2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro, <bir7@leland.Stanford.Edu>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
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
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
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
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
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
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
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.
55 * This module is effectively the top level interface to the BSD socket
60 #include <linux/config.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>
74 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
75 #include <linux/kmod.h>
78 #include <asm/uaccess.h>
80 #include <linux/inet.h>
86 #include <linux/netfilter.h>
88 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
89 static loff_t
sock_lseek(struct file
*file
, loff_t offset
, int whence
);
90 static ssize_t
sock_read(struct file
*file
, char *buf
,
91 size_t size
, loff_t
*ppos
);
92 static ssize_t
sock_write(struct file
*file
, const char *buf
,
93 size_t size
, loff_t
*ppos
);
94 static int sock_mmap(struct file
*file
, struct vm_area_struct
* vma
);
96 static int sock_close(struct inode
*inode
, struct file
*file
);
97 static unsigned int sock_poll(struct file
*file
,
98 struct poll_table_struct
*wait
);
99 static int sock_ioctl(struct inode
*inode
, struct file
*file
,
100 unsigned int cmd
, unsigned long arg
);
101 static int sock_fasync(int fd
, struct file
*filp
, int on
);
102 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
103 unsigned long count
, loff_t
*ppos
);
104 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
105 unsigned long count
, loff_t
*ppos
);
109 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
110 * in the operation structures but are done directly via the socketcall() multiplexor.
113 static struct file_operations socket_file_ops
= {
120 open
: sock_no_open
, /* special open code to disallow open via /proc */
128 * The protocol list. Each protocol is registered in here.
131 static struct net_proto_family
*net_families
[NPROTO
];
134 static atomic_t net_family_lockct
= ATOMIC_INIT(0);
135 static spinlock_t net_family_lock
= SPIN_LOCK_UNLOCKED
;
137 /* The strategy is: modifications net_family vector are short, do not
138 sleep and veeery rare, but read access should be free of any exclusive
142 static void net_family_write_lock(void)
144 spin_lock(&net_family_lock
);
145 while (atomic_read(&net_family_lockct
) != 0) {
146 spin_unlock(&net_family_lock
);
148 current
->policy
|= SCHED_YIELD
;
151 spin_lock(&net_family_lock
);
155 static __inline__
void net_family_write_unlock(void)
157 spin_unlock(&net_family_lock
);
160 static __inline__
void net_family_read_lock(void)
162 atomic_inc(&net_family_lockct
);
163 spin_unlock_wait(&net_family_lock
);
166 static __inline__
void net_family_read_unlock(void)
168 atomic_dec(&net_family_lockct
);
172 #define net_family_write_lock() do { } while(0)
173 #define net_family_write_unlock() do { } while(0)
174 #define net_family_read_lock() do { } while(0)
175 #define net_family_read_unlock() do { } while(0)
180 * Statistics counters of the socket lists
185 char __pad
[SMP_CACHE_BYTES
];
186 } sockets_in_use
[NR_CPUS
] __cacheline_aligned
= {{0}};
189 * Support routines. Move socket addresses back and forth across the kernel/user
190 * divide and look after the messy bits.
193 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
194 16 for IP, 16 for IPX,
197 must be at least one bigger than
198 the AF_UNIX size (see net/unix/af_unix.c
203 * move_addr_to_kernel - copy a socket address into kernel space
204 * @uaddr: Address in user space
205 * @kaddr: Address in kernel space
206 * @ulen: Length in user space
208 * The address is copied into kernel space. If the provided address is
209 * too long an error code of -EINVAL is returned. If the copy gives
210 * invalid addresses -EFAULT is returned. On a success 0 is returned.
213 int move_addr_to_kernel(void *uaddr
, int ulen
, void *kaddr
)
215 if(ulen
<0||ulen
>MAX_SOCK_ADDR
)
219 if(copy_from_user(kaddr
,uaddr
,ulen
))
225 * move_addr_to_user - copy an address to user space
226 * @kaddr: kernel space address
227 * @klen: length of address in kernel
228 * @uaddr: user space address
229 * @ulen: pointer to user length field
231 * The value pointed to by ulen on entry is the buffer length available.
