blob: 4147fe4bf41d12adeb0a08eeb1a2ee39fdc7203f [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>
80#include <linux/divert.h>
81#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010086#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010087#include <linux/wireless.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
89#include <asm/uaccess.h>
90#include <asm/unistd.h>
91
92#include <net/compat.h>
93
94#include <net/sock.h>
95#include <linux/netfilter.h>
96
97static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
98static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
Stephen Hemminger89bddce2006-09-01 00:19:31 -070099 size_t size, loff_t pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700101 size_t size, loff_t pos);
102static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
104static int sock_close(struct inode *inode, struct file *file);
105static unsigned int sock_poll(struct file *file,
106 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700107static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800108#ifdef CONFIG_COMPAT
109static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700110 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800111#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112static int sock_fasync(int fd, struct file *filp, int on);
113static ssize_t sock_readv(struct file *file, const struct iovec *vector,
114 unsigned long count, loff_t *ppos);
115static ssize_t sock_writev(struct file *file, const struct iovec *vector,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700116 unsigned long count, loff_t *ppos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117static ssize_t sock_sendpage(struct file *file, struct page *page,
118 int offset, size_t size, loff_t *ppos, int more);
119
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120/*
121 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
122 * in the operation structures but are done directly via the socketcall() multiplexor.
123 */
124
125static struct file_operations socket_file_ops = {
126 .owner = THIS_MODULE,
127 .llseek = no_llseek,
128 .aio_read = sock_aio_read,
129 .aio_write = sock_aio_write,
130 .poll = sock_poll,
131 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800132#ifdef CONFIG_COMPAT
133 .compat_ioctl = compat_sock_ioctl,
134#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 .mmap = sock_mmap,
136 .open = sock_no_open, /* special open code to disallow open via /proc */
137 .release = sock_close,
138 .fasync = sock_fasync,
139 .readv = sock_readv,
140 .writev = sock_writev,
Jens Axboe5274f052006-03-30 15:15:30 +0200141 .sendpage = sock_sendpage,
142 .splice_write = generic_splice_sendpage,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143};
144
145/*
146 * The protocol list. Each protocol is registered in here.
147 */
148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149static DEFINE_SPINLOCK(net_family_lock);
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -0700150static const struct net_proto_family *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152/*
153 * Statistics counters of the socket lists
154 */
155
156static DEFINE_PER_CPU(int, sockets_in_use) = 0;
157
158/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700159 * Support routines.
160 * Move socket addresses back and forth across the kernel/user
161 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 */
163
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700164#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 16 for IP, 16 for IPX,
166 24 for IPv6,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700167 about 80 for AX.25
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 must be at least one bigger than
169 the AF_UNIX size (see net/unix/af_unix.c
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700170 :unix_mkname()).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173/**
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
178 *
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 */
183
184int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
185{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700186 if (ulen < 0 || ulen > MAX_SOCK_ADDR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700188 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700190 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100192 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193}
194
195/**
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
201 *
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
206 * accessible.
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
210 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700211
212int move_addr_to_user(void *kaddr, int klen, void __user *uaddr,
213 int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
215 int err;
216 int len;
217
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700218 err = get_user(len, ulen);
219 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700221 if (len > klen)
222 len = klen;
223 if (len < 0 || len > MAX_SOCK_ADDR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500226 if (audit_sockaddr(klen, kaddr))
227 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return -EFAULT;
230 }
231 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700232 * "fromlen shall refer to the value before truncation.."
233 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 */
235 return __put_user(klen, ulen);
236}
237
238#define SOCKFS_MAGIC 0x534F434B
239
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700240static kmem_cache_t *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
242static struct inode *sock_alloc_inode(struct super_block *sb)
243{
244 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700245
246 ei = kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 if (!ei)
248 return NULL;
249 init_waitqueue_head(&ei->socket.wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700250
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 ei->socket.fasync_list = NULL;
252 ei->socket.state = SS_UNCONNECTED;
253 ei->socket.flags = 0;
254 ei->socket.ops = NULL;
255 ei->socket.sk = NULL;
256 ei->socket.file = NULL;
257 ei->socket.flags = 0;
258
259 return &ei->vfs_inode;
260}
261
262static void sock_destroy_inode(struct inode *inode)
263{
264 kmem_cache_free(sock_inode_cachep,
265 container_of(inode, struct socket_alloc, vfs_inode));
266}
267
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700268static void init_once(void *foo, kmem_cache_t *cachep, unsigned long flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700270 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700272 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR))
273 == SLAB_CTOR_CONSTRUCTOR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 inode_init_once(&ei->vfs_inode);
275}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277static int init_inodecache(void)
278{
279 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700280 sizeof(struct socket_alloc),
281 0,
282 (SLAB_HWCACHE_ALIGN |
283 SLAB_RECLAIM_ACCOUNT |
284 SLAB_MEM_SPREAD),
285 init_once,
286 NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 if (sock_inode_cachep == NULL)
288 return -ENOMEM;
289 return 0;
290}
291
292static struct super_operations sockfs_ops = {
293 .alloc_inode = sock_alloc_inode,
294 .destroy_inode =sock_destroy_inode,
295 .statfs = simple_statfs,
296};
297
David Howells454e2392006-06-23 02:02:57 -0700298static int sockfs_get_sb(struct file_system_type *fs_type,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700299 int flags, const char *dev_name, void *data,
300 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301{
David Howells454e2392006-06-23 02:02:57 -0700302 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
303 mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304}
305
Eric Dumazetba899662005-08-26 12:05:31 -0700306static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307
308static struct file_system_type sock_fs_type = {
309 .name = "sockfs",
310 .get_sb = sockfs_get_sb,
311 .kill_sb = kill_anon_super,
312};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700313
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314static int sockfs_delete_dentry(struct dentry *dentry)
315{
316 return 1;
317}
318static struct dentry_operations sockfs_dentry_operations = {
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700319 .d_delete = sockfs_delete_dentry,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320};
321
322/*
323 * Obtains the first available file descriptor and sets it up for use.
324 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800325 * These functions create file structures and maps them to fd space
326 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 * and file struct implicitly stored in sock->file.
328 * Note that another thread may close file descriptor before we return
329 * from this function. We use the fact that now we do not refer
330 * to socket after mapping. If one day we will need it, this
331 * function will increment ref. count on file by 1.
332 *
333 * In any case returned fd MAY BE not valid!
334 * This race condition is unavoidable
335 * with shared fd spaces, we cannot solve it inside kernel,
336 * but we take care of internal coherence yet.
