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> |
c019bbc6 | 72 | #include <linux/thread_info.h> |
1da177e4 LT |
73 | #include <linux/wanrouter.h> |
74 | #include <linux/if_bridge.h> | |
20380731 ACM |
75 | #include <linux/if_frad.h> |
76 | #include <linux/if_vlan.h> | |
1da177e4 LT |
77 | #include <linux/init.h> |
78 | #include <linux/poll.h> | |
79 | #include <linux/cache.h> | |
80 | #include <linux/module.h> | |
81 | #include <linux/highmem.h> | |
1da177e4 LT |
82 | #include <linux/mount.h> |
83 | #include <linux/security.h> | |
84 | #include <linux/syscalls.h> | |
85 | #include <linux/compat.h> | |
86 | #include <linux/kmod.h> | |
3ec3b2fb | 87 | #include <linux/audit.h> |
d86b5e0e | 88 | #include <linux/wireless.h> |
1b8d7ae4 | 89 | #include <linux/nsproxy.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 | ||
239 | #define SOCKFS_MAGIC 0x534F434B | |
240 | ||
e18b890b | 241 | static struct kmem_cache *sock_inode_cachep __read_mostly; |
1da177e4 LT |
242 | |
243 | static struct inode *sock_alloc_inode(struct super_block *sb) | |
244 | { | |
245 | struct socket_alloc *ei; | |
89bddce5 | 246 | |
e94b1766 | 247 | ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); |
1da177e4 LT |
248 | if (!ei) |
249 | return NULL; | |
250 | init_waitqueue_head(&ei->socket.wait); | |
89bddce5 | 251 | |
1da177e4 LT |
252 | ei->socket.fasync_list = NULL; |
253 | ei->socket.state = SS_UNCONNECTED; | |
254 | ei->socket.flags = 0; | |
255 | ei->socket.ops = NULL; | |
256 | ei->socket.sk = NULL; | |
257 | ei->socket.file = NULL; | |
1da177e4 LT |
258 | |
259 | return &ei->vfs_inode; | |
260 | } | |
261 | ||
262 | static void sock_destroy_inode(struct inode *inode) | |
263 | { | |
264 | kmem_cache_free(sock_inode_cachep, | |
265 | container_of(inode, struct socket_alloc, vfs_inode)); | |
266 | } | |
267 | ||
4ba9b9d0 | 268 | static void init_once(struct kmem_cache *cachep, void *foo) |
1da177e4 | 269 | { |
89bddce5 | 270 | struct socket_alloc *ei = (struct socket_alloc *)foo; |
1da177e4 | 271 | |
a35afb83 | 272 | inode_init_once(&ei->vfs_inode); |
1da177e4 | 273 | } |
89bddce5 | 274 | |
1da177e4 LT |
275 | static int init_inodecache(void) |
276 | { | |
277 | sock_inode_cachep = kmem_cache_create("sock_inode_cache", | |
89bddce5 SH |
278 | sizeof(struct socket_alloc), |
279 | 0, | |
280 | (SLAB_HWCACHE_ALIGN | | |
281 | SLAB_RECLAIM_ACCOUNT | | |
282 | SLAB_MEM_SPREAD), | |
20c2df83 | 283 | init_once); |
1da177e4 LT |
284 | if (sock_inode_cachep == NULL) |
285 | return -ENOMEM; | |
286 | return 0; | |
287 | } | |
288 | ||
289 | static struct super_operations sockfs_ops = { | |
290 | .alloc_inode = sock_alloc_inode, | |
291 | .destroy_inode =sock_destroy_inode, | |
292 | .statfs = simple_statfs, | |
293 | }; | |
294 | ||
454e2398 | 295 | static int sockfs_get_sb(struct file_system_type *fs_type, |
89bddce5 SH |
296 | int flags, const char *dev_name, void *data, |
297 | struct vfsmount *mnt) | |
1da177e4 | 298 | { |
454e2398 DH |
299 | return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC, |
300 | mnt); | |
1da177e4 LT |
301 | } |
302 | ||
ba89966c | 303 | static struct vfsmount *sock_mnt __read_mostly; |
1da177e4 LT |
304 | |
305 | static struct file_system_type sock_fs_type = { | |
306 | .name = "sockfs", | |
307 | .get_sb = sockfs_get_sb, | |
308 | .kill_sb = kill_anon_super, | |
309 | }; | |
89bddce5 | 310 | |
1da177e4 LT |
311 | static int sockfs_delete_dentry(struct dentry *dentry) |
312 | { | |
304e61e6 ED |
313 | /* |
314 | * At creation time, we pretended this dentry was hashed | |
315 | * (by clearing DCACHE_UNHASHED bit in d_flags) | |
316 | * At delete time, we restore the truth : not hashed. | |
317 | * (so that dput() can proceed correctly) | |
318 | */ | |
319 | dentry->d_flags |= DCACHE_UNHASHED; | |
320 | return 0; | |
1da177e4 | 321 | } |
c23fbb6b ED |
322 | |
323 | /* | |
324 | * sockfs_dname() is called from d_path(). | |
325 | */ | |
326 | static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) | |
327 | { | |
328 | return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", | |
329 | dentry->d_inode->i_ino); | |
330 | } | |
331 | ||
1da177e4 | 332 | static struct dentry_operations sockfs_dentry_operations = { |
89bddce5 | 333 | .d_delete = sockfs_delete_dentry, |
c23fbb6b | 334 | .d_dname = sockfs_dname, |
1da177e4 LT |
335 | }; |
336 | ||
337 | /* | |
338 | * Obtains the first available file descriptor and sets it up for use. | |
339 | * | |
39d8c1b6 DM |
340 | * These functions create file structures and maps them to fd space |
341 | * of the current process. On success it returns file descriptor | |
1da177e4 LT |
342 | * and file struct implicitly stored in sock->file. |
343 | * Note that another thread may close file descriptor before we return | |
344 | * from this function. We use the fact that now we do not refer | |
345 | * to socket after mapping. If one day we will need it, this | |
346 | * function will increment ref. count on file by 1. | |
347 | * | |
348 | * In any case returned fd MAY BE not valid! | |
349 | * This race condition is unavoidable | |
350 | * with shared fd spaces, we cannot solve it inside kernel, | |
351 | * but we take care of internal coherence yet. | |
352 | */ | |
353 | ||
a677a039 | 354 | static int sock_alloc_fd(struct file **filep, int flags) |
1da177e4 LT |
355 | { |
356 | int fd; | |
1da177e4 | 357 | |
a677a039 | 358 | fd = get_unused_fd_flags(flags); |
39d8c1b6 | 359 | if (likely(fd >= 0)) { |
1da177e4 LT |
360 | struct file *file = get_empty_filp(); |
361 | ||
39d8c1b6 DM |
362 | *filep = file; |
363 | if (unlikely(!file)) { | |
1da177e4 | 364 | put_unused_fd(fd); |
39d8c1b6 | 365 | return -ENFILE; |
1da177e4 | 366 | } |
39d8c1b6 DM |
367 | } else |
368 | *filep = NULL; | |
369 | return fd; | |
370 | } | |
1da177e4 | 371 | |
77d27200 | 372 | static int sock_attach_fd(struct socket *sock, struct file *file, int flags) |
39d8c1b6 | 373 | { |
ce8d2cdf | 374 | struct dentry *dentry; |
c23fbb6b | 375 | struct qstr name = { .name = "" }; |
39d8c1b6 | 376 | |
ce8d2cdf DH |
377 | dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name); |
378 | if (unlikely(!dentry)) | |
39d8c1b6 DM |
379 | return -ENOMEM; |
380 | ||
ce8d2cdf | 381 | dentry->d_op = &sockfs_dentry_operations; |
304e61e6 ED |
382 | /* |
383 | * We dont want to push this dentry into global dentry hash table. | |
384 | * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED | |
385 | * This permits a working /proc/$pid/fd/XXX on sockets | |
386 | */ | |
ce8d2cdf DH |
387 | dentry->d_flags &= ~DCACHE_UNHASHED; |
388 | d_instantiate(dentry, SOCK_INODE(sock)); | |
39d8c1b6 DM |
389 | |
390 | sock->file = file; | |
ce8d2cdf DH |
391 | init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE, |
392 | &socket_file_ops); | |
393 | SOCK_INODE(sock)->i_fop = &socket_file_ops; | |
77d27200 | 394 | file->f_flags = O_RDWR | (flags & O_NONBLOCK); |
39d8c1b6 DM |
395 | file->f_pos = 0; |
396 | file->private_data = sock; | |
1da177e4 | 397 | |
39d8c1b6 DM |
398 | return 0; |
399 | } | |
400 | ||
a677a039 | 401 | int sock_map_fd(struct socket *sock, int flags) |
39d8c1b6 DM |
402 | { |
403 | struct file *newfile; | |
a677a039 | 404 | int fd = sock_alloc_fd(&newfile, flags); |
39d8c1b6 DM |
405 | |
406 | if (likely(fd >= 0)) { | |
77d27200 | 407 | int err = sock_attach_fd(sock, newfile, flags); |
39d8c1b6 DM |
408 | |
409 | if (unlikely(err < 0)) { | |
410 | put_filp(newfile); | |
1da177e4 | 411 | put_unused_fd(fd); |
39d8c1b6 | 412 | return err; |
1da177e4 | 413 | } |
39d8c1b6 | 414 | fd_install(fd, newfile); |
1da177e4 | 415 | } |
1da177e4 LT |
416 | return fd; |
417 | } | |
418 | ||
6cb153ca BL |
419 | static struct socket *sock_from_file(struct file *file, int *err) |
420 | { | |
6cb153ca BL |
421 | if (file->f_op == &socket_file_ops) |
422 | return file->private_data; /* set in sock_map_fd */ | |
423 | ||
23bb80d2 ED |
424 | *err = -ENOTSOCK; |
425 | return NULL; | |
6cb153ca BL |
426 | } |
427 | ||
1da177e4 LT |
428 | /** |
429 | * sockfd_lookup - Go from a file number to its socket slot | |
430 | * @fd: file handle | |
431 | * @err: pointer to an error code return | |
432 | * | |
433 | * The file handle passed in is locked and the socket it is bound | |
434 | * too is returned. If an error occurs the err pointer is overwritten | |
435 | * with a negative errno code and NULL is returned. The function checks | |
436 | * for both invalid handles and passing a handle which is not a socket. | |
437 | * | |
438 | * On a success the socket object pointer is returned. | |
439 | */ | |
440 | ||
441 | struct socket *sockfd_lookup(int fd, int *err) | |
442 | { | |
443 | struct file *file; | |
1da177e4 LT |
444 | struct socket *sock; |
445 | ||
89bddce5 SH |
446 | file = fget(fd); |
447 | if (!file) { | |
1da177e4 LT |
448 | *err = -EBADF; |
449 | return NULL; | |
450 | } | |
89bddce5 | 451 | |
6cb153ca BL |
452 | sock = sock_from_file(file, err); |
453 | if (!sock) | |
1da177e4 | 454 | fput(file); |
6cb153ca BL |
455 | return sock; |
456 | } | |
1da177e4 | 457 | |
6cb153ca BL |
458 | static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) |
459 | { | |
460 | struct file *file; | |
461 | struct socket *sock; | |
462 | ||
3672558c | 463 | *err = -EBADF; |
6cb153ca BL |
464 | file = fget_light(fd, fput_needed); |
465 | if (file) { | |
466 | sock = sock_from_file(file, err); | |
467 | if (sock) | |
468 | return sock; | |
469 | fput_light(file, *fput_needed); | |
1da177e4 | 470 | } |
6cb153ca | 471 | return NULL; |
1da177e4 LT |
472 | } |
473 | ||
474 | /** | |
475 | * sock_alloc - allocate a socket | |
89bddce5 | 476 | * |
1da177e4 LT |
477 | * Allocate a new inode and socket object. The two are bound together |
478 | * and initialised. The socket is then returned. If we are out of inodes | |
479 | * NULL is returned. | |
480 | */ | |
481 | ||
482 | static struct socket *sock_alloc(void) | |
483 | { | |
89bddce5 SH |
484 | struct inode *inode; |
485 | struct socket *sock; | |
1da177e4 LT |
486 | |
487 | inode = new_inode(sock_mnt->mnt_sb); | |
488 | if (!inode) | |
489 | return NULL; | |
490 | ||
491 | sock = SOCKET_I(inode); | |
492 | ||
89bddce5 | 493 | inode->i_mode = S_IFSOCK | S_IRWXUGO; |
