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/netlink.h>
71 #include <linux/rtnetlink.h>
72 #include <linux/init.h>
73 #include <linux/poll.h>
74 #include <linux/cache.h>
75 #include <linux/module.h>
76 #include <linux/highmem.h>
78 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
79 #include <linux/kmod.h>
82 #include <asm/uaccess.h>
86 #include <linux/netfilter.h>
88 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
89 static ssize_t
sock_read(struct file
*file
, char *buf
,
90 size_t size
, loff_t
*ppos
);
91 static ssize_t
sock_write(struct file
*file
, const char *buf
,
92 size_t size
, loff_t
*ppos
);
93 static int sock_mmap(struct file
*file
, struct vm_area_struct
* vma
);
95 static int sock_close(struct inode
*inode
, struct file
*file
);
96 static unsigned int sock_poll(struct file
*file
,
97 struct poll_table_struct
*wait
);
98 static int sock_ioctl(struct inode
*inode
, struct file
*file
,
99 unsigned int cmd
, unsigned long arg
);
100 static int sock_fasync(int fd
, struct file
*filp
, int on
);
101 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
102 unsigned long count
, loff_t
*ppos
);
103 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
104 unsigned long count
, loff_t
*ppos
);
105 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
106 int offset
, size_t size
, loff_t
*ppos
, int more
);
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.
114 static struct file_operations socket_file_ops
= {
121 open
: sock_no_open
, /* special open code to disallow open via /proc */
126 sendpage
: sock_sendpage
130 * The protocol list. Each protocol is registered in here.
133 static struct net_proto_family
*net_families
[NPROTO
];
136 static atomic_t net_family_lockct
= ATOMIC_INIT(0);
137 static spinlock_t net_family_lock
= SPIN_LOCK_UNLOCKED
;
139 /* The strategy is: modifications net_family vector are short, do not
140 sleep and veeery rare, but read access should be free of any exclusive
144 static void net_family_write_lock(void)
146 spin_lock(&net_family_lock
);
147 while (atomic_read(&net_family_lockct
) != 0) {
148 spin_unlock(&net_family_lock
);
152 spin_lock(&net_family_lock
);
156 static __inline__
void net_family_write_unlock(void)
158 spin_unlock(&net_family_lock
);
161 static __inline__
void net_family_read_lock(void)
163 atomic_inc(&net_family_lockct
);
164 spin_unlock_wait(&net_family_lock
);
167 static __inline__
void net_family_read_unlock(void)
169 atomic_dec(&net_family_lockct
);
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)
181 * Statistics counters of the socket lists
186 char __pad
[SMP_CACHE_BYTES
];
187 } sockets_in_use
[NR_CPUS
] __cacheline_aligned
= {{0}};
190 * Support routines. Move socket addresses back and forth across the kernel/user
191 * divide and look after the messy bits.
194 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
195 16 for IP, 16 for IPX,
198 must be at least one bigger than
199 the AF_UNIX size (see net/unix/af_unix.c
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
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.
214 int move_addr_to_kernel(void *uaddr
, int ulen
, void *kaddr
)
216 if(ulen
<0||ulen
>MAX_SOCK_ADDR
)
220 if(copy_from_user(kaddr
,uaddr
,ulen
))
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
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
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.
242 int move_addr_to_user(void *kaddr
, int klen
, void *uaddr
, int *ulen
)
247 if((err
=get_user(len
, ulen
)))
251 if(len
<0 || len
> MAX_SOCK_ADDR
)
255 if(copy_to_user(uaddr
,kaddr
,len
))
259 * "fromlen shall refer to the value before truncation.."
