iov_len; | |
1da177e4 | 779 | |
ce1d4d3e CH |
780 | msg->msg_name = NULL; |
781 | msg->msg_namelen = 0; | |
782 | msg->msg_control = NULL; | |
783 | msg->msg_controllen = 0; | |
89bddce5 | 784 | msg->msg_iov = (struct iovec *)iov; |
ce1d4d3e CH |
785 | msg->msg_iovlen = nr_segs; |
786 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
787 | ||
788 | return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags); | |
789 | } | |
790 | ||
027445c3 BP |
791 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, |
792 | unsigned long nr_segs, loff_t pos) | |
ce1d4d3e CH |
793 | { |
794 | struct sock_iocb siocb, *x; | |
795 | ||
1da177e4 LT |
796 | if (pos != 0) |
797 | return -ESPIPE; | |
027445c3 BP |
798 | |
799 | if (iocb->ki_left == 0) /* Match SYS5 behaviour */ | |
1da177e4 LT |
800 | return 0; |
801 | ||
027445c3 BP |
802 | |
803 | x = alloc_sock_iocb(iocb, &siocb); | |
ce1d4d3e CH |
804 | if (!x) |
805 | return -ENOMEM; | |
027445c3 | 806 | return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
1da177e4 LT |
807 | } |
808 | ||
ce1d4d3e | 809 | static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb, |
027445c3 BP |
810 | struct file *file, const struct iovec *iov, |
811 | unsigned long nr_segs) | |
1da177e4 | 812 | { |
ce1d4d3e CH |
813 | struct socket *sock = file->private_data; |
814 | size_t size = 0; | |
815 | int i; | |
1da177e4 | 816 | |
89bddce5 SH |
817 | for (i = 0; i < nr_segs; i++) |
818 | size += iov[i].iov_len; | |
1da177e4 | 819 | |
ce1d4d3e CH |
820 | msg->msg_name = NULL; |
821 | msg->msg_namelen = 0; | |
822 | msg->msg_control = NULL; | |
823 | msg->msg_controllen = 0; | |
89bddce5 | 824 | msg->msg_iov = (struct iovec *)iov; |
ce1d4d3e CH |
825 | msg->msg_iovlen = nr_segs; |
826 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
827 | if (sock->type == SOCK_SEQPACKET) | |
828 | msg->msg_flags |= MSG_EOR; | |
1da177e4 | 829 | |
ce1d4d3e | 830 | return __sock_sendmsg(iocb, sock, msg, size); |
1da177e4 LT |
831 | } |
832 | ||
027445c3 BP |
833 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, |
834 | unsigned long nr_segs, loff_t pos) | |
ce1d4d3e CH |
835 | { |
836 | struct sock_iocb siocb, *x; | |
1da177e4 | 837 | |
ce1d4d3e CH |
838 | if (pos != 0) |
839 | return -ESPIPE; | |
027445c3 | 840 | |
027445c3 | 841 | x = alloc_sock_iocb(iocb, &siocb); |
ce1d4d3e CH |
842 | if (!x) |
843 | return -ENOMEM; | |
1da177e4 | 844 | |
027445c3 | 845 | return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
1da177e4 LT |
846 | } |
847 | ||
1da177e4 LT |
848 | /* |
849 | * Atomic setting of ioctl hooks to avoid race | |
850 | * with module unload. | |
851 | */ | |
852 | ||
4a3e2f71 | 853 | static DEFINE_MUTEX(br_ioctl_mutex); |
881d966b | 854 | static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL; |
1da177e4 | 855 | |
881d966b | 856 | void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) |
1da177e4 | 857 | { |
4a3e2f71 | 858 | mutex_lock(&br_ioctl_mutex); |
1da177e4 | 859 | br_ioctl_hook = hook; |
4a3e2f71 | 860 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 | 861 | } |
89bddce5 | 862 | |
1da177e4 LT |
863 | EXPORT_SYMBOL(brioctl_set); |
864 | ||
4a3e2f71 | 865 | static DEFINE_MUTEX(vlan_ioctl_mutex); |
881d966b | 866 | static int (*vlan_ioctl_hook) (struct net *, void __user *arg); |
1da177e4 | 867 | |
881d966b | 868 | void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) |
1da177e4 | 869 | { |
4a3e2f71 | 870 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 871 | vlan_ioctl_hook = hook; |
4a3e2f71 | 872 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 873 | } |
89bddce5 | 874 | |
1da177e4 LT |
875 | EXPORT_SYMBOL(vlan_ioctl_set); |
876 | ||
4a3e2f71 | 877 | static DEFINE_MUTEX(dlci_ioctl_mutex); |
89bddce5 | 878 | static int (*dlci_ioctl_hook) (unsigned int, void __user *); |
1da177e4 | 879 | |
89bddce5 | 880 | void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) |
1da177e4 | 881 | { |
4a3e2f71 | 882 | mutex_lock(&dlci_ioctl_mutex); |
1da177e4 | 883 | dlci_ioctl_hook = hook; |
4a3e2f71 | 884 | mutex_unlock(&dlci_ioctl_mutex); |
1da177e4 | 885 | } |
89bddce5 | 886 | |
1da177e4 LT |
887 | EXPORT_SYMBOL(dlci_ioctl_set); |
888 | ||
889 | /* | |
890 | * With an ioctl, arg may well be a user mode pointer, but we don't know | |
891 | * what to do with it - that's up to the protocol still. | |
892 | */ | |
893 | ||
894 | static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) | |
895 | { | |
896 | struct socket *sock; | |
881d966b | 897 | struct sock *sk; |
1da177e4 LT |
898 | void __user *argp = (void __user *)arg; |
899 | int pid, err; | |
881d966b | 900 | struct net *net; |
1da177e4 | 901 | |
b69aee04 | 902 | sock = file->private_data; |
881d966b | 903 | sk = sock->sk; |
3b1e0a65 | 904 | net = sock_net(sk); |
1da177e4 | 905 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) { |
881d966b | 906 | err = dev_ioctl(net, cmd, argp); |
1da177e4 | 907 | } else |
d86b5e0e | 908 | #ifdef CONFIG_WIRELESS_EXT |
1da177e4 | 909 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { |
881d966b | 910 | err = dev_ioctl(net, cmd, argp); |
1da177e4 | 911 | } else |
89bddce5 SH |
912 | #endif /* CONFIG_WIRELESS_EXT */ |
913 | switch (cmd) { | |
1da177e4 LT |
914 | case FIOSETOWN: |
915 | case SIOCSPGRP: | |
916 | err = -EFAULT; | |
917 | if (get_user(pid, (int __user *)argp)) | |
918 | break; | |
919 | err = f_setown(sock->file, pid, 1); | |
920 | break; | |
921 | case FIOGETOWN: | |
922 | case SIOCGPGRP: | |
609d7fa9 | 923 | err = put_user(f_getown(sock->file), |
89bddce5 | 924 | (int __user *)argp); |
1da177e4 LT |
925 | break; |
926 | case SIOCGIFBR: | |
927 | case SIOCSIFBR: | |
928 | case SIOCBRADDBR: | |
929 | case SIOCBRDELBR: | |
930 | err = -ENOPKG; | |
931 | if (!br_ioctl_hook) | |
932 | request_module("bridge"); | |
933 | ||
4a3e2f71 | 934 | mutex_lock(&br_ioctl_mutex); |
89bddce5 | 935 | if (br_ioctl_hook) |
881d966b | 936 | err = br_ioctl_hook(net, cmd, argp); |
4a3e2f71 | 937 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 LT |
938 | break; |
939 | case SIOCGIFVLAN: | |
940 | case SIOCSIFVLAN: | |
941 | err = -ENOPKG; | |
942 | if (!vlan_ioctl_hook) | |
943 | request_module("8021q"); | |
944 | ||
4a3e2f71 | 945 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 946 | if (vlan_ioctl_hook) |
881d966b | 947 | err = vlan_ioctl_hook(net, argp); |
4a3e2f71 | 948 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 949 | break; |
1da177e4 LT |
950 | case SIOCADDDLCI: |
951 | case SIOCDELDLCI: | |
952 | err = -ENOPKG; | |
953 | if (!dlci_ioctl_hook) | |
954 | request_module("dlci"); | |
955 | ||
7512cbf6 PE |
956 | mutex_lock(&dlci_ioctl_mutex); |
957 | if (dlci_ioctl_hook) | |
1da177e4 | 958 | err = dlci_ioctl_hook(cmd, argp); |
7512cbf6 | 959 | mutex_unlock(&dlci_ioctl_mutex); |
1da177e4 LT |
960 | break; |
961 | default: | |
962 | err = sock->ops->ioctl(sock, cmd, arg); | |
b5e5fa5e CH |
963 | |
964 | /* | |
965 | * If this ioctl is unknown try to hand it down | |
966 | * to the NIC driver. | |
967 | */ | |
968 | if (err == -ENOIOCTLCMD) | |
881d966b | 969 | err = dev_ioctl(net, cmd, argp); |
1da177e4 | 970 | break; |
89bddce5 | 971 | } |
1da177e4 LT |
972 | return err; |
973 | } | |
974 | ||
975 | int sock_create_lite(int family, int type, int protocol, struct socket **res) | |
976 | { | |
977 | int err; | |
978 | struct socket *sock = NULL; | |
89bddce5 | 979 | |
1da177e4 LT |
980 | err = security_socket_create(family, type, protocol, 1); |
981 | if (err) | |
982 | goto out; | |
983 | ||
984 | sock = sock_alloc(); | |
985 | if (!sock) { | |
986 | err = -ENOMEM; | |
987 | goto out; | |
988 | } | |
989 | ||
1da177e4 | 990 | sock->type = type; |