337 */
338
David S. Miller39d8c1b2006-03-20 17:13:49 -0800339static int sock_alloc_fd(struct file **filep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340{
341 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800342
343 fd = get_unused_fd();
344 if (likely(fd >= 0)) {
345 struct file *file = get_empty_filp();
346
347 *filep = file;
348 if (unlikely(!file)) {
349 put_unused_fd(fd);
350 return -ENFILE;
351 }
352 } else
353 *filep = NULL;
354 return fd;
355}
356
357static int sock_attach_fd(struct socket *sock, struct file *file)
358{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 struct qstr this;
360 char name[32];
361
David S. Miller39d8c1b2006-03-20 17:13:49 -0800362 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
363 this.name = name;
364 this.hash = SOCK_INODE(sock)->i_ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
David S. Miller39d8c1b2006-03-20 17:13:49 -0800366 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
367 if (unlikely(!file->f_dentry))
368 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369
David S. Miller39d8c1b2006-03-20 17:13:49 -0800370 file->f_dentry->d_op = &sockfs_dentry_operations;
371 d_add(file->f_dentry, SOCK_INODE(sock));
372 file->f_vfsmnt = mntget(sock_mnt);
373 file->f_mapping = file->f_dentry->d_inode->i_mapping;
374
375 sock->file = file;
376 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
377 file->f_mode = FMODE_READ | FMODE_WRITE;
378 file->f_flags = O_RDWR;
379 file->f_pos = 0;
380 file->private_data = sock;
381
382 return 0;
383}
384
385int sock_map_fd(struct socket *sock)
386{
387 struct file *newfile;
388 int fd = sock_alloc_fd(&newfile);
389
390 if (likely(fd >= 0)) {
391 int err = sock_attach_fd(sock, newfile);
392
393 if (unlikely(err < 0)) {
394 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800396 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800398 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 return fd;
401}
402
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800403static struct socket *sock_from_file(struct file *file, int *err)
404{
405 struct inode *inode;
406 struct socket *sock;
407
408 if (file->f_op == &socket_file_ops)
409 return file->private_data; /* set in sock_map_fd */
410
411 inode = file->f_dentry->d_inode;
412 if (!S_ISSOCK(inode->i_mode)) {
413 *err = -ENOTSOCK;
414 return NULL;
415 }
416
417 sock = SOCKET_I(inode);
418 if (sock->file != file) {
419 printk(KERN_ERR "socki_lookup: socket file changed!\n");
420 sock->file = file;
421 }
422 return sock;
423}
424
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425/**
426 * sockfd_lookup - Go from a file number to its socket slot
427 * @fd: file handle
428 * @err: pointer to an error code return
429 *
430 * The file handle passed in is locked and the socket it is bound
431 * too is returned. If an error occurs the err pointer is overwritten
432 * with a negative errno code and NULL is returned. The function checks
433 * for both invalid handles and passing a handle which is not a socket.
434 *
435 * On a success the socket object pointer is returned.
436 */
437
438struct socket *sockfd_lookup(int fd, int *err)
439{
440 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 struct socket *sock;
442
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700443 file = fget(fd);
444 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 *err = -EBADF;
446 return NULL;
447 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700448
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800449 sock = sock_from_file(file, err);
450 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 return sock;
453}
454
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800455static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
456{
457 struct file *file;
458 struct socket *sock;
459
Hua Zhong36725582006-04-19 15:25:02 -0700460 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800461 file = fget_light(fd, fput_needed);
462 if (file) {
463 sock = sock_from_file(file, err);
464 if (sock)
465 return sock;
466 fput_light(file, *fput_needed);
467 }
468 return NULL;
469}
470
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471/**
472 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700473 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 * Allocate a new inode and socket object. The two are bound together
475 * and initialised. The socket is then returned. If we are out of inodes
476 * NULL is returned.
477 */
478
479static struct socket *sock_alloc(void)
480{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700481 struct inode *inode;
482 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
484 inode = new_inode(sock_mnt->mnt_sb);
485 if (!inode)
486 return NULL;
487
488 sock = SOCKET_I(inode);
489
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700490 inode->i_mode = S_IFSOCK | S_IRWXUGO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 inode->i_uid = current->fsuid;
492 inode->i_gid = current->fsgid;
493
494 get_cpu_var(sockets_in_use)++;
495 put_cpu_var(sockets_in_use);
496 return sock;
497}
498
499/*
500 * In theory you can't get an open on this inode, but /proc provides
501 * a back door. Remember to keep it shut otherwise you'll let the
502 * creepy crawlies in.
503 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
506{
507 return -ENXIO;
508}
509
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800510const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 .owner = THIS_MODULE,
512 .open = sock_no_open,
513};
514
515/**
516 * sock_release - close a socket
517 * @sock: socket to close
518 *
519 * The socket is released from the protocol stack if it has a release
520 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700521 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700523
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524void sock_release(struct socket *sock)
525{
526 if (sock->ops) {
527 struct module *owner = sock->ops->owner;
528
529 sock->ops->release(sock);
530 sock->ops = NULL;
531 module_put(owner);
532 }
533
534 if (sock->fasync_list)
535 printk(KERN_ERR "sock_release: fasync list not empty!\n");
536
537 get_cpu_var(sockets_in_use)--;
538 put_cpu_var(sockets_in_use);
539 if (!sock->file) {
540 iput(SOCK_INODE(sock));
541 return;
542 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700543 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544}
545
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700546static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 struct msghdr *msg, size_t size)
548{
549 struct sock_iocb *si = kiocb_to_siocb(iocb);
550 int err;
551
552 si->sock = sock;
553 si->scm = NULL;
554 si->msg = msg;
555 si->size = size;
556
557 err = security_socket_sendmsg(sock, msg, size);
558 if (err)
559 return err;
560
561 return sock->ops->sendmsg(iocb, sock, msg, size);
562}
563
564int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
565{
566 struct kiocb iocb;
567 struct sock_iocb siocb;
568 int ret;
569
570 init_sync_kiocb(&iocb, NULL);
571 iocb.private = &siocb;
572 ret = __sock_sendmsg(&iocb, sock, msg, size);
573 if (-EIOCBQUEUED == ret)
574 ret = wait_on_sync_kiocb(&iocb);
575 return ret;
576}
577
578int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
579 struct kvec *vec, size_t num, size_t size)
580{
581 mm_segment_t oldfs = get_fs();
582 int result;
583
584 set_fs(KERNEL_DS);
585 /*
586 * the following is safe, since for compiler definitions of kvec and
587 * iovec are identical, yielding the same in-core layout and alignment
588 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700589 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 msg->msg_iovlen = num;
591 result = sock_sendmsg(sock, msg, size);
592 set_fs(oldfs);
593 return result;
594}
595
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700596static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 struct msghdr *msg, size_t size, int flags)
598{
599 int err;
600 struct sock_iocb *si = kiocb_to_siocb(iocb);
601
602 si->sock = sock;
603 si->scm = NULL;
604 si->msg = msg;
605 si->size = size;
606 si->flags = flags;
607
608 err = security_socket_recvmsg(sock, msg, size, flags);
609 if (err)
610 return err;
611
612 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
613}
614
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700615int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 size_t size, int flags)
617{
618 struct kiocb iocb;
619 struct sock_iocb siocb;
620 int ret;
621
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700622 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 iocb.private = &siocb;
624 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
625 if (-EIOCBQUEUED == ret)
626 ret = wait_on_sync_kiocb(&iocb);
627 return ret;
628}
629
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700630int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
631 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632{
633 mm_segment_t oldfs = get_fs();
634 int result;
635
636 set_fs(KERNEL_DS);
637 /*
638 * the following is safe, since for compiler definitions of kvec and
639 * iovec are identical, yielding the same in-core layout and alignment
640 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700641 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 result = sock_recvmsg(sock, msg, size, flags);
643 set_fs(oldfs);
644 return result;
645}
646
647static void sock_aio_dtor(struct kiocb *iocb)
648{