232 * This is overwritten with the buffer space used. -EINVAL is returned
233 * if an overlong buffer is specified or a negative buffer size. -EFAULT
234 * is returned if either the buffer or the length field are not
236 * After copying the data up to the limit the user specifies, the true
237 * length of the data is written over the length limit the user
238 * specified. Zero is returned for a success.
241 int move_addr_to_user(void *kaddr
, int klen
, void *uaddr
, int *ulen
)
246 if((err
=get_user(len
, ulen
)))
250 if(len
<0 || len
> MAX_SOCK_ADDR
)
254 if(copy_to_user(uaddr
,kaddr
,len
))
258 * "fromlen shall refer to the value before truncation.."
261 return __put_user(klen
, ulen
);
265 * Obtains the first available file descriptor and sets it up for use.
267 * This functions creates file structure and maps it to fd space
268 * of current process. On success it returns file descriptor
269 * and file struct implicitly stored in sock->file.
270 * Note that another thread may close file descriptor before we return
271 * from this function. We use the fact that now we do not refer
272 * to socket after mapping. If one day we will need it, this
273 * function will inincrement ref. count on file by 1.
275 * In any case returned fd MAY BE not valid!
276 * This race condition is inavoidable
277 * with shared fd spaces, we cannot solve is inside kernel,
278 * but we take care of internal coherence yet.
281 static int sock_map_fd(struct socket
*sock
)
286 * Find a file descriptor suitable for return to the user.
289 fd
= get_unused_fd();
291 struct file
*file
= get_empty_filp();
299 file
->f_dentry
= d_alloc_root(sock
->inode
);
300 /* MOUNT_REWRITE: set to sockfs internal vfsmnt */
301 file
->f_vfsmnt
= NULL
;
302 if (!file
->f_dentry
) {
310 file
->f_op
= &socket_file_ops
;
312 file
->f_flags
= O_RDWR
;
314 fd_install(fd
, file
);
321 extern __inline__
struct socket
*socki_lookup(struct inode
*inode
)
323 return &inode
->u
.socket_i
;
327 * sockfd_lookup - Go from a file number to its socket slot
329 * @err: pointer to an error code return
331 * The file handle passed in is locked and the socket it is bound
332 * too is returned. If an error occurs the err pointer is overwritten
333 * with a negative errno code and NULL is returned. The function checks
334 * for both invalid handles and passing a handle which is not a socket.
336 * On a success the socket object pointer is returned.
339 struct socket
*sockfd_lookup(int fd
, int *err
)
345 if (!(file
= fget(fd
)))
351 inode
= file
->f_dentry
->d_inode
;
352 if (!inode
|| !inode
->i_sock
|| !(sock
= socki_lookup(inode
)))
359 if (sock
->file
!= file
) {
360 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
366 extern __inline__
void sockfd_put(struct socket
*sock
)
372 * sock_alloc - allocate a socket
374 * Allocate a new inode and socket object. The two are bound together
375 * and initialised. The socket is then returned. If we are out of inodes
379 struct socket
*sock_alloc(void)
381 struct inode
* inode
;
382 struct socket
* sock
;
384 inode
= get_empty_inode();
388 sock
= socki_lookup(inode
);
390 inode
->i_mode
= S_IFSOCK
|S_IRWXUGO
;
392 inode
->i_uid
= current
->fsuid
;
393 inode
->i_gid
= current
->fsgid
;
396 init_waitqueue_head(&sock
->wait
);
397 sock
->fasync_list
= NULL
;
398 sock
->state
= SS_UNCONNECTED
;
404 sockets_in_use
[smp_processor_id()].counter
++;
409 * In theory you can't get an open on this inode, but /proc provides
410 * a back door. Remember to keep it shut otherwise you'll let the
411 * creepy crawlies in.
