Root/net/socket.c

1/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
56 * paradigm.
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
61#include <linux/mm.h>
62#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
66#include <linux/thread_info.h>
67#include <linux/rcupdate.h>
68#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
71#include <linux/mutex.h>
72#include <linux/wanrouter.h>
73#include <linux/if_bridge.h>
74#include <linux/if_frad.h>
75#include <linux/if_vlan.h>
76#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
81#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
86#include <linux/audit.h>
87#include <linux/wireless.h>
88#include <linux/nsproxy.h>
89#include <linux/magic.h>
90#include <linux/slab.h>
91
92#include <asm/uaccess.h>
93#include <asm/unistd.h>
94
95#include <net/compat.h>
96#include <net/wext.h>
97
98#include <net/sock.h>
99#include <linux/netfilter.h>
100
101#include <linux/if_tun.h>
102#include <linux/ipv6_route.h>
103#include <linux/route.h>
104#include <linux/sockios.h>
105#include <linux/atalk.h>
106
107static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
108static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
109             unsigned long nr_segs, loff_t pos);
110static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
111              unsigned long nr_segs, loff_t pos);
112static int sock_mmap(struct file *file, struct vm_area_struct *vma);
113
114static int sock_close(struct inode *inode, struct file *file);
115static unsigned int sock_poll(struct file *file,
116                  struct poll_table_struct *wait);
117static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
118#ifdef CONFIG_COMPAT
119static long compat_sock_ioctl(struct file *file,
120                  unsigned int cmd, unsigned long arg);
121#endif
122static int sock_fasync(int fd, struct file *filp, int on);
123static ssize_t sock_sendpage(struct file *file, struct page *page,
124                 int offset, size_t size, loff_t *ppos, int more);
125static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
126                    struct pipe_inode_info *pipe, size_t len,
127                unsigned int flags);
128
129/*
130 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
131 * in the operation structures but are done directly via the socketcall() multiplexor.
132 */
133
134static const struct file_operations socket_file_ops = {
135    .owner = THIS_MODULE,
136    .llseek = no_llseek,
137    .aio_read = sock_aio_read,
138    .aio_write = sock_aio_write,
139    .poll = sock_poll,
140    .unlocked_ioctl = sock_ioctl,
141#ifdef CONFIG_COMPAT
142    .compat_ioctl = compat_sock_ioctl,
143#endif
144    .mmap = sock_mmap,
145    .open = sock_no_open, /* special open code to disallow open via /proc */
146    .release = sock_close,
147    .fasync = sock_fasync,
148    .sendpage = sock_sendpage,
149    .splice_write = generic_splice_sendpage,
150    .splice_read = sock_splice_read,
151};
152
153/*
154 * The protocol list. Each protocol is registered in here.
155 */
156
157static DEFINE_SPINLOCK(net_family_lock);
158static const struct net_proto_family *net_families[NPROTO] __read_mostly;
159
160/*
161 * Statistics counters of the socket lists
162 */
163
164static DEFINE_PER_CPU(int, sockets_in_use) = 0;
165
166/*
167 * Support routines.
168 * Move socket addresses back and forth across the kernel/user
169 * divide and look after the messy bits.
170 */
171
172#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
173                       16 for IP, 16 for IPX,
174                       24 for IPv6,
175                       about 80 for AX.25
176                       must be at least one bigger than
177                       the AF_UNIX size (see net/unix/af_unix.c
178                       :unix_mkname()).
179                     */
180
181/**
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
186 *
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
190 */
191
192int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
193{
194    if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
195        return -EINVAL;
196    if (ulen == 0)
197        return 0;
198    if (copy_from_user(kaddr, uaddr, ulen))
199        return -EFAULT;
200    return audit_sockaddr(ulen, kaddr);
201}
202
203/**
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
209 *
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
214 * accessible.
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
218 */
219
220int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
221              int __user *ulen)
222{
223    int err;
224    int len;
225
226    err = get_user(len, ulen);
227    if (err)
228        return err;
229    if (len > klen)
230        len = klen;
231    if (len < 0 || len > sizeof(struct sockaddr_storage))
232        return -EINVAL;
233    if (len) {
234        if (audit_sockaddr(klen, kaddr))
235            return -ENOMEM;
236        if (copy_to_user(uaddr, kaddr, len))
237            return -EFAULT;
238    }
239    /*
240     * "fromlen shall refer to the value before truncation.."
241     * 1003.1g
242     */
243    return __put_user(klen, ulen);
244}
245
246static struct kmem_cache *sock_inode_cachep __read_mostly;
247
248static struct inode *sock_alloc_inode(struct super_block *sb)
249{
250    struct socket_alloc *ei;
251
252    ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
253    if (!ei)
254        return NULL;
255    init_waitqueue_head(&ei->socket.wait);
256
257    ei->socket.fasync_list = NULL;
258    ei->socket.state = SS_UNCONNECTED;
259    ei->socket.flags = 0;
260    ei->socket.ops = NULL;
261    ei->socket.sk = NULL;
262    ei->socket.file = NULL;
263
264    return &ei->vfs_inode;
265}
266
267static void sock_destroy_inode(struct inode *inode)
268{
269    kmem_cache_free(sock_inode_cachep,
270            container_of(inode, struct socket_alloc, vfs_inode));
271}
272
273static void init_once(void *foo)
274{
275    struct socket_alloc *ei = (struct socket_alloc *)foo;
276
277    inode_init_once(&ei->vfs_inode);
278}
279
280static int init_inodecache(void)
281{
282    sock_inode_cachep = kmem_cache_create("sock_inode_cache",
283                          sizeof(struct socket_alloc),
284                          0,
285                          (SLAB_HWCACHE_ALIGN |
286                           SLAB_RECLAIM_ACCOUNT |
287                           SLAB_MEM_SPREAD),
288                          init_once);
289    if (sock_inode_cachep == NULL)
290        return -ENOMEM;
291    return 0;
292}
293
294static const struct super_operations sockfs_ops = {
295    .alloc_inode = sock_alloc_inode,
296    .destroy_inode =sock_destroy_inode,
297    .statfs = simple_statfs,
298};
299
300static int sockfs_get_sb(struct file_system_type *fs_type,
301             int flags, const char *dev_name, void *data,
302             struct vfsmount *mnt)
303{
304    return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
305                 mnt);
306}
307
308static struct vfsmount *sock_mnt __read_mostly;
309
310static struct file_system_type sock_fs_type = {
311    .name = "sockfs",
312    .get_sb = sockfs_get_sb,
313    .kill_sb = kill_anon_super,
314};
315
316/*
317 * sockfs_dname() is called from d_path().
318 */
319static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
320{
321    return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
322                dentry->d_inode->i_ino);
323}
324
325static const struct dentry_operations sockfs_dentry_operations = {
326    .d_dname = sockfs_dname,
327};
328
329/*
330 * Obtains the first available file descriptor and sets it up for use.
331 *
332 * These functions create file structures and maps them to fd space
333 * of the current process. On success it returns file descriptor
334 * and file struct implicitly stored in sock->file.
335 * Note that another thread may close file descriptor before we return
336 * from this function. We use the fact that now we do not refer
337 * to socket after mapping. If one day we will need it, this
338 * function will increment ref. count on file by 1.
339 *
340 * In any case returned fd MAY BE not valid!
341 * This race condition is unavoidable
342 * with shared fd spaces, we cannot solve it inside kernel,
343 * but we take care of internal coherence yet.
344 */
345
346static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
347{
348    struct qstr name = { .name = "" };
349    struct path path;
350    struct file *file;
351    int fd;
352
353    fd = get_unused_fd_flags(flags);
354    if (unlikely(fd < 0))
355        return fd;
356
357    path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
358    if (unlikely(!path.dentry)) {
359        put_unused_fd(fd);
360        return -ENOMEM;
361    }
362    path.mnt = mntget(sock_mnt);
363
364    path.dentry->d_op = &sockfs_dentry_operations;
365    d_instantiate(path.dentry, SOCK_INODE(sock));
366    SOCK_INODE(sock)->i_fop = &socket_file_ops;
367
368    file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
369          &socket_file_ops);
370    if (unlikely(!file)) {
371        /* drop dentry, keep inode */
372        atomic_inc(&path.dentry->d_inode->i_count);
373        path_put(&path);
374        put_unused_fd(fd);
375        return -ENFILE;
376    }
377
378    sock->file = file;
379    file->f_flags = O_RDWR | (flags & O_NONBLOCK);
380    file->f_pos = 0;
381    file->private_data = sock;
382
383    *f = file;
384    return fd;
385}
386
387int sock_map_fd(struct socket *sock, int flags)
388{
389    struct file *newfile;
390    int fd = sock_alloc_file(sock, &newfile, flags);
391
392    if (likely(fd >= 0))
393        fd_install(fd, newfile);
394
395    return fd;
396}
397
398static struct socket *sock_from_file(struct file *file, int *err)
399{
400    if (file->f_op == &socket_file_ops)
401        return file->private_data; /* set in sock_map_fd */
402
403    *err = -ENOTSOCK;
404    return NULL;
405}
406
407/**
408 * sockfd_lookup - Go from a file number to its socket slot
409 * @fd: file handle
410 * @err: pointer to an error code return
411 *
412 * The file handle passed in is locked and the socket it is bound
413 * too is returned. If an error occurs the err pointer is overwritten
414 * with a negative errno code and NULL is returned. The function checks
415 * for both invalid handles and passing a handle which is not a socket.
416 *
417 * On a success the socket object pointer is returned.
418 */
419
420struct socket *sockfd_lookup(int fd, int *err)
421{
422    struct file *file;
423    struct socket *sock;
424
425    file = fget(fd);
426    if (!file) {
427        *err = -EBADF;
428        return NULL;
429    }
430
431    sock = sock_from_file(file, err);
432    if (!sock)
433        fput(file);
434    return sock;
435}
436
437static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
438{
439    struct file *file;
440    struct socket *sock;
441
442    *err = -EBADF;
443    file = fget_light(fd, fput_needed);
444    if (file) {
445        sock = sock_from_file(file, err);
446        if (sock)
447            return sock;
448        fput_light(file, *fput_needed);
449    }
450    return NULL;
451}
452
453/**
454 * sock_alloc - allocate a socket
455 *
456 * Allocate a new inode and socket object. The two are bound together
457 * and initialised. The socket is then returned. If we are out of inodes
458 * NULL is returned.
459 */
460
461static struct socket *sock_alloc(void)
462{
463    struct inode *inode;
464    struct socket *sock;
465
466    inode = new_inode(sock_mnt->mnt_sb);
467    if (!inode)
468        return NULL;
469
470    sock = SOCKET_I(inode);
471
472    kmemcheck_annotate_bitfield(sock, type);
473    inode->i_mode = S_IFSOCK | S_IRWXUGO;
474    inode->i_uid = current_fsuid();
475    inode->i_gid = current_fsgid();
476
477    percpu_add(sockets_in_use, 1);
478    return sock;
479}
480
481/*
482 * In theory you can't get an open on this inode, but /proc provides
483 * a back door. Remember to keep it shut otherwise you'll let the
484 * creepy crawlies in.