7420ed23 VY |
991 | err = security_socket_post_create(sock, family, type, protocol, 1); |
992 | if (err) | |
993 | goto out_release; | |
994 | ||
1da177e4 LT |
995 | out: |
996 | *res = sock; | |
997 | return err; | |
7420ed23 VY |
998 | out_release: |
999 | sock_release(sock); | |
1000 | sock = NULL; | |
1001 | goto out; | |
1da177e4 LT |
1002 | } |
1003 | ||
1004 | /* No kernel lock held - perfect */ | |
89bddce5 | 1005 | static unsigned int sock_poll(struct file *file, poll_table *wait) |
1da177e4 LT |
1006 | { |
1007 | struct socket *sock; | |
1008 | ||
1009 | /* | |
89bddce5 | 1010 | * We can't return errors to poll, so it's either yes or no. |
1da177e4 | 1011 | */ |
b69aee04 | 1012 | sock = file->private_data; |
1da177e4 LT |
1013 | return sock->ops->poll(file, sock, wait); |
1014 | } | |
1015 | ||
89bddce5 | 1016 | static int sock_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 1017 | { |
b69aee04 | 1018 | struct socket *sock = file->private_data; |
1da177e4 LT |
1019 | |
1020 | return sock->ops->mmap(file, sock, vma); | |
1021 | } | |
1022 | ||
20380731 | 1023 | static int sock_close(struct inode *inode, struct file *filp) |
1da177e4 LT |
1024 | { |
1025 | /* | |
89bddce5 SH |
1026 | * It was possible the inode is NULL we were |
1027 | * closing an unfinished socket. | |
1da177e4 LT |
1028 | */ |
1029 | ||
89bddce5 | 1030 | if (!inode) { |
1da177e4 LT |
1031 | printk(KERN_DEBUG "sock_close: NULL inode\n"); |
1032 | return 0; | |
1033 | } | |
1da177e4 LT |
1034 | sock_release(SOCKET_I(inode)); |
1035 | return 0; | |
1036 | } | |
1037 | ||
1038 | /* | |
1039 | * Update the socket async list | |
1040 | * | |
1041 | * Fasync_list locking strategy. | |
1042 | * | |
1043 | * 1. fasync_list is modified only under process context socket lock | |
1044 | * i.e. under semaphore. | |
1045 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) | |
1046 | * or under socket lock. | |
1047 | * 3. fasync_list can be used from softirq context, so that | |
1048 | * modification under socket lock have to be enhanced with | |
1049 | * write_lock_bh(&sk->sk_callback_lock). | |
1050 | * --ANK (990710) | |
1051 | */ | |
1052 | ||
1053 | static int sock_fasync(int fd, struct file *filp, int on) | |
1054 | { | |
89bddce5 | 1055 | struct fasync_struct *fa, *fna = NULL, **prev; |
1da177e4 LT |
1056 | struct socket *sock; |
1057 | struct sock *sk; | |
1058 | ||
89bddce5 | 1059 | if (on) { |
8b3a7005 | 1060 | fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL); |
89bddce5 | 1061 | if (fna == NULL) |
1da177e4 LT |
1062 | return -ENOMEM; |
1063 | } | |
1064 | ||
b69aee04 | 1065 | sock = filp->private_data; |
1da177e4 | 1066 | |
89bddce5 SH |
1067 | sk = sock->sk; |
1068 | if (sk == NULL) { | |
1da177e4 LT |
1069 | kfree(fna); |
1070 | return -EINVAL; | |
1071 | } | |
1072 | ||
1073 | lock_sock(sk); | |
1074 | ||
76398425 JC |
1075 | spin_lock(&filp->f_lock); |
1076 | if (on) | |
1077 | filp->f_flags |= FASYNC; | |
1078 | else | |
1079 | filp->f_flags &= ~FASYNC; | |
1080 | spin_unlock(&filp->f_lock); | |
1081 | ||
89bddce5 | 1082 | prev = &(sock->fasync_list); |
1da177e4 | 1083 | |
89bddce5 SH |
1084 | for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev) |
1085 | if (fa->fa_file == filp) | |
1da177e4 LT |
1086 | break; |
1087 | ||
89bddce5 SH |
1088 | if (on) { |
1089 | if (fa != NULL) { | |
1da177e4 | 1090 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1091 | fa->fa_fd = fd; |
1da177e4 LT |
1092 | write_unlock_bh(&sk->sk_callback_lock); |
1093 | ||
1094 | kfree(fna); | |
1095 | goto out; | |
1096 | } | |
89bddce5 SH |
1097 | fna->fa_file = filp; |
1098 | fna->fa_fd = fd; | |
1099 | fna->magic = FASYNC_MAGIC; | |
1100 | fna->fa_next = sock->fasync_list; | |
1da177e4 | 1101 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1102 | sock->fasync_list = fna; |
1da177e4 | 1103 | write_unlock_bh(&sk->sk_callback_lock); |
89bddce5 SH |
1104 | } else { |
1105 | if (fa != NULL) { | |
1da177e4 | 1106 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1107 | *prev = fa->fa_next; |
1da177e4 LT |
1108 | write_unlock_bh(&sk->sk_callback_lock); |
1109 | kfree(fa); | |
1110 | } | |
1111 | } | |
1112 | ||
1113 | out: | |
1114 | release_sock(sock->sk); | |
1115 | return 0; | |
1116 | } | |
1117 | ||
1118 | /* This function may be called only under socket lock or callback_lock */ | |
1119 | ||
1120 | int sock_wake_async(struct socket *sock, int how, int band) | |
1121 | { | |
1122 | if (!sock || !sock->fasync_list) | |
1123 | return -1; | |
89bddce5 | 1124 | switch (how) { |
8d8ad9d7 | 1125 | case SOCK_WAKE_WAITD: |
1da177e4 LT |
1126 | if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags)) |
1127 | break; | |
1128 | goto call_kill; | |
8d8ad9d7 | 1129 | case SOCK_WAKE_SPACE: |
1da177e4 LT |
1130 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags)) |
1131 | break; | |
1132 | /* fall through */ | |
8d8ad9d7 | 1133 | case SOCK_WAKE_IO: |
89bddce5 | 1134 | call_kill: |
1da177e4 LT |
1135 | __kill_fasync(sock->fasync_list, SIGIO, band); |
1136 | break; | |
8d8ad9d7 | 1137 | case SOCK_WAKE_URG: |
1da177e4 LT |
1138 | __kill_fasync(sock->fasync_list, SIGURG, band); |
1139 | } | |
1140 | return 0; | |
1141 | } | |
1142 | ||
1b8d7ae4 | 1143 | static int __sock_create(struct net *net, int family, int type, int protocol, |
89bddce5 | 1144 | struct socket **res, int kern) |
1da177e4 LT |
1145 | { |
1146 | int err; | |
1147 | struct socket *sock; | |
55737fda | 1148 | const struct net_proto_family *pf; |
1da177e4 LT |
1149 | |
1150 | /* | |
89bddce5 | 1151 | * Check protocol is in range |
1da177e4 LT |
1152 | */ |
1153 | if (family < 0 || family >= NPROTO) | |
1154 | return -EAFNOSUPPORT; | |
1155 | if (type < 0 || type >= SOCK_MAX) | |
1156 | return -EINVAL; | |
1157 | ||
1158 | /* Compatibility. | |
1159 | ||
1160 | This uglymoron is moved from INET layer to here to avoid | |
1161 | deadlock in module load. | |
1162 | */ | |
1163 | if (family == PF_INET && type == SOCK_PACKET) { | |
89bddce5 | 1164 | static int warned; |
1da177e4 LT |
1165 | if (!warned) { |
1166 | warned = 1; | |
89bddce5 SH |
1167 | printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", |
1168 | current->comm); | |
1da177e4 LT |
1169 | } |
1170 | family = PF_PACKET; | |
1171 | } | |
1172 | ||
1173 | err = security_socket_create(family, type, protocol, kern); | |
1174 | if (err) | |
1175 | return err; | |
89bddce5 | 1176 | |
55737fda SH |
1177 | /* |
1178 | * Allocate the socket and allow the family to set things up. if | |
1179 | * the protocol is 0, the family is instructed to select an appropriate | |
1180 | * default. | |
1181 | */ | |
1182 | sock = sock_alloc(); | |
1183 | if (!sock) { | |
1184 | if (net_ratelimit()) | |
1185 | printk(KERN_WARNING "socket: no more sockets\n"); | |
1186 | return -ENFILE; /* Not exactly a match, but its the | |
1187 | closest posix thing */ | |
1188 | } | |
1189 | ||
1190 | sock->type = type; | |
1191 | ||
95a5afca | 1192 | #ifdef CONFIG_MODULES |
89bddce5 SH |
1193 | /* Attempt to load a protocol module if the find failed. |
1194 | * | |
1195 | * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user | |
1da177e4 LT |
1196 | * requested real, full-featured networking support upon configuration. |
1197 | * Otherwise module support will break! | |
1198 | */ | |
55737fda | 1199 | if (net_families[family] == NULL) |
89bddce5 | 1200 | request_module("net-pf-%d", family); |
1da177e4 LT |
1201 | #endif |
1202 | ||
55737fda SH |
1203 | rcu_read_lock(); |
1204 | pf = rcu_dereference(net_families[family]); | |
1205 | err = -EAFNOSUPPORT; | |
1206 | if (!pf) | |
1207 | goto out_release; | |
1da177e4 LT |
