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, <bir7@leland.Stanford.Edu>
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 */
59
60 #include <linux/config.h>
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/wanrouter.h>
70 #include <linux/init.h>
71 #include <linux/poll.h>
72 #include <linux/cache.h>
73 #include <linux/module.h>
74
75 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
76 #include <linux/kmod.h>
77 #endif
78
79 #include <asm/uaccess.h>
80
81 #include <linux/inet.h>
82 #include <net/ip.h>
83 #include <net/sock.h>
84 #include <net/tcp.h>
85 #include <net/udp.h>
86 #include <net/scm.h>
87 #include <linux/netfilter.h>
88
89 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
90 static loff_t sock_lseek(struct file *file, loff_t offset, int whence);
91 static ssize_t sock_read(struct file *file, char *buf,
92 size_t size, loff_t *ppos);
93 static ssize_t sock_write(struct file *file, const char *buf,
94 size_t size, loff_t *ppos);
95 static int sock_mmap(struct file *file, struct vm_area_struct * vma);
96
97 static int sock_close(struct inode *inode, struct file *file);
98 static unsigned int sock_poll(struct file *file,
99 struct poll_table_struct *wait);
100 static int sock_ioctl(struct inode *inode, struct file *file,
101 unsigned int cmd, unsigned long arg);
102 static int sock_fasync(int fd, struct file *filp, int on);
103 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
104 unsigned long count, loff_t *ppos);
105 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
106 unsigned long count, loff_t *ppos);
107
108
109 /*
110 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
111 * in the operation structures but are done directly via the socketcall() multiplexor.
112 */
113
114 static struct file_operations socket_file_ops = {
115 llseek: sock_lseek,
116 read: sock_read,
117 write: sock_write,
118 poll: sock_poll,
119 ioctl: sock_ioctl,
120 mmap: sock_mmap,
121 open: sock_no_open, /* special open code to disallow open via /proc */
122 release: sock_close,
123 fasync: sock_fasync,
124 readv: sock_readv,
125 writev: sock_writev
126 };
127
128 /*
129 * The protocol list. Each protocol is registered in here.
130 */
131
132 static struct net_proto_family *net_families[NPROTO];
133
134 #ifdef CONFIG_SMP
135 static atomic_t net_family_lockct = ATOMIC_INIT(0);
136 static spinlock_t net_family_lock = SPIN_LOCK_UNLOCKED;
137
138 /* The strategy is: modifications net_family vector are short, do not
139 sleep and veeery rare, but read access should be free of any exclusive
140 locks.
141 */
142
143 static void net_family_write_lock(void)
144 {
145 spin_lock(&net_family_lock);
146 while (atomic_read(&net_family_lockct) != 0) {
147 spin_unlock(&net_family_lock);
148
149 current->policy |= SCHED_YIELD;
150 schedule();
151
152 spin_lock(&net_family_lock);
153 }
154 }
155
156 static __inline__ void net_family_write_unlock(void)
157 {
158 spin_unlock(&net_family_lock);
159 }
160
161 static __inline__ void net_family_read_lock(void)
162 {
163 atomic_inc(&net_family_lockct);
164 spin_unlock_wait(&net_family_lock);
165 }
166
167 static __inline__ void net_family_read_unlock(void)
168 {
169 atomic_dec(&net_family_lockct);
170 }
171
172 #else
173 #define net_family_write_lock() do { } while(0)
174 #define net_family_write_unlock() do { } while(0)
175 #define net_family_read_lock() do { } while(0)
176 #define net_family_read_unlock() do { } while(0)
177 #endif
178
179
180 /*
181 * Statistics counters of the socket lists
182 */
183
184 static union {
185 int counter;
186 char __pad[SMP_CACHE_BYTES];
187 } sockets_in_use[NR_CPUS] __cacheline_aligned = {{0}};
188
189 /*
190 * Support routines. Move socket addresses back and forth across the kernel/user
191 * divide and look after the messy bits.
192 */
193
194 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
195 16 for IP, 16 for IPX,
196 24 for IPv6,
197 about 80 for AX.25
198 must be at least one bigger than
199 the AF_UNIX size (see net/unix/af_unix.c
200 :unix_mkname()).
201 */
202
203 /**
204 * move_addr_to_kernel - copy a socket address into kernel space
205 * @uaddr: Address in user space
206 * @kaddr: Address in kernel space
207 * @ulen: Length in user space
208 *
209 * The address is copied into kernel space. If the provided address is
210 * too long an error code of -EINVAL is returned. If the copy gives
211 * invalid addresses -EFAULT is returned. On a success 0 is returned.
212 */
213
214 int move_addr_to_kernel(void *uaddr, int ulen, void *kaddr)
215 {
216 if(ulen<0||ulen>MAX_SOCK_ADDR)
217 return -EINVAL;
218 if(ulen==0)
219 return 0;
220 if(copy_from_user(kaddr,uaddr,ulen))
221 return -EFAULT;
222 return 0;
223 }
224
225 /**
226 * move_addr_to_user - copy an address to user space
227 * @kaddr: kernel space address
228 * @klen: length of address in kernel
229 * @uaddr: user space address
230 * @ulen: pointer to user length field
231 *
232 * The value pointed to by ulen on entry is the buffer length available.
233 * This is overwritten with the buffer space used. -EINVAL is returned
234 * if an overlong buffer is specified or a negative buffer size. -EFAULT
235 * is returned if either the buffer or the length field are not
236 * accessible.
237 * After copying the data up to the limit the user specifies, the true
238 * length of the data is written over the length limit the user
239 * specified. Zero is returned for a success.
240 */
241
242 int move_addr_to_user(void *kaddr, int klen, void *uaddr, int *ulen)
243 {
244 int err;
245 int len;
246
247 if((err=get_user(len, ulen)))
248 return err;
249 if(len>klen)
250 len=klen;
251 if(len<0 || len> MAX_SOCK_ADDR)
252 return -EINVAL;
253 if(len)
254 {
255 if(copy_to_user(uaddr,kaddr,len))
256 return -EFAULT;
257 }
258 /*
259 * "fromlen shall refer to the value before truncation.."