414 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
420 * sock_release - close a socket
421 * @sock: socket to close
423 * The socket is released from the protocol stack if it has a release
424 * callback, and the inode is then released if the socket is bound to
425 * an inode not a file.
428 void sock_release(struct socket
*sock
)
431 sock
->ops
->release(sock
);
433 if (sock
->fasync_list
)
434 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
436 sockets_in_use
[smp_processor_id()].counter
--;
444 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, int size
)
447 struct scm_cookie scm
;
449 err
= scm_send(sock
, msg
, &scm
);
451 err
= sock
->ops
->sendmsg(sock
, msg
, size
, &scm
);
457 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int size
, int flags
)
459 struct scm_cookie scm
;
461 memset(&scm
, 0, sizeof(scm
));
463 size
= sock
->ops
->recvmsg(sock
, msg
, size
, flags
, &scm
);
465 scm_recv(sock
, msg
, &scm
, flags
);
472 * Sockets are not seekable.
475 static loff_t
sock_lseek(struct file
*file
, loff_t offset
, int whence
)
481 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
482 * area ubuf...ubuf+size-1 is writable before asking the protocol.
485 static ssize_t
sock_read(struct file
*file
, char *ubuf
,
486 size_t size
, loff_t
*ppos
)
493 if (ppos
!= &file
->f_pos
)
495 if (size
==0) /* Match SYS5 behaviour */
498 sock
= socki_lookup(file
->f_dentry
->d_inode
);
504 msg
.msg_control
=NULL
;
505 msg
.msg_controllen
=0;
508 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
510 return sock_recvmsg(sock
, &msg
, size
, flags
);
515 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
516 * is readable by the user process.
519 static ssize_t
sock_write(struct file
*file
, const char *ubuf
,
520 size_t size
, loff_t
*ppos
)
526 if (ppos
!= &file
->f_pos
)
528 if(size
==0) /* Match SYS5 behaviour */
531 sock
= socki_lookup(file
->f_dentry
->d_inode
);
537 msg
.msg_control
=NULL
;
538 msg
.msg_controllen
=0;
539 msg
.msg_flags
=!(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
540 if (sock
->type
== SOCK_SEQPACKET
)
541 msg
.msg_flags
|= MSG_EOR
;
542 iov
.iov_base
=(void *)ubuf
;
545 return sock_sendmsg(sock
, &msg
, size
);
548 int sock_readv_writev(int type
, struct inode
* inode
, struct file
* file
,
549 const struct iovec
* iov
, long count
, long size
)
554 sock
= socki_lookup(inode
);
558 msg
.msg_control
= NULL
;
559 msg
.msg_controllen
= 0;
560 msg
.msg_iov
= (struct iovec
*) iov
;
561 msg
.msg_iovlen
= count
;
562 msg
.msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
564 /* read() does a VERIFY_WRITE */
565 if (type
== VERIFY_WRITE
)
566 return sock_recvmsg(sock
, &msg
, size
, msg
.msg_flags
);
568 if (sock
->type
== SOCK_SEQPACKET
)
569 msg
.msg_flags
|= MSG_EOR
;
571 return sock_sendmsg(sock
, &msg
, size
);
574 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
575 unsigned long count
, loff_t
*ppos
)
579 for (i
= 0 ; i
< count
; i
++)
580 tot_len
+= vector
[i
].iov_len
;
581 return sock_readv_writev(VERIFY_WRITE
, file
->f_dentry
->d_inode
,
582 file
, vector
, count
, tot_len
);
585 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
586 unsigned long count
, loff_t
*ppos
)
590 for (i
= 0 ; i
< count
; i
++)
591 tot_len
+= vector
[i
].iov_len
;
592 return sock_readv_writev(VERIFY_READ
, file
->f_dentry
->d_inode
,
593 file
, vector
, count
, tot_len
);
597 * With an ioctl arg may well be a user mode pointer, but we don't know what to do
598 * with it - that's up to the protocol still.