1208 | |
1209 | /* | |
1210 | * We will call the ->create function, that possibly is in a loadable | |
1211 | * module, so we have to bump that loadable module refcnt first. | |
1212 | */ | |
55737fda | 1213 | if (!try_module_get(pf->owner)) |
1da177e4 LT |
1214 | goto out_release; |
1215 | ||
55737fda SH |
1216 | /* Now protected by module ref count */ |
1217 | rcu_read_unlock(); | |
1218 | ||
1b8d7ae4 | 1219 | err = pf->create(net, sock, protocol); |
55737fda | 1220 | if (err < 0) |
1da177e4 | 1221 | goto out_module_put; |
a79af59e | 1222 | |
1da177e4 LT |
1223 | /* |
1224 | * Now to bump the refcnt of the [loadable] module that owns this | |
1225 | * socket at sock_release time we decrement its refcnt. | |
1226 | */ | |
55737fda SH |
1227 | if (!try_module_get(sock->ops->owner)) |
1228 | goto out_module_busy; | |
1229 | ||
1da177e4 LT |
1230 | /* |
1231 | * Now that we're done with the ->create function, the [loadable] | |
1232 | * module can have its refcnt decremented | |
1233 | */ | |
55737fda | 1234 | module_put(pf->owner); |
7420ed23 VY |
1235 | err = security_socket_post_create(sock, family, type, protocol, kern); |
1236 | if (err) | |
3b185525 | 1237 | goto out_sock_release; |
55737fda | 1238 | *res = sock; |
1da177e4 | 1239 | |
55737fda SH |
1240 | return 0; |
1241 | ||
1242 | out_module_busy: | |
1243 | err = -EAFNOSUPPORT; | |
1da177e4 | 1244 | out_module_put: |
55737fda SH |
1245 | sock->ops = NULL; |
1246 | module_put(pf->owner); | |
1247 | out_sock_release: | |
1da177e4 | 1248 | sock_release(sock); |
55737fda SH |
1249 | return err; |
1250 | ||
1251 | out_release: | |
1252 | rcu_read_unlock(); | |
1253 | goto out_sock_release; | |
1da177e4 LT |
1254 | } |
1255 | ||
1256 | int sock_create(int family, int type, int protocol, struct socket **res) | |
1257 | { | |
1b8d7ae4 | 1258 | return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); |
1da177e4 LT |
1259 | } |
1260 | ||
1261 | int sock_create_kern(int family, int type, int protocol, struct socket **res) | |
1262 | { | |
1b8d7ae4 | 1263 | return __sock_create(&init_net, family, type, protocol, res, 1); |
1da177e4 LT |
1264 | } |
1265 | ||
3e0fa65f | 1266 | SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) |
1da177e4 LT |
1267 | { |
1268 | int retval; | |
1269 | struct socket *sock; | |
a677a039 UD |
1270 | int flags; |
1271 | ||
e38b36f3 UD |
1272 | /* Check the SOCK_* constants for consistency. */ |
1273 | BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); | |
1274 | BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); | |
1275 | BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); | |
1276 | BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); | |
1277 | ||
a677a039 | 1278 | flags = type & ~SOCK_TYPE_MASK; |
77d27200 | 1279 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
a677a039 UD |
1280 | return -EINVAL; |
1281 | type &= SOCK_TYPE_MASK; | |
1da177e4 | 1282 | |
aaca0bdc UD |
1283 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1284 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1285 | ||
1da177e4 LT |
1286 | retval = sock_create(family, type, protocol, &sock); |
1287 | if (retval < 0) | |
1288 | goto out; | |
1289 | ||
77d27200 | 1290 | retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); |
1da177e4 LT |
1291 | if (retval < 0) |
1292 | goto out_release; | |
1293 | ||
1294 | out: | |
1295 | /* It may be already another descriptor 8) Not kernel problem. */ | |
1296 | return retval; | |
1297 | ||
1298 | out_release: | |
1299 | sock_release(sock); | |
1300 | return retval; | |
1301 | } | |
1302 | ||
1303 | /* | |
1304 | * Create a pair of connected sockets. | |
1305 | */ | |
1306 | ||
3e0fa65f HC |
1307 | SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, |
1308 | int __user *, usockvec) | |
1da177e4 LT |
1309 | { |
1310 | struct socket *sock1, *sock2; | |
1311 | int fd1, fd2, err; | |
db349509 | 1312 | struct file *newfile1, *newfile2; |
a677a039 UD |
1313 | int flags; |
1314 | ||
1315 | flags = type & ~SOCK_TYPE_MASK; | |
77d27200 | 1316 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
a677a039 UD |
1317 | return -EINVAL; |
1318 | type &= SOCK_TYPE_MASK; | |
1da177e4 | 1319 | |
aaca0bdc UD |
1320 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1321 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1322 | ||
1da177e4 LT |
1323 | /* |
1324 | * Obtain the first socket and check if the underlying protocol | |
1325 | * supports the socketpair call. | |
1326 | */ | |
1327 | ||
1328 | err = sock_create(family, type, protocol, &sock1); | |
1329 | if (err < 0) | |
1330 | goto out; | |
1331 | ||
1332 | err = sock_create(family, type, protocol, &sock2); | |
1333 | if (err < 0) | |
1334 | goto out_release_1; | |
1335 | ||
1336 | err = sock1->ops->socketpair(sock1, sock2); | |
89bddce5 | 1337 | if (err < 0) |
1da177e4 LT |
1338 | goto out_release_both; |
1339 | ||
a677a039 | 1340 | fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC); |
bf3c23d1 DM |
1341 | if (unlikely(fd1 < 0)) { |
1342 | err = fd1; | |
db349509 | 1343 | goto out_release_both; |
bf3c23d1 | 1344 | } |
1da177e4 | 1345 | |
a677a039 | 1346 | fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC); |
db349509 | 1347 | if (unlikely(fd2 < 0)) { |
bf3c23d1 | 1348 | err = fd2; |
db349509 AV |
1349 | put_filp(newfile1); |
1350 | put_unused_fd(fd1); | |
1da177e4 | 1351 | goto out_release_both; |
db349509 | 1352 | } |
1da177e4 | 1353 | |
77d27200 | 1354 | err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK); |
db349509 AV |
1355 | if (unlikely(err < 0)) { |
1356 | goto out_fd2; | |
1357 | } | |
1358 | ||
77d27200 | 1359 | err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK); |
db349509 AV |
1360 | if (unlikely(err < 0)) { |
1361 | fput(newfile1); | |
1362 | goto out_fd1; | |
1363 | } | |
1364 | ||
157cf649 | 1365 | audit_fd_pair(fd1, fd2); |
db349509 AV |
1366 | fd_install(fd1, newfile1); |
1367 | fd_install(fd2, newfile2); | |
1da177e4 LT |
1368 | /* fd1 and fd2 may be already another descriptors. |
1369 | * Not kernel problem. | |
1370 | */ | |
1371 | ||
89bddce5 | 1372 | err = put_user(fd1, &usockvec[0]); |
1da177e4 LT |
1373 | if (!err) |
1374 | err = put_user(fd2, &usockvec[1]); | |
1375 | if (!err) | |
1376 | return 0; | |
1377 | ||
1378 | sys_close(fd2); | |
1379 | sys_close(fd1); | |
1380 | return err; | |
1381 | ||
1da177e4 | 1382 | out_release_both: |
89bddce5 | 1383 | sock_release(sock2); |
1da177e4 | 1384 | out_release_1: |
89bddce5 | 1385 | sock_release(sock1); |
1da177e4 LT |
1386 | out: |
1387 | return err; | |
db349509 AV |
1388 | |
1389 | out_fd2: | |
1390 | put_filp(newfile1); | |
1391 | sock_release(sock1); | |
1392 | out_fd1: | |
1393 | put_filp(newfile2); | |
1394 | sock_release(sock2); | |
db349509 AV |
1395 | put_unused_fd(fd1); |
1396 | put_unused_fd(fd2); | |
1397 | goto out; | |
1da177e4 LT |
1398 | } |
1399 | ||
1da177e4 LT |
1400 | /* |
1401 | * Bind a name to a socket. Nothing much to do here since it's | |
1402 | * the protocol's responsibility to handle the local address. | |
1403 | * | |
1404 | * We move the socket address to kernel space before we call | |
1405 | * the protocol layer (having also checked the address is ok). | |
1406 | */ | |
1407 | ||
20f37034 | 1408 | SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) |
1da177e4 LT |
1409 | { |
1410 | struct socket *sock; | |
230b1839 | 1411 | struct sockaddr_storage address; |
6cb153ca | 1412 | int err, fput_needed; |
1da177e4 | 1413 | |
89bddce5 | 1414 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
e71a4783 | 1415 | if (sock) { |
230b1839 | 1416 | err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address); |