1da177e4 LT |
494 | inode->i_uid = current->fsuid; |
495 | inode->i_gid = current->fsgid; | |
496 | ||
497 | get_cpu_var(sockets_in_use)++; | |
498 | put_cpu_var(sockets_in_use); | |
499 | return sock; | |
500 | } | |
501 | ||
502 | /* | |
503 | * In theory you can't get an open on this inode, but /proc provides | |
504 | * a back door. Remember to keep it shut otherwise you'll let the | |
505 | * creepy crawlies in. | |
506 | */ | |
89bddce5 | 507 | |
1da177e4 LT |
508 | static int sock_no_open(struct inode *irrelevant, struct file *dontcare) |
509 | { | |
510 | return -ENXIO; | |
511 | } | |
512 | ||
4b6f5d20 | 513 | const struct file_operations bad_sock_fops = { |
1da177e4 LT |
514 | .owner = THIS_MODULE, |
515 | .open = sock_no_open, | |
516 | }; | |
517 | ||
518 | /** | |
519 | * sock_release - close a socket | |
520 | * @sock: socket to close | |
521 | * | |
522 | * The socket is released from the protocol stack if it has a release | |
523 | * callback, and the inode is then released if the socket is bound to | |
89bddce5 | 524 | * an inode not a file. |
1da177e4 | 525 | */ |
89bddce5 | 526 | |
1da177e4 LT |
527 | void sock_release(struct socket *sock) |
528 | { | |
529 | if (sock->ops) { | |
530 | struct module *owner = sock->ops->owner; | |
531 | ||
532 | sock->ops->release(sock); | |
533 | sock->ops = NULL; | |
534 | module_put(owner); | |
535 | } | |
536 | ||
537 | if (sock->fasync_list) | |
538 | printk(KERN_ERR "sock_release: fasync list not empty!\n"); | |
539 | ||
540 | get_cpu_var(sockets_in_use)--; | |
541 | put_cpu_var(sockets_in_use); | |
542 | if (!sock->file) { | |
543 | iput(SOCK_INODE(sock)); | |
544 | return; | |
545 | } | |
89bddce5 | 546 | sock->file = NULL; |
1da177e4 LT |
547 | } |
548 | ||
89bddce5 | 549 | static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
550 | struct msghdr *msg, size_t size) |
551 | { | |
552 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
553 | int err; | |
554 | ||
555 | si->sock = sock; | |
556 | si->scm = NULL; | |
557 | si->msg = msg; | |
558 | si->size = size; | |
559 | ||
560 | err = security_socket_sendmsg(sock, msg, size); | |
561 | if (err) | |
562 | return err; | |
563 | ||
564 | return sock->ops->sendmsg(iocb, sock, msg, size); | |
565 | } | |
566 | ||
567 | int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) | |
568 | { | |
569 | struct kiocb iocb; | |
570 | struct sock_iocb siocb; | |
571 | int ret; | |
572 | ||
573 | init_sync_kiocb(&iocb, NULL); | |
574 | iocb.private = &siocb; | |
575 | ret = __sock_sendmsg(&iocb, sock, msg, size); | |
576 | if (-EIOCBQUEUED == ret) | |
577 | ret = wait_on_sync_kiocb(&iocb); | |
578 | return ret; | |
579 | } | |
580 | ||
581 | int kernel_sendmsg(struct socket *sock, struct msghdr *msg, | |
582 | struct kvec *vec, size_t num, size_t size) | |
583 | { | |
584 | mm_segment_t oldfs = get_fs(); | |
585 | int result; | |
586 | ||
587 | set_fs(KERNEL_DS); | |
588 | /* | |
589 | * the following is safe, since for compiler definitions of kvec and | |
590 | * iovec are identical, yielding the same in-core layout and alignment | |
591 | */ | |
89bddce5 | 592 | msg->msg_iov = (struct iovec *)vec; |
1da177e4 LT |
593 | msg->msg_iovlen = num; |
594 | result = sock_sendmsg(sock, msg, size); | |
595 | set_fs(oldfs); | |
596 | return result; | |
597 | } | |
598 | ||
92f37fd2 ED |
599 | /* |
600 | * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) | |
601 | */ | |
602 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, | |
603 | struct sk_buff *skb) | |
604 | { | |
605 | ktime_t kt = skb->tstamp; | |
606 | ||
607 | if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { | |
608 | struct timeval tv; | |
609 | /* Race occurred between timestamp enabling and packet | |
610 | receiving. Fill in the current time for now. */ | |
611 | if (kt.tv64 == 0) | |
612 | kt = ktime_get_real(); | |
613 | skb->tstamp = kt; | |
614 | tv = ktime_to_timeval(kt); | |
615 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv); | |
616 | } else { | |
617 | struct timespec ts; | |
618 | /* Race occurred between timestamp enabling and packet | |
619 | receiving. Fill in the current time for now. */ | |
620 | if (kt.tv64 == 0) | |
621 | kt = ktime_get_real(); | |
622 | skb->tstamp = kt; | |
623 | ts = ktime_to_timespec(kt); | |
624 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts); | |
625 | } | |
626 | } | |
627 | ||
7c81fd8b ACM |
628 | EXPORT_SYMBOL_GPL(__sock_recv_timestamp); |
629 | ||
89bddce5 | 630 | static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 LT |
631 | struct msghdr *msg, size_t size, int flags) |
632 | { | |
633 | int err; | |
634 | struct sock_iocb *si = kiocb_to_siocb(iocb); | |
635 | ||
636 | si->sock = sock; | |
637 | si->scm = NULL; | |
638 | si->msg = msg; | |
639 | si->size = size; | |
640 | si->flags = flags; | |
641 | ||
642 | err = security_socket_recvmsg(sock, msg, size, flags); | |
643 | if (err) | |
644 | return err; | |
645 | ||
646 | return sock->ops->recvmsg(iocb, sock, msg, size, flags); | |
647 | } | |
648 | ||
89bddce5 | 649 | int sock_recvmsg(struct socket *sock, struct msghdr *msg, |
1da177e4 LT |
650 | size_t size, int flags) |
651 | { | |
652 | struct kiocb iocb; | |
653 | struct sock_iocb siocb; | |
654 | int ret; | |
655 | ||
89bddce5 | 656 | init_sync_kiocb(&iocb, NULL); |
1da177e4 LT |
657 | iocb.private = &siocb; |
658 | ret = __sock_recvmsg(&iocb, sock, msg, size, flags); | |
659 | if (-EIOCBQUEUED == ret) | |
660 | ret = wait_on_sync_kiocb(&iocb); | |
661 | return ret; | |
662 | } | |
663 | ||
89bddce5 SH |
664 | int kernel_recvmsg(struct socket *sock, struct msghdr *msg, |
665 | struct kvec *vec, size_t num, size_t size, int flags) | |
1da177e4 LT |
666 | { |
667 | mm_segment_t oldfs = get_fs(); | |
668 | int result; | |
669 | ||
670 | set_fs(KERNEL_DS); | |
671 | /* | |
672 | * the following is safe, since for compiler definitions of kvec and | |
673 | * iovec are identical, yielding the same in-core layout and alignment | |
674 | */ | |
89bddce5 | 675 | msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num; |
1da177e4 LT |
676 | result = sock_recvmsg(sock, msg, size, flags); |
677 | set_fs(oldfs); | |
678 | return result; | |
679 | } | |
680 | ||
681 | static void sock_aio_dtor(struct kiocb *iocb) | |
682 | { | |
683 | kfree(iocb->private); | |
684 | } | |
685 | ||
ce1d4d3e CH |
686 | static ssize_t sock_sendpage(struct file *file, struct page *page, |
687 | int offset, size_t size, loff_t *ppos, int more) | |
1da177e4 | 688 | { |
1da177e4 LT |
689 | struct socket *sock; |
690 | int flags; | |
691 | ||
ce1d4d3e CH |
692 | sock = file->private_data; |
693 | ||
694 | flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT; | |
695 | if (more) | |
696 | flags |= MSG_MORE; | |
697 | ||
698 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
699 | } | |
1da177e4 | 700 | |
9c55e01c JA |
701 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, |
702 | struct pipe_inode_info *pipe, size_t len, | |
703 | unsigned int flags) | |
704 | { | |
705 | struct socket *sock = file->private_data; | |
706 | ||
997b37da RDC |
707 | if (unlikely(!sock->ops->splice_read)) |
708 | return -EINVAL; | |
709 | ||
9c55e01c JA |
710 | return sock->ops->splice_read(sock, ppos, pipe, len, flags); |
711 | } | |
712 | ||
ce1d4d3e | 713 | static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb, |
89bddce5 | 714 | struct sock_iocb *siocb) |
ce1d4d3e CH |
715 | { |
716 | if (!is_sync_kiocb(iocb)) { | |
717 | siocb = kmalloc(sizeof(*siocb), GFP_KERNEL); | |
718 | if (!siocb) | |
719 | return NULL; | |
1da177e4 LT |
720 | iocb->ki_dtor = sock_aio_dtor; |
721 | } | |
1da177e4 | 722 | |
ce1d4d3e | 723 | siocb->kiocb = iocb; |
ce1d4d3e CH |
724 | iocb->private = siocb; |
725 | return siocb; | |
1da177e4 LT |
726 | } |
727 | ||
ce1d4d3e | 728 | static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb, |
027445c3 BP |
729 | struct file *file, const struct iovec *iov, |
730 | unsigned long nr_segs) | |
ce1d4d3e CH |
731 | { |
732 | struct socket *sock = file->private_data; | |
733 | size_t size = 0; | |
734 | int i; | |
1da177e4 | 735 | |
89bddce5 SH |
736 | for (i = 0; i < nr_segs; i++) |
737 | size += iov[i].iov_len; | |
1da177e4 | 738 | |
ce1d4d3e CH |
739 | msg->msg_name = NULL; |
740 | msg->msg_namelen = 0; | |
741 | msg->msg_control = NULL; | |
742 | msg->msg_controllen = 0; | |
89bddce5 | 743 | msg->msg_iov = (struct iovec *)iov; |
ce1d4d3e CH |
744 | msg->msg_iovlen = nr_segs; |
745 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
746 | ||
747 | return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags); | |
748 | } | |
749 | ||
027445c3 BP |
750 | static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, |
751 | unsigned long nr_segs, loff_t pos) | |
ce1d4d3e CH |
752 | { |
753 | struct sock_iocb siocb, *x; | |
754 | ||
1da177e4 LT |
755 | if (pos != 0) |
756 | return -ESPIPE; | |
027445c3 BP |
757 | |
758 | if (iocb->ki_left == 0) /* Match SYS5 behaviour */ | |
1da177e4 LT |
759 | return 0; |
760 | ||
027445c3 BP |
761 | |
762 | x = alloc_sock_iocb(iocb, &siocb); | |
ce1d4d3e CH |
763 | if (!x) |
764 | return -ENOMEM; | |
027445c3 | 765 | return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
1da177e4 LT |
766 | } |
767 | ||
ce1d4d3e | 768 | static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb, |
027445c3 BP |
769 | struct file *file, const struct iovec *iov, |
770 | unsigned long nr_segs) | |
1da177e4 | 771 | { |
ce1d4d3e CH |
772 | struct socket *sock = file->private_data; |
773 | size_t size = 0; | |
774 | int i; | |
1da177e4 | 775 | |
89bddce5 SH |
776 | for (i = 0; i < nr_segs; i++) |
777 | size += iov[i].iov_len; | |
1da177e4 | 778 | |
ce1d4d3e CH |
779 | msg->msg_name = NULL; |
780 | msg->msg_namelen = 0; | |
781 | msg->msg_control = NULL; | |
782 | msg->msg_controllen = 0; | |
89bddce5 | 783 | msg->msg_iov = (struct iovec *)iov; |
ce1d4d3e CH |
784 | msg->msg_iovlen = nr_segs; |
785 | msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; | |
786 | if (sock->type == SOCK_SEQPACKET) | |
787 | msg->msg_flags |= MSG_EOR; | |
1da177e4 | 788 | |
ce1d4d3e | 789 | return __sock_sendmsg(iocb, sock, msg, size); |
