blob: 29ea1de43ecb3ec8bfe6c7ef13369c1e557e6530 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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>
Jesper Juhl02c30a82005-05-05 16:16:16 -07007 * Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * 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)
Stephen Hemminger89bddce2006-09-01 00:19:31 -070045 * Tigran Aivazian : Made listen(2) backlog sanity checks
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 * 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
Stephen Hemminger89bddce2006-09-01 00:19:31 -070056 * paradigm.
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070066#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/netdevice.h>
68#include <linux/proc_fs.h>
69#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080070#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <linux/wanrouter.h>
72#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030073#include <linux/if_frad.h>
74#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#include <linux/init.h>
76#include <linux/poll.h>
77#include <linux/cache.h>
78#include <linux/module.h>
79#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070080#include <linux/mount.h>
81#include <linux/security.h>
82#include <linux/syscalls.h>
83#include <linux/compat.h>
84#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010085#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010086#include <linux/wireless.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070087
88#include <asm/uaccess.h>
89#include <asm/unistd.h>
90
91#include <net/compat.h>
92
93#include <net/sock.h>
94#include <linux/netfilter.h>
95
96static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -070097static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
98 unsigned long nr_segs, loff_t pos);
99static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
100 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700101static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102
103static int sock_close(struct inode *inode, struct file *file);
104static unsigned int sock_poll(struct file *file,
105 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700106static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800107#ifdef CONFIG_COMPAT
108static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700109 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800110#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112static ssize_t sock_sendpage(struct file *file, struct page *page,
113 int offset, size_t size, loff_t *ppos, int more);
114
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115/*
116 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
117 * in the operation structures but are done directly via the socketcall() multiplexor.
118 */
119
120static struct file_operations socket_file_ops = {
121 .owner = THIS_MODULE,
122 .llseek = no_llseek,
123 .aio_read = sock_aio_read,
124 .aio_write = sock_aio_write,
125 .poll = sock_poll,
126 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800127#ifdef CONFIG_COMPAT
128 .compat_ioctl = compat_sock_ioctl,
129#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 .mmap = sock_mmap,
131 .open = sock_no_open, /* special open code to disallow open via /proc */
132 .release = sock_close,
133 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200134 .sendpage = sock_sendpage,
135 .splice_write = generic_splice_sendpage,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136};
137
138/*
139 * The protocol list. Each protocol is registered in here.
140 */
141
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142static DEFINE_SPINLOCK(net_family_lock);
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -0700143static const struct net_proto_family *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145/*
146 * Statistics counters of the socket lists
147 */
148
149static DEFINE_PER_CPU(int, sockets_in_use) = 0;
150
151/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700152 * Support routines.
153 * Move socket addresses back and forth across the kernel/user
154 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 */
156
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700157#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 16 for IP, 16 for IPX,
159 24 for IPv6,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700160 about 80 for AX.25
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 must be at least one bigger than
162 the AF_UNIX size (see net/unix/af_unix.c
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700163 :unix_mkname()).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700165
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166/**
167 * move_addr_to_kernel - copy a socket address into kernel space
168 * @uaddr: Address in user space
169 * @kaddr: Address in kernel space
170 * @ulen: Length in user space
171 *
172 * The address is copied into kernel space. If the provided address is
173 * too long an error code of -EINVAL is returned. If the copy gives
174 * invalid addresses -EFAULT is returned. On a success 0 is returned.
175 */
176
177int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
178{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700179 if (ulen < 0 || ulen > MAX_SOCK_ADDR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700181 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700183 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100185 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186}
187
188/**
189 * move_addr_to_user - copy an address to user space
190 * @kaddr: kernel space address
191 * @klen: length of address in kernel
192 * @uaddr: user space address
193 * @ulen: pointer to user length field
194 *
195 * The value pointed to by ulen on entry is the buffer length available.
196 * This is overwritten with the buffer space used. -EINVAL is returned
197 * if an overlong buffer is specified or a negative buffer size. -EFAULT
198 * is returned if either the buffer or the length field are not
199 * accessible.
200 * After copying the data up to the limit the user specifies, the true
201 * length of the data is written over the length limit the user
202 * specified. Zero is returned for a success.
203 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700204
205int move_addr_to_user(void *kaddr, int klen, void __user *uaddr,
206 int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207{
208 int err;
209 int len;
210
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700211 err = get_user(len, ulen);
212 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700214 if (len > klen)
215 len = klen;
216 if (len < 0 || len > MAX_SOCK_ADDR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700218 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500219 if (audit_sockaddr(klen, kaddr))
220 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700221 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 return -EFAULT;
223 }
224 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 * "fromlen shall refer to the value before truncation.."
226 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 */
228 return __put_user(klen, ulen);
229}
230
231#define SOCKFS_MAGIC 0x534F434B
232
Christoph Lametere18b8902006-12-06 20:33:20 -0800233static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234
235static struct inode *sock_alloc_inode(struct super_block *sb)
236{
237 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700238
Christoph Lametere94b1762006-12-06 20:33:17 -0800239 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 if (!ei)
241 return NULL;
242 init_waitqueue_head(&ei->socket.wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700243
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 ei->socket.fasync_list = NULL;
245 ei->socket.state = SS_UNCONNECTED;
246 ei->socket.flags = 0;
247 ei->socket.ops = NULL;
248 ei->socket.sk = NULL;
249 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250
251 return &ei->vfs_inode;
252}
253
254static void sock_destroy_inode(struct inode *inode)
255{
256 kmem_cache_free(sock_inode_cachep,
257 container_of(inode, struct socket_alloc, vfs_inode));
258}
259
Christoph Lametere18b8902006-12-06 20:33:20 -0800260static void init_once(void *foo, struct kmem_cache *cachep, unsigned long flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700262 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700264 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR))
265 == SLAB_CTOR_CONSTRUCTOR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 inode_init_once(&ei->vfs_inode);
267}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700268
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269static int init_inodecache(void)
270{
271 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700272 sizeof(struct socket_alloc),
273 0,
274 (SLAB_HWCACHE_ALIGN |
275 SLAB_RECLAIM_ACCOUNT |
276 SLAB_MEM_SPREAD),
277 init_once,
278 NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 if (sock_inode_cachep == NULL)
280 return -ENOMEM;
281 return 0;
282}
283
284static struct super_operations sockfs_ops = {
285 .alloc_inode = sock_alloc_inode,
286 .destroy_inode =sock_destroy_inode,
287 .statfs = simple_statfs,
288};
289
David Howells454e2392006-06-23 02:02:57 -0700290static int sockfs_get_sb(struct file_system_type *fs_type,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700291 int flags, const char *dev_name, void *data,
292 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293{
David Howells454e2392006-06-23 02:02:57 -0700294 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
295 mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296}
297
Eric Dumazetba899662005-08-26 12:05:31 -0700298static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
300static struct file_system_type sock_fs_type = {
301 .name = "sockfs",
302 .get_sb = sockfs_get_sb,
303 .kill_sb = kill_anon_super,
304};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700305
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306static int sockfs_delete_dentry(struct dentry *dentry)
307{
Eric Dumazet304e61e2006-12-06 20:38:49 -0800308 /*
309 * At creation time, we pretended this dentry was hashed
310 * (by clearing DCACHE_UNHASHED bit in d_flags)
311 * At delete time, we restore the truth : not hashed.