485 */
486
487static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
488{
489    return -ENXIO;
490}
491
492const struct file_operations bad_sock_fops = {
493    .owner = THIS_MODULE,
494    .open = sock_no_open,
495};
496
497/**
498 * sock_release - close a socket
499 * @sock: socket to close
500 *
501 * The socket is released from the protocol stack if it has a release
502 * callback, and the inode is then released if the socket is bound to
503 * an inode not a file.
504 */
505
506void sock_release(struct socket *sock)
507{
508    if (sock->ops) {
509        struct module *owner = sock->ops->owner;
510
511        sock->ops->release(sock);
512        sock->ops = NULL;
513        module_put(owner);
514    }
515
516    if (sock->fasync_list)
517        printk(KERN_ERR "sock_release: fasync list not empty!\n");
518
519    percpu_sub(sockets_in_use, 1);
520    if (!sock->file) {
521        iput(SOCK_INODE(sock));
522        return;
523    }
524    sock->file = NULL;
525}
526
527int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
528              union skb_shared_tx *shtx)
529{
530    shtx->flags = 0;
531    if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
532        shtx->hardware = 1;
533    if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
534        shtx->software = 1;
535    return 0;
536}
537EXPORT_SYMBOL(sock_tx_timestamp);
538
539static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
540                 struct msghdr *msg, size_t size)
541{
542    struct sock_iocb *si = kiocb_to_siocb(iocb);
543    int err;
544
545    si->sock = sock;
546    si->scm = NULL;
547    si->msg = msg;
548    si->size = size;
549
550    err = security_socket_sendmsg(sock, msg, size);
551    if (err)
552        return err;
553
554    return sock->ops->sendmsg(iocb, sock, msg, size);
555}
556
557int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
558{
559    struct kiocb iocb;
560    struct sock_iocb siocb;
561    int ret;
562
563    init_sync_kiocb(&iocb, NULL);
564    iocb.private = &siocb;
565    ret = __sock_sendmsg(&iocb, sock, msg, size);
566    if (-EIOCBQUEUED == ret)
567        ret = wait_on_sync_kiocb(&iocb);
568    return ret;
569}
570
571int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
572           struct kvec *vec, size_t num, size_t size)
573{
574    mm_segment_t oldfs = get_fs();
575    int result;
576
577    set_fs(KERNEL_DS);
578    /*
579     * the following is safe, since for compiler definitions of kvec and
580     * iovec are identical, yielding the same in-core layout and alignment
581     */
582    msg->msg_iov = (struct iovec *)vec;
583    msg->msg_iovlen = num;
584    result = sock_sendmsg(sock, msg, size);
585    set_fs(oldfs);
586    return result;
587}
588
589static int ktime2ts(ktime_t kt, struct timespec *ts)
590{
591    if (kt.tv64) {
592        *ts = ktime_to_timespec(kt);
593        return 1;
594    } else {
595        return 0;
596    }
597}
598
599/*
600 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
601 */
602void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
603    struct sk_buff *skb)
604{
605    int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
606    struct timespec ts[3];
607    int empty = 1;
608    struct skb_shared_hwtstamps *shhwtstamps =
609        skb_hwtstamps(skb);
610
611    /* Race occurred between timestamp enabling and packet
612       receiving. Fill in the current time for now. */
613    if (need_software_tstamp && skb->tstamp.tv64 == 0)
614        __net_timestamp(skb);
615
616    if (need_software_tstamp) {
617        if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
618            struct timeval tv;
619            skb_get_timestamp(skb, &tv);
620            put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
621                 sizeof(tv), &tv);
622        } else {
623            struct timespec ts;
624            skb_get_timestampns(skb, &ts);
625            put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
626                 sizeof(ts), &ts);
627        }
628    }
629
630
631    memset(ts, 0, sizeof(ts));
632    if (skb->tstamp.tv64 &&
633        sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
634        skb_get_timestampns(skb, ts + 0);
635        empty = 0;
636    }
637    if (shhwtstamps) {
638        if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
639            ktime2ts(shhwtstamps->syststamp, ts + 1))
640            empty = 0;
641        if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
642            ktime2ts(shhwtstamps->hwtstamp, ts + 2))
643            empty = 0;
644    }
645    if (!empty)
646        put_cmsg(msg, SOL_SOCKET,
647             SCM_TIMESTAMPING, sizeof(ts), &ts);
648}
649
650EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
651
652inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
653{
654    if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
655        put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
656            sizeof(__u32), &skb->dropcount);
657}
658
659void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
660    struct sk_buff *skb)
661{
662    sock_recv_timestamp(msg, sk, skb);
663    sock_recv_drops(msg, sk, skb);
664}
665EXPORT_SYMBOL_GPL(sock_recv_ts_and_drops);
666
667static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
668                       struct msghdr *msg, size_t size, int flags)
669{
670    struct sock_iocb *si = kiocb_to_siocb(iocb);
671
672    si->sock = sock;
673    si->scm = NULL;
674    si->msg = msg;
675    si->size = size;
676    si->flags = flags;
677
678    return sock->ops->recvmsg(iocb, sock, msg, size, flags);
679}
680
681static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
682                 struct msghdr *msg, size_t size, int flags)
683{
684    int err = security_socket_recvmsg(sock, msg, size, flags);
685
686    return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
687}
688
689int sock_recvmsg(struct socket *sock, struct msghdr *msg,
690         size_t size, int flags)
691{
692    struct kiocb iocb;
693    struct sock_iocb siocb;
694    int ret;
695
696    init_sync_kiocb(&iocb, NULL);
697    iocb.private = &siocb;
698    ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
699    if (-EIOCBQUEUED == ret)
700        ret = wait_on_sync_kiocb(&iocb);
701    return ret;
702}
703
704static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
705                  size_t size, int flags)
706{
707    struct kiocb iocb;
708    struct sock_iocb siocb;
709    int ret;
710
711    init_sync_kiocb(&iocb, NULL);
712    iocb.private = &siocb;
713    ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
714    if (-EIOCBQUEUED == ret)
715        ret = wait_on_sync_kiocb(&iocb);
716    return ret;
717}
718
719int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
720           struct kvec *vec, size_t num, size_t size, int flags)
721{
722    mm_segment_t oldfs = get_fs();
723    int result;
724
725    set_fs(KERNEL_DS);
726    /*
727     * the following is safe, since for compiler definitions of kvec and
728     * iovec are identical, yielding the same in-core layout and alignment
729     */
730    msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
731    result = sock_recvmsg(sock, msg, size, flags);
732    set_fs(oldfs);
733    return result;
734}
735
736static void sock_aio_dtor(struct kiocb *iocb)
737{
738    kfree(iocb->private);
739}
740
741static ssize_t sock_sendpage(struct file *file, struct page *page,
742                 int offset, size_t size, loff_t *ppos, int more)
743{
744    struct socket *sock;
745    int flags;
746
747    sock = file->private_data;
748
749    flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
750    if (more)
751        flags |= MSG_MORE;
752
753    return kernel_sendpage(sock, page, offset, size, flags);
754}
755
756static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
757                    struct pipe_inode_info *pipe, size_t len,
758                unsigned int flags)
759{
760    struct socket *sock = file->private_data;
761
762    if (unlikely(!sock->ops->splice_read))
763        return -EINVAL;
764
765    return sock->ops->splice_read(sock, ppos, pipe, len, flags);
766}
767
768static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
769                     struct sock_iocb *siocb)
770{
771    if (!is_sync_kiocb(iocb)) {
772        siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
773        if (!siocb)
774            return NULL;
775        iocb->ki_dtor = sock_aio_dtor;
776    }
777
778    siocb->kiocb = iocb;
779    iocb->private = siocb;
780    return siocb;
781}
782
783static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
784        struct file *file, const struct iovec *iov,
785        unsigned long nr_segs)
786{
787    struct socket *sock = file->private_data;
788    size_t size = 0;
789    int i;
790
791    for (i = 0; i < nr_segs; i++)
792        size += iov[i].iov_len;
793
794    msg->msg_name = NULL;
795    msg->msg_namelen = 0;
796    msg->msg_control = NULL;
797    msg->msg_controllen = 0;
798    msg->msg_iov = (struct iovec *)iov;
799    msg->msg_iovlen = nr_segs;
800    msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
801
802    return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
803}
804
805static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
806                unsigned long nr_segs, loff_t pos)
807{
808    struct sock_iocb siocb, *x;
809
810    if (pos != 0)
811        return -ESPIPE;
812
813    if (iocb->ki_left == 0) /* Match SYS5 behaviour */
814        return 0;
815
816
817    x = alloc_sock_iocb(iocb, &siocb);
818    if (!x)
819        return -ENOMEM;
820    return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
821}
822
823static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
824            struct file *file, const struct iovec *iov,
825            unsigned long nr_segs)
826{
827    struct socket *sock = file->private_data;
828    size_t size = 0;
829    int i;
830
831    for (i = 0; i < nr_segs; i++)
832        size += iov[i].iov_len;
833
834    msg->msg_name = NULL;
835    msg->msg_namelen = 0;
836    msg->msg_control = NULL;
837    msg->msg_controllen = 0;
838    msg->msg_iov = (struct iovec *)iov;
839    msg->msg_iovlen = nr_segs;
840    msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
841    if (sock->type == SOCK_SEQPACKET)
842        msg->msg_flags |= MSG_EOR;
843
844    return __sock_sendmsg(iocb, sock, msg, size);
845}
846
847static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
848              unsigned long nr_segs, loff_t pos)
849{
850    struct sock_iocb siocb, *x;
851
852    if (pos != 0)
853        return -ESPIPE;
854
855    x = alloc_sock_iocb(iocb, &siocb);
856    if (!x)
857        return -ENOMEM;
858
859    return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
860}
861
862/*
863 * Atomic setting of ioctl hooks to avoid race
864 * with module unload.
865 */
866
867static DEFINE_MUTEX(br_ioctl_mutex);
868static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
869
870void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
871{
872    mutex_lock(&br_ioctl_mutex);
873    br_ioctl_hook = hook;
874    mutex_unlock(&br_ioctl_mutex);
875}
876
877EXPORT_SYMBOL(brioctl_set);
878
879static DEFINE_MUTEX(vlan_ioctl_mutex);
880static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
881
882void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
883{
884    mutex_lock(&vlan_ioctl_mutex);
885    vlan_ioctl_hook = hook;
886    mutex_unlock(&vlan_ioctl_mutex);
887}
888
889EXPORT_SYMBOL(vlan_ioctl_set);
890
891static DEFINE_MUTEX(dlci_ioctl_mutex);
892static int (*dlci_ioctl_hook) (unsigned int, void __user *);
893
894void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
895{
896    mutex_lock(&dlci_ioctl_mutex);
897    dlci_ioctl_hook = hook;
898    mutex_unlock(&dlci_ioctl_mutex);
899}
900
901EXPORT_SYMBOL(dlci_ioctl_set);
902
903static long sock_do_ioctl(struct net *net, struct socket *sock,
904                 unsigned int cmd, unsigned long arg)
905{
906    int err;
907    void __user *argp = (void __user *)arg;
908
909    err = sock->ops->ioctl(sock, cmd, arg);
910
911    /*
912     * If this ioctl is unknown try to hand it down
913     * to the NIC driver.