262 return __put_user(klen
, ulen
);
265 #define SOCKFS_MAGIC 0x534F434B
266 static int sockfs_statfs(struct super_block
*sb
, struct statfs
*buf
)
268 buf
->f_type
= SOCKFS_MAGIC
;
270 buf
->f_namelen
= 255;
274 static struct super_operations sockfs_ops
= {
275 statfs
: sockfs_statfs
,
278 static struct super_block
* sockfs_read_super(struct super_block
*sb
, void *data
, int silent
)
280 struct inode
*root
= new_inode(sb
);
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 });
295 sb
->s_root
->d_sb
= sb
;
296 sb
->s_root
->d_parent
= sb
->s_root
;
297 d_instantiate(sb
->s_root
, root
);
301 static struct vfsmount
*sock_mnt
;
302 static DECLARE_FSTYPE(sock_fs_type
, "sockfs", sockfs_read_super
, FS_NOMOUNT
);
303 static int sockfs_delete_dentry(struct dentry
*dentry
)
307 static struct dentry_operations sockfs_dentry_operations
= {
308 d_delete
: sockfs_delete_dentry
,
312 * Obtains the first available file descriptor and sets it up for use.
314 * This functions creates file structure and maps it to fd space
315 * of current process. On success it returns file descriptor
316 * and file struct implicitly stored in sock->file.
317 * Note that another thread may close file descriptor before we return
318 * from this function. We use the fact that now we do not refer
319 * to socket after mapping. If one day we will need it, this
320 * function will inincrement ref. count on file by 1.
322 * In any case returned fd MAY BE not valid!
323 * This race condition is inavoidable
324 * with shared fd spaces, we cannot solve is inside kernel,
325 * but we take care of internal coherence yet.
328 static int sock_map_fd(struct socket
*sock
)
335 * Find a file descriptor suitable for return to the user.
338 fd
= get_unused_fd();
340 struct file
*file
= get_empty_filp();
348 sprintf(name
, "[%lu]", sock
->inode
->i_ino
);
350 this.len
= strlen(name
);
351 this.hash
= sock
->inode
->i_ino
;
353 file
->f_dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &this);
354 if (!file
->f_dentry
) {
360 file
->f_dentry
->d_op
= &sockfs_dentry_operations
;
361 d_add(file
->f_dentry
, sock
->inode
);
362 file
->f_vfsmnt
= mntget(sock_mnt
);
365 file
->f_op
= sock
->inode
->i_fop
= &socket_file_ops
;
367 file
->f_flags
= O_RDWR
;
369 fd_install(fd
, file
);
376 extern __inline__
struct socket
*socki_lookup(struct inode
*inode
)
378 return &inode
->u
.socket_i
;
382 * sockfd_lookup - Go from a file number to its socket slot
384 * @err: pointer to an error code return
386 * The file handle passed in is locked and the socket it is bound
387 * too is returned. If an error occurs the err pointer is overwritten
388 * with a negative errno code and NULL is returned. The function checks
389 * for both invalid handles and passing a handle which is not a socket.
391 * On a success the socket object pointer is returned.
394 struct socket
*sockfd_lookup(int fd
, int *err
)
400 if (!(file
= fget(fd
)))
406 inode
= file
->f_dentry
->d_inode
;
407 if (!inode
->i_sock
|| !(sock
= socki_lookup(inode
)))
414 if (sock
->file
!= file
) {
415 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
421 extern __inline__
void sockfd_put(struct socket
*sock
)
427 * sock_alloc - allocate a socket
429 * Allocate a new inode and socket object. The two are bound together
430 * and initialised. The socket is then returned. If we are out of inodes
434 struct socket
*sock_alloc(void)
436 struct inode
* inode
;
437 struct socket
* sock
;
439 inode
= get_empty_inode();
443 inode
->i_sb
= sock_mnt
->mnt_sb
;
444 sock
= socki_lookup(inode
);
446 inode
->i_mode
= S_IFSOCK
|S_IRWXUGO
;
448 inode
->i_uid
= current
->fsuid
;
449 inode
->i_gid
= current
->fsgid
;
452 init_waitqueue_head(&sock
->wait
);
453 sock
->fasync_list
= NULL
;
454 sock
->state
= SS_UNCONNECTED
;
460 sockets_in_use
[smp_processor_id()].counter
++;
465 * In theory you can't get an open on this inode, but /proc provides
466 * a back door. Remember to keep it shut otherwise you'll let the
467 * creepy crawlies in.