649 kfree(iocb->private);
650}
651
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300652static ssize_t sock_sendpage(struct file *file, struct page *page,
653 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654{
655 struct socket *sock;
656 int flags;
657
Eric Dumazetb69aee02005-09-06 14:42:45 -0700658 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
660 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
661 if (more)
662 flags |= MSG_MORE;
663
664 return sock->ops->sendpage(sock, page, offset, size, flags);
665}
666
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800667static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700668 char __user *ubuf, size_t size,
669 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800670{
671 if (!is_sync_kiocb(iocb)) {
672 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
673 if (!siocb)
674 return NULL;
675 iocb->ki_dtor = sock_aio_dtor;
676 }
677
678 siocb->kiocb = iocb;
679 siocb->async_iov.iov_base = ubuf;
680 siocb->async_iov.iov_len = size;
681
682 iocb->private = siocb;
683 return siocb;
684}
685
686static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700687 struct file *file, struct iovec *iov,
688 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800689{
690 struct socket *sock = file->private_data;
691 size_t size = 0;
692 int i;
693
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700694 for (i = 0; i < nr_segs; i++)
695 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800696
697 msg->msg_name = NULL;
698 msg->msg_namelen = 0;
699 msg->msg_control = NULL;
700 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700701 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800702 msg->msg_iovlen = nr_segs;
703 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
704
705 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
706}
707
708static ssize_t sock_readv(struct file *file, const struct iovec *iov,
709 unsigned long nr_segs, loff_t *ppos)
710{
711 struct kiocb iocb;
712 struct sock_iocb siocb;
713 struct msghdr msg;
714 int ret;
715
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700716 init_sync_kiocb(&iocb, NULL);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800717 iocb.private = &siocb;
718
719 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
720 if (-EIOCBQUEUED == ret)
721 ret = wait_on_sync_kiocb(&iocb);
722 return ret;
723}
724
725static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700726 size_t count, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800727{
728 struct sock_iocb siocb, *x;
729
730 if (pos != 0)
731 return -ESPIPE;
732 if (count == 0) /* Match SYS5 behaviour */
733 return 0;
734
735 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
736 if (!x)
737 return -ENOMEM;
738 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700739 &x->async_iov, 1);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800740}
741
742static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700743 struct file *file, struct iovec *iov,
744 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800745{
746 struct socket *sock = file->private_data;
747 size_t size = 0;
748 int i;
749
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700750 for (i = 0; i < nr_segs; i++)
751 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800752
753 msg->msg_name = NULL;
754 msg->msg_namelen = 0;
755 msg->msg_control = NULL;
756 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700757 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800758 msg->msg_iovlen = nr_segs;
759 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
760 if (sock->type == SOCK_SEQPACKET)
761 msg->msg_flags |= MSG_EOR;
762
763 return __sock_sendmsg(iocb, sock, msg, size);
764}
765
766static ssize_t sock_writev(struct file *file, const struct iovec *iov,
767 unsigned long nr_segs, loff_t *ppos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768{
769 struct msghdr msg;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800770 struct kiocb iocb;
771 struct sock_iocb siocb;
772 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800774 init_sync_kiocb(&iocb, NULL);
775 iocb.private = &siocb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800777 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
778 if (-EIOCBQUEUED == ret)
779 ret = wait_on_sync_kiocb(&iocb);
780 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781}
782
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800783static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700784 size_t count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800786 struct sock_iocb siocb, *x;
787
788 if (pos != 0)
789 return -ESPIPE;
790 if (count == 0) /* Match SYS5 behaviour */
791 return 0;
792
793 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
794 if (!x)
795 return -ENOMEM;
796
797 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700798 &x->async_iov, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799}
800
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801/*
802 * Atomic setting of ioctl hooks to avoid race
803 * with module unload.
804 */
805
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800806static DEFINE_MUTEX(br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700807static int (*br_ioctl_hook) (unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700809void brioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800811 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800813 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700815
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816EXPORT_SYMBOL(brioctl_set);
817
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800818static DEFINE_MUTEX(vlan_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700819static int (*vlan_ioctl_hook) (void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700821void vlan_ioctl_set(int (*hook) (void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800823 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800825 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700827
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828EXPORT_SYMBOL(vlan_ioctl_set);
829
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800830static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700831static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700833void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800835 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800837 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700839
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840EXPORT_SYMBOL(dlci_ioctl_set);
841
842/*
843 * With an ioctl, arg may well be a user mode pointer, but we don't know
844 * what to do with it - that's up to the protocol still.
845 */
846
847static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
848{
849 struct socket *sock;
850 void __user *argp = (void __user *)arg;
851 int pid, err;
852
Eric Dumazetb69aee02005-09-06 14:42:45 -0700853 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
855 err = dev_ioctl(cmd, argp);
856 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100857#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
859 err = dev_ioctl(cmd, argp);
860 } else
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700861#endif /* CONFIG_WIRELESS_EXT */
862 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 case FIOSETOWN:
864 case SIOCSPGRP:
865 err = -EFAULT;
866 if (get_user(pid, (int __user *)argp))
867 break;
868 err = f_setown(sock->file, pid, 1);
869 break;
870 case FIOGETOWN:
871 case SIOCGPGRP:
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700872 err = put_user(sock->file->f_owner.pid,
873 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 break;
875 case SIOCGIFBR:
876 case SIOCSIFBR:
877 case SIOCBRADDBR:
878 case SIOCBRDELBR:
879 err = -ENOPKG;
880 if (!br_ioctl_hook)
881 request_module("bridge");
882
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800883 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700884 if (br_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 err = br_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800886 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 break;
888 case SIOCGIFVLAN:
889 case SIOCSIFVLAN:
890 err = -ENOPKG;
891 if (!vlan_ioctl_hook)
892 request_module("8021q");
893
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800894 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 if (vlan_ioctl_hook)
896 err = vlan_ioctl_hook(argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800897 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 break;
899 case SIOCGIFDIVERT:
900 case SIOCSIFDIVERT:
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700901 /* Convert this to call through a hook */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 err = divert_ioctl(cmd, argp);
903 break;
904 case SIOCADDDLCI:
905 case SIOCDELDLCI:
906 err = -ENOPKG;
907 if (!dlci_ioctl_hook)
908 request_module("dlci");
909
910 if (dlci_ioctl_hook) {
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800911 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 err = dlci_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800913 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 }
915 break;
916 default:
917 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800918
919 /*
920 * If this ioctl is unknown try to hand it down
921 * to the NIC driver.