914     */
915    if (err == -ENOIOCTLCMD)
916        err = dev_ioctl(net, cmd, argp);
917
918    return err;
919}
920
921/*
922 * With an ioctl, arg may well be a user mode pointer, but we don't know
923 * what to do with it - that's up to the protocol still.
924 */
925
926static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
927{
928    struct socket *sock;
929    struct sock *sk;
930    void __user *argp = (void __user *)arg;
931    int pid, err;
932    struct net *net;
933
934    sock = file->private_data;
935    sk = sock->sk;
936    net = sock_net(sk);
937    if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
938        err = dev_ioctl(net, cmd, argp);
939    } else
940#ifdef CONFIG_WEXT_CORE
941    if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
942        err = dev_ioctl(net, cmd, argp);
943    } else
944#endif
945        switch (cmd) {
946        case FIOSETOWN:
947        case SIOCSPGRP:
948            err = -EFAULT;
949            if (get_user(pid, (int __user *)argp))
950                break;
951            err = f_setown(sock->file, pid, 1);
952            break;
953        case FIOGETOWN:
954        case SIOCGPGRP:
955            err = put_user(f_getown(sock->file),
956                       (int __user *)argp);
957            break;
958        case SIOCGIFBR:
959        case SIOCSIFBR:
960        case SIOCBRADDBR:
961        case SIOCBRDELBR:
962            err = -ENOPKG;
963            if (!br_ioctl_hook)
964                request_module("bridge");
965
966            mutex_lock(&br_ioctl_mutex);
967            if (br_ioctl_hook)
968                err = br_ioctl_hook(net, cmd, argp);
969            mutex_unlock(&br_ioctl_mutex);
970            break;
971        case SIOCGIFVLAN:
972        case SIOCSIFVLAN:
973            err = -ENOPKG;
974            if (!vlan_ioctl_hook)
975                request_module("8021q");
976
977            mutex_lock(&vlan_ioctl_mutex);
978            if (vlan_ioctl_hook)
979                err = vlan_ioctl_hook(net, argp);
980            mutex_unlock(&vlan_ioctl_mutex);
981            break;
982        case SIOCADDDLCI:
983        case SIOCDELDLCI:
984            err = -ENOPKG;
985            if (!dlci_ioctl_hook)
986                request_module("dlci");
987
988            mutex_lock(&dlci_ioctl_mutex);
989            if (dlci_ioctl_hook)
990                err = dlci_ioctl_hook(cmd, argp);
991            mutex_unlock(&dlci_ioctl_mutex);
992            break;
993        default:
994            err = sock_do_ioctl(net, sock, cmd, arg);
995            break;
996        }
997    return err;
998}
999
1000int sock_create_lite(int family, int type, int protocol, struct socket **res)
1001{
1002    int err;
1003    struct socket *sock = NULL;
1004
1005    err = security_socket_create(family, type, protocol, 1);
1006    if (err)
1007        goto out;
1008
1009    sock = sock_alloc();
1010    if (!sock) {
1011        err = -ENOMEM;
1012        goto out;
1013    }
1014
1015    sock->type = type;
1016    err = security_socket_post_create(sock, family, type, protocol, 1);
1017    if (err)
1018        goto out_release;
1019
1020out:
1021    *res = sock;
1022    return err;
1023out_release:
1024    sock_release(sock);
1025    sock = NULL;
1026    goto out;
1027}
1028
1029/* No kernel lock held - perfect */
1030static unsigned int sock_poll(struct file *file, poll_table *wait)
1031{
1032    struct socket *sock;
1033
1034    /*
1035     * We can't return errors to poll, so it's either yes or no.
1036     */
1037    sock = file->private_data;
1038    return sock->ops->poll(file, sock, wait);
1039}
1040
1041static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1042{
1043    struct socket *sock = file->private_data;
1044
1045    return sock->ops->mmap(file, sock, vma);
1046}
1047
1048static int sock_close(struct inode *inode, struct file *filp)
1049{
1050    /*
1051     * It was possible the inode is NULL we were
1052     * closing an unfinished socket.
1053     */
1054
1055    if (!inode) {
1056        printk(KERN_DEBUG "sock_close: NULL inode\n");
1057        return 0;
1058    }
1059    sock_release(SOCKET_I(inode));
1060    return 0;
1061}
1062
1063/*
1064 * Update the socket async list
1065 *
1066 * Fasync_list locking strategy.
1067 *
1068 * 1. fasync_list is modified only under process context socket lock
1069 * i.e. under semaphore.
1070 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1071 * or under socket lock.
1072 * 3. fasync_list can be used from softirq context, so that
1073 * modification under socket lock have to be enhanced with
1074 * write_lock_bh(&sk->sk_callback_lock).
1075 * --ANK (990710)
1076 */
1077
1078static int sock_fasync(int fd, struct file *filp, int on)
1079{
1080    struct fasync_struct *fa, *fna = NULL, **prev;
1081    struct socket *sock;
1082    struct sock *sk;
1083
1084    if (on) {
1085        fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1086        if (fna == NULL)
1087            return -ENOMEM;
1088    }
1089
1090    sock = filp->private_data;
1091
1092    sk = sock->sk;
1093    if (sk == NULL) {
1094        kfree(fna);
1095        return -EINVAL;
1096    }
1097
1098    lock_sock(sk);
1099
1100    spin_lock(&filp->f_lock);
1101    if (on)
1102        filp->f_flags |= FASYNC;
1103    else
1104        filp->f_flags &= ~FASYNC;
1105    spin_unlock(&filp->f_lock);
1106
1107    prev = &(sock->fasync_list);
1108
1109    for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1110        if (fa->fa_file == filp)
1111            break;
1112
1113    if (on) {
1114        if (fa != NULL) {
1115            write_lock_bh(&sk->sk_callback_lock);
1116            fa->fa_fd = fd;
1117            write_unlock_bh(&sk->sk_callback_lock);
1118
1119            kfree(fna);
1120            goto out;
1121        }
1122        fna->fa_file = filp;
1123        fna->fa_fd = fd;
1124        fna->magic = FASYNC_MAGIC;
1125        fna->fa_next = sock->fasync_list;
1126        write_lock_bh(&sk->sk_callback_lock);
1127        sock->fasync_list = fna;
1128        sock_set_flag(sk, SOCK_FASYNC);
1129        write_unlock_bh(&sk->sk_callback_lock);
1130    } else {
1131        if (fa != NULL) {
1132            write_lock_bh(&sk->sk_callback_lock);
1133            *prev = fa->fa_next;
1134            if (!sock->fasync_list)
1135                sock_reset_flag(sk, SOCK_FASYNC);
1136            write_unlock_bh(&sk->sk_callback_lock);
1137            kfree(fa);
1138        }
1139    }
1140
1141out:
1142    release_sock(sock->sk);
1143    return 0;
1144}
1145
1146/* This function may be called only under socket lock or callback_lock */
1147
1148int sock_wake_async(struct socket *sock, int how, int band)
1149{
1150    if (!sock || !sock->fasync_list)
1151        return -1;
1152    switch (how) {
1153    case SOCK_WAKE_WAITD:
1154        if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1155            break;
1156        goto call_kill;
1157    case SOCK_WAKE_SPACE:
1158        if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1159            break;
1160        /* fall through */
1161    case SOCK_WAKE_IO:
1162call_kill:
1163        __kill_fasync(sock->fasync_list, SIGIO, band);
1164        break;
1165    case SOCK_WAKE_URG:
1166        __kill_fasync(sock->fasync_list, SIGURG, band);
1167    }
1168    return 0;
1169}
1170
1171static int __sock_create(struct net *net, int family, int type, int protocol,
1172             struct socket **res, int kern)
1173{
1174    int err;
1175    struct socket *sock;
1176    const struct net_proto_family *pf;
1177
1178    /*
1179     * Check protocol is in range
1180     */
1181    if (family < 0 || family >= NPROTO)
1182        return -EAFNOSUPPORT;
1183    if (type < 0 || type >= SOCK_MAX)
1184        return -EINVAL;
1185
1186    /* Compatibility.
1187
1188       This uglymoron is moved from INET layer to here to avoid
1189       deadlock in module load.
1190     */
1191    if (family == PF_INET && type == SOCK_PACKET) {
1192        static int warned;
1193        if (!warned) {
1194            warned = 1;
1195            printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1196                   current->comm);
1197        }
1198        family = PF_PACKET;
1199    }
1200
1201    err = security_socket_create(family, type, protocol, kern);
1202    if (err)
1203        return err;
1204
1205    /*
1206     * Allocate the socket and allow the family to set things up. if
1207     * the protocol is 0, the family is instructed to select an appropriate
1208     * default.
1209     */
1210    sock = sock_alloc();
1211    if (!sock) {
1212        if (net_ratelimit())
1213            printk(KERN_WARNING "socket: no more sockets\n");
1214        return -ENFILE; /* Not exactly a match, but its the
1215                   closest posix thing */
1216    }
1217
1218    sock->type = type;
1219
1220#ifdef CONFIG_MODULES
1221    /* Attempt to load a protocol module if the find failed.
1222     *
1223     * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1224     * requested real, full-featured networking support upon configuration.
1225     * Otherwise module support will break!
1226     */
1227    if (net_families[family] == NULL)
1228        request_module("net-pf-%d", family);
1229#endif
1230
1231    rcu_read_lock();
1232    pf = rcu_dereference(net_families[family]);
1233    err = -EAFNOSUPPORT;
1234    if (!pf)
1235        goto out_release;
1236
1237    /*
1238     * We will call the ->create function, that possibly is in a loadable
1239     * module, so we have to bump that loadable module refcnt first.
1240     */
1241    if (!try_module_get(pf->owner))
1242        goto out_release;
1243
1244    /* Now protected by module ref count */
1245    rcu_read_unlock();
1246
1247    err = pf->create(net, sock, protocol, kern);
1248    if (err < 0)
1249        goto out_module_put;
1250
1251    /*
1252     * Now to bump the refcnt of the [loadable] module that owns this
1253     * socket at sock_release time we decrement its refcnt.