312 * (so that dput() can proceed correctly)
313 */
314 dentry->d_flags |= DCACHE_UNHASHED;
315 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316}
317static struct dentry_operations sockfs_dentry_operations = {
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700318 .d_delete = sockfs_delete_dentry,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319};
320
321/*
322 * Obtains the first available file descriptor and sets it up for use.
323 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800324 * These functions create file structures and maps them to fd space
325 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 * and file struct implicitly stored in sock->file.
327 * Note that another thread may close file descriptor before we return
328 * from this function. We use the fact that now we do not refer
329 * to socket after mapping. If one day we will need it, this
330 * function will increment ref. count on file by 1.
331 *
332 * In any case returned fd MAY BE not valid!
333 * This race condition is unavoidable
334 * with shared fd spaces, we cannot solve it inside kernel,
335 * but we take care of internal coherence yet.
336 */
337
David S. Miller39d8c1b2006-03-20 17:13:49 -0800338static int sock_alloc_fd(struct file **filep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339{
340 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800341
342 fd = get_unused_fd();
343 if (likely(fd >= 0)) {
344 struct file *file = get_empty_filp();
345
346 *filep = file;
347 if (unlikely(!file)) {
348 put_unused_fd(fd);
349 return -ENFILE;
350 }
351 } else
352 *filep = NULL;
353 return fd;
354}
355
356static int sock_attach_fd(struct socket *sock, struct file *file)
357{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 struct qstr this;
359 char name[32];
360
David S. Miller39d8c1b2006-03-20 17:13:49 -0800361 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
362 this.name = name;
Eric Dumazet304e61e2006-12-06 20:38:49 -0800363 this.hash = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
David S. Miller39d8c1b2006-03-20 17:13:49 -0800365 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
366 if (unlikely(!file->f_dentry))
367 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368
David S. Miller39d8c1b2006-03-20 17:13:49 -0800369 file->f_dentry->d_op = &sockfs_dentry_operations;
Eric Dumazet304e61e2006-12-06 20:38:49 -0800370 /*
371 * We dont want to push this dentry into global dentry hash table.
372 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
373 * This permits a working /proc/$pid/fd/XXX on sockets
374 */
375 file->f_dentry->d_flags &= ~DCACHE_UNHASHED;
376 d_instantiate(file->f_dentry, SOCK_INODE(sock));
David S. Miller39d8c1b2006-03-20 17:13:49 -0800377 file->f_vfsmnt = mntget(sock_mnt);
378 file->f_mapping = file->f_dentry->d_inode->i_mapping;
379
380 sock->file = file;
381 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
382 file->f_mode = FMODE_READ | FMODE_WRITE;
383 file->f_flags = O_RDWR;
384 file->f_pos = 0;
385 file->private_data = sock;
386
387 return 0;
388}
389
390int sock_map_fd(struct socket *sock)
391{
392 struct file *newfile;
393 int fd = sock_alloc_fd(&newfile);
394
395 if (likely(fd >= 0)) {
396 int err = sock_attach_fd(sock, newfile);
397
398 if (unlikely(err < 0)) {
399 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800401 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800403 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 return fd;
406}
407
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800408static struct socket *sock_from_file(struct file *file, int *err)
409{
410 struct inode *inode;
411 struct socket *sock;
412
413 if (file->f_op == &socket_file_ops)
414 return file->private_data; /* set in sock_map_fd */
415
416 inode = file->f_dentry->d_inode;
417 if (!S_ISSOCK(inode->i_mode)) {
418 *err = -ENOTSOCK;
419 return NULL;
420 }
421
422 sock = SOCKET_I(inode);
423 if (sock->file != file) {
424 printk(KERN_ERR "socki_lookup: socket file changed!\n");
425 sock->file = file;
426 }
427 return sock;
428}
429
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430/**
431 * sockfd_lookup - Go from a file number to its socket slot
432 * @fd: file handle
433 * @err: pointer to an error code return
434 *
435 * The file handle passed in is locked and the socket it is bound
436 * too is returned. If an error occurs the err pointer is overwritten
437 * with a negative errno code and NULL is returned. The function checks
438 * for both invalid handles and passing a handle which is not a socket.
439 *
440 * On a success the socket object pointer is returned.
441 */
442
443struct socket *sockfd_lookup(int fd, int *err)
444{
445 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 struct socket *sock;
447
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700448 file = fget(fd);
449 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 *err = -EBADF;
451 return NULL;
452 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700453
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800454 sock = sock_from_file(file, err);
455 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 return sock;
458}
459
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800460static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
461{
462 struct file *file;
463 struct socket *sock;
464
Hua Zhong36725582006-04-19 15:25:02 -0700465 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800466 file = fget_light(fd, fput_needed);
467 if (file) {
468 sock = sock_from_file(file, err);
469 if (sock)
470 return sock;
471 fput_light(file, *fput_needed);
472 }
473 return NULL;
474}
475
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476/**
477 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700478 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 * Allocate a new inode and socket object. The two are bound together
480 * and initialised. The socket is then returned. If we are out of inodes
481 * NULL is returned.
482 */
483
484static struct socket *sock_alloc(void)
485{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700486 struct inode *inode;
487 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488
489 inode = new_inode(sock_mnt->mnt_sb);
490 if (!inode)
491 return NULL;
492
493 sock = SOCKET_I(inode);
494
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700495 inode->i_mode = S_IFSOCK | S_IRWXUGO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 inode->i_uid = current->fsuid;
497 inode->i_gid = current->fsgid;
498
499 get_cpu_var(sockets_in_use)++;
500 put_cpu_var(sockets_in_use);
501 return sock;
502}
503
504/*
505 * In theory you can't get an open on this inode, but /proc provides
506 * a back door. Remember to keep it shut otherwise you'll let the
507 * creepy crawlies in.
