blob: 0778c54424117ee48c6337714f3fe72403d4f7ae [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
Arjan van de Venda7071d2007-02-12 00:55:36 -0800120static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121 .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
Josef Sipek3126a422006-12-08 02:37:23 -0800365 file->f_path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
366 if (unlikely(!file->f_path.dentry))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800367 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368
Josef Sipek3126a422006-12-08 02:37:23 -0800369 file->f_path.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 */
Josef Sipek3126a422006-12-08 02:37:23 -0800375 file->f_path.dentry->d_flags &= ~DCACHE_UNHASHED;
376 d_instantiate(file->f_path.dentry, SOCK_INODE(sock));
377 file->f_path.mnt = mntget(sock_mnt);
378 file->f_mapping = file->f_path.dentry->d_inode->i_mapping;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800379
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{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800410 if (file->f_op == &socket_file_ops)
411 return file->private_data; /* set in sock_map_fd */
412
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800413 *err = -ENOTSOCK;
414 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800415}
416
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417/**
418 * sockfd_lookup - Go from a file number to its socket slot
419 * @fd: file handle
420 * @err: pointer to an error code return
421 *
422 * The file handle passed in is locked and the socket it is bound
423 * too is returned. If an error occurs the err pointer is overwritten
424 * with a negative errno code and NULL is returned. The function checks
425 * for both invalid handles and passing a handle which is not a socket.
426 *
427 * On a success the socket object pointer is returned.
428 */
429
430struct socket *sockfd_lookup(int fd, int *err)
431{
432 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 struct socket *sock;
434
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700435 file = fget(fd);
436 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 *err = -EBADF;
438 return NULL;
439 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700440
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800441 sock = sock_from_file(file, err);
442 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 return sock;
445}
446
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800447static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
448{
449 struct file *file;
450 struct socket *sock;
451
Hua Zhong36725582006-04-19 15:25:02 -0700452 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800453 file = fget_light(fd, fput_needed);
454 if (file) {
455 sock = sock_from_file(file, err);
456 if (sock)
457 return sock;
458 fput_light(file, *fput_needed);
459 }
460 return NULL;
461}
462
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463/**
464 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700465 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 * Allocate a new inode and socket object. The two are bound together
467 * and initialised. The socket is then returned. If we are out of inodes
468 * NULL is returned.
469 */
470
471static struct socket *sock_alloc(void)
472{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700473 struct inode *inode;
474 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475
476 inode = new_inode(sock_mnt->mnt_sb);
477 if (!inode)
478 return NULL;
479
480 sock = SOCKET_I(inode);
481
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700482 inode->i_mode = S_IFSOCK | S_IRWXUGO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 inode->i_uid = current->fsuid;
484 inode->i_gid = current->fsgid;
485
486 get_cpu_var(sockets_in_use)++;
487 put_cpu_var(sockets_in_use);
488 return sock;
489}
490
491/*
492 * In theory you can't get an open on this inode, but /proc provides
493 * a back door. Remember to keep it shut otherwise you'll let the
494 * creepy crawlies in.
495 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
498{
499 return -ENXIO;
500}
501
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800502const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 .owner = THIS_MODULE,
504 .open = sock_no_open,
505};
506
507/**
508 * sock_release - close a socket
509 * @sock: socket to close
510 *
511 * The socket is released from the protocol stack if it has a release
512 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700513 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516void sock_release(struct socket *sock)
517{
518 if (sock->ops) {
519 struct module *owner = sock->ops->owner;
520
521 sock->ops->release(sock);
522 sock->ops = NULL;
523 module_put(owner);
524 }
525
526 if (sock->fasync_list)
527 printk(KERN_ERR "sock_release: fasync list not empty!\n");
528
529 get_cpu_var(sockets_in_use)--;
530 put_cpu_var(sockets_in_use);
531 if (!sock->file) {
532 iput(SOCK_INODE(sock));
533 return;
534 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700535 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536}
537
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700538static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 struct msghdr *msg, size_t size)
540{
541 struct sock_iocb *si = kiocb_to_siocb(iocb);
542 int err;
543
544 si->sock = sock;
545 si->scm = NULL;
546 si->msg = msg;
547 si->size = size;
548
549 err = security_socket_sendmsg(sock, msg, size);
550 if (err)
551 return err;
552
553 return sock->ops->sendmsg(iocb, sock, msg, size);
554}
555
556int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
557{
558 struct kiocb iocb;
559 struct sock_iocb siocb;
560 int ret;
561
562 init_sync_kiocb(&iocb, NULL);
563 iocb.private = &siocb;
564 ret = __sock_sendmsg(&iocb, sock, msg, size);
565 if (-EIOCBQUEUED == ret)
566 ret = wait_on_sync_kiocb(&iocb);
567 return ret;
568}
569
570int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
571 struct kvec *vec, size_t num, size_t size)
572{
573 mm_segment_t oldfs = get_fs();
574 int result;
575
576 set_fs(KERNEL_DS);
577 /*
578 * the following is safe, since for compiler definitions of kvec and
579 * iovec are identical, yielding the same in-core layout and alignment
580 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700581 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 msg->msg_iovlen = num;
583 result = sock_sendmsg(sock, msg, size);
584 set_fs(oldfs);
585 return result;
586}
587
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700588static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 struct msghdr *msg, size_t size, int flags)
590{
591 int err;
592 struct sock_iocb *si = kiocb_to_siocb(iocb);
593
594 si->sock = sock;
595 si->scm = NULL;
596 si->msg = msg;
597 si->size = size;
598 si->flags = flags;
599
600 err = security_socket_recvmsg(sock, msg, size, flags);
601 if (err)
602 return err;
603
604 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
605}
606
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700607int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 size_t size, int flags)
609{
610 struct kiocb iocb;
611 struct sock_iocb siocb;
612 int ret;
613
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700614 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 iocb.private = &siocb;
616 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
617 if (-EIOCBQUEUED == ret)
618 ret = wait_on_sync_kiocb(&iocb);
619 return ret;
620}
621
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700622int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
623 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624{
625 mm_segment_t oldfs = get_fs();
626 int result;
627
628 set_fs(KERNEL_DS);
629 /*
630 * the following is safe, since for compiler definitions of kvec and
631 * iovec are identical, yielding the same in-core layout and alignment
632 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700633 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 result = sock_recvmsg(sock, msg, size, flags);
635 set_fs(oldfs);
636 return result;
637}
638