89bddce5 SH |
1417 | if (err >= 0) { |
1418 | err = security_socket_bind(sock, | |
230b1839 | 1419 | (struct sockaddr *)&address, |
89bddce5 | 1420 | addrlen); |
6cb153ca BL |
1421 | if (!err) |
1422 | err = sock->ops->bind(sock, | |
89bddce5 | 1423 | (struct sockaddr *) |
230b1839 | 1424 | &address, addrlen); |
1da177e4 | 1425 | } |
6cb153ca | 1426 | fput_light(sock->file, fput_needed); |
89bddce5 | 1427 | } |
1da177e4 LT |
1428 | return err; |
1429 | } | |
1430 | ||
1da177e4 LT |
1431 | /* |
1432 | * Perform a listen. Basically, we allow the protocol to do anything | |
1433 | * necessary for a listen, and if that works, we mark the socket as | |
1434 | * ready for listening. | |
1435 | */ | |
1436 | ||
3e0fa65f | 1437 | SYSCALL_DEFINE2(listen, int, fd, int, backlog) |
1da177e4 LT |
1438 | { |
1439 | struct socket *sock; | |
6cb153ca | 1440 | int err, fput_needed; |
b8e1f9b5 | 1441 | int somaxconn; |
89bddce5 SH |
1442 | |
1443 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1444 | if (sock) { | |
8efa6e93 | 1445 | somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; |
b8e1f9b5 PE |
1446 | if ((unsigned)backlog > somaxconn) |
1447 | backlog = somaxconn; | |
1da177e4 LT |
1448 | |
1449 | err = security_socket_listen(sock, backlog); | |
6cb153ca BL |
1450 | if (!err) |
1451 | err = sock->ops->listen(sock, backlog); | |
1da177e4 | 1452 | |
6cb153ca | 1453 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1454 | } |
1455 | return err; | |
1456 | } | |
1457 | ||
1da177e4 LT |
1458 | /* |
1459 | * For accept, we attempt to create a new socket, set up the link | |
1460 | * with the client, wake up the client, then return the new | |
1461 | * connected fd. We collect the address of the connector in kernel | |
1462 | * space and move it to user at the very end. This is unclean because | |
1463 | * we open the socket then return an error. | |
1464 | * | |
1465 | * 1003.1g adds the ability to recvmsg() to query connection pending | |
1466 | * status to recvmsg. We need to add that support in a way thats | |
1467 | * clean when we restucture accept also. | |
1468 | */ | |
1469 | ||
20f37034 HC |
1470 | SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, |
1471 | int __user *, upeer_addrlen, int, flags) | |
1da177e4 LT |
1472 | { |
1473 | struct socket *sock, *newsock; | |
39d8c1b6 | 1474 | struct file *newfile; |
6cb153ca | 1475 | int err, len, newfd, fput_needed; |
230b1839 | 1476 | struct sockaddr_storage address; |
1da177e4 | 1477 | |
77d27200 | 1478 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
aaca0bdc UD |
1479 | return -EINVAL; |
1480 | ||
1481 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) | |
1482 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1483 | ||
6cb153ca | 1484 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1485 | if (!sock) |
1486 | goto out; | |
1487 | ||
1488 | err = -ENFILE; | |
89bddce5 | 1489 | if (!(newsock = sock_alloc())) |
1da177e4 LT |
1490 | goto out_put; |
1491 | ||
1492 | newsock->type = sock->type; | |
1493 | newsock->ops = sock->ops; | |
1494 | ||
1da177e4 LT |
1495 | /* |
1496 | * We don't need try_module_get here, as the listening socket (sock) | |
1497 | * has the protocol module (sock->ops->owner) held. | |
1498 | */ | |
1499 | __module_get(newsock->ops->owner); | |
1500 | ||
aaca0bdc | 1501 | newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC); |
39d8c1b6 DM |
1502 | if (unlikely(newfd < 0)) { |
1503 | err = newfd; | |
9a1875e6 DM |
1504 | sock_release(newsock); |
1505 | goto out_put; | |
39d8c1b6 DM |
1506 | } |
1507 | ||
77d27200 | 1508 | err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK); |
39d8c1b6 | 1509 | if (err < 0) |
79f4f642 | 1510 | goto out_fd_simple; |
39d8c1b6 | 1511 | |
a79af59e FF |
1512 | err = security_socket_accept(sock, newsock); |
1513 | if (err) | |
39d8c1b6 | 1514 | goto out_fd; |
a79af59e | 1515 | |
1da177e4 LT |
1516 | err = sock->ops->accept(sock, newsock, sock->file->f_flags); |
1517 | if (err < 0) | |
39d8c1b6 | 1518 | goto out_fd; |
1da177e4 LT |
1519 | |
1520 | if (upeer_sockaddr) { | |
230b1839 | 1521 | if (newsock->ops->getname(newsock, (struct sockaddr *)&address, |
89bddce5 | 1522 | &len, 2) < 0) { |
1da177e4 | 1523 | err = -ECONNABORTED; |
39d8c1b6 | 1524 | goto out_fd; |
1da177e4 | 1525 | } |
230b1839 YH |
1526 | err = move_addr_to_user((struct sockaddr *)&address, |
1527 | len, upeer_sockaddr, upeer_addrlen); | |
1da177e4 | 1528 | if (err < 0) |
39d8c1b6 | 1529 | goto out_fd; |
1da177e4 LT |
1530 | } |
1531 | ||
1532 | /* File flags are not inherited via accept() unlike another OSes. */ | |
1533 | ||
39d8c1b6 DM |
1534 | fd_install(newfd, newfile); |
1535 | err = newfd; | |
1da177e4 | 1536 | |
1da177e4 | 1537 | out_put: |
6cb153ca | 1538 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1539 | out: |
1540 | return err; | |
79f4f642 AD |
1541 | out_fd_simple: |
1542 | sock_release(newsock); | |
1543 | put_filp(newfile); | |
1544 | put_unused_fd(newfd); | |
1545 | goto out_put; | |
39d8c1b6 | 1546 | out_fd: |
9606a216 | 1547 | fput(newfile); |
39d8c1b6 | 1548 | put_unused_fd(newfd); |
1da177e4 LT |
1549 | goto out_put; |
1550 | } | |
1551 | ||
20f37034 HC |
1552 | SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, |
1553 | int __user *, upeer_addrlen) | |
aaca0bdc | 1554 | { |
de11defe | 1555 | return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); |
aaca0bdc UD |
1556 | } |
1557 | ||
1da177e4 LT |
1558 | /* |
1559 | * Attempt to connect to a socket with the server address. The address | |
1560 | * is in user space so we verify it is OK and move it to kernel space. | |
1561 | * | |
1562 | * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to | |
1563 | * break bindings | |
1564 | * | |
1565 | * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and | |
1566 | * other SEQPACKET protocols that take time to connect() as it doesn't | |
1567 | * include the -EINPROGRESS status for such sockets. | |
1568 | */ | |
1569 | ||
20f37034 HC |
1570 | SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, |
1571 | int, addrlen) | |
1da177e4 LT |
1572 | { |
1573 | struct socket *sock; | |
230b1839 | 1574 | struct sockaddr_storage address; |
6cb153ca | 1575 | int err, fput_needed; |
1da177e4 | 1576 | |
6cb153ca | 1577 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1578 | if (!sock) |
1579 | goto out; | |
230b1839 | 1580 | err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address); |
1da177e4 LT |
1581 | if (err < 0) |
1582 | goto out_put; | |
1583 | ||
89bddce5 | 1584 | err = |
230b1839 | 1585 | security_socket_connect(sock, (struct sockaddr *)&address, addrlen); |
1da177e4 LT |
1586 | if (err) |
1587 | goto out_put; | |
1588 | ||
230b1839 | 1589 | err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, |
1da177e4 LT |
1590 | sock->file->f_flags); |
1591 | out_put: | |
6cb153ca | 1592 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1593 | out: |
1594 | return err; | |
1595 | } | |
1596 | ||
1597 | /* | |
1598 | * Get the local address ('name') of a socket object. Move the obtained | |
1599 | * name to user space. | |
1600 | */ | |
1601 | ||
20f37034 HC |
1602 | SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, |
1603 | int __user *, usockaddr_len) | |
1da177e4 LT |
1604 | { |
1605 | struct socket *sock; | |
230b1839 | 1606 | struct sockaddr_storage address; |
6cb153ca | 1607 | int len, err, fput_needed; |
89bddce5 | 1608 | |
6cb153ca | 1609 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1610 | if (!sock) |
1611 | goto out; | |
1612 | ||
1613 | err = security_socket_getsockname(sock); | |
1614 | if (err) | |
1615 | goto out_put; | |
1616 | ||