508 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700509
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
511{
512 return -ENXIO;
513}
514
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800515const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 .owner = THIS_MODULE,
517 .open = sock_no_open,
518};
519
520/**
521 * sock_release - close a socket
522 * @sock: socket to close
523 *
524 * The socket is released from the protocol stack if it has a release
525 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700526 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700528
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529void sock_release(struct socket *sock)
530{
531 if (sock->ops) {
532 struct module *owner = sock->ops->owner;
533
534 sock->ops->release(sock);
535 sock->ops = NULL;
536 module_put(owner);
537 }
538
539 if (sock->fasync_list)
540 printk(KERN_ERR "sock_release: fasync list not empty!\n");
541
542 get_cpu_var(sockets_in_use)--;
543 put_cpu_var(sockets_in_use);
544 if (!sock->file) {
545 iput(SOCK_INODE(sock));
546 return;
547 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700548 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549}
550
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700551static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 struct msghdr *msg, size_t size)
553{
554 struct sock_iocb *si = kiocb_to_siocb(iocb);
555 int err;
556
557 si->sock = sock;
558 si->scm = NULL;
559 si->msg = msg;
560 si->size = size;
561
562 err = security_socket_sendmsg(sock, msg, size);
563 if (err)
564 return err;
565
566 return sock->ops->sendmsg(iocb, sock, msg, size);
567}
568
569int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
570{
571 struct kiocb iocb;
572 struct sock_iocb siocb;
573 int ret;
574
575 init_sync_kiocb(&iocb, NULL);
576 iocb.private = &siocb;
577 ret = __sock_sendmsg(&iocb, sock, msg, size);
578 if (-EIOCBQUEUED == ret)
579 ret = wait_on_sync_kiocb(&iocb);
580 return ret;
581}
582
583int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
584 struct kvec *vec, size_t num, size_t size)
585{
586 mm_segment_t oldfs = get_fs();
587 int result;
588
589 set_fs(KERNEL_DS);
590 /*
591 * the following is safe, since for compiler definitions of kvec and
592 * iovec are identical, yielding the same in-core layout and alignment
593 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700594 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 msg->msg_iovlen = num;
596 result = sock_sendmsg(sock, msg, size);
597 set_fs(oldfs);
598 return result;
599}
600
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700601static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 struct msghdr *msg, size_t size, int flags)
603{
604 int err;
605 struct sock_iocb *si = kiocb_to_siocb(iocb);
606
607 si->sock = sock;
608 si->scm = NULL;
609 si->msg = msg;
610 si->size = size;
611 si->flags = flags;
612
613 err = security_socket_recvmsg(sock, msg, size, flags);
614 if (err)
615 return err;
616
617 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
618}
619
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700620int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 size_t size, int flags)
622{
623 struct kiocb iocb;
624 struct sock_iocb siocb;
625 int ret;
626
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700627 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 iocb.private = &siocb;
629 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
630 if (-EIOCBQUEUED == ret)
631 ret = wait_on_sync_kiocb(&iocb);
632 return ret;
633}
634
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700635int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
636 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637{
638 mm_segment_t oldfs = get_fs();
639 int result;
640
641 set_fs(KERNEL_DS);
642 /*
643 * the following is safe, since for compiler definitions of kvec and
644 * iovec are identical, yielding the same in-core layout and alignment
645 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700646 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 result = sock_recvmsg(sock, msg, size, flags);
648 set_fs(oldfs);
649 return result;
650}
651
652static void sock_aio_dtor(struct kiocb *iocb)
653{
654 kfree(iocb->private);
655}
656
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300657static ssize_t sock_sendpage(struct file *file, struct page *page,
658 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659{
660 struct socket *sock;
661 int flags;
662
Eric Dumazetb69aee02005-09-06 14:42:45 -0700663 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
665 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
666 if (more)
667 flags |= MSG_MORE;
668
669 return sock->ops->sendpage(sock, page, offset, size, flags);
670}
671
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800672static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700673 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800674{
675 if (!is_sync_kiocb(iocb)) {
676 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
677 if (!siocb)
678 return NULL;
679 iocb->ki_dtor = sock_aio_dtor;
680 }
681
682 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800683 iocb->private = siocb;
684 return siocb;
685}
686
687static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700688 struct file *file, const struct iovec *iov,
689 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800690{
691 struct socket *sock = file->private_data;
692 size_t size = 0;
693 int i;
694
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700695 for (i = 0; i < nr_segs; i++)
696 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800697
698 msg->msg_name = NULL;
699 msg->msg_namelen = 0;
700 msg->msg_control = NULL;
701 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700702 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800703 msg->msg_iovlen = nr_segs;
704 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
705
706 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
707}
708
Badari Pulavarty027445c2006-09-30 23:28:46 -0700709static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
710 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800711{
712 struct sock_iocb siocb, *x;
713
714 if (pos != 0)
715 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700716
717 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800718 return 0;
719
Badari Pulavarty027445c2006-09-30 23:28:46 -0700720
721 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800722 if (!x)
723 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700724 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800725}
726
727static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700728 struct file *file, const struct iovec *iov,
729 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800730{
731 struct socket *sock = file->private_data;
732 size_t size = 0;
733 int i;
734
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700735 for (i = 0; i < nr_segs; i++)
736 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800737
738 msg->msg_name = NULL;
739 msg->msg_namelen = 0;
740 msg->msg_control = NULL;
741 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700742 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800743 msg->msg_iovlen = nr_segs;
744 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
745 if (sock->type == SOCK_SEQPACKET)
746 msg->msg_flags |= MSG_EOR;
747
748 return __sock_sendmsg(iocb, sock, msg, size);
749}
750
Badari Pulavarty027445c2006-09-30 23:28:46 -0700751static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
752 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800754 struct sock_iocb siocb, *x;
755
756 if (pos != 0)
757 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700758
759 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800760 return 0;
761
Badari Pulavarty027445c2006-09-30 23:28:46 -0700762 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800763 if (!x)
764 return -ENOMEM;
765
Badari Pulavarty027445c2006-09-30 23:28:46 -0700766 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767}
768
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769/*
770 * Atomic setting of ioctl hooks to avoid race
771 * with module unload.
772 */
773
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800774static DEFINE_MUTEX(br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700775static int (*br_ioctl_hook) (unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700777void brioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800779 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800781 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700783
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784EXPORT_SYMBOL(brioctl_set);
785
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800786static DEFINE_MUTEX(vlan_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700787static int (*vlan_ioctl_hook) (void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700789void vlan_ioctl_set(int (*hook) (void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800791 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800793 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700795
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796EXPORT_SYMBOL(vlan_ioctl_set);
797
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800798static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700799static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700801void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800803 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800805 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700807
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808EXPORT_SYMBOL(dlci_ioctl_set);
809
810/*
811 * With an ioctl, arg may well be a user mode pointer, but we don't know
812 * what to do with it - that's up to the protocol still.