639static void sock_aio_dtor(struct kiocb *iocb)
640{
641 kfree(iocb->private);
642}
643
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300644static ssize_t sock_sendpage(struct file *file, struct page *page,
645 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646{
647 struct socket *sock;
648 int flags;
649
Eric Dumazetb69aee02005-09-06 14:42:45 -0700650 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
652 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
653 if (more)
654 flags |= MSG_MORE;
655
656 return sock->ops->sendpage(sock, page, offset, size, flags);
657}
658
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800659static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700660 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800661{
662 if (!is_sync_kiocb(iocb)) {
663 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
664 if (!siocb)
665 return NULL;
666 iocb->ki_dtor = sock_aio_dtor;
667 }
668
669 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800670 iocb->private = siocb;
671 return siocb;
672}
673
674static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700675 struct file *file, const struct iovec *iov,
676 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800677{
678 struct socket *sock = file->private_data;
679 size_t size = 0;
680 int i;
681
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700682 for (i = 0; i < nr_segs; i++)
683 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800684
685 msg->msg_name = NULL;
686 msg->msg_namelen = 0;
687 msg->msg_control = NULL;
688 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700689 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800690 msg->msg_iovlen = nr_segs;
691 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
692
693 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
694}
695
Badari Pulavarty027445c2006-09-30 23:28:46 -0700696static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
697 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800698{
699 struct sock_iocb siocb, *x;
700
701 if (pos != 0)
702 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700703
704 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800705 return 0;
706
Badari Pulavarty027445c2006-09-30 23:28:46 -0700707
708 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800709 if (!x)
710 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700711 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800712}
713
714static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700715 struct file *file, const struct iovec *iov,
716 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800717{
718 struct socket *sock = file->private_data;
719 size_t size = 0;
720 int i;
721
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700722 for (i = 0; i < nr_segs; i++)
723 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800724
725 msg->msg_name = NULL;
726 msg->msg_namelen = 0;
727 msg->msg_control = NULL;
728 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700729 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800730 msg->msg_iovlen = nr_segs;
731 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
732 if (sock->type == SOCK_SEQPACKET)
733 msg->msg_flags |= MSG_EOR;
734
735 return __sock_sendmsg(iocb, sock, msg, size);
736}
737
Badari Pulavarty027445c2006-09-30 23:28:46 -0700738static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
739 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800741 struct sock_iocb siocb, *x;
742
743 if (pos != 0)
744 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700745
746 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800747 return 0;
748
Badari Pulavarty027445c2006-09-30 23:28:46 -0700749 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800750 if (!x)
751 return -ENOMEM;
752
Badari Pulavarty027445c2006-09-30 23:28:46 -0700753 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754}
755
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756/*
757 * Atomic setting of ioctl hooks to avoid race
758 * with module unload.
759 */
760
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800761static DEFINE_MUTEX(br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700762static int (*br_ioctl_hook) (unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700764void brioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800766 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800768 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700770
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771EXPORT_SYMBOL(brioctl_set);
772
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800773static DEFINE_MUTEX(vlan_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700774static int (*vlan_ioctl_hook) (void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700776void vlan_ioctl_set(int (*hook) (void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800778 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800780 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700782
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783EXPORT_SYMBOL(vlan_ioctl_set);
784
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800785static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700786static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700788void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800790 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800792 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700794
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795EXPORT_SYMBOL(dlci_ioctl_set);
796
797/*
798 * With an ioctl, arg may well be a user mode pointer, but we don't know
799 * what to do with it - that's up to the protocol still.
800 */
801
802static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
803{
804 struct socket *sock;
805 void __user *argp = (void __user *)arg;
806 int pid, err;
807
Eric Dumazetb69aee02005-09-06 14:42:45 -0700808 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
810 err = dev_ioctl(cmd, argp);
811 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100812#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
814 err = dev_ioctl(cmd, argp);
815 } else
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700816#endif /* CONFIG_WIRELESS_EXT */
817 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 case FIOSETOWN:
819 case SIOCSPGRP:
820 err = -EFAULT;
821 if (get_user(pid, (int __user *)argp))
822 break;
823 err = f_setown(sock->file, pid, 1);
824 break;
825 case FIOGETOWN:
826 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700827 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700828 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 break;
830 case SIOCGIFBR:
831 case SIOCSIFBR:
832 case SIOCBRADDBR:
833 case SIOCBRDELBR:
834 err = -ENOPKG;
835 if (!br_ioctl_hook)
836 request_module("bridge");
837
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800838 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700839 if (br_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 err = br_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800841 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 break;
843 case SIOCGIFVLAN:
844 case SIOCSIFVLAN:
845 err = -ENOPKG;
846 if (!vlan_ioctl_hook)
847 request_module("8021q");
848
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800849 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 if (vlan_ioctl_hook)
851 err = vlan_ioctl_hook(argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800852 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 case SIOCADDDLCI:
855 case SIOCDELDLCI:
856 err = -ENOPKG;
857 if (!dlci_ioctl_hook)
858 request_module("dlci");
859
860 if (dlci_ioctl_hook) {
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800861 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 err = dlci_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800863 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 }
865 break;
866 default:
867 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800868
869 /*
870 * If this ioctl is unknown try to hand it down
871 * to the NIC driver.