1da177e4 LT |
790 | } |
791 | ||
027445c3 BP |
792 | static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, |
793 | unsigned long nr_segs, loff_t pos) | |
ce1d4d3e CH |
794 | { |
795 | struct sock_iocb siocb, *x; | |
1da177e4 | 796 | |
ce1d4d3e CH |
797 | if (pos != 0) |
798 | return -ESPIPE; | |
027445c3 | 799 | |
027445c3 | 800 | x = alloc_sock_iocb(iocb, &siocb); |
ce1d4d3e CH |
801 | if (!x) |
802 | return -ENOMEM; | |
1da177e4 | 803 | |
027445c3 | 804 | return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); |
1da177e4 LT |
805 | } |
806 | ||
1da177e4 LT |
807 | /* |
808 | * Atomic setting of ioctl hooks to avoid race | |
809 | * with module unload. | |
810 | */ | |
811 | ||
4a3e2f71 | 812 | static DEFINE_MUTEX(br_ioctl_mutex); |
881d966b | 813 | static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL; |
1da177e4 | 814 | |
881d966b | 815 | void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) |
1da177e4 | 816 | { |
4a3e2f71 | 817 | mutex_lock(&br_ioctl_mutex); |
1da177e4 | 818 | br_ioctl_hook = hook; |
4a3e2f71 | 819 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 | 820 | } |
89bddce5 | 821 | |
1da177e4 LT |
822 | EXPORT_SYMBOL(brioctl_set); |
823 | ||
4a3e2f71 | 824 | static DEFINE_MUTEX(vlan_ioctl_mutex); |
881d966b | 825 | static int (*vlan_ioctl_hook) (struct net *, void __user *arg); |
1da177e4 | 826 | |
881d966b | 827 | void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) |
1da177e4 | 828 | { |
4a3e2f71 | 829 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 830 | vlan_ioctl_hook = hook; |
4a3e2f71 | 831 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 832 | } |
89bddce5 | 833 | |
1da177e4 LT |
834 | EXPORT_SYMBOL(vlan_ioctl_set); |
835 | ||
4a3e2f71 | 836 | static DEFINE_MUTEX(dlci_ioctl_mutex); |
89bddce5 | 837 | static int (*dlci_ioctl_hook) (unsigned int, void __user *); |
1da177e4 | 838 | |
89bddce5 | 839 | void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) |
1da177e4 | 840 | { |
4a3e2f71 | 841 | mutex_lock(&dlci_ioctl_mutex); |
1da177e4 | 842 | dlci_ioctl_hook = hook; |
4a3e2f71 | 843 | mutex_unlock(&dlci_ioctl_mutex); |
1da177e4 | 844 | } |
89bddce5 | 845 | |
1da177e4 LT |
846 | EXPORT_SYMBOL(dlci_ioctl_set); |
847 | ||
848 | /* | |
849 | * With an ioctl, arg may well be a user mode pointer, but we don't know | |
850 | * what to do with it - that's up to the protocol still. | |
851 | */ | |
852 | ||
853 | static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) | |
854 | { | |
855 | struct socket *sock; | |
881d966b | 856 | struct sock *sk; |
1da177e4 LT |
857 | void __user *argp = (void __user *)arg; |
858 | int pid, err; | |
881d966b | 859 | struct net *net; |
1da177e4 | 860 | |
b69aee04 | 861 | sock = file->private_data; |
881d966b | 862 | sk = sock->sk; |
3b1e0a65 | 863 | net = sock_net(sk); |
1da177e4 | 864 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) { |
881d966b | 865 | err = dev_ioctl(net, cmd, argp); |
1da177e4 | 866 | } else |
d86b5e0e | 867 | #ifdef CONFIG_WIRELESS_EXT |
1da177e4 | 868 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { |
881d966b | 869 | err = dev_ioctl(net, cmd, argp); |
1da177e4 | 870 | } else |
89bddce5 SH |
871 | #endif /* CONFIG_WIRELESS_EXT */ |
872 | switch (cmd) { | |
1da177e4 LT |
873 | case FIOSETOWN: |
874 | case SIOCSPGRP: | |
875 | err = -EFAULT; | |
876 | if (get_user(pid, (int __user *)argp)) | |
877 | break; | |
878 | err = f_setown(sock->file, pid, 1); | |
879 | break; | |
880 | case FIOGETOWN: | |
881 | case SIOCGPGRP: | |
609d7fa9 | 882 | err = put_user(f_getown(sock->file), |
89bddce5 | 883 | (int __user *)argp); |
1da177e4 LT |
884 | break; |
885 | case SIOCGIFBR: | |
886 | case SIOCSIFBR: | |
887 | case SIOCBRADDBR: | |
888 | case SIOCBRDELBR: | |
889 | err = -ENOPKG; | |
890 | if (!br_ioctl_hook) | |
891 | request_module("bridge"); | |
892 | ||
4a3e2f71 | 893 | mutex_lock(&br_ioctl_mutex); |
89bddce5 | 894 | if (br_ioctl_hook) |
881d966b | 895 | err = br_ioctl_hook(net, cmd, argp); |
4a3e2f71 | 896 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 LT |
897 | break; |
898 | case SIOCGIFVLAN: | |
899 | case SIOCSIFVLAN: | |
900 | err = -ENOPKG; | |
901 | if (!vlan_ioctl_hook) | |
902 | request_module("8021q"); | |
903 | ||
4a3e2f71 | 904 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 905 | if (vlan_ioctl_hook) |
881d966b | 906 | err = vlan_ioctl_hook(net, argp); |
4a3e2f71 | 907 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 908 | break; |
1da177e4 LT |
909 | case SIOCADDDLCI: |
910 | case SIOCDELDLCI: | |
911 | err = -ENOPKG; | |
912 | if (!dlci_ioctl_hook) | |
913 | request_module("dlci"); | |
914 | ||
7512cbf6 PE |
915 | mutex_lock(&dlci_ioctl_mutex); |
916 | if (dlci_ioctl_hook) | |
1da177e4 | 917 | err = dlci_ioctl_hook(cmd, argp); |
7512cbf6 | 918 | mutex_unlock(&dlci_ioctl_mutex); |
1da177e4 LT |
919 | break; |
920 | default: | |
921 | err = sock->ops->ioctl(sock, cmd, arg); | |
b5e5fa5e CH |
922 | |
923 | /* | |
924 | * If this ioctl is unknown try to hand it down | |
925 | * to the NIC driver. | |
926 | */ | |
927 | if (err == -ENOIOCTLCMD) | |
881d966b | 928 | err = dev_ioctl(net, cmd, argp); |
1da177e4 | 929 | break; |
89bddce5 | 930 | } |
1da177e4 LT |
931 | return err; |
932 | } | |
933 | ||
934 | int sock_create_lite(int family, int type, int protocol, struct socket **res) | |
935 | { | |
936 | int err; | |
937 | struct socket *sock = NULL; | |
89bddce5 | 938 | |
1da177e4 LT |
939 | err = security_socket_create(family, type, protocol, 1); |
940 | if (err) | |
941 | goto out; | |
942 | ||
943 | sock = sock_alloc(); | |
944 | if (!sock) { | |
945 | err = -ENOMEM; | |
946 | goto out; | |
947 | } | |
948 | ||
1da177e4 | 949 | sock->type = type; |
7420ed23 VY |
950 | err = security_socket_post_create(sock, family, type, protocol, 1); |
951 | if (err) | |
952 | goto out_release; | |
953 | ||
1da177e4 LT |
954 | out: |
955 | *res = sock; | |
956 | return err; | |
7420ed23 VY |
957 | out_release: |
958 | sock_release(sock); | |
959 | sock = NULL; | |
960 | goto out; | |
1da177e4 LT |
961 | } |
962 | ||
963 | /* No kernel lock held - perfect */ | |
89bddce5 | 964 | static unsigned int sock_poll(struct file *file, poll_table *wait) |
1da177e4 LT |
965 | { |
966 | struct socket *sock; | |
967 | ||
968 | /* | |
89bddce5 | 969 | * We can't return errors to poll, so it's either yes or no. |
1da177e4 | 970 | */ |
b69aee04 | 971 | sock = file->private_data; |
1da177e4 LT |
972 | return sock->ops->poll(file, sock, wait); |
973 | } | |
974 | ||
89bddce5 | 975 | static int sock_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 976 | { |
b69aee04 | 977 | struct socket *sock = file->private_data; |
1da177e4 LT |
978 | |
979 | return sock->ops->mmap(file, sock, vma); | |
980 | } | |
981 | ||
20380731 | 982 | static int sock_close(struct inode *inode, struct file *filp) |
1da177e4 LT |
983 | { |
984 | /* | |
89bddce5 SH |
985 | * It was possible the inode is NULL we were |
986 | * closing an unfinished socket. | |
1da177e4 LT |
987 | */ |
988 | ||
89bddce5 | 989 | if (!inode) { |
1da177e4 LT |
990 | printk(KERN_DEBUG "sock_close: NULL inode\n"); |
991 | return 0; | |
992 | } | |
993 | sock_fasync(-1, filp, 0); | |
994 | sock_release(SOCKET_I(inode)); | |
995 | return 0; | |
996 | } | |
997 | ||
998 | /* | |
999 | * Update the socket async list | |
1000 | * | |
1001 | * Fasync_list locking strategy. | |
1002 | * | |
1003 | * 1. fasync_list is modified only under process context socket lock | |
1004 | * i.e. under semaphore. | |
1005 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) | |
1006 | * or under socket lock. | |
1007 | * 3. fasync_list can be used from softirq context, so that | |
1008 | * modification under socket lock have to be enhanced with | |
1009 | * write_lock_bh(&sk->sk_callback_lock). | |
1010 | * --ANK (990710) | |
1011 | */ | |
1012 | ||
1013 | static int sock_fasync(int fd, struct file *filp, int on) | |
1014 | { | |
89bddce5 | 1015 | struct fasync_struct *fa, *fna = NULL, **prev; |
1da177e4 LT |
1016 | struct socket *sock; |
1017 | struct sock *sk; | |
1018 | ||
89bddce5 | 1019 | if (on) { |
8b3a7005 | 1020 | fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL); |
89bddce5 | 1021 | if (fna == NULL) |
1da177e4 LT |
1022 | return -ENOMEM; |
1023 | } | |
1024 | ||
b69aee04 | 1025 | sock = filp->private_data; |
1da177e4 | 1026 | |
89bddce5 SH |
1027 | sk = sock->sk; |
1028 | if (sk == NULL) { | |
1da177e4 LT |
1029 | kfree(fna); |
1030 | return -EINVAL; | |
1031 | } | |
1032 | ||
1033 | lock_sock(sk); | |
1034 | ||
89bddce5 | 1035 | prev = &(sock->fasync_list); |
1da177e4 | 1036 | |
89bddce5 SH |
1037 | for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev) |
1038 | if (fa->fa_file == filp) | |
1da177e4 LT |
1039 | break; |
1040 | ||
89bddce5 SH |
1041 | if (on) { |
1042 | if (fa != NULL) { | |
1da177e4 | 1043 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1044 | fa->fa_fd = fd; |
1da177e4 LT |
1045 | write_unlock_bh(&sk->sk_callback_lock); |
1046 | ||
1047 | kfree(fna); | |
1048 | goto out; | |
1049 | } | |
89bddce5 SH |
1050 | fna->fa_file = filp; |
1051 | fna->fa_fd = fd; | |
1052 | fna->magic = FASYNC_MAGIC; | |
1053 | fna->fa_next = sock->fasync_list; | |
1da177e4 | 1054 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1055 | sock->fasync_list = fna; |
1da177e4 | 1056 | write_unlock_bh(&sk->sk_callback_lock); |
89bddce5 SH |
1057 | } else { |
1058 | if (fa != NULL) { | |
1da177e4 | 1059 | write_lock_bh(&sk->sk_callback_lock); |
89bddce5 | 1060 | *prev = fa->fa_next; |
1da177e4 LT |
1061 | write_unlock_bh(&sk->sk_callback_lock); |
1062 | kfree(fa); | |
1063 | } | |
1064 | } | |
1065 | ||
1066 | out: | |
1067 | release_sock(sock->sk); | |
1068 | return 0; | |
1069 | } | |
1070 | ||
1071 | /* This function may be called only under socket lock or callback_lock */ | |
1072 | ||
1073 | int sock_wake_async(struct socket *sock, int how, int band) | |
1074 | { | |
1075 | if (!sock || !sock->fasync_list) | |
1076 | return -1; | |
89bddce5 | 1077 | switch (how) { |
8d8ad9d7 | 1078 | case SOCK_WAKE_WAITD: |
1da177e4 LT |
1079 | if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags)) |