813 */
814
815static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
816{
817 struct socket *sock;
818 void __user *argp = (void __user *)arg;
819 int pid, err;
820
Eric Dumazetb69aee02005-09-06 14:42:45 -0700821 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
823 err = dev_ioctl(cmd, argp);
824 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100825#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
827 err = dev_ioctl(cmd, argp);
828 } else
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700829#endif /* CONFIG_WIRELESS_EXT */
830 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 case FIOSETOWN:
832 case SIOCSPGRP:
833 err = -EFAULT;
834 if (get_user(pid, (int __user *)argp))
835 break;
836 err = f_setown(sock->file, pid, 1);
837 break;
838 case FIOGETOWN:
839 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700840 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700841 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 break;
843 case SIOCGIFBR:
844 case SIOCSIFBR:
845 case SIOCBRADDBR:
846 case SIOCBRDELBR:
847 err = -ENOPKG;
848 if (!br_ioctl_hook)
849 request_module("bridge");
850
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800851 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700852 if (br_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 err = br_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800854 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 break;
856 case SIOCGIFVLAN:
857 case SIOCSIFVLAN:
858 err = -ENOPKG;
859 if (!vlan_ioctl_hook)
860 request_module("8021q");
861
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800862 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 if (vlan_ioctl_hook)
864 err = vlan_ioctl_hook(argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800865 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 case SIOCADDDLCI:
868 case SIOCDELDLCI:
869 err = -ENOPKG;
870 if (!dlci_ioctl_hook)
871 request_module("dlci");
872
873 if (dlci_ioctl_hook) {
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800874 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 err = dlci_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800876 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 }
878 break;
879 default:
880 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800881
882 /*
883 * If this ioctl is unknown try to hand it down
884 * to the NIC driver.
885 */
886 if (err == -ENOIOCTLCMD)
887 err = dev_ioctl(cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700889 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 return err;
891}
892
893int sock_create_lite(int family, int type, int protocol, struct socket **res)
894{
895 int err;
896 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700897
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 err = security_socket_create(family, type, protocol, 1);
899 if (err)
900 goto out;
901
902 sock = sock_alloc();
903 if (!sock) {
904 err = -ENOMEM;
905 goto out;
906 }
907
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700909 err = security_socket_post_create(sock, family, type, protocol, 1);
910 if (err)
911 goto out_release;
912
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913out:
914 *res = sock;
915 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700916out_release:
917 sock_release(sock);
918 sock = NULL;
919 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920}
921
922/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700923static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924{
925 struct socket *sock;
926
927 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700928 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700930 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 return sock->ops->poll(file, sock, wait);
932}
933
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700934static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700936 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937
938 return sock->ops->mmap(file, sock, vma);
939}
940
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300941static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942{
943 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700944 * It was possible the inode is NULL we were
945 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 */
947
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700948 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 printk(KERN_DEBUG "sock_close: NULL inode\n");
950 return 0;
951 }
952 sock_fasync(-1, filp, 0);
953 sock_release(SOCKET_I(inode));
954 return 0;
955}
956
957/*
958 * Update the socket async list
959 *
960 * Fasync_list locking strategy.
961 *
962 * 1. fasync_list is modified only under process context socket lock
963 * i.e. under semaphore.
964 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
965 * or under socket lock.
966 * 3. fasync_list can be used from softirq context, so that
967 * modification under socket lock have to be enhanced with
968 * write_lock_bh(&sk->sk_callback_lock).
969 * --ANK (990710)
970 */
971
972static int sock_fasync(int fd, struct file *filp, int on)
973{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700974 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 struct socket *sock;
976 struct sock *sk;
977
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700978 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800979 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700980 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 return -ENOMEM;
982 }
983
Eric Dumazetb69aee02005-09-06 14:42:45 -0700984 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700986 sk = sock->sk;
987 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 kfree(fna);
989 return -EINVAL;
990 }
991
992 lock_sock(sk);
993
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700994 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700996 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
997 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 break;
999
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001000 if (on) {
1001 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001003 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 write_unlock_bh(&sk->sk_callback_lock);
1005
1006 kfree(fna);
1007 goto out;
1008 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001009 fna->fa_file = filp;
1010 fna->fa_fd = fd;
1011 fna->magic = FASYNC_MAGIC;
1012 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001014 sock->fasync_list = fna;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001016 } else {
1017 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001019 *prev = fa->fa_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 write_unlock_bh(&sk->sk_callback_lock);
1021 kfree(fa);
1022 }
1023 }
1024
1025out:
1026 release_sock(sock->sk);
1027 return 0;
1028}
1029
1030/* This function may be called only under socket lock or callback_lock */
1031
1032int sock_wake_async(struct socket *sock, int how, int band)
1033{
1034 if (!sock || !sock->fasync_list)
1035 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001036 switch (how) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 case 1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001038
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1040 break;
1041 goto call_kill;
1042 case 2:
1043 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1044 break;
1045 /* fall through */
1046 case 0:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001047call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 __kill_fasync(sock->fasync_list, SIGIO, band);
1049 break;
1050 case 3:
1051 __kill_fasync(sock->fasync_list, SIGURG, band);
1052 }
1053 return 0;
1054}
1055
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001056static int __sock_create(int family, int type, int protocol,
1057 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058{
1059 int err;
1060 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001061 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
1063 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001064 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 */
1066 if (family < 0 || family >= NPROTO)
1067 return -EAFNOSUPPORT;
1068 if (type < 0 || type >= SOCK_MAX)
1069 return -EINVAL;
1070
1071 /* Compatibility.
1072
1073 This uglymoron is moved from INET layer to here to avoid
1074 deadlock in module load.
1075 */
1076 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001077 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 if (!warned) {
1079 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001080 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1081 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 }
1083 family = PF_PACKET;
1084 }
1085
1086 err = security_socket_create(family, type, protocol, kern);
1087 if (err)
1088 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001089
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001090 /*
1091 * Allocate the socket and allow the family to set things up. if
1092 * the protocol is 0, the family is instructed to select an appropriate
1093 * default.
1094 */
1095 sock = sock_alloc();
1096 if (!sock) {
1097 if (net_ratelimit())
1098 printk(KERN_WARNING "socket: no more sockets\n");
1099 return -ENFILE; /* Not exactly a match, but its the
1100 closest posix thing */
1101 }
1102
1103 sock->type = type;
1104
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105#if defined(CONFIG_KMOD)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001106 /* Attempt to load a protocol module if the find failed.
1107 *
1108 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 * requested real, full-featured networking support upon configuration.
1110 * Otherwise module support will break!
1111 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001112 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001113 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114#endif
1115
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001116 rcu_read_lock();
1117 pf = rcu_dereference(net_families[family]);
1118 err = -EAFNOSUPPORT;
1119 if (!pf)
1120 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121
1122 /*
1123 * We will call the ->create function, that possibly is in a loadable
1124 * module, so we have to bump that loadable module refcnt first.