601 int sock_ioctl(struct inode
*inode
, struct file
*file
, unsigned int cmd
,
608 sock
= socki_lookup(inode
);
609 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
616 /* No kernel lock held - perfect */
617 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
622 * We can't return errors to poll, so it's either yes or no.
624 sock
= socki_lookup(file
->f_dentry
->d_inode
);
625 return sock
->ops
->poll(file
, sock
, wait
);
628 static int sock_mmap(struct file
* file
, struct vm_area_struct
* vma
)
630 struct socket
*sock
= socki_lookup(file
->f_dentry
->d_inode
);
632 return sock
->ops
->mmap(file
, sock
, vma
);
635 int sock_close(struct inode
*inode
, struct file
*filp
)
638 * It was possible the inode is NULL we were
639 * closing an unfinished socket.
644 printk(KERN_DEBUG
"sock_close: NULL inode\n");
648 sock_fasync(-1, filp
, 0);
649 sock_release(socki_lookup(inode
));
655 * Update the socket async list
657 * Fasync_list locking strategy.
659 * 1. fasync_list is modified only under process context socket lock
660 * i.e. under semaphore.
661 * 2. fasync_list is used under read_lock(&sk->callback_lock)
662 * or under socket lock.
663 * 3. fasync_list can be used from softirq context, so that
664 * modification under socket lock have to be enhanced with
665 * write_lock_bh(&sk->callback_lock).
669 static int sock_fasync(int fd
, struct file
*filp
, int on
)
671 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
677 fna
=(struct fasync_struct
*)kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
683 sock
= socki_lookup(filp
->f_dentry
->d_inode
);
685 if ((sk
=sock
->sk
) == NULL
)
690 prev
=&(sock
->fasync_list
);
692 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
693 if (fa
->fa_file
==filp
)
700 write_lock_bh(&sk
->callback_lock
);
702 write_unlock_bh(&sk
->callback_lock
);
704 kfree_s(fna
,sizeof(struct fasync_struct
));
709 fna
->magic
=FASYNC_MAGIC
;
710 fna
->fa_next
=sock
->fasync_list
;
711 write_lock_bh(&sk
->callback_lock
);
712 sock
->fasync_list
=fna
;
713 write_unlock_bh(&sk
->callback_lock
);
719 write_lock_bh(&sk
->callback_lock
);
721 write_unlock_bh(&sk
->callback_lock
);
722 kfree_s(fa
,sizeof(struct fasync_struct
));
727 release_sock(sock
->sk
);
731 /* This function may be called only under socket lock or callback_lock */
733 int sock_wake_async(struct socket
*sock
, int how
, int band
)
735 if (!sock
|| !sock
->fasync_list
)
741 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
745 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
750 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
753 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
759 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
765 * Check protocol is in range
767 if(family
<0 || family
>=NPROTO
)
772 This uglymoron is moved from INET layer to here to avoid
773 deadlock in module load.
775 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
779 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n", current
->comm
);
784 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
785 /* Attempt to load a protocol module if the find failed.
787 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
788 * requested real, full-featured networking support upon configuration.
789 * Otherwise module support will break!
791 if (net_families
[family
]==NULL
)
793 char module_name
[30];
794 sprintf(module_name
,"net-pf-%d",family
);
795 request_module(module_name
);
799 net_family_read_lock();
800 if (net_families
[family
] == NULL
) {
806 * Allocate the socket and allow the family to set things up. if
807 * the protocol is 0, the family is instructed to select an appropriate
811 if (!(sock
= sock_alloc()))
813 printk(KERN_WARNING
"socket: no more sockets\n");
814 i
= -ENFILE
; /* Not exactly a match, but its the
815 closest posix thing */
821 if ((i
= net_families
[family
]->create(sock
, protocol
)) < 0)
830 net_family_read_unlock();
834 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
839 retval
= sock_create(family
, type
, protocol
, &sock
);
843 retval
= sock_map_fd(sock
);
848 /* It may be already another descriptor 8) Not kernel problem. */
857 * Create a pair of connected sockets.
