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#include <net/cls_cgroup.h>
98
99#include <net/sock.h>
100#include <linux/netfilter.h>
101
102#include <linux/if_tun.h>
103#include <linux/ipv6_route.h>
104#include <linux/route.h>
105#include <linux/sockios.h>
106#include <linux/atalk.h>
107
108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110             unsigned long nr_segs, loff_t pos);
111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112              unsigned long nr_segs, loff_t pos);
113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
114
115static int sock_close(struct inode *inode, struct file *file);
116static unsigned int sock_poll(struct file *file,
117                  struct poll_table_struct *wait);
118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
119#ifdef CONFIG_COMPAT
120static long compat_sock_ioctl(struct file *file,
121                  unsigned int cmd, unsigned long arg);
122#endif
123static int sock_fasync(int fd, struct file *filp, int on);
124static ssize_t sock_sendpage(struct file *file, struct page *page,
125                 int offset, size_t size, loff_t *ppos, int more);
126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
127                struct pipe_inode_info *pipe, size_t len,
128                unsigned int flags);
129
130/*
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
133 */
134
135static const struct file_operations socket_file_ops = {
136    .owner = THIS_MODULE,
137    .llseek = no_llseek,
138    .aio_read = sock_aio_read,
139    .aio_write = sock_aio_write,
140    .poll = sock_poll,
141    .unlocked_ioctl = sock_ioctl,
142#ifdef CONFIG_COMPAT
143    .compat_ioctl = compat_sock_ioctl,
144#endif
145    .mmap = sock_mmap,
146    .open = sock_no_open, /* special open code to disallow open via /proc */
147    .release = sock_close,
148    .fasync = sock_fasync,
149    .sendpage = sock_sendpage,
150    .splice_write = generic_splice_sendpage,
151    .splice_read = sock_splice_read,
152};
153
154/*
155 * The protocol list. Each protocol is registered in here.
156 */
157
158static DEFINE_SPINLOCK(net_family_lock);
159static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
160
161/*
162 * Statistics counters of the socket lists
163 */
164
165static DEFINE_PER_CPU(int, sockets_in_use);
166
167/*
168 * Support routines.
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
171 */
172
173/**
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
178 *
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 */
183
184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
185{
186    if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
187        return -EINVAL;
188    if (ulen == 0)
189        return 0;
190    if (copy_from_user(kaddr, uaddr, ulen))
191        return -EFAULT;
192    return audit_sockaddr(ulen, kaddr);
193}
194
195/**
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
201 *
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
206 * accessible.
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
210 */
211
212static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
213                 void __user *uaddr, int __user *ulen)
214{
215    int err;
216    int len;
217
218    err = get_user(len, ulen);
219    if (err)
220        return err;
221    if (len > klen)
222        len = klen;
223    if (len < 0 || len > sizeof(struct sockaddr_storage))
224        return -EINVAL;
225    if (len) {
226        if (audit_sockaddr(klen, kaddr))
227            return -ENOMEM;
228        if (copy_to_user(uaddr, kaddr, len))
229            return -EFAULT;
230    }
231    /*
232     * "fromlen shall refer to the value before truncation.."
233     * 1003.1g
234     */
235    return __put_user(klen, ulen);
236}
237
238static struct kmem_cache *sock_inode_cachep __read_mostly;
239
240static struct inode *sock_alloc_inode(struct super_block *sb)
241{
242    struct socket_alloc *ei;
243    struct socket_wq *wq;
244
245    ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
246    if (!ei)
247        return NULL;
248    wq = kmalloc(sizeof(*wq), GFP_KERNEL);
249    if (!wq) {
250        kmem_cache_free(sock_inode_cachep, ei);
251        return NULL;
252    }
253    init_waitqueue_head(&wq->wait);
254    wq->fasync_list = NULL;
255    RCU_INIT_POINTER(ei->socket.wq, wq);
256
257    ei->socket.state = SS_UNCONNECTED;
258    ei->socket.flags = 0;
259    ei->socket.ops = NULL;
260    ei->socket.sk = NULL;
261    ei->socket.file = NULL;
262
263    return &ei->vfs_inode;
264}
265
266static void sock_destroy_inode(struct inode *inode)
267{
268    struct socket_alloc *ei;
269    struct socket_wq *wq;
270
271    ei = container_of(inode, struct socket_alloc, vfs_inode);
272    wq = rcu_dereference_protected(ei->socket.wq, 1);
273    kfree_rcu(wq, rcu);
274    kmem_cache_free(sock_inode_cachep, ei);
275}
276
277static void init_once(void *foo)
278{
279    struct socket_alloc *ei = (struct socket_alloc *)foo;
280
281    inode_init_once(&ei->vfs_inode);
282}
283
284static int init_inodecache(void)
285{
286    sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287                          sizeof(struct socket_alloc),
288                          0,
289                          (SLAB_HWCACHE_ALIGN |
290                           SLAB_RECLAIM_ACCOUNT |
291                           SLAB_MEM_SPREAD),
292                          init_once);
293    if (sock_inode_cachep == NULL)
294        return -ENOMEM;
295    return 0;
296}
297
298static const struct super_operations sockfs_ops = {
299    .alloc_inode = sock_alloc_inode,
300    .destroy_inode = sock_destroy_inode,
301    .statfs = simple_statfs,
302};
303
304/*
305 * sockfs_dname() is called from d_path().
