ia64/linux-2.6.18-xen.hg

view net/socket.c @ 871:9cbcc9008446

xen/x86: don't initialize cpu_data[]'s apicid field on generic code

Afaict, this is not only redundant with the intialization done in
drivers/xen/core/smpboot.c, but actually results - at least for
secondary CPUs - in the Xen-specific value written to be later
overwritten with whatever the generic code determines (with no
guarantee that the two values are identical).

Signed-off-by: Jan Beulich <jbeulich@novell.com>
author Keir Fraser <keir.fraser@citrix.com>
date Thu May 14 10:09:15 2009 +0100 (2009-05-14)
parents 831230e53067
children
line source
1 /*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
56 * paradigm.
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
61 #include <linux/mm.h>
62 #include <linux/smp_lock.h>
63 #include <linux/socket.h>
64 #include <linux/file.h>
65 #include <linux/net.h>
66 #include <linux/interrupt.h>
67 #include <linux/netdevice.h>
68 #include <linux/proc_fs.h>
69 #include <linux/seq_file.h>
70 #include <linux/mutex.h>
71 #include <linux/wanrouter.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #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>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
89 #include <asm/uaccess.h>
90 #include <asm/unistd.h>
92 #include <net/compat.h>
94 #include <net/sock.h>
95 #include <linux/netfilter.h>
97 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
98 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
99 size_t size, loff_t pos);
100 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
101 size_t size, loff_t pos);
102 static int sock_mmap(struct file *file, struct vm_area_struct * vma);
104 static int sock_close(struct inode *inode, struct file *file);
105 static unsigned int sock_poll(struct file *file,
106 struct poll_table_struct *wait);
107 static long sock_ioctl(struct file *file,
108 unsigned int cmd, unsigned long arg);
109 #ifdef CONFIG_COMPAT
110 static long compat_sock_ioctl(struct file *file,
111 unsigned int cmd, unsigned long arg);
112 #endif
113 static int sock_fasync(int fd, struct file *filp, int on);
114 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
115 unsigned long count, loff_t *ppos);
116 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
117 unsigned long count, loff_t *ppos);
118 static ssize_t sock_sendpage(struct file *file, struct page *page,
119 int offset, size_t size, loff_t *ppos, int more);
121 /*
122 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
123 * in the operation structures but are done directly via the socketcall() multiplexor.
124 */
126 static struct file_operations socket_file_ops = {
127 .owner = THIS_MODULE,
128 .llseek = no_llseek,
129 .aio_read = sock_aio_read,
130 .aio_write = sock_aio_write,
131 .poll = sock_poll,
132 .unlocked_ioctl = sock_ioctl,
133 #ifdef CONFIG_COMPAT
134 .compat_ioctl = compat_sock_ioctl,
135 #endif
136 .mmap = sock_mmap,
137 .open = sock_no_open, /* special open code to disallow open via /proc */
138 .release = sock_close,
139 .fasync = sock_fasync,
140 .readv = sock_readv,
141 .writev = sock_writev,
142 .sendpage = sock_sendpage,
143 .splice_write = generic_splice_sendpage,
144 };
146 /*
147 * The protocol list. Each protocol is registered in here.
148 */
150 static struct net_proto_family *net_families[NPROTO];
152 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
153 static atomic_t net_family_lockct = ATOMIC_INIT(0);
154 static DEFINE_SPINLOCK(net_family_lock);
156 /* The strategy is: modifications net_family vector are short, do not
157 sleep and veeery rare, but read access should be free of any exclusive
158 locks.
159 */
161 static void net_family_write_lock(void)
162 {
163 spin_lock(&net_family_lock);
164 while (atomic_read(&net_family_lockct) != 0) {
165 spin_unlock(&net_family_lock);
167 yield();
169 spin_lock(&net_family_lock);
170 }
171 }
173 static __inline__ void net_family_write_unlock(void)
174 {
175 spin_unlock(&net_family_lock);
176 }
178 static __inline__ void net_family_read_lock(void)
179 {
180 atomic_inc(&net_family_lockct);
181 spin_unlock_wait(&net_family_lock);
182 }
184 static __inline__ void net_family_read_unlock(void)
185 {
186 atomic_dec(&net_family_lockct);
187 }
189 #else
190 #define net_family_write_lock() do { } while(0)
191 #define net_family_write_unlock() do { } while(0)
192 #define net_family_read_lock() do { } while(0)
193 #define net_family_read_unlock() do { } while(0)
194 #endif
197 /*
198 * Statistics counters of the socket lists
199 */
201 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
203 /*
204 * Support routines. Move socket addresses back and forth across the kernel/user
205 * divide and look after the messy bits.
206 */
208 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
209 16 for IP, 16 for IPX,
210 24 for IPv6,
211 about 80 for AX.25
212 must be at least one bigger than
213 the AF_UNIX size (see net/unix/af_unix.c
214 :unix_mkname()).
215 */
217 /**
218 * move_addr_to_kernel - copy a socket address into kernel space
219 * @uaddr: Address in user space
220 * @kaddr: Address in kernel space
221 * @ulen: Length in user space
222 *
223 * The address is copied into kernel space. If the provided address is
224 * too long an error code of -EINVAL is returned. If the copy gives
225 * invalid addresses -EFAULT is returned. On a success 0 is returned.
226 */
228 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
229 {
230 if(ulen<0||ulen>MAX_SOCK_ADDR)
231 return -EINVAL;
232 if(ulen==0)
233 return 0;
234 if(copy_from_user(kaddr,uaddr,ulen))
235 return -EFAULT;
236 return audit_sockaddr(ulen, kaddr);
237 }
239 /**
240 * move_addr_to_user - copy an address to user space
241 * @kaddr: kernel space address
242 * @klen: length of address in kernel
243 * @uaddr: user space address
244 * @ulen: pointer to user length field
245 *
246 * The value pointed to by ulen on entry is the buffer length available.
247 * This is overwritten with the buffer space used. -EINVAL is returned
248 * if an overlong buffer is specified or a negative buffer size. -EFAULT
249 * is returned if either the buffer or the length field are not
250 * accessible.
251 * After copying the data up to the limit the user specifies, the true
252 * length of the data is written over the length limit the user
253 * specified. Zero is returned for a success.
254 */
256 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
257 {
258 int err;
259 int len;
261 if((err=get_user(len, ulen)))
262 return err;
263 if(len>klen)
264 len=klen;
265 if(len<0 || len> MAX_SOCK_ADDR)
266 return -EINVAL;
267 if(len)
268 {
269 if (audit_sockaddr(klen, kaddr))
270 return -ENOMEM;
271 if(copy_to_user(uaddr,kaddr,len))
272 return -EFAULT;
273 }
274 /*
275 * "fromlen shall refer to the value before truncation.."