1254     */
1255    if (!try_module_get(sock->ops->owner))
1256        goto out_module_busy;
1257
1258    /*
1259     * Now that we're done with the ->create function, the [loadable]
1260     * module can have its refcnt decremented
1261     */
1262    module_put(pf->owner);
1263    err = security_socket_post_create(sock, family, type, protocol, kern);
1264    if (err)
1265        goto out_sock_release;
1266    *res = sock;
1267
1268    return 0;
1269
1270out_module_busy:
1271    err = -EAFNOSUPPORT;
1272out_module_put:
1273    sock->ops = NULL;
1274    module_put(pf->owner);
1275out_sock_release:
1276    sock_release(sock);
1277    return err;
1278
1279out_release:
1280    rcu_read_unlock();
1281    goto out_sock_release;
1282}
1283
1284int sock_create(int family, int type, int protocol, struct socket **res)
1285{
1286    return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1287}
1288
1289int sock_create_kern(int family, int type, int protocol, struct socket **res)
1290{
1291    return __sock_create(&init_net, family, type, protocol, res, 1);
1292}
1293
1294SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1295{
1296    int retval;
1297    struct socket *sock;
1298    int flags;
1299
1300    /* Check the SOCK_* constants for consistency. */
1301    BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1302    BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1303    BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1304    BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1305
1306    flags = type & ~SOCK_TYPE_MASK;
1307    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1308        return -EINVAL;
1309    type &= SOCK_TYPE_MASK;
1310
1311    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1312        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1313
1314    retval = sock_create(family, type, protocol, &sock);
1315    if (retval < 0)
1316        goto out;
1317
1318    retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1319    if (retval < 0)
1320        goto out_release;
1321
1322out:
1323    /* It may be already another descriptor 8) Not kernel problem. */
1324    return retval;
1325
1326out_release:
1327    sock_release(sock);
1328    return retval;
1329}
1330
1331/*
1332 * Create a pair of connected sockets.
1333 */
1334
1335SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1336        int __user *, usockvec)
1337{
1338    struct socket *sock1, *sock2;
1339    int fd1, fd2, err;
1340    struct file *newfile1, *newfile2;
1341    int flags;
1342
1343    flags = type & ~SOCK_TYPE_MASK;
1344    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1345        return -EINVAL;
1346    type &= SOCK_TYPE_MASK;
1347
1348    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1349        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1350
1351    /*
1352     * Obtain the first socket and check if the underlying protocol
1353     * supports the socketpair call.
1354     */
1355
1356    err = sock_create(family, type, protocol, &sock1);
1357    if (err < 0)
1358        goto out;
1359
1360    err = sock_create(family, type, protocol, &sock2);
1361    if (err < 0)
1362        goto out_release_1;
1363
1364    err = sock1->ops->socketpair(sock1, sock2);
1365    if (err < 0)
1366        goto out_release_both;
1367
1368    fd1 = sock_alloc_file(sock1, &newfile1, flags);
1369    if (unlikely(fd1 < 0)) {
1370        err = fd1;
1371        goto out_release_both;
1372    }
1373
1374    fd2 = sock_alloc_file(sock2, &newfile2, flags);
1375    if (unlikely(fd2 < 0)) {
1376        err = fd2;
1377        fput(newfile1);
1378        put_unused_fd(fd1);
1379        sock_release(sock2);
1380        goto out;
1381    }
1382
1383    audit_fd_pair(fd1, fd2);
1384    fd_install(fd1, newfile1);
1385    fd_install(fd2, newfile2);
1386    /* fd1 and fd2 may be already another descriptors.
1387     * Not kernel problem.
1388     */
1389
1390    err = put_user(fd1, &usockvec[0]);
1391    if (!err)
1392        err = put_user(fd2, &usockvec[1]);
1393    if (!err)
1394        return 0;
1395
1396    sys_close(fd2);
1397    sys_close(fd1);
1398    return err;
1399
1400out_release_both:
1401    sock_release(sock2);
1402out_release_1:
1403    sock_release(sock1);
1404out:
1405    return err;
1406}
1407
1408/*
1409 * Bind a name to a socket. Nothing much to do here since it's
1410 * the protocol's responsibility to handle the local address.
1411 *
1412 * We move the socket address to kernel space before we call
1413 * the protocol layer (having also checked the address is ok).
1414 */
1415
1416SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1417{
1418    struct socket *sock;
1419    struct sockaddr_storage address;
1420    int err, fput_needed;
1421
1422    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1423    if (sock) {
1424        err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1425        if (err >= 0) {
1426            err = security_socket_bind(sock,
1427                           (struct sockaddr *)&address,
1428                           addrlen);
1429            if (!err)
1430                err = sock->ops->bind(sock,
1431                              (struct sockaddr *)
1432                              &address, addrlen);
1433        }
1434        fput_light(sock->file, fput_needed);
1435    }
1436    return err;
1437}
1438
1439/*
1440 * Perform a listen. Basically, we allow the protocol to do anything
1441 * necessary for a listen, and if that works, we mark the socket as
1442 * ready for listening.
1443 */
1444
1445SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1446{
1447    struct socket *sock;
1448    int err, fput_needed;
1449    int somaxconn;
1450
1451    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1452    if (sock) {
1453        somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1454        if ((unsigned)backlog > somaxconn)
1455            backlog = somaxconn;
1456
1457        err = security_socket_listen(sock, backlog);
1458        if (!err)
1459            err = sock->ops->listen(sock, backlog);
1460
1461        fput_light(sock->file, fput_needed);
1462    }
1463    return err;
1464}
1465
1466/*
1467 * For accept, we attempt to create a new socket, set up the link
1468 * with the client, wake up the client, then return the new
1469 * connected fd. We collect the address of the connector in kernel
1470 * space and move it to user at the very end. This is unclean because
1471 * we open the socket then return an error.
1472 *
1473 * 1003.1g adds the ability to recvmsg() to query connection pending
1474 * status to recvmsg. We need to add that support in a way thats
1475 * clean when we restucture accept also.
1476 */
1477
1478SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1479        int __user *, upeer_addrlen, int, flags)
1480{
1481    struct socket *sock, *newsock;
1482    struct file *newfile;
1483    int err, len, newfd, fput_needed;
1484    struct sockaddr_storage address;
1485
1486    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1487        return -EINVAL;
1488
1489    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1490        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1491
1492    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1493    if (!sock)
1494        goto out;
1495
1496    err = -ENFILE;
1497    if (!(newsock = sock_alloc()))
1498        goto out_put;
1499
1500    newsock->type = sock->type;
1501    newsock->ops = sock->ops;
1502
1503    /*
1504     * We don't need try_module_get here, as the listening socket (sock)
1505     * has the protocol module (sock->ops->owner) held.
1506     */
1507    __module_get(newsock->ops->owner);
1508
1509    newfd = sock_alloc_file(newsock, &newfile, flags);
1510    if (unlikely(newfd < 0)) {
1511        err = newfd;
1512        sock_release(newsock);
1513        goto out_put;
1514    }
1515
1516    err = security_socket_accept(sock, newsock);
1517    if (err)
1518        goto out_fd;
1519
1520    err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1521    if (err < 0)
1522        goto out_fd;
1523
1524    if (upeer_sockaddr) {
1525        if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1526                      &len, 2) < 0) {
1527            err = -ECONNABORTED;
1528            goto out_fd;
1529        }
1530        err = move_addr_to_user((struct sockaddr *)&address,
1531                    len, upeer_sockaddr, upeer_addrlen);
1532        if (err < 0)
1533            goto out_fd;
1534    }
1535
1536    /* File flags are not inherited via accept() unlike another OSes. */
1537
1538    fd_install(newfd, newfile);
1539    err = newfd;
1540
1541out_put:
1542    fput_light(sock->file, fput_needed);
1543out:
1544    return err;
1545out_fd:
1546    fput(newfile);
1547    put_unused_fd(newfd);
1548    goto out_put;
1549}
1550
1551SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1552        int __user *, upeer_addrlen)
1553{
1554    return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1555}
1556
1557/*
1558 * Attempt to connect to a socket with the server address. The address
1559 * is in user space so we verify it is OK and move it to kernel space.
1560 *
1561 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1562 * break bindings
1563 *
1564 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1565 * other SEQPACKET protocols that take time to connect() as it doesn't
1566 * include the -EINPROGRESS status for such sockets.
1567 */
1568
1569SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1570        int, addrlen)
1571{
1572    struct socket *sock;
1573    struct sockaddr_storage address;
1574    int err, fput_needed;
1575
1576    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1577    if (!sock)
1578        goto out;
1579    err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1580    if (err < 0)
1581        goto out_put;
1582
1583    err =
1584        security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1585    if (err)
1586        goto out_put;
1587
1588    err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1589                 sock->file->f_flags);
1590out_put:
1591    fput_light(sock->file, fput_needed);
1592out:
1593    return err;
1594}
1595
1596/*
1597 * Get the local address ('name') of a socket object. Move the obtained
1598 * name to user space.
1599 */
1600
1601SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1602        int __user *, usockaddr_len)
1603{
1604    struct socket *sock;
1605    struct sockaddr_storage address;
1606    int len, err, fput_needed;
1607
1608    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1609    if (!sock)
1610        goto out;
1611
1612    err = security_socket_getsockname(sock);
1613    if (err)
1614        goto out_put;
1615
1616    err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1617    if (err)
1618        goto out_put;
1619    err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1620
1621out_put:
1622    fput_light(sock->file, fput_needed);
1623out:
1624    return err;
1625}
1626
1627/*
1628 * Get the remote address ('name') of a socket object. Move the obtained
1629 * name to user space.
1630 */
1631
1632SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1633        int __user *, usockaddr_len)
1634{
1635    struct socket *sock;
1636    struct sockaddr_storage address;
1637    int len, err, fput_needed;
1638
1639    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1640    if (sock != NULL) {
1641        err = security_socket_getpeername(sock);
1642        if (err) {
1643            fput_light(sock->file, fput_needed);
1644            return err;
1645        }
1646
1647        err =
1648            sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1649                       1);
1650        if (!err)
1651            err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1652                        usockaddr_len);
1653        fput_light(sock->file, fput_needed);
1654    }
1655    return err;
1656}
1657
1658/*
1659 * Send a datagram to a given address. We move the address into kernel
1660 * space and check the user space data area is readable before invoking
1661 * the protocol.
1662 */
1663
1664SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1665        unsigned, flags, struct sockaddr __user *, addr,
1666        int, addr_len)
1667{
1668    struct socket *sock;
1669    struct sockaddr_storage address;
1670    int err;
1671    struct msghdr msg;
1672    struct iovec iov;
1673    int fput_needed;
1674
1675    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1676    if (!sock)
1677        goto out;
1678
1679    iov.iov_base = buff;
1680    iov.iov_len = len;
1681    msg.msg_name = NULL;
1682    msg.msg_iov = &iov;
1683    msg.msg_iovlen = 1;
1684    msg.msg_control = NULL;
1685    msg.msg_controllen = 0;
1686    msg.msg_namelen = 0;
1687    if (addr) {
1688        err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1689        if (err < 0)
1690            goto out_put;
1691        msg.msg_name = (struct sockaddr *)&address;
1692        msg.msg_namelen = addr_len;
1693    }
1694    if (sock->file->f_flags & O_NONBLOCK)
1695        flags |= MSG_DONTWAIT;
1696    msg.msg_flags = flags;
1697    err = sock_sendmsg(sock, &msg, len);
1698
1699out_put:
1700    fput_light(sock->file, fput_needed);
1701out:
1702    return err;
1703}
1704
1705/*
1706 * Send a datagram down a socket.