306 */
307static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
308{
309    return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
310                dentry->d_inode->i_ino);
311}
312
313static const struct dentry_operations sockfs_dentry_operations = {
314    .d_dname = sockfs_dname,
315};
316
317static struct dentry *sockfs_mount(struct file_system_type *fs_type,
318             int flags, const char *dev_name, void *data)
319{
320    return mount_pseudo(fs_type, "socket:", &sockfs_ops,
321        &sockfs_dentry_operations, SOCKFS_MAGIC);
322}
323
324static struct vfsmount *sock_mnt __read_mostly;
325
326static struct file_system_type sock_fs_type = {
327    .name = "sockfs",
328    .mount = sockfs_mount,
329    .kill_sb = kill_anon_super,
330};
331
332/*
333 * Obtains the first available file descriptor and sets it up for use.
334 *
335 * These functions create file structures and maps them to fd space
336 * of the current process. On success it returns file descriptor
337 * and file struct implicitly stored in sock->file.
338 * Note that another thread may close file descriptor before we return
339 * from this function. We use the fact that now we do not refer
340 * to socket after mapping. If one day we will need it, this
341 * function will increment ref. count on file by 1.
342 *
343 * In any case returned fd MAY BE not valid!
344 * This race condition is unavoidable
345 * with shared fd spaces, we cannot solve it inside kernel,
346 * but we take care of internal coherence yet.
347 */
348
349static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
350{
351    struct qstr name = { .name = "" };
352    struct path path;
353    struct file *file;
354    int fd;
355
356    fd = get_unused_fd_flags(flags);
357    if (unlikely(fd < 0))
358        return fd;
359
360    path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
361    if (unlikely(!path.dentry)) {
362        put_unused_fd(fd);
363        return -ENOMEM;
364    }
365    path.mnt = mntget(sock_mnt);
366
367    d_instantiate(path.dentry, SOCK_INODE(sock));
368    SOCK_INODE(sock)->i_fop = &socket_file_ops;
369
370    file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
371          &socket_file_ops);
372    if (unlikely(!file)) {
373        /* drop dentry, keep inode */
374        ihold(path.dentry->d_inode);
375        path_put(&path);
376        put_unused_fd(fd);
377        return -ENFILE;
378    }
379
380    sock->file = file;
381    file->f_flags = O_RDWR | (flags & O_NONBLOCK);
382    file->f_pos = 0;
383    file->private_data = sock;
384
385    *f = file;
386    return fd;
387}
388
389int sock_map_fd(struct socket *sock, int flags)
390{
391    struct file *newfile;
392    int fd = sock_alloc_file(sock, &newfile, flags);
393
394    if (likely(fd >= 0))
395        fd_install(fd, newfile);
396
397    return fd;
398}
399EXPORT_SYMBOL(sock_map_fd);
400
401static struct socket *sock_from_file(struct file *file, int *err)
402{
403    if (file->f_op == &socket_file_ops)
404        return file->private_data; /* set in sock_map_fd */
405
406    *err = -ENOTSOCK;
407    return NULL;
408}
409
410/**
411 * sockfd_lookup - Go from a file number to its socket slot
412 * @fd: file handle
413 * @err: pointer to an error code return
414 *
415 * The file handle passed in is locked and the socket it is bound
416 * too is returned. If an error occurs the err pointer is overwritten
417 * with a negative errno code and NULL is returned. The function checks
418 * for both invalid handles and passing a handle which is not a socket.
419 *
420 * On a success the socket object pointer is returned.
421 */
422
423struct socket *sockfd_lookup(int fd, int *err)
424{
425    struct file *file;
426    struct socket *sock;
427
428    file = fget(fd);
429    if (!file) {
430        *err = -EBADF;
431        return NULL;
432    }
433
434    sock = sock_from_file(file, err);
435    if (!sock)
436        fput(file);
437    return sock;
438}
439EXPORT_SYMBOL(sockfd_lookup);
440
441static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
442{
443    struct file *file;
444    struct socket *sock;
445
446    *err = -EBADF;
447    file = fget_light(fd, fput_needed);
448    if (file) {
449        sock = sock_from_file(file, err);
450        if (sock)
451            return sock;
452        fput_light(file, *fput_needed);
453    }
454    return NULL;
455}
456
457/**
458 * sock_alloc - allocate a socket
459 *
460 * Allocate a new inode and socket object. The two are bound together
461 * and initialised. The socket is then returned. If we are out of inodes
462 * NULL is returned.
463 */
464
465static struct socket *sock_alloc(void)
466{
467    struct inode *inode;
468    struct socket *sock;
469
470    inode = new_inode_pseudo(sock_mnt->mnt_sb);
471    if (!inode)
472        return NULL;
473
474    sock = SOCKET_I(inode);
475
476    kmemcheck_annotate_bitfield(sock, type);
477    inode->i_ino = get_next_ino();
478    inode->i_mode = S_IFSOCK | S_IRWXUGO;
479    inode->i_uid = current_fsuid();
480    inode->i_gid = current_fsgid();
481
482    percpu_add(sockets_in_use, 1);
483    return sock;
484}
485
486/*
487 * In theory you can't get an open on this inode, but /proc provides
488 * a back door. Remember to keep it shut otherwise you'll let the
489 * creepy crawlies in.
490 */
491
492static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
493{
494    return -ENXIO;
495}
496
497const struct file_operations bad_sock_fops = {
498    .owner = THIS_MODULE,
499    .open = sock_no_open,
500    .llseek = noop_llseek,
501};
502
503/**
504 * sock_release - close a socket
505 * @sock: socket to close
506 *
507 * The socket is released from the protocol stack if it has a release
508 * callback, and the inode is then released if the socket is bound to
509 * an inode not a file.