276 * 1003.1g
277 */
278 return __put_user(klen, ulen);
279 }
281 #define SOCKFS_MAGIC 0x534F434B
283 static kmem_cache_t * sock_inode_cachep __read_mostly;
285 static struct inode *sock_alloc_inode(struct super_block *sb)
286 {
287 struct socket_alloc *ei;
288 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
289 if (!ei)
290 return NULL;
291 init_waitqueue_head(&ei->socket.wait);
293 ei->socket.fasync_list = NULL;
294 ei->socket.state = SS_UNCONNECTED;
295 ei->socket.flags = 0;
296 ei->socket.ops = NULL;
297 ei->socket.sk = NULL;
298 ei->socket.file = NULL;
299 ei->socket.flags = 0;
301 return &ei->vfs_inode;
302 }
304 static void sock_destroy_inode(struct inode *inode)
305 {
306 kmem_cache_free(sock_inode_cachep,
307 container_of(inode, struct socket_alloc, vfs_inode));
308 }
310 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
311 {
312 struct socket_alloc *ei = (struct socket_alloc *) foo;
314 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
315 SLAB_CTOR_CONSTRUCTOR)
316 inode_init_once(&ei->vfs_inode);
317 }
319 static int init_inodecache(void)
320 {
321 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
322 sizeof(struct socket_alloc),
323 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
324 SLAB_MEM_SPREAD),
325 init_once, NULL);
326 if (sock_inode_cachep == NULL)
327 return -ENOMEM;
328 return 0;
329 }
331 static struct super_operations sockfs_ops = {
332 .alloc_inode = sock_alloc_inode,
333 .destroy_inode =sock_destroy_inode,
334 .statfs = simple_statfs,
335 };
337 static int sockfs_get_sb(struct file_system_type *fs_type,
338 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
339 {
340 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
341 mnt);
342 }
344 static struct vfsmount *sock_mnt __read_mostly;
346 static struct file_system_type sock_fs_type = {
347 .name = "sockfs",
348 .get_sb = sockfs_get_sb,
349 .kill_sb = kill_anon_super,
350 };
351 static int sockfs_delete_dentry(struct dentry *dentry)
352 {
353 return 1;
354 }
355 static struct dentry_operations sockfs_dentry_operations = {
356 .d_delete = sockfs_delete_dentry,
357 };
359 /*
360 * Obtains the first available file descriptor and sets it up for use.
361 *
362 * These functions create file structures and maps them to fd space
363 * of the current process. On success it returns file descriptor
364 * and file struct implicitly stored in sock->file.
365 * Note that another thread may close file descriptor before we return
366 * from this function. We use the fact that now we do not refer
367 * to socket after mapping. If one day we will need it, this
368 * function will increment ref. count on file by 1.
369 *
370 * In any case returned fd MAY BE not valid!
371 * This race condition is unavoidable
372 * with shared fd spaces, we cannot solve it inside kernel,
373 * but we take care of internal coherence yet.
374 */
376 static int sock_alloc_fd(struct file **filep)
377 {
378 int fd;
380 fd = get_unused_fd();
381 if (likely(fd >= 0)) {
382 struct file *file = get_empty_filp();
384 *filep = file;
385 if (unlikely(!file)) {
386 put_unused_fd(fd);
387 return -ENFILE;
388 }
389 } else
390 *filep = NULL;
391 return fd;
392 }
394 static int sock_attach_fd(struct socket *sock, struct file *file)
395 {
396 struct qstr this;
397 char name[32];
399 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
400 this.name = name;
401 this.hash = SOCK_INODE(sock)->i_ino;
403 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
404 if (unlikely(!file->f_dentry))
405 return -ENOMEM;
407 file->f_dentry->d_op = &sockfs_dentry_operations;
408 d_add(file->f_dentry, SOCK_INODE(sock));
409 file->f_vfsmnt = mntget(sock_mnt);
410 file->f_mapping = file->f_dentry->d_inode->i_mapping;
412 sock->file = file;
413 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
414 file->f_mode = FMODE_READ | FMODE_WRITE;
415 file->f_flags = O_RDWR;
416 file->f_pos = 0;
417 file->private_data = sock;
419 return 0;
420 }
422 int sock_map_fd(struct socket *sock)
423 {
424 struct file *newfile;
425 int fd = sock_alloc_fd(&newfile);
427 if (likely(fd >= 0)) {
428 int err = sock_attach_fd(sock, newfile);
430 if (unlikely(err < 0)) {
431 put_filp(newfile);
432 put_unused_fd(fd);
433 return err;
434 }
435 fd_install(fd, newfile);
436 }
437 return fd;
438 }
440 static struct socket *sock_from_file(struct file *file, int *err)
441 {
442 struct inode *inode;
443 struct socket *sock;
445 if (file->f_op == &socket_file_ops)
446 return file->private_data; /* set in sock_map_fd */
448 inode = file->f_dentry->d_inode;
449 if (!S_ISSOCK(inode->i_mode)) {
450 *err = -ENOTSOCK;
451 return NULL;
452 }
454 sock = SOCKET_I(inode);
455 if (sock->file != file) {
456 printk(KERN_ERR "socki_lookup: socket file changed!\n");
457 sock->file = file;
458 }
459 return sock;
460 }
462 /**
463 * sockfd_lookup - Go from a file number to its socket slot
464 * @fd: file handle
465 * @err: pointer to an error code return
466 *
467 * The file handle passed in is locked and the socket it is bound
468 * too is returned. If an error occurs the err pointer is overwritten
469 * with a negative errno code and NULL is returned. The function checks
470 * for both invalid handles and passing a handle which is not a socket.
471 *
472 * On a success the socket object pointer is returned.
473 */
475 struct socket *sockfd_lookup(int fd, int *err)
476 {
477 struct file *file;
478 struct socket *sock;
480 if (!(file = fget(fd))) {
481 *err = -EBADF;
482 return NULL;
483 }
484 sock = sock_from_file(file, err);
485 if (!sock)
486 fput(file);
487 return sock;
488 }
490 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
491 {
492 struct file *file;
493 struct socket *sock;
495 *err = -EBADF;
496 file = fget_light(fd, fput_needed);
497 if (file) {
498 sock = sock_from_file(file, err);
499 if (sock)
500 return sock;
501 fput_light(file, *fput_needed);
502 }
503 return NULL;
504 }
506 /**
507 * sock_alloc - allocate a socket
508 *
509 * Allocate a new inode and socket object. The two are bound together
510 * and initialised. The socket is then returned. If we are out of inodes
511 * NULL is returned.
512 */
514 static struct socket *sock_alloc(void)
515 {
516 struct inode * inode;
517 struct socket * sock;
519 inode = new_inode(sock_mnt->mnt_sb);
520 if (!inode)
521 return NULL;
523 sock = SOCKET_I(inode);
525 inode->i_mode = S_IFSOCK|S_IRWXUGO;
526 inode->i_uid = current->fsuid;
527 inode->i_gid = current->fsgid;
529 get_cpu_var(sockets_in_use)++;
530 put_cpu_var(sockets_in_use);
531 return sock;
532 }
534 /*
535 * In theory you can't get an open on this inode, but /proc provides
536 * a back door. Remember to keep it shut otherwise you'll let the
537 * creepy crawlies in.