230b1839 | 1617 | err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0); |
1da177e4 LT |
1618 | if (err) |
1619 | goto out_put; | |
230b1839 | 1620 | err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len); |
1da177e4 LT |
1621 | |
1622 | out_put: | |
6cb153ca | 1623 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1624 | out: |
1625 | return err; | |
1626 | } | |
1627 | ||
1628 | /* | |
1629 | * Get the remote address ('name') of a socket object. Move the obtained | |
1630 | * name to user space. | |
1631 | */ | |
1632 | ||
20f37034 HC |
1633 | SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, |
1634 | int __user *, usockaddr_len) | |
1da177e4 LT |
1635 | { |
1636 | struct socket *sock; | |
230b1839 | 1637 | struct sockaddr_storage address; |
6cb153ca | 1638 | int len, err, fput_needed; |
1da177e4 | 1639 | |
89bddce5 SH |
1640 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1641 | if (sock != NULL) { | |
1da177e4 LT |
1642 | err = security_socket_getpeername(sock); |
1643 | if (err) { | |
6cb153ca | 1644 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1645 | return err; |
1646 | } | |
1647 | ||
89bddce5 | 1648 | err = |
230b1839 | 1649 | sock->ops->getname(sock, (struct sockaddr *)&address, &len, |
89bddce5 | 1650 | 1); |
1da177e4 | 1651 | if (!err) |
230b1839 | 1652 | err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, |
89bddce5 | 1653 | usockaddr_len); |
6cb153ca | 1654 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1655 | } |
1656 | return err; | |
1657 | } | |
1658 | ||
1659 | /* | |
1660 | * Send a datagram to a given address. We move the address into kernel | |
1661 | * space and check the user space data area is readable before invoking | |
1662 | * the protocol. | |
1663 | */ | |
1664 | ||
3e0fa65f HC |
1665 | SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, |
1666 | unsigned, flags, struct sockaddr __user *, addr, | |
1667 | int, addr_len) | |
1da177e4 LT |
1668 | { |
1669 | struct socket *sock; | |
230b1839 | 1670 | struct sockaddr_storage address; |
1da177e4 LT |
1671 | int err; |
1672 | struct msghdr msg; | |
1673 | struct iovec iov; | |
6cb153ca | 1674 | int fput_needed; |
6cb153ca | 1675 | |
de0fa95c PE |
1676 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1677 | if (!sock) | |
4387ff75 | 1678 | goto out; |
6cb153ca | 1679 | |
89bddce5 SH |
1680 | iov.iov_base = buff; |
1681 | iov.iov_len = len; | |
1682 | msg.msg_name = NULL; | |
1683 | msg.msg_iov = &iov; | |
1684 | msg.msg_iovlen = 1; | |
1685 | msg.msg_control = NULL; | |
1686 | msg.msg_controllen = 0; | |
1687 | msg.msg_namelen = 0; | |
6cb153ca | 1688 | if (addr) { |
230b1839 | 1689 | err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address); |
1da177e4 LT |
1690 | if (err < 0) |
1691 | goto out_put; | |
230b1839 | 1692 | msg.msg_name = (struct sockaddr *)&address; |
89bddce5 | 1693 | msg.msg_namelen = addr_len; |
1da177e4 LT |
1694 | } |
1695 | if (sock->file->f_flags & O_NONBLOCK) | |
1696 | flags |= MSG_DONTWAIT; | |
1697 | msg.msg_flags = flags; | |
1698 | err = sock_sendmsg(sock, &msg, len); | |
1699 | ||
89bddce5 | 1700 | out_put: |
de0fa95c | 1701 | fput_light(sock->file, fput_needed); |
4387ff75 | 1702 | out: |
1da177e4 LT |
1703 | return err; |
1704 | } | |
1705 | ||
1706 | /* | |
89bddce5 | 1707 | * Send a datagram down a socket. |
1da177e4 LT |
1708 | */ |
1709 | ||
3e0fa65f HC |
1710 | SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, |
1711 | unsigned, flags) | |
1da177e4 LT |
1712 | { |
1713 | return sys_sendto(fd, buff, len, flags, NULL, 0); | |
1714 | } | |
1715 | ||
1716 | /* | |
89bddce5 | 1717 | * Receive a frame from the socket and optionally record the address of the |
1da177e4 LT |
1718 | * sender. We verify the buffers are writable and if needed move the |
1719 | * sender address from kernel to user space. | |
1720 | */ | |
1721 | ||
3e0fa65f HC |
1722 | SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, |
1723 | unsigned, flags, struct sockaddr __user *, addr, | |
1724 | int __user *, addr_len) | |
1da177e4 LT |
1725 | { |
1726 | struct socket *sock; | |
1727 | struct iovec iov; | |
1728 | struct msghdr msg; | |
230b1839 | 1729 | struct sockaddr_storage address; |
89bddce5 | 1730 | int err, err2; |
6cb153ca BL |
1731 | int fput_needed; |
1732 | ||
de0fa95c | 1733 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 | 1734 | if (!sock) |
de0fa95c | 1735 | goto out; |
1da177e4 | 1736 | |
89bddce5 SH |
1737 | msg.msg_control = NULL; |
1738 | msg.msg_controllen = 0; | |
1739 | msg.msg_iovlen = 1; | |
1740 | msg.msg_iov = &iov; | |
1741 | iov.iov_len = size; | |
1742 | iov.iov_base = ubuf; | |
230b1839 YH |
1743 | msg.msg_name = (struct sockaddr *)&address; |
1744 | msg.msg_namelen = sizeof(address); | |
1da177e4 LT |
1745 | if (sock->file->f_flags & O_NONBLOCK) |
1746 | flags |= MSG_DONTWAIT; | |
89bddce5 | 1747 | err = sock_recvmsg(sock, &msg, size, flags); |
1da177e4 | 1748 | |
89bddce5 | 1749 | if (err >= 0 && addr != NULL) { |
230b1839 YH |
1750 | err2 = move_addr_to_user((struct sockaddr *)&address, |
1751 | msg.msg_namelen, addr, addr_len); | |
89bddce5 SH |
1752 | if (err2 < 0) |
1753 | err = err2; | |
1da177e4 | 1754 | } |
de0fa95c PE |
1755 | |
1756 | fput_light(sock->file, fput_needed); | |
4387ff75 | 1757 | out: |
1da177e4 LT |
1758 | return err; |
1759 | } | |
1760 | ||
1761 | /* | |
89bddce5 | 1762 | * Receive a datagram from a socket. |
1da177e4 LT |
1763 | */ |
1764 | ||
89bddce5 SH |
1765 | asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size, |
1766 | unsigned flags) | |
1da177e4 LT |
1767 | { |
1768 | return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); | |
1769 | } | |
1770 | ||
1771 | /* | |
1772 | * Set a socket option. Because we don't know the option lengths we have | |
1773 | * to pass the user mode parameter for the protocols to sort out. | |
1774 | */ | |
1775 | ||
20f37034 HC |
1776 | SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, |
1777 | char __user *, optval, int, optlen) | |
1da177e4 | 1778 | { |
6cb153ca | 1779 | int err, fput_needed; |
1da177e4 LT |
1780 | struct socket *sock; |
1781 | ||
1782 | if (optlen < 0) | |
1783 | return -EINVAL; | |
89bddce5 SH |
1784 | |
1785 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1786 | if (sock != NULL) { | |
1787 | err = security_socket_setsockopt(sock, level, optname); | |
6cb153ca BL |
1788 | if (err) |
1789 | goto out_put; | |
1da177e4 LT |
1790 | |
1791 | if (level == SOL_SOCKET) | |
89bddce5 SH |
1792 | err = |
1793 | sock_setsockopt(sock, level, optname, optval, | |
1794 | optlen); | |
1da177e4 | 1795 | else |
89bddce5 SH |
1796 | err = |
1797 | sock->ops->setsockopt(sock, level, optname, optval, | |
1798 | optlen); | |
6cb153ca BL |
1799 | out_put: |
1800 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1801 | } |
1802 | return err; | |
1803 | } | |
1804 | ||
1805 | /* | |
1806 | * Get a socket option. Because we don't know the option lengths we have | |
1807 | * to pass a user mode parameter for the protocols to sort out. | |
1808 | */ | |
1809 | ||
20f37034 HC |
1810 | SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, |
1811 | char __user *, optval, int __user *, optlen) | |
1da177e4 | 1812 | { |
6cb153ca | 1813 | int err, fput_needed; |
1da177e4 LT |
1814 | struct socket *sock; |
1815 | ||
89bddce5 SH |
1816 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1817 | if (sock != NULL) { | |
6cb153ca BL |
1818 | err = security_socket_getsockopt(sock, level, optname); |
1819 | if (err) | |
1820 | goto out_put; | |
1da177e4 LT |
1821 | |
1822 | if (level == SOL_SOCKET) | |
89bddce5 SH |
1823 | err = |
1824 | sock_getsockopt(sock, level, optname, optval, | |