510 */
511
512void sock_release(struct socket *sock)
513{
514    if (sock->ops) {
515        struct module *owner = sock->ops->owner;
516
517        sock->ops->release(sock);
518        sock->ops = NULL;
519        module_put(owner);
520    }
521
522    if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
523        printk(KERN_ERR "sock_release: fasync list not empty!\n");
524
525    percpu_sub(sockets_in_use, 1);
526    if (!sock->file) {
527        iput(SOCK_INODE(sock));
528        return;
529    }
530    sock->file = NULL;
531}
532EXPORT_SYMBOL(sock_release);
533
534int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
535{
536    *tx_flags = 0;
537    if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
538        *tx_flags |= SKBTX_HW_TSTAMP;
539    if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
540        *tx_flags |= SKBTX_SW_TSTAMP;
541    if (sock_flag(sk, SOCK_WIFI_STATUS))
542        *tx_flags |= SKBTX_WIFI_STATUS;
543    return 0;
544}
545EXPORT_SYMBOL(sock_tx_timestamp);
546
547static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
548                       struct msghdr *msg, size_t size)
549{
550    struct sock_iocb *si = kiocb_to_siocb(iocb);
551
552    sock_update_classid(sock->sk);
553
554    sock_update_netprioidx(sock->sk);
555
556    si->sock = sock;
557    si->scm = NULL;
558    si->msg = msg;
559    si->size = size;
560
561    return sock->ops->sendmsg(iocb, sock, msg, size);
562}
563
564static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
565                 struct msghdr *msg, size_t size)
566{
567    int err = security_socket_sendmsg(sock, msg, size);
568
569    return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
570}
571
572int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
573{
574    struct kiocb iocb;
575    struct sock_iocb siocb;
576    int ret;
577
578    init_sync_kiocb(&iocb, NULL);
579    iocb.private = &siocb;
580    ret = __sock_sendmsg(&iocb, sock, msg, size);
581    if (-EIOCBQUEUED == ret)
582        ret = wait_on_sync_kiocb(&iocb);
583    return ret;
584}
585EXPORT_SYMBOL(sock_sendmsg);
586
587static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
588{
589    struct kiocb iocb;
590    struct sock_iocb siocb;
591    int ret;
592
593    init_sync_kiocb(&iocb, NULL);
594    iocb.private = &siocb;
595    ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
596    if (-EIOCBQUEUED == ret)
597        ret = wait_on_sync_kiocb(&iocb);
598    return ret;
599}
600
601int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
602           struct kvec *vec, size_t num, size_t size)
603{
604    mm_segment_t oldfs = get_fs();
605    int result;
606
607    set_fs(KERNEL_DS);
608    /*
609     * the following is safe, since for compiler definitions of kvec and
610     * iovec are identical, yielding the same in-core layout and alignment
611     */
612    msg->msg_iov = (struct iovec *)vec;
613    msg->msg_iovlen = num;
614    result = sock_sendmsg(sock, msg, size);
615    set_fs(oldfs);
616    return result;
617}
618EXPORT_SYMBOL(kernel_sendmsg);
619
620static int ktime2ts(ktime_t kt, struct timespec *ts)
621{
622    if (kt.tv64) {
623        *ts = ktime_to_timespec(kt);
624        return 1;
625    } else {
626        return 0;
627    }
628}
629
630/*
631 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
632 */
633void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
634    struct sk_buff *skb)
635{
636    int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
637    struct timespec ts[3];
638    int empty = 1;
639    struct skb_shared_hwtstamps *shhwtstamps =
640        skb_hwtstamps(skb);
641
642    /* Race occurred between timestamp enabling and packet
643       receiving. Fill in the current time for now. */
644    if (need_software_tstamp && skb->tstamp.tv64 == 0)
645        __net_timestamp(skb);
646
647    if (need_software_tstamp) {
648        if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
649            struct timeval tv;
650            skb_get_timestamp(skb, &tv);
651            put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
652                 sizeof(tv), &tv);
653        } else {
654            skb_get_timestampns(skb, &ts[0]);
655            put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
656                 sizeof(ts[0]), &ts[0]);
657        }
658    }
659
660
661    memset(ts, 0, sizeof(ts));
662    if (skb->tstamp.tv64 &&
663        sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
664        skb_get_timestampns(skb, ts + 0);
665        empty = 0;
666    }
667    if (shhwtstamps) {
668        if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
669            ktime2ts(shhwtstamps->syststamp, ts + 1))
670            empty = 0;
671        if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
672            ktime2ts(shhwtstamps->hwtstamp, ts + 2))
673            empty = 0;
674    }
675    if (!empty)
676        put_cmsg(msg, SOL_SOCKET,
677             SCM_TIMESTAMPING, sizeof(ts), &ts);
678}
679EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
680
681void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
682    struct sk_buff *skb)
683{
684    int ack;
685
686    if (!sock_flag(sk, SOCK_WIFI_STATUS))
687        return;
688    if (!skb->wifi_acked_valid)
689        return;
690
691    ack = skb->wifi_acked;
692
693    put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
694}
695EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
696
697static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
698                   struct sk_buff *skb)
699{
700    if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
701        put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
702            sizeof(__u32), &skb->dropcount);
703}
704
705void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
706    struct sk_buff *skb)
707{
708    sock_recv_timestamp(msg, sk, skb);
709    sock_recv_drops(msg, sk, skb);
710}
711EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
712
713static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
714                       struct msghdr *msg, size_t size, int flags)
715{
716    struct sock_iocb *si = kiocb_to_siocb(iocb);
717
718    sock_update_classid(sock->sk);
719
720    si->sock = sock;
721    si->scm = NULL;
722    si->msg = msg;
723    si->size = size;
724    si->flags = flags;
725
726    return sock->ops->recvmsg(iocb, sock, msg, size, flags);
727}
728
729static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
730                 struct msghdr *msg, size_t size, int flags)
731{
732    int err = security_socket_recvmsg(sock, msg, size, flags);
733
734    return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
735}
736
737int sock_recvmsg(struct socket *sock, struct msghdr *msg,
738         size_t size, int flags)
739{
740    struct kiocb iocb;
741    struct sock_iocb siocb;
742    int ret;
743
744    init_sync_kiocb(&iocb, NULL);
745    iocb.private = &siocb;
746    ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
747    if (-EIOCBQUEUED == ret)
748        ret = wait_on_sync_kiocb(&iocb);
749    return ret;
750}
751EXPORT_SYMBOL(sock_recvmsg);
752
753static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
754                  size_t size, int flags)
755{
756    struct kiocb iocb;
757    struct sock_iocb siocb;
758    int ret;
759
760    init_sync_kiocb(&iocb, NULL);
761    iocb.private = &siocb;
762    ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
763    if (-EIOCBQUEUED == ret)
764        ret = wait_on_sync_kiocb(&iocb);
765    return ret;
766}
767
768/**
769 * kernel_recvmsg - Receive a message from a socket (kernel space)
770 * @sock: The socket to receive the message from
771 * @msg: Received message
772 * @vec: Input s/g array for message data
773 * @num: Size of input s/g array
774 * @size: Number of bytes to read
775 * @flags: Message flags (MSG_DONTWAIT, etc...)