922 */
923 if (err == -ENOIOCTLCMD)
924 err = dev_ioctl(cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700926 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 return err;
928}
929
930int sock_create_lite(int family, int type, int protocol, struct socket **res)
931{
932 int err;
933 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700934
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 err = security_socket_create(family, type, protocol, 1);
936 if (err)
937 goto out;
938
939 sock = sock_alloc();
940 if (!sock) {
941 err = -ENOMEM;
942 goto out;
943 }
944
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700946 err = security_socket_post_create(sock, family, type, protocol, 1);
947 if (err)
948 goto out_release;
949
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950out:
951 *res = sock;
952 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700953out_release:
954 sock_release(sock);
955 sock = NULL;
956 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957}
958
959/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700960static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961{
962 struct socket *sock;
963
964 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700965 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700967 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 return sock->ops->poll(file, sock, wait);
969}
970
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700971static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700973 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974
975 return sock->ops->mmap(file, sock, vma);
976}
977
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300978static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979{
980 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700981 * It was possible the inode is NULL we were
982 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 */
984
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700985 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 printk(KERN_DEBUG "sock_close: NULL inode\n");
987 return 0;
988 }
989 sock_fasync(-1, filp, 0);
990 sock_release(SOCKET_I(inode));
991 return 0;
992}
993
994/*
995 * Update the socket async list
996 *
997 * Fasync_list locking strategy.
998 *
999 * 1. fasync_list is modified only under process context socket lock
1000 * i.e. under semaphore.
1001 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1002 * or under socket lock.
1003 * 3. fasync_list can be used from softirq context, so that
1004 * modification under socket lock have to be enhanced with
1005 * write_lock_bh(&sk->sk_callback_lock).
1006 * --ANK (990710)
1007 */
1008
1009static int sock_fasync(int fd, struct file *filp, int on)
1010{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001011 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 struct socket *sock;
1013 struct sock *sk;
1014
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001015 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001016 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001017 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 return -ENOMEM;
1019 }
1020
Eric Dumazetb69aee02005-09-06 14:42:45 -07001021 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001023 sk = sock->sk;
1024 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 kfree(fna);
1026 return -EINVAL;
1027 }
1028
1029 lock_sock(sk);
1030
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001031 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001033 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1034 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 break;
1036
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001037 if (on) {
1038 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001040 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 write_unlock_bh(&sk->sk_callback_lock);
1042
1043 kfree(fna);
1044 goto out;
1045 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001046 fna->fa_file = filp;
1047 fna->fa_fd = fd;
1048 fna->magic = FASYNC_MAGIC;
1049 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001051 sock->fasync_list = fna;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001053 } else {
1054 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001056 *prev = fa->fa_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 write_unlock_bh(&sk->sk_callback_lock);
1058 kfree(fa);
1059 }
1060 }
1061
1062out:
1063 release_sock(sock->sk);
1064 return 0;
1065}
1066
1067/* This function may be called only under socket lock or callback_lock */
1068
1069int sock_wake_async(struct socket *sock, int how, int band)
1070{
1071 if (!sock || !sock->fasync_list)
1072 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001073 switch (how) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 case 1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001075
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1077 break;
1078 goto call_kill;
1079 case 2:
1080 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1081 break;
1082 /* fall through */
1083 case 0:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001084call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 __kill_fasync(sock->fasync_list, SIGIO, band);
1086 break;
1087 case 3:
1088 __kill_fasync(sock->fasync_list, SIGURG, band);
1089 }
1090 return 0;
1091}
1092
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001093static int __sock_create(int family, int type, int protocol,
1094 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095{
1096 int err;
1097 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001098 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099
1100 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001101 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 */
1103 if (family < 0 || family >= NPROTO)
1104 return -EAFNOSUPPORT;
1105 if (type < 0 || type >= SOCK_MAX)
1106 return -EINVAL;
1107
1108 /* Compatibility.
1109
1110 This uglymoron is moved from INET layer to here to avoid
1111 deadlock in module load.
1112 */
1113 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001114 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 if (!warned) {
1116 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001117 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1118 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 }
1120 family = PF_PACKET;
1121 }
1122
1123 err = security_socket_create(family, type, protocol, kern);
1124 if (err)
1125 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001126
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001127 /*
1128 * Allocate the socket and allow the family to set things up. if
1129 * the protocol is 0, the family is instructed to select an appropriate
1130 * default.
1131 */
1132 sock = sock_alloc();
1133 if (!sock) {
1134 if (net_ratelimit())
1135 printk(KERN_WARNING "socket: no more sockets\n");
1136 return -ENFILE; /* Not exactly a match, but its the
1137 closest posix thing */
1138 }
1139
1140 sock->type = type;
1141
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142#if defined(CONFIG_KMOD)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001143 /* Attempt to load a protocol module if the find failed.
1144 *
1145 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 * requested real, full-featured networking support upon configuration.
1147 * Otherwise module support will break!
1148 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001149 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001150 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151#endif
1152
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001153 rcu_read_lock();
1154 pf = rcu_dereference(net_families[family]);
1155 err = -EAFNOSUPPORT;
1156 if (!pf)
1157 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158
1159 /*
1160 * We will call the ->create function, that possibly is in a loadable
1161 * module, so we have to bump that loadable module refcnt first.
1162 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001163 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 goto out_release;
1165
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001166 /* Now protected by module ref count */
1167 rcu_read_unlock();
1168
1169 err = pf->create(sock, protocol);
1170 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001172
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 /*
1174 * Now to bump the refcnt of the [loadable] module that owns this
1175 * socket at sock_release time we decrement its refcnt.