1080 | break; | |
1081 | goto call_kill; | |
8d8ad9d7 | 1082 | case SOCK_WAKE_SPACE: |
1da177e4 LT |
1083 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags)) |
1084 | break; | |
1085 | /* fall through */ | |
8d8ad9d7 | 1086 | case SOCK_WAKE_IO: |
89bddce5 | 1087 | call_kill: |
1da177e4 LT |
1088 | __kill_fasync(sock->fasync_list, SIGIO, band); |
1089 | break; | |
8d8ad9d7 | 1090 | case SOCK_WAKE_URG: |
1da177e4 LT |
1091 | __kill_fasync(sock->fasync_list, SIGURG, band); |
1092 | } | |
1093 | return 0; | |
1094 | } | |
1095 | ||
1b8d7ae4 | 1096 | static int __sock_create(struct net *net, int family, int type, int protocol, |
89bddce5 | 1097 | struct socket **res, int kern) |
1da177e4 LT |
1098 | { |
1099 | int err; | |
1100 | struct socket *sock; | |
55737fda | 1101 | const struct net_proto_family *pf; |
1da177e4 LT |
1102 | |
1103 | /* | |
89bddce5 | 1104 | * Check protocol is in range |
1da177e4 LT |
1105 | */ |
1106 | if (family < 0 || family >= NPROTO) | |
1107 | return -EAFNOSUPPORT; | |
1108 | if (type < 0 || type >= SOCK_MAX) | |
1109 | return -EINVAL; | |
1110 | ||
1111 | /* Compatibility. | |
1112 | ||
1113 | This uglymoron is moved from INET layer to here to avoid | |
1114 | deadlock in module load. | |
1115 | */ | |
1116 | if (family == PF_INET && type == SOCK_PACKET) { | |
89bddce5 | 1117 | static int warned; |
1da177e4 LT |
1118 | if (!warned) { |
1119 | warned = 1; | |
89bddce5 SH |
1120 | printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", |
1121 | current->comm); | |
1da177e4 LT |
1122 | } |
1123 | family = PF_PACKET; | |
1124 | } | |
1125 | ||
1126 | err = security_socket_create(family, type, protocol, kern); | |
1127 | if (err) | |
1128 | return err; | |
89bddce5 | 1129 | |
55737fda SH |
1130 | /* |
1131 | * Allocate the socket and allow the family to set things up. if | |
1132 | * the protocol is 0, the family is instructed to select an appropriate | |
1133 | * default. | |
1134 | */ | |
1135 | sock = sock_alloc(); | |
1136 | if (!sock) { | |
1137 | if (net_ratelimit()) | |
1138 | printk(KERN_WARNING "socket: no more sockets\n"); | |
1139 | return -ENFILE; /* Not exactly a match, but its the | |
1140 | closest posix thing */ | |
1141 | } | |
1142 | ||
1143 | sock->type = type; | |
1144 | ||
1da177e4 | 1145 | #if defined(CONFIG_KMOD) |
89bddce5 SH |
1146 | /* Attempt to load a protocol module if the find failed. |
1147 | * | |
1148 | * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user | |
1da177e4 LT |
1149 | * requested real, full-featured networking support upon configuration. |
1150 | * Otherwise module support will break! | |
1151 | */ | |
55737fda | 1152 | if (net_families[family] == NULL) |
89bddce5 | 1153 | request_module("net-pf-%d", family); |
1da177e4 LT |
1154 | #endif |
1155 | ||
55737fda SH |
1156 | rcu_read_lock(); |
1157 | pf = rcu_dereference(net_families[family]); | |
1158 | err = -EAFNOSUPPORT; | |
1159 | if (!pf) | |
1160 | goto out_release; | |
1da177e4 LT |
1161 | |
1162 | /* | |
1163 | * We will call the ->create function, that possibly is in a loadable | |
1164 | * module, so we have to bump that loadable module refcnt first. | |
1165 | */ | |
55737fda | 1166 | if (!try_module_get(pf->owner)) |
1da177e4 LT |
1167 | goto out_release; |
1168 | ||
55737fda SH |
1169 | /* Now protected by module ref count */ |
1170 | rcu_read_unlock(); | |
1171 | ||
1b8d7ae4 | 1172 | err = pf->create(net, sock, protocol); |
55737fda | 1173 | if (err < 0) |
1da177e4 | 1174 | goto out_module_put; |
a79af59e | 1175 | |
1da177e4 LT |
1176 | /* |
1177 | * Now to bump the refcnt of the [loadable] module that owns this | |
1178 | * socket at sock_release time we decrement its refcnt. | |
1179 | */ | |
55737fda SH |
1180 | if (!try_module_get(sock->ops->owner)) |
1181 | goto out_module_busy; | |
1182 | ||
1da177e4 LT |
1183 | /* |
1184 | * Now that we're done with the ->create function, the [loadable] | |
1185 | * module can have its refcnt decremented | |
1186 | */ | |
55737fda | 1187 | module_put(pf->owner); |
7420ed23 VY |
1188 | err = security_socket_post_create(sock, family, type, protocol, kern); |
1189 | if (err) | |
3b185525 | 1190 | goto out_sock_release; |
55737fda | 1191 | *res = sock; |
1da177e4 | 1192 | |
55737fda SH |
1193 | return 0; |
1194 | ||
1195 | out_module_busy: | |
1196 | err = -EAFNOSUPPORT; | |
1da177e4 | 1197 | out_module_put: |
55737fda SH |
1198 | sock->ops = NULL; |
1199 | module_put(pf->owner); | |
1200 | out_sock_release: | |
1da177e4 | 1201 | sock_release(sock); |
55737fda SH |
1202 | return err; |
1203 | ||
1204 | out_release: | |
1205 | rcu_read_unlock(); | |
1206 | goto out_sock_release; | |
1da177e4 LT |
1207 | } |
1208 | ||
1209 | int sock_create(int family, int type, int protocol, struct socket **res) | |
1210 | { | |
1b8d7ae4 | 1211 | return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); |
1da177e4 LT |
1212 | } |
1213 | ||
1214 | int sock_create_kern(int family, int type, int protocol, struct socket **res) | |
1215 | { | |
1b8d7ae4 | 1216 | return __sock_create(&init_net, family, type, protocol, res, 1); |
1da177e4 LT |
1217 | } |
1218 | ||
1219 | asmlinkage long sys_socket(int family, int type, int protocol) | |
1220 | { | |
1221 | int retval; | |
1222 | struct socket *sock; | |
a677a039 UD |
1223 | int flags; |
1224 | ||
e38b36f3 UD |
1225 | /* Check the SOCK_* constants for consistency. */ |
1226 | BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); | |
1227 | BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); | |
1228 | BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); | |
1229 | BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); | |
1230 | ||
a677a039 | 1231 | flags = type & ~SOCK_TYPE_MASK; |
77d27200 | 1232 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
a677a039 UD |
1233 | return -EINVAL; |
1234 | type &= SOCK_TYPE_MASK; | |
1da177e4 | 1235 | |
aaca0bdc UD |
1236 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1237 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1238 | ||
1da177e4 LT |
1239 | retval = sock_create(family, type, protocol, &sock); |
1240 | if (retval < 0) | |
1241 | goto out; | |
1242 | ||
77d27200 | 1243 | retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); |
1da177e4 LT |
1244 | if (retval < 0) |
1245 | goto out_release; | |
1246 | ||
1247 | out: | |
1248 | /* It may be already another descriptor 8) Not kernel problem. */ | |
1249 | return retval; | |
1250 | ||
1251 | out_release: | |
1252 | sock_release(sock); | |
1253 | return retval; | |
1254 | } | |
1255 | ||
1256 | /* | |
1257 | * Create a pair of connected sockets. | |
1258 | */ | |
1259 | ||
89bddce5 SH |
1260 | asmlinkage long sys_socketpair(int family, int type, int protocol, |
1261 | int __user *usockvec) | |
1da177e4 LT |
1262 | { |
1263 | struct socket *sock1, *sock2; | |
1264 | int fd1, fd2, err; | |
db349509 | 1265 | struct file *newfile1, *newfile2; |
a677a039 UD |
1266 | int flags; |
1267 | ||
1268 | flags = type & ~SOCK_TYPE_MASK; | |
77d27200 | 1269 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
a677a039 UD |
1270 | return -EINVAL; |
1271 | type &= SOCK_TYPE_MASK; | |
1da177e4 | 1272 | |
aaca0bdc UD |
1273 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1274 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1275 | ||
1da177e4 LT |
1276 | /* |
1277 | * Obtain the first socket and check if the underlying protocol | |
1278 | * supports the socketpair call. | |
1279 | */ | |
1280 | ||
1281 | err = sock_create(family, type, protocol, &sock1); | |
1282 | if (err < 0) | |
1283 | goto out; | |
1284 | ||
1285 | err = sock_create(family, type, protocol, &sock2); | |
1286 | if (err < 0) | |
1287 | goto out_release_1; | |
1288 | ||
1289 | err = sock1->ops->socketpair(sock1, sock2); | |
89bddce5 | 1290 | if (err < 0) |
1da177e4 LT |
1291 | goto out_release_both; |
1292 | ||
a677a039 | 1293 | fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC); |
bf3c23d1 DM |
1294 | if (unlikely(fd1 < 0)) { |
1295 | err = fd1; | |
db349509 | 1296 | goto out_release_both; |
bf3c23d1 | 1297 | } |
1da177e4 | 1298 | |
a677a039 | 1299 | fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC); |
db349509 | 1300 | if (unlikely(fd2 < 0)) { |
bf3c23d1 | 1301 | err = fd2; |
db349509 AV |
1302 | put_filp(newfile1); |
1303 | put_unused_fd(fd1); | |
1da177e4 | 1304 | goto out_release_both; |
db349509 | 1305 | } |
1da177e4 | 1306 | |
77d27200 | 1307 | err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK); |
db349509 AV |
1308 | if (unlikely(err < 0)) { |
1309 | goto out_fd2; | |
1310 | } | |
1311 | ||
77d27200 | 1312 | err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK); |
db349509 AV |
1313 | if (unlikely(err < 0)) { |
1314 | fput(newfile1); | |
1315 | goto out_fd1; | |
1316 | } | |
1317 | ||
1318 | err = audit_fd_pair(fd1, fd2); | |
1319 | if (err < 0) { | |
1320 | fput(newfile1); | |
1321 | fput(newfile2); | |
1322 | goto out_fd; | |
1323 | } | |
1da177e4 | 1324 | |
db349509 AV |
1325 | fd_install(fd1, newfile1); |
1326 | fd_install(fd2, newfile2); | |
1da177e4 LT |
1327 | /* fd1 and fd2 may be already another descriptors. |
1328 | * Not kernel problem. | |
1329 | */ | |
1330 | ||
89bddce5 | 1331 | err = put_user(fd1, &usockvec[0]); |
1da177e4 LT |
1332 | if (!err) |
1333 | err = put_user(fd2, &usockvec[1]); | |
1334 | if (!err) | |
1335 | return 0; | |
1336 | ||
1337 | sys_close(fd2); | |
1338 | sys_close(fd1); | |
1339 | return err; | |
1340 | ||
1da177e4 | 1341 | out_release_both: |
89bddce5 | 1342 | sock_release(sock2); |
1da177e4 | 1343 | out_release_1: |
89bddce5 | 1344 | sock_release(sock1); |
1da177e4 LT |
1345 | out: |
1346 | return err; | |
db349509 AV |
1347 | |
1348 | out_fd2: | |
1349 | put_filp(newfile1); | |
1350 | sock_release(sock1); | |
1351 | out_fd1: | |
1352 | put_filp(newfile2); | |
1353 | sock_release(sock2); | |
1354 | out_fd: | |
1355 | put_unused_fd(fd1); | |
1356 | put_unused_fd(fd2); | |
1357 | goto out; | |
1da177e4 LT |
1358 | } |
1359 | ||
1da177e4 LT |
1360 | /* |
1361 | * Bind a name to a socket. Nothing much to do here since it's | |
1362 | * the protocol's responsibility to handle the local address. | |
1363 | * | |