1125 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001126 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 goto out_release;
1128
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001129 /* Now protected by module ref count */
1130 rcu_read_unlock();
1131
1132 err = pf->create(sock, protocol);
1133 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001135
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 /*
1137 * Now to bump the refcnt of the [loadable] module that owns this
1138 * socket at sock_release time we decrement its refcnt.
1139 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001140 if (!try_module_get(sock->ops->owner))
1141 goto out_module_busy;
1142
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 /*
1144 * Now that we're done with the ->create function, the [loadable]
1145 * module can have its refcnt decremented
1146 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001147 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001148 err = security_socket_post_create(sock, family, type, protocol, kern);
1149 if (err)
1150 goto out_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001151 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001153 return 0;
1154
1155out_module_busy:
1156 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001158 sock->ops = NULL;
1159 module_put(pf->owner);
1160out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001162 return err;
1163
1164out_release:
1165 rcu_read_unlock();
1166 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167}
1168
1169int sock_create(int family, int type, int protocol, struct socket **res)
1170{
1171 return __sock_create(family, type, protocol, res, 0);
1172}
1173
1174int sock_create_kern(int family, int type, int protocol, struct socket **res)
1175{
1176 return __sock_create(family, type, protocol, res, 1);
1177}
1178
1179asmlinkage long sys_socket(int family, int type, int protocol)
1180{
1181 int retval;
1182 struct socket *sock;
1183
1184 retval = sock_create(family, type, protocol, &sock);
1185 if (retval < 0)
1186 goto out;
1187
1188 retval = sock_map_fd(sock);
1189 if (retval < 0)
1190 goto out_release;
1191
1192out:
1193 /* It may be already another descriptor 8) Not kernel problem. */
1194 return retval;
1195
1196out_release:
1197 sock_release(sock);
1198 return retval;
1199}
1200
1201/*
1202 * Create a pair of connected sockets.
1203 */
1204
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001205asmlinkage long sys_socketpair(int family, int type, int protocol,
1206 int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207{
1208 struct socket *sock1, *sock2;
1209 int fd1, fd2, err;
1210
1211 /*
1212 * Obtain the first socket and check if the underlying protocol
1213 * supports the socketpair call.
1214 */
1215
1216 err = sock_create(family, type, protocol, &sock1);
1217 if (err < 0)
1218 goto out;
1219
1220 err = sock_create(family, type, protocol, &sock2);
1221 if (err < 0)
1222 goto out_release_1;
1223
1224 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001225 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 goto out_release_both;
1227
1228 fd1 = fd2 = -1;
1229
1230 err = sock_map_fd(sock1);
1231 if (err < 0)
1232 goto out_release_both;
1233 fd1 = err;
1234
1235 err = sock_map_fd(sock2);
1236 if (err < 0)
1237 goto out_close_1;
1238 fd2 = err;
1239
1240 /* fd1 and fd2 may be already another descriptors.
1241 * Not kernel problem.
1242 */
1243
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001244 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 if (!err)
1246 err = put_user(fd2, &usockvec[1]);
1247 if (!err)
1248 return 0;
1249
1250 sys_close(fd2);
1251 sys_close(fd1);
1252 return err;
1253
1254out_close_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001255 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 sys_close(fd1);
1257 return err;
1258
1259out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001260 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001262 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263out:
1264 return err;
1265}
1266
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267/*
1268 * Bind a name to a socket. Nothing much to do here since it's
1269 * the protocol's responsibility to handle the local address.
1270 *
1271 * We move the socket address to kernel space before we call
1272 * the protocol layer (having also checked the address is ok).
1273 */
1274
1275asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1276{
1277 struct socket *sock;
1278 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001279 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001281 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1282 if(sock) {
1283 err = move_addr_to_kernel(umyaddr, addrlen, address);
1284 if (err >= 0) {
1285 err = security_socket_bind(sock,
1286 (struct sockaddr *)address,
1287 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001288 if (!err)
1289 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001290 (struct sockaddr *)
1291 address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001293 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001294 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 return err;
1296}
1297
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298/*
1299 * Perform a listen. Basically, we allow the protocol to do anything
1300 * necessary for a listen, and if that works, we mark the socket as
1301 * ready for listening.
1302 */
1303
Brian Haley7a42c212006-08-31 15:03:02 -07001304int sysctl_somaxconn __read_mostly = SOMAXCONN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305
1306asmlinkage long sys_listen(int fd, int backlog)
1307{
1308 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001309 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001310
1311 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1312 if (sock) {
1313 if ((unsigned)backlog > sysctl_somaxconn)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 backlog = sysctl_somaxconn;
1315
1316 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001317 if (!err)
1318 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001320 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 }
1322 return err;
1323}
1324
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325/*
1326 * For accept, we attempt to create a new socket, set up the link
1327 * with the client, wake up the client, then return the new
1328 * connected fd. We collect the address of the connector in kernel
1329 * space and move it to user at the very end. This is unclean because
1330 * we open the socket then return an error.
1331 *
1332 * 1003.1g adds the ability to recvmsg() to query connection pending
1333 * status to recvmsg. We need to add that support in a way thats
1334 * clean when we restucture accept also.
1335 */
1336
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001337asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1338 int __user *upeer_addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339{
1340 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001341 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001342 int err, len, newfd, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 char address[MAX_SOCK_ADDR];
1344
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001345 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 if (!sock)
1347 goto out;
1348
1349 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001350 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 goto out_put;
1352
1353 newsock->type = sock->type;
1354 newsock->ops = sock->ops;
1355
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 /*
1357 * We don't need try_module_get here, as the listening socket (sock)
1358 * has the protocol module (sock->ops->owner) held.
1359 */
1360 __module_get(newsock->ops->owner);
1361
David S. Miller39d8c1b2006-03-20 17:13:49 -08001362 newfd = sock_alloc_fd(&newfile);
1363 if (unlikely(newfd < 0)) {
1364 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001365 sock_release(newsock);
1366 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001367 }
1368
1369 err = sock_attach_fd(newsock, newfile);
1370 if (err < 0)
1371 goto out_fd;
1372
Frank Filza79af592005-09-27 15:23:38 -07001373 err = security_socket_accept(sock, newsock);
1374 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001375 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001376
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1378 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001379 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
1381 if (upeer_sockaddr) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001382 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1383 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001385 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001387 err = move_addr_to_user(address, len, upeer_sockaddr,
1388 upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001390 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 }
1392
1393 /* File flags are not inherited via accept() unlike another OSes. */
1394
David S. Miller39d8c1b2006-03-20 17:13:49 -08001395 fd_install(newfd, newfile);
1396 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397
1398 security_socket_post_accept(sock, newsock);
1399
1400out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001401 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402out:
1403 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001404out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001405 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001406 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 goto out_put;
1408}
1409
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410/*
1411 * Attempt to connect to a socket with the server address. The address
1412 * is in user space so we verify it is OK and move it to kernel space.
