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.
46 * This program is free software; you can redistribute it and/or
47 * modify it under the terms of the GNU General Public License
48 * as published by the Free Software Foundation; either version
49 * 2 of the License, or (at your option) any later version.
52 * This module is effectively the top level interface to the BSD socket
57 #include <linux/config.h>
58 #include <linux/signal.h>
59 #include <linux/errno.h>
60 #include <linux/sched.h>
62 #include <linux/smp.h>
63 #include <linux/smp_lock.h>
64 #include <linux/kernel.h>
65 #include <linux/major.h>
66 #include <linux/stat.h>
67 #include <linux/socket.h>
68 #include <linux/fcntl.h>
69 #include <linux/file.h>
70 #include <linux/net.h>
71 #include <linux/interrupt.h>
72 #include <linux/netdevice.h>
73 #include <linux/proc_fs.h>
74 #include <linux/firewall.h>
75 #include <linux/wanrouter.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
79 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
80 #include <linux/kmod.h>
83 #include <asm/system.h>
84 #include <asm/uaccess.h>
86 #include <linux/inet.h>
88 #include <net/protocol.h>
92 #include <linux/skbuff.h>
97 static long long sock_lseek(struct file
*file
, long long offset
, int whence
);
98 static ssize_t
sock_read(struct file
*file
, char *buf
,
99 size_t size
, loff_t
*ppos
);
100 static ssize_t
sock_write(struct file
*file
, const char *buf
,
101 size_t size
, loff_t
*ppos
);
103 static int sock_close(struct inode
*inode
, struct file
*file
);
104 static unsigned int sock_poll(struct file
*file
,
105 struct poll_table_struct
*wait
);
106 static int sock_ioctl(struct inode
*inode
, struct file
*file
,
107 unsigned int cmd
, unsigned long arg
);
108 static int sock_fasync(int fd
, struct file
*filp
, int on
);
112 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
113 * in the operation structures but are done directly via the socketcall() multiplexor.
116 static struct file_operations socket_file_ops
= {
124 NULL
, /* no special open code... */
132 * The protocol list. Each protocol is registered in here.
135 struct net_proto_family
*net_families
[NPROTO
];
138 * Statistics counters of the socket lists
141 static int sockets_in_use
= 0;
144 * Support routines. Move socket addresses back and forth across the kernel/user
145 * divide and look after the messy bits.
148 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
149 16 for IP, 16 for IPX,
152 must be at least one bigger than
153 the AF_UNIX size (see net/unix/af_unix.c
157 int move_addr_to_kernel(void *uaddr
, int ulen
, void *kaddr
)
159 if(ulen
<0||ulen
>MAX_SOCK_ADDR
)
163 if(copy_from_user(kaddr
,uaddr
,ulen
))
168 int move_addr_to_user(void *kaddr
, int klen
, void *uaddr
, int *ulen
)
173 if((err
=get_user(len
, ulen
)))
177 if(len
<0 || len
> MAX_SOCK_ADDR
)
181 if(copy_to_user(uaddr
,kaddr
,len
))
185 * "fromlen shall refer to the value before truncation.."
188 return __put_user(klen
, ulen
);
192 * Obtains the first available file descriptor and sets it up for use.
195 static int get_fd(struct inode
*inode
)
200 * Find a file descriptor suitable for return to the user.
203 fd
= get_unused_fd();
205 struct file
*file
= get_empty_filp();
212 file
->f_dentry
= d_alloc_root(inode
, NULL
);
213 if (!file
->f_dentry
) {
220 * The socket maintains a reference to the inode, so we
221 * have to increment the count.
225 fd_install(fd
, file
);
226 file
->f_op
= &socket_file_ops
;
228 file
->f_flags
= O_RDWR
;
234 extern __inline__
struct socket
*socki_lookup(struct inode
*inode
)
236 return &inode
->u
.socket_i
;
240 * Go from a file number to its socket slot.