470 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
476 * sock_release - close a socket
477 * @sock: socket to close
479 * The socket is released from the protocol stack if it has a release
480 * callback, and the inode is then released if the socket is bound to
481 * an inode not a file.
484 void sock_release(struct socket
*sock
)
487 sock
->ops
->release(sock
);
489 if (sock
->fasync_list
)
490 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
492 sockets_in_use
[smp_processor_id()].counter
--;
500 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, int size
)
503 struct scm_cookie scm
;
505 err
= scm_send(sock
, msg
, &scm
);
507 err
= sock
->ops
->sendmsg(sock
, msg
, size
, &scm
);
513 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int size
, int flags
)
515 struct scm_cookie scm
;
517 memset(&scm
, 0, sizeof(scm
));
519 size
= sock
->ops
->recvmsg(sock
, msg
, size
, flags
, &scm
);
521 scm_recv(sock
, msg
, &scm
, flags
);
528 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
529 * area ubuf...ubuf+size-1 is writable before asking the protocol.
532 static ssize_t
sock_read(struct file
*file
, char *ubuf
,
533 size_t size
, loff_t
*ppos
)
540 if (ppos
!= &file
->f_pos
)
542 if (size
==0) /* Match SYS5 behaviour */
545 sock
= socki_lookup(file
->f_dentry
->d_inode
);
551 msg
.msg_control
=NULL
;
552 msg
.msg_controllen
=0;
555 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
557 return sock_recvmsg(sock
, &msg
, size
, flags
);
562 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
563 * is readable by the user process.
566 static ssize_t
sock_write(struct file
*file
, const char *ubuf
,
567 size_t size
, loff_t
*ppos
)
573 if (ppos
!= &file
->f_pos
)
575 if(size
==0) /* Match SYS5 behaviour */
578 sock
= socki_lookup(file
->f_dentry
->d_inode
);
584 msg
.msg_control
=NULL
;
585 msg
.msg_controllen
=0;
586 msg
.msg_flags
=!(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
587 if (sock
->type
== SOCK_SEQPACKET
)
588 msg
.msg_flags
|= MSG_EOR
;
589 iov
.iov_base
=(void *)ubuf
;
592 return sock_sendmsg(sock
, &msg
, size
);
595 ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
596 int offset
, size_t size
, loff_t
*ppos
, int more
)
601 if (ppos
!= &file
->f_pos
)
604 sock
= socki_lookup(file
->f_dentry
->d_inode
);
606 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
610 if (!sock
->ops
->sendpage
)
611 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
613 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
616 int sock_readv_writev(int type
, struct inode
* inode
, struct file
* file
,
617 const struct iovec
* iov
, long count
, long size
)
622 sock
= socki_lookup(inode
);
626 msg
.msg_control
= NULL
;
627 msg
.msg_controllen
= 0;
628 msg
.msg_iov
= (struct iovec
*) iov
;
629 msg
.msg_iovlen
= count
;
630 msg
.msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
632 /* read() does a VERIFY_WRITE */
633 if (type
== VERIFY_WRITE
)
634 return sock_recvmsg(sock
, &msg
, size
, msg
.msg_flags
);
636 if (sock
->type
== SOCK_SEQPACKET
)
637 msg
.msg_flags
|= MSG_EOR
;
639 return sock_sendmsg(sock
, &msg
, size
);
642 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
643 unsigned long count
, loff_t
*ppos
)
647 for (i
= 0 ; i
< count
; i
++)
648 tot_len
+= vector
[i
].iov_len
;
649 return sock_readv_writev(VERIFY_WRITE
, file
->f_dentry
->d_inode
,
650 file
, vector
, count
, tot_len
);
653 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
654 unsigned long count
, loff_t
*ppos
)
658 for (i
= 0 ; i
< count
; i
++)
659 tot_len
+= vector
[i
].iov_len
;
660 return sock_readv_writev(VERIFY_READ
, file
->f_dentry
->d_inode
,
661 file
, vector
, count
, tot_len
);
665 * With an ioctl arg may well be a user mode pointer, but we don't know what to do
666 * with it - that's up to the protocol still.