860 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
, int usockvec
[2])
862 struct socket
*sock1
, *sock2
;
866 * Obtain the first socket and check if the underlying protocol
867 * supports the socketpair call.
870 err
= sock_create(family
, type
, protocol
, &sock1
);
874 err
= sock_create(family
, type
, protocol
, &sock2
);
878 err
= sock1
->ops
->socketpair(sock1
, sock2
);
880 goto out_release_both
;
884 err
= sock_map_fd(sock1
);
886 goto out_release_both
;
889 err
= sock_map_fd(sock2
);
894 /* fd1 and fd2 may be already another descriptors.
895 * Not kernel problem.
898 err
= put_user(fd1
, &usockvec
[0]);
900 err
= put_user(fd2
, &usockvec
[1]);
923 * Bind a name to a socket. Nothing much to do here since it's
924 * the protocol's responsibility to handle the local address.
926 * We move the socket address to kernel space before we call
927 * the protocol layer (having also checked the address is ok).
930 asmlinkage
long sys_bind(int fd
, struct sockaddr
*umyaddr
, int addrlen
)
933 char address
[MAX_SOCK_ADDR
];
936 if((sock
= sockfd_lookup(fd
,&err
))!=NULL
)
938 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0)
939 err
= sock
->ops
->bind(sock
, (struct sockaddr
*)address
, addrlen
);
947 * Perform a listen. Basically, we allow the protocol to do anything
948 * necessary for a listen, and if that works, we mark the socket as
949 * ready for listening.
952 asmlinkage
long sys_listen(int fd
, int backlog
)
957 if ((sock
= sockfd_lookup(fd
, &err
)) != NULL
) {
958 if ((unsigned) backlog
> SOMAXCONN
)
960 err
=sock
->ops
->listen(sock
, backlog
);
968 * For accept, we attempt to create a new socket, set up the link
969 * with the client, wake up the client, then return the new
970 * connected fd. We collect the address of the connector in kernel
971 * space and move it to user at the very end. This is unclean because
972 * we open the socket then return an error.
974 * 1003.1g adds the ability to recvmsg() to query connection pending
975 * status to recvmsg. We need to add that support in a way thats
976 * clean when we restucture accept also.
979 asmlinkage
long sys_accept(int fd
, struct sockaddr
*upeer_sockaddr
, int *upeer_addrlen
)
981 struct socket
*sock
, *newsock
;
983 char address
[MAX_SOCK_ADDR
];
985 sock
= sockfd_lookup(fd
, &err
);
990 if (!(newsock
= sock_alloc()))
993 newsock
->type
= sock
->type
;
994 newsock
->ops
= sock
->ops
;
996 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1000 if (upeer_sockaddr
) {
1001 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 2)<0) {
1002 err
= -ECONNABORTED
;
1005 err
= move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
1010 /* File flags are not inherited via accept() unlike another OSes. */
1012 if ((err
= sock_map_fd(newsock
)) < 0)
1021 sock_release(newsock
);
1027 * Attempt to connect to a socket with the server address. The address
1028 * is in user space so we verify it is OK and move it to kernel space.
1030 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1033 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1034 * other SEQPACKET protocols that take time to connect() as it doesn't
1035 * include the -EINPROGRESS status for such sockets.
1038 asmlinkage
long sys_connect(int fd
, struct sockaddr
*uservaddr
, int addrlen
)
1040 struct socket
*sock
;
1041 char address
[MAX_SOCK_ADDR
];
1044 sock
= sockfd_lookup(fd
, &err
);
1047 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1050 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
1051 sock
->file
->f_flags
);
1059 * Get the local address ('name') of a socket object. Move the obtained
1060 * name to user space.
1063 asmlinkage
long sys_getsockname(int fd
, struct sockaddr
*usockaddr
, int *usockaddr_len
)
1065 struct socket
*sock
;
1066 char address
[MAX_SOCK_ADDR
];
1069 sock
= sockfd_lookup(fd
, &err
);
1072 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1075 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1084 * Get the remote address ('name') of a socket object. Move the obtained
1085 * name to user space.