538 */
540 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
541 {
542 return -ENXIO;
543 }
545 const struct file_operations bad_sock_fops = {
546 .owner = THIS_MODULE,
547 .open = sock_no_open,
548 };
550 /**
551 * sock_release - close a socket
552 * @sock: socket to close
553 *
554 * The socket is released from the protocol stack if it has a release
555 * callback, and the inode is then released if the socket is bound to
556 * an inode not a file.
557 */
559 void sock_release(struct socket *sock)
560 {
561 if (sock->ops) {
562 struct module *owner = sock->ops->owner;
564 sock->ops->release(sock);
565 sock->ops = NULL;
566 module_put(owner);
567 }
569 if (sock->fasync_list)
570 printk(KERN_ERR "sock_release: fasync list not empty!\n");
572 get_cpu_var(sockets_in_use)--;
573 put_cpu_var(sockets_in_use);
574 if (!sock->file) {
575 iput(SOCK_INODE(sock));
576 return;
577 }
578 sock->file=NULL;
579 }
581 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
582 struct msghdr *msg, size_t size)
583 {
584 struct sock_iocb *si = kiocb_to_siocb(iocb);
585 int err;
587 si->sock = sock;
588 si->scm = NULL;
589 si->msg = msg;
590 si->size = size;
592 err = security_socket_sendmsg(sock, msg, size);
593 if (err)
594 return err;
596 return sock->ops->sendmsg(iocb, sock, msg, size);
597 }
599 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
600 {
601 struct kiocb iocb;
602 struct sock_iocb siocb;
603 int ret;
605 init_sync_kiocb(&iocb, NULL);
606 iocb.private = &siocb;
607 ret = __sock_sendmsg(&iocb, sock, msg, size);
608 if (-EIOCBQUEUED == ret)
609 ret = wait_on_sync_kiocb(&iocb);
610 return ret;
611 }
613 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
614 struct kvec *vec, size_t num, size_t size)
615 {
616 mm_segment_t oldfs = get_fs();
617 int result;
619 set_fs(KERNEL_DS);
620 /*
621 * the following is safe, since for compiler definitions of kvec and
622 * iovec are identical, yielding the same in-core layout and alignment
623 */
624 msg->msg_iov = (struct iovec *)vec,
625 msg->msg_iovlen = num;
626 result = sock_sendmsg(sock, msg, size);
627 set_fs(oldfs);
628 return result;
629 }
631 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
632 struct msghdr *msg, size_t size, int flags)
633 {
634 int err;
635 struct sock_iocb *si = kiocb_to_siocb(iocb);
637 si->sock = sock;
638 si->scm = NULL;
639 si->msg = msg;
640 si->size = size;
641 si->flags = flags;
643 err = security_socket_recvmsg(sock, msg, size, flags);
644 if (err)
645 return err;
647 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
648 }
650 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
651 size_t size, int flags)
652 {
653 struct kiocb iocb;
654 struct sock_iocb siocb;
655 int ret;
657 init_sync_kiocb(&iocb, NULL);
658 iocb.private = &siocb;
659 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
660 if (-EIOCBQUEUED == ret)
661 ret = wait_on_sync_kiocb(&iocb);
662 return ret;
663 }
665 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
666 struct kvec *vec, size_t num,
667 size_t size, int flags)
668 {
669 mm_segment_t oldfs = get_fs();
670 int result;
672 set_fs(KERNEL_DS);
673 /*
674 * the following is safe, since for compiler definitions of kvec and
675 * iovec are identical, yielding the same in-core layout and alignment
676 */
677 msg->msg_iov = (struct iovec *)vec,
678 msg->msg_iovlen = num;
679 result = sock_recvmsg(sock, msg, size, flags);
680 set_fs(oldfs);
681 return result;
682 }
684 static void sock_aio_dtor(struct kiocb *iocb)
685 {
686 kfree(iocb->private);
687 }
689 static ssize_t sock_sendpage(struct file *file, struct page *page,
690 int offset, size_t size, loff_t *ppos, int more)
691 {
692 struct socket *sock;
693 int flags;
695 sock = file->private_data;
697 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
698 if (more)
699 flags |= MSG_MORE;
701 return sock->ops->sendpage(sock, page, offset, size, flags);
702 }
704 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
705 char __user *ubuf, size_t size, struct sock_iocb *siocb)
706 {
707 if (!is_sync_kiocb(iocb)) {
708 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
709 if (!siocb)
710 return NULL;
711 iocb->ki_dtor = sock_aio_dtor;
712 }
714 siocb->kiocb = iocb;
715 siocb->async_iov.iov_base = ubuf;
716 siocb->async_iov.iov_len = size;
718 iocb->private = siocb;
719 return siocb;
720 }
722 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
723 struct file *file, struct iovec *iov, unsigned long nr_segs)
724 {
725 struct socket *sock = file->private_data;
726 size_t size = 0;
727 int i;
729 for (i = 0 ; i < nr_segs ; i++)
730 size += iov[i].iov_len;
732 msg->msg_name = NULL;
733 msg->msg_namelen = 0;
734 msg->msg_control = NULL;
735 msg->msg_controllen = 0;
736 msg->msg_iov = (struct iovec *) iov;
737 msg->msg_iovlen = nr_segs;
738 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
740 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
741 }
743 static ssize_t sock_readv(struct file *file, const struct iovec *iov,
744 unsigned long nr_segs, loff_t *ppos)
745 {
746 struct kiocb iocb;
747 struct sock_iocb siocb;
748 struct msghdr msg;
749 int ret;
751 init_sync_kiocb(&iocb, NULL);
752 iocb.private = &siocb;
754 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
755 if (-EIOCBQUEUED == ret)
756 ret = wait_on_sync_kiocb(&iocb);
757 return ret;
758 }
760 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
761 size_t count, loff_t pos)
762 {
763 struct sock_iocb siocb, *x;
765 if (pos != 0)
766 return -ESPIPE;
767 if (count == 0) /* Match SYS5 behaviour */
768 return 0;
770 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
771 if (!x)
772 return -ENOMEM;
773 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
774 &x->async_iov, 1);
775 }
777 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
778 struct file *file, struct iovec *iov, unsigned long nr_segs)
779 {
780 struct socket *sock = file->private_data;
781 size_t size = 0;
782 int i;
784 for (i = 0 ; i < nr_segs ; i++)
785 size += iov[i].iov_len;
787 msg->msg_name = NULL;
788 msg->msg_namelen = 0;
789 msg->msg_control = NULL;
790 msg->msg_controllen = 0;
791 msg->msg_iov = (struct iovec *) iov;
792 msg->msg_iovlen = nr_segs;
793 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
794 if (sock->type == SOCK_SEQPACKET)
795 msg->msg_flags |= MSG_EOR;
797 return __sock_sendmsg(iocb, sock, msg, size);
798 }
800 static ssize_t sock_writev(struct file *file, const struct iovec *iov,
801 unsigned long nr_segs, loff_t *ppos)
802 {
803 struct msghdr msg;
804 struct kiocb iocb;
805 struct sock_iocb siocb;
806 int ret;
808 init_sync_kiocb(&iocb, NULL);
809 iocb.private = &siocb;
811 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
812 if (-EIOCBQUEUED == ret)
813 ret = wait_on_sync_kiocb(&iocb);
814 return ret;
815 }
817 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
818 size_t count, loff_t pos)
819 {
820 struct sock_iocb siocb, *x;
822 if (pos != 0)
823 return -ESPIPE;
824 if (count == 0) /* Match SYS5 behaviour */
825 return 0;
827 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
828 if (!x)
829 return -ENOMEM;
831 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
832 &x->async_iov, 1);
833 }
836 /*
837 * Atomic setting of ioctl hooks to avoid race
838 * with module unload.