260 * 1003.1g
261 */
262 return __put_user(klen, ulen);
263 }
264
265 #define SOCKFS_MAGIC 0x534F434B
266 static int sockfs_statfs(struct super_block *sb, struct statfs *buf)
267 {
268 buf->f_type = SOCKFS_MAGIC;
269 buf->f_bsize = 1024;
270 buf->f_namelen = 255;
271 return 0;
272 }
273
274 static struct super_operations sockfs_ops = {
275 statfs: sockfs_statfs,
276 };
277
278 static struct super_block * sockfs_read_super(struct super_block *sb, void *data, int silent)
279 {
280 struct inode *root = new_inode(sb);
281 if (!root)
282 return NULL;
283 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
284 root->i_uid = root->i_gid = 0;
285 root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
286 sb->s_blocksize = 1024;
287 sb->s_blocksize_bits = 10;
288 sb->s_magic = SOCKFS_MAGIC;
289 sb->s_op = &sockfs_ops;
290 sb->s_root = d_alloc(NULL, &(const struct qstr) { "socket:", 7, 0 });
291 if (!sb->s_root) {
292 iput(root);
293 return NULL;
294 }
295 sb->s_root->d_sb = sb;
296 sb->s_root->d_parent = sb->s_root;
297 d_instantiate(sb->s_root, root);
298 return sb;
299 }
300
301 static struct vfsmount *sock_mnt;
302 static DECLARE_FSTYPE(sock_fs_type, "sockfs", sockfs_read_super,
303 FS_NOMOUNT|FS_SINGLE);
304 static int sockfs_delete_dentry(struct dentry *dentry)
305 {
306 return 1;
307 }
308 static struct dentry_operations sockfs_dentry_operations = {
309 d_delete: sockfs_delete_dentry,
310 };
311
312 /*
313 * Obtains the first available file descriptor and sets it up for use.
314 *
315 * This functions creates file structure and maps it to fd space
316 * of current process. On success it returns file descriptor
317 * and file struct implicitly stored in sock->file.
318 * Note that another thread may close file descriptor before we return
319 * from this function. We use the fact that now we do not refer
320 * to socket after mapping. If one day we will need it, this
321 * function will inincrement ref. count on file by 1.
322 *
323 * In any case returned fd MAY BE not valid!
324 * This race condition is inavoidable
325 * with shared fd spaces, we cannot solve is inside kernel,
326 * but we take care of internal coherence yet.
327 */
328
329 static int sock_map_fd(struct socket *sock)
330 {
331 int fd;
332 struct qstr this;
333 char name[32];
334
335 /*
336 * Find a file descriptor suitable for return to the user.
337 */
338
339 fd = get_unused_fd();
340 if (fd >= 0) {
341 struct file *file = get_empty_filp();
342
343 if (!file) {
344 put_unused_fd(fd);
345 fd = -ENFILE;
346 goto out;
347 }
348
349 sprintf(name, "[%lu]", sock->inode->i_ino);
350 this.name = name;
351 this.len = strlen(name);
352 this.hash = sock->inode->i_ino;
353
354 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
355 if (!file->f_dentry) {
356 put_filp(file);
357 put_unused_fd(fd);
358 fd = -ENOMEM;
359 goto out;
360 }
361 file->f_dentry->d_op = &sockfs_dentry_operations;
362 d_add(file->f_dentry, sock->inode);
363 file->f_vfsmnt = mntget(sock_mnt);
364
365 sock->file = file;
366 file->f_op = sock->inode->i_fop = &socket_file_ops;
367 file->f_mode = 3;
368 file->f_flags = O_RDWR;
369 file->f_pos = 0;
370 fd_install(fd, file);
371 }
372
373 out:
374 return fd;
375 }
376
377 extern __inline__ struct socket *socki_lookup(struct inode *inode)
378 {
379 return &inode->u.socket_i;
380 }
381
382 /**
383 * sockfd_lookup - Go from a file number to its socket slot
384 * @fd: file handle
385 * @err: pointer to an error code return
386 *
387 * The file handle passed in is locked and the socket it is bound
388 * too is returned. If an error occurs the err pointer is overwritten
389 * with a negative errno code and NULL is returned. The function checks
390 * for both invalid handles and passing a handle which is not a socket.
391 *
392 * On a success the socket object pointer is returned.
393 */
394
395 struct socket *sockfd_lookup(int fd, int *err)
396 {
397 struct file *file;
398 struct inode *inode;
399 struct socket *sock;
400
401 if (!(file = fget(fd)))
402 {
403 *err = -EBADF;
404 return NULL;
405 }
406
407 inode = file->f_dentry->d_inode;
408 if (!inode->i_sock || !(sock = socki_lookup(inode)))
409 {
410 *err = -ENOTSOCK;
411 fput(file);
412 return NULL;
413 }
414
415 if (sock->file != file) {
416 printk(KERN_ERR "socki_lookup: socket file changed!\n");
417 sock->file = file;
418 }
419 return sock;
420 }
421
422 extern __inline__ void sockfd_put(struct socket *sock)
423 {
424 fput(sock->file);
425 }
426
427 /**
428 * sock_alloc - allocate a socket
429 *
430 * Allocate a new inode and socket object. The two are bound together
431 * and initialised. The socket is then returned. If we are out of inodes
432 * NULL is returned.
433 */
434
435 struct socket *sock_alloc(void)
436 {
437 struct inode * inode;
438 struct socket * sock;
439
440 inode = get_empty_inode();
441 if (!inode)
442 return NULL;
443
444 inode->i_sb = sock_mnt->mnt_sb;
445 sock = socki_lookup(inode);
446
447 inode->i_mode = S_IFSOCK|S_IRWXUGO;
448 inode->i_sock = 1;
449 inode->i_uid = current->fsuid;
450 inode->i_gid = current->fsgid;
451
452 sock->inode = inode;
453 init_waitqueue_head(&sock->wait);
454 sock->fasync_list = NULL;
455 sock->state = SS_UNCONNECTED;
456 sock->flags = 0;
457 sock->ops = NULL;
458 sock->sk = NULL;
459 sock->file = NULL;
460
461 sockets_in_use[smp_processor_id()].counter++;
462 return sock;
463 }
464
465 /*
466 * In theory you can't get an open on this inode, but /proc provides
467 * a back door. Remember to keep it shut otherwise you'll let the
468 * creepy crawlies in.