872 */
873 if (err == -ENOIOCTLCMD)
874 err = dev_ioctl(cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700876 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 return err;
878}
879
880int sock_create_lite(int family, int type, int protocol, struct socket **res)
881{
882 int err;
883 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700884
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 err = security_socket_create(family, type, protocol, 1);
886 if (err)
887 goto out;
888
889 sock = sock_alloc();
890 if (!sock) {
891 err = -ENOMEM;
892 goto out;
893 }
894
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700896 err = security_socket_post_create(sock, family, type, protocol, 1);
897 if (err)
898 goto out_release;
899
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900out:
901 *res = sock;
902 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700903out_release:
904 sock_release(sock);
905 sock = NULL;
906 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907}
908
909/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700910static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911{
912 struct socket *sock;
913
914 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700915 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700917 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 return sock->ops->poll(file, sock, wait);
919}
920
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700921static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700923 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924
925 return sock->ops->mmap(file, sock, vma);
926}
927
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300928static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929{
930 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700931 * It was possible the inode is NULL we were
932 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 */
934
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700935 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 printk(KERN_DEBUG "sock_close: NULL inode\n");
937 return 0;
938 }
939 sock_fasync(-1, filp, 0);
940 sock_release(SOCKET_I(inode));
941 return 0;
942}
943
944/*
945 * Update the socket async list
946 *
947 * Fasync_list locking strategy.
948 *
949 * 1. fasync_list is modified only under process context socket lock
950 * i.e. under semaphore.
951 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
952 * or under socket lock.
953 * 3. fasync_list can be used from softirq context, so that
954 * modification under socket lock have to be enhanced with
955 * write_lock_bh(&sk->sk_callback_lock).
956 * --ANK (990710)
957 */
958
959static int sock_fasync(int fd, struct file *filp, int on)
960{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700961 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 struct socket *sock;
963 struct sock *sk;
964
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700965 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800966 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700967 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 return -ENOMEM;
969 }
970
Eric Dumazetb69aee02005-09-06 14:42:45 -0700971 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700973 sk = sock->sk;
974 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 kfree(fna);
976 return -EINVAL;
977 }
978
979 lock_sock(sk);
980
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700981 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700983 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
984 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 break;
986
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700987 if (on) {
988 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700990 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 write_unlock_bh(&sk->sk_callback_lock);
992
993 kfree(fna);
994 goto out;
995 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700996 fna->fa_file = filp;
997 fna->fa_fd = fd;
998 fna->magic = FASYNC_MAGIC;
999 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001001 sock->fasync_list = fna;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001003 } else {
1004 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001006 *prev = fa->fa_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 write_unlock_bh(&sk->sk_callback_lock);
1008 kfree(fa);
1009 }
1010 }
1011
1012out:
1013 release_sock(sock->sk);
1014 return 0;
1015}
1016
1017/* This function may be called only under socket lock or callback_lock */
1018
1019int sock_wake_async(struct socket *sock, int how, int band)
1020{
1021 if (!sock || !sock->fasync_list)
1022 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001023 switch (how) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 case 1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1027 break;
1028 goto call_kill;
1029 case 2:
1030 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1031 break;
1032 /* fall through */
1033 case 0:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001034call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 __kill_fasync(sock->fasync_list, SIGIO, band);
1036 break;
1037 case 3:
1038 __kill_fasync(sock->fasync_list, SIGURG, band);
1039 }
1040 return 0;
1041}
1042
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001043static int __sock_create(int family, int type, int protocol,
1044 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045{
1046 int err;
1047 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001048 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049
1050 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001051 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 */
1053 if (family < 0 || family >= NPROTO)
1054 return -EAFNOSUPPORT;
1055 if (type < 0 || type >= SOCK_MAX)
1056 return -EINVAL;
1057
1058 /* Compatibility.
1059
1060 This uglymoron is moved from INET layer to here to avoid
1061 deadlock in module load.
1062 */
1063 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001064 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 if (!warned) {
1066 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001067 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1068 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 }
1070 family = PF_PACKET;
1071 }
1072
1073 err = security_socket_create(family, type, protocol, kern);
1074 if (err)
1075 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001076
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001077 /*
1078 * Allocate the socket and allow the family to set things up. if
1079 * the protocol is 0, the family is instructed to select an appropriate
1080 * default.
1081 */
1082 sock = sock_alloc();
1083 if (!sock) {
1084 if (net_ratelimit())
1085 printk(KERN_WARNING "socket: no more sockets\n");
1086 return -ENFILE; /* Not exactly a match, but its the
1087 closest posix thing */
1088 }
1089
1090 sock->type = type;
1091
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092#if defined(CONFIG_KMOD)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001093 /* Attempt to load a protocol module if the find failed.
1094 *
1095 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 * requested real, full-featured networking support upon configuration.
1097 * Otherwise module support will break!