1825 | optlen); | |
1da177e4 | 1826 | else |
89bddce5 SH |
1827 | err = |
1828 | sock->ops->getsockopt(sock, level, optname, optval, | |
1829 | optlen); | |
6cb153ca BL |
1830 | out_put: |
1831 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1832 | } |
1833 | return err; | |
1834 | } | |
1835 | ||
1da177e4 LT |
1836 | /* |
1837 | * Shutdown a socket. | |
1838 | */ | |
1839 | ||
754fe8d2 | 1840 | SYSCALL_DEFINE2(shutdown, int, fd, int, how) |
1da177e4 | 1841 | { |
6cb153ca | 1842 | int err, fput_needed; |
1da177e4 LT |
1843 | struct socket *sock; |
1844 | ||
89bddce5 SH |
1845 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1846 | if (sock != NULL) { | |
1da177e4 | 1847 | err = security_socket_shutdown(sock, how); |
6cb153ca BL |
1848 | if (!err) |
1849 | err = sock->ops->shutdown(sock, how); | |
1850 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1851 | } |
1852 | return err; | |
1853 | } | |
1854 | ||
89bddce5 | 1855 | /* A couple of helpful macros for getting the address of the 32/64 bit |
1da177e4 LT |
1856 | * fields which are the same type (int / unsigned) on our platforms. |
1857 | */ | |
1858 | #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) | |
1859 | #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) | |
1860 | #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) | |
1861 | ||
1da177e4 LT |
1862 | /* |
1863 | * BSD sendmsg interface | |
1864 | */ | |
1865 | ||
3e0fa65f | 1866 | SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags) |
1da177e4 | 1867 | { |
89bddce5 SH |
1868 | struct compat_msghdr __user *msg_compat = |
1869 | (struct compat_msghdr __user *)msg; | |
1da177e4 | 1870 | struct socket *sock; |
230b1839 | 1871 | struct sockaddr_storage address; |
1da177e4 | 1872 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; |
b9d717a7 | 1873 | unsigned char ctl[sizeof(struct cmsghdr) + 20] |
89bddce5 SH |
1874 | __attribute__ ((aligned(sizeof(__kernel_size_t)))); |
1875 | /* 20 is size of ipv6_pktinfo */ | |
1da177e4 LT |
1876 | unsigned char *ctl_buf = ctl; |
1877 | struct msghdr msg_sys; | |
1878 | int err, ctl_len, iov_size, total_len; | |
6cb153ca | 1879 | int fput_needed; |
89bddce5 | 1880 | |
1da177e4 LT |
1881 | err = -EFAULT; |
1882 | if (MSG_CMSG_COMPAT & flags) { | |
1883 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1884 | return -EFAULT; | |
89bddce5 SH |
1885 | } |
1886 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1da177e4 LT |
1887 | return -EFAULT; |
1888 | ||
6cb153ca | 1889 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
89bddce5 | 1890 | if (!sock) |
1da177e4 LT |
1891 | goto out; |
1892 | ||
1893 | /* do not move before msg_sys is valid */ | |
1894 | err = -EMSGSIZE; | |
1895 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
1896 | goto out_put; | |
1897 | ||
89bddce5 | 1898 | /* Check whether to allocate the iovec area */ |
1da177e4 LT |
1899 | err = -ENOMEM; |
1900 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
1901 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
1902 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
1903 | if (!iov) | |
1904 | goto out_put; | |
1905 | } | |
1906 | ||
1907 | /* This will also move the address data into kernel space */ | |
1908 | if (MSG_CMSG_COMPAT & flags) { | |
230b1839 YH |
1909 | err = verify_compat_iovec(&msg_sys, iov, |
1910 | (struct sockaddr *)&address, | |
1911 | VERIFY_READ); | |
1da177e4 | 1912 | } else |
230b1839 YH |
1913 | err = verify_iovec(&msg_sys, iov, |
1914 | (struct sockaddr *)&address, | |
1915 | VERIFY_READ); | |
89bddce5 | 1916 | if (err < 0) |
1da177e4 LT |
1917 | goto out_freeiov; |
1918 | total_len = err; | |
1919 | ||
1920 | err = -ENOBUFS; | |
1921 | ||
1922 | if (msg_sys.msg_controllen > INT_MAX) | |
1923 | goto out_freeiov; | |
89bddce5 | 1924 | ctl_len = msg_sys.msg_controllen; |
1da177e4 | 1925 | if ((MSG_CMSG_COMPAT & flags) && ctl_len) { |
89bddce5 SH |
1926 | err = |
1927 | cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, | |
1928 | sizeof(ctl)); | |
1da177e4 LT |
1929 | if (err) |
1930 | goto out_freeiov; | |
1931 | ctl_buf = msg_sys.msg_control; | |
8920e8f9 | 1932 | ctl_len = msg_sys.msg_controllen; |
1da177e4 | 1933 | } else if (ctl_len) { |
89bddce5 | 1934 | if (ctl_len > sizeof(ctl)) { |
1da177e4 | 1935 | ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); |
89bddce5 | 1936 | if (ctl_buf == NULL) |
1da177e4 LT |
1937 | goto out_freeiov; |
1938 | } | |
1939 | err = -EFAULT; | |
1940 | /* | |
1941 | * Careful! Before this, msg_sys.msg_control contains a user pointer. | |
1942 | * Afterwards, it will be a kernel pointer. Thus the compiler-assisted | |
1943 | * checking falls down on this. | |
1944 | */ | |
89bddce5 SH |
1945 | if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control, |
1946 | ctl_len)) | |
1da177e4 LT |
1947 | goto out_freectl; |
1948 | msg_sys.msg_control = ctl_buf; | |
1949 | } | |
1950 | msg_sys.msg_flags = flags; | |
1951 | ||
1952 | if (sock->file->f_flags & O_NONBLOCK) | |
1953 | msg_sys.msg_flags |= MSG_DONTWAIT; | |
1954 | err = sock_sendmsg(sock, &msg_sys, total_len); | |
1955 | ||
1956 | out_freectl: | |
89bddce5 | 1957 | if (ctl_buf != ctl) |
1da177e4 LT |
1958 | sock_kfree_s(sock->sk, ctl_buf, ctl_len); |
1959 | out_freeiov: | |
1960 | if (iov != iovstack) | |
1961 | sock_kfree_s(sock->sk, iov, iov_size); | |
1962 | out_put: | |
6cb153ca | 1963 | fput_light(sock->file, fput_needed); |
89bddce5 | 1964 | out: |
1da177e4 LT |
1965 | return err; |
1966 | } | |
1967 | ||
1968 | /* | |
1969 | * BSD recvmsg interface | |
1970 | */ | |
1971 | ||
3e0fa65f HC |
1972 | SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg, |
1973 | unsigned int, flags) | |
1da177e4 | 1974 | { |
89bddce5 SH |
1975 | struct compat_msghdr __user *msg_compat = |
1976 | (struct compat_msghdr __user *)msg; | |
1da177e4 LT |
1977 | struct socket *sock; |
1978 | struct iovec iovstack[UIO_FASTIOV]; | |
89bddce5 | 1979 | struct iovec *iov = iovstack; |
1da177e4 LT |
1980 | struct msghdr msg_sys; |
1981 | unsigned long cmsg_ptr; | |
1982 | int err, iov_size, total_len, len; | |
6cb153ca | 1983 | int fput_needed; |
1da177e4 LT |
1984 | |
1985 | /* kernel mode address */ | |
230b1839 | 1986 | struct sockaddr_storage addr; |
1da177e4 LT |
1987 | |
1988 | /* user mode address pointers */ | |
1989 | struct sockaddr __user *uaddr; | |
1990 | int __user *uaddr_len; | |
89bddce5 | 1991 | |
1da177e4 LT |
1992 | if (MSG_CMSG_COMPAT & flags) { |
1993 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1994 | return -EFAULT; | |
89bddce5 SH |
1995 | } |
1996 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1997 | return -EFAULT; | |
1da177e4 | 1998 | |
6cb153ca | 1999 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
2000 | if (!sock) |
2001 | goto out; | |
2002 | ||
2003 | err = -EMSGSIZE; | |
2004 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
2005 | goto out_put; | |
89bddce5 SH |
2006 | |
2007 | /* Check whether to allocate the iovec area */ | |
1da177e4 LT |
2008 | err = -ENOMEM; |
2009 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
2010 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
2011 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
2012 | if (!iov) | |
2013 | goto out_put; | |
2014 | } | |
2015 | ||
2016 | /* | |
89bddce5 SH |
2017 | * Save the user-mode address (verify_iovec will change the |
2018 | * kernel msghdr to use the kernel address space) | |
1da177e4 | 2019 | */ |
89bddce5 | 2020 | |
cfcabdcc | 2021 | uaddr = (__force void __user *)msg_sys.msg_name; |
1da177e4 LT |
2022 | uaddr_len = COMPAT_NAMELEN(msg); |
2023 | if (MSG_CMSG_COMPAT & flags) { | |
230b1839 YH |