469 */
470
471 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
472 {
473 return -ENXIO;
474 }
475
476 /**
477 * sock_release - close a socket
478 * @sock: socket to close
479 *
480 * The socket is released from the protocol stack if it has a release
481 * callback, and the inode is then released if the socket is bound to
482 * an inode not a file.
483 */
484
485 void sock_release(struct socket *sock)
486 {
487 if (sock->ops)
488 sock->ops->release(sock);
489
490 if (sock->fasync_list)
491 printk(KERN_ERR "sock_release: fasync list not empty!\n");
492
493 sockets_in_use[smp_processor_id()].counter--;
494 if (!sock->file) {
495 iput(sock->inode);
496 return;
497 }
498 sock->file=NULL;
499 }
500
501 int sock_sendmsg(struct socket *sock, struct msghdr *msg, int size)
502 {
503 int err;
504 struct scm_cookie scm;
505
506 err = scm_send(sock, msg, &scm);
507 if (err >= 0) {
508 err = sock->ops->sendmsg(sock, msg, size, &scm);
509 scm_destroy(&scm);
510 }
511 return err;
512 }
513
514 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int size, int flags)
515 {
516 struct scm_cookie scm;
517
518 memset(&scm, 0, sizeof(scm));
519
520 size = sock->ops->recvmsg(sock, msg, size, flags, &scm);
521 if (size >= 0)
522 scm_recv(sock, msg, &scm, flags);
523
524 return size;
525 }
526
527
528 /*
529 * Sockets are not seekable.
530 */
531
532 static loff_t sock_lseek(struct file *file, loff_t offset, int whence)
533 {
534 return -ESPIPE;
535 }
536
537 /*
538 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
539 * area ubuf...ubuf+size-1 is writable before asking the protocol.
540 */
541
542 static ssize_t sock_read(struct file *file, char *ubuf,
543 size_t size, loff_t *ppos)
544 {
545 struct socket *sock;
546 struct iovec iov;
547 struct msghdr msg;
548 int flags;
549
550 if (ppos != &file->f_pos)
551 return -ESPIPE;
552 if (size==0) /* Match SYS5 behaviour */
553 return 0;
554
555 sock = socki_lookup(file->f_dentry->d_inode);
556
557 msg.msg_name=NULL;
558 msg.msg_namelen=0;
559 msg.msg_iov=&iov;
560 msg.msg_iovlen=1;
561 msg.msg_control=NULL;
562 msg.msg_controllen=0;
563 iov.iov_base=ubuf;
564 iov.iov_len=size;
565 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
566
567 return sock_recvmsg(sock, &msg, size, flags);
568 }
569
570
571 /*
572 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
573 * is readable by the user process.
574 */
575
576 static ssize_t sock_write(struct file *file, const char *ubuf,
577 size_t size, loff_t *ppos)
578 {
579 struct socket *sock;
580 struct msghdr msg;
581 struct iovec iov;
582
583 if (ppos != &file->f_pos)
584 return -ESPIPE;
585 if(size==0) /* Match SYS5 behaviour */
586 return 0;
587
588 sock = socki_lookup(file->f_dentry->d_inode);
589
590 msg.msg_name=NULL;
591 msg.msg_namelen=0;
592 msg.msg_iov=&iov;
593 msg.msg_iovlen=1;
594 msg.msg_control=NULL;
595 msg.msg_controllen=0;
596 msg.msg_flags=!(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
597 if (sock->type == SOCK_SEQPACKET)
598 msg.msg_flags |= MSG_EOR;
599 iov.iov_base=(void *)ubuf;
600 iov.iov_len=size;
601
602 return sock_sendmsg(sock, &msg, size);
603 }
604
605 int sock_readv_writev(int type, struct inode * inode, struct file * file,
606 const struct iovec * iov, long count, long size)
607 {
608 struct msghdr msg;
609 struct socket *sock;
610
611 sock = socki_lookup(inode);
612
613 msg.msg_name = NULL;
614 msg.msg_namelen = 0;
615 msg.msg_control = NULL;
616 msg.msg_controllen = 0;
617 msg.msg_iov = (struct iovec *) iov;
618 msg.msg_iovlen = count;
619 msg.msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
620
621 /* read() does a VERIFY_WRITE */
622 if (type == VERIFY_WRITE)
623 return sock_recvmsg(sock, &msg, size, msg.msg_flags);
624
625 if (sock->type == SOCK_SEQPACKET)
626 msg.msg_flags |= MSG_EOR;
627
628 return sock_sendmsg(sock, &msg, size);
629 }
630
631 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
632 unsigned long count, loff_t *ppos)
633 {
634 size_t tot_len = 0;
635 int i;
636 for (i = 0 ; i < count ; i++)
637 tot_len += vector[i].iov_len;
638 return sock_readv_writev(VERIFY_WRITE, file->f_dentry->d_inode,
639 file, vector, count, tot_len);
640 }
641
642 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
643 unsigned long count, loff_t *ppos)
644 {
645 size_t tot_len = 0;
646 int i;
647 for (i = 0 ; i < count ; i++)
648 tot_len += vector[i].iov_len;
649 return sock_readv_writev(VERIFY_READ, file->f_dentry->d_inode,
650 file, vector, count, tot_len);
651 }
652
653 /*
654 * With an ioctl arg may well be a user mode pointer, but we don't know what to do
655 * with it - that's up to the protocol still.
656 */
657
658 int sock_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
659 unsigned long arg)
660 {
661 struct socket *sock;
662 int err;
663
664 unlock_kernel();
665 sock = socki_lookup(inode);
666 err = sock->ops->ioctl(sock, cmd, arg);
667 lock_kernel();
668
669 return err;
670 }
671
672
673 /* No kernel lock held - perfect */
674 static unsigned int sock_poll(struct file *file, poll_table * wait)
675 {
676 struct socket *sock;
677
678 /*
679 * We can't return errors to poll, so it's either yes or no.