1088 asmlinkage
long sys_getpeername(int fd
, struct sockaddr
*usockaddr
, int *usockaddr_len
)
1090 struct socket
*sock
;
1091 char address
[MAX_SOCK_ADDR
];
1094 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1096 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
1098 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
1105 * Send a datagram to a given address. We move the address into kernel
1106 * space and check the user space data area is readable before invoking
1110 asmlinkage
long sys_sendto(int fd
, void * buff
, size_t len
, unsigned flags
,
1111 struct sockaddr
*addr
, int addr_len
)
1113 struct socket
*sock
;
1114 char address
[MAX_SOCK_ADDR
];
1119 sock
= sockfd_lookup(fd
, &err
);
1127 msg
.msg_control
=NULL
;
1128 msg
.msg_controllen
=0;
1129 msg
.msg_namelen
=addr_len
;
1132 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1135 msg
.msg_name
=address
;
1137 if (sock
->file
->f_flags
& O_NONBLOCK
)
1138 flags
|= MSG_DONTWAIT
;
1139 msg
.msg_flags
= flags
;
1140 err
= sock_sendmsg(sock
, &msg
, len
);
1149 * Send a datagram down a socket.
1152 asmlinkage
long sys_send(int fd
, void * buff
, size_t len
, unsigned flags
)
1154 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1158 * Receive a frame from the socket and optionally record the address of the
1159 * sender. We verify the buffers are writable and if needed move the
1160 * sender address from kernel to user space.
1163 asmlinkage
long sys_recvfrom(int fd
, void * ubuf
, size_t size
, unsigned flags
,
1164 struct sockaddr
*addr
, int *addr_len
)
1166 struct socket
*sock
;
1169 char address
[MAX_SOCK_ADDR
];
1172 sock
= sockfd_lookup(fd
, &err
);
1176 msg
.msg_control
=NULL
;
1177 msg
.msg_controllen
=0;
1182 msg
.msg_name
=address
;
1183 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1184 if (sock
->file
->f_flags
& O_NONBLOCK
)
1185 flags
|= MSG_DONTWAIT
;
1186 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1188 if(err
>= 0 && addr
!= NULL
&& msg
.msg_namelen
)
1190 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1200 * Receive a datagram from a socket.
1203 asmlinkage
long sys_recv(int fd
, void * ubuf
, size_t size
, unsigned flags
)
1205 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1209 * Set a socket option. Because we don't know the option lengths we have
1210 * to pass the user mode parameter for the protocols to sort out.
1213 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
, char *optval
, int optlen
)
1216 struct socket
*sock
;
1218 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1220 if (level
== SOL_SOCKET
)
1221 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1223 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1230 * Get a socket option. Because we don't know the option lengths we have
1231 * to pass a user mode parameter for the protocols to sort out.
1234 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
, char *optval
, int *optlen
)
1237 struct socket
*sock
;
1239 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1241 if (level
== SOL_SOCKET
)
1242 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1244 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1252 * Shutdown a socket.