243 extern struct socket
*sockfd_lookup(int fd
, int *err
)
249 if (!(file
= fget(fd
)))
255 inode
= file
->f_dentry
->d_inode
;
256 if (!inode
|| !inode
->i_sock
|| !(sock
= socki_lookup(inode
)))
263 if (sock
->file
!= file
) {
264 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
270 extern __inline__
void sockfd_put(struct socket
*sock
)
279 struct socket
*sock_alloc(void)
281 struct inode
* inode
;
282 struct socket
* sock
;
284 inode
= get_empty_inode();
288 sock
= socki_lookup(inode
);
290 inode
->i_mode
= S_IFSOCK
;
292 inode
->i_uid
= current
->uid
;
293 inode
->i_gid
= current
->gid
;
296 init_waitqueue(&sock
->wait
);
297 sock
->fasync_list
= NULL
;
298 sock
->state
= SS_UNCONNECTED
;
308 void sock_release(struct socket
*sock
)
310 if (sock
->state
!= SS_UNCONNECTED
)
311 sock
->state
= SS_DISCONNECTING
;
314 sock
->ops
->release(sock
, NULL
);
316 if (sock
->fasync_list
)
317 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
319 --sockets_in_use
; /* Bookkeeping.. */
324 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, int size
)
327 struct scm_cookie scm
;
329 err
= scm_send(sock
, msg
, &scm
);
331 err
= sock
->ops
->sendmsg(sock
, msg
, size
, &scm
);
337 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int size
, int flags
)
339 struct scm_cookie scm
;
341 memset(&scm
, 0, sizeof(scm
));
343 size
= sock
->ops
->recvmsg(sock
, msg
, size
, flags
, &scm
);
345 scm_recv(sock
, msg
, &scm
, flags
);
352 * Sockets are not seekable.
355 static long long sock_lseek(struct file
*file
,long long offset
, int whence
)
361 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
362 * area ubuf...ubuf+size-1 is writable before asking the protocol.
365 static ssize_t
sock_read(struct file
*file
, char *ubuf
,
366 size_t size
, loff_t
*ppos
)
372 if (ppos
!= &file
->f_pos
)
374 if (size
==0) /* Match SYS5 behaviour */
377 sock
= socki_lookup(file
->f_dentry
->d_inode
);
383 msg
.msg_control
=NULL
;
384 msg
.msg_controllen
=0;
388 return sock_recvmsg(sock
, &msg
, size
,
389 !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
);
394 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
395 * is readable by the user process.
398 static ssize_t
sock_write(struct file
*file
, const char *ubuf
,
399 size_t size
, loff_t
*ppos
)
405 if (ppos
!= &file
->f_pos
)
407 if(size
==0) /* Match SYS5 behaviour */
410 sock
= socki_lookup(file
->f_dentry
->d_inode
);
416 msg
.msg_control
=NULL
;
417 msg
.msg_controllen
=0;
418 msg
.msg_flags
=!(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
419 iov
.iov_base
=(void *)ubuf
;
422 return sock_sendmsg(sock
, &msg
, size
);
425 int sock_readv_writev(int type
, struct inode
* inode
, struct file
* file
,
426 const struct iovec
* iov
, long count
, long size
)
431 sock
= socki_lookup(inode
);
435 msg
.msg_control
= NULL
;
436 msg
.msg_controllen
= 0;
437 msg
.msg_iov
= (struct iovec
*) iov
;
438 msg
.msg_iovlen
= count
;
439 msg
.msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
441 /* read() does a VERIFY_WRITE */
442 if (type
== VERIFY_WRITE
)
443 return sock_recvmsg(sock
, &msg
, size
, msg
.msg_flags
);
444 return sock_sendmsg(sock
, &msg
, size
);
449 * With an ioctl arg may well be a user mode pointer, but we don't know what to do
450 * with it - that's up to the protocol still.
453 int sock_ioctl(struct inode
*inode
, struct file
*file
, unsigned int cmd
,
456 struct socket
*sock
= socki_lookup(inode
);
457 return sock
->ops
->ioctl(sock
, cmd
, arg
);
461 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
465 sock
= socki_lookup(file
->f_dentry
->d_inode
);
468 * We can't return errors to poll, so it's either yes or no.
471 return sock
->ops
->poll(file
, sock
, wait
);
475 int sock_close(struct inode
*inode
, struct file
*filp
)
478 * It was possible the inode is NULL we were
479 * closing an unfinished socket.