669 int sock_ioctl(struct inode
*inode
, struct file
*file
, unsigned int cmd
,
676 sock
= socki_lookup(inode
);
677 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
684 /* No kernel lock held - perfect */
685 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
690 * We can't return errors to poll, so it's either yes or no.
692 sock
= socki_lookup(file
->f_dentry
->d_inode
);
693 return sock
->ops
->poll(file
, sock
, wait
);
696 static int sock_mmap(struct file
* file
, struct vm_area_struct
* vma
)
698 struct socket
*sock
= socki_lookup(file
->f_dentry
->d_inode
);
700 return sock
->ops
->mmap(file
, sock
, vma
);
703 int sock_close(struct inode
*inode
, struct file
*filp
)
706 * It was possible the inode is NULL we were
707 * closing an unfinished socket.
712 printk(KERN_DEBUG
"sock_close: NULL inode\n");
715 sock_fasync(-1, filp
, 0);
716 sock_release(socki_lookup(inode
));
721 * Update the socket async list
723 * Fasync_list locking strategy.
725 * 1. fasync_list is modified only under process context socket lock
726 * i.e. under semaphore.
727 * 2. fasync_list is used under read_lock(&sk->callback_lock)
728 * or under socket lock.
729 * 3. fasync_list can be used from softirq context, so that
730 * modification under socket lock have to be enhanced with
731 * write_lock_bh(&sk->callback_lock).
735 static int sock_fasync(int fd
, struct file
*filp
, int on
)
737 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
743 fna
=(struct fasync_struct
*)kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
748 sock
= socki_lookup(filp
->f_dentry
->d_inode
);
750 if ((sk
=sock
->sk
) == NULL
) {
758 prev
=&(sock
->fasync_list
);
760 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
761 if (fa
->fa_file
==filp
)
768 write_lock_bh(&sk
->callback_lock
);
770 write_unlock_bh(&sk
->callback_lock
);
777 fna
->magic
=FASYNC_MAGIC
;
778 fna
->fa_next
=sock
->fasync_list
;
779 write_lock_bh(&sk
->callback_lock
);
780 sock
->fasync_list
=fna
;
781 write_unlock_bh(&sk
->callback_lock
);
787 write_lock_bh(&sk
->callback_lock
);
789 write_unlock_bh(&sk
->callback_lock
);
795 release_sock(sock
->sk
);
799 /* This function may be called only under socket lock or callback_lock */
801 int sock_wake_async(struct socket
*sock
, int how
, int band
)
803 if (!sock
|| !sock
->fasync_list
)
809 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
813 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
818 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
821 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
827 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
833 * Check protocol is in range
835 if (family
< 0 || family
>= NPROTO
)
836 return -EAFNOSUPPORT
;
837 if (type
< 0 || type
>= SOCK_MAX
)
842 This uglymoron is moved from INET layer to here to avoid
843 deadlock in module load.
845 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
849 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n", current
->comm
);
854 #if defined(CONFIG_KMOD) && defined(CONFIG_NET)
855 /* Attempt to load a protocol module if the find failed.
857 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
858 * requested real, full-featured networking support upon configuration.
859 * Otherwise module support will break!
861 if (net_families
[family
]==NULL
)
863 char module_name
[30];
864 sprintf(module_name
,"net-pf-%d",family
);
865 request_module(module_name
);
869 net_family_read_lock();
870 if (net_families
[family
] == NULL
) {
876 * Allocate the socket and allow the family to set things up. if
877 * the protocol is 0, the family is instructed to select an appropriate
881 if (!(sock
= sock_alloc()))
883 printk(KERN_WARNING
"socket: no more sockets\n");
884 i
= -ENFILE
; /* Not exactly a match, but its the
885 closest posix thing */
891 if ((i
= net_families
[family
]->create(sock
, protocol
)) < 0)
900 net_family_read_unlock();
904 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
909 retval
= sock_create(family
, type
, protocol
, &sock
);
913 retval
= sock_map_fd(sock
);
918 /* It may be already another descriptor 8) Not kernel problem. */
927 * Create a pair of connected sockets.