1255 asmlinkage
long sys_shutdown(int fd
, int how
)
1258 struct socket
*sock
;
1260 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1262 err
=sock
->ops
->shutdown(sock
, how
);
1269 * BSD sendmsg interface
1272 asmlinkage
long sys_sendmsg(int fd
, struct msghdr
*msg
, unsigned flags
)
1274 struct socket
*sock
;
1275 char address
[MAX_SOCK_ADDR
];
1276 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1277 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]; /* 20 is size of ipv6_pktinfo */
1278 unsigned char *ctl_buf
= ctl
;
1279 struct msghdr msg_sys
;
1280 int err
, ctl_len
, iov_size
, total_len
;
1283 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1286 sock
= sockfd_lookup(fd
, &err
);
1290 /* do not move before msg_sys is valid */
1292 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1295 /* Check whether to allocate the iovec area*/
1297 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1298 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1299 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1304 /* This will also move the address data into kernel space */
1305 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1312 if (msg_sys
.msg_controllen
> INT_MAX
)
1314 ctl_len
= msg_sys
.msg_controllen
;
1317 if (ctl_len
> sizeof(ctl
))
1320 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1321 if (ctl_buf
== NULL
)
1325 if (copy_from_user(ctl_buf
, msg_sys
.msg_control
, ctl_len
))
1327 msg_sys
.msg_control
= ctl_buf
;
1329 msg_sys
.msg_flags
= flags
;
1331 if (sock
->file
->f_flags
& O_NONBLOCK
)
1332 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1333 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1337 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1339 if (iov
!= iovstack
)
1340 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1348 * BSD recvmsg interface
1351 asmlinkage
long sys_recvmsg(int fd
, struct msghdr
*msg
, unsigned int flags
)
1353 struct socket
*sock
;
1354 struct iovec iovstack
[UIO_FASTIOV
];
1355 struct iovec
*iov
=iovstack
;
1356 struct msghdr msg_sys
;
1357 unsigned long cmsg_ptr
;
1358 int err
, iov_size
, total_len
, len
;
1360 /* kernel mode address */
1361 char addr
[MAX_SOCK_ADDR
];
1363 /* user mode address pointers */
1364 struct sockaddr
*uaddr
;
1368 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1371 sock
= sockfd_lookup(fd
, &err
);
1376 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1379 /* Check whether to allocate the iovec area*/
1381 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1382 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1383 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1389 * Save the user-mode address (verify_iovec will change the
1390 * kernel msghdr to use the kernel address space)
1393 uaddr
= msg_sys
.msg_name
;
1394 uaddr_len
= &msg
->msg_namelen
;
1395 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1400 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1401 msg_sys
.msg_flags
= 0;
1403 if (sock
->file
->f_flags
& O_NONBLOCK
)
1404 flags
|= MSG_DONTWAIT
;
1405 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1410 if (uaddr
!= NULL
&& msg_sys
.msg_namelen
) {
1411 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1415 err
= __put_user(msg_sys
.msg_flags
, &msg
->msg_flags
);
1418 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1419 &msg
->msg_controllen
);
1425 if (iov
!= iovstack
)
1426 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1435 * Perform a file control on a socket file descriptor.
1437 * Doesn't aquire a fd lock, because no network fcntl
1438 * function sleeps currently.
1441 int sock_fcntl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
1443 struct socket
*sock
;
1445 sock
= socki_lookup (filp
->f_dentry
->d_inode
);
1446 if (sock
&& sock
->ops
)
1447 return sock_no_fcntl(sock
, cmd
, arg
);
1451 /* Argument list sizes for sys_socketcall */
1452 #define AL(x) ((x) * sizeof(unsigned long))
1453 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1454 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1455 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1459 * System call vectors.
1461 * Argument checking cleaned up. Saved 20% in size.
1462 * This function doesn't need to set the kernel lock because
1463 * it is set by the callees.