1098 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001099 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001100 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101#endif
1102
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001103 rcu_read_lock();
1104 pf = rcu_dereference(net_families[family]);
1105 err = -EAFNOSUPPORT;
1106 if (!pf)
1107 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
1109 /*
1110 * We will call the ->create function, that possibly is in a loadable
1111 * module, so we have to bump that loadable module refcnt first.
1112 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001113 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 goto out_release;
1115
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001116 /* Now protected by module ref count */
1117 rcu_read_unlock();
1118
1119 err = pf->create(sock, protocol);
1120 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001122
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 /*
1124 * Now to bump the refcnt of the [loadable] module that owns this
1125 * socket at sock_release time we decrement its refcnt.
1126 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001127 if (!try_module_get(sock->ops->owner))
1128 goto out_module_busy;
1129
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 /*
1131 * Now that we're done with the ->create function, the [loadable]
1132 * module can have its refcnt decremented
1133 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001134 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001135 err = security_socket_post_create(sock, family, type, protocol, kern);
1136 if (err)
1137 goto out_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001138 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001140 return 0;
1141
1142out_module_busy:
1143 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001145 sock->ops = NULL;
1146 module_put(pf->owner);
1147out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001149 return err;
1150
1151out_release:
1152 rcu_read_unlock();
1153 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154}
1155
1156int sock_create(int family, int type, int protocol, struct socket **res)
1157{
1158 return __sock_create(family, type, protocol, res, 0);
1159}
1160
1161int sock_create_kern(int family, int type, int protocol, struct socket **res)
1162{
1163 return __sock_create(family, type, protocol, res, 1);
1164}
1165
1166asmlinkage long sys_socket(int family, int type, int protocol)
1167{
1168 int retval;
1169 struct socket *sock;
1170
1171 retval = sock_create(family, type, protocol, &sock);
1172 if (retval < 0)
1173 goto out;
1174
1175 retval = sock_map_fd(sock);
1176 if (retval < 0)
1177 goto out_release;
1178
1179out:
1180 /* It may be already another descriptor 8) Not kernel problem. */
1181 return retval;
1182
1183out_release:
1184 sock_release(sock);
1185 return retval;
1186}
1187
1188/*
1189 * Create a pair of connected sockets.
1190 */
1191
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001192asmlinkage long sys_socketpair(int family, int type, int protocol,
1193 int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194{
1195 struct socket *sock1, *sock2;
1196 int fd1, fd2, err;
1197
1198 /*
1199 * Obtain the first socket and check if the underlying protocol
1200 * supports the socketpair call.
1201 */
1202
1203 err = sock_create(family, type, protocol, &sock1);
1204 if (err < 0)
1205 goto out;
1206
1207 err = sock_create(family, type, protocol, &sock2);
1208 if (err < 0)
1209 goto out_release_1;
1210
1211 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001212 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 goto out_release_both;
1214
1215 fd1 = fd2 = -1;
1216
1217 err = sock_map_fd(sock1);
1218 if (err < 0)
1219 goto out_release_both;
1220 fd1 = err;
1221
1222 err = sock_map_fd(sock2);
1223 if (err < 0)
1224 goto out_close_1;
1225 fd2 = err;
1226
1227 /* fd1 and fd2 may be already another descriptors.
1228 * Not kernel problem.
1229 */
1230
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001231 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 if (!err)
1233 err = put_user(fd2, &usockvec[1]);
1234 if (!err)
1235 return 0;
1236
1237 sys_close(fd2);
1238 sys_close(fd1);
1239 return err;
1240
1241out_close_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001242 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 sys_close(fd1);
1244 return err;
1245
1246out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001247 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001249 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250out:
1251 return err;
1252}
1253
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254/*
1255 * Bind a name to a socket. Nothing much to do here since it's
1256 * the protocol's responsibility to handle the local address.
1257 *
1258 * We move the socket address to kernel space before we call
1259 * the protocol layer (having also checked the address is ok).
1260 */
1261
1262asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1263{
1264 struct socket *sock;
1265 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001266 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001268 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1269 if(sock) {
1270 err = move_addr_to_kernel(umyaddr, addrlen, address);
1271 if (err >= 0) {
1272 err = security_socket_bind(sock,
1273 (struct sockaddr *)address,
1274 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001275 if (!err)
1276 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001277 (struct sockaddr *)
1278 address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001280 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001281 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 return err;
1283}
1284
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285/*
1286 * Perform a listen. Basically, we allow the protocol to do anything
1287 * necessary for a listen, and if that works, we mark the socket as
1288 * ready for listening.
1289 */
1290
Brian Haley7a42c212006-08-31 15:03:02 -07001291int sysctl_somaxconn __read_mostly = SOMAXCONN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292
1293asmlinkage long sys_listen(int fd, int backlog)
1294{
1295 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001296 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001297
1298 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1299 if (sock) {
1300 if ((unsigned)backlog > sysctl_somaxconn)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 backlog = sysctl_somaxconn;
1302
1303 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001304 if (!err)
1305 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001307 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 }
1309 return err;
1310}
1311
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312/*
1313 * For accept, we attempt to create a new socket, set up the link
1314 * with the client, wake up the client, then return the new
1315 * connected fd. We collect the address of the connector in kernel
1316 * space and move it to user at the very end. This is unclean because
1317 * we open the socket then return an error.
1318 *
1319 * 1003.1g adds the ability to recvmsg() to query connection pending
1320 * status to recvmsg. We need to add that support in a way thats
1321 * clean when we restucture accept also.