680 */
681 sock = socki_lookup(file->f_dentry->d_inode);
682 return sock->ops->poll(file, sock, wait);
683 }
684
685 static int sock_mmap(struct file * file, struct vm_area_struct * vma)
686 {
687 struct socket *sock = socki_lookup(file->f_dentry->d_inode);
688
689 return sock->ops->mmap(file, sock, vma);
690 }
691
692 int sock_close(struct inode *inode, struct file *filp)
693 {
694 /*
695 * It was possible the inode is NULL we were
696 * closing an unfinished socket.
697 */
698
699 if (!inode)
700 {
701 printk(KERN_DEBUG "sock_close: NULL inode\n");
702 return 0;
703 }
704 sock_fasync(-1, filp, 0);
705 sock_release(socki_lookup(inode));
706 return 0;
707 }
708
709 /*
710 * Update the socket async list
711 *
712 * Fasync_list locking strategy.
713 *
714 * 1. fasync_list is modified only under process context socket lock
715 * i.e. under semaphore.
716 * 2. fasync_list is used under read_lock(&sk->callback_lock)
717 * or under socket lock.
718 * 3. fasync_list can be used from softirq context, so that
719 * modification under socket lock have to be enhanced with
720 * write_lock_bh(&sk->callback_lock).
721 * --ANK (990710)
722 */
723
724 static int sock_fasync(int fd, struct file *filp, int on)
725 {
726 struct fasync_struct *fa, *fna=NULL, **prev;
727 struct socket *sock;
728 struct sock *sk;
729
730 if (on)
731 {
732 fna=(struct fasync_struct *)kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
733 if(fna==NULL)
734 return -ENOMEM;
735 }
736
737
738 sock = socki_lookup(filp->f_dentry->d_inode);
739
740 if ((sk=sock->sk) == NULL)
741 return -EINVAL;
742
743 lock_sock(sk);
744
745 prev=&(sock->fasync_list);
746
747 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
748 if (fa->fa_file==filp)
749 break;
750
751 if(on)
752 {
753 if(fa!=NULL)
754 {
755 write_lock_bh(&sk->callback_lock);
756 fa->fa_fd=fd;
757 write_unlock_bh(&sk->callback_lock);
758
759 kfree(fna);
760 goto out;
761 }
762 fna->fa_file=filp;
763 fna->fa_fd=fd;
764 fna->magic=FASYNC_MAGIC;
765 fna->fa_next=sock->fasync_list;
766 write_lock_bh(&sk->callback_lock);
767 sock->fasync_list=fna;
768 write_unlock_bh(&sk->callback_lock);
769 }
770 else
771 {
772 if (fa!=NULL)
773 {
774 write_lock_bh(&sk->callback_lock);
775 *prev=fa->fa_next;
776 write_unlock_bh(&sk->callback_lock);
777 kfree(fa);
778 }
779 }
780
781 out:
782 release_sock(sock->sk);
783 return 0;
784 }
785
786 /* This function may be called only under socket lock or callback_lock */
787
788 int sock_wake_async(struct socket *sock, int how, int band)
789 {
790 if (!sock || !sock->fasync_list)
791 return -1;
792 switch (how)
793 {
794 case 1:
795
796 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
797 break;
798 goto call_kill;
799 case 2:
800 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
801 break;
802 /* fall through */
803 case 0:
804 call_kill:
805 __kill_fasync(sock->fasync_list, SIGIO, band);
806 break;
807 case 3:
808 __kill_fasync(sock->fasync_list, SIGURG, band);
809 }
810 return 0;
811 }
812
813
814 int sock_create(int family, int type, int protocol, struct socket **res)
815 {
816 int i;
817 struct socket *sock;
818
819 /*
820 * Check protocol is in range
821 */
822 if(family<0 || family>=NPROTO)
823 return -EAFNOSUPPORT;
824
825 /* Compatibility.
826
827 This uglymoron is moved from INET layer to here to avoid
828 deadlock in module load.
829 */
830 if (family == PF_INET && type == SOCK_PACKET) {
831 static int warned;
832 if (!warned) {
833 warned = 1;
834 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
835 }
836 family = PF_PACKET;
837 }
838
839 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
840 /* Attempt to load a protocol module if the find failed.
841 *
842 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
843 * requested real, full-featured networking support upon configuration.
844 * Otherwise module support will break!
845 */
846 if (net_families[family]==NULL)
847 {
848 char module_name[30];
849 sprintf(module_name,"net-pf-%d",family);
850 request_module(module_name);
851 }
852 #endif
853
854 net_family_read_lock();
855 if (net_families[family] == NULL) {
856 i = -EAFNOSUPPORT;
857 goto out;
858 }
859
860 /*
861 * Allocate the socket and allow the family to set things up. if
862 * the protocol is 0, the family is instructed to select an appropriate
863 * default.
864 */
865
866 if (!(sock = sock_alloc()))
867 {
868 printk(KERN_WARNING "socket: no more sockets\n");
869 i = -ENFILE; /* Not exactly a match, but its the
870 closest posix thing */
871 goto out;
872 }
873
874 sock->type = type;
875
876 if ((i = net_families[family]->create(sock, protocol)) < 0)
877 {
878 sock_release(sock);
879 goto out;
880 }
881
882 *res = sock;
883
884 out:
885 net_family_read_unlock();
886 return i;
887 }
888
889 asmlinkage long sys_socket(int family, int type, int protocol)
890 {
891 int retval;
892 struct socket *sock;
893
894 retval = sock_create(family, type, protocol, &sock);
895 if (retval < 0)
896 goto out;
897
898 retval = sock_map_fd(sock);
899 if (retval < 0)
900 goto out_release;
901
902 out:
903 /* It may be already another descriptor 8) Not kernel problem. */
904 return retval;
905
906 out_release:
907 sock_release(sock);
908 return retval;
909 }
910
911 /*
912 * Create a pair of connected sockets.