839 */
841 static DEFINE_MUTEX(br_ioctl_mutex);
842 static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
844 void brioctl_set(int (*hook)(unsigned int, void __user *))
845 {
846 mutex_lock(&br_ioctl_mutex);
847 br_ioctl_hook = hook;
848 mutex_unlock(&br_ioctl_mutex);
849 }
850 EXPORT_SYMBOL(brioctl_set);
852 static DEFINE_MUTEX(vlan_ioctl_mutex);
853 static int (*vlan_ioctl_hook)(void __user *arg);
855 void vlan_ioctl_set(int (*hook)(void __user *))
856 {
857 mutex_lock(&vlan_ioctl_mutex);
858 vlan_ioctl_hook = hook;
859 mutex_unlock(&vlan_ioctl_mutex);
860 }
861 EXPORT_SYMBOL(vlan_ioctl_set);
863 static DEFINE_MUTEX(dlci_ioctl_mutex);
864 static int (*dlci_ioctl_hook)(unsigned int, void __user *);
866 void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
867 {
868 mutex_lock(&dlci_ioctl_mutex);
869 dlci_ioctl_hook = hook;
870 mutex_unlock(&dlci_ioctl_mutex);
871 }
872 EXPORT_SYMBOL(dlci_ioctl_set);
874 /*
875 * With an ioctl, arg may well be a user mode pointer, but we don't know
876 * what to do with it - that's up to the protocol still.
877 */
879 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
880 {
881 struct socket *sock;
882 void __user *argp = (void __user *)arg;
883 int pid, err;
885 sock = file->private_data;
886 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
887 err = dev_ioctl(cmd, argp);
888 } else
889 #ifdef CONFIG_WIRELESS_EXT
890 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
891 err = dev_ioctl(cmd, argp);
892 } else
893 #endif /* CONFIG_WIRELESS_EXT */
894 switch (cmd) {
895 case FIOSETOWN:
896 case SIOCSPGRP:
897 err = -EFAULT;
898 if (get_user(pid, (int __user *)argp))
899 break;
900 err = f_setown(sock->file, pid, 1);
901 break;
902 case FIOGETOWN:
903 case SIOCGPGRP:
904 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
905 break;
906 case SIOCGIFBR:
907 case SIOCSIFBR:
908 case SIOCBRADDBR:
909 case SIOCBRDELBR:
910 err = -ENOPKG;
911 if (!br_ioctl_hook)
912 request_module("bridge");
914 mutex_lock(&br_ioctl_mutex);
915 if (br_ioctl_hook)
916 err = br_ioctl_hook(cmd, argp);
917 mutex_unlock(&br_ioctl_mutex);
918 break;
919 case SIOCGIFVLAN:
920 case SIOCSIFVLAN:
921 err = -ENOPKG;
922 if (!vlan_ioctl_hook)
923 request_module("8021q");
925 mutex_lock(&vlan_ioctl_mutex);
926 if (vlan_ioctl_hook)
927 err = vlan_ioctl_hook(argp);
928 mutex_unlock(&vlan_ioctl_mutex);
929 break;
930 case SIOCGIFDIVERT:
931 case SIOCSIFDIVERT:
932 /* Convert this to call through a hook */
933 err = divert_ioctl(cmd, argp);
934 break;
935 case SIOCADDDLCI:
936 case SIOCDELDLCI:
937 err = -ENOPKG;
938 if (!dlci_ioctl_hook)
939 request_module("dlci");
941 if (dlci_ioctl_hook) {
942 mutex_lock(&dlci_ioctl_mutex);
943 err = dlci_ioctl_hook(cmd, argp);
944 mutex_unlock(&dlci_ioctl_mutex);
945 }
946 break;
947 default:
948 err = sock->ops->ioctl(sock, cmd, arg);
950 /*
951 * If this ioctl is unknown try to hand it down
952 * to the NIC driver.
953 */
954 if (err == -ENOIOCTLCMD)
955 err = dev_ioctl(cmd, argp);
956 break;
957 }
958 return err;
959 }
961 int sock_create_lite(int family, int type, int protocol, struct socket **res)
962 {
963 int err;
964 struct socket *sock = NULL;
966 err = security_socket_create(family, type, protocol, 1);
967 if (err)
968 goto out;
970 sock = sock_alloc();
971 if (!sock) {
972 err = -ENOMEM;
973 goto out;
974 }
976 security_socket_post_create(sock, family, type, protocol, 1);
977 sock->type = type;
978 out:
979 *res = sock;
980 return err;
981 }
983 /* No kernel lock held - perfect */
984 static unsigned int sock_poll(struct file *file, poll_table * wait)
985 {
986 struct socket *sock;
988 /*
989 * We can't return errors to poll, so it's either yes or no.
990 */
991 sock = file->private_data;
992 return sock->ops->poll(file, sock, wait);
993 }
995 static int sock_mmap(struct file * file, struct vm_area_struct * vma)
996 {
997 struct socket *sock = file->private_data;
999 return sock->ops->mmap(file, sock, vma);
1002 static int sock_close(struct inode *inode, struct file *filp)
1004 /*
1005 * It was possible the inode is NULL we were
1006 * closing an unfinished socket.
1007 */
1009 if (!inode)
1011 printk(KERN_DEBUG "sock_close: NULL inode\n");
1012 return 0;
1014 sock_fasync(-1, filp, 0);
1015 sock_release(SOCKET_I(inode));
1016 return 0;
1019 /*
1020 * Update the socket async list
1022 * Fasync_list locking strategy.
1024 * 1. fasync_list is modified only under process context socket lock
1025 * i.e. under semaphore.
1026 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1027 * or under socket lock.
1028 * 3. fasync_list can be used from softirq context, so that
1029 * modification under socket lock have to be enhanced with
1030 * write_lock_bh(&sk->sk_callback_lock).