1413 *
1414 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1415 * break bindings
1416 *
1417 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1418 * other SEQPACKET protocols that take time to connect() as it doesn't
1419 * include the -EINPROGRESS status for such sockets.
1420 */
1421
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001422asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1423 int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424{
1425 struct socket *sock;
1426 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001427 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001429 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 if (!sock)
1431 goto out;
1432 err = move_addr_to_kernel(uservaddr, addrlen, address);
1433 if (err < 0)
1434 goto out_put;
1435
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001436 err =
1437 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 if (err)
1439 goto out_put;
1440
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001441 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 sock->file->f_flags);
1443out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001444 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445out:
1446 return err;
1447}
1448
1449/*
1450 * Get the local address ('name') of a socket object. Move the obtained
1451 * name to user space.
1452 */
1453
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001454asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1455 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456{
1457 struct socket *sock;
1458 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001459 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001460
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001461 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 if (!sock)
1463 goto out;
1464
1465 err = security_socket_getsockname(sock);
1466 if (err)
1467 goto out_put;
1468
1469 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1470 if (err)
1471 goto out_put;
1472 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1473
1474out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001475 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476out:
1477 return err;
1478}
1479
1480/*
1481 * Get the remote address ('name') of a socket object. Move the obtained
1482 * name to user space.
1483 */
1484
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001485asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1486 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487{
1488 struct socket *sock;
1489 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001490 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001492 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1493 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 err = security_socket_getpeername(sock);
1495 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001496 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 return err;
1498 }
1499
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001500 err =
1501 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1502 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 if (!err)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001504 err = move_addr_to_user(address, len, usockaddr,
1505 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001506 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 }
1508 return err;
1509}
1510
1511/*
1512 * Send a datagram to a given address. We move the address into kernel
1513 * space and check the user space data area is readable before invoking
1514 * the protocol.
1515 */
1516
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001517asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1518 unsigned flags, struct sockaddr __user *addr,
1519 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520{
1521 struct socket *sock;
1522 char address[MAX_SOCK_ADDR];
1523 int err;
1524 struct msghdr msg;
1525 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001526 int fput_needed;
1527 struct file *sock_file;
1528
1529 sock_file = fget_light(fd, &fput_needed);
1530 if (!sock_file)
1531 return -EBADF;
1532
1533 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 if (!sock)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001535 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001536 iov.iov_base = buff;
1537 iov.iov_len = len;
1538 msg.msg_name = NULL;
1539 msg.msg_iov = &iov;
1540 msg.msg_iovlen = 1;
1541 msg.msg_control = NULL;
1542 msg.msg_controllen = 0;
1543 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001544 if (addr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 err = move_addr_to_kernel(addr, addr_len, address);
1546 if (err < 0)
1547 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001548 msg.msg_name = address;
1549 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 }
1551 if (sock->file->f_flags & O_NONBLOCK)
1552 flags |= MSG_DONTWAIT;
1553 msg.msg_flags = flags;
1554 err = sock_sendmsg(sock, &msg, len);
1555
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001556out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001557 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 return err;
1559}
1560
1561/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001562 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 */
1564
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001565asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566{
1567 return sys_sendto(fd, buff, len, flags, NULL, 0);
1568}
1569
1570/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001571 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572 * sender. We verify the buffers are writable and if needed move the
1573 * sender address from kernel to user space.
1574 */
1575
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001576asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1577 unsigned flags, struct sockaddr __user *addr,
1578 int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579{
1580 struct socket *sock;
1581 struct iovec iov;
1582 struct msghdr msg;
1583 char address[MAX_SOCK_ADDR];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001584 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001585 struct file *sock_file;
1586 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001588 sock_file = fget_light(fd, &fput_needed);
1589 if (!sock_file)
1590 return -EBADF;
1591
1592 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 if (!sock)
1594 goto out;
1595
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001596 msg.msg_control = NULL;
1597 msg.msg_controllen = 0;
1598 msg.msg_iovlen = 1;
1599 msg.msg_iov = &iov;
1600 iov.iov_len = size;
1601 iov.iov_base = ubuf;
1602 msg.msg_name = address;
1603 msg.msg_namelen = MAX_SOCK_ADDR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 if (sock->file->f_flags & O_NONBLOCK)
1605 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001606 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001608 if (err >= 0 && addr != NULL) {
1609 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1610 if (err2 < 0)
1611 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613out:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001614 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 return err;
1616}
1617
1618/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001619 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 */
1621
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001622asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1623 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624{
1625 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1626}
1627
1628/*
1629 * Set a socket option. Because we don't know the option lengths we have
1630 * to pass the user mode parameter for the protocols to sort out.
1631 */
1632
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001633asmlinkage long sys_setsockopt(int fd, int level, int optname,
1634 char __user *optval, int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001636 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 struct socket *sock;
1638
1639 if (optlen < 0)
1640 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001641
1642 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1643 if (sock != NULL) {
1644 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001645 if (err)
1646 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
1648 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001649 err =
1650 sock_setsockopt(sock, level, optname, optval,
1651 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001653 err =
1654 sock->ops->setsockopt(sock, level, optname, optval,
1655 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001656out_put:
1657 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 }
1659 return err;
1660}
1661
1662/*
1663 * Get a socket option. Because we don't know the option lengths we have
1664 * to pass a user mode parameter for the protocols to sort out.
1665 */
1666
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001667asmlinkage long sys_getsockopt(int fd, int level, int optname,
1668 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001670 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 struct socket *sock;
1672
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001673 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1674 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001675 err = security_socket_getsockopt(sock, level, optname);
1676 if (err)
1677 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678
1679 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001680 err =
1681 sock_getsockopt(sock, level, optname, optval,
1682 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001684 err =
1685 sock->ops->getsockopt(sock, level, optname, optval,
1686 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001687out_put:
1688 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 }
1690 return err;
1691}
1692
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693/*
1694 * Shutdown a socket.
1695 */
1696
1697asmlinkage long sys_shutdown(int fd, int how)
1698{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001699 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 struct socket *sock;
1701
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001702 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1703 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001705 if (!err)
1706 err = sock->ops->shutdown(sock, how);
1707 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 }
1709 return err;
1710}
1711
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001712/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 * fields which are the same type (int / unsigned) on our platforms.