776 *
777 * On return the msg structure contains the scatter/gather array passed in the
778 * vec argument. The array is modified so that it consists of the unfilled
779 * portion of the original array.
780 *
781 * The returned value is the total number of bytes received, or an error.
782 */
783int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
784           struct kvec *vec, size_t num, size_t size, int flags)
785{
786    mm_segment_t oldfs = get_fs();
787    int result;
788
789    set_fs(KERNEL_DS);
790    /*
791     * the following is safe, since for compiler definitions of kvec and
792     * iovec are identical, yielding the same in-core layout and alignment
793     */
794    msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
795    result = sock_recvmsg(sock, msg, size, flags);
796    set_fs(oldfs);
797    return result;
798}
799EXPORT_SYMBOL(kernel_recvmsg);
800
801static void sock_aio_dtor(struct kiocb *iocb)
802{
803    kfree(iocb->private);
804}
805
806static ssize_t sock_sendpage(struct file *file, struct page *page,
807                 int offset, size_t size, loff_t *ppos, int more)
808{
809    struct socket *sock;
810    int flags;
811
812    sock = file->private_data;
813
814    flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
815    /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
816    flags |= more;
817
818    return kernel_sendpage(sock, page, offset, size, flags);
819}
820
821static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
822                struct pipe_inode_info *pipe, size_t len,
823                unsigned int flags)
824{
825    struct socket *sock = file->private_data;
826
827    if (unlikely(!sock->ops->splice_read))
828        return -EINVAL;
829
830    sock_update_classid(sock->sk);
831
832    return sock->ops->splice_read(sock, ppos, pipe, len, flags);
833}
834
835static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
836                     struct sock_iocb *siocb)
837{
838    if (!is_sync_kiocb(iocb)) {
839        siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
840        if (!siocb)
841            return NULL;
842        iocb->ki_dtor = sock_aio_dtor;
843    }
844
845    siocb->kiocb = iocb;
846    iocb->private = siocb;
847    return siocb;
848}
849
850static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
851        struct file *file, const struct iovec *iov,
852        unsigned long nr_segs)
853{
854    struct socket *sock = file->private_data;
855    size_t size = 0;
856    int i;
857
858    for (i = 0; i < nr_segs; i++)
859        size += iov[i].iov_len;
860
861    msg->msg_name = NULL;
862    msg->msg_namelen = 0;
863    msg->msg_control = NULL;
864    msg->msg_controllen = 0;
865    msg->msg_iov = (struct iovec *)iov;
866    msg->msg_iovlen = nr_segs;
867    msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
868
869    return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
870}
871
872static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
873                unsigned long nr_segs, loff_t pos)
874{
875    struct sock_iocb siocb, *x;
876
877    if (pos != 0)
878        return -ESPIPE;
879
880    if (iocb->ki_left == 0) /* Match SYS5 behaviour */
881        return 0;
882
883
884    x = alloc_sock_iocb(iocb, &siocb);
885    if (!x)
886        return -ENOMEM;
887    return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
888}
889
890static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
891            struct file *file, const struct iovec *iov,
892            unsigned long nr_segs)
893{
894    struct socket *sock = file->private_data;
895    size_t size = 0;
896    int i;
897
898    for (i = 0; i < nr_segs; i++)
899        size += iov[i].iov_len;
900
901    msg->msg_name = NULL;
902    msg->msg_namelen = 0;
903    msg->msg_control = NULL;
904    msg->msg_controllen = 0;
905    msg->msg_iov = (struct iovec *)iov;
906    msg->msg_iovlen = nr_segs;
907    msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
908    if (sock->type == SOCK_SEQPACKET)
909        msg->msg_flags |= MSG_EOR;
910
911    return __sock_sendmsg(iocb, sock, msg, size);
912}
913
914static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
915              unsigned long nr_segs, loff_t pos)
916{
917    struct sock_iocb siocb, *x;
918
919    if (pos != 0)
920        return -ESPIPE;
921
922    x = alloc_sock_iocb(iocb, &siocb);
923    if (!x)
924        return -ENOMEM;
925
926    return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
927}
928
929/*
930 * Atomic setting of ioctl hooks to avoid race
931 * with module unload.