930 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
, int usockvec
[2])
932 struct socket
*sock1
, *sock2
;
936 * Obtain the first socket and check if the underlying protocol
937 * supports the socketpair call.
940 err
= sock_create(family
, type
, protocol
, &sock1
);
944 err
= sock_create(family
, type
, protocol
, &sock2
);
948 err
= sock1
->ops
->socketpair(sock1
, sock2
);
950 goto out_release_both
;
954 err
= sock_map_fd(sock1
);
956 goto out_release_both
;
959 err
= sock_map_fd(sock2
);
964 /* fd1 and fd2 may be already another descriptors.
965 * Not kernel problem.
968 err
= put_user(fd1
, &usockvec
[0]);
970 err
= put_user(fd2
, &usockvec
[1]);
993 * Bind a name to a socket. Nothing much to do here since it's
994 * the protocol's responsibility to handle the local address.
996 * We move the socket address to kernel space before we call
997 * the protocol layer (having also checked the address is ok).
1000 asmlinkage
long sys_bind(int fd
, struct sockaddr
*umyaddr
, int addrlen
)
1002 struct socket
*sock
;
1003 char address
[MAX_SOCK_ADDR
];
1006 if((sock
= sockfd_lookup(fd
,&err
))!=NULL
)
1008 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0)
1009 err
= sock
->ops
->bind(sock
, (struct sockaddr
*)address
, addrlen
);
1017 * Perform a listen. Basically, we allow the protocol to do anything
1018 * necessary for a listen, and if that works, we mark the socket as
1019 * ready for listening.
1022 asmlinkage
long sys_listen(int fd
, int backlog
)
1024 struct socket
*sock
;
1027 if ((sock
= sockfd_lookup(fd
, &err
)) != NULL
) {
1028 if ((unsigned) backlog
> SOMAXCONN
)
1029 backlog
= SOMAXCONN
;
1030 err
=sock
->ops
->listen(sock
, backlog
);
1038 * For accept, we attempt to create a new socket, set up the link
1039 * with the client, wake up the client, then return the new
1040 * connected fd. We collect the address of the connector in kernel
1041 * space and move it to user at the very end. This is unclean because
1042 * we open the socket then return an error.
1044 * 1003.1g adds the ability to recvmsg() to query connection pending
1045 * status to recvmsg. We need to add that support in a way thats
1046 * clean when we restucture accept also.
1049 asmlinkage
long sys_accept(int fd
, struct sockaddr
*upeer_sockaddr
, int *upeer_addrlen
)
1051 struct socket
*sock
, *newsock
;
1053 char address
[MAX_SOCK_ADDR
];
1055 sock
= sockfd_lookup(fd
, &err
);
1060 if (!(newsock
= sock_alloc()))
1063 newsock
->type
= sock
->type
;
1064 newsock
->ops
= sock
->ops
;
1066 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1070 if (upeer_sockaddr
) {
1071 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 2)<0) {
1072 err
= -ECONNABORTED
;
1075 err
= move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
1080 /* File flags are not inherited via accept() unlike another OSes. */
1082 if ((err
= sock_map_fd(newsock
)) < 0)
1091 sock_release(newsock
);
1097 * Attempt to connect to a socket with the server address. The address
1098 * is in user space so we verify it is OK and move it to kernel space.
1100 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1103 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1104 * other SEQPACKET protocols that take time to connect() as it doesn't
1105 * include the -EINPROGRESS status for such sockets.
1108 asmlinkage
long sys_connect(int fd
, struct sockaddr
*uservaddr
, int addrlen
)
1110 struct socket
*sock
;
1111 char address
[MAX_SOCK_ADDR
];
1114 sock
= sockfd_lookup(fd
, &err
);
1117 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1120 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
1121 sock
->file
->f_flags
);
1129 * Get the local address ('name') of a socket object. Move the obtained
1130 * name to user space.