1031 * --ANK (990710)
1032 */
1034 static int sock_fasync(int fd, struct file *filp, int on)
1036 struct fasync_struct *fa, *fna=NULL, **prev;
1037 struct socket *sock;
1038 struct sock *sk;
1040 if (on)
1042 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1043 if(fna==NULL)
1044 return -ENOMEM;
1047 sock = filp->private_data;
1049 if ((sk=sock->sk) == NULL) {
1050 kfree(fna);
1051 return -EINVAL;
1054 lock_sock(sk);
1056 prev=&(sock->fasync_list);
1058 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1059 if (fa->fa_file==filp)
1060 break;
1062 if(on)
1064 if(fa!=NULL)
1066 write_lock_bh(&sk->sk_callback_lock);
1067 fa->fa_fd=fd;
1068 write_unlock_bh(&sk->sk_callback_lock);
1070 kfree(fna);
1071 goto out;
1073 fna->fa_file=filp;
1074 fna->fa_fd=fd;
1075 fna->magic=FASYNC_MAGIC;
1076 fna->fa_next=sock->fasync_list;
1077 write_lock_bh(&sk->sk_callback_lock);
1078 sock->fasync_list=fna;
1079 write_unlock_bh(&sk->sk_callback_lock);
1081 else
1083 if (fa!=NULL)
1085 write_lock_bh(&sk->sk_callback_lock);
1086 *prev=fa->fa_next;
1087 write_unlock_bh(&sk->sk_callback_lock);
1088 kfree(fa);
1092 out:
1093 release_sock(sock->sk);
1094 return 0;
1097 /* This function may be called only under socket lock or callback_lock */
1099 int sock_wake_async(struct socket *sock, int how, int band)
1101 if (!sock || !sock->fasync_list)
1102 return -1;
1103 switch (how)
1105 case 1:
1107 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1108 break;
1109 goto call_kill;
1110 case 2:
1111 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1112 break;
1113 /* fall through */
1114 case 0:
1115 call_kill:
1116 __kill_fasync(sock->fasync_list, SIGIO, band);
1117 break;
1118 case 3:
1119 __kill_fasync(sock->fasync_list, SIGURG, band);
1121 return 0;
1124 static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1126 int err;
1127 struct socket *sock;
1129 /*
1130 * Check protocol is in range
1131 */
1132 if (family < 0 || family >= NPROTO)
1133 return -EAFNOSUPPORT;
1134 if (type < 0 || type >= SOCK_MAX)
1135 return -EINVAL;
1137 /* Compatibility.
1139 This uglymoron is moved from INET layer to here to avoid
1140 deadlock in module load.
1141 */
1142 if (family == PF_INET && type == SOCK_PACKET) {
1143 static int warned;
1144 if (!warned) {
1145 warned = 1;
1146 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1148 family = PF_PACKET;
1151 err = security_socket_create(family, type, protocol, kern);
1152 if (err)
1153 return err;
1155 #if defined(CONFIG_KMOD)
1156 /* Attempt to load a protocol module if the find failed.
1158 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1159 * requested real, full-featured networking support upon configuration.
1160 * Otherwise module support will break!
1161 */
1162 if (net_families[family]==NULL)
1164 request_module("net-pf-%d",family);
1166 #endif
1168 net_family_read_lock();
1169 if (net_families[family] == NULL) {
1170 err = -EAFNOSUPPORT;
1171 goto out;
1174 /*
1175 * Allocate the socket and allow the family to set things up. if
1176 * the protocol is 0, the family is instructed to select an appropriate
1177 * default.
1178 */
1180 if (!(sock = sock_alloc())) {
1181 if (net_ratelimit())
1182 printk(KERN_WARNING "socket: no more sockets\n");
1183 err = -ENFILE; /* Not exactly a match, but its the
1184 closest posix thing */
1185 goto out;
1188 sock->type = type;
1190 /*
1191 * We will call the ->create function, that possibly is in a loadable
1192 * module, so we have to bump that loadable module refcnt first.
1193 */
1194 err = -EAFNOSUPPORT;
1195 if (!try_module_get(net_families[family]->owner))
1196 goto out_release;
1198 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1199 sock->ops = NULL;
1200 goto out_module_put;
1203 /*
1204 * Now to bump the refcnt of the [loadable] module that owns this
1205 * socket at sock_release time we decrement its refcnt.
1206 */
1207 if (!try_module_get(sock->ops->owner)) {
1208 sock->ops = NULL;
1209 goto out_module_put;
1211 /*
1212 * Now that we're done with the ->create function, the [loadable]
1213 * module can have its refcnt decremented
1214 */
1215 module_put(net_families[family]->owner);
1216 *res = sock;
1217 security_socket_post_create(sock, family, type, protocol, kern);
1219 out:
1220 net_family_read_unlock();
1221 return err;
1222 out_module_put:
1223 module_put(net_families[family]->owner);
1224 out_release:
1225 sock_release(sock);
1226 goto out;
1229 int sock_create(int family, int type, int protocol, struct socket **res)
1231 return __sock_create(family, type, protocol, res, 0);
1234 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1236 return __sock_create(family, type, protocol, res, 1);
1239 asmlinkage long sys_socket(int family, int type, int protocol)
1241 int retval;
1242 struct socket *sock;
1244 retval = sock_create(family, type, protocol, &sock);
1245 if (retval < 0)
1246 goto out;
1248 retval = sock_map_fd(sock);
1249 if (retval < 0)
1250 goto out_release;
1252 out:
1253 /* It may be already another descriptor 8) Not kernel problem. */
1254 return retval;
1256 out_release:
1257 sock_release(sock);
1258 return retval;
1261 /*
1262 * Create a pair of connected sockets.
1263 */
1265 asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1267 struct socket *sock1, *sock2;
1268 int fd1, fd2, err;
1270 /*
1271 * Obtain the first socket and check if the underlying protocol
1272 * supports the socketpair call.
1273 */
1275 err = sock_create(family, type, protocol, &sock1);
1276 if (err < 0)
1277 goto out;
1279 err = sock_create(family, type, protocol, &sock2);
1280 if (err < 0)
1281 goto out_release_1;
1283 err = sock1->ops->socketpair(sock1, sock2);
1284 if (err < 0)
1285 goto out_release_both;
1287 fd1 = fd2 = -1;
1289 err = sock_map_fd(sock1);
1290 if (err < 0)
1291 goto out_release_both;
1292 fd1 = err;
1294 err = sock_map_fd(sock2);
1295 if (err < 0)
1296 goto out_close_1;
1297 fd2 = err;
1299 /* fd1 and fd2 may be already another descriptors.
1300 * Not kernel problem.
1301 */
1303 err = put_user(fd1, &usockvec[0]);
1304 if (!err)
1305 err = put_user(fd2, &usockvec[1]);
1306 if (!err)
1307 return 0;
1309 sys_close(fd2);
1310 sys_close(fd1);
1311 return err;
1313 out_close_1:
1314 sock_release(sock2);
1315 sys_close(fd1);
1316 return err;
1318 out_release_both:
1319 sock_release(sock2);
1320 out_release_1:
1321 sock_release(sock1);
1322 out:
1323 return err;
1327 /*
1328 * Bind a name to a socket. Nothing much to do here since it's
1329 * the protocol's responsibility to handle the local address.
1331 * We move the socket address to kernel space before we call
1332 * the protocol layer (having also checked the address is ok).