1176 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001177 if (!try_module_get(sock->ops->owner))
1178 goto out_module_busy;
1179
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 /*
1181 * Now that we're done with the ->create function, the [loadable]
1182 * module can have its refcnt decremented
1183 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001184 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001185 err = security_socket_post_create(sock, family, type, protocol, kern);
1186 if (err)
1187 goto out_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001188 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001190 return 0;
1191
1192out_module_busy:
1193 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001195 sock->ops = NULL;
1196 module_put(pf->owner);
1197out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001199 return err;
1200
1201out_release:
1202 rcu_read_unlock();
1203 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204}
1205
1206int sock_create(int family, int type, int protocol, struct socket **res)
1207{
1208 return __sock_create(family, type, protocol, res, 0);
1209}
1210
1211int sock_create_kern(int family, int type, int protocol, struct socket **res)
1212{
1213 return __sock_create(family, type, protocol, res, 1);
1214}
1215
1216asmlinkage long sys_socket(int family, int type, int protocol)
1217{
1218 int retval;
1219 struct socket *sock;
1220
1221 retval = sock_create(family, type, protocol, &sock);
1222 if (retval < 0)
1223 goto out;
1224
1225 retval = sock_map_fd(sock);
1226 if (retval < 0)
1227 goto out_release;
1228
1229out:
1230 /* It may be already another descriptor 8) Not kernel problem. */
1231 return retval;
1232
1233out_release:
1234 sock_release(sock);
1235 return retval;
1236}
1237
1238/*
1239 * Create a pair of connected sockets.
1240 */
1241
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001242asmlinkage long sys_socketpair(int family, int type, int protocol,
1243 int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244{
1245 struct socket *sock1, *sock2;
1246 int fd1, fd2, err;
1247
1248 /*
1249 * Obtain the first socket and check if the underlying protocol
1250 * supports the socketpair call.
1251 */
1252
1253 err = sock_create(family, type, protocol, &sock1);
1254 if (err < 0)
1255 goto out;
1256
1257 err = sock_create(family, type, protocol, &sock2);
1258 if (err < 0)
1259 goto out_release_1;
1260
1261 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001262 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 goto out_release_both;
1264
1265 fd1 = fd2 = -1;
1266
1267 err = sock_map_fd(sock1);
1268 if (err < 0)
1269 goto out_release_both;
1270 fd1 = err;
1271
1272 err = sock_map_fd(sock2);
1273 if (err < 0)
1274 goto out_close_1;
1275 fd2 = err;
1276
1277 /* fd1 and fd2 may be already another descriptors.
1278 * Not kernel problem.
1279 */
1280
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001281 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 if (!err)
1283 err = put_user(fd2, &usockvec[1]);
1284 if (!err)
1285 return 0;
1286
1287 sys_close(fd2);
1288 sys_close(fd1);
1289 return err;
1290
1291out_close_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001292 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 sys_close(fd1);
1294 return err;
1295
1296out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001297 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001299 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300out:
1301 return err;
1302}
1303
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304/*
1305 * Bind a name to a socket. Nothing much to do here since it's
1306 * the protocol's responsibility to handle the local address.
1307 *
1308 * We move the socket address to kernel space before we call
1309 * the protocol layer (having also checked the address is ok).
1310 */
1311
1312asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1313{
1314 struct socket *sock;
1315 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001316 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001318 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1319 if(sock) {
1320 err = move_addr_to_kernel(umyaddr, addrlen, address);
1321 if (err >= 0) {
1322 err = security_socket_bind(sock,
1323 (struct sockaddr *)address,
1324 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001325 if (!err)
1326 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001327 (struct sockaddr *)
1328 address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001330 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001331 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 return err;
1333}
1334
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335/*
1336 * Perform a listen. Basically, we allow the protocol to do anything
1337 * necessary for a listen, and if that works, we mark the socket as
1338 * ready for listening.
1339 */
1340
1341int sysctl_somaxconn = SOMAXCONN;
1342
1343asmlinkage long sys_listen(int fd, int backlog)
1344{
1345 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001346 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001347
1348 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1349 if (sock) {
1350 if ((unsigned)backlog > sysctl_somaxconn)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 backlog = sysctl_somaxconn;
1352
1353 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001354 if (!err)
1355 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001357 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 }
1359 return err;
1360}
1361
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362/*
1363 * For accept, we attempt to create a new socket, set up the link
1364 * with the client, wake up the client, then return the new
1365 * connected fd. We collect the address of the connector in kernel
1366 * space and move it to user at the very end. This is unclean because
1367 * we open the socket then return an error.
1368 *
1369 * 1003.1g adds the ability to recvmsg() to query connection pending
1370 * status to recvmsg. We need to add that support in a way thats
1371 * clean when we restucture accept also.
1372 */
1373
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001374asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1375 int __user *upeer_addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376{
1377 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001378 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001379 int err, len, newfd, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 char address[MAX_SOCK_ADDR];
1381
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001382 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 if (!sock)
1384 goto out;
1385
1386 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001387 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 goto out_put;
1389
1390 newsock->type = sock->type;
1391 newsock->ops = sock->ops;
1392
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 /*
1394 * We don't need try_module_get here, as the listening socket (sock)
1395 * has the protocol module (sock->ops->owner) held.
1396 */
1397 __module_get(newsock->ops->owner);
1398
David S. Miller39d8c1b2006-03-20 17:13:49 -08001399 newfd = sock_alloc_fd(&newfile);
1400 if (unlikely(newfd < 0)) {
1401 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001402 sock_release(newsock);
1403 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001404 }
1405
1406 err = sock_attach_fd(newsock, newfile);
1407 if (err < 0)
1408 goto out_fd;
1409
Frank Filza79af592005-09-27 15:23:38 -07001410 err = security_socket_accept(sock, newsock);
1411 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001412 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001413
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1415 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001416 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417
1418 if (upeer_sockaddr) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001419 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1420 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001422 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001424 err = move_addr_to_user(address, len, upeer_sockaddr,
1425 upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001427 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 }
1429
1430 /* File flags are not inherited via accept() unlike another OSes. */
1431
David S. Miller39d8c1b2006-03-20 17:13:49 -08001432 fd_install(newfd, newfile);
1433 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434
1435 security_socket_post_accept(sock, newsock);
1436
1437out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001438 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439out:
1440 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001441out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001442 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001443 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 goto out_put;
1445}
1446
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447/*
1448 * Attempt to connect to a socket with the server address. The address
1449 * is in user space so we verify it is OK and move it to kernel space.
1450 *
1451 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1452 * break bindings
1453 *
1454 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1455 * other SEQPACKET protocols that take time to connect() as it doesn't
1456 * include the -EINPROGRESS status for such sockets.