2024 | err = verify_compat_iovec(&msg_sys, iov, |
2025 | (struct sockaddr *)&addr, | |
2026 | VERIFY_WRITE); | |
1da177e4 | 2027 | } else |
230b1839 YH |
2028 | err = verify_iovec(&msg_sys, iov, |
2029 | (struct sockaddr *)&addr, | |
2030 | VERIFY_WRITE); | |
1da177e4 LT |
2031 | if (err < 0) |
2032 | goto out_freeiov; | |
89bddce5 | 2033 | total_len = err; |
1da177e4 LT |
2034 | |
2035 | cmsg_ptr = (unsigned long)msg_sys.msg_control; | |
4a19542e | 2036 | msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); |
89bddce5 | 2037 | |
1da177e4 LT |
2038 | if (sock->file->f_flags & O_NONBLOCK) |
2039 | flags |= MSG_DONTWAIT; | |
2040 | err = sock_recvmsg(sock, &msg_sys, total_len, flags); | |
2041 | if (err < 0) | |
2042 | goto out_freeiov; | |
2043 | len = err; | |
2044 | ||
2045 | if (uaddr != NULL) { | |
230b1839 YH |
2046 | err = move_addr_to_user((struct sockaddr *)&addr, |
2047 | msg_sys.msg_namelen, uaddr, | |
89bddce5 | 2048 | uaddr_len); |
1da177e4 LT |
2049 | if (err < 0) |
2050 | goto out_freeiov; | |
2051 | } | |
37f7f421 DM |
2052 | err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT), |
2053 | COMPAT_FLAGS(msg)); | |
1da177e4 LT |
2054 | if (err) |
2055 | goto out_freeiov; | |
2056 | if (MSG_CMSG_COMPAT & flags) | |
89bddce5 | 2057 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, |
1da177e4 LT |
2058 | &msg_compat->msg_controllen); |
2059 | else | |
89bddce5 | 2060 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, |
1da177e4 LT |
2061 | &msg->msg_controllen); |
2062 | if (err) | |
2063 | goto out_freeiov; | |
2064 | err = len; | |
2065 | ||
2066 | out_freeiov: | |
2067 | if (iov != iovstack) | |
2068 | sock_kfree_s(sock->sk, iov, iov_size); | |
2069 | out_put: | |
6cb153ca | 2070 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2071 | out: |
2072 | return err; | |
2073 | } | |
2074 | ||
2075 | #ifdef __ARCH_WANT_SYS_SOCKETCALL | |
2076 | ||
2077 | /* Argument list sizes for sys_socketcall */ | |
2078 | #define AL(x) ((x) * sizeof(unsigned long)) | |
aaca0bdc | 2079 | static const unsigned char nargs[19]={ |
89bddce5 SH |
2080 | AL(0),AL(3),AL(3),AL(3),AL(2),AL(3), |
2081 | AL(3),AL(3),AL(4),AL(4),AL(4),AL(6), | |
aaca0bdc | 2082 | AL(6),AL(2),AL(5),AL(5),AL(3),AL(3), |
de11defe | 2083 | AL(4) |
89bddce5 SH |
2084 | }; |
2085 | ||
1da177e4 LT |
2086 | #undef AL |
2087 | ||
2088 | /* | |
89bddce5 | 2089 | * System call vectors. |
1da177e4 LT |
2090 | * |
2091 | * Argument checking cleaned up. Saved 20% in size. | |
2092 | * This function doesn't need to set the kernel lock because | |
89bddce5 | 2093 | * it is set by the callees. |
1da177e4 LT |
2094 | */ |
2095 | ||
3e0fa65f | 2096 | SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) |
1da177e4 LT |
2097 | { |
2098 | unsigned long a[6]; | |
89bddce5 | 2099 | unsigned long a0, a1; |
1da177e4 | 2100 | int err; |
47379052 | 2101 | unsigned int len; |
1da177e4 | 2102 | |
de11defe | 2103 | if (call < 1 || call > SYS_ACCEPT4) |
1da177e4 LT |
2104 | return -EINVAL; |
2105 | ||
47379052 AV |
2106 | len = nargs[call]; |
2107 | if (len > sizeof(a)) | |
2108 | return -EINVAL; | |
2109 | ||
1da177e4 | 2110 | /* copy_from_user should be SMP safe. */ |
47379052 | 2111 | if (copy_from_user(a, args, len)) |
1da177e4 | 2112 | return -EFAULT; |
3ec3b2fb | 2113 | |
f3298dc4 | 2114 | audit_socketcall(nargs[call] / sizeof(unsigned long), a); |
3ec3b2fb | 2115 | |
89bddce5 SH |
2116 | a0 = a[0]; |
2117 | a1 = a[1]; | |
2118 | ||
2119 | switch (call) { | |
2120 | case SYS_SOCKET: | |
2121 | err = sys_socket(a0, a1, a[2]); | |
2122 | break; | |
2123 | case SYS_BIND: | |
2124 | err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]); | |
2125 | break; | |
2126 | case SYS_CONNECT: | |
2127 | err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]); | |
2128 | break; | |
2129 | case SYS_LISTEN: | |
2130 | err = sys_listen(a0, a1); | |
2131 | break; | |
2132 | case SYS_ACCEPT: | |
de11defe UD |
2133 | err = sys_accept4(a0, (struct sockaddr __user *)a1, |
2134 | (int __user *)a[2], 0); | |
89bddce5 SH |
2135 | break; |
2136 | case SYS_GETSOCKNAME: | |
2137 | err = | |
2138 | sys_getsockname(a0, (struct sockaddr __user *)a1, | |
2139 | (int __user *)a[2]); | |
2140 | break; | |
2141 | case SYS_GETPEERNAME: | |
2142 | err = | |
2143 | sys_getpeername(a0, (struct sockaddr __user *)a1, | |
2144 | (int __user *)a[2]); | |
2145 | break; | |
2146 | case SYS_SOCKETPAIR: | |
2147 | err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]); | |
2148 | break; | |
2149 | case SYS_SEND: | |
2150 | err = sys_send(a0, (void __user *)a1, a[2], a[3]); | |
2151 | break; | |
2152 | case SYS_SENDTO: | |
2153 | err = sys_sendto(a0, (void __user *)a1, a[2], a[3], | |
2154 | (struct sockaddr __user *)a[4], a[5]); | |
2155 | break; | |
2156 | case SYS_RECV: | |
2157 | err = sys_recv(a0, (void __user *)a1, a[2], a[3]); | |
2158 | break; | |
2159 | case SYS_RECVFROM: | |
2160 | err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3], | |
2161 | (struct sockaddr __user *)a[4], | |
2162 | (int __user *)a[5]); | |
2163 | break; | |
2164 | case SYS_SHUTDOWN: | |
2165 | err = sys_shutdown(a0, a1); | |
2166 | break; | |
2167 | case SYS_SETSOCKOPT: | |
2168 | err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]); | |
2169 | break; | |
2170 | case SYS_GETSOCKOPT: | |
2171 | err = | |
2172 | sys_getsockopt(a0, a1, a[2], (char __user *)a[3], | |
2173 | (int __user *)a[4]); | |
2174 | break; | |
2175 | case SYS_SENDMSG: | |
2176 | err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2177 | break; | |
2178 | case SYS_RECVMSG: | |
2179 | err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2180 | break; | |
de11defe UD |
2181 | case SYS_ACCEPT4: |
2182 | err = sys_accept4(a0, (struct sockaddr __user *)a1, | |
2183 | (int __user *)a[2], a[3]); | |
aaca0bdc | 2184 | break; |
89bddce5 SH |
2185 | default: |
2186 | err = -EINVAL; | |
2187 | break; | |
1da177e4 LT |
2188 | } |
2189 | return err; | |
2190 | } | |
2191 | ||
89bddce5 | 2192 | #endif /* __ARCH_WANT_SYS_SOCKETCALL */ |
1da177e4 | 2193 | |
55737fda SH |
2194 | /** |
2195 | * sock_register - add a socket protocol handler | |
2196 | * @ops: description of protocol | |
2197 | * | |
1da177e4 LT |
2198 | * This function is called by a protocol handler that wants to |
2199 | * advertise its address family, and have it linked into the | |
55737fda SH |
2200 | * socket interface. The value ops->family coresponds to the |
2201 | * socket system call protocol family. | |
1da177e4 | 2202 | */ |
f0fd27d4 | 2203 | int sock_register(const struct net_proto_family *ops) |
1da177e4 LT |
2204 | { |
2205 | int err; | |
2206 | ||
2207 | if (ops->family >= NPROTO) { | |
89bddce5 SH |
2208 | printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, |
2209 | NPROTO); | |
1da177e4 LT |
2210 | return -ENOBUFS; |
2211 | } | |
55737fda SH |
2212 | |
2213 | spin_lock(&net_family_lock); | |
2214 | if (net_families[ops->family]) | |
2215 | err = -EEXIST; | |
2216 | else { | |
89bddce5 | 2217 | net_families[ops->family] = ops; |
1da177e4 LT |
2218 | err = 0; |
2219 | } | |
55737fda SH |
2220 | spin_unlock(&net_family_lock); |
2221 | ||
89bddce5 | 2222 | printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family); |
1da177e4 LT |
2223 | return err; |
2224 | } | |
2225 | ||
55737fda SH |
2226 | /** |
2227 | * sock_unregister - remove a protocol handler | |
2228 | * @family: protocol family to remove | |
2229 | * | |
1da177e4 LT |
2230 | * This function is called by a protocol handler that wants to |
2231 | * remove its address family, and have it unlinked from the | |
55737fda SH |
2232 | * new socket creation. |
2233 | * | |