1333 */
1335 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1337 struct socket *sock;
1338 char address[MAX_SOCK_ADDR];
1339 int err, fput_needed;
1341 if((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1343 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1344 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1345 if (!err)
1346 err = sock->ops->bind(sock,
1347 (struct sockaddr *)address, addrlen);
1349 fput_light(sock->file, fput_needed);
1351 return err;
1355 /*
1356 * Perform a listen. Basically, we allow the protocol to do anything
1357 * necessary for a listen, and if that works, we mark the socket as
1358 * ready for listening.
1359 */
1361 int sysctl_somaxconn = SOMAXCONN;
1363 asmlinkage long sys_listen(int fd, int backlog)
1365 struct socket *sock;
1366 int err, fput_needed;
1368 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1369 if ((unsigned) backlog > sysctl_somaxconn)
1370 backlog = sysctl_somaxconn;
1372 err = security_socket_listen(sock, backlog);
1373 if (!err)
1374 err = sock->ops->listen(sock, backlog);
1376 fput_light(sock->file, fput_needed);
1378 return err;
1382 /*
1383 * For accept, we attempt to create a new socket, set up the link
1384 * with the client, wake up the client, then return the new
1385 * connected fd. We collect the address of the connector in kernel
1386 * space and move it to user at the very end. This is unclean because
1387 * we open the socket then return an error.
1389 * 1003.1g adds the ability to recvmsg() to query connection pending
1390 * status to recvmsg. We need to add that support in a way thats
1391 * clean when we restucture accept also.
1392 */
1394 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1396 struct socket *sock, *newsock;
1397 struct file *newfile;
1398 int err, len, newfd, fput_needed;
1399 char address[MAX_SOCK_ADDR];
1401 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1402 if (!sock)
1403 goto out;
1405 err = -ENFILE;
1406 if (!(newsock = sock_alloc()))
1407 goto out_put;
1409 newsock->type = sock->type;
1410 newsock->ops = sock->ops;
1412 /*
1413 * We don't need try_module_get here, as the listening socket (sock)
1414 * has the protocol module (sock->ops->owner) held.
1415 */
1416 __module_get(newsock->ops->owner);
1418 newfd = sock_alloc_fd(&newfile);
1419 if (unlikely(newfd < 0)) {
1420 err = newfd;
1421 sock_release(newsock);
1422 goto out_put;
1425 err = sock_attach_fd(newsock, newfile);
1426 if (err < 0)
1427 goto out_fd;
1429 err = security_socket_accept(sock, newsock);
1430 if (err)
1431 goto out_fd;
1433 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1434 if (err < 0)
1435 goto out_fd;
1437 if (upeer_sockaddr) {
1438 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1439 err = -ECONNABORTED;
1440 goto out_fd;
1442 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1443 if (err < 0)
1444 goto out_fd;
1447 /* File flags are not inherited via accept() unlike another OSes. */
1449 fd_install(newfd, newfile);
1450 err = newfd;
1452 security_socket_post_accept(sock, newsock);
1454 out_put:
1455 fput_light(sock->file, fput_needed);
1456 out:
1457 return err;
1458 out_fd:
1459 fput(newfile);
1460 put_unused_fd(newfd);
1461 goto out_put;
1465 /*
1466 * Attempt to connect to a socket with the server address. The address
1467 * is in user space so we verify it is OK and move it to kernel space.
1469 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1470 * break bindings
1472 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1473 * other SEQPACKET protocols that take time to connect() as it doesn't
1474 * include the -EINPROGRESS status for such sockets.
1475 */
1477 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1479 struct socket *sock;
1480 char address[MAX_SOCK_ADDR];
1481 int err, fput_needed;
1483 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1484 if (!sock)
1485 goto out;
1486 err = move_addr_to_kernel(uservaddr, addrlen, address);
1487 if (err < 0)
1488 goto out_put;
1490 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1491 if (err)
1492 goto out_put;
1494 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1495 sock->file->f_flags);
1496 out_put:
1497 fput_light(sock->file, fput_needed);
1498 out:
1499 return err;
1502 /*
1503 * Get the local address ('name') of a socket object. Move the obtained
1504 * name to user space.
1505 */
1507 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1509 struct socket *sock;
1510 char address[MAX_SOCK_ADDR];
1511 int len, err, fput_needed;
1513 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1514 if (!sock)
1515 goto out;
1517 err = security_socket_getsockname(sock);
1518 if (err)
1519 goto out_put;
1521 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1522 if (err)
1523 goto out_put;
1524 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1526 out_put:
1527 fput_light(sock->file, fput_needed);
1528 out:
1529 return err;
1532 /*
1533 * Get the remote address ('name') of a socket object. Move the obtained
1534 * name to user space.
1535 */
1537 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1539 struct socket *sock;
1540 char address[MAX_SOCK_ADDR];
1541 int len, err, fput_needed;
1543 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1544 err = security_socket_getpeername(sock);
1545 if (err) {
1546 fput_light(sock->file, fput_needed);
1547 return err;
1550 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1551 if (!err)
1552 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1553 fput_light(sock->file, fput_needed);
1555 return err;
1558 /*
1559 * Send a datagram to a given address. We move the address into kernel
1560 * space and check the user space data area is readable before invoking
1561 * the protocol.
1562 */
1564 asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1565 struct sockaddr __user *addr, int addr_len)
1567 struct socket *sock;
1568 char address[MAX_SOCK_ADDR];
1569 int err;
1570 struct msghdr msg;
1571 struct iovec iov;
1572 int fput_needed;
1573 struct file *sock_file;
1575 sock_file = fget_light(fd, &fput_needed);
1576 if (!sock_file)
1577 return -EBADF;
1579 sock = sock_from_file(sock_file, &err);
1580 if (!sock)
1581 goto out_put;
1582 iov.iov_base=buff;
1583 iov.iov_len=len;
1584 msg.msg_name=NULL;
1585 msg.msg_iov=&iov;
1586 msg.msg_iovlen=1;
1587 msg.msg_control=NULL;
1588 msg.msg_controllen=0;
1589 msg.msg_namelen=0;
1590 if (addr) {
1591 err = move_addr_to_kernel(addr, addr_len, address);
1592 if (err < 0)
1593 goto out_put;
1594 msg.msg_name=address;
1595 msg.msg_namelen=addr_len;
1597 if (sock->file->f_flags & O_NONBLOCK)
1598 flags |= MSG_DONTWAIT;
1599 msg.msg_flags = flags;
1600 err = sock_sendmsg(sock, &msg, len);
1602 out_put:
1603 fput_light(sock_file, fput_needed);
1604 return err;
1607 /*
1608 * Send a datagram down a socket.
1609 */
1611 asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1613 return sys_sendto(fd, buff, len, flags, NULL, 0);
1616 /*
1617 * Receive a frame from the socket and optionally record the address of the
1618 * sender. We verify the buffers are writable and if needed move the
1619 * sender address from kernel to user space.