1457 */
1458
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001459asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1460 int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461{
1462 struct socket *sock;
1463 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001464 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001466 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 if (!sock)
1468 goto out;
1469 err = move_addr_to_kernel(uservaddr, addrlen, address);
1470 if (err < 0)
1471 goto out_put;
1472
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001473 err =
1474 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 if (err)
1476 goto out_put;
1477
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001478 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 sock->file->f_flags);
1480out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001481 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482out:
1483 return err;
1484}
1485
1486/*
1487 * Get the local address ('name') of a socket object. Move the obtained
1488 * name to user space.
1489 */
1490
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001491asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1492 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493{
1494 struct socket *sock;
1495 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001496 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001497
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001498 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 if (!sock)
1500 goto out;
1501
1502 err = security_socket_getsockname(sock);
1503 if (err)
1504 goto out_put;
1505
1506 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1507 if (err)
1508 goto out_put;
1509 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1510
1511out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001512 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513out:
1514 return err;
1515}
1516
1517/*
1518 * Get the remote address ('name') of a socket object. Move the obtained
1519 * name to user space.
1520 */
1521
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001522asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1523 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524{
1525 struct socket *sock;
1526 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001527 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001529 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1530 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 err = security_socket_getpeername(sock);
1532 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001533 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 return err;
1535 }
1536
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001537 err =
1538 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1539 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 if (!err)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001541 err = move_addr_to_user(address, len, usockaddr,
1542 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001543 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 }
1545 return err;
1546}
1547
1548/*
1549 * Send a datagram to a given address. We move the address into kernel
1550 * space and check the user space data area is readable before invoking
1551 * the protocol.
1552 */
1553
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001554asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1555 unsigned flags, struct sockaddr __user *addr,
1556 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557{
1558 struct socket *sock;
1559 char address[MAX_SOCK_ADDR];
1560 int err;
1561 struct msghdr msg;
1562 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001563 int fput_needed;
1564 struct file *sock_file;
1565
1566 sock_file = fget_light(fd, &fput_needed);
1567 if (!sock_file)
1568 return -EBADF;
1569
1570 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 if (!sock)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001572 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001573 iov.iov_base = buff;
1574 iov.iov_len = len;
1575 msg.msg_name = NULL;
1576 msg.msg_iov = &iov;
1577 msg.msg_iovlen = 1;
1578 msg.msg_control = NULL;
1579 msg.msg_controllen = 0;
1580 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001581 if (addr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 err = move_addr_to_kernel(addr, addr_len, address);
1583 if (err < 0)
1584 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001585 msg.msg_name = address;
1586 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 }
1588 if (sock->file->f_flags & O_NONBLOCK)
1589 flags |= MSG_DONTWAIT;
1590 msg.msg_flags = flags;
1591 err = sock_sendmsg(sock, &msg, len);
1592
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001593out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001594 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 return err;
1596}
1597
1598/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001599 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 */
1601
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001602asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603{
1604 return sys_sendto(fd, buff, len, flags, NULL, 0);
1605}
1606
1607/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001608 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 * sender. We verify the buffers are writable and if needed move the
1610 * sender address from kernel to user space.
1611 */
1612
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001613asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1614 unsigned flags, struct sockaddr __user *addr,
1615 int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616{
1617 struct socket *sock;
1618 struct iovec iov;
1619 struct msghdr msg;
1620 char address[MAX_SOCK_ADDR];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001621 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001622 struct file *sock_file;
1623 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001625 sock_file = fget_light(fd, &fput_needed);
1626 if (!sock_file)
1627 return -EBADF;
1628
1629 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 if (!sock)
1631 goto out;
1632
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001633 msg.msg_control = NULL;
1634 msg.msg_controllen = 0;
1635 msg.msg_iovlen = 1;
1636 msg.msg_iov = &iov;
1637 iov.iov_len = size;
1638 iov.iov_base = ubuf;
1639 msg.msg_name = address;
1640 msg.msg_namelen = MAX_SOCK_ADDR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 if (sock->file->f_flags & O_NONBLOCK)
1642 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001643 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001645 if (err >= 0 && addr != NULL) {
1646 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1647 if (err2 < 0)
1648 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650out:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001651 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 return err;
1653}
1654
1655/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001656 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657 */
1658
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001659asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1660 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661{
1662 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1663}
1664
1665/*
1666 * Set a socket option. Because we don't know the option lengths we have
1667 * to pass the user mode parameter for the protocols to sort out.
1668 */
1669
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001670asmlinkage long sys_setsockopt(int fd, int level, int optname,
1671 char __user *optval, int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001673 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 struct socket *sock;
1675
1676 if (optlen < 0)
1677 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001678
1679 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1680 if (sock != NULL) {
1681 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001682 if (err)
1683 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684
1685 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001686 err =
1687 sock_setsockopt(sock, level, optname, optval,
1688 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001690 err =
1691 sock->ops->setsockopt(sock, level, optname, optval,
1692 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001693out_put:
1694 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 }
1696 return err;
1697}
1698
1699/*
1700 * Get a socket option. Because we don't know the option lengths we have
1701 * to pass a user mode parameter for the protocols to sort out.
1702 */
1703
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001704asmlinkage long sys_getsockopt(int fd, int level, int optname,
1705 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001707 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 struct socket *sock;
1709
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001710 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1711 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001712 err = security_socket_getsockopt(sock, level, optname);
1713 if (err)
1714 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715
1716 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001717 err =
1718 sock_getsockopt(sock, level, optname, optval,
1719 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001721 err =
1722 sock->ops->getsockopt(sock, level, optname, optval,
1723 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001724out_put:
1725 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 }
1727 return err;
1728}
1729
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730/*
1731 * Shutdown a socket.
1732 */
1733
1734asmlinkage long sys_shutdown(int fd, int how)
1735{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001736 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 struct socket *sock;
1738
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001739 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1740 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001742 if (!err)
1743 err = sock->ops->shutdown(sock, how);
1744 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 }
1746 return err;
1747}
1748
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001749/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 * fields which are the same type (int / unsigned) on our platforms.