484 printk(KERN_DEBUG
"sock_close: NULL inode\n");
487 sock_fasync(-1, filp
, 0);
488 sock_release(socki_lookup(inode
));
493 * Update the socket async list
496 static int sock_fasync(int fd
, struct file
*filp
, int on
)
498 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
504 fna
=(struct fasync_struct
*)kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
509 sock
= socki_lookup(filp
->f_dentry
->d_inode
);
511 prev
=&(sock
->fasync_list
);
516 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
517 if (fa
->fa_file
==filp
)
525 kfree_s(fna
,sizeof(struct fasync_struct
));
526 restore_flags(flags
);
531 fna
->magic
=FASYNC_MAGIC
;
532 fna
->fa_next
=sock
->fasync_list
;
533 sock
->fasync_list
=fna
;
540 kfree_s(fa
,sizeof(struct fasync_struct
));
543 restore_flags(flags
);
547 int sock_wake_async(struct socket
*sock
, int how
)
549 if (!sock
|| !sock
->fasync_list
)
554 if (sock
->flags
& SO_WAITDATA
)
558 if (!(sock
->flags
& SO_NOSPACE
))
560 sock
->flags
&= ~SO_NOSPACE
;
564 kill_fasync(sock
->fasync_list
, SIGIO
);
571 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
577 * Check protocol is in range
579 if(family
<0||family
>=NPROTO
)
582 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
583 /* Attempt to load a protocol module if the find failed.
585 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
586 * requested real, full-featured networking support upon configuration.
587 * Otherwise module support will break!
589 if (net_families
[family
]==NULL
)
591 char module_name
[30];
592 sprintf(module_name
,"net-pf-%d",family
);
593 request_module(module_name
);
597 if (net_families
[family
]==NULL
)
601 * Check that this is a type that we know how to manipulate and
602 * the protocol makes sense here. The family can still reject the
606 if ((type
!= SOCK_STREAM
&& type
!= SOCK_DGRAM
&&
607 type
!= SOCK_SEQPACKET
&& type
!= SOCK_RAW
&& type
!= SOCK_RDM
&&
611 type
!= SOCK_PACKET
) || protocol
< 0)
615 * Allocate the socket and allow the family to set things up. if
616 * the protocol is 0, the family is instructed to select an appropriate
620 if (!(sock
= sock_alloc()))
622 printk(KERN_WARNING
"socket: no more sockets\n");
623 return -ENFILE
; /* Not exactly a match, but its the
624 closest posix thing */
629 if ((i
= net_families
[family
]->create(sock
, protocol
)) < 0)
639 asmlinkage
int sys_socket(int family
, int type
, int protocol
)
646 retval
= sock_create(family
, type
, protocol
, &sock
);
650 retval
= get_fd(sock
->inode
);
653 sock
->file
= fcheck(retval
);
665 * Create a pair of connected sockets.
668 asmlinkage
int sys_socketpair(int family
, int type
, int protocol
, int usockvec
[2])
670 struct socket
*sock1
, *sock2
;
676 * Obtain the first socket and check if the underlying protocol
677 * supports the socketpair call.
680 err
= sys_socket(family
, type
, protocol
);
686 * Now grab another socket
689 fd2
= sys_socket(family
, type
, protocol
);
694 * Get the sockets for the two fd's
696 sock1
= sockfd_lookup(fd1
, &err
);
699 sock2
= sockfd_lookup(fd2
, &err
);
703 /* try to connect the two sockets together */
704 err
= sock1
->ops
->socketpair(sock1
, sock2
);
708 err
= put_user(fd1
, &usockvec
[0]);
711 err
= put_user(fd2
, &usockvec
[1]);
731 * Bind a name to a socket. Nothing much to do here since it's
732 * the protocol's responsibility to handle the local address.
734 * We move the socket address to kernel space before we call
735 * the protocol layer (having also checked the address is ok).
738 asmlinkage
int sys_bind(int fd
, struct sockaddr
*umyaddr
, int addrlen
)
741 char address
[MAX_SOCK_ADDR
];
745 if((sock
= sockfd_lookup(fd
,&err
))!=NULL
)
747 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0)
748 err
= sock
->ops
->bind(sock
, (struct sockaddr
*)address
, addrlen
);
757 * Perform a listen. Basically, we allow the protocol to do anything
758 * necessary for a listen, and if that works, we mark the socket as
759 * ready for listening.
762 asmlinkage
int sys_listen(int fd
, int backlog
)
768 if((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
770 err
=sock
->ops
->listen(sock
, backlog
);
779 * For accept, we attempt to create a new socket, set up the link
780 * with the client, wake up the client, then return the new
781 * connected fd. We collect the address of the connector in kernel
782 * space and move it to user at the very end. This is unclean because
783 * we open the socket then return an error.
785 * 1003.1g adds the ability to recvmsg() to query connection pending
786 * status to recvmsg. We need to add that support in a way thats
787 * clean when we restucture accept also.