1620 */
1622 asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1623 struct sockaddr __user *addr, int __user *addr_len)
1625 struct socket *sock;
1626 struct iovec iov;
1627 struct msghdr msg;
1628 char address[MAX_SOCK_ADDR];
1629 int err,err2;
1630 struct file *sock_file;
1631 int fput_needed;
1633 sock_file = fget_light(fd, &fput_needed);
1634 if (!sock_file)
1635 return -EBADF;
1637 sock = sock_from_file(sock_file, &err);
1638 if (!sock)
1639 goto out;
1641 msg.msg_control=NULL;
1642 msg.msg_controllen=0;
1643 msg.msg_iovlen=1;
1644 msg.msg_iov=&iov;
1645 iov.iov_len=size;
1646 iov.iov_base=ubuf;
1647 msg.msg_name=address;
1648 msg.msg_namelen=MAX_SOCK_ADDR;
1649 if (sock->file->f_flags & O_NONBLOCK)
1650 flags |= MSG_DONTWAIT;
1651 err=sock_recvmsg(sock, &msg, size, flags);
1653 if(err >= 0 && addr != NULL)
1655 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1656 if(err2<0)
1657 err=err2;
1659 out:
1660 fput_light(sock_file, fput_needed);
1661 return err;
1664 /*
1665 * Receive a datagram from a socket.
1666 */
1668 asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1670 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1673 /*
1674 * Set a socket option. Because we don't know the option lengths we have
1675 * to pass the user mode parameter for the protocols to sort out.
1676 */
1678 asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1680 int err, fput_needed;
1681 struct socket *sock;
1683 if (optlen < 0)
1684 return -EINVAL;
1686 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL)
1688 err = security_socket_setsockopt(sock,level,optname);
1689 if (err)
1690 goto out_put;
1692 if (level == SOL_SOCKET)
1693 err=sock_setsockopt(sock,level,optname,optval,optlen);
1694 else
1695 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1696 out_put:
1697 fput_light(sock->file, fput_needed);
1699 return err;
1702 /*
1703 * Get a socket option. Because we don't know the option lengths we have
1704 * to pass a user mode parameter for the protocols to sort out.
1705 */
1707 asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1709 int err, fput_needed;
1710 struct socket *sock;
1712 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1713 err = security_socket_getsockopt(sock, level, optname);
1714 if (err)
1715 goto out_put;
1717 if (level == SOL_SOCKET)
1718 err=sock_getsockopt(sock,level,optname,optval,optlen);
1719 else
1720 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1721 out_put:
1722 fput_light(sock->file, fput_needed);
1724 return err;
1728 /*
1729 * Shutdown a socket.
1730 */
1732 asmlinkage long sys_shutdown(int fd, int how)
1734 int err, fput_needed;
1735 struct socket *sock;
1737 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1739 err = security_socket_shutdown(sock, how);
1740 if (!err)
1741 err = sock->ops->shutdown(sock, how);
1742 fput_light(sock->file, fput_needed);
1744 return err;
1747 /* A couple of helpful macros for getting the address of the 32/64 bit
1748 * fields which are the same type (int / unsigned) on our platforms.
1749 */
1750 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1751 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1752 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1755 /*
1756 * BSD sendmsg interface
1757 */
1759 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1761 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1762 struct socket *sock;
1763 char address[MAX_SOCK_ADDR];
1764 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1765 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1766 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1767 /* 20 is size of ipv6_pktinfo */
1768 unsigned char *ctl_buf = ctl;
1769 struct msghdr msg_sys;
1770 int err, ctl_len, iov_size, total_len;
1771 int fput_needed;
1773 err = -EFAULT;
1774 if (MSG_CMSG_COMPAT & flags) {
1775 if (get_compat_msghdr(&msg_sys, msg_compat))
1776 return -EFAULT;
1777 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1778 return -EFAULT;
1780 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1781 if (!sock)
1782 goto out;
1784 /* do not move before msg_sys is valid */
1785 err = -EMSGSIZE;
1786 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1787 goto out_put;
1789 /* Check whether to allocate the iovec area*/
1790 err = -ENOMEM;
1791 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1792 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1793 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1794 if (!iov)
1795 goto out_put;
1798 /* This will also move the address data into kernel space */
1799 if (MSG_CMSG_COMPAT & flags) {
1800 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1801 } else
1802 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1803 if (err < 0)
1804 goto out_freeiov;
1805 total_len = err;
1807 err = -ENOBUFS;
1809 if (msg_sys.msg_controllen > INT_MAX)
1810 goto out_freeiov;
1811 ctl_len = msg_sys.msg_controllen;
1812 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1813 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
1814 if (err)
1815 goto out_freeiov;
1816 ctl_buf = msg_sys.msg_control;
1817 ctl_len = msg_sys.msg_controllen;
1818 } else if (ctl_len) {
1819 if (ctl_len > sizeof(ctl))
1821 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1822 if (ctl_buf == NULL)
1823 goto out_freeiov;
1825 err = -EFAULT;
1826 /*
1827 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1828 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1829 * checking falls down on this.
1830 */
1831 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1832 goto out_freectl;
1833 msg_sys.msg_control = ctl_buf;
1835 msg_sys.msg_flags = flags;
1837 if (sock->file->f_flags & O_NONBLOCK)
1838 msg_sys.msg_flags |= MSG_DONTWAIT;
1839 err = sock_sendmsg(sock, &msg_sys, total_len);
1841 out_freectl:
1842 if (ctl_buf != ctl)
1843 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1844 out_freeiov:
1845 if (iov != iovstack)
1846 sock_kfree_s(sock->sk, iov, iov_size);
1847 out_put:
1848 fput_light(sock->file, fput_needed);
1849 out:
1850 return err;
1853 /*
1854 * BSD recvmsg interface
1855 */
1857 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1859 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1860 struct socket *sock;
1861 struct iovec iovstack[UIO_FASTIOV];
1862 struct iovec *iov=iovstack;
1863 struct msghdr msg_sys;
1864 unsigned long cmsg_ptr;
1865 int err, iov_size, total_len, len;
1866 int fput_needed;
1868 /* kernel mode address */
1869 char addr[MAX_SOCK_ADDR];
1871 /* user mode address pointers */
1872 struct sockaddr __user *uaddr;
1873 int __user *uaddr_len;
1875 if (MSG_CMSG_COMPAT & flags) {
1876 if (get_compat_msghdr(&msg_sys, msg_compat))
1877 return -EFAULT;
1878 } else
1879 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1880 return -EFAULT;
1882 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1883 if (!sock)
1884 goto out;
1886 err = -EMSGSIZE;
1887 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1888 goto out_put;
1890 /* Check whether to allocate the iovec area*/
1891 err = -ENOMEM;
1892 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1893 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1894 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1895 if (!iov)
1896 goto out_put;
1899 /*
1900 * Save the user-mode address (verify_iovec will change the
1901 * kernel msghdr to use the kernel address space)
1902 */
1904 uaddr = (void __user *) msg_sys.msg_name;
1905 uaddr_len = COMPAT_NAMELEN(msg);
1906 if (MSG_CMSG_COMPAT & flags) {
1907 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1908 } else
1909 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1910 if (err < 0)
1911 goto out_freeiov;
1912 total_len=err;
1914 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1915 msg_sys.msg_flags = 0;
1916 if (MSG_CMSG_COMPAT & flags)
1917 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1919 if (sock->file->f_flags & O_NONBLOCK)
1920 flags |= MSG_DONTWAIT;
1921 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1922 if (err < 0)
1923 goto out_freeiov;
1924 len = err;
1926 if (uaddr != NULL) {
1927 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1928 if (err < 0)
1929 goto out_freeiov;
1931 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1932 COMPAT_FLAGS(msg));
1933 if (err)
1934 goto out_freeiov;
1935 if (MSG_CMSG_COMPAT & flags)
1936 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1937 &msg_compat->msg_controllen);
1938 else
1939 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1940 &msg->msg_controllen);
1941 if (err)
1942 goto out_freeiov;
1943 err = len;
1945 out_freeiov:
1946 if (iov != iovstack)
1947 sock_kfree_s(sock->sk, iov, iov_size);
1948 out_put:
1949 fput_light(sock->file, fput_needed);
1950 out:
1951 return err;
1954 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1956 /* Argument list sizes for sys_socketcall */
1957 #define AL(x) ((x) * sizeof(unsigned long))
1958 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1959 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1960 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1961 #undef AL
1963 /*
1964 * System call vectors.