1133 asmlinkage
long sys_getsockname(int fd
, struct sockaddr
*usockaddr
, int *usockaddr_len
)
1135 struct socket
*sock
;
1136 char address
[MAX_SOCK_ADDR
];
1139 sock
= sockfd_lookup(fd
, &err
);
1142 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1145 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1154 * Get the remote address ('name') of a socket object. Move the obtained
1155 * name to user space.
1158 asmlinkage
long sys_getpeername(int fd
, struct sockaddr
*usockaddr
, int *usockaddr_len
)
1160 struct socket
*sock
;
1161 char address
[MAX_SOCK_ADDR
];
1164 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1166 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
1168 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
1175 * Send a datagram to a given address. We move the address into kernel
1176 * space and check the user space data area is readable before invoking
1180 asmlinkage
long sys_sendto(int fd
, void * buff
, size_t len
, unsigned flags
,
1181 struct sockaddr
*addr
, int addr_len
)
1183 struct socket
*sock
;
1184 char address
[MAX_SOCK_ADDR
];
1189 sock
= sockfd_lookup(fd
, &err
);
1197 msg
.msg_control
=NULL
;
1198 msg
.msg_controllen
=0;
1202 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1205 msg
.msg_name
=address
;
1206 msg
.msg_namelen
=addr_len
;
1208 if (sock
->file
->f_flags
& O_NONBLOCK
)
1209 flags
|= MSG_DONTWAIT
;
1210 msg
.msg_flags
= flags
;
1211 err
= sock_sendmsg(sock
, &msg
, len
);
1220 * Send a datagram down a socket.
1223 asmlinkage
long sys_send(int fd
, void * buff
, size_t len
, unsigned flags
)
1225 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1229 * Receive a frame from the socket and optionally record the address of the
1230 * sender. We verify the buffers are writable and if needed move the
1231 * sender address from kernel to user space.
1234 asmlinkage
long sys_recvfrom(int fd
, void * ubuf
, size_t size
, unsigned flags
,
1235 struct sockaddr
*addr
, int *addr_len
)
1237 struct socket
*sock
;
1240 char address
[MAX_SOCK_ADDR
];
1243 sock
= sockfd_lookup(fd
, &err
);
1247 msg
.msg_control
=NULL
;
1248 msg
.msg_controllen
=0;
1253 msg
.msg_name
=address
;
1254 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1255 if (sock
->file
->f_flags
& O_NONBLOCK
)
1256 flags
|= MSG_DONTWAIT
;
1257 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1259 if(err
>= 0 && addr
!= NULL
&& msg
.msg_namelen
)
1261 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1271 * Receive a datagram from a socket.
1274 asmlinkage
long sys_recv(int fd
, void * ubuf
, size_t size
, unsigned flags
)
1276 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1280 * Set a socket option. Because we don't know the option lengths we have
1281 * to pass the user mode parameter for the protocols to sort out.
1284 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
, char *optval
, int optlen
)
1287 struct socket
*sock
;
1292 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1294 if (level
== SOL_SOCKET
)
1295 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1297 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1304 * Get a socket option. Because we don't know the option lengths we have
1305 * to pass a user mode parameter for the protocols to sort out.
1308 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
, char *optval
, int *optlen
)
1311 struct socket
*sock
;
1313 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1315 if (level
== SOL_SOCKET
)
1316 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1318 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1326 * Shutdown a socket.