1707 */
1708
1709SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1710        unsigned, flags)
1711{
1712    return sys_sendto(fd, buff, len, flags, NULL, 0);
1713}
1714
1715/*
1716 * Receive a frame from the socket and optionally record the address of the
1717 * sender. We verify the buffers are writable and if needed move the
1718 * sender address from kernel to user space.
1719 */
1720
1721SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1722        unsigned, flags, struct sockaddr __user *, addr,
1723        int __user *, addr_len)
1724{
1725    struct socket *sock;
1726    struct iovec iov;
1727    struct msghdr msg;
1728    struct sockaddr_storage address;
1729    int err, err2;
1730    int fput_needed;
1731
1732    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1733    if (!sock)
1734        goto out;
1735
1736    msg.msg_control = NULL;
1737    msg.msg_controllen = 0;
1738    msg.msg_iovlen = 1;
1739    msg.msg_iov = &iov;
1740    iov.iov_len = size;
1741    iov.iov_base = ubuf;
1742    msg.msg_name = (struct sockaddr *)&address;
1743    msg.msg_namelen = sizeof(address);
1744    if (sock->file->f_flags & O_NONBLOCK)
1745        flags |= MSG_DONTWAIT;
1746    err = sock_recvmsg(sock, &msg, size, flags);
1747
1748    if (err >= 0 && addr != NULL) {
1749        err2 = move_addr_to_user((struct sockaddr *)&address,
1750                     msg.msg_namelen, addr, addr_len);
1751        if (err2 < 0)
1752            err = err2;
1753    }
1754
1755    fput_light(sock->file, fput_needed);
1756out:
1757    return err;
1758}
1759
1760/*
1761 * Receive a datagram from a socket.
1762 */
1763
1764asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1765             unsigned flags)
1766{
1767    return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1768}
1769
1770/*
1771 * Set a socket option. Because we don't know the option lengths we have
1772 * to pass the user mode parameter for the protocols to sort out.
1773 */
1774
1775SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1776        char __user *, optval, int, optlen)
1777{
1778    int err, fput_needed;
1779    struct socket *sock;
1780
1781    if (optlen < 0)
1782        return -EINVAL;
1783
1784    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1785    if (sock != NULL) {
1786        err = security_socket_setsockopt(sock, level, optname);
1787        if (err)
1788            goto out_put;
1789
1790        if (level == SOL_SOCKET)
1791            err =
1792                sock_setsockopt(sock, level, optname, optval,
1793                        optlen);
1794        else
1795            err =
1796                sock->ops->setsockopt(sock, level, optname, optval,
1797                          optlen);
1798out_put:
1799        fput_light(sock->file, fput_needed);
1800    }
1801    return err;
1802}
1803
1804/*
1805 * Get a socket option. Because we don't know the option lengths we have
1806 * to pass a user mode parameter for the protocols to sort out.
1807 */
1808
1809SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1810        char __user *, optval, int __user *, optlen)
1811{
1812    int err, fput_needed;
1813    struct socket *sock;
1814
1815    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1816    if (sock != NULL) {
1817        err = security_socket_getsockopt(sock, level, optname);
1818        if (err)
1819            goto out_put;
1820
1821        if (level == SOL_SOCKET)
1822            err =
1823                sock_getsockopt(sock, level, optname, optval,
1824                        optlen);
1825        else
1826            err =
1827                sock->ops->getsockopt(sock, level, optname, optval,
1828                          optlen);
1829out_put:
1830        fput_light(sock->file, fput_needed);
1831    }
1832    return err;
1833}
1834
1835/*
1836 * Shutdown a socket.
1837 */
1838
1839SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1840{
1841    int err, fput_needed;
1842    struct socket *sock;
1843
1844    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1845    if (sock != NULL) {
1846        err = security_socket_shutdown(sock, how);
1847        if (!err)
1848            err = sock->ops->shutdown(sock, how);
1849        fput_light(sock->file, fput_needed);
1850    }
1851    return err;
1852}
1853
1854/* A couple of helpful macros for getting the address of the 32/64 bit
1855 * fields which are the same type (int / unsigned) on our platforms.
1856 */
1857#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1858#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1859#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1860
1861/*
1862 * BSD sendmsg interface
1863 */
1864
1865SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1866{
1867    struct compat_msghdr __user *msg_compat =
1868        (struct compat_msghdr __user *)msg;
1869    struct socket *sock;
1870    struct sockaddr_storage address;
1871    struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1872    unsigned char ctl[sizeof(struct cmsghdr) + 20]
1873        __attribute__ ((aligned(sizeof(__kernel_size_t))));
1874    /* 20 is size of ipv6_pktinfo */
1875    unsigned char *ctl_buf = ctl;
1876    struct msghdr msg_sys;
1877    int err, ctl_len, iov_size, total_len;
1878    int fput_needed;
1879
1880    err = -EFAULT;
1881    if (MSG_CMSG_COMPAT & flags) {
1882        if (get_compat_msghdr(&msg_sys, msg_compat))
1883            return -EFAULT;
1884    }
1885    else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1886        return -EFAULT;
1887
1888    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1889    if (!sock)
1890        goto out;
1891
1892    /* do not move before msg_sys is valid */
1893    err = -EMSGSIZE;
1894    if (msg_sys.msg_iovlen > UIO_MAXIOV)
1895        goto out_put;
1896
1897    /* Check whether to allocate the iovec area */
1898    err = -ENOMEM;
1899    iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1900    if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1901        iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1902        if (!iov)
1903            goto out_put;
1904    }
1905
1906    /* This will also move the address data into kernel space */
1907    if (MSG_CMSG_COMPAT & flags) {
1908        err = verify_compat_iovec(&msg_sys, iov,
1909                      (struct sockaddr *)&address,
1910                      VERIFY_READ);
1911    } else
1912        err = verify_iovec(&msg_sys, iov,
1913                   (struct sockaddr *)&address,
1914                   VERIFY_READ);
1915    if (err < 0)
1916        goto out_freeiov;
1917    total_len = err;
1918
1919    err = -ENOBUFS;
1920
1921    if (msg_sys.msg_controllen > INT_MAX)
1922        goto out_freeiov;
1923    ctl_len = msg_sys.msg_controllen;
1924    if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1925        err =
1926            cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1927                             sizeof(ctl));
1928        if (err)
1929            goto out_freeiov;
1930        ctl_buf = msg_sys.msg_control;
1931        ctl_len = msg_sys.msg_controllen;
1932    } else if (ctl_len) {
1933        if (ctl_len > sizeof(ctl)) {
1934            ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1935            if (ctl_buf == NULL)
1936                goto out_freeiov;
1937        }
1938        err = -EFAULT;
1939        /*
1940         * Careful! Before this, msg_sys.msg_control contains a user pointer.
1941         * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1942         * checking falls down on this.
1943         */
1944        if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1945                   ctl_len))
1946            goto out_freectl;
1947        msg_sys.msg_control = ctl_buf;
1948    }
1949    msg_sys.msg_flags = flags;
1950
1951    if (sock->file->f_flags & O_NONBLOCK)
1952        msg_sys.msg_flags |= MSG_DONTWAIT;
1953    err = sock_sendmsg(sock, &msg_sys, total_len);
1954
1955out_freectl:
1956    if (ctl_buf != ctl)
1957        sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1958out_freeiov:
1959    if (iov != iovstack)
1960        sock_kfree_s(sock->sk, iov, iov_size);
1961out_put:
1962    fput_light(sock->file, fput_needed);
1963out:
1964    return err;
1965}
1966
1967static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1968             struct msghdr *msg_sys, unsigned flags, int nosec)
1969{
1970    struct compat_msghdr __user *msg_compat =
1971        (struct compat_msghdr __user *)msg;
1972    struct iovec iovstack[UIO_FASTIOV];
1973    struct iovec *iov = iovstack;
1974    unsigned long cmsg_ptr;
1975    int err, iov_size, total_len, len;
1976
1977    /* kernel mode address */
1978    struct sockaddr_storage addr;
1979
1980    /* user mode address pointers */
1981    struct sockaddr __user *uaddr;
1982    int __user *uaddr_len;
1983
1984    if (MSG_CMSG_COMPAT & flags) {
1985        if (get_compat_msghdr(msg_sys, msg_compat))
1986            return -EFAULT;
1987    }
1988    else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1989        return -EFAULT;
1990
1991    err = -EMSGSIZE;
1992    if (msg_sys->msg_iovlen > UIO_MAXIOV)
1993        goto out;
1994
1995    /* Check whether to allocate the iovec area */
1996    err = -ENOMEM;
1997    iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1998    if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1999        iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2000        if (!iov)
2001            goto out;
2002    }
2003
2004    /*
2005     * Save the user-mode address (verify_iovec will change the
2006     * kernel msghdr to use the kernel address space)
2007     */
2008
2009    uaddr = (__force void __user *)msg_sys->msg_name;
2010    uaddr_len = COMPAT_NAMELEN(msg);
2011    if (MSG_CMSG_COMPAT & flags) {
2012        err = verify_compat_iovec(msg_sys, iov,
2013                      (struct sockaddr *)&addr,
2014                      VERIFY_WRITE);
2015    } else
2016        err = verify_iovec(msg_sys, iov,
2017                   (struct sockaddr *)&addr,
2018                   VERIFY_WRITE);
2019    if (err < 0)
2020        goto out_freeiov;
2021    total_len = err;
2022
2023    cmsg_ptr = (unsigned long)msg_sys->msg_control;
2024    msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2025
2026    if (sock->file->f_flags & O_NONBLOCK)
2027        flags |= MSG_DONTWAIT;
2028    err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2029                              total_len, flags);
2030    if (err < 0)
2031        goto out_freeiov;
2032    len = err;
2033
2034    if (uaddr != NULL) {
2035        err = move_addr_to_user((struct sockaddr *)&addr,
2036                    msg_sys->msg_namelen, uaddr,
2037                    uaddr_len);
2038        if (err < 0)
2039            goto out_freeiov;
2040    }
2041    err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2042             COMPAT_FLAGS(msg));
2043    if (err)
2044        goto out_freeiov;
2045    if (MSG_CMSG_COMPAT & flags)
2046        err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2047                 &msg_compat->msg_controllen);
2048    else
2049        err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2050                 &msg->msg_controllen);
2051    if (err)
2052        goto out_freeiov;
2053    err = len;
2054
2055out_freeiov:
2056    if (iov != iovstack)
2057        sock_kfree_s(sock->sk, iov, iov_size);
2058out:
2059    return err;
2060}
2061
2062/*
2063 * BSD recvmsg interface
2064 */
2065
2066SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2067        unsigned int, flags)
2068{
2069    int fput_needed, err;
2070    struct msghdr msg_sys;
2071    struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2072
2073    if (!sock)
2074        goto out;
2075
2076    err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2077
2078    fput_light(sock->file, fput_needed);
2079out:
2080    return err;
2081}
2082
2083/*
2084 * Linux recvmmsg interface
2085 */
2086
2087int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2088           unsigned int flags, struct timespec *timeout)
2089{
2090    int fput_needed, err, datagrams;
2091    struct socket *sock;
2092    struct mmsghdr __user *entry;
2093    struct compat_mmsghdr __user *compat_entry;
2094    struct msghdr msg_sys;
2095    struct timespec end_time;
2096
2097    if (timeout &&
2098        poll_select_set_timeout(&end_time, timeout->tv_sec,
2099                    timeout->tv_nsec))
2100        return -EINVAL;
2101
2102    datagrams = 0;
2103
2104    sock = sockfd_lookup_light(fd, &err, &fput_needed);
2105    if (!sock)
2106        return err;
2107
2108    err = sock_error(sock->sk);
2109    if (err)
2110        goto out_put;
2111
2112    entry = mmsg;
2113    compat_entry = (struct compat_mmsghdr __user *)mmsg;
2114
2115    while (datagrams < vlen) {
2116        /*
2117         * No need to ask LSM for more than the first datagram.