913 */
914
915 asmlinkage long sys_socketpair(int family, int type, int protocol, int usockvec[2])
916 {
917 struct socket *sock1, *sock2;
918 int fd1, fd2, err;
919
920 /*
921 * Obtain the first socket and check if the underlying protocol
922 * supports the socketpair call.
923 */
924
925 err = sock_create(family, type, protocol, &sock1);
926 if (err < 0)
927 goto out;
928
929 err = sock_create(family, type, protocol, &sock2);
930 if (err < 0)
931 goto out_release_1;
932
933 err = sock1->ops->socketpair(sock1, sock2);
934 if (err < 0)
935 goto out_release_both;
936
937 fd1 = fd2 = -1;
938
939 err = sock_map_fd(sock1);
940 if (err < 0)
941 goto out_release_both;
942 fd1 = err;
943
944 err = sock_map_fd(sock2);
945 if (err < 0)
946 goto out_close_1;
947 fd2 = err;
948
949 /* fd1 and fd2 may be already another descriptors.
950 * Not kernel problem.
951 */
952
953 err = put_user(fd1, &usockvec[0]);
954 if (!err)
955 err = put_user(fd2, &usockvec[1]);
956 if (!err)
957 return 0;
958
959 sys_close(fd2);
960 sys_close(fd1);
961 return err;
962
963 out_close_1:
964 sock_release(sock2);
965 sys_close(fd1);
966 return err;
967
968 out_release_both:
969 sock_release(sock2);
970 out_release_1:
971 sock_release(sock1);
972 out:
973 return err;
974 }
975
976
977 /*
978 * Bind a name to a socket. Nothing much to do here since it's
979 * the protocol's responsibility to handle the local address.
980 *
981 * We move the socket address to kernel space before we call
982 * the protocol layer (having also checked the address is ok).
983 */
984
985 asmlinkage long sys_bind(int fd, struct sockaddr *umyaddr, int addrlen)
986 {
987 struct socket *sock;
988 char address[MAX_SOCK_ADDR];
989 int err;
990
991 if((sock = sockfd_lookup(fd,&err))!=NULL)
992 {
993 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0)
994 err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
995 sockfd_put(sock);
996 }
997 return err;
998 }
999
1000
1001 /*
1002 * Perform a listen. Basically, we allow the protocol to do anything
1003 * necessary for a listen, and if that works, we mark the socket as
1004 * ready for listening.
1005 */
1006
1007 asmlinkage long sys_listen(int fd, int backlog)
1008 {
1009 struct socket *sock;
1010 int err;
1011
1012 if ((sock = sockfd_lookup(fd, &err)) != NULL) {
1013 if ((unsigned) backlog > SOMAXCONN)
1014 backlog = SOMAXCONN;
1015 err=sock->ops->listen(sock, backlog);
1016 sockfd_put(sock);
1017 }
1018 return err;
1019 }
1020
1021
1022 /*
1023 * For accept, we attempt to create a new socket, set up the link
1024 * with the client, wake up the client, then return the new
1025 * connected fd. We collect the address of the connector in kernel
1026 * space and move it to user at the very end. This is unclean because
1027 * we open the socket then return an error.
1028 *
1029 * 1003.1g adds the ability to recvmsg() to query connection pending
1030 * status to recvmsg. We need to add that support in a way thats
1031 * clean when we restucture accept also.
1032 */
1033
1034 asmlinkage long sys_accept(int fd, struct sockaddr *upeer_sockaddr, int *upeer_addrlen)
1035 {
1036 struct socket *sock, *newsock;
1037 int err, len;
1038 char address[MAX_SOCK_ADDR];
1039
1040 sock = sockfd_lookup(fd, &err);
1041 if (!sock)
1042 goto out;
1043
1044 err = -EMFILE;
1045 if (!(newsock = sock_alloc()))
1046 goto out_put;
1047
1048 newsock->type = sock->type;
1049 newsock->ops = sock->ops;
1050
1051 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1052 if (err < 0)
1053 goto out_release;
1054
1055 if (upeer_sockaddr) {
1056 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1057 err = -ECONNABORTED;
1058 goto out_release;
1059 }
1060 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1061 if (err < 0)
1062 goto out_release;
1063 }
1064
1065 /* File flags are not inherited via accept() unlike another OSes. */
1066
1067 if ((err = sock_map_fd(newsock)) < 0)
1068 goto out_release;
1069
1070 out_put:
1071 sockfd_put(sock);
1072 out:
1073 return err;
1074
1075 out_release:
1076 sock_release(newsock);
1077 goto out_put;
1078 }
1079
1080
1081 /*
1082 * Attempt to connect to a socket with the server address. The address
1083 * is in user space so we verify it is OK and move it to kernel space.
1084 *
1085 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1086 * break bindings
1087 *
1088 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1089 * other SEQPACKET protocols that take time to connect() as it doesn't
1090 * include the -EINPROGRESS status for such sockets.
1091 */
1092
1093 asmlinkage long sys_connect(int fd, struct sockaddr *uservaddr, int addrlen)
1094 {
1095 struct socket *sock;
1096 char address[MAX_SOCK_ADDR];
1097 int err;
1098
1099 sock = sockfd_lookup(fd, &err);
1100 if (!sock)
1101 goto out;
1102 err = move_addr_to_kernel(uservaddr, addrlen, address);
1103 if (err < 0)
1104 goto out_put;
1105 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1106 sock->file->f_flags);
1107 out_put:
1108 sockfd_put(sock);
1109 out:
1110 return err;
1111 }
1112
1113 /*
1114 * Get the local address ('name') of a socket object. Move the obtained
1115 * name to user space.
1116 */
1117
1118 asmlinkage long sys_getsockname(int fd, struct sockaddr *usockaddr, int *usockaddr_len)
1119 {
1120 struct socket *sock;
1121 char address[MAX_SOCK_ADDR];
1122 int len, err;
1123
1124 sock = sockfd_lookup(fd, &err);
1125 if (!sock)
1126 goto out;
1127 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1128 if (err)
1129 goto out_put;
1130 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1131
1132 out_put:
1133 sockfd_put(sock);
1134 out:
1135 return err;
1136 }
1137
1138 /*
1139 * Get the remote address ('name') of a socket object. Move the obtained
1140 * name to user space.