932 */
933
934static DEFINE_MUTEX(br_ioctl_mutex);
935static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
936
937void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
938{
939    mutex_lock(&br_ioctl_mutex);
940    br_ioctl_hook = hook;
941    mutex_unlock(&br_ioctl_mutex);
942}
943EXPORT_SYMBOL(brioctl_set);
944
945static DEFINE_MUTEX(vlan_ioctl_mutex);
946static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
947
948void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
949{
950    mutex_lock(&vlan_ioctl_mutex);
951    vlan_ioctl_hook = hook;
952    mutex_unlock(&vlan_ioctl_mutex);
953}
954EXPORT_SYMBOL(vlan_ioctl_set);
955
956static DEFINE_MUTEX(dlci_ioctl_mutex);
957static int (*dlci_ioctl_hook) (unsigned int, void __user *);
958
959void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
960{
961    mutex_lock(&dlci_ioctl_mutex);
962    dlci_ioctl_hook = hook;
963    mutex_unlock(&dlci_ioctl_mutex);
964}
965EXPORT_SYMBOL(dlci_ioctl_set);
966
967static long sock_do_ioctl(struct net *net, struct socket *sock,
968                 unsigned int cmd, unsigned long arg)
969{
970    int err;
971    void __user *argp = (void __user *)arg;
972
973    err = sock->ops->ioctl(sock, cmd, arg);
974
975    /*
976     * If this ioctl is unknown try to hand it down
977     * to the NIC driver.
978     */
979    if (err == -ENOIOCTLCMD)
980        err = dev_ioctl(net, cmd, argp);
981
982    return err;
983}
984
985/*
986 * With an ioctl, arg may well be a user mode pointer, but we don't know
987 * what to do with it - that's up to the protocol still.
988 */
989
990static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
991{
992    struct socket *sock;
993    struct sock *sk;
994    void __user *argp = (void __user *)arg;
995    int pid, err;
996    struct net *net;
997
998    sock = file->private_data;
999    sk = sock->sk;
1000    net = sock_net(sk);
1001    if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1002        err = dev_ioctl(net, cmd, argp);
1003    } else
1004#ifdef CONFIG_WEXT_CORE
1005    if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1006        err = dev_ioctl(net, cmd, argp);
1007    } else
1008#endif
1009        switch (cmd) {
1010        case FIOSETOWN:
1011        case SIOCSPGRP:
1012            err = -EFAULT;
1013            if (get_user(pid, (int __user *)argp))
1014                break;
1015            err = f_setown(sock->file, pid, 1);
1016            break;
1017        case FIOGETOWN:
1018        case SIOCGPGRP:
1019            err = put_user(f_getown(sock->file),
1020                       (int __user *)argp);
1021            break;
1022        case SIOCGIFBR:
1023        case SIOCSIFBR:
1024        case SIOCBRADDBR:
1025        case SIOCBRDELBR:
1026            err = -ENOPKG;
1027            if (!br_ioctl_hook)
1028                request_module("bridge");
1029
1030            mutex_lock(&br_ioctl_mutex);
1031            if (br_ioctl_hook)
1032                err = br_ioctl_hook(net, cmd, argp);
1033            mutex_unlock(&br_ioctl_mutex);
1034            break;
1035        case SIOCGIFVLAN:
1036        case SIOCSIFVLAN:
1037            err = -ENOPKG;
1038            if (!vlan_ioctl_hook)
1039                request_module("8021q");
1040
1041            mutex_lock(&vlan_ioctl_mutex);
1042            if (vlan_ioctl_hook)
1043                err = vlan_ioctl_hook(net, argp);
1044            mutex_unlock(&vlan_ioctl_mutex);
1045            break;
1046        case SIOCADDDLCI:
1047        case SIOCDELDLCI:
1048            err = -ENOPKG;
1049            if (!dlci_ioctl_hook)
1050                request_module("dlci");
1051
1052            mutex_lock(&dlci_ioctl_mutex);
1053            if (dlci_ioctl_hook)
1054                err = dlci_ioctl_hook(cmd, argp);
1055            mutex_unlock(&dlci_ioctl_mutex);
1056            break;
1057        default:
1058            err = sock_do_ioctl(net, sock, cmd, arg);
1059            break;
1060        }
1061    return err;
1062}
1063
1064int sock_create_lite(int family, int type, int protocol, struct socket **res)
1065{
1066    int err;
1067    struct socket *sock = NULL;
1068
1069    err = security_socket_create(family, type, protocol, 1);
1070    if (err)
1071        goto out;
1072
1073    sock = sock_alloc();
1074    if (!sock) {
1075        err = -ENOMEM;
1076        goto out;
1077    }
1078
1079    sock->type = type;
1080    err = security_socket_post_create(sock, family, type, protocol, 1);
1081    if (err)
1082        goto out_release;
1083
1084out:
1085    *res = sock;
1086    return err;
1087out_release:
1088    sock_release(sock);
1089    sock = NULL;
1090    goto out;
1091}
1092EXPORT_SYMBOL(sock_create_lite);
1093
1094/* No kernel lock held - perfect */
1095static unsigned int sock_poll(struct file *file, poll_table *wait)
1096{
1097    struct socket *sock;
1098
1099    /*
1100     * We can't return errors to poll, so it's either yes or no.
1101     */
1102    sock = file->private_data;
1103    return sock->ops->poll(file, sock, wait);
1104}
1105
1106static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1107{
1108    struct socket *sock = file->private_data;
1109
1110    return sock->ops->mmap(file, sock, vma);
1111}
1112
1113static int sock_close(struct inode *inode, struct file *filp)
1114{
1115    /*
1116     * It was possible the inode is NULL we were
1117     * closing an unfinished socket.