2118         */
2119        if (MSG_CMSG_COMPAT & flags) {
2120            err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2121                        &msg_sys, flags, datagrams);
2122            if (err < 0)
2123                break;
2124            err = __put_user(err, &compat_entry->msg_len);
2125            ++compat_entry;
2126        } else {
2127            err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2128                        &msg_sys, flags, datagrams);
2129            if (err < 0)
2130                break;
2131            err = put_user(err, &entry->msg_len);
2132            ++entry;
2133        }
2134
2135        if (err)
2136            break;
2137        ++datagrams;
2138
2139        /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2140        if (flags & MSG_WAITFORONE)
2141            flags |= MSG_DONTWAIT;
2142
2143        if (timeout) {
2144            ktime_get_ts(timeout);
2145            *timeout = timespec_sub(end_time, *timeout);
2146            if (timeout->tv_sec < 0) {
2147                timeout->tv_sec = timeout->tv_nsec = 0;
2148                break;
2149            }
2150
2151            /* Timeout, return less than vlen datagrams */
2152            if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2153                break;
2154        }
2155
2156        /* Out of band data, return right away */
2157        if (msg_sys.msg_flags & MSG_OOB)
2158            break;
2159    }
2160
2161out_put:
2162    fput_light(sock->file, fput_needed);
2163
2164    if (err == 0)
2165        return datagrams;
2166
2167    if (datagrams != 0) {
2168        /*
2169         * We may return less entries than requested (vlen) if the
2170         * sock is non block and there aren't enough datagrams...
2171         */
2172        if (err != -EAGAIN) {
2173            /*
2174             * ... or if recvmsg returns an error after we
2175             * received some datagrams, where we record the
2176             * error to return on the next call or if the
2177             * app asks about it using getsockopt(SO_ERROR).
2178             */
2179            sock->sk->sk_err = -err;
2180        }
2181
2182        return datagrams;
2183    }
2184
2185    return err;
2186}
2187
2188SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2189        unsigned int, vlen, unsigned int, flags,
2190        struct timespec __user *, timeout)
2191{
2192    int datagrams;
2193    struct timespec timeout_sys;
2194
2195    if (!timeout)
2196        return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2197
2198    if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2199        return -EFAULT;
2200
2201    datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2202
2203    if (datagrams > 0 &&
2204        copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2205        datagrams = -EFAULT;
2206
2207    return datagrams;
2208}
2209
2210#ifdef __ARCH_WANT_SYS_SOCKETCALL
2211/* Argument list sizes for sys_socketcall */
2212#define AL(x) ((x) * sizeof(unsigned long))
2213static const unsigned char nargs[20] = {
2214    AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2215    AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2216    AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2217    AL(4),AL(5)
2218};
2219
2220#undef AL
2221
2222/*
2223 * System call vectors.
2224 *
2225 * Argument checking cleaned up. Saved 20% in size.
2226 * This function doesn't need to set the kernel lock because
2227 * it is set by the callees.
2228 */
2229
2230SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2231{
2232    unsigned long a[6];
2233    unsigned long a0, a1;
2234    int err;
2235    unsigned int len;
2236
2237    if (call < 1 || call > SYS_RECVMMSG)
2238        return -EINVAL;
2239
2240    len = nargs[call];
2241    if (len > sizeof(a))
2242        return -EINVAL;
2243
2244    /* copy_from_user should be SMP safe. */
2245    if (copy_from_user(a, args, len))
2246        return -EFAULT;
2247
2248    audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2249
2250    a0 = a[0];
2251    a1 = a[1];
2252
2253    switch (call) {
2254    case SYS_SOCKET:
2255        err = sys_socket(a0, a1, a[2]);
2256        break;
2257    case SYS_BIND:
2258        err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2259        break;
2260    case SYS_CONNECT:
2261        err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2262        break;
2263    case SYS_LISTEN:
2264        err = sys_listen(a0, a1);
2265        break;
2266    case SYS_ACCEPT:
2267        err = sys_accept4(a0, (struct sockaddr __user *)a1,
2268                  (int __user *)a[2], 0);
2269        break;
2270    case SYS_GETSOCKNAME:
2271        err =
2272            sys_getsockname(a0, (struct sockaddr __user *)a1,
2273                    (int __user *)a[2]);
2274        break;
2275    case SYS_GETPEERNAME:
2276        err =
2277            sys_getpeername(a0, (struct sockaddr __user *)a1,
2278                    (int __user *)a[2]);
2279        break;
2280    case SYS_SOCKETPAIR:
2281        err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2282        break;
2283    case SYS_SEND:
2284        err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2285        break;
2286    case SYS_SENDTO:
2287        err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2288                 (struct sockaddr __user *)a[4], a[5]);
2289        break;
2290    case SYS_RECV:
2291        err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2292        break;
2293    case SYS_RECVFROM:
2294        err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2295                   (struct sockaddr __user *)a[4],
2296                   (int __user *)a[5]);
2297        break;
2298    case SYS_SHUTDOWN:
2299        err = sys_shutdown(a0, a1);
2300        break;
2301    case SYS_SETSOCKOPT:
2302        err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2303        break;
2304    case SYS_GETSOCKOPT:
2305        err =
2306            sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2307                   (int __user *)a[4]);
2308        break;
2309    case SYS_SENDMSG:
2310        err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2311        break;
2312    case SYS_RECVMSG:
2313        err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2314        break;
2315    case SYS_RECVMMSG:
2316        err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2317                   (struct timespec __user *)a[4]);
2318        break;
2319    case SYS_ACCEPT4:
2320        err = sys_accept4(a0, (struct sockaddr __user *)a1,
2321                  (int __user *)a[2], a[3]);
2322        break;
2323    default:
2324        err = -EINVAL;
2325        break;
2326    }
2327    return err;
2328}
2329
2330#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2331
2332/**
2333 * sock_register - add a socket protocol handler
2334 * @ops: description of protocol
2335 *
2336 * This function is called by a protocol handler that wants to
2337 * advertise its address family, and have it linked into the
2338 * socket interface. The value ops->family coresponds to the
2339 * socket system call protocol family.
2340 */
2341int sock_register(const struct net_proto_family *ops)
2342{
2343    int err;
2344
2345    if (ops->family >= NPROTO) {
2346        printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2347               NPROTO);
2348        return -ENOBUFS;
2349    }
2350
2351    spin_lock(&net_family_lock);
2352    if (net_families[ops->family])
2353        err = -EEXIST;
2354    else {
2355        net_families[ops->family] = ops;
2356        err = 0;
2357    }
2358    spin_unlock(&net_family_lock);
2359
2360    printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2361    return err;
2362}
2363
2364/**
2365 * sock_unregister - remove a protocol handler
2366 * @family: protocol family to remove
2367 *
2368 * This function is called by a protocol handler that wants to
2369 * remove its address family, and have it unlinked from the
2370 * new socket creation.
2371 *
2372 * If protocol handler is a module, then it can use module reference
2373 * counts to protect against new references. If protocol handler is not
2374 * a module then it needs to provide its own protection in
2375 * the ops->create routine.
2376 */
2377void sock_unregister(int family)
2378{
2379    BUG_ON(family < 0 || family >= NPROTO);
2380
2381    spin_lock(&net_family_lock);
2382    net_families[family] = NULL;
2383    spin_unlock(&net_family_lock);
2384
2385    synchronize_rcu();
2386
2387    printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2388}
2389
2390static int __init sock_init(void)
2391{
2392    /*
2393     * Initialize sock SLAB cache.
2394     */
2395
2396    sk_init();
2397
2398    /*
2399     * Initialize skbuff SLAB cache
2400     */
2401    skb_init();
2402
2403    /*
2404     * Initialize the protocols module.
2405     */
2406
2407    init_inodecache();
2408    register_filesystem(&sock_fs_type);
2409    sock_mnt = kern_mount(&sock_fs_type);
2410
2411    /* The real protocol initialization is performed in later initcalls.