1329 asmlinkage
long sys_shutdown(int fd
, int how
)
1332 struct socket
*sock
;
1334 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1336 err
=sock
->ops
->shutdown(sock
, how
);
1343 * BSD sendmsg interface
1346 asmlinkage
long sys_sendmsg(int fd
, struct msghdr
*msg
, unsigned flags
)
1348 struct socket
*sock
;
1349 char address
[MAX_SOCK_ADDR
];
1350 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1351 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]; /* 20 is size of ipv6_pktinfo */
1352 unsigned char *ctl_buf
= ctl
;
1353 struct msghdr msg_sys
;
1354 int err
, ctl_len
, iov_size
, total_len
;
1357 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1360 sock
= sockfd_lookup(fd
, &err
);
1364 /* do not move before msg_sys is valid */
1366 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1369 /* Check whether to allocate the iovec area*/
1371 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1372 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1373 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1378 /* This will also move the address data into kernel space */
1379 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1386 if (msg_sys
.msg_controllen
> INT_MAX
)
1388 ctl_len
= msg_sys
.msg_controllen
;
1391 if (ctl_len
> sizeof(ctl
))
1393 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1394 if (ctl_buf
== NULL
)
1398 if (copy_from_user(ctl_buf
, msg_sys
.msg_control
, ctl_len
))
1400 msg_sys
.msg_control
= ctl_buf
;
1402 msg_sys
.msg_flags
= flags
;
1404 if (sock
->file
->f_flags
& O_NONBLOCK
)
1405 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1406 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1410 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1412 if (iov
!= iovstack
)
1413 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1421 * BSD recvmsg interface
1424 asmlinkage
long sys_recvmsg(int fd
, struct msghdr
*msg
, unsigned int flags
)
1426 struct socket
*sock
;
1427 struct iovec iovstack
[UIO_FASTIOV
];
1428 struct iovec
*iov
=iovstack
;
1429 struct msghdr msg_sys
;
1430 unsigned long cmsg_ptr
;
1431 int err
, iov_size
, total_len
, len
;
1433 /* kernel mode address */
1434 char addr
[MAX_SOCK_ADDR
];
1436 /* user mode address pointers */
1437 struct sockaddr
*uaddr
;
1441 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1444 sock
= sockfd_lookup(fd
, &err
);
1449 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1452 /* Check whether to allocate the iovec area*/
1454 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1455 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1456 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1462 * Save the user-mode address (verify_iovec will change the
1463 * kernel msghdr to use the kernel address space)
1466 uaddr
= msg_sys
.msg_name
;
1467 uaddr_len
= &msg
->msg_namelen
;
1468 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1473 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1474 msg_sys
.msg_flags
= 0;
1476 if (sock
->file
->f_flags
& O_NONBLOCK
)
1477 flags
|= MSG_DONTWAIT
;
1478 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1483 if (uaddr
!= NULL
&& msg_sys
.msg_namelen
) {
1484 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1488 err
= __put_user(msg_sys
.msg_flags
, &msg
->msg_flags
);
1491 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1492 &msg
->msg_controllen
);
1498 if (iov
!= iovstack
)
1499 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1508 * Perform a file control on a socket file descriptor.
1510 * Doesn't acquire a fd lock, because no network fcntl
1511 * function sleeps currently.
1514 int sock_fcntl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
1516 struct socket
*sock
;
1518 sock
= socki_lookup (filp
->f_dentry
->d_inode
);
1519 if (sock
&& sock
->ops
)
1520 return sock_no_fcntl(sock
, cmd
, arg
);
1524 /* Argument list sizes for sys_socketcall */
1525 #define AL(x) ((x) * sizeof(unsigned long))
1526 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1527 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1528 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1532 * System call vectors.
1534 * Argument checking cleaned up. Saved 20% in size.
1535 * This function doesn't need to set the kernel lock because
1536 * it is set by the callees.