1466 asmlinkage
long sys_socketcall(int call
, unsigned long *args
)
1469 unsigned long a0
,a1
;
1472 if(call
<1||call
>SYS_RECVMSG
)
1475 /* copy_from_user should be SMP safe. */
1476 if (copy_from_user(a
, args
, nargs
[call
]))
1485 err
= sys_socket(a0
,a1
,a
[2]);
1488 err
= sys_bind(a0
,(struct sockaddr
*)a1
, a
[2]);
1491 err
= sys_connect(a0
, (struct sockaddr
*)a1
, a
[2]);
1494 err
= sys_listen(a0
,a1
);
1497 err
= sys_accept(a0
,(struct sockaddr
*)a1
, (int *)a
[2]);
1499 case SYS_GETSOCKNAME
:
1500 err
= sys_getsockname(a0
,(struct sockaddr
*)a1
, (int *)a
[2]);
1502 case SYS_GETPEERNAME
:
1503 err
= sys_getpeername(a0
, (struct sockaddr
*)a1
, (int *)a
[2]);
1505 case SYS_SOCKETPAIR
:
1506 err
= sys_socketpair(a0
,a1
, a
[2], (int *)a
[3]);
1509 err
= sys_send(a0
, (void *)a1
, a
[2], a
[3]);
1512 err
= sys_sendto(a0
,(void *)a1
, a
[2], a
[3],
1513 (struct sockaddr
*)a
[4], a
[5]);
1516 err
= sys_recv(a0
, (void *)a1
, a
[2], a
[3]);
1519 err
= sys_recvfrom(a0
, (void *)a1
, a
[2], a
[3],
1520 (struct sockaddr
*)a
[4], (int *)a
[5]);
1523 err
= sys_shutdown(a0
,a1
);
1525 case SYS_SETSOCKOPT
:
1526 err
= sys_setsockopt(a0
, a1
, a
[2], (char *)a
[3], a
[4]);
1528 case SYS_GETSOCKOPT
:
1529 err
= sys_getsockopt(a0
, a1
, a
[2], (char *)a
[3], (int *)a
[4]);
1532 err
= sys_sendmsg(a0
, (struct msghdr
*) a1
, a
[2]);
1535 err
= sys_recvmsg(a0
, (struct msghdr
*) a1
, a
[2]);
1545 * This function is called by a protocol handler that wants to
1546 * advertise its address family, and have it linked into the
1550 int sock_register(struct net_proto_family
*ops
)
1554 if (ops
->family
>= NPROTO
) {
1555 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
1558 net_family_write_lock();
1560 if (net_families
[ops
->family
] == NULL
) {
1561 net_families
[ops
->family
]=ops
;
1564 net_family_write_unlock();
1569 * This function is called by a protocol handler that wants to
1570 * remove its address family, and have it unlinked from the
1574 int sock_unregister(int family
)
1576 if (family
< 0 || family
>= NPROTO
)
1579 net_family_write_lock();
1580 net_families
[family
]=NULL
;
1581 net_family_write_unlock();
1585 void __init
proto_init(void)
1587 extern struct net_proto protocols
[]; /* Network protocols */
1588 struct net_proto
*pro
;
1590 /* Kick all configured protocols. */
1592 while (pro
->name
!= NULL
)
1594 (*pro
->init_func
)(pro
);
1597 /* We're all done... */
1600 extern void sk_init(void);
1602 #ifdef CONFIG_BRIDGE
1603 extern int br_init(void);
1606 #ifdef CONFIG_WAN_ROUTER
1607 extern void wanrouter_init(void);
1610 void __init
sock_init(void)
1614 printk(KERN_INFO
"Linux NET4.0 for Linux 2.3\n");
1615 printk(KERN_INFO
"Based upon Swansea University Computer Society NET3.039\n");
1618 * Initialize all address (protocol) families.
1621 for (i
= 0; i
< NPROTO
; i
++)
1622 net_families
[i
] = NULL
;
1625 * Initialize sock SLAB cache.
1632 * Initialize skbuff SLAB cache
1638 * Ethernet bridge layer.
1641 #ifdef CONFIG_BRIDGE
1649 #ifdef CONFIG_WAN_ROUTER
1654 * Initialize the protocols module.
1660 * The netlink device handler may be needed early.
1663 #ifdef CONFIG_RTNETLINK
1666 #ifdef CONFIG_NETLINK_DEV
1669 #ifdef CONFIG_NETFILTER
1674 int socket_get_info(char *buffer
, char **start
, off_t offset
, int length
)
1679 for (cpu
=0; cpu
<smp_num_cpus
; cpu
++)
1680 counter
+= sockets_in_use
[cpu_logical_map(cpu
)].counter
;
1682 /* It can be negative, by the way. 8) */
1686 len
= sprintf(buffer
, "sockets: used %d\n", counter
);
1692 *start
= buffer
+ offset
;