2234 | * If protocol handler is a module, then it can use module reference | |
2235 | * counts to protect against new references. If protocol handler is not | |
2236 | * a module then it needs to provide its own protection in | |
2237 | * the ops->create routine. | |
1da177e4 | 2238 | */ |
f0fd27d4 | 2239 | void sock_unregister(int family) |
1da177e4 | 2240 | { |
f0fd27d4 | 2241 | BUG_ON(family < 0 || family >= NPROTO); |
1da177e4 | 2242 | |
55737fda | 2243 | spin_lock(&net_family_lock); |
89bddce5 | 2244 | net_families[family] = NULL; |
55737fda SH |
2245 | spin_unlock(&net_family_lock); |
2246 | ||
2247 | synchronize_rcu(); | |
2248 | ||
89bddce5 | 2249 | printk(KERN_INFO "NET: Unregistered protocol family %d\n", family); |
1da177e4 LT |
2250 | } |
2251 | ||
77d76ea3 | 2252 | static int __init sock_init(void) |
1da177e4 LT |
2253 | { |
2254 | /* | |
89bddce5 | 2255 | * Initialize sock SLAB cache. |
1da177e4 | 2256 | */ |
89bddce5 | 2257 | |
1da177e4 LT |
2258 | sk_init(); |
2259 | ||
1da177e4 | 2260 | /* |
89bddce5 | 2261 | * Initialize skbuff SLAB cache |
1da177e4 LT |
2262 | */ |
2263 | skb_init(); | |
1da177e4 LT |
2264 | |
2265 | /* | |
89bddce5 | 2266 | * Initialize the protocols module. |
1da177e4 LT |
2267 | */ |
2268 | ||
2269 | init_inodecache(); | |
2270 | register_filesystem(&sock_fs_type); | |
2271 | sock_mnt = kern_mount(&sock_fs_type); | |
77d76ea3 AK |
2272 | |
2273 | /* The real protocol initialization is performed in later initcalls. | |
1da177e4 LT |
2274 | */ |
2275 | ||
2276 | #ifdef CONFIG_NETFILTER | |
2277 | netfilter_init(); | |
2278 | #endif | |
cbeb321a DM |
2279 | |
2280 | return 0; | |
1da177e4 LT |
2281 | } |
2282 | ||
77d76ea3 AK |
2283 | core_initcall(sock_init); /* early initcall */ |
2284 | ||
1da177e4 LT |
2285 | #ifdef CONFIG_PROC_FS |
2286 | void socket_seq_show(struct seq_file *seq) | |
2287 | { | |
2288 | int cpu; | |
2289 | int counter = 0; | |
2290 | ||
6f912042 | 2291 | for_each_possible_cpu(cpu) |
89bddce5 | 2292 | counter += per_cpu(sockets_in_use, cpu); |
1da177e4 LT |
2293 | |
2294 | /* It can be negative, by the way. 8) */ | |
2295 | if (counter < 0) | |
2296 | counter = 0; | |
2297 | ||
2298 | seq_printf(seq, "sockets: used %d\n", counter); | |
2299 | } | |
89bddce5 | 2300 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 2301 | |
89bbfc95 SP |
2302 | #ifdef CONFIG_COMPAT |
2303 | static long compat_sock_ioctl(struct file *file, unsigned cmd, | |
89bddce5 | 2304 | unsigned long arg) |
89bbfc95 SP |
2305 | { |
2306 | struct socket *sock = file->private_data; | |
2307 | int ret = -ENOIOCTLCMD; | |
87de87d5 DM |
2308 | struct sock *sk; |
2309 | struct net *net; | |
2310 | ||
2311 | sk = sock->sk; | |
2312 | net = sock_net(sk); | |
89bbfc95 SP |
2313 | |
2314 | if (sock->ops->compat_ioctl) | |
2315 | ret = sock->ops->compat_ioctl(sock, cmd, arg); | |
2316 | ||
87de87d5 DM |
2317 | if (ret == -ENOIOCTLCMD && |
2318 | (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) | |
2319 | ret = compat_wext_handle_ioctl(net, cmd, arg); | |
2320 | ||
89bbfc95 SP |
2321 | return ret; |
2322 | } | |
2323 | #endif | |
2324 | ||
ac5a488e SS |
2325 | int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) |
2326 | { | |
2327 | return sock->ops->bind(sock, addr, addrlen); | |
2328 | } | |
2329 | ||
2330 | int kernel_listen(struct socket *sock, int backlog) | |
2331 | { | |
2332 | return sock->ops->listen(sock, backlog); | |
2333 | } | |
2334 | ||
2335 | int kernel_accept(struct socket *sock, struct socket **newsock, int flags) | |
2336 | { | |
2337 | struct sock *sk = sock->sk; | |
2338 | int err; | |
2339 | ||
2340 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
2341 | newsock); | |
2342 | if (err < 0) | |
2343 | goto done; | |
2344 | ||
2345 | err = sock->ops->accept(sock, *newsock, flags); | |
2346 | if (err < 0) { | |
2347 | sock_release(*newsock); | |
fa8705b0 | 2348 | *newsock = NULL; |
ac5a488e SS |
2349 | goto done; |
2350 | } | |
2351 | ||
2352 | (*newsock)->ops = sock->ops; | |
1b08534e | 2353 | __module_get((*newsock)->ops->owner); |
ac5a488e SS |
2354 | |
2355 | done: | |
2356 | return err; | |
2357 | } | |
2358 | ||
2359 | int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, | |
4768fbcb | 2360 | int flags) |
ac5a488e SS |
2361 | { |
2362 | return sock->ops->connect(sock, addr, addrlen, flags); | |
2363 | } | |
2364 | ||
2365 | int kernel_getsockname(struct socket *sock, struct sockaddr *addr, | |
2366 | int *addrlen) | |
2367 | { | |
2368 | return sock->ops->getname(sock, addr, addrlen, 0); | |
2369 | } | |
2370 | ||
2371 | int kernel_getpeername(struct socket *sock, struct sockaddr *addr, | |
2372 | int *addrlen) | |
2373 | { | |
2374 | return sock->ops->getname(sock, addr, addrlen, 1); | |
2375 | } | |
2376 | ||
2377 | int kernel_getsockopt(struct socket *sock, int level, int optname, | |
2378 | char *optval, int *optlen) | |
2379 | { | |
2380 | mm_segment_t oldfs = get_fs(); | |
2381 | int err; | |
2382 | ||
2383 | set_fs(KERNEL_DS); | |
2384 | if (level == SOL_SOCKET) | |
2385 | err = sock_getsockopt(sock, level, optname, optval, optlen); | |
2386 | else | |
2387 | err = sock->ops->getsockopt(sock, level, optname, optval, | |
2388 | optlen); | |
2389 | set_fs(oldfs); | |
2390 | return err; | |
2391 | } | |
2392 | ||
2393 | int kernel_setsockopt(struct socket *sock, int level, int optname, | |
2394 | char *optval, int optlen) | |
2395 | { | |
2396 | mm_segment_t oldfs = get_fs(); | |
2397 | int err; | |
2398 | ||
2399 | set_fs(KERNEL_DS); | |
2400 | if (level == SOL_SOCKET) | |
2401 | err = sock_setsockopt(sock, level, optname, optval, optlen); | |
2402 | else | |
2403 | err = sock->ops->setsockopt(sock, level, optname, optval, | |
2404 | optlen); | |
2405 | set_fs(oldfs); | |
2406 | return err; | |
2407 | } | |
2408 | ||
2409 | int kernel_sendpage(struct socket *sock, struct page *page, int offset, | |
2410 | size_t size, int flags) | |
2411 | { | |
2412 | if (sock->ops->sendpage) | |
2413 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
2414 | ||
2415 | return sock_no_sendpage(sock, page, offset, size, flags); | |
2416 | } | |
2417 | ||
2418 | int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg) | |
2419 | { | |
2420 | mm_segment_t oldfs = get_fs(); | |
2421 | int err; | |
2422 | ||
2423 | set_fs(KERNEL_DS); | |
2424 | err = sock->ops->ioctl(sock, cmd, arg); | |
2425 | set_fs(oldfs); | |
2426 | ||
2427 | return err; | |
2428 | } | |
2429 | ||
91cf45f0 TM |
2430 | int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) |
2431 | { | |
2432 | return sock->ops->shutdown(sock, how); | |
2433 | } | |
2434 | ||
1da177e4 LT |
2435 | EXPORT_SYMBOL(sock_create); |
2436 | EXPORT_SYMBOL(sock_create_kern); | |
2437 | EXPORT_SYMBOL(sock_create_lite); | |
2438 | EXPORT_SYMBOL(sock_map_fd); | |
2439 | EXPORT_SYMBOL(sock_recvmsg); | |
2440 | EXPORT_SYMBOL(sock_register); | |
2441 | EXPORT_SYMBOL(sock_release); | |
2442 | EXPORT_SYMBOL(sock_sendmsg); | |
2443 | EXPORT_SYMBOL(sock_unregister); | |
2444 | EXPORT_SYMBOL(sock_wake_async); | |
2445 | EXPORT_SYMBOL(sockfd_lookup); | |
2446 | EXPORT_SYMBOL(kernel_sendmsg); | |
2447 | EXPORT_SYMBOL(kernel_recvmsg); | |
ac5a488e SS |
2448 | EXPORT_SYMBOL(kernel_bind); |
2449 | EXPORT_SYMBOL(kernel_listen); | |
2450 | EXPORT_SYMBOL(kernel_accept); | |
2451 | EXPORT_SYMBOL(kernel_connect); | |
2452 | EXPORT_SYMBOL(kernel_getsockname); | |
2453 | EXPORT_SYMBOL(kernel_getpeername); | |
2454 | EXPORT_SYMBOL(kernel_getsockopt); | |
2455 | EXPORT_SYMBOL(kernel_setsockopt); | |
2456 | EXPORT_SYMBOL(kernel_sendpage); | |
2457 | EXPORT_SYMBOL(kernel_sock_ioctl); | |
91cf45f0 | 2458 | EXPORT_SYMBOL(kernel_sock_shutdown); |