2412     */
2413
2414#ifdef CONFIG_NETFILTER
2415    netfilter_init();
2416#endif
2417
2418    return 0;
2419}
2420
2421core_initcall(sock_init); /* early initcall */
2422
2423#ifdef CONFIG_PROC_FS
2424void socket_seq_show(struct seq_file *seq)
2425{
2426    int cpu;
2427    int counter = 0;
2428
2429    for_each_possible_cpu(cpu)
2430        counter += per_cpu(sockets_in_use, cpu);
2431
2432    /* It can be negative, by the way. 8) */
2433    if (counter < 0)
2434        counter = 0;
2435
2436    seq_printf(seq, "sockets: used %d\n", counter);
2437}
2438#endif /* CONFIG_PROC_FS */
2439
2440#ifdef CONFIG_COMPAT
2441static int do_siocgstamp(struct net *net, struct socket *sock,
2442             unsigned int cmd, struct compat_timeval __user *up)
2443{
2444    mm_segment_t old_fs = get_fs();
2445    struct timeval ktv;
2446    int err;
2447
2448    set_fs(KERNEL_DS);
2449    err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2450    set_fs(old_fs);
2451    if (!err) {
2452        err = put_user(ktv.tv_sec, &up->tv_sec);
2453        err |= __put_user(ktv.tv_usec, &up->tv_usec);
2454    }
2455    return err;
2456}
2457
2458static int do_siocgstampns(struct net *net, struct socket *sock,
2459             unsigned int cmd, struct compat_timespec __user *up)
2460{
2461    mm_segment_t old_fs = get_fs();
2462    struct timespec kts;
2463    int err;
2464
2465    set_fs(KERNEL_DS);
2466    err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2467    set_fs(old_fs);
2468    if (!err) {
2469        err = put_user(kts.tv_sec, &up->tv_sec);
2470        err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2471    }
2472    return err;
2473}
2474
2475static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2476{
2477    struct ifreq __user *uifr;
2478    int err;
2479
2480    uifr = compat_alloc_user_space(sizeof(struct ifreq));
2481    if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2482        return -EFAULT;
2483
2484    err = dev_ioctl(net, SIOCGIFNAME, uifr);
2485    if (err)
2486        return err;
2487
2488    if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2489        return -EFAULT;
2490
2491    return 0;
2492}
2493
2494static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2495{
2496    struct compat_ifconf ifc32;
2497    struct ifconf ifc;
2498    struct ifconf __user *uifc;
2499    struct compat_ifreq __user *ifr32;
2500    struct ifreq __user *ifr;
2501    unsigned int i, j;
2502    int err;
2503
2504    if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2505        return -EFAULT;
2506
2507    if (ifc32.ifcbuf == 0) {
2508        ifc32.ifc_len = 0;
2509        ifc.ifc_len = 0;
2510        ifc.ifc_req = NULL;
2511        uifc = compat_alloc_user_space(sizeof(struct ifconf));
2512    } else {
2513        size_t len =((ifc32.ifc_len / sizeof (struct compat_ifreq)) + 1) *
2514            sizeof (struct ifreq);
2515        uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2516        ifc.ifc_len = len;
2517        ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2518        ifr32 = compat_ptr(ifc32.ifcbuf);
2519        for (i = 0; i < ifc32.ifc_len; i += sizeof (struct compat_ifreq)) {
2520            if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2521                return -EFAULT;
2522            ifr++;
2523            ifr32++;
2524        }
2525    }
2526    if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2527        return -EFAULT;
2528
2529    err = dev_ioctl(net, SIOCGIFCONF, uifc);
2530    if (err)
2531        return err;
2532
2533    if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2534        return -EFAULT;
2535
2536    ifr = ifc.ifc_req;
2537    ifr32 = compat_ptr(ifc32.ifcbuf);
2538    for (i = 0, j = 0;
2539             i + sizeof (struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2540         i += sizeof (struct compat_ifreq), j += sizeof (struct ifreq)) {
2541        if (copy_in_user(ifr32, ifr, sizeof (struct compat_ifreq)))
2542            return -EFAULT;
2543        ifr32++;
2544        ifr++;
2545    }
2546
2547    if (ifc32.ifcbuf == 0) {
2548        /* Translate from 64-bit structure multiple to
2549         * a 32-bit one.
2550         */
2551        i = ifc.ifc_len;
2552        i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2553        ifc32.ifc_len = i;
2554    } else {
2555        ifc32.ifc_len = i;
2556    }
2557    if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2558        return -EFAULT;
2559
2560    return 0;
2561}
2562
2563static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2564{
2565    struct ifreq __user *ifr;
2566    u32 data;
2567    void __user *datap;
2568
2569    ifr = compat_alloc_user_space(sizeof(*ifr));
2570
2571    if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2572        return -EFAULT;
2573
2574    if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2575        return -EFAULT;
2576
2577    datap = compat_ptr(data);
2578    if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2579        return -EFAULT;
2580
2581    return dev_ioctl(net, SIOCETHTOOL, ifr);
2582}
2583
2584static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2585{
2586    void __user *uptr;
2587    compat_uptr_t uptr32;
2588    struct ifreq __user *uifr;
2589
2590    uifr = compat_alloc_user_space(sizeof (*uifr));
2591    if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2592        return -EFAULT;
2593
2594    if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2595        return -EFAULT;
2596
2597    uptr = compat_ptr(uptr32);
2598
2599    if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2600        return -EFAULT;
2601
2602    return dev_ioctl(net, SIOCWANDEV, uifr);
2603}
2604
2605static int bond_ioctl(struct net *net, unsigned int cmd,
2606             struct compat_ifreq __user *ifr32)
2607{
2608    struct ifreq kifr;
2609    struct ifreq __user *uifr;
2610    mm_segment_t old_fs;
2611    int err;
2612    u32 data;
2613    void __user *datap;
2614
2615    switch (cmd) {
2616    case SIOCBONDENSLAVE:
2617    case SIOCBONDRELEASE:
2618    case SIOCBONDSETHWADDR:
2619    case SIOCBONDCHANGEACTIVE:
2620        if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2621            return -EFAULT;
2622
2623        old_fs = get_fs();
2624        set_fs (KERNEL_DS);
2625        err = dev_ioctl(net, cmd, &kifr);
2626        set_fs (old_fs);
2627
2628        return err;
2629    case SIOCBONDSLAVEINFOQUERY:
2630    case SIOCBONDINFOQUERY:
2631        uifr = compat_alloc_user_space(sizeof(*uifr));
2632        if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2633            return -EFAULT;
2634
2635        if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2636            return -EFAULT;
2637
2638        datap = compat_ptr(data);
2639        if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2640            return -EFAULT;
2641
2642        return dev_ioctl(net, cmd, uifr);
2643    default:
2644        return -EINVAL;
2645    };
2646}
2647
2648static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2649                 struct compat_ifreq __user *u_ifreq32)
2650{
2651    struct ifreq __user *u_ifreq64;
2652    char tmp_buf[IFNAMSIZ];
2653    void __user *data64;
2654    u32 data32;
2655
2656    if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2657               IFNAMSIZ))
2658        return -EFAULT;
2659    if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2660        return -EFAULT;
2661    data64 = compat_ptr(data32);
2662
2663    u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2664
2665    /* Don't check these user accesses, just let that get trapped
2666     * in the ioctl handler instead.
2667     */
2668    if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2669             IFNAMSIZ))
2670        return -EFAULT;
2671    if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2672        return -EFAULT;
2673
2674    return dev_ioctl(net, cmd, u_ifreq64);
2675}
2676
2677static int dev_ifsioc(struct net *net, struct socket *sock,
2678             unsigned int cmd, struct compat_ifreq __user *uifr32)
2679{
2680    struct ifreq __user *uifr;
2681    int err;
2682
2683    uifr = compat_alloc_user_space(sizeof(*uifr));
2684    if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2685        return -EFAULT;
2686
2687    err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2688
2689    if (!err) {
2690        switch (cmd) {
2691        case SIOCGIFFLAGS:
2692        case SIOCGIFMETRIC:
2693        case SIOCGIFMTU:
2694        case SIOCGIFMEM:
2695        case SIOCGIFHWADDR:
2696        case SIOCGIFINDEX:
2697        case SIOCGIFADDR:
2698        case SIOCGIFBRDADDR:
2699        case SIOCGIFDSTADDR:
2700        case SIOCGIFNETMASK:
2701        case SIOCGIFPFLAGS:
2702        case SIOCGIFTXQLEN:
2703        case SIOCGMIIPHY:
2704        case SIOCGMIIREG:
2705            if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2706                err = -EFAULT;
2707            break;
2708        }
2709    }
2710    return err;
2711}
2712
2713static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2714            struct compat_ifreq __user *uifr32)
2715{
2716    struct ifreq ifr;
2717    struct compat_ifmap __user *uifmap32;
2718    mm_segment_t old_fs;
2719    int err;
2720
2721    uifmap32 = &uifr32->ifr_ifru.ifru_map;
2722    err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2723    err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2724    err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2725    err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2726    err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2727    err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2728    err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2729    if (err)
2730        return -EFAULT;
2731
2732    old_fs = get_fs();
2733    set_fs (KERNEL_DS);
2734    err = dev_ioctl(net, cmd, (void __user *)&ifr);
2735    set_fs (old_fs);
2736
2737    if (cmd == SIOCGIFMAP && !err) {
2738        err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2739        err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2740        err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2741        err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2742        err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2743        err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2744        err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2745        if (err)
2746            err = -EFAULT;
2747    }
2748    return err;
2749}
2750
2751static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2752{
2753    void __user *uptr;
2754    compat_uptr_t uptr32;
2755    struct ifreq __user *uifr;
2756
2757    uifr = compat_alloc_user_space(sizeof (*uifr));
2758    if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2759        return -EFAULT;
2760
2761    if (get_user(uptr32, &uifr32->ifr_data))
2762        return -EFAULT;
2763
2764    uptr = compat_ptr(uptr32);
2765
2766    if (put_user(uptr, &uifr->ifr_data))
2767        return -EFAULT;
2768
2769    return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2770}
2771
2772struct rtentry32 {
2773    u32 rt_pad1;
2774    struct sockaddr rt_dst; /* target address */
2775    struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2776    struct sockaddr rt_genmask; /* target network mask (IP) */