790 asmlinkage
int sys_accept(int fd
, struct sockaddr
*upeer_sockaddr
, int *upeer_addrlen
)
793 struct socket
*sock
, *newsock
;
795 char address
[MAX_SOCK_ADDR
];
798 sock
= sockfd_lookup(fd
, &err
);
804 if (!(newsock
= sock_alloc()))
807 inode
= newsock
->inode
;
808 newsock
->type
= sock
->type
;
810 err
= sock
->ops
->dup(newsock
, sock
);
814 err
= newsock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
817 newsock
= socki_lookup(inode
);
819 if ((err
= get_fd(inode
)) < 0)
821 newsock
->file
= fcheck(err
);
825 /* Handle the race where the accept works and we
826 then getname after it has closed again */
827 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 1)<0)
832 /* N.B. Should check for errors here */
833 move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
843 sock_release(newsock
);
849 * Attempt to connect to a socket with the server address. The address
850 * is in user space so we verify it is OK and move it to kernel space.
852 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
855 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
856 * other SEQPACKET protocols that take time to connect() as it doesn't
857 * include the -EINPROGRESS status for such sockets.
860 asmlinkage
int sys_connect(int fd
, struct sockaddr
*uservaddr
, int addrlen
)
863 char address
[MAX_SOCK_ADDR
];
867 sock
= sockfd_lookup(fd
, &err
);
870 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
873 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
874 sock
->file
->f_flags
);
883 * Get the local address ('name') of a socket object. Move the obtained
884 * name to user space.
887 asmlinkage
int sys_getsockname(int fd
, struct sockaddr
*usockaddr
, int *usockaddr_len
)
890 char address
[MAX_SOCK_ADDR
];
894 sock
= sockfd_lookup(fd
, &err
);
897 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
900 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
910 * Get the remote address ('name') of a socket object. Move the obtained
911 * name to user space.
914 asmlinkage
int sys_getpeername(int fd
, struct sockaddr
*usockaddr
, int *usockaddr_len
)
917 char address
[MAX_SOCK_ADDR
];
921 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
923 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
925 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
933 * Send a datagram down a socket. The datagram as with write() is
934 * in user space. We check it can be read.
937 asmlinkage
int sys_send(int fd
, void * buff
, size_t len
, unsigned flags
)
945 sock
= sockfd_lookup(fd
, &err
);
953 msg
.msg_control
=NULL
;
954 msg
.msg_controllen
=0;
955 if (sock
->file
->f_flags
& O_NONBLOCK
)
956 flags
|= MSG_DONTWAIT
;
957 msg
.msg_flags
= flags
;
958 err
= sock_sendmsg(sock
, &msg
, len
);
967 * Send a datagram to a given address. We move the address into kernel
968 * space and check the user space data area is readable before invoking
972 asmlinkage
int sys_sendto(int fd
, void * buff
, size_t len
, unsigned flags
,
973 struct sockaddr
*addr
, int addr_len
)
976 char address
[MAX_SOCK_ADDR
];
982 sock
= sockfd_lookup(fd
, &err
);
990 msg
.msg_control
=NULL
;
991 msg
.msg_controllen
=0;
992 msg
.msg_namelen
=addr_len
;
995 err
= move_addr_to_kernel(addr
, addr_len
, address
);
998 msg
.msg_name
=address
;
1000 if (sock
->file
->f_flags
& O_NONBLOCK
)
1001 flags
|= MSG_DONTWAIT
;
1002 msg
.msg_flags
= flags
;
1003 err
= sock_sendmsg(sock
, &msg
, len
);
1014 * Receive a frame from the socket and optionally record the address of the
1015 * sender. We verify the buffers are writable and if needed move the
1016 * sender address from kernel to user space.
1019 asmlinkage
int sys_recvfrom(int fd
, void * ubuf
, size_t size
, unsigned flags
,
1020 struct sockaddr
*addr
, int *addr_len
)
1022 struct socket
*sock
;
1025 char address
[MAX_SOCK_ADDR
];
1029 sock
= sockfd_lookup(fd
, &err
);
1033 msg
.msg_control
=NULL
;
1034 msg
.msg_controllen
=0;
1039 msg
.msg_name
=address
;
1040 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1041 if (sock
->file
->f_flags
& O_NONBLOCK
)
1042 flags
|= MSG_DONTWAIT
;
1043 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1045 if(err
>= 0 && addr
!= NULL
)
1047 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1058 * Receive a datagram from a socket.
1061 asmlinkage
int sys_recv(int fd
, void * ubuf
, size_t size
, unsigned flags
)
1063 return sys_recvfrom(fd
,ubuf
,size
,flags
, NULL
, NULL
);
1067 * Set a socket option. Because we don't know the option lengths we have
1068 * to pass the user mode parameter for the protocols to sort out.