1141 */
1142
1143 asmlinkage long sys_getpeername(int fd, struct sockaddr *usockaddr, int *usockaddr_len)
1144 {
1145 struct socket *sock;
1146 char address[MAX_SOCK_ADDR];
1147 int len, err;
1148
1149 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1150 {
1151 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1152 if (!err)
1153 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1154 sockfd_put(sock);
1155 }
1156 return err;
1157 }
1158
1159 /*
1160 * Send a datagram to a given address. We move the address into kernel
1161 * space and check the user space data area is readable before invoking
1162 * the protocol.
1163 */
1164
1165 asmlinkage long sys_sendto(int fd, void * buff, size_t len, unsigned flags,
1166 struct sockaddr *addr, int addr_len)
1167 {
1168 struct socket *sock;
1169 char address[MAX_SOCK_ADDR];
1170 int err;
1171 struct msghdr msg;
1172 struct iovec iov;
1173
1174 sock = sockfd_lookup(fd, &err);
1175 if (!sock)
1176 goto out;
1177 iov.iov_base=buff;
1178 iov.iov_len=len;
1179 msg.msg_name=NULL;
1180 msg.msg_iov=&iov;
1181 msg.msg_iovlen=1;
1182 msg.msg_control=NULL;
1183 msg.msg_controllen=0;
1184 msg.msg_namelen=addr_len;
1185 if(addr)
1186 {
1187 err = move_addr_to_kernel(addr, addr_len, address);
1188 if (err < 0)
1189 goto out_put;
1190 msg.msg_name=address;
1191 }
1192 if (sock->file->f_flags & O_NONBLOCK)
1193 flags |= MSG_DONTWAIT;
1194 msg.msg_flags = flags;
1195 err = sock_sendmsg(sock, &msg, len);
1196
1197 out_put:
1198 sockfd_put(sock);
1199 out:
1200 return err;
1201 }
1202
1203 /*
1204 * Send a datagram down a socket.
1205 */
1206
1207 asmlinkage long sys_send(int fd, void * buff, size_t len, unsigned flags)
1208 {
1209 return sys_sendto(fd, buff, len, flags, NULL, 0);
1210 }
1211
1212 /*
1213 * Receive a frame from the socket and optionally record the address of the
1214 * sender. We verify the buffers are writable and if needed move the
1215 * sender address from kernel to user space.
1216 */
1217
1218 asmlinkage long sys_recvfrom(int fd, void * ubuf, size_t size, unsigned flags,
1219 struct sockaddr *addr, int *addr_len)
1220 {
1221 struct socket *sock;
1222 struct iovec iov;
1223 struct msghdr msg;
1224 char address[MAX_SOCK_ADDR];
1225 int err,err2;
1226
1227 sock = sockfd_lookup(fd, &err);
1228 if (!sock)
1229 goto out;
1230
1231 msg.msg_control=NULL;
1232 msg.msg_controllen=0;
1233 msg.msg_iovlen=1;
1234 msg.msg_iov=&iov;
1235 iov.iov_len=size;
1236 iov.iov_base=ubuf;
1237 msg.msg_name=address;
1238 msg.msg_namelen=MAX_SOCK_ADDR;
1239 if (sock->file->f_flags & O_NONBLOCK)
1240 flags |= MSG_DONTWAIT;
1241 err=sock_recvmsg(sock, &msg, size, flags);
1242
1243 if(err >= 0 && addr != NULL && msg.msg_namelen)
1244 {
1245 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1246 if(err2<0)
1247 err=err2;
1248 }
1249 sockfd_put(sock);
1250 out:
1251 return err;
1252 }
1253
1254 /*
1255 * Receive a datagram from a socket.
1256 */
1257
1258 asmlinkage long sys_recv(int fd, void * ubuf, size_t size, unsigned flags)
1259 {
1260 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1261 }
1262
1263 /*
1264 * Set a socket option. Because we don't know the option lengths we have
1265 * to pass the user mode parameter for the protocols to sort out.
1266 */
1267
1268 asmlinkage long sys_setsockopt(int fd, int level, int optname, char *optval, int optlen)
1269 {
1270 int err;
1271 struct socket *sock;
1272
1273 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1274 {
1275 if (level == SOL_SOCKET)
1276 err=sock_setsockopt(sock,level,optname,optval,optlen);
1277 else
1278 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1279 sockfd_put(sock);
1280 }
1281 return err;
1282 }
1283
1284 /*
1285 * Get a socket option. Because we don't know the option lengths we have
1286 * to pass a user mode parameter for the protocols to sort out.
1287 */
1288
1289 asmlinkage long sys_getsockopt(int fd, int level, int optname, char *optval, int *optlen)
1290 {
1291 int err;
1292 struct socket *sock;
1293
1294 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1295 {
1296 if (level == SOL_SOCKET)
1297 err=sock_getsockopt(sock,level,optname,optval,optlen);
1298 else
1299 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1300 sockfd_put(sock);
1301 }
1302 return err;
1303 }
1304
1305
1306 /*
1307 * Shutdown a socket.