1364 | * We move the socket address to kernel space before we call | |
1365 | * the protocol layer (having also checked the address is ok). | |
1366 | */ | |
1367 | ||
1368 | asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) | |
1369 | { | |
1370 | struct socket *sock; | |
230b1839 | 1371 | struct sockaddr_storage address; |
6cb153ca | 1372 | int err, fput_needed; |
1da177e4 | 1373 | |
89bddce5 | 1374 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
e71a4783 | 1375 | if (sock) { |
230b1839 | 1376 | err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address); |
89bddce5 SH |
1377 | if (err >= 0) { |
1378 | err = security_socket_bind(sock, | |
230b1839 | 1379 | (struct sockaddr *)&address, |
89bddce5 | 1380 | addrlen); |
6cb153ca BL |
1381 | if (!err) |
1382 | err = sock->ops->bind(sock, | |
89bddce5 | 1383 | (struct sockaddr *) |
230b1839 | 1384 | &address, addrlen); |
1da177e4 | 1385 | } |
6cb153ca | 1386 | fput_light(sock->file, fput_needed); |
89bddce5 | 1387 | } |
1da177e4 LT |
1388 | return err; |
1389 | } | |
1390 | ||
1da177e4 LT |
1391 | /* |
1392 | * Perform a listen. Basically, we allow the protocol to do anything | |
1393 | * necessary for a listen, and if that works, we mark the socket as | |
1394 | * ready for listening. | |
1395 | */ | |
1396 | ||
1da177e4 LT |
1397 | asmlinkage long sys_listen(int fd, int backlog) |
1398 | { | |
1399 | struct socket *sock; | |
6cb153ca | 1400 | int err, fput_needed; |
b8e1f9b5 | 1401 | int somaxconn; |
89bddce5 SH |
1402 | |
1403 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1404 | if (sock) { | |
8efa6e93 | 1405 | somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; |
b8e1f9b5 PE |
1406 | if ((unsigned)backlog > somaxconn) |
1407 | backlog = somaxconn; | |
1da177e4 LT |
1408 | |
1409 | err = security_socket_listen(sock, backlog); | |
6cb153ca BL |
1410 | if (!err) |
1411 | err = sock->ops->listen(sock, backlog); | |
1da177e4 | 1412 | |
6cb153ca | 1413 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1414 | } |
1415 | return err; | |
1416 | } | |
1417 | ||
1da177e4 LT |
1418 | /* |
1419 | * For accept, we attempt to create a new socket, set up the link | |
1420 | * with the client, wake up the client, then return the new | |
1421 | * connected fd. We collect the address of the connector in kernel | |
1422 | * space and move it to user at the very end. This is unclean because | |
1423 | * we open the socket then return an error. | |
1424 | * | |
1425 | * 1003.1g adds the ability to recvmsg() to query connection pending | |
1426 | * status to recvmsg. We need to add that support in a way thats | |
1427 | * clean when we restucture accept also. | |
1428 | */ | |
1429 | ||
aaca0bdc UD |
1430 | long do_accept(int fd, struct sockaddr __user *upeer_sockaddr, |
1431 | int __user *upeer_addrlen, int flags) | |
1da177e4 LT |
1432 | { |
1433 | struct socket *sock, *newsock; | |
39d8c1b6 | 1434 | struct file *newfile; |
6cb153ca | 1435 | int err, len, newfd, fput_needed; |
230b1839 | 1436 | struct sockaddr_storage address; |
1da177e4 | 1437 | |
77d27200 | 1438 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
aaca0bdc UD |
1439 | return -EINVAL; |
1440 | ||
1441 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) | |
1442 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1443 | ||
6cb153ca | 1444 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1445 | if (!sock) |
1446 | goto out; | |
1447 | ||
1448 | err = -ENFILE; | |
89bddce5 | 1449 | if (!(newsock = sock_alloc())) |
1da177e4 LT |
1450 | goto out_put; |
1451 | ||
1452 | newsock->type = sock->type; | |
1453 | newsock->ops = sock->ops; | |
1454 | ||
1da177e4 LT |
1455 | /* |
1456 | * We don't need try_module_get here, as the listening socket (sock) | |
1457 | * has the protocol module (sock->ops->owner) held. | |
1458 | */ | |
1459 | __module_get(newsock->ops->owner); | |
1460 | ||
aaca0bdc | 1461 | newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC); |
39d8c1b6 DM |
1462 | if (unlikely(newfd < 0)) { |
1463 | err = newfd; | |
9a1875e6 DM |
1464 | sock_release(newsock); |
1465 | goto out_put; | |
39d8c1b6 DM |
1466 | } |
1467 | ||
77d27200 | 1468 | err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK); |
39d8c1b6 | 1469 | if (err < 0) |
79f4f642 | 1470 | goto out_fd_simple; |
39d8c1b6 | 1471 | |
a79af59e FF |
1472 | err = security_socket_accept(sock, newsock); |
1473 | if (err) | |
39d8c1b6 | 1474 | goto out_fd; |
a79af59e | 1475 | |
1da177e4 LT |
1476 | err = sock->ops->accept(sock, newsock, sock->file->f_flags); |
1477 | if (err < 0) | |
39d8c1b6 | 1478 | goto out_fd; |
1da177e4 LT |
1479 | |
1480 | if (upeer_sockaddr) { | |
230b1839 | 1481 | if (newsock->ops->getname(newsock, (struct sockaddr *)&address, |
89bddce5 | 1482 | &len, 2) < 0) { |
1da177e4 | 1483 | err = -ECONNABORTED; |
39d8c1b6 | 1484 | goto out_fd; |
1da177e4 | 1485 | } |
230b1839 YH |
1486 | err = move_addr_to_user((struct sockaddr *)&address, |
1487 | len, upeer_sockaddr, upeer_addrlen); | |
1da177e4 | 1488 | if (err < 0) |
39d8c1b6 | 1489 | goto out_fd; |
1da177e4 LT |
1490 | } |
1491 | ||
1492 | /* File flags are not inherited via accept() unlike another OSes. */ | |
1493 | ||
39d8c1b6 DM |
1494 | fd_install(newfd, newfile); |
1495 | err = newfd; | |
1da177e4 LT |
1496 | |
1497 | security_socket_post_accept(sock, newsock); | |
1498 | ||
1499 | out_put: | |
6cb153ca | 1500 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1501 | out: |
1502 | return err; | |
79f4f642 AD |
1503 | out_fd_simple: |
1504 | sock_release(newsock); | |
1505 | put_filp(newfile); | |
1506 | put_unused_fd(newfd); | |
1507 | goto out_put; | |
39d8c1b6 | 1508 | out_fd: |
9606a216 | 1509 | fput(newfile); |
39d8c1b6 | 1510 | put_unused_fd(newfd); |
1da177e4 LT |
1511 | goto out_put; |
1512 | } | |
1513 | ||
c019bbc6 | 1514 | #ifdef HAVE_SET_RESTORE_SIGMASK |
aaca0bdc UD |
1515 | asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr, |
1516 | int __user *upeer_addrlen, | |
1517 | const sigset_t __user *sigmask, | |
1518 | size_t sigsetsize, int flags) | |
1519 | { | |
1520 | sigset_t ksigmask, sigsaved; | |
1521 | int ret; | |
1522 | ||
1523 | if (sigmask) { | |
1524 | /* XXX: Don't preclude handling different sized sigset_t's. */ | |
1525 | if (sigsetsize != sizeof(sigset_t)) | |
1526 | return -EINVAL; | |
1527 | if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask))) | |
1528 | return -EFAULT; | |
1529 | ||
1530 | sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1531 | sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved); | |
1532 | } | |
1533 | ||
1534 | ret = do_accept(fd, upeer_sockaddr, upeer_addrlen, flags); | |
1535 | ||
1536 | if (ret < 0 && signal_pending(current)) { | |
1537 | /* | |
1538 | * Don't restore the signal mask yet. Let do_signal() deliver | |
1539 | * the signal on the way back to userspace, before the signal | |
1540 | * mask is restored. | |
1541 | */ | |
1542 | if (sigmask) { | |
1543 | memcpy(¤t->saved_sigmask, &sigsaved, | |
1544 | sizeof(sigsaved)); | |
1545 | set_restore_sigmask(); | |
1546 | } | |
1547 | } else if (sigmask) | |
1548 | sigprocmask(SIG_SETMASK, &sigsaved, NULL); | |
1549 | ||
1550 | return ret; | |
1551 | } | |
c019bbc6 UD |
1552 | #else |
1553 | asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr, | |
1554 | int __user *upeer_addrlen, | |
1555 | const sigset_t __user *sigmask, | |
1556 | size_t sigsetsize, int flags) | |
1557 | { | |
1558 | /* The platform does not support restoring the signal mask in the | |
1559 | * return path. So we do not allow using paccept() with a signal | |
1560 | * mask. */ | |
1561 | if (sigmask) | |
1562 | return -EINVAL; | |
1563 | ||
1564 | return do_accept(fd, upeer_sockaddr, upeer_addrlen, flags); | |
1565 | } | |
1566 | #endif | |
aaca0bdc UD |
1567 | |
1568 | asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, | |
1569 | int __user *upeer_addrlen) | |
1570 | { | |
1571 | return do_accept(fd, upeer_sockaddr, upeer_addrlen, 0); | |
1572 | } | |
1573 | ||
1da177e4 LT |
1574 | /* |
1575 | * Attempt to connect to a socket with the server address. The address | |
1576 | * is in user space so we verify it is OK and move it to kernel space. | |
1577 | * | |
1578 | * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to | |
1579 | * break bindings | |
1580 | * | |
1581 | * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and | |
1582 | * other SEQPACKET protocols that take time to connect() as it doesn't | |
1583 | * include the -EINPROGRESS status for such sockets. | |
1584 | */ | |
1585 | ||
89bddce5 SH |
1586 | asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, |
1587 | int addrlen) | |
1da177e4 LT |
1588 | { |
1589 | struct socket *sock; | |
230b1839 | 1590 | struct sockaddr_storage address; |
6cb153ca | 1591 | int err, fput_needed; |
1da177e4 | 1592 | |
6cb153ca | 1593 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1594 | if (!sock) |
1595 | goto out; | |
230b1839 | 1596 | err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address); |
1da177e4 LT |
1597 | if (err < 0) |
1598 | goto out_put; | |
1599 | ||
89bddce5 | 1600 | err = |
230b1839 | 1601 | security_socket_connect(sock, (struct sockaddr *)&address, addrlen); |
1da177e4 LT |
1602 | if (err) |
1603 | goto out_put; | |
1604 | ||
230b1839 | 1605 | err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, |
1da177e4 LT |
1606 | sock->file->f_flags); |
1607 | out_put: | |
6cb153ca | 1608 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1609 | out: |
1610 | return err; | |
1611 | } | |
1612 | ||
1613 | /* | |
1614 | * Get the local address ('name') of a socket object. Move the obtained | |
1615 | * name to user space. | |
1616 | */ | |
1617 | ||
89bddce5 SH |
1618 | asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, |
1619 | int __user *usockaddr_len) | |
1da177e4 LT |
1620 | { |
1621 | struct socket *sock; | |
230b1839 | 1622 | struct sockaddr_storage address; |
6cb153ca | 1623 | int len, err, fput_needed; |
89bddce5 | 1624 | |
6cb153ca | 1625 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
1626 | if (!sock) |
1627 | goto out; | |
1628 | ||
1629 | err = security_socket_getsockname(sock); | |