1322 */
1323
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001324asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1325 int __user *upeer_addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326{
1327 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001328 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001329 int err, len, newfd, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 char address[MAX_SOCK_ADDR];
1331
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001332 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 if (!sock)
1334 goto out;
1335
1336 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001337 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 goto out_put;
1339
1340 newsock->type = sock->type;
1341 newsock->ops = sock->ops;
1342
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 /*
1344 * We don't need try_module_get here, as the listening socket (sock)
1345 * has the protocol module (sock->ops->owner) held.
1346 */
1347 __module_get(newsock->ops->owner);
1348
David S. Miller39d8c1b2006-03-20 17:13:49 -08001349 newfd = sock_alloc_fd(&newfile);
1350 if (unlikely(newfd < 0)) {
1351 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001352 sock_release(newsock);
1353 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001354 }
1355
1356 err = sock_attach_fd(newsock, newfile);
1357 if (err < 0)
1358 goto out_fd;
1359
Frank Filza79af592005-09-27 15:23:38 -07001360 err = security_socket_accept(sock, newsock);
1361 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001362 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001363
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1365 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001366 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
1368 if (upeer_sockaddr) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001369 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1370 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001372 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001374 err = move_addr_to_user(address, len, upeer_sockaddr,
1375 upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001377 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 }
1379
1380 /* File flags are not inherited via accept() unlike another OSes. */
1381
David S. Miller39d8c1b2006-03-20 17:13:49 -08001382 fd_install(newfd, newfile);
1383 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384
1385 security_socket_post_accept(sock, newsock);
1386
1387out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001388 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389out:
1390 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001391out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001392 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001393 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 goto out_put;
1395}
1396
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397/*
1398 * Attempt to connect to a socket with the server address. The address
1399 * is in user space so we verify it is OK and move it to kernel space.
1400 *
1401 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1402 * break bindings
1403 *
1404 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1405 * other SEQPACKET protocols that take time to connect() as it doesn't
1406 * include the -EINPROGRESS status for such sockets.
1407 */
1408
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001409asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1410 int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411{
1412 struct socket *sock;
1413 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001414 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001416 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 if (!sock)
1418 goto out;
1419 err = move_addr_to_kernel(uservaddr, addrlen, address);
1420 if (err < 0)
1421 goto out_put;
1422
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001423 err =
1424 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 if (err)
1426 goto out_put;
1427
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001428 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 sock->file->f_flags);
1430out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001431 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432out:
1433 return err;
1434}
1435
1436/*
1437 * Get the local address ('name') of a socket object. Move the obtained
1438 * name to user space.
1439 */
1440
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001441asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1442 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443{
1444 struct socket *sock;
1445 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001446 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001447
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001448 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 if (!sock)
1450 goto out;
1451
1452 err = security_socket_getsockname(sock);
1453 if (err)
1454 goto out_put;
1455
1456 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1457 if (err)
1458 goto out_put;
1459 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1460
1461out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001462 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463out:
1464 return err;
1465}
1466
1467/*
1468 * Get the remote address ('name') of a socket object. Move the obtained
1469 * name to user space.
1470 */
1471
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001472asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1473 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474{
1475 struct socket *sock;
1476 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001477 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001479 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1480 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 err = security_socket_getpeername(sock);
1482 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001483 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 return err;
1485 }
1486
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001487 err =
1488 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1489 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 if (!err)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001491 err = move_addr_to_user(address, len, usockaddr,
1492 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001493 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 }
1495 return err;
1496}
1497
1498/*
1499 * Send a datagram to a given address. We move the address into kernel
1500 * space and check the user space data area is readable before invoking
1501 * the protocol.
1502 */
1503
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001504asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1505 unsigned flags, struct sockaddr __user *addr,
1506 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507{
1508 struct socket *sock;
1509 char address[MAX_SOCK_ADDR];
1510 int err;
1511 struct msghdr msg;
1512 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001513 int fput_needed;
1514 struct file *sock_file;
1515
1516 sock_file = fget_light(fd, &fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001517 err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001518 if (!sock_file)
David S. Miller4387ff72007-02-08 15:06:08 -08001519 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001520
1521 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 if (!sock)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001523 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001524 iov.iov_base = buff;
1525 iov.iov_len = len;
1526 msg.msg_name = NULL;
1527 msg.msg_iov = &iov;
1528 msg.msg_iovlen = 1;
1529 msg.msg_control = NULL;
1530 msg.msg_controllen = 0;
1531 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001532 if (addr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 err = move_addr_to_kernel(addr, addr_len, address);
1534 if (err < 0)
1535 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001536 msg.msg_name = address;
1537 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 }
1539 if (sock->file->f_flags & O_NONBLOCK)
1540 flags |= MSG_DONTWAIT;
1541 msg.msg_flags = flags;
1542 err = sock_sendmsg(sock, &msg, len);
1543
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001544out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001545 fput_light(sock_file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001546out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 return err;
1548}
1549
1550/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001551 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 */
1553
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001554asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555{
1556 return sys_sendto(fd, buff, len, flags, NULL, 0);
1557}
1558
1559/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001560 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 * sender. We verify the buffers are writable and if needed move the
1562 * sender address from kernel to user space.