1714 */
1715#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1716#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1717#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1718
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719/*
1720 * BSD sendmsg interface
1721 */
1722
1723asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1724{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001725 struct compat_msghdr __user *msg_compat =
1726 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 struct socket *sock;
1728 char address[MAX_SOCK_ADDR];
1729 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001730 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001731 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1732 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 unsigned char *ctl_buf = ctl;
1734 struct msghdr msg_sys;
1735 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001736 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001737
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 err = -EFAULT;
1739 if (MSG_CMSG_COMPAT & flags) {
1740 if (get_compat_msghdr(&msg_sys, msg_compat))
1741 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001742 }
1743 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 return -EFAULT;
1745
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001746 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001747 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 goto out;
1749
1750 /* do not move before msg_sys is valid */
1751 err = -EMSGSIZE;
1752 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1753 goto out_put;
1754
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001755 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756 err = -ENOMEM;
1757 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1758 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1759 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1760 if (!iov)
1761 goto out_put;
1762 }
1763
1764 /* This will also move the address data into kernel space */
1765 if (MSG_CMSG_COMPAT & flags) {
1766 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1767 } else
1768 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001769 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 goto out_freeiov;
1771 total_len = err;
1772
1773 err = -ENOBUFS;
1774
1775 if (msg_sys.msg_controllen > INT_MAX)
1776 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001777 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001779 err =
1780 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1781 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 if (err)
1783 goto out_freeiov;
1784 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001785 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001787 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001789 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 goto out_freeiov;
1791 }
1792 err = -EFAULT;
1793 /*
1794 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1795 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1796 * checking falls down on this.
1797 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001798 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1799 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 goto out_freectl;
1801 msg_sys.msg_control = ctl_buf;
1802 }
1803 msg_sys.msg_flags = flags;
1804
1805 if (sock->file->f_flags & O_NONBLOCK)
1806 msg_sys.msg_flags |= MSG_DONTWAIT;
1807 err = sock_sendmsg(sock, &msg_sys, total_len);
1808
1809out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001810 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1812out_freeiov:
1813 if (iov != iovstack)
1814 sock_kfree_s(sock->sk, iov, iov_size);
1815out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001816 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001817out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 return err;
1819}
1820
1821/*
1822 * BSD recvmsg interface
1823 */
1824
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001825asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1826 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001828 struct compat_msghdr __user *msg_compat =
1829 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 struct socket *sock;
1831 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001832 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 struct msghdr msg_sys;
1834 unsigned long cmsg_ptr;
1835 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001836 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837
1838 /* kernel mode address */
1839 char addr[MAX_SOCK_ADDR];
1840
1841 /* user mode address pointers */
1842 struct sockaddr __user *uaddr;
1843 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001844
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 if (MSG_CMSG_COMPAT & flags) {
1846 if (get_compat_msghdr(&msg_sys, msg_compat))
1847 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001848 }
1849 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1850 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001852 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 if (!sock)
1854 goto out;
1855
1856 err = -EMSGSIZE;
1857 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1858 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001859
1860 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 err = -ENOMEM;
1862 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1863 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1864 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1865 if (!iov)
1866 goto out_put;
1867 }
1868
1869 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001870 * Save the user-mode address (verify_iovec will change the
1871 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001873
1874 uaddr = (void __user *)msg_sys.msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 uaddr_len = COMPAT_NAMELEN(msg);
1876 if (MSG_CMSG_COMPAT & flags) {
1877 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1878 } else
1879 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1880 if (err < 0)
1881 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001882 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883
1884 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1885 msg_sys.msg_flags = 0;
1886 if (MSG_CMSG_COMPAT & flags)
1887 msg_sys.msg_flags = MSG_CMSG_COMPAT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001888
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 if (sock->file->f_flags & O_NONBLOCK)
1890 flags |= MSG_DONTWAIT;
1891 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1892 if (err < 0)
1893 goto out_freeiov;
1894 len = err;
1895
1896 if (uaddr != NULL) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001897 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1898 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 if (err < 0)
1900 goto out_freeiov;
1901 }
David S. Miller37f7f422005-09-16 16:51:01 -07001902 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1903 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 if (err)
1905 goto out_freeiov;
1906 if (MSG_CMSG_COMPAT & flags)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001907 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908 &msg_compat->msg_controllen);
1909 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001910 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 &msg->msg_controllen);
1912 if (err)
1913 goto out_freeiov;
1914 err = len;
1915
1916out_freeiov:
1917 if (iov != iovstack)
1918 sock_kfree_s(sock->sk, iov, iov_size);
1919out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001920 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921out:
1922 return err;
1923}
1924
1925#ifdef __ARCH_WANT_SYS_SOCKETCALL
1926
1927/* Argument list sizes for sys_socketcall */
1928#define AL(x) ((x) * sizeof(unsigned long))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001929static const unsigned char nargs[18]={
1930 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1931 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1932 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1933};
1934
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935#undef AL
1936
1937/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001938 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 *
1940 * Argument checking cleaned up. Saved 20% in size.
1941 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001942 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 */
1944
1945asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1946{
1947 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001948 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 int err;
1950
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001951 if (call < 1 || call > SYS_RECVMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 return -EINVAL;
1953
1954 /* copy_from_user should be SMP safe. */
1955 if (copy_from_user(a, args, nargs[call]))
1956 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001957
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001958 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001959 if (err)
1960 return err;
1961
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001962 a0 = a[0];
1963 a1 = a[1];
1964
1965 switch (call) {
1966 case SYS_SOCKET:
1967 err = sys_socket(a0, a1, a[2]);
1968 break;
1969 case SYS_BIND:
1970 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
1971 break;
1972 case SYS_CONNECT:
1973 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1974 break;
1975 case SYS_LISTEN:
1976 err = sys_listen(a0, a1);
1977 break;
1978 case SYS_ACCEPT:
1979 err =
1980 sys_accept(a0, (struct sockaddr __user *)a1,
1981 (int __user *)a[2]);
1982 break;
1983 case SYS_GETSOCKNAME:
1984 err =
1985 sys_getsockname(a0, (struct sockaddr __user *)a1,
1986 (int __user *)a[2]);
1987 break;
1988 case SYS_GETPEERNAME:
1989 err =
1990 sys_getpeername(a0, (struct sockaddr __user *)a1,
1991 (int __user *)a[2]);
1992 break;
1993 case SYS_SOCKETPAIR:
1994 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
1995 break;
1996 case SYS_SEND:
1997 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1998 break;
1999 case SYS_SENDTO:
2000 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2001 (struct sockaddr __user *)a[4], a[5]);
2002 break;
2003 case SYS_RECV:
2004 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2005 break;
2006 case SYS_RECVFROM:
2007 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2008 (struct sockaddr __user *)a[4],
2009 (int __user *)a[5]);
2010 break;
2011 case SYS_SHUTDOWN:
2012 err = sys_shutdown(a0, a1);
2013 break;
2014 case SYS_SETSOCKOPT:
2015 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2016 break;
2017 case SYS_GETSOCKOPT:
2018 err =
2019 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2020 (int __user *)a[4]);
2021 break;
2022 case SYS_SENDMSG:
2023 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2024 break;
2025 case SYS_RECVMSG:
2026 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2027 break;
2028 default:
2029 err = -EINVAL;
2030 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 }
2032 return err;
2033}
2034
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002035#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002037/**
2038 * sock_register - add a socket protocol handler
2039 * @ops: description of protocol
2040 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 * This function is called by a protocol handler that wants to
2042 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002043 * socket interface. The value ops->family coresponds to the
2044 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002046int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047{
2048 int err;
2049
2050 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002051 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2052 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053 return -ENOBUFS;
2054 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002055
2056 spin_lock(&net_family_lock);
2057 if (net_families[ops->family])
2058 err = -EEXIST;
2059 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002060 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 err = 0;
2062 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002063 spin_unlock(&net_family_lock);
2064
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002065 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 return err;
2067}
2068
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002069/**
2070 * sock_unregister - remove a protocol handler
2071 * @family: protocol family to remove
2072 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073 * This function is called by a protocol handler that wants to
2074 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002075 * new socket creation.