1630 | if (err) | |
1631 | goto out_put; | |
1632 | ||
230b1839 | 1633 | err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0); |
1da177e4 LT |
1634 | if (err) |
1635 | goto out_put; | |
230b1839 | 1636 | err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len); |
1da177e4 LT |
1637 | |
1638 | out_put: | |
6cb153ca | 1639 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1640 | out: |
1641 | return err; | |
1642 | } | |
1643 | ||
1644 | /* | |
1645 | * Get the remote address ('name') of a socket object. Move the obtained | |
1646 | * name to user space. | |
1647 | */ | |
1648 | ||
89bddce5 SH |
1649 | asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, |
1650 | int __user *usockaddr_len) | |
1da177e4 LT |
1651 | { |
1652 | struct socket *sock; | |
230b1839 | 1653 | struct sockaddr_storage address; |
6cb153ca | 1654 | int len, err, fput_needed; |
1da177e4 | 1655 | |
89bddce5 SH |
1656 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1657 | if (sock != NULL) { | |
1da177e4 LT |
1658 | err = security_socket_getpeername(sock); |
1659 | if (err) { | |
6cb153ca | 1660 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1661 | return err; |
1662 | } | |
1663 | ||
89bddce5 | 1664 | err = |
230b1839 | 1665 | sock->ops->getname(sock, (struct sockaddr *)&address, &len, |
89bddce5 | 1666 | 1); |
1da177e4 | 1667 | if (!err) |
230b1839 | 1668 | err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, |
89bddce5 | 1669 | usockaddr_len); |
6cb153ca | 1670 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1671 | } |
1672 | return err; | |
1673 | } | |
1674 | ||
1675 | /* | |
1676 | * Send a datagram to a given address. We move the address into kernel | |
1677 | * space and check the user space data area is readable before invoking | |
1678 | * the protocol. | |
1679 | */ | |
1680 | ||
89bddce5 SH |
1681 | asmlinkage long sys_sendto(int fd, void __user *buff, size_t len, |
1682 | unsigned flags, struct sockaddr __user *addr, | |
1683 | int addr_len) | |
1da177e4 LT |
1684 | { |
1685 | struct socket *sock; | |
230b1839 | 1686 | struct sockaddr_storage address; |
1da177e4 LT |
1687 | int err; |
1688 | struct msghdr msg; | |
1689 | struct iovec iov; | |
6cb153ca | 1690 | int fput_needed; |
6cb153ca | 1691 | |
de0fa95c PE |
1692 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1693 | if (!sock) | |
4387ff75 | 1694 | goto out; |
6cb153ca | 1695 | |
89bddce5 SH |
1696 | iov.iov_base = buff; |
1697 | iov.iov_len = len; | |
1698 | msg.msg_name = NULL; | |
1699 | msg.msg_iov = &iov; | |
1700 | msg.msg_iovlen = 1; | |
1701 | msg.msg_control = NULL; | |
1702 | msg.msg_controllen = 0; | |
1703 | msg.msg_namelen = 0; | |
6cb153ca | 1704 | if (addr) { |
230b1839 | 1705 | err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address); |
1da177e4 LT |
1706 | if (err < 0) |
1707 | goto out_put; | |
230b1839 | 1708 | msg.msg_name = (struct sockaddr *)&address; |
89bddce5 | 1709 | msg.msg_namelen = addr_len; |
1da177e4 LT |
1710 | } |
1711 | if (sock->file->f_flags & O_NONBLOCK) | |
1712 | flags |= MSG_DONTWAIT; | |
1713 | msg.msg_flags = flags; | |
1714 | err = sock_sendmsg(sock, &msg, len); | |
1715 | ||
89bddce5 | 1716 | out_put: |
de0fa95c | 1717 | fput_light(sock->file, fput_needed); |
4387ff75 | 1718 | out: |
1da177e4 LT |
1719 | return err; |
1720 | } | |
1721 | ||
1722 | /* | |
89bddce5 | 1723 | * Send a datagram down a socket. |
1da177e4 LT |
1724 | */ |
1725 | ||
89bddce5 | 1726 | asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags) |
1da177e4 LT |
1727 | { |
1728 | return sys_sendto(fd, buff, len, flags, NULL, 0); | |
1729 | } | |
1730 | ||
1731 | /* | |
89bddce5 | 1732 | * Receive a frame from the socket and optionally record the address of the |
1da177e4 LT |
1733 | * sender. We verify the buffers are writable and if needed move the |
1734 | * sender address from kernel to user space. | |
1735 | */ | |
1736 | ||
89bddce5 SH |
1737 | asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size, |
1738 | unsigned flags, struct sockaddr __user *addr, | |
1739 | int __user *addr_len) | |
1da177e4 LT |
1740 | { |
1741 | struct socket *sock; | |
1742 | struct iovec iov; | |
1743 | struct msghdr msg; | |
230b1839 | 1744 | struct sockaddr_storage address; |
89bddce5 | 1745 | int err, err2; |
6cb153ca BL |
1746 | int fput_needed; |
1747 | ||
de0fa95c | 1748 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 | 1749 | if (!sock) |
de0fa95c | 1750 | goto out; |
1da177e4 | 1751 | |
89bddce5 SH |
1752 | msg.msg_control = NULL; |
1753 | msg.msg_controllen = 0; | |
1754 | msg.msg_iovlen = 1; | |
1755 | msg.msg_iov = &iov; | |
1756 | iov.iov_len = size; | |
1757 | iov.iov_base = ubuf; | |
230b1839 YH |
1758 | msg.msg_name = (struct sockaddr *)&address; |
1759 | msg.msg_namelen = sizeof(address); | |
1da177e4 LT |
1760 | if (sock->file->f_flags & O_NONBLOCK) |
1761 | flags |= MSG_DONTWAIT; | |
89bddce5 | 1762 | err = sock_recvmsg(sock, &msg, size, flags); |
1da177e4 | 1763 | |
89bddce5 | 1764 | if (err >= 0 && addr != NULL) { |
230b1839 YH |
1765 | err2 = move_addr_to_user((struct sockaddr *)&address, |
1766 | msg.msg_namelen, addr, addr_len); | |
89bddce5 SH |
1767 | if (err2 < 0) |
1768 | err = err2; | |
1da177e4 | 1769 | } |
de0fa95c PE |
1770 | |
1771 | fput_light(sock->file, fput_needed); | |
4387ff75 | 1772 | out: |
1da177e4 LT |
1773 | return err; |
1774 | } | |
1775 | ||
1776 | /* | |
89bddce5 | 1777 | * Receive a datagram from a socket. |
1da177e4 LT |
1778 | */ |
1779 | ||
89bddce5 SH |
1780 | asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size, |
1781 | unsigned flags) | |
1da177e4 LT |
1782 | { |
1783 | return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); | |
1784 | } | |
1785 | ||
1786 | /* | |
1787 | * Set a socket option. Because we don't know the option lengths we have | |
1788 | * to pass the user mode parameter for the protocols to sort out. | |
1789 | */ | |
1790 | ||
89bddce5 SH |
1791 | asmlinkage long sys_setsockopt(int fd, int level, int optname, |
1792 | char __user *optval, int optlen) | |
1da177e4 | 1793 | { |
6cb153ca | 1794 | int err, fput_needed; |
1da177e4 LT |
1795 | struct socket *sock; |
1796 | ||
1797 | if (optlen < 0) | |
1798 | return -EINVAL; | |
89bddce5 SH |
1799 | |
1800 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1801 | if (sock != NULL) { | |
1802 | err = security_socket_setsockopt(sock, level, optname); | |
6cb153ca BL |
1803 | if (err) |
1804 | goto out_put; | |
1da177e4 LT |
1805 | |
1806 | if (level == SOL_SOCKET) | |
89bddce5 SH |
1807 | err = |
1808 | sock_setsockopt(sock, level, optname, optval, | |
1809 | optlen); | |
1da177e4 | 1810 | else |
89bddce5 SH |
1811 | err = |
1812 | sock->ops->setsockopt(sock, level, optname, optval, | |
1813 | optlen); | |
6cb153ca BL |
1814 | out_put: |
1815 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1816 | } |
1817 | return err; | |
1818 | } | |
1819 | ||
1820 | /* | |
1821 | * Get a socket option. Because we don't know the option lengths we have | |
1822 | * to pass a user mode parameter for the protocols to sort out. | |
1823 | */ | |
1824 | ||
89bddce5 SH |
1825 | asmlinkage long sys_getsockopt(int fd, int level, int optname, |
1826 | char __user *optval, int __user *optlen) | |
1da177e4 | 1827 | { |
6cb153ca | 1828 | int err, fput_needed; |
1da177e4 LT |
1829 | struct socket *sock; |
1830 | ||
89bddce5 SH |
1831 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1832 | if (sock != NULL) { | |
6cb153ca BL |
1833 | err = security_socket_getsockopt(sock, level, optname); |
1834 | if (err) | |
1835 | goto out_put; | |
1da177e4 LT |
1836 | |
1837 | if (level == SOL_SOCKET) | |
89bddce5 SH |
1838 | err = |
1839 | sock_getsockopt(sock, level, optname, optval, | |
1840 | optlen); | |
1da177e4 | 1841 | else |
89bddce5 SH |
1842 | err = |
1843 | sock->ops->getsockopt(sock, level, optname, optval, | |
1844 | optlen); | |
6cb153ca BL |
1845 | out_put: |
1846 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1847 | } |
1848 | return err; | |
1849 | } | |
1850 | ||
1da177e4 LT |
1851 | /* |
1852 | * Shutdown a socket. | |
1853 | */ | |
1854 | ||
1855 | asmlinkage long sys_shutdown(int fd, int how) | |
1856 | { | |
6cb153ca | 1857 | int err, fput_needed; |
1da177e4 LT |
1858 | struct socket *sock; |
1859 | ||
89bddce5 SH |
1860 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1861 | if (sock != NULL) { | |
1da177e4 | 1862 | err = security_socket_shutdown(sock, how); |
6cb153ca BL |
1863 | if (!err) |
1864 | err = sock->ops->shutdown(sock, how); | |
1865 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
1866 | } |
1867 | return err; | |
1868 | } | |
1869 | ||
89bddce5 | 1870 | /* A couple of helpful macros for getting the address of the 32/64 bit |
1da177e4 LT |
1871 | * fields which are the same type (int / unsigned) on our platforms. |
1872 | */ | |
1873 | #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) | |
1874 | #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) | |
1875 | #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) | |
1876 | ||
1da177e4 LT |
1877 | /* |
1878 | * BSD sendmsg interface | |
1879 | */ | |
1880 | ||
1881 | asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags) | |
1882 | { | |
89bddce5 SH |
1883 | struct compat_msghdr __user *msg_compat = |
1884 | (struct compat_msghdr __user *)msg; | |
1da177e4 | 1885 | struct socket *sock; |
230b1839 | 1886 | struct sockaddr_storage address; |
1da177e4 | 1887 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; |
b9d717a7 | 1888 | unsigned char ctl[sizeof(struct cmsghdr) + 20] |
89bddce5 SH |
1889 | __attribute__ ((aligned(sizeof(__kernel_size_t)))); |
1890 | /* 20 is size of ipv6_pktinfo */ | |
1da177e4 LT |
1891 | unsigned char *ctl_buf = ctl; |
1892 | struct msghdr msg_sys; | |
1893 | int err, ctl_len, iov_size, total_len; | |
6cb153ca | 1894 | int fput_needed; |
89bddce5 | 1895 | |
1da177e4 LT |
1896 | err = -EFAULT; |
1897 | if (MSG_CMSG_COMPAT & flags) { | |
1898 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
1899 | return -EFAULT; | |
89bddce5 SH |
1900 | } |
1901 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
1da177e4 LT |
1902 | return -EFAULT; |
1903 | ||