1308 */
1309
1310 asmlinkage long sys_shutdown(int fd, int how)
1311 {
1312 int err;
1313 struct socket *sock;
1314
1315 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1316 {
1317 err=sock->ops->shutdown(sock, how);
1318 sockfd_put(sock);
1319 }
1320 return err;
1321 }
1322
1323 /*
1324 * BSD sendmsg interface
1325 */
1326
1327 asmlinkage long sys_sendmsg(int fd, struct msghdr *msg, unsigned flags)
1328 {
1329 struct socket *sock;
1330 char address[MAX_SOCK_ADDR];
1331 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1332 unsigned char ctl[sizeof(struct cmsghdr) + 20]; /* 20 is size of ipv6_pktinfo */
1333 unsigned char *ctl_buf = ctl;
1334 struct msghdr msg_sys;
1335 int err, ctl_len, iov_size, total_len;
1336
1337 err = -EFAULT;
1338 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1339 goto out;
1340
1341 sock = sockfd_lookup(fd, &err);
1342 if (!sock)
1343 goto out;
1344
1345 /* do not move before msg_sys is valid */
1346 err = -EINVAL;
1347 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1348 goto out_put;
1349
1350 /* Check whether to allocate the iovec area*/
1351 err = -ENOMEM;
1352 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1353 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1354 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1355 if (!iov)
1356 goto out_put;
1357 }
1358
1359 /* This will also move the address data into kernel space */
1360 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1361 if (err < 0)
1362 goto out_freeiov;
1363 total_len = err;
1364
1365 err = -ENOBUFS;
1366
1367 if (msg_sys.msg_controllen > INT_MAX)
1368 goto out_freeiov;
1369 ctl_len = msg_sys.msg_controllen;
1370 if (ctl_len)
1371 {
1372 if (ctl_len > sizeof(ctl))
1373 {
1374 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1375 if (ctl_buf == NULL)
1376 goto out_freeiov;
1377 }
1378 err = -EFAULT;
1379 if (copy_from_user(ctl_buf, msg_sys.msg_control, ctl_len))
1380 goto out_freectl;
1381 msg_sys.msg_control = ctl_buf;
1382 }
1383 msg_sys.msg_flags = flags;
1384
1385 if (sock->file->f_flags & O_NONBLOCK)
1386 msg_sys.msg_flags |= MSG_DONTWAIT;
1387 err = sock_sendmsg(sock, &msg_sys, total_len);
1388
1389 out_freectl:
1390 if (ctl_buf != ctl)
1391 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1392 out_freeiov:
1393 if (iov != iovstack)
1394 sock_kfree_s(sock->sk, iov, iov_size);
1395 out_put:
1396 sockfd_put(sock);
1397 out:
1398 return err;
1399 }
1400
1401 /*
1402 * BSD recvmsg interface
1403 */
1404
1405 asmlinkage long sys_recvmsg(int fd, struct msghdr *msg, unsigned int flags)
1406 {
1407 struct socket *sock;
1408 struct iovec iovstack[UIO_FASTIOV];
1409 struct iovec *iov=iovstack;
1410 struct msghdr msg_sys;
1411 unsigned long cmsg_ptr;
1412 int err, iov_size, total_len, len;
1413
1414 /* kernel mode address */
1415 char addr[MAX_SOCK_ADDR];
1416
1417 /* user mode address pointers */
1418 struct sockaddr *uaddr;
1419 int *uaddr_len;
1420
1421 err=-EFAULT;
1422 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1423 goto out;
1424
1425 sock = sockfd_lookup(fd, &err);
1426 if (!sock)
1427 goto out;
1428
1429 err = -EINVAL;
1430 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1431 goto out_put;
1432
1433 /* Check whether to allocate the iovec area*/
1434 err = -ENOMEM;
1435 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1436 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1437 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1438 if (!iov)
1439 goto out_put;
1440 }
1441
1442 /*
1443 * Save the user-mode address (verify_iovec will change the
1444 * kernel msghdr to use the kernel address space)
1445 */
1446
1447 uaddr = msg_sys.msg_name;
1448 uaddr_len = &msg->msg_namelen;
1449 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1450 if (err < 0)
1451 goto out_freeiov;
1452 total_len=err;
1453
1454 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1455 msg_sys.msg_flags = 0;
1456
1457 if (sock->file->f_flags & O_NONBLOCK)
1458 flags |= MSG_DONTWAIT;
1459 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1460 if (err < 0)
1461 goto out_freeiov;
1462 len = err;
1463
1464 if (uaddr != NULL && msg_sys.msg_namelen) {
1465 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1466 if (err < 0)
1467 goto out_freeiov;
1468 }
1469 err = __put_user(msg_sys.msg_flags, &msg->msg_flags);
1470 if (err)
1471 goto out_freeiov;
1472 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1473 &msg->msg_controllen);
1474 if (err)
1475 goto out_freeiov;
1476 err = len;
1477
1478 out_freeiov:
1479 if (iov != iovstack)
1480 sock_kfree_s(sock->sk, iov, iov_size);
1481 out_put:
1482 sockfd_put(sock);
1483 out:
1484 return err;
1485 }
1486
1487
1488 /*
1489 * Perform a file control on a socket file descriptor.
1490 *
1491 * Doesn't acquire a fd lock, because no network fcntl
1492 * function sleeps currently.
1493 */
1494
1495 int sock_fcntl(struct file *filp, unsigned int cmd, unsigned long arg)
1496 {
1497 struct socket *sock;
1498
1499 sock = socki_lookup (filp->f_dentry->d_inode);
1500 if (sock && sock->ops)
1501 return sock_no_fcntl(sock, cmd, arg);
1502 return(-EINVAL);
1503 }
1504
1505 /* Argument list sizes for sys_socketcall */
1506 #define AL(x) ((x) * sizeof(unsigned long))
1507 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1508 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1509 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1510 #undef AL
1511
1512 /*
1513 * System call vectors.
1514 *
1515 * Argument checking cleaned up. Saved 20% in size.
1516 * This function doesn't need to set the kernel lock because
1517 * it is set by the callees.