1966 * Argument checking cleaned up. Saved 20% in size.
1967 * This function doesn't need to set the kernel lock because
1968 * it is set by the callees.
1969 */
1971 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1973 unsigned long a[6];
1974 unsigned long a0,a1;
1975 int err;
1977 if(call<1||call>SYS_RECVMSG)
1978 return -EINVAL;
1980 /* copy_from_user should be SMP safe. */
1981 if (copy_from_user(a, args, nargs[call]))
1982 return -EFAULT;
1984 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
1985 if (err)
1986 return err;
1988 a0=a[0];
1989 a1=a[1];
1991 switch(call)
1993 case SYS_SOCKET:
1994 err = sys_socket(a0,a1,a[2]);
1995 break;
1996 case SYS_BIND:
1997 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1998 break;
1999 case SYS_CONNECT:
2000 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2001 break;
2002 case SYS_LISTEN:
2003 err = sys_listen(a0,a1);
2004 break;
2005 case SYS_ACCEPT:
2006 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2007 break;
2008 case SYS_GETSOCKNAME:
2009 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2010 break;
2011 case SYS_GETPEERNAME:
2012 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
2013 break;
2014 case SYS_SOCKETPAIR:
2015 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
2016 break;
2017 case SYS_SEND:
2018 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2019 break;
2020 case SYS_SENDTO:
2021 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2022 (struct sockaddr __user *)a[4], a[5]);
2023 break;
2024 case SYS_RECV:
2025 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2026 break;
2027 case SYS_RECVFROM:
2028 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2029 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2030 break;
2031 case SYS_SHUTDOWN:
2032 err = sys_shutdown(a0,a1);
2033 break;
2034 case SYS_SETSOCKOPT:
2035 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2036 break;
2037 case SYS_GETSOCKOPT:
2038 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2039 break;
2040 case SYS_SENDMSG:
2041 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2042 break;
2043 case SYS_RECVMSG:
2044 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2045 break;
2046 default:
2047 err = -EINVAL;
2048 break;
2050 return err;
2053 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2055 /*
2056 * This function is called by a protocol handler that wants to
2057 * advertise its address family, and have it linked into the
2058 * SOCKET module.
2059 */
2061 int sock_register(struct net_proto_family *ops)
2063 int err;
2065 if (ops->family >= NPROTO) {
2066 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2067 return -ENOBUFS;
2069 net_family_write_lock();
2070 err = -EEXIST;
2071 if (net_families[ops->family] == NULL) {
2072 net_families[ops->family]=ops;
2073 err = 0;
2075 net_family_write_unlock();
2076 printk(KERN_INFO "NET: Registered protocol family %d\n",
2077 ops->family);
2078 return err;
2081 /*
2082 * This function is called by a protocol handler that wants to
2083 * remove its address family, and have it unlinked from the
2084 * SOCKET module.
2085 */
2087 int sock_unregister(int family)
2089 if (family < 0 || family >= NPROTO)
2090 return -1;
2092 net_family_write_lock();
2093 net_families[family]=NULL;
2094 net_family_write_unlock();
2095 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2096 family);
2097 return 0;
2100 static int __init sock_init(void)
2102 /*
2103 * Initialize sock SLAB cache.
2104 */
2106 sk_init();
2108 /*
2109 * Initialize skbuff SLAB cache
2110 */
2111 skb_init();
2113 /*
2114 * Initialize the protocols module.
2115 */
2117 init_inodecache();
2118 register_filesystem(&sock_fs_type);
2119 sock_mnt = kern_mount(&sock_fs_type);
2121 /* The real protocol initialization is performed in later initcalls.
2122 */
2124 #ifdef CONFIG_NETFILTER
2125 netfilter_init();
2126 #endif
2128 return 0;
2131 core_initcall(sock_init); /* early initcall */
2133 #ifdef CONFIG_PROC_FS
2134 void socket_seq_show(struct seq_file *seq)
2136 int cpu;
2137 int counter = 0;
2139 for_each_possible_cpu(cpu)
2140 counter += per_cpu(sockets_in_use, cpu);
2142 /* It can be negative, by the way. 8) */
2143 if (counter < 0)
2144 counter = 0;
2146 seq_printf(seq, "sockets: used %d\n", counter);
2148 #endif /* CONFIG_PROC_FS */
2150 #ifdef CONFIG_COMPAT
2151 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2152 unsigned long arg)
2154 struct socket *sock = file->private_data;
2155 int ret = -ENOIOCTLCMD;
2157 if (sock->ops->compat_ioctl)
2158 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2160 return ret;
2162 #endif
2164 /* ABI emulation layers need these two */
2165 EXPORT_SYMBOL(move_addr_to_kernel);
2166 EXPORT_SYMBOL(move_addr_to_user);
2167 EXPORT_SYMBOL(sock_create);
2168 EXPORT_SYMBOL(sock_create_kern);
2169 EXPORT_SYMBOL(sock_create_lite);
2170 EXPORT_SYMBOL(sock_map_fd);
2171 EXPORT_SYMBOL(sock_recvmsg);
2172 EXPORT_SYMBOL(sock_register);
2173 EXPORT_SYMBOL(sock_release);
2174 EXPORT_SYMBOL(sock_sendmsg);
2175 EXPORT_SYMBOL(sock_unregister);
2176 EXPORT_SYMBOL(sock_wake_async);
2177 EXPORT_SYMBOL(sockfd_lookup);
2178 EXPORT_SYMBOL(kernel_sendmsg);
2179 EXPORT_SYMBOL(kernel_recvmsg);