1118     */
1119
1120    if (!inode) {
1121        printk(KERN_DEBUG "sock_close: NULL inode\n");
1122        return 0;
1123    }
1124    sock_release(SOCKET_I(inode));
1125    return 0;
1126}
1127
1128/*
1129 * Update the socket async list
1130 *
1131 * Fasync_list locking strategy.
1132 *
1133 * 1. fasync_list is modified only under process context socket lock
1134 * i.e. under semaphore.
1135 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1136 * or under socket lock
1137 */
1138
1139static int sock_fasync(int fd, struct file *filp, int on)
1140{
1141    struct socket *sock = filp->private_data;
1142    struct sock *sk = sock->sk;
1143    struct socket_wq *wq;
1144
1145    if (sk == NULL)
1146        return -EINVAL;
1147
1148    lock_sock(sk);
1149    wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1150    fasync_helper(fd, filp, on, &wq->fasync_list);
1151
1152    if (!wq->fasync_list)
1153        sock_reset_flag(sk, SOCK_FASYNC);
1154    else
1155        sock_set_flag(sk, SOCK_FASYNC);
1156
1157    release_sock(sk);
1158    return 0;
1159}
1160
1161/* This function may be called only under socket lock or callback_lock or rcu_lock */
1162
1163int sock_wake_async(struct socket *sock, int how, int band)
1164{
1165    struct socket_wq *wq;
1166
1167    if (!sock)
1168        return -1;
1169    rcu_read_lock();
1170    wq = rcu_dereference(sock->wq);
1171    if (!wq || !wq->fasync_list) {
1172        rcu_read_unlock();
1173        return -1;
1174    }
1175    switch (how) {
1176    case SOCK_WAKE_WAITD:
1177        if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1178            break;
1179        goto call_kill;
1180    case SOCK_WAKE_SPACE:
1181        if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1182            break;
1183        /* fall through */
1184    case SOCK_WAKE_IO:
1185call_kill:
1186        kill_fasync(&wq->fasync_list, SIGIO, band);
1187        break;
1188    case SOCK_WAKE_URG:
1189        kill_fasync(&wq->fasync_list, SIGURG, band);
1190    }
1191    rcu_read_unlock();
1192    return 0;
1193}
1194EXPORT_SYMBOL(sock_wake_async);
1195
1196int __sock_create(struct net *net, int family, int type, int protocol,
1197             struct socket **res, int kern)
1198{
1199    int err;
1200    struct socket *sock;
1201    const struct net_proto_family *pf;
1202
1203    /*
1204     * Check protocol is in range
1205     */
1206    if (family < 0 || family >= NPROTO)
1207        return -EAFNOSUPPORT;
1208    if (type < 0 || type >= SOCK_MAX)
1209        return -EINVAL;
1210
1211    /* Compatibility.
1212
1213       This uglymoron is moved from INET layer to here to avoid
1214       deadlock in module load.
1215     */
1216    if (family == PF_INET && type == SOCK_PACKET) {
1217        static int warned;
1218        if (!warned) {
1219            warned = 1;
1220            printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1221                   current->comm);
1222        }
1223        family = PF_PACKET;
1224    }
1225
1226    err = security_socket_create(family, type, protocol, kern);
1227    if (err)
1228        return err;
1229
1230    /*
1231     * Allocate the socket and allow the family to set things up. if
1232     * the protocol is 0, the family is instructed to select an appropriate
1233     * default.
1234     */
1235    sock = sock_alloc();
1236    if (!sock) {
1237        if (net_ratelimit())
1238            printk(KERN_WARNING "socket: no more sockets\n");
1239        return -ENFILE; /* Not exactly a match, but its the
1240                   closest posix thing */
1241    }
1242
1243    sock->type = type;
1244
1245#ifdef CONFIG_MODULES
1246    /* Attempt to load a protocol module if the find failed.
1247     *
1248     * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1249     * requested real, full-featured networking support upon configuration.
1250     * Otherwise module support will break!
1251     */
1252    if (rcu_access_pointer(net_families[family]) == NULL)
1253        request_module("net-pf-%d", family);
1254#endif
1255
1256    rcu_read_lock();
1257    pf = rcu_dereference(net_families[family]);
1258    err = -EAFNOSUPPORT;
1259    if (!pf)
1260        goto out_release;
1261
1262    /*
1263     * We will call the ->create function, that possibly is in a loadable
1264     * module, so we have to bump that loadable module refcnt first.
1265     */
1266    if (!try_module_get(pf->owner))
1267        goto out_release;
1268
1269    /* Now protected by module ref count */
1270    rcu_read_unlock();
1271
1272    err = pf->create(net, sock, protocol, kern);
1273    if (err < 0)
1274        goto out_module_put;
1275
1276    /*
1277     * Now to bump the refcnt of the [loadable] module that owns this
1278     * socket at sock_release time we decrement its refcnt.