1563 */
1564
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001565asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1566 unsigned flags, struct sockaddr __user *addr,
1567 int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568{
1569 struct socket *sock;
1570 struct iovec iov;
1571 struct msghdr msg;
1572 char address[MAX_SOCK_ADDR];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001573 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001574 struct file *sock_file;
1575 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001577 sock_file = fget_light(fd, &fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001578 err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001579 if (!sock_file)
David S. Miller4387ff72007-02-08 15:06:08 -08001580 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001581
1582 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001584 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001586 msg.msg_control = NULL;
1587 msg.msg_controllen = 0;
1588 msg.msg_iovlen = 1;
1589 msg.msg_iov = &iov;
1590 iov.iov_len = size;
1591 iov.iov_base = ubuf;
1592 msg.msg_name = address;
1593 msg.msg_namelen = MAX_SOCK_ADDR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 if (sock->file->f_flags & O_NONBLOCK)
1595 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001596 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001598 if (err >= 0 && addr != NULL) {
1599 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1600 if (err2 < 0)
1601 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 }
David S. Miller4387ff72007-02-08 15:06:08 -08001603out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001604 fput_light(sock_file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001605out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 return err;
1607}
1608
1609/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001610 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 */
1612
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001613asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1614 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615{
1616 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1617}
1618
1619/*
1620 * Set a socket option. Because we don't know the option lengths we have
1621 * to pass the user mode parameter for the protocols to sort out.
1622 */
1623
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001624asmlinkage long sys_setsockopt(int fd, int level, int optname,
1625 char __user *optval, int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001627 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 struct socket *sock;
1629
1630 if (optlen < 0)
1631 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001632
1633 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1634 if (sock != NULL) {
1635 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001636 if (err)
1637 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638
1639 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001640 err =
1641 sock_setsockopt(sock, level, optname, optval,
1642 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001644 err =
1645 sock->ops->setsockopt(sock, level, optname, optval,
1646 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001647out_put:
1648 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 }
1650 return err;
1651}
1652
1653/*
1654 * Get a socket option. Because we don't know the option lengths we have
1655 * to pass a user mode parameter for the protocols to sort out.
1656 */
1657
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001658asmlinkage long sys_getsockopt(int fd, int level, int optname,
1659 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001661 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 struct socket *sock;
1663
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001664 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1665 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001666 err = security_socket_getsockopt(sock, level, optname);
1667 if (err)
1668 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669
1670 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001671 err =
1672 sock_getsockopt(sock, level, optname, optval,
1673 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001675 err =
1676 sock->ops->getsockopt(sock, level, optname, optval,
1677 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001678out_put:
1679 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 }
1681 return err;
1682}
1683
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684/*
1685 * Shutdown a socket.
1686 */
1687
1688asmlinkage long sys_shutdown(int fd, int how)
1689{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001690 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 struct socket *sock;
1692
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001693 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1694 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001696 if (!err)
1697 err = sock->ops->shutdown(sock, how);
1698 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 }
1700 return err;
1701}
1702
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001703/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 * fields which are the same type (int / unsigned) on our platforms.
1705 */
1706#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1707#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1708#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1709
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710/*
1711 * BSD sendmsg interface
1712 */
1713
1714asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1715{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001716 struct compat_msghdr __user *msg_compat =
1717 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 struct socket *sock;
1719 char address[MAX_SOCK_ADDR];
1720 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001721 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001722 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1723 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 unsigned char *ctl_buf = ctl;
1725 struct msghdr msg_sys;
1726 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001727 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001728
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 err = -EFAULT;
1730 if (MSG_CMSG_COMPAT & flags) {
1731 if (get_compat_msghdr(&msg_sys, msg_compat))
1732 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001733 }
1734 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 return -EFAULT;
1736
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001737 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001738 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 goto out;
1740
1741 /* do not move before msg_sys is valid */
1742 err = -EMSGSIZE;
1743 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1744 goto out_put;
1745
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001746 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 err = -ENOMEM;
1748 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1749 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1750 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1751 if (!iov)
1752 goto out_put;
1753 }
1754
1755 /* This will also move the address data into kernel space */
1756 if (MSG_CMSG_COMPAT & flags) {
1757 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1758 } else
1759 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001760 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 goto out_freeiov;
1762 total_len = err;
1763
1764 err = -ENOBUFS;
1765
1766 if (msg_sys.msg_controllen > INT_MAX)
1767 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001768 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001770 err =
1771 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1772 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773 if (err)
1774 goto out_freeiov;
1775 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001776 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001778 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001780 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 goto out_freeiov;
1782 }
1783 err = -EFAULT;
1784 /*
1785 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1786 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1787 * checking falls down on this.