1071 asmlinkage
int sys_setsockopt(int fd
, int level
, int optname
, char *optval
, int optlen
)
1074 struct socket
*sock
;
1077 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1079 if (level
== SOL_SOCKET
)
1080 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1082 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1090 * Get a socket option. Because we don't know the option lengths we have
1091 * to pass a user mode parameter for the protocols to sort out.
1094 asmlinkage
int sys_getsockopt(int fd
, int level
, int optname
, char *optval
, int *optlen
)
1097 struct socket
*sock
;
1100 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1102 if (level
== SOL_SOCKET
)
1103 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1105 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1114 * Shutdown a socket.
1117 asmlinkage
int sys_shutdown(int fd
, int how
)
1120 struct socket
*sock
;
1123 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1125 err
=sock
->ops
->shutdown(sock
, how
);
1133 * BSD sendmsg interface
1136 asmlinkage
int sys_sendmsg(int fd
, struct msghdr
*msg
, unsigned flags
)
1138 struct socket
*sock
;
1139 char address
[MAX_SOCK_ADDR
];
1140 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1141 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]; /* 20 is size of ipv6_pktinfo */
1142 unsigned char *ctl_buf
= ctl
;
1143 struct msghdr msg_sys
;
1144 int err
, ctl_len
, iov_size
, total_len
;
1149 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1152 sock
= sockfd_lookup(fd
, &err
);
1156 /* do not move before msg_sys is valid */
1158 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1161 /* Check whether to allocate the iovec area*/
1163 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1164 if (msg_sys
.msg_iovlen
> 1 /* UIO_FASTIOV */) {
1165 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1170 /* This will also move the address data into kernel space */
1171 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1176 ctl_len
= msg_sys
.msg_controllen
;
1179 if (ctl_len
> sizeof(ctl
))
1181 /* Suggested by the Advanced Sockets API for IPv6 draft:
1182 * Limit the msg_controllen size by the SO_SNDBUF size.
1184 /* Note - when this code becomes multithreaded on
1185 * SMP machines you have a race to fix here.
1188 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1189 if (ctl_buf
== NULL
)
1193 if (copy_from_user(ctl_buf
, msg_sys
.msg_control
, ctl_len
))
1195 msg_sys
.msg_control
= ctl_buf
;
1197 msg_sys
.msg_flags
= flags
;
1199 if (sock
->file
->f_flags
& O_NONBLOCK
)
1200 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1201 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1205 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1207 if (iov
!= iovstack
)
1208 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1217 * BSD recvmsg interface
1220 asmlinkage
int sys_recvmsg(int fd
, struct msghdr
*msg
, unsigned int flags
)
1222 struct socket
*sock
;
1223 struct iovec iovstack
[UIO_FASTIOV
];
1224 struct iovec
*iov
=iovstack
;
1225 struct msghdr msg_sys
;
1226 unsigned long cmsg_ptr
;
1227 int err
, iov_size
, total_len
, len
;
1229 /* kernel mode address */
1230 char addr
[MAX_SOCK_ADDR
];
1232 /* user mode address pointers */
1233 struct sockaddr
*uaddr
;
1238 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1241 sock
= sockfd_lookup(fd
, &err
);
1246 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1249 /* Check whether to allocate the iovec area*/
1251 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1252 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1253 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1259 * Save the user-mode address (verify_iovec will change the
1260 * kernel msghdr to use the kernel address space)
1263 uaddr
= msg_sys
.msg_name
;
1264 uaddr_len
= &msg
->msg_namelen
;
1265 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1270 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1271 msg_sys
.msg_flags
= 0;
1273 if (sock
->file
->f_flags
& O_NONBLOCK
)
1274 flags
|= MSG_DONTWAIT
;
1275 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1280 if (uaddr
!= NULL
) {
1281 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1285 err
= __put_user(msg_sys
.msg_flags
, &msg
->msg_flags
);
1288 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1289 &msg
->msg_controllen
);
1295 if (iov
!= iovstack
)
1296 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1306 * Perform a file control on a socket file descriptor.
1308 * FIXME: does this need an fd lock ?
1311 int sock_fcntl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
1313 struct socket
*sock
;
1315 sock
= socki_lookup (filp
->f_dentry
->d_inode
);
1316 if (sock
&& sock
->ops
)
1317 return sock
->ops
->fcntl(sock
, cmd
, arg
);
1321 /* Argument list sizes for sys_socketcall */
1322 #define AL(x) ((x) * sizeof(unsigned long))
1323 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1324 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1325 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1329 * System call vectors.