6cb153ca | 1904 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
89bddce5 | 1905 | if (!sock) |
1da177e4 LT |
1906 | goto out; |
1907 | ||
1908 | /* do not move before msg_sys is valid */ | |
1909 | err = -EMSGSIZE; | |
1910 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
1911 | goto out_put; | |
1912 | ||
89bddce5 | 1913 | /* Check whether to allocate the iovec area */ |
1da177e4 LT |
1914 | err = -ENOMEM; |
1915 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
1916 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
1917 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
1918 | if (!iov) | |
1919 | goto out_put; | |
1920 | } | |
1921 | ||
1922 | /* This will also move the address data into kernel space */ | |
1923 | if (MSG_CMSG_COMPAT & flags) { | |
230b1839 YH |
1924 | err = verify_compat_iovec(&msg_sys, iov, |
1925 | (struct sockaddr *)&address, | |
1926 | VERIFY_READ); | |
1da177e4 | 1927 | } else |
230b1839 YH |
1928 | err = verify_iovec(&msg_sys, iov, |
1929 | (struct sockaddr *)&address, | |
1930 | VERIFY_READ); | |
89bddce5 | 1931 | if (err < 0) |
1da177e4 LT |
1932 | goto out_freeiov; |
1933 | total_len = err; | |
1934 | ||
1935 | err = -ENOBUFS; | |
1936 | ||
1937 | if (msg_sys.msg_controllen > INT_MAX) | |
1938 | goto out_freeiov; | |
89bddce5 | 1939 | ctl_len = msg_sys.msg_controllen; |
1da177e4 | 1940 | if ((MSG_CMSG_COMPAT & flags) && ctl_len) { |
89bddce5 SH |
1941 | err = |
1942 | cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, | |
1943 | sizeof(ctl)); | |
1da177e4 LT |
1944 | if (err) |
1945 | goto out_freeiov; | |
1946 | ctl_buf = msg_sys.msg_control; | |
8920e8f9 | 1947 | ctl_len = msg_sys.msg_controllen; |
1da177e4 | 1948 | } else if (ctl_len) { |
89bddce5 | 1949 | if (ctl_len > sizeof(ctl)) { |
1da177e4 | 1950 | ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); |
89bddce5 | 1951 | if (ctl_buf == NULL) |
1da177e4 LT |
1952 | goto out_freeiov; |
1953 | } | |
1954 | err = -EFAULT; | |
1955 | /* | |
1956 | * Careful! Before this, msg_sys.msg_control contains a user pointer. | |
1957 | * Afterwards, it will be a kernel pointer. Thus the compiler-assisted | |
1958 | * checking falls down on this. | |
1959 | */ | |
89bddce5 SH |
1960 | if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control, |
1961 | ctl_len)) | |
1da177e4 LT |
1962 | goto out_freectl; |
1963 | msg_sys.msg_control = ctl_buf; | |
1964 | } | |
1965 | msg_sys.msg_flags = flags; | |
1966 | ||
1967 | if (sock->file->f_flags & O_NONBLOCK) | |
1968 | msg_sys.msg_flags |= MSG_DONTWAIT; | |
1969 | err = sock_sendmsg(sock, &msg_sys, total_len); | |
1970 | ||
1971 | out_freectl: | |
89bddce5 | 1972 | if (ctl_buf != ctl) |
1da177e4 LT |
1973 | sock_kfree_s(sock->sk, ctl_buf, ctl_len); |
1974 | out_freeiov: | |
1975 | if (iov != iovstack) | |
1976 | sock_kfree_s(sock->sk, iov, iov_size); | |
1977 | out_put: | |
6cb153ca | 1978 | fput_light(sock->file, fput_needed); |
89bddce5 | 1979 | out: |
1da177e4 LT |
1980 | return err; |
1981 | } | |
1982 | ||
1983 | /* | |
1984 | * BSD recvmsg interface | |
1985 | */ | |
1986 | ||
89bddce5 SH |
1987 | asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, |
1988 | unsigned int flags) | |
1da177e4 | 1989 | { |
89bddce5 SH |
1990 | struct compat_msghdr __user *msg_compat = |
1991 | (struct compat_msghdr __user *)msg; | |
1da177e4 LT |
1992 | struct socket *sock; |
1993 | struct iovec iovstack[UIO_FASTIOV]; | |
89bddce5 | 1994 | struct iovec *iov = iovstack; |
1da177e4 LT |
1995 | struct msghdr msg_sys; |
1996 | unsigned long cmsg_ptr; | |
1997 | int err, iov_size, total_len, len; | |
6cb153ca | 1998 | int fput_needed; |
1da177e4 LT |
1999 | |
2000 | /* kernel mode address */ | |
230b1839 | 2001 | struct sockaddr_storage addr; |
1da177e4 LT |
2002 | |
2003 | /* user mode address pointers */ | |
2004 | struct sockaddr __user *uaddr; | |
2005 | int __user *uaddr_len; | |
89bddce5 | 2006 | |
1da177e4 LT |
2007 | if (MSG_CMSG_COMPAT & flags) { |
2008 | if (get_compat_msghdr(&msg_sys, msg_compat)) | |
2009 | return -EFAULT; | |
89bddce5 SH |
2010 | } |
2011 | else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) | |
2012 | return -EFAULT; | |
1da177e4 | 2013 | |
6cb153ca | 2014 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
2015 | if (!sock) |
2016 | goto out; | |
2017 | ||
2018 | err = -EMSGSIZE; | |
2019 | if (msg_sys.msg_iovlen > UIO_MAXIOV) | |
2020 | goto out_put; | |
89bddce5 SH |
2021 | |
2022 | /* Check whether to allocate the iovec area */ | |
1da177e4 LT |
2023 | err = -ENOMEM; |
2024 | iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); | |
2025 | if (msg_sys.msg_iovlen > UIO_FASTIOV) { | |
2026 | iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); | |
2027 | if (!iov) | |
2028 | goto out_put; | |
2029 | } | |
2030 | ||
2031 | /* | |
89bddce5 SH |
2032 | * Save the user-mode address (verify_iovec will change the |
2033 | * kernel msghdr to use the kernel address space) | |
1da177e4 | 2034 | */ |
89bddce5 | 2035 | |
cfcabdcc | 2036 | uaddr = (__force void __user *)msg_sys.msg_name; |
1da177e4 LT |
2037 | uaddr_len = COMPAT_NAMELEN(msg); |
2038 | if (MSG_CMSG_COMPAT & flags) { | |
230b1839 YH |
2039 | err = verify_compat_iovec(&msg_sys, iov, |
2040 | (struct sockaddr *)&addr, | |
2041 | VERIFY_WRITE); | |
1da177e4 | 2042 | } else |
230b1839 YH |
2043 | err = verify_iovec(&msg_sys, iov, |
2044 | (struct sockaddr *)&addr, | |
2045 | VERIFY_WRITE); | |
1da177e4 LT |
2046 | if (err < 0) |
2047 | goto out_freeiov; | |
89bddce5 | 2048 | total_len = err; |
1da177e4 LT |
2049 | |
2050 | cmsg_ptr = (unsigned long)msg_sys.msg_control; | |
4a19542e | 2051 | msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); |
89bddce5 | 2052 | |
1da177e4 LT |
2053 | if (sock->file->f_flags & O_NONBLOCK) |
2054 | flags |= MSG_DONTWAIT; | |
2055 | err = sock_recvmsg(sock, &msg_sys, total_len, flags); | |
2056 | if (err < 0) | |
2057 | goto out_freeiov; | |
2058 | len = err; | |
2059 | ||
2060 | if (uaddr != NULL) { | |
230b1839 YH |
2061 | err = move_addr_to_user((struct sockaddr *)&addr, |
2062 | msg_sys.msg_namelen, uaddr, | |
89bddce5 | 2063 | uaddr_len); |
1da177e4 LT |
2064 | if (err < 0) |
2065 | goto out_freeiov; | |
2066 | } | |
37f7f421 DM |
2067 | err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT), |
2068 | COMPAT_FLAGS(msg)); | |
1da177e4 LT |
2069 | if (err) |
2070 | goto out_freeiov; | |
2071 | if (MSG_CMSG_COMPAT & flags) | |
89bddce5 | 2072 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, |
1da177e4 LT |
2073 | &msg_compat->msg_controllen); |
2074 | else | |
89bddce5 | 2075 | err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr, |
1da177e4 LT |
2076 | &msg->msg_controllen); |
2077 | if (err) | |
2078 | goto out_freeiov; | |
2079 | err = len; | |
2080 | ||
2081 | out_freeiov: | |
2082 | if (iov != iovstack) | |
2083 | sock_kfree_s(sock->sk, iov, iov_size); | |
2084 | out_put: | |
6cb153ca | 2085 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2086 | out: |
2087 | return err; | |
2088 | } | |
2089 | ||
2090 | #ifdef __ARCH_WANT_SYS_SOCKETCALL | |
2091 | ||
2092 | /* Argument list sizes for sys_socketcall */ | |
2093 | #define AL(x) ((x) * sizeof(unsigned long)) | |
aaca0bdc | 2094 | static const unsigned char nargs[19]={ |
89bddce5 SH |
2095 | AL(0),AL(3),AL(3),AL(3),AL(2),AL(3), |
2096 | AL(3),AL(3),AL(4),AL(4),AL(4),AL(6), | |
aaca0bdc UD |
2097 | AL(6),AL(2),AL(5),AL(5),AL(3),AL(3), |
2098 | AL(6) | |
89bddce5 SH |
2099 | }; |
2100 | ||
1da177e4 LT |
2101 | #undef AL |
2102 | ||
2103 | /* | |
89bddce5 | 2104 | * System call vectors. |
1da177e4 LT |
2105 | * |
2106 | * Argument checking cleaned up. Saved 20% in size. | |
2107 | * This function doesn't need to set the kernel lock because | |
89bddce5 | 2108 | * it is set by the callees. |
1da177e4 LT |
2109 | */ |
2110 | ||
2111 | asmlinkage long sys_socketcall(int call, unsigned long __user *args) | |
2112 | { | |
2113 | unsigned long a[6]; | |
89bddce5 | 2114 | unsigned long a0, a1; |
1da177e4 LT |
2115 | int err; |
2116 | ||
aaca0bdc | 2117 | if (call < 1 || call > SYS_PACCEPT) |
1da177e4 LT |
2118 | return -EINVAL; |
2119 | ||
2120 | /* copy_from_user should be SMP safe. */ | |
2121 | if (copy_from_user(a, args, nargs[call])) | |
2122 | return -EFAULT; | |
3ec3b2fb | 2123 | |
89bddce5 | 2124 | err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); |
3ec3b2fb DW |
2125 | if (err) |
2126 | return err; | |
2127 | ||
89bddce5 SH |
2128 | a0 = a[0]; |
2129 | a1 = a[1]; | |
2130 | ||
2131 | switch (call) { | |
2132 | case SYS_SOCKET: | |
2133 | err = sys_socket(a0, a1, a[2]); | |
2134 | break; | |
2135 | case SYS_BIND: | |
2136 | err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]); | |
2137 | break; | |
2138 | case SYS_CONNECT: | |
2139 | err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]); | |
2140 | break; | |
2141 | case SYS_LISTEN: | |
2142 | err = sys_listen(a0, a1); | |
2143 | break; | |
2144 | case SYS_ACCEPT: | |
2145 | err = | |
aaca0bdc UD |
2146 | do_accept(a0, (struct sockaddr __user *)a1, |
2147 | (int __user *)a[2], 0); | |
89bddce5 SH |
2148 | break; |
2149 | case SYS_GETSOCKNAME: | |
2150 | err = | |
2151 | sys_getsockname(a0, (struct sockaddr __user *)a1, | |
2152 | (int __user *)a[2]); | |
2153 | break; | |
2154 | case SYS_GETPEERNAME: | |
2155 | err = | |
2156 | sys_getpeername(a0, (struct sockaddr __user *)a1, | |
2157 | (int __user *)a[2]); | |
2158 | break; | |
2159 | case SYS_SOCKETPAIR: | |
2160 | err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]); | |
2161 | break; | |
2162 | case SYS_SEND: | |
2163 | err = sys_send(a0, (void __user *)a1, a[2], a[3]); | |
2164 | break; | |
2165 | case SYS_SENDTO: | |
2166 | err = sys_sendto(a0, (void __user *)a1, a[2], a[3], | |
2167 | (struct sockaddr __user *)a[4], a[5]); | |
2168 | break; | |
2169 | case SYS_RECV: | |
2170 | err = sys_recv(a0, (void __user *)a1, a[2], a[3]); | |
2171 | break; | |
2172 | case SYS_RECVFROM: | |
2173 | err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3], | |
2174 | (struct sockaddr __user *)a[4], | |
2175 | (int __user *)a[5]); | |
2176 | break; | |
2177 | case SYS_SHUTDOWN: | |
2178 | err = sys_shutdown(a0, a1); | |