1751 */
1752#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1753#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1754#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1755
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756/*
1757 * BSD sendmsg interface
1758 */
1759
1760asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1761{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001762 struct compat_msghdr __user *msg_compat =
1763 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 struct socket *sock;
1765 char address[MAX_SOCK_ADDR];
1766 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001767 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001768 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1769 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 unsigned char *ctl_buf = ctl;
1771 struct msghdr msg_sys;
1772 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001773 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001774
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 err = -EFAULT;
1776 if (MSG_CMSG_COMPAT & flags) {
1777 if (get_compat_msghdr(&msg_sys, msg_compat))
1778 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001779 }
1780 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 return -EFAULT;
1782
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001783 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001784 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 goto out;
1786
1787 /* do not move before msg_sys is valid */
1788 err = -EMSGSIZE;
1789 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1790 goto out_put;
1791
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001792 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 err = -ENOMEM;
1794 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1795 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1796 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1797 if (!iov)
1798 goto out_put;
1799 }
1800
1801 /* This will also move the address data into kernel space */
1802 if (MSG_CMSG_COMPAT & flags) {
1803 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1804 } else
1805 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001806 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 goto out_freeiov;
1808 total_len = err;
1809
1810 err = -ENOBUFS;
1811
1812 if (msg_sys.msg_controllen > INT_MAX)
1813 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001814 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001816 err =
1817 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1818 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819 if (err)
1820 goto out_freeiov;
1821 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001822 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001824 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001826 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 goto out_freeiov;
1828 }
1829 err = -EFAULT;
1830 /*
1831 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1832 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1833 * checking falls down on this.
1834 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001835 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1836 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 goto out_freectl;
1838 msg_sys.msg_control = ctl_buf;
1839 }
1840 msg_sys.msg_flags = flags;
1841
1842 if (sock->file->f_flags & O_NONBLOCK)
1843 msg_sys.msg_flags |= MSG_DONTWAIT;
1844 err = sock_sendmsg(sock, &msg_sys, total_len);
1845
1846out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001847 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1849out_freeiov:
1850 if (iov != iovstack)
1851 sock_kfree_s(sock->sk, iov, iov_size);
1852out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001853 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001854out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 return err;
1856}
1857
1858/*
1859 * BSD recvmsg interface
1860 */
1861
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001862asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1863 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001865 struct compat_msghdr __user *msg_compat =
1866 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867 struct socket *sock;
1868 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001869 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870 struct msghdr msg_sys;
1871 unsigned long cmsg_ptr;
1872 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001873 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874
1875 /* kernel mode address */
1876 char addr[MAX_SOCK_ADDR];
1877
1878 /* user mode address pointers */
1879 struct sockaddr __user *uaddr;
1880 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001881
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 if (MSG_CMSG_COMPAT & flags) {
1883 if (get_compat_msghdr(&msg_sys, msg_compat))
1884 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001885 }
1886 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1887 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001889 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890 if (!sock)
1891 goto out;
1892
1893 err = -EMSGSIZE;
1894 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1895 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001896
1897 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 err = -ENOMEM;
1899 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1900 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1901 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1902 if (!iov)
1903 goto out_put;
1904 }
1905
1906 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001907 * Save the user-mode address (verify_iovec will change the
1908 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001910
1911 uaddr = (void __user *)msg_sys.msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 uaddr_len = COMPAT_NAMELEN(msg);
1913 if (MSG_CMSG_COMPAT & flags) {
1914 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1915 } else
1916 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1917 if (err < 0)
1918 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001919 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920
1921 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1922 msg_sys.msg_flags = 0;
1923 if (MSG_CMSG_COMPAT & flags)
1924 msg_sys.msg_flags = MSG_CMSG_COMPAT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001925
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 if (sock->file->f_flags & O_NONBLOCK)
1927 flags |= MSG_DONTWAIT;
1928 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1929 if (err < 0)
1930 goto out_freeiov;
1931 len = err;
1932
1933 if (uaddr != NULL) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001934 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1935 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 if (err < 0)
1937 goto out_freeiov;
1938 }
David S. Miller37f7f422005-09-16 16:51:01 -07001939 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1940 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 if (err)
1942 goto out_freeiov;
1943 if (MSG_CMSG_COMPAT & flags)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001944 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 &msg_compat->msg_controllen);
1946 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001947 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 &msg->msg_controllen);
1949 if (err)
1950 goto out_freeiov;
1951 err = len;
1952
1953out_freeiov:
1954 if (iov != iovstack)
1955 sock_kfree_s(sock->sk, iov, iov_size);
1956out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001957 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958out:
1959 return err;
1960}
1961
1962#ifdef __ARCH_WANT_SYS_SOCKETCALL
1963
1964/* Argument list sizes for sys_socketcall */
1965#define AL(x) ((x) * sizeof(unsigned long))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001966static const unsigned char nargs[18]={
1967 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1968 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1969 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1970};
1971
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972#undef AL
1973
1974/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001975 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 *
1977 * Argument checking cleaned up. Saved 20% in size.
1978 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001979 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 */
1981
1982asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1983{
1984 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001985 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 int err;
1987
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001988 if (call < 1 || call > SYS_RECVMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 return -EINVAL;
1990
1991 /* copy_from_user should be SMP safe. */
1992 if (copy_from_user(a, args, nargs[call]))
1993 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001994
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001995 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001996 if (err)
1997 return err;
1998
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001999 a0 = a[0];
2000 a1 = a[1];
2001
2002 switch (call) {
2003 case SYS_SOCKET:
2004 err = sys_socket(a0, a1, a[2]);
2005 break;
2006 case SYS_BIND:
2007 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2008 break;
2009 case SYS_CONNECT:
2010 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2011 break;
2012 case SYS_LISTEN:
2013 err = sys_listen(a0, a1);
2014 break;
2015 case SYS_ACCEPT:
2016 err =
2017 sys_accept(a0, (struct sockaddr __user *)a1,
2018 (int __user *)a[2]);
2019 break;
2020 case SYS_GETSOCKNAME:
2021 err =
2022 sys_getsockname(a0, (struct sockaddr __user *)a1,
2023 (int __user *)a[2]);
2024 break;
2025 case SYS_GETPEERNAME:
2026 err =
2027 sys_getpeername(a0, (struct sockaddr __user *)a1,
2028 (int __user *)a[2]);
2029 break;
2030 case SYS_SOCKETPAIR:
2031 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2032 break;
2033 case SYS_SEND:
2034 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2035 break;
2036 case SYS_SENDTO:
2037 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2038 (struct sockaddr __user *)a[4], a[5]);
2039 break;
2040 case SYS_RECV:
2041 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2042 break;
2043 case SYS_RECVFROM:
2044 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2045 (struct sockaddr __user *)a[4],
2046 (int __user *)a[5]);
2047 break;
2048 case SYS_SHUTDOWN:
2049 err = sys_shutdown(a0, a1);
2050 break;
2051 case SYS_SETSOCKOPT:
2052 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2053 break;
2054 case SYS_GETSOCKOPT:
2055 err =
2056 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2057 (int __user *)a[4]);
2058 break;
2059 case SYS_SENDMSG:
2060 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2061 break;
2062 case SYS_RECVMSG:
2063 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2064 break;
2065 default:
2066 err = -EINVAL;
2067 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068 }
2069 return err;
2070}
2071
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002072#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002074/**
2075 * sock_register - add a socket protocol handler
2076 * @ops: description of protocol
2077 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 * This function is called by a protocol handler that wants to
2079 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002080 * socket interface. The value ops->family coresponds to the
2081 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002083int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084{
2085 int err;
2086
2087 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002088 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2089 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 return -ENOBUFS;
2091 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002092
2093 spin_lock(&net_family_lock);
2094 if (net_families[ops->family])
2095 err = -EEXIST;
2096 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002097 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098 err = 0;
2099 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002100 spin_unlock(&net_family_lock);
2101
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002102 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 return err;
2104}
2105
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002106/**
2107 * sock_unregister - remove a protocol handler
2108 * @family: protocol family to remove
2109 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 * This function is called by a protocol handler that wants to
2111 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002112 * new socket creation.