1331 * Argument checking cleaned up. Saved 20% in size.
1332 * This function doesn't need to set the kernel lock because
1333 * it is set by the callees.
1336 asmlinkage
int sys_socketcall(int call
, unsigned long *args
)
1339 unsigned long a0
,a1
;
1342 if(call
<1||call
>SYS_RECVMSG
)
1345 /* copy_from_user should be SMP safe. */
1346 if (copy_from_user(a
, args
, nargs
[call
]))
1355 err
= sys_socket(a0
,a1
,a
[2]);
1358 err
= sys_bind(a0
,(struct sockaddr
*)a1
, a
[2]);
1361 err
= sys_connect(a0
, (struct sockaddr
*)a1
, a
[2]);
1364 err
= sys_listen(a0
,a1
);
1367 err
= sys_accept(a0
,(struct sockaddr
*)a1
, (int *)a
[2]);
1369 case SYS_GETSOCKNAME
:
1370 err
= sys_getsockname(a0
,(struct sockaddr
*)a1
, (int *)a
[2]);
1372 case SYS_GETPEERNAME
:
1373 err
= sys_getpeername(a0
, (struct sockaddr
*)a1
, (int *)a
[2]);
1375 case SYS_SOCKETPAIR
:
1376 err
= sys_socketpair(a0
,a1
, a
[2], (int *)a
[3]);
1379 err
= sys_send(a0
, (void *)a1
, a
[2], a
[3]);
1382 err
= sys_sendto(a0
,(void *)a1
, a
[2], a
[3],
1383 (struct sockaddr
*)a
[4], a
[5]);
1386 err
= sys_recv(a0
, (void *)a1
, a
[2], a
[3]);
1389 err
= sys_recvfrom(a0
, (void *)a1
, a
[2], a
[3],
1390 (struct sockaddr
*)a
[4], (int *)a
[5]);
1393 err
= sys_shutdown(a0
,a1
);
1395 case SYS_SETSOCKOPT
:
1396 err
= sys_setsockopt(a0
, a1
, a
[2], (char *)a
[3], a
[4]);
1398 case SYS_GETSOCKOPT
:
1399 err
= sys_getsockopt(a0
, a1
, a
[2], (char *)a
[3], (int *)a
[4]);
1402 err
= sys_sendmsg(a0
, (struct msghdr
*) a1
, a
[2]);
1405 err
= sys_recvmsg(a0
, (struct msghdr
*) a1
, a
[2]);
1415 * This function is called by a protocol handler that wants to
1416 * advertise its address family, and have it linked into the
1420 int sock_register(struct net_proto_family
*ops
)
1422 if (ops
->family
>= NPROTO
) {
1423 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
1426 net_families
[ops
->family
]=ops
;
1431 * This function is called by a protocol handler that wants to
1432 * remove its address family, and have it unlinked from the
1436 int sock_unregister(int family
)
1438 if (family
< 0 || family
>= NPROTO
)
1441 net_families
[family
]=NULL
;
1445 void __init
proto_init(void)
1447 extern struct net_proto protocols
[]; /* Network protocols */
1448 struct net_proto
*pro
;
1450 /* Kick all configured protocols. */
1452 while (pro
->name
!= NULL
)
1454 (*pro
->init_func
)(pro
);
1457 /* We're all done... */
1460 extern void sk_init(void);
1461 #ifdef CONFIG_WAN_ROUTER
1462 extern void wanrouter_init(void);
1465 void __init
sock_init(void)
1469 printk(KERN_INFO
"Swansea University Computer Society NET3.039 for Linux 2.1\n");
1472 * Initialize all address (protocol) families.
1475 for (i
= 0; i
< NPROTO
; i
++)
1476 net_families
[i
] = NULL
;
1479 * Initialize sock SLAB cache.
1486 * Initialize skbuff SLAB cache
1496 #ifdef CONFIG_WAN_ROUTER
1501 * Attach the firewall module if configured
1504 #ifdef CONFIG_FIREWALL
1509 * Initialize the protocols module.
1515 * The netlink device handler may be needed early.
1518 #ifdef CONFIG_RTNETLINK
1521 #ifdef CONFIG_NETLINK_DEV
1526 int socket_get_info(char *buffer
, char **start
, off_t offset
, int length
)
1528 int len
= sprintf(buffer
, "sockets: used %d\n", sockets_in_use
);
1534 *start
= buffer
+ offset
;