2777    unsigned short rt_flags;
2778    short rt_pad2;
2779    u32 rt_pad3;
2780    unsigned char rt_tos;
2781    unsigned char rt_class;
2782    short rt_pad4;
2783    short rt_metric; /* +1 for binary compatibility! */
2784    /* char * */ u32 rt_dev; /* forcing the device at add */
2785    u32 rt_mtu; /* per route MTU/Window */
2786    u32 rt_window; /* Window clamping */
2787    unsigned short rt_irtt; /* Initial RTT */
2788};
2789
2790struct in6_rtmsg32 {
2791    struct in6_addr rtmsg_dst;
2792    struct in6_addr rtmsg_src;
2793    struct in6_addr rtmsg_gateway;
2794    u32 rtmsg_type;
2795    u16 rtmsg_dst_len;
2796    u16 rtmsg_src_len;
2797    u32 rtmsg_metric;
2798    u32 rtmsg_info;
2799    u32 rtmsg_flags;
2800    s32 rtmsg_ifindex;
2801};
2802
2803static int routing_ioctl(struct net *net, struct socket *sock,
2804             unsigned int cmd, void __user *argp)
2805{
2806    int ret;
2807    void *r = NULL;
2808    struct in6_rtmsg r6;
2809    struct rtentry r4;
2810    char devname[16];
2811    u32 rtdev;
2812    mm_segment_t old_fs = get_fs();
2813
2814    if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2815        struct in6_rtmsg32 __user *ur6 = argp;
2816        ret = copy_from_user (&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2817            3 * sizeof(struct in6_addr));
2818        ret |= __get_user (r6.rtmsg_type, &(ur6->rtmsg_type));
2819        ret |= __get_user (r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2820        ret |= __get_user (r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2821        ret |= __get_user (r6.rtmsg_metric, &(ur6->rtmsg_metric));
2822        ret |= __get_user (r6.rtmsg_info, &(ur6->rtmsg_info));
2823        ret |= __get_user (r6.rtmsg_flags, &(ur6->rtmsg_flags));
2824        ret |= __get_user (r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2825
2826        r = (void *) &r6;
2827    } else { /* ipv4 */
2828        struct rtentry32 __user *ur4 = argp;
2829        ret = copy_from_user (&r4.rt_dst, &(ur4->rt_dst),
2830                    3 * sizeof(struct sockaddr));
2831        ret |= __get_user (r4.rt_flags, &(ur4->rt_flags));
2832        ret |= __get_user (r4.rt_metric, &(ur4->rt_metric));
2833        ret |= __get_user (r4.rt_mtu, &(ur4->rt_mtu));
2834        ret |= __get_user (r4.rt_window, &(ur4->rt_window));
2835        ret |= __get_user (r4.rt_irtt, &(ur4->rt_irtt));
2836        ret |= __get_user (rtdev, &(ur4->rt_dev));
2837        if (rtdev) {
2838            ret |= copy_from_user (devname, compat_ptr(rtdev), 15);
2839            r4.rt_dev = devname; devname[15] = 0;
2840        } else
2841            r4.rt_dev = NULL;
2842
2843        r = (void *) &r4;
2844    }
2845
2846    if (ret) {
2847        ret = -EFAULT;
2848        goto out;
2849    }
2850
2851    set_fs (KERNEL_DS);
2852    ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
2853    set_fs (old_fs);
2854
2855out:
2856    return ret;
2857}
2858
2859/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2860 * for some operations; this forces use of the newer bridge-utils that
2861 * use compatiable ioctls
2862 */
2863static int old_bridge_ioctl(compat_ulong_t __user *argp)
2864{
2865    compat_ulong_t tmp;
2866
2867    if (get_user(tmp, argp))
2868        return -EFAULT;
2869    if (tmp == BRCTL_GET_VERSION)
2870        return BRCTL_VERSION + 1;
2871    return -EINVAL;
2872}
2873
2874static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2875             unsigned int cmd, unsigned long arg)
2876{
2877    void __user *argp = compat_ptr(arg);
2878    struct sock *sk = sock->sk;
2879    struct net *net = sock_net(sk);
2880
2881    if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2882        return siocdevprivate_ioctl(net, cmd, argp);
2883
2884    switch (cmd) {
2885    case SIOCSIFBR:
2886    case SIOCGIFBR:
2887        return old_bridge_ioctl(argp);
2888    case SIOCGIFNAME:
2889        return dev_ifname32(net, argp);
2890    case SIOCGIFCONF:
2891        return dev_ifconf(net, argp);
2892    case SIOCETHTOOL:
2893        return ethtool_ioctl(net, argp);
2894    case SIOCWANDEV:
2895        return compat_siocwandev(net, argp);
2896    case SIOCGIFMAP:
2897    case SIOCSIFMAP:
2898        return compat_sioc_ifmap(net, cmd, argp);
2899    case SIOCBONDENSLAVE:
2900    case SIOCBONDRELEASE:
2901    case SIOCBONDSETHWADDR:
2902    case SIOCBONDSLAVEINFOQUERY:
2903    case SIOCBONDINFOQUERY:
2904    case SIOCBONDCHANGEACTIVE:
2905        return bond_ioctl(net, cmd, argp);
2906    case SIOCADDRT:
2907    case SIOCDELRT:
2908        return routing_ioctl(net, sock, cmd, argp);
2909    case SIOCGSTAMP:
2910        return do_siocgstamp(net, sock, cmd, argp);
2911    case SIOCGSTAMPNS:
2912        return do_siocgstampns(net, sock, cmd, argp);
2913    case SIOCSHWTSTAMP:
2914        return compat_siocshwtstamp(net, argp);
2915
2916    case FIOSETOWN:
2917    case SIOCSPGRP:
2918    case FIOGETOWN:
2919    case SIOCGPGRP:
2920    case SIOCBRADDBR:
2921    case SIOCBRDELBR:
2922    case SIOCGIFVLAN:
2923    case SIOCSIFVLAN:
2924    case SIOCADDDLCI:
2925    case SIOCDELDLCI:
2926        return sock_ioctl(file, cmd, arg);
2927
2928    case SIOCGIFFLAGS:
2929    case SIOCSIFFLAGS:
2930    case SIOCGIFMETRIC:
2931    case SIOCSIFMETRIC:
2932    case SIOCGIFMTU:
2933    case SIOCSIFMTU:
2934    case SIOCGIFMEM:
2935    case SIOCSIFMEM:
2936    case SIOCGIFHWADDR:
2937    case SIOCSIFHWADDR:
2938    case SIOCADDMULTI:
2939    case SIOCDELMULTI:
2940    case SIOCGIFINDEX:
2941    case SIOCGIFADDR:
2942    case SIOCSIFADDR:
2943    case SIOCSIFHWBROADCAST:
2944    case SIOCDIFADDR:
2945    case SIOCGIFBRDADDR:
2946    case SIOCSIFBRDADDR:
2947    case SIOCGIFDSTADDR:
2948    case SIOCSIFDSTADDR:
2949    case SIOCGIFNETMASK:
2950    case SIOCSIFNETMASK:
2951    case SIOCSIFPFLAGS:
2952    case SIOCGIFPFLAGS:
2953    case SIOCGIFTXQLEN:
2954    case SIOCSIFTXQLEN:
2955    case SIOCBRADDIF:
2956    case SIOCBRDELIF:
2957    case SIOCSIFNAME:
2958    case SIOCGMIIPHY:
2959    case SIOCGMIIREG:
2960    case SIOCSMIIREG:
2961        return dev_ifsioc(net, sock, cmd, argp);
2962
2963    case SIOCSARP:
2964    case SIOCGARP:
2965    case SIOCDARP:
2966    case SIOCATMARK:
2967        return sock_do_ioctl(net, sock, cmd, arg);
2968    }
2969
2970    /* Prevent warning from compat_sys_ioctl, these always
2971     * result in -EINVAL in the native case anyway. */
2972    switch (cmd) {
2973    case SIOCRTMSG:
2974    case SIOCGIFCOUNT:
2975    case SIOCSRARP:
2976    case SIOCGRARP:
2977    case SIOCDRARP:
2978    case SIOCSIFLINK:
2979    case SIOCGIFSLAVE:
2980    case SIOCSIFSLAVE:
2981        return -EINVAL;
2982    }
2983
2984    return -ENOIOCTLCMD;
2985}
2986
2987static long compat_sock_ioctl(struct file *file, unsigned cmd,
2988                  unsigned long arg)
2989{
2990    struct socket *sock = file->private_data;
2991    int ret = -ENOIOCTLCMD;
2992    struct sock *sk;
2993    struct net *net;
2994
2995    sk = sock->sk;
2996    net = sock_net(sk);
2997
2998    if (sock->ops->compat_ioctl)
2999        ret = sock->ops->compat_ioctl(sock, cmd, arg);
3000
3001    if (ret == -ENOIOCTLCMD &&
3002        (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3003        ret = compat_wext_handle_ioctl(net, cmd, arg);
3004
3005    if (ret == -ENOIOCTLCMD)
3006        ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3007
3008    return ret;
3009}
3010#endif
3011
3012int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3013{
3014    return sock->ops->bind(sock, addr, addrlen);
3015}
3016
3017int kernel_listen(struct socket *sock, int backlog)
3018{
3019    return sock->ops->listen(sock, backlog);
3020}
3021
3022int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3023{
3024    struct sock *sk = sock->sk;
3025    int err;
3026
3027    err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3028                   newsock);
3029    if (err < 0)
3030        goto done;
3031
3032    err = sock->ops->accept(sock, *newsock, flags);
3033    if (err < 0) {
3034        sock_release(*newsock);
3035        *newsock = NULL;
3036        goto done;
3037    }
3038
3039    (*newsock)->ops = sock->ops;
3040    __module_get((*newsock)->ops->owner);
3041
3042done:
3043    return err;
3044}
3045
3046int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3047           int flags)
3048{
3049    return sock->ops->connect(sock, addr, addrlen, flags);
3050}
3051
3052int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3053             int *addrlen)
3054{
3055    return sock->ops->getname(sock, addr, addrlen, 0);
3056}
3057
3058int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3059             int *addrlen)
3060{
3061    return sock->ops->getname(sock, addr, addrlen, 1);
3062}
3063
3064int kernel_getsockopt(struct socket *sock, int level, int optname,
3065            char *optval, int *optlen)
3066{
3067    mm_segment_t oldfs = get_fs();
3068    int err;
3069
3070    set_fs(KERNEL_DS);
3071    if (level == SOL_SOCKET)
3072        err = sock_getsockopt(sock, level, optname, optval, optlen);
3073    else
3074        err = sock->ops->getsockopt(sock, level, optname, optval,
3075                        optlen);
3076    set_fs(oldfs);
3077    return err;
3078}
3079
3080int kernel_setsockopt(struct socket *sock, int level, int optname,
3081            char *optval, unsigned int optlen)
3082{
3083    mm_segment_t oldfs = get_fs();
3084    int err;
3085
3086    set_fs(KERNEL_DS);
3087    if (level == SOL_SOCKET)
3088        err = sock_setsockopt(sock, level, optname, optval, optlen);
3089    else
3090        err = sock->ops->setsockopt(sock, level, optname, optval,
3091                        optlen);
3092    set_fs(oldfs);
3093    return err;
3094}
3095
3096int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3097            size_t size, int flags)
3098{
3099    if (sock->ops->sendpage)
3100        return sock->ops->sendpage(sock, page, offset, size, flags);
3101
3102    return sock_no_sendpage(sock, page, offset, size, flags);
3103}
3104
3105int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3106{
3107    mm_segment_t oldfs = get_fs();
3108    int err;
3109
3110    set_fs(KERNEL_DS);
3111    err = sock->ops->ioctl(sock, cmd, arg);
3112    set_fs(oldfs);
3113
3114    return err;
3115}
3116
3117int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3118{
3119    return sock->ops->shutdown(sock, how);
3120}
3121
3122EXPORT_SYMBOL(sock_create);
3123EXPORT_SYMBOL(sock_create_kern);
3124EXPORT_SYMBOL(sock_create_lite);
3125EXPORT_SYMBOL(sock_map_fd);
3126EXPORT_SYMBOL(sock_recvmsg);
3127EXPORT_SYMBOL(sock_register);
3128EXPORT_SYMBOL(sock_release);
3129EXPORT_SYMBOL(sock_sendmsg);
3130EXPORT_SYMBOL(sock_unregister);
3131EXPORT_SYMBOL(sock_wake_async);
3132EXPORT_SYMBOL(sockfd_lookup);
3133EXPORT_SYMBOL(kernel_sendmsg);
3134EXPORT_SYMBOL(kernel_recvmsg);
3135EXPORT_SYMBOL(kernel_bind);
3136EXPORT_SYMBOL(kernel_listen);
3137EXPORT_SYMBOL(kernel_accept);
3138EXPORT_SYMBOL(kernel_connect);
3139EXPORT_SYMBOL(kernel_getsockname);
3140EXPORT_SYMBOL(kernel_getpeername);
3141EXPORT_SYMBOL(kernel_getsockopt);
3142EXPORT_SYMBOL(kernel_setsockopt);
3143EXPORT_SYMBOL(kernel_sendpage);
3144EXPORT_SYMBOL(kernel_sock_ioctl);
3145EXPORT_SYMBOL(kernel_sock_shutdown);
3146

Archive Download this file



interactive