1279     */
1280    if (!try_module_get(sock->ops->owner))
1281        goto out_module_busy;
1282
1283    /*
1284     * Now that we're done with the ->create function, the [loadable]
1285     * module can have its refcnt decremented
1286     */
1287    module_put(pf->owner);
1288    err = security_socket_post_create(sock, family, type, protocol, kern);
1289    if (err)
1290        goto out_sock_release;
1291    *res = sock;
1292
1293    return 0;
1294
1295out_module_busy:
1296    err = -EAFNOSUPPORT;
1297out_module_put:
1298    sock->ops = NULL;
1299    module_put(pf->owner);
1300out_sock_release:
1301    sock_release(sock);
1302    return err;
1303
1304out_release:
1305    rcu_read_unlock();
1306    goto out_sock_release;
1307}
1308EXPORT_SYMBOL(__sock_create);
1309
1310int sock_create(int family, int type, int protocol, struct socket **res)
1311{
1312    return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1313}
1314EXPORT_SYMBOL(sock_create);
1315
1316int sock_create_kern(int family, int type, int protocol, struct socket **res)
1317{
1318    return __sock_create(&init_net, family, type, protocol, res, 1);
1319}
1320EXPORT_SYMBOL(sock_create_kern);
1321
1322SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1323{
1324    int retval;
1325    struct socket *sock;
1326    int flags;
1327
1328    /* Check the SOCK_* constants for consistency. */
1329    BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1330    BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1331    BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1332    BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1333
1334    flags = type & ~SOCK_TYPE_MASK;
1335    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1336        return -EINVAL;
1337    type &= SOCK_TYPE_MASK;
1338
1339    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1340        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1341
1342    retval = sock_create(family, type, protocol, &sock);
1343    if (retval < 0)
1344        goto out;
1345
1346    retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1347    if (retval < 0)
1348        goto out_release;
1349
1350out:
1351    /* It may be already another descriptor 8) Not kernel problem. */
1352    return retval;
1353
1354out_release:
1355    sock_release(sock);
1356    return retval;
1357}
1358
1359/*
1360 * Create a pair of connected sockets.
1361 */
1362
1363SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1364        int __user *, usockvec)
1365{
1366    struct socket *sock1, *sock2;
1367    int fd1, fd2, err;
1368    struct file *newfile1, *newfile2;
1369    int flags;
1370
1371    flags = type & ~SOCK_TYPE_MASK;
1372    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1373        return -EINVAL;
1374    type &= SOCK_TYPE_MASK;
1375
1376    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1377        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1378
1379    /*
1380     * Obtain the first socket and check if the underlying protocol
1381     * supports the socketpair call.
1382     */
1383
1384    err = sock_create(family, type, protocol, &sock1);
1385    if (err < 0)
1386        goto out;
1387
1388    err = sock_create(family, type, protocol, &sock2);
1389    if (err < 0)
1390        goto out_release_1;
1391
1392    err = sock1->ops->socketpair(sock1, sock2);
1393    if (err < 0)
1394        goto out_release_both;
1395
1396    fd1 = sock_alloc_file(sock1, &newfile1, flags);
1397    if (unlikely(fd1 < 0)) {
1398        err = fd1;
1399        goto out_release_both;
1400    }
1401
1402    fd2 = sock_alloc_file(sock2, &newfile2, flags);
1403    if (unlikely(fd2 < 0)) {
1404        err = fd2;
1405        fput(newfile1);
1406        put_unused_fd(fd1);
1407        sock_release(sock2);
1408        goto out;
1409    }
1410
1411    audit_fd_pair(fd1, fd2);
1412    fd_install(fd1, newfile1);
1413    fd_install(fd2, newfile2);
1414    /* fd1 and fd2 may be already another descriptors.
1415     * Not kernel problem.
1416     */
1417
1418    err = put_user(fd1, &usockvec[0]);
1419    if (!err)
1420        err = put_user(fd2, &usockvec[1]);
1421    if (!err)
1422        return 0;
1423
1424    sys_close(fd2);
1425    sys_close(fd1);
1426    return err;
1427
1428out_release_both:
1429    sock_release(sock2);
1430out_release_1:
1431    sock_release(sock1);
1432out:
1433    return err;
1434}
1435
1436/*
1437 * Bind a name to a socket. Nothing much to do here since it's
1438 * the protocol's responsibility to handle the local address.
1439 *
1440 * We move the socket address to kernel space before we call
1441 * the protocol layer (having also checked the address is ok).
1442 */
1443
1444SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1445{
1446    struct socket *sock;
1447    struct sockaddr_storage address;
1448    int err, fput_needed;
1449
1450    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1451    if (sock) {
1452        err = move_addr_to_kernel(umyaddr, addrlen, &address);
1453        if (err >= 0) {
1454            err = security_socket_bind(sock,
1455                           (struct sockaddr *)&address,
1456                           addrlen);
1457            if (!err)
1458                err = sock->ops->bind(sock,
1459                              (struct sockaddr *)
1460                              &address, addrlen);
1461        }
1462        fput_light(sock->file, fput_needed);
1463    }
1464    return err;
1465}
1466
1467/*
1468 * Perform a listen. Basically, we allow the protocol to do anything
1469 * necessary for a listen, and if that works, we mark the socket as
1470 * ready for listening.
1471 */
1472
1473SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1474{
1475    struct socket *sock;
1476    int err, fput_needed;
1477    int somaxconn;
1478
1479    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1480    if (sock) {
1481        somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1482        if ((unsigned)backlog > somaxconn)
1483            backlog = somaxconn;
1484
1485        err = security_socket_listen(sock, backlog);
1486        if (!err)
1487            err = sock->ops->listen(sock, backlog);
1488
1489        fput_light(sock->file, fput_needed);
1490    }
1491    return err;
1492}
1493
1494/*
1495 * For accept, we attempt to create a new socket, set up the link
1496 * with the client, wake up the client, then return the new
1497 * connected fd. We collect the address of the connector in kernel
1498 * space and move it to user at the very end. This is unclean because
1499 * we open the socket then return an error.
1500 *
1501 * 1003.1g adds the ability to recvmsg() to query connection pending
1502 * status to recvmsg. We need to add that support in a way thats
1503 * clean when we restucture accept also.