2076 *
2077 * If protocol handler is a module, then it can use module reference
2078 * counts to protect against new references. If protocol handler is not
2079 * a module then it needs to provide its own protection in
2080 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002082void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002084 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002086 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002087 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002088 spin_unlock(&net_family_lock);
2089
2090 synchronize_rcu();
2091
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002092 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093}
2094
Andi Kleen77d76ea2005-12-22 12:43:42 -08002095static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096{
2097 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002098 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002100
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 sk_init();
2102
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002104 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 */
2106 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107
2108 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002109 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 */
2111
2112 init_inodecache();
2113 register_filesystem(&sock_fs_type);
2114 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002115
2116 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 */
2118
2119#ifdef CONFIG_NETFILTER
2120 netfilter_init();
2121#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002122
2123 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124}
2125
Andi Kleen77d76ea2005-12-22 12:43:42 -08002126core_initcall(sock_init); /* early initcall */
2127
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128#ifdef CONFIG_PROC_FS
2129void socket_seq_show(struct seq_file *seq)
2130{
2131 int cpu;
2132 int counter = 0;
2133
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002134 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002135 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136
2137 /* It can be negative, by the way. 8) */
2138 if (counter < 0)
2139 counter = 0;
2140
2141 seq_printf(seq, "sockets: used %d\n", counter);
2142}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002143#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002145#ifdef CONFIG_COMPAT
2146static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002147 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002148{
2149 struct socket *sock = file->private_data;
2150 int ret = -ENOIOCTLCMD;
2151
2152 if (sock->ops->compat_ioctl)
2153 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2154
2155 return ret;
2156}
2157#endif
2158
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002159int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2160{
2161 return sock->ops->bind(sock, addr, addrlen);
2162}
2163
2164int kernel_listen(struct socket *sock, int backlog)
2165{
2166 return sock->ops->listen(sock, backlog);
2167}
2168
2169int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2170{
2171 struct sock *sk = sock->sk;
2172 int err;
2173
2174 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2175 newsock);
2176 if (err < 0)
2177 goto done;
2178
2179 err = sock->ops->accept(sock, *newsock, flags);
2180 if (err < 0) {
2181 sock_release(*newsock);
2182 goto done;
2183 }
2184
2185 (*newsock)->ops = sock->ops;
2186
2187done:
2188 return err;
2189}
2190
2191int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2192 int flags)
2193{
2194 return sock->ops->connect(sock, addr, addrlen, flags);
2195}
2196
2197int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2198 int *addrlen)
2199{
2200 return sock->ops->getname(sock, addr, addrlen, 0);
2201}
2202
2203int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2204 int *addrlen)
2205{
2206 return sock->ops->getname(sock, addr, addrlen, 1);
2207}
2208
2209int kernel_getsockopt(struct socket *sock, int level, int optname,
2210 char *optval, int *optlen)
2211{
2212 mm_segment_t oldfs = get_fs();
2213 int err;
2214
2215 set_fs(KERNEL_DS);
2216 if (level == SOL_SOCKET)
2217 err = sock_getsockopt(sock, level, optname, optval, optlen);
2218 else
2219 err = sock->ops->getsockopt(sock, level, optname, optval,
2220 optlen);
2221 set_fs(oldfs);
2222 return err;
2223}
2224
2225int kernel_setsockopt(struct socket *sock, int level, int optname,
2226 char *optval, int optlen)
2227{
2228 mm_segment_t oldfs = get_fs();
2229 int err;
2230
2231 set_fs(KERNEL_DS);
2232 if (level == SOL_SOCKET)
2233 err = sock_setsockopt(sock, level, optname, optval, optlen);
2234 else
2235 err = sock->ops->setsockopt(sock, level, optname, optval,
2236 optlen);
2237 set_fs(oldfs);
2238 return err;
2239}
2240
2241int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2242 size_t size, int flags)
2243{
2244 if (sock->ops->sendpage)
2245 return sock->ops->sendpage(sock, page, offset, size, flags);
2246
2247 return sock_no_sendpage(sock, page, offset, size, flags);
2248}
2249
2250int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2251{
2252 mm_segment_t oldfs = get_fs();
2253 int err;
2254
2255 set_fs(KERNEL_DS);
2256 err = sock->ops->ioctl(sock, cmd, arg);
2257 set_fs(oldfs);
2258
2259 return err;
2260}
2261
Linus Torvalds1da177e2005-04-16 15:20:36 -07002262/* ABI emulation layers need these two */
2263EXPORT_SYMBOL(move_addr_to_kernel);
2264EXPORT_SYMBOL(move_addr_to_user);
2265EXPORT_SYMBOL(sock_create);
2266EXPORT_SYMBOL(sock_create_kern);
2267EXPORT_SYMBOL(sock_create_lite);
2268EXPORT_SYMBOL(sock_map_fd);
2269EXPORT_SYMBOL(sock_recvmsg);
2270EXPORT_SYMBOL(sock_register);
2271EXPORT_SYMBOL(sock_release);
2272EXPORT_SYMBOL(sock_sendmsg);
2273EXPORT_SYMBOL(sock_unregister);
2274EXPORT_SYMBOL(sock_wake_async);
2275EXPORT_SYMBOL(sockfd_lookup);
2276EXPORT_SYMBOL(kernel_sendmsg);
2277EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002278EXPORT_SYMBOL(kernel_bind);
2279EXPORT_SYMBOL(kernel_listen);
2280EXPORT_SYMBOL(kernel_accept);
2281EXPORT_SYMBOL(kernel_connect);
2282EXPORT_SYMBOL(kernel_getsockname);
2283EXPORT_SYMBOL(kernel_getpeername);
2284EXPORT_SYMBOL(kernel_getsockopt);
2285EXPORT_SYMBOL(kernel_setsockopt);
2286EXPORT_SYMBOL(kernel_sendpage);
2287EXPORT_SYMBOL(kernel_sock_ioctl);