1788 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001789 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1790 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 goto out_freectl;
1792 msg_sys.msg_control = ctl_buf;
1793 }
1794 msg_sys.msg_flags = flags;
1795
1796 if (sock->file->f_flags & O_NONBLOCK)
1797 msg_sys.msg_flags |= MSG_DONTWAIT;
1798 err = sock_sendmsg(sock, &msg_sys, total_len);
1799
1800out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001801 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1803out_freeiov:
1804 if (iov != iovstack)
1805 sock_kfree_s(sock->sk, iov, iov_size);
1806out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001807 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001808out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 return err;
1810}
1811
1812/*
1813 * BSD recvmsg interface
1814 */
1815
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001816asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1817 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001819 struct compat_msghdr __user *msg_compat =
1820 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 struct socket *sock;
1822 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001823 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 struct msghdr msg_sys;
1825 unsigned long cmsg_ptr;
1826 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001827 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828
1829 /* kernel mode address */
1830 char addr[MAX_SOCK_ADDR];
1831
1832 /* user mode address pointers */
1833 struct sockaddr __user *uaddr;
1834 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001835
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 if (MSG_CMSG_COMPAT & flags) {
1837 if (get_compat_msghdr(&msg_sys, msg_compat))
1838 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001839 }
1840 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1841 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001843 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 if (!sock)
1845 goto out;
1846
1847 err = -EMSGSIZE;
1848 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1849 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001850
1851 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 err = -ENOMEM;
1853 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1854 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1855 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1856 if (!iov)
1857 goto out_put;
1858 }
1859
1860 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001861 * Save the user-mode address (verify_iovec will change the
1862 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001864
1865 uaddr = (void __user *)msg_sys.msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866 uaddr_len = COMPAT_NAMELEN(msg);
1867 if (MSG_CMSG_COMPAT & flags) {
1868 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1869 } else
1870 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1871 if (err < 0)
1872 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001873 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874
1875 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1876 msg_sys.msg_flags = 0;
1877 if (MSG_CMSG_COMPAT & flags)
1878 msg_sys.msg_flags = MSG_CMSG_COMPAT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001879
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 if (sock->file->f_flags & O_NONBLOCK)
1881 flags |= MSG_DONTWAIT;
1882 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1883 if (err < 0)
1884 goto out_freeiov;
1885 len = err;
1886
1887 if (uaddr != NULL) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001888 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1889 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890 if (err < 0)
1891 goto out_freeiov;
1892 }
David S. Miller37f7f422005-09-16 16:51:01 -07001893 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1894 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 if (err)
1896 goto out_freeiov;
1897 if (MSG_CMSG_COMPAT & flags)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001898 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 &msg_compat->msg_controllen);
1900 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001901 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902 &msg->msg_controllen);
1903 if (err)
1904 goto out_freeiov;
1905 err = len;
1906
1907out_freeiov:
1908 if (iov != iovstack)
1909 sock_kfree_s(sock->sk, iov, iov_size);
1910out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001911 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912out:
1913 return err;
1914}
1915
1916#ifdef __ARCH_WANT_SYS_SOCKETCALL
1917
1918/* Argument list sizes for sys_socketcall */
1919#define AL(x) ((x) * sizeof(unsigned long))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001920static const unsigned char nargs[18]={
1921 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1922 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1923 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1924};
1925
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926#undef AL
1927
1928/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001929 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 *
1931 * Argument checking cleaned up. Saved 20% in size.
1932 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001933 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 */
1935
1936asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1937{
1938 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001939 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 int err;
1941
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001942 if (call < 1 || call > SYS_RECVMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 return -EINVAL;
1944
1945 /* copy_from_user should be SMP safe. */
1946 if (copy_from_user(a, args, nargs[call]))
1947 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001948
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001949 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001950 if (err)
1951 return err;
1952
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001953 a0 = a[0];
1954 a1 = a[1];
1955
1956 switch (call) {
1957 case SYS_SOCKET:
1958 err = sys_socket(a0, a1, a[2]);
1959 break;
1960 case SYS_BIND:
1961 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
1962 break;
1963 case SYS_CONNECT:
1964 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1965 break;
1966 case SYS_LISTEN:
1967 err = sys_listen(a0, a1);
1968 break;
1969 case SYS_ACCEPT:
1970 err =
1971 sys_accept(a0, (struct sockaddr __user *)a1,
1972 (int __user *)a[2]);
1973 break;
1974 case SYS_GETSOCKNAME:
1975 err =
1976 sys_getsockname(a0, (struct sockaddr __user *)a1,
1977 (int __user *)a[2]);
1978 break;
1979 case SYS_GETPEERNAME:
1980 err =
1981 sys_getpeername(a0, (struct sockaddr __user *)a1,
1982 (int __user *)a[2]);
1983 break;
1984 case SYS_SOCKETPAIR:
1985 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
1986 break;
1987 case SYS_SEND:
1988 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1989 break;
1990 case SYS_SENDTO:
1991 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
1992 (struct sockaddr __user *)a[4], a[5]);
1993 break;
1994 case SYS_RECV:
1995 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
1996 break;
1997 case SYS_RECVFROM:
1998 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
1999 (struct sockaddr __user *)a[4],
2000 (int __user *)a[5]);
2001 break;
2002 case SYS_SHUTDOWN:
2003 err = sys_shutdown(a0, a1);
2004 break;
2005 case SYS_SETSOCKOPT:
2006 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2007 break;
2008 case SYS_GETSOCKOPT:
2009 err =
2010 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2011 (int __user *)a[4]);
2012 break;
2013 case SYS_SENDMSG:
2014 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2015 break;
2016 case SYS_RECVMSG:
2017 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2018 break;
2019 default:
2020 err = -EINVAL;
2021 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 }
2023 return err;
2024}
2025
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002026#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002028/**
2029 * sock_register - add a socket protocol handler
2030 * @ops: description of protocol
2031 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 * This function is called by a protocol handler that wants to
2033 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002034 * socket interface. The value ops->family coresponds to the
2035 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002037int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038{
2039 int err;
2040
2041 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002042 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2043 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 return -ENOBUFS;
2045 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002046
2047 spin_lock(&net_family_lock);
2048 if (net_families[ops->family])
2049 err = -EEXIST;
2050 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002051 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 err = 0;
2053 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002054 spin_unlock(&net_family_lock);
2055
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002056 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 return err;
2058}
2059
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002060/**
2061 * sock_unregister - remove a protocol handler
2062 * @family: protocol family to remove
2063 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064 * This function is called by a protocol handler that wants to
2065 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002066 * new socket creation.