1518 */
1519
1520 asmlinkage long sys_socketcall(int call, unsigned long *args)
1521 {
1522 unsigned long a[6];
1523 unsigned long a0,a1;
1524 int err;
1525
1526 if(call<1||call>SYS_RECVMSG)
1527 return -EINVAL;
1528
1529 /* copy_from_user should be SMP safe. */
1530 if (copy_from_user(a, args, nargs[call]))
1531 return -EFAULT;
1532
1533 a0=a[0];
1534 a1=a[1];
1535
1536 switch(call)
1537 {
1538 case SYS_SOCKET:
1539 err = sys_socket(a0,a1,a[2]);
1540 break;
1541 case SYS_BIND:
1542 err = sys_bind(a0,(struct sockaddr *)a1, a[2]);
1543 break;
1544 case SYS_CONNECT:
1545 err = sys_connect(a0, (struct sockaddr *)a1, a[2]);
1546 break;
1547 case SYS_LISTEN:
1548 err = sys_listen(a0,a1);
1549 break;
1550 case SYS_ACCEPT:
1551 err = sys_accept(a0,(struct sockaddr *)a1, (int *)a[2]);
1552 break;
1553 case SYS_GETSOCKNAME:
1554 err = sys_getsockname(a0,(struct sockaddr *)a1, (int *)a[2]);
1555 break;
1556 case SYS_GETPEERNAME:
1557 err = sys_getpeername(a0, (struct sockaddr *)a1, (int *)a[2]);
1558 break;
1559 case SYS_SOCKETPAIR:
1560 err = sys_socketpair(a0,a1, a[2], (int *)a[3]);
1561 break;
1562 case SYS_SEND:
1563 err = sys_send(a0, (void *)a1, a[2], a[3]);
1564 break;
1565 case SYS_SENDTO:
1566 err = sys_sendto(a0,(void *)a1, a[2], a[3],
1567 (struct sockaddr *)a[4], a[5]);
1568 break;
1569 case SYS_RECV:
1570 err = sys_recv(a0, (void *)a1, a[2], a[3]);
1571 break;
1572 case SYS_RECVFROM:
1573 err = sys_recvfrom(a0, (void *)a1, a[2], a[3],
1574 (struct sockaddr *)a[4], (int *)a[5]);
1575 break;
1576 case SYS_SHUTDOWN:
1577 err = sys_shutdown(a0,a1);
1578 break;
1579 case SYS_SETSOCKOPT:
1580 err = sys_setsockopt(a0, a1, a[2], (char *)a[3], a[4]);
1581 break;
1582 case SYS_GETSOCKOPT:
1583 err = sys_getsockopt(a0, a1, a[2], (char *)a[3], (int *)a[4]);
1584 break;
1585 case SYS_SENDMSG:
1586 err = sys_sendmsg(a0, (struct msghdr *) a1, a[2]);
1587 break;
1588 case SYS_RECVMSG:
1589 err = sys_recvmsg(a0, (struct msghdr *) a1, a[2]);
1590 break;
1591 default:
1592 err = -EINVAL;
1593 break;
1594 }
1595 return err;
1596 }
1597
1598 /*
1599 * This function is called by a protocol handler that wants to
1600 * advertise its address family, and have it linked into the
1601 * SOCKET module.
1602 */
1603
1604 int sock_register(struct net_proto_family *ops)
1605 {
1606 int err;
1607
1608 if (ops->family >= NPROTO) {
1609 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
1610 return -ENOBUFS;
1611 }
1612 net_family_write_lock();
1613 err = -EEXIST;
1614 if (net_families[ops->family] == NULL) {
1615 net_families[ops->family]=ops;
1616 err = 0;
1617 }
1618 net_family_write_unlock();
1619 return err;
1620 }
1621
1622 /*
1623 * This function is called by a protocol handler that wants to
1624 * remove its address family, and have it unlinked from the
1625 * SOCKET module.
1626 */
1627
1628 int sock_unregister(int family)
1629 {
1630 if (family < 0 || family >= NPROTO)
1631 return -1;
1632
1633 net_family_write_lock();
1634 net_families[family]=NULL;
1635 net_family_write_unlock();
1636 return 0;
1637 }
1638
1639
1640 extern void sk_init(void);
1641
1642 #ifdef CONFIG_WAN_ROUTER
1643 extern void wanrouter_init(void);
1644 #endif
1645
1646 void __init sock_init(void)
1647 {
1648 int i;
1649
1650 printk(KERN_INFO "Linux NET4.0 for Linux 2.4\n");
1651 printk(KERN_INFO "Based upon Swansea University Computer Society NET3.039\n");
1652
1653 /*
1654 * Initialize all address (protocol) families.
1655 */
1656
1657 for (i = 0; i < NPROTO; i++)
1658 net_families[i] = NULL;
1659
1660 /*
1661 * Initialize sock SLAB cache.
1662 */
1663
1664 sk_init();
1665
1666 #ifdef SLAB_SKB
1667 /*
1668 * Initialize skbuff SLAB cache
1669 */
1670 skb_init();
1671 #endif
1672
1673 /*
1674 * Wan router layer.
1675 */
1676
1677 #ifdef CONFIG_WAN_ROUTER
1678 wanrouter_init();
1679 #endif
1680
1681 /*
1682 * Initialize the protocols module.
1683 */
1684
1685 register_filesystem(&sock_fs_type);
1686 sock_mnt = kern_mount(&sock_fs_type);
1687 /* The real protocol initialization is performed when
1688 * do_initcalls is run.
1689 */
1690
1691
1692 /*
1693 * The netlink device handler may be needed early.
1694 */
1695
1696 #ifdef CONFIG_RTNETLINK
1697 rtnetlink_init();
1698 #endif
1699 #ifdef CONFIG_NETLINK_DEV
1700 init_netlink();
1701 #endif
1702 #ifdef CONFIG_NETFILTER
1703 netfilter_init();
1704 #endif
1705 }
1706
1707 int socket_get_info(char *buffer, char **start, off_t offset, int length)
1708 {
1709 int len, cpu;
1710 int counter = 0;
1711
1712 for (cpu=0; cpu<smp_num_cpus; cpu++)
1713 counter += sockets_in_use[cpu_logical_map(cpu)].counter;
1714
1715 /* It can be negative, by the way. 8) */
1716 if (counter < 0)
1717 counter = 0;
1718
1719 len = sprintf(buffer, "sockets: used %d\n", counter);
1720 if (offset >= len)
1721 {
1722 *start = buffer;
1723 return 0;
1724 }
1725 *start = buffer + offset;
1726 len -= offset;
1727 if (len > length)
1728 len = length;
1729 if (len < 0)
1730 len = 0;
1731 return len;
1732 }
1733
This page was automatically generated by the
LXR engine.
Visit the LXR main site for more
information.