1539 asmlinkage
long sys_socketcall(int call
, unsigned long *args
)
1542 unsigned long a0
,a1
;
1545 if(call
<1||call
>SYS_RECVMSG
)
1548 /* copy_from_user should be SMP safe. */
1549 if (copy_from_user(a
, args
, nargs
[call
]))
1558 err
= sys_socket(a0
,a1
,a
[2]);
1561 err
= sys_bind(a0
,(struct sockaddr
*)a1
, a
[2]);
1564 err
= sys_connect(a0
, (struct sockaddr
*)a1
, a
[2]);
1567 err
= sys_listen(a0
,a1
);
1570 err
= sys_accept(a0
,(struct sockaddr
*)a1
, (int *)a
[2]);
1572 case SYS_GETSOCKNAME
:
1573 err
= sys_getsockname(a0
,(struct sockaddr
*)a1
, (int *)a
[2]);
1575 case SYS_GETPEERNAME
:
1576 err
= sys_getpeername(a0
, (struct sockaddr
*)a1
, (int *)a
[2]);
1578 case SYS_SOCKETPAIR
:
1579 err
= sys_socketpair(a0
,a1
, a
[2], (int *)a
[3]);
1582 err
= sys_send(a0
, (void *)a1
, a
[2], a
[3]);
1585 err
= sys_sendto(a0
,(void *)a1
, a
[2], a
[3],
1586 (struct sockaddr
*)a
[4], a
[5]);
1589 err
= sys_recv(a0
, (void *)a1
, a
[2], a
[3]);
1592 err
= sys_recvfrom(a0
, (void *)a1
, a
[2], a
[3],
1593 (struct sockaddr
*)a
[4], (int *)a
[5]);
1596 err
= sys_shutdown(a0
,a1
);
1598 case SYS_SETSOCKOPT
:
1599 err
= sys_setsockopt(a0
, a1
, a
[2], (char *)a
[3], a
[4]);
1601 case SYS_GETSOCKOPT
:
1602 err
= sys_getsockopt(a0
, a1
, a
[2], (char *)a
[3], (int *)a
[4]);
1605 err
= sys_sendmsg(a0
, (struct msghdr
*) a1
, a
[2]);
1608 err
= sys_recvmsg(a0
, (struct msghdr
*) a1
, a
[2]);
1618 * This function is called by a protocol handler that wants to
1619 * advertise its address family, and have it linked into the
1623 int sock_register(struct net_proto_family
*ops
)
1627 if (ops
->family
>= NPROTO
) {
1628 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
1631 net_family_write_lock();
1633 if (net_families
[ops
->family
] == NULL
) {
1634 net_families
[ops
->family
]=ops
;
1637 net_family_write_unlock();
1642 * This function is called by a protocol handler that wants to
1643 * remove its address family, and have it unlinked from the
1647 int sock_unregister(int family
)
1649 if (family
< 0 || family
>= NPROTO
)
1652 net_family_write_lock();
1653 net_families
[family
]=NULL
;
1654 net_family_write_unlock();
1659 extern void sk_init(void);
1661 #ifdef CONFIG_WAN_ROUTER
1662 extern void wanrouter_init(void);
1666 extern void bluez_init(void);
1669 void __init
sock_init(void)
1673 printk(KERN_INFO
"Linux NET4.0 for Linux 2.4\n");
1674 printk(KERN_INFO
"Based upon Swansea University Computer Society NET3.039\n");
1677 * Initialize all address (protocol) families.
1680 for (i
= 0; i
< NPROTO
; i
++)
1681 net_families
[i
] = NULL
;
1684 * Initialize sock SLAB cache.
1691 * Initialize skbuff SLAB cache
1700 #ifdef CONFIG_WAN_ROUTER
1705 * Initialize the protocols module.
1708 register_filesystem(&sock_fs_type
);
1709 sock_mnt
= kern_mount(&sock_fs_type
);
1710 /* The real protocol initialization is performed when
1711 * do_initcalls is run.
1716 * The netlink device handler may be needed early.
1722 #ifdef CONFIG_NETLINK_DEV
1725 #ifdef CONFIG_NETFILTER
1734 int socket_get_info(char *buffer
, char **start
, off_t offset
, int length
)
1739 for (cpu
=0; cpu
<smp_num_cpus
; cpu
++)
1740 counter
+= sockets_in_use
[cpu_logical_map(cpu
)].counter
;
1742 /* It can be negative, by the way. 8) */
1746 len
= sprintf(buffer
, "sockets: used %d\n", counter
);
1752 *start
= buffer
+ offset
;