2179 | break; | |
2180 | case SYS_SETSOCKOPT: | |
2181 | err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]); | |
2182 | break; | |
2183 | case SYS_GETSOCKOPT: | |
2184 | err = | |
2185 | sys_getsockopt(a0, a1, a[2], (char __user *)a[3], | |
2186 | (int __user *)a[4]); | |
2187 | break; | |
2188 | case SYS_SENDMSG: | |
2189 | err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2190 | break; | |
2191 | case SYS_RECVMSG: | |
2192 | err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]); | |
2193 | break; | |
aaca0bdc UD |
2194 | case SYS_PACCEPT: |
2195 | err = | |
2196 | sys_paccept(a0, (struct sockaddr __user *)a1, | |
2197 | (int __user *)a[2], | |
2198 | (const sigset_t __user *) a[3], | |
2199 | a[4], a[5]); | |
2200 | break; | |
89bddce5 SH |
2201 | default: |
2202 | err = -EINVAL; | |
2203 | break; | |
1da177e4 LT |
2204 | } |
2205 | return err; | |
2206 | } | |
2207 | ||
89bddce5 | 2208 | #endif /* __ARCH_WANT_SYS_SOCKETCALL */ |
1da177e4 | 2209 | |
55737fda SH |
2210 | /** |
2211 | * sock_register - add a socket protocol handler | |
2212 | * @ops: description of protocol | |
2213 | * | |
1da177e4 LT |
2214 | * This function is called by a protocol handler that wants to |
2215 | * advertise its address family, and have it linked into the | |
55737fda SH |
2216 | * socket interface. The value ops->family coresponds to the |
2217 | * socket system call protocol family. | |
1da177e4 | 2218 | */ |
f0fd27d4 | 2219 | int sock_register(const struct net_proto_family *ops) |
1da177e4 LT |
2220 | { |
2221 | int err; | |
2222 | ||
2223 | if (ops->family >= NPROTO) { | |
89bddce5 SH |
2224 | printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, |
2225 | NPROTO); | |
1da177e4 LT |
2226 | return -ENOBUFS; |
2227 | } | |
55737fda SH |
2228 | |
2229 | spin_lock(&net_family_lock); | |
2230 | if (net_families[ops->family]) | |
2231 | err = -EEXIST; | |
2232 | else { | |
89bddce5 | 2233 | net_families[ops->family] = ops; |
1da177e4 LT |
2234 | err = 0; |
2235 | } | |
55737fda SH |
2236 | spin_unlock(&net_family_lock); |
2237 | ||
89bddce5 | 2238 | printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family); |
1da177e4 LT |
2239 | return err; |
2240 | } | |
2241 | ||
55737fda SH |
2242 | /** |
2243 | * sock_unregister - remove a protocol handler | |
2244 | * @family: protocol family to remove | |
2245 | * | |
1da177e4 LT |
2246 | * This function is called by a protocol handler that wants to |
2247 | * remove its address family, and have it unlinked from the | |
55737fda SH |
2248 | * new socket creation. |
2249 | * | |
2250 | * If protocol handler is a module, then it can use module reference | |
2251 | * counts to protect against new references. If protocol handler is not | |
2252 | * a module then it needs to provide its own protection in | |
2253 | * the ops->create routine. | |
1da177e4 | 2254 | */ |
f0fd27d4 | 2255 | void sock_unregister(int family) |
1da177e4 | 2256 | { |
f0fd27d4 | 2257 | BUG_ON(family < 0 || family >= NPROTO); |
1da177e4 | 2258 | |
55737fda | 2259 | spin_lock(&net_family_lock); |
89bddce5 | 2260 | net_families[family] = NULL; |
55737fda SH |
2261 | spin_unlock(&net_family_lock); |
2262 | ||
2263 | synchronize_rcu(); | |
2264 | ||
89bddce5 | 2265 | printk(KERN_INFO "NET: Unregistered protocol family %d\n", family); |
1da177e4 LT |
2266 | } |
2267 | ||
77d76ea3 | 2268 | static int __init sock_init(void) |
1da177e4 LT |
2269 | { |
2270 | /* | |
89bddce5 | 2271 | * Initialize sock SLAB cache. |
1da177e4 | 2272 | */ |
89bddce5 | 2273 | |
1da177e4 LT |
2274 | sk_init(); |
2275 | ||
1da177e4 | 2276 | /* |
89bddce5 | 2277 | * Initialize skbuff SLAB cache |
1da177e4 LT |
2278 | */ |
2279 | skb_init(); | |
1da177e4 LT |
2280 | |
2281 | /* | |
89bddce5 | 2282 | * Initialize the protocols module. |
1da177e4 LT |
2283 | */ |
2284 | ||
2285 | init_inodecache(); | |
2286 | register_filesystem(&sock_fs_type); | |
2287 | sock_mnt = kern_mount(&sock_fs_type); | |
77d76ea3 AK |
2288 | |
2289 | /* The real protocol initialization is performed in later initcalls. | |
1da177e4 LT |
2290 | */ |
2291 | ||
2292 | #ifdef CONFIG_NETFILTER | |
2293 | netfilter_init(); | |
2294 | #endif | |
cbeb321a DM |
2295 | |
2296 | return 0; | |
1da177e4 LT |
2297 | } |
2298 | ||
77d76ea3 AK |
2299 | core_initcall(sock_init); /* early initcall */ |
2300 | ||
1da177e4 LT |
2301 | #ifdef CONFIG_PROC_FS |
2302 | void socket_seq_show(struct seq_file *seq) | |
2303 | { | |
2304 | int cpu; | |
2305 | int counter = 0; | |
2306 | ||
6f912042 | 2307 | for_each_possible_cpu(cpu) |
89bddce5 | 2308 | counter += per_cpu(sockets_in_use, cpu); |
1da177e4 LT |
2309 | |
2310 | /* It can be negative, by the way. 8) */ | |
2311 | if (counter < 0) | |
2312 | counter = 0; | |
2313 | ||
2314 | seq_printf(seq, "sockets: used %d\n", counter); | |
2315 | } | |
89bddce5 | 2316 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 2317 | |
89bbfc95 SP |
2318 | #ifdef CONFIG_COMPAT |
2319 | static long compat_sock_ioctl(struct file *file, unsigned cmd, | |
89bddce5 | 2320 | unsigned long arg) |
89bbfc95 SP |
2321 | { |
2322 | struct socket *sock = file->private_data; | |
2323 | int ret = -ENOIOCTLCMD; | |
87de87d5 DM |
2324 | struct sock *sk; |
2325 | struct net *net; | |
2326 | ||
2327 | sk = sock->sk; | |
2328 | net = sock_net(sk); | |
89bbfc95 SP |
2329 | |
2330 | if (sock->ops->compat_ioctl) | |
2331 | ret = sock->ops->compat_ioctl(sock, cmd, arg); | |
2332 | ||
87de87d5 DM |
2333 | if (ret == -ENOIOCTLCMD && |
2334 | (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) | |
2335 | ret = compat_wext_handle_ioctl(net, cmd, arg); | |
2336 | ||
89bbfc95 SP |
2337 | return ret; |
2338 | } | |
2339 | #endif | |
2340 | ||
ac5a488e SS |
2341 | int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) |
2342 | { | |
2343 | return sock->ops->bind(sock, addr, addrlen); | |
2344 | } | |
2345 | ||
2346 | int kernel_listen(struct socket *sock, int backlog) | |
2347 | { | |
2348 | return sock->ops->listen(sock, backlog); | |
2349 | } | |
2350 | ||
2351 | int kernel_accept(struct socket *sock, struct socket **newsock, int flags) | |
2352 | { | |
2353 | struct sock *sk = sock->sk; | |
2354 | int err; | |
2355 | ||
2356 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
2357 | newsock); | |
2358 | if (err < 0) | |
2359 | goto done; | |
2360 | ||
2361 | err = sock->ops->accept(sock, *newsock, flags); | |
2362 | if (err < 0) { | |
2363 | sock_release(*newsock); | |
fa8705b0 | 2364 | *newsock = NULL; |
ac5a488e SS |
2365 | goto done; |
2366 | } | |
2367 | ||
2368 | (*newsock)->ops = sock->ops; | |
2369 | ||
2370 | done: | |
2371 | return err; | |
2372 | } | |
2373 | ||
2374 | int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, | |
4768fbcb | 2375 | int flags) |
ac5a488e SS |
2376 | { |
2377 | return sock->ops->connect(sock, addr, addrlen, flags); | |
2378 | } | |
2379 | ||
2380 | int kernel_getsockname(struct socket *sock, struct sockaddr *addr, | |
2381 | int *addrlen) | |
2382 | { | |
2383 | return sock->ops->getname(sock, addr, addrlen, 0); | |
2384 | } | |
2385 | ||
2386 | int kernel_getpeername(struct socket *sock, struct sockaddr *addr, | |
2387 | int *addrlen) | |
2388 | { | |
2389 | return sock->ops->getname(sock, addr, addrlen, 1); | |
2390 | } | |
2391 | ||
2392 | int kernel_getsockopt(struct socket *sock, int level, int optname, | |
2393 | char *optval, int *optlen) | |
2394 | { | |
2395 | mm_segment_t oldfs = get_fs(); | |
2396 | int err; | |
2397 | ||
2398 | set_fs(KERNEL_DS); | |
2399 | if (level == SOL_SOCKET) | |
2400 | err = sock_getsockopt(sock, level, optname, optval, optlen); | |
2401 | else | |
2402 | err = sock->ops->getsockopt(sock, level, optname, optval, | |
2403 | optlen); | |
2404 | set_fs(oldfs); | |
2405 | return err; | |
2406 | } | |
2407 | ||
2408 | int kernel_setsockopt(struct socket *sock, int level, int optname, | |
2409 | char *optval, int optlen) | |
2410 | { | |
2411 | mm_segment_t oldfs = get_fs(); | |
2412 | int err; | |
2413 | ||
2414 | set_fs(KERNEL_DS); | |
2415 | if (level == SOL_SOCKET) | |
2416 | err = sock_setsockopt(sock, level, optname, optval, optlen); | |
2417 | else | |
2418 | err = sock->ops->setsockopt(sock, level, optname, optval, | |
2419 | optlen); | |
2420 | set_fs(oldfs); | |
2421 | return err; | |
2422 | } | |
2423 | ||
2424 | int kernel_sendpage(struct socket *sock, struct page *page, int offset, | |
2425 | size_t size, int flags) | |
2426 | { | |
2427 | if (sock->ops->sendpage) | |
2428 | return sock->ops->sendpage(sock, page, offset, size, flags); | |
2429 | ||
2430 | return sock_no_sendpage(sock, page, offset, size, flags); | |
2431 | } | |
2432 | ||
2433 | int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg) | |
2434 | { | |
2435 | mm_segment_t oldfs = get_fs(); | |
2436 | int err; | |
2437 | ||
2438 | set_fs(KERNEL_DS); | |
2439 | err = sock->ops->ioctl(sock, cmd, arg); | |
2440 | set_fs(oldfs); | |
2441 | ||
2442 | return err; | |
2443 | } | |
2444 | ||
91cf45f0 TM |
2445 | int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) |
2446 | { | |
2447 | return sock->ops->shutdown(sock, how); | |
2448 | } | |
2449 | ||
1da177e4 LT |
2450 | EXPORT_SYMBOL(sock_create); |
2451 | EXPORT_SYMBOL(sock_create_kern); | |
2452 | EXPORT_SYMBOL(sock_create_lite); | |
2453 | EXPORT_SYMBOL(sock_map_fd); | |
2454 | EXPORT_SYMBOL(sock_recvmsg); | |
2455 | EXPORT_SYMBOL(sock_register); | |
2456 | EXPORT_SYMBOL(sock_release); | |
2457 | EXPORT_SYMBOL(sock_sendmsg); | |
2458 | EXPORT_SYMBOL(sock_unregister); | |
2459 | EXPORT_SYMBOL(sock_wake_async); | |
2460 | EXPORT_SYMBOL(sockfd_lookup); | |
2461 | EXPORT_SYMBOL(kernel_sendmsg); | |
2462 | EXPORT_SYMBOL(kernel_recvmsg); | |
ac5a488e SS |
2463 | EXPORT_SYMBOL(kernel_bind); |
2464 | EXPORT_SYMBOL(kernel_listen); | |
2465 | EXPORT_SYMBOL(kernel_accept); | |
2466 | EXPORT_SYMBOL(kernel_connect); | |
2467 | EXPORT_SYMBOL(kernel_getsockname); | |
2468 | EXPORT_SYMBOL(kernel_getpeername); | |
2469 | EXPORT_SYMBOL(kernel_getsockopt); | |
2470 | EXPORT_SYMBOL(kernel_setsockopt); | |
2471 | EXPORT_SYMBOL(kernel_sendpage); | |
2472 | EXPORT_SYMBOL(kernel_sock_ioctl); | |
91cf45f0 | 2473 | EXPORT_SYMBOL(kernel_sock_shutdown); |