2113 *
2114 * If protocol handler is a module, then it can use module reference
2115 * counts to protect against new references. If protocol handler is not
2116 * a module then it needs to provide its own protection in
2117 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002119void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002121 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002123 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002124 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002125 spin_unlock(&net_family_lock);
2126
2127 synchronize_rcu();
2128
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002129 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130}
2131
Andi Kleen77d76ea2005-12-22 12:43:42 -08002132static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133{
2134 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002135 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002137
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138 sk_init();
2139
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002141 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 */
2143 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144
2145 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002146 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147 */
2148
2149 init_inodecache();
2150 register_filesystem(&sock_fs_type);
2151 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002152
2153 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 */
2155
2156#ifdef CONFIG_NETFILTER
2157 netfilter_init();
2158#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002159
2160 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161}
2162
Andi Kleen77d76ea2005-12-22 12:43:42 -08002163core_initcall(sock_init); /* early initcall */
2164
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165#ifdef CONFIG_PROC_FS
2166void socket_seq_show(struct seq_file *seq)
2167{
2168 int cpu;
2169 int counter = 0;
2170
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002171 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002172 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173
2174 /* It can be negative, by the way. 8) */
2175 if (counter < 0)
2176 counter = 0;
2177
2178 seq_printf(seq, "sockets: used %d\n", counter);
2179}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002180#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002182#ifdef CONFIG_COMPAT
2183static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002184 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002185{
2186 struct socket *sock = file->private_data;
2187 int ret = -ENOIOCTLCMD;
2188
2189 if (sock->ops->compat_ioctl)
2190 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2191
2192 return ret;
2193}
2194#endif
2195
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002196int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2197{
2198 return sock->ops->bind(sock, addr, addrlen);
2199}
2200
2201int kernel_listen(struct socket *sock, int backlog)
2202{
2203 return sock->ops->listen(sock, backlog);
2204}
2205
2206int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2207{
2208 struct sock *sk = sock->sk;
2209 int err;
2210
2211 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2212 newsock);
2213 if (err < 0)
2214 goto done;
2215
2216 err = sock->ops->accept(sock, *newsock, flags);
2217 if (err < 0) {
2218 sock_release(*newsock);
2219 goto done;
2220 }
2221
2222 (*newsock)->ops = sock->ops;
2223
2224done:
2225 return err;
2226}
2227
2228int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2229 int flags)
2230{
2231 return sock->ops->connect(sock, addr, addrlen, flags);
2232}
2233
2234int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2235 int *addrlen)
2236{
2237 return sock->ops->getname(sock, addr, addrlen, 0);
2238}
2239
2240int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2241 int *addrlen)
2242{
2243 return sock->ops->getname(sock, addr, addrlen, 1);
2244}
2245
2246int kernel_getsockopt(struct socket *sock, int level, int optname,
2247 char *optval, int *optlen)
2248{
2249 mm_segment_t oldfs = get_fs();
2250 int err;
2251
2252 set_fs(KERNEL_DS);
2253 if (level == SOL_SOCKET)
2254 err = sock_getsockopt(sock, level, optname, optval, optlen);
2255 else
2256 err = sock->ops->getsockopt(sock, level, optname, optval,
2257 optlen);
2258 set_fs(oldfs);
2259 return err;
2260}
2261
2262int kernel_setsockopt(struct socket *sock, int level, int optname,
2263 char *optval, int optlen)
2264{
2265 mm_segment_t oldfs = get_fs();
2266 int err;
2267
2268 set_fs(KERNEL_DS);
2269 if (level == SOL_SOCKET)
2270 err = sock_setsockopt(sock, level, optname, optval, optlen);
2271 else
2272 err = sock->ops->setsockopt(sock, level, optname, optval,
2273 optlen);
2274 set_fs(oldfs);
2275 return err;
2276}
2277
2278int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2279 size_t size, int flags)
2280{
2281 if (sock->ops->sendpage)
2282 return sock->ops->sendpage(sock, page, offset, size, flags);
2283
2284 return sock_no_sendpage(sock, page, offset, size, flags);
2285}
2286
2287int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2288{
2289 mm_segment_t oldfs = get_fs();
2290 int err;
2291
2292 set_fs(KERNEL_DS);
2293 err = sock->ops->ioctl(sock, cmd, arg);
2294 set_fs(oldfs);
2295
2296 return err;
2297}
2298
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299/* ABI emulation layers need these two */
2300EXPORT_SYMBOL(move_addr_to_kernel);
2301EXPORT_SYMBOL(move_addr_to_user);
2302EXPORT_SYMBOL(sock_create);
2303EXPORT_SYMBOL(sock_create_kern);
2304EXPORT_SYMBOL(sock_create_lite);
2305EXPORT_SYMBOL(sock_map_fd);
2306EXPORT_SYMBOL(sock_recvmsg);
2307EXPORT_SYMBOL(sock_register);
2308EXPORT_SYMBOL(sock_release);
2309EXPORT_SYMBOL(sock_sendmsg);
2310EXPORT_SYMBOL(sock_unregister);
2311EXPORT_SYMBOL(sock_wake_async);
2312EXPORT_SYMBOL(sockfd_lookup);
2313EXPORT_SYMBOL(kernel_sendmsg);
2314EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002315EXPORT_SYMBOL(kernel_bind);
2316EXPORT_SYMBOL(kernel_listen);
2317EXPORT_SYMBOL(kernel_accept);
2318EXPORT_SYMBOL(kernel_connect);
2319EXPORT_SYMBOL(kernel_getsockname);
2320EXPORT_SYMBOL(kernel_getpeername);
2321EXPORT_SYMBOL(kernel_getsockopt);
2322EXPORT_SYMBOL(kernel_setsockopt);
2323EXPORT_SYMBOL(kernel_sendpage);
2324EXPORT_SYMBOL(kernel_sock_ioctl);