2067 *
2068 * If protocol handler is a module, then it can use module reference
2069 * counts to protect against new references. If protocol handler is not
2070 * a module then it needs to provide its own protection in
2071 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002073void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002075 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002077 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002078 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002079 spin_unlock(&net_family_lock);
2080
2081 synchronize_rcu();
2082
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002083 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084}
2085
Andi Kleen77d76ea2005-12-22 12:43:42 -08002086static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087{
2088 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002089 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002091
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 sk_init();
2093
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002095 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 */
2097 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098
2099 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002100 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 */
2102
2103 init_inodecache();
2104 register_filesystem(&sock_fs_type);
2105 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002106
2107 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 */
2109
2110#ifdef CONFIG_NETFILTER
2111 netfilter_init();
2112#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002113
2114 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115}
2116
Andi Kleen77d76ea2005-12-22 12:43:42 -08002117core_initcall(sock_init); /* early initcall */
2118
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119#ifdef CONFIG_PROC_FS
2120void socket_seq_show(struct seq_file *seq)
2121{
2122 int cpu;
2123 int counter = 0;
2124
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002125 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002126 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127
2128 /* It can be negative, by the way. 8) */
2129 if (counter < 0)
2130 counter = 0;
2131
2132 seq_printf(seq, "sockets: used %d\n", counter);
2133}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002134#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002136#ifdef CONFIG_COMPAT
2137static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002138 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002139{
2140 struct socket *sock = file->private_data;
2141 int ret = -ENOIOCTLCMD;
2142
2143 if (sock->ops->compat_ioctl)
2144 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2145
2146 return ret;
2147}
2148#endif
2149
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002150int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2151{
2152 return sock->ops->bind(sock, addr, addrlen);
2153}
2154
2155int kernel_listen(struct socket *sock, int backlog)
2156{
2157 return sock->ops->listen(sock, backlog);
2158}
2159
2160int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2161{
2162 struct sock *sk = sock->sk;
2163 int err;
2164
2165 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2166 newsock);
2167 if (err < 0)
2168 goto done;
2169
2170 err = sock->ops->accept(sock, *newsock, flags);
2171 if (err < 0) {
2172 sock_release(*newsock);
2173 goto done;
2174 }
2175
2176 (*newsock)->ops = sock->ops;
2177
2178done:
2179 return err;
2180}
2181
2182int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09002183 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002184{
2185 return sock->ops->connect(sock, addr, addrlen, flags);
2186}
2187
2188int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2189 int *addrlen)
2190{
2191 return sock->ops->getname(sock, addr, addrlen, 0);
2192}
2193
2194int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2195 int *addrlen)
2196{
2197 return sock->ops->getname(sock, addr, addrlen, 1);
2198}
2199
2200int kernel_getsockopt(struct socket *sock, int level, int optname,
2201 char *optval, int *optlen)
2202{
2203 mm_segment_t oldfs = get_fs();
2204 int err;
2205
2206 set_fs(KERNEL_DS);
2207 if (level == SOL_SOCKET)
2208 err = sock_getsockopt(sock, level, optname, optval, optlen);
2209 else
2210 err = sock->ops->getsockopt(sock, level, optname, optval,
2211 optlen);
2212 set_fs(oldfs);
2213 return err;
2214}
2215
2216int kernel_setsockopt(struct socket *sock, int level, int optname,
2217 char *optval, int optlen)
2218{
2219 mm_segment_t oldfs = get_fs();
2220 int err;
2221
2222 set_fs(KERNEL_DS);
2223 if (level == SOL_SOCKET)
2224 err = sock_setsockopt(sock, level, optname, optval, optlen);
2225 else
2226 err = sock->ops->setsockopt(sock, level, optname, optval,
2227 optlen);
2228 set_fs(oldfs);
2229 return err;
2230}
2231
2232int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2233 size_t size, int flags)
2234{
2235 if (sock->ops->sendpage)
2236 return sock->ops->sendpage(sock, page, offset, size, flags);
2237
2238 return sock_no_sendpage(sock, page, offset, size, flags);
2239}
2240
2241int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2242{
2243 mm_segment_t oldfs = get_fs();
2244 int err;
2245
2246 set_fs(KERNEL_DS);
2247 err = sock->ops->ioctl(sock, cmd, arg);
2248 set_fs(oldfs);
2249
2250 return err;
2251}
2252
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253/* ABI emulation layers need these two */
2254EXPORT_SYMBOL(move_addr_to_kernel);
2255EXPORT_SYMBOL(move_addr_to_user);
2256EXPORT_SYMBOL(sock_create);
2257EXPORT_SYMBOL(sock_create_kern);
2258EXPORT_SYMBOL(sock_create_lite);
2259EXPORT_SYMBOL(sock_map_fd);
2260EXPORT_SYMBOL(sock_recvmsg);
2261EXPORT_SYMBOL(sock_register);
2262EXPORT_SYMBOL(sock_release);
2263EXPORT_SYMBOL(sock_sendmsg);
2264EXPORT_SYMBOL(sock_unregister);
2265EXPORT_SYMBOL(sock_wake_async);
2266EXPORT_SYMBOL(sockfd_lookup);
2267EXPORT_SYMBOL(kernel_sendmsg);
2268EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002269EXPORT_SYMBOL(kernel_bind);
2270EXPORT_SYMBOL(kernel_listen);
2271EXPORT_SYMBOL(kernel_accept);
2272EXPORT_SYMBOL(kernel_connect);
2273EXPORT_SYMBOL(kernel_getsockname);
2274EXPORT_SYMBOL(kernel_getpeername);
2275EXPORT_SYMBOL(kernel_getsockopt);
2276EXPORT_SYMBOL(kernel_setsockopt);
2277EXPORT_SYMBOL(kernel_sendpage);
2278EXPORT_SYMBOL(kernel_sock_ioctl);