1504 */
1505
1506SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1507        int __user *, upeer_addrlen, int, flags)
1508{
1509    struct socket *sock, *newsock;
1510    struct file *newfile;
1511    int err, len, newfd, fput_needed;
1512    struct sockaddr_storage address;
1513
1514    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1515        return -EINVAL;
1516
1517    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1518        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1519
1520    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1521    if (!sock)
1522        goto out;
1523
1524    err = -ENFILE;
1525    newsock = sock_alloc();
1526    if (!newsock)
1527        goto out_put;
1528
1529    newsock->type = sock->type;
1530    newsock->ops = sock->ops;
1531
1532    /*
1533     * We don't need try_module_get here, as the listening socket (sock)
1534     * has the protocol module (sock->ops->owner) held.
1535     */
1536    __module_get(newsock->ops->owner);
1537
1538    newfd = sock_alloc_file(newsock, &newfile, flags);
1539    if (unlikely(newfd < 0)) {
1540        err = newfd;
1541        sock_release(newsock);
1542        goto out_put;
1543    }
1544
1545    err = security_socket_accept(sock, newsock);
1546    if (err)
1547        goto out_fd;
1548
1549    err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1550    if (err < 0)
1551        goto out_fd;
1552
1553    if (upeer_sockaddr) {
1554        if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1555                      &len, 2) < 0) {
1556            err = -ECONNABORTED;
1557            goto out_fd;
1558        }
1559        err = move_addr_to_user(&address,
1560                    len, upeer_sockaddr, upeer_addrlen);
1561        if (err < 0)
1562            goto out_fd;
1563    }
1564
1565    /* File flags are not inherited via accept() unlike another OSes. */
1566
1567    fd_install(newfd, newfile);
1568    err = newfd;
1569
1570out_put:
1571    fput_light(sock->file, fput_needed);
1572out:
1573    return err;
1574out_fd:
1575    fput(newfile);
1576    put_unused_fd(newfd);
1577    goto out_put;
1578}
1579
1580SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1581        int __user *, upeer_addrlen)
1582{
1583    return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1584}
1585
1586/*
1587 * Attempt to connect to a socket with the server address. The address
1588 * is in user space so we verify it is OK and move it to kernel space.
1589 *
1590 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1591 * break bindings
1592 *
1593 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1594 * other SEQPACKET protocols that take time to connect() as it doesn't
1595 * include the -EINPROGRESS status for such sockets.
1596 */
1597
1598SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1599        int, addrlen)
1600{
1601    struct socket *sock;
1602    struct sockaddr_storage address;
1603    int err, fput_needed;
1604
1605    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1606    if (!sock)
1607        goto out;
1608    err = move_addr_to_kernel(uservaddr, addrlen, &address);
1609    if (err < 0)
1610        goto out_put;
1611
1612    err =
1613        security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1614    if (err)
1615        goto out_put;
1616
1617    err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1618                 sock->file->f_flags);
1619out_put:
1620    fput_light(sock->file, fput_needed);
1621out:
1622    return err;
1623}
1624
1625/*
1626 * Get the local address ('name') of a socket object. Move the obtained
1627 * name to user space.
1628 */
1629
1630SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1631        int __user *, usockaddr_len)
1632{
1633    struct socket *sock;
1634    struct sockaddr_storage address;
1635    int len, err, fput_needed;
1636
1637    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1638    if (!sock)
1639        goto out;
1640
1641    err = security_socket_getsockname(sock);
1642    if (err)
1643        goto out_put;
1644
1645    err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1646    if (err)
1647        goto out_put;
1648    err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1649
1650out_put:
1651    fput_light(sock->file, fput_needed);
1652out:
1653    return err;
1654}
1655
1656/*
1657 * Get the remote address ('name') of a socket object. Move the obtained
1658 * name to user space.
1659 */
1660
1661SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1662        int __user *, usockaddr_len)
1663{
1664    struct socket *sock;
1665    struct sockaddr_storage address;
1666    int len, err, fput_needed;
1667
1668    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1669    if (sock != NULL) {
1670        err = security_socket_getpeername(sock);
1671        if (err) {
1672            fput_light(sock->file, fput_needed);
1673            return err;
1674        }
1675
1676        err =
1677            sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1678                       1);
1679        if (!err)
1680            err = move_addr_to_user(&address, len, usockaddr,
1681                        usockaddr_len);
1682        fput_light(sock->file, fput_needed);
1683    }
1684    return err;
1685}
1686
1687/*
1688 * Send a datagram to a given address. We move the address into kernel
1689 * space and check the user space data area is readable before invoking
1690 * the protocol.
1691 */
1692
1693SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1694        unsigned, flags, struct sockaddr __user *, addr,
1695        int, addr_len)
1696{
1697    struct socket *sock;
1698    struct sockaddr_storage address;
1699    int err;
1700    struct msghdr msg;
1701    struct iovec iov;
1702    int fput_needed;
1703
1704    if (len > INT_MAX)
1705        len = INT_MAX;
1706    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1707    if (!sock)
1708        goto out;
1709
1710    iov.iov_base = buff;
1711    iov.iov_len = len;
1712    msg.msg_name = NULL;
1713    msg.msg_iov = &iov;
1714    msg.msg_iovlen = 1;
1715    msg.msg_control = NULL;
1716    msg.msg_controllen = 0;
1717    msg.msg_namelen = 0;
1718    if (addr) {
1719        err = move_addr_to_kernel(addr, addr_len, &address);
1720        if (err < 0)
1721            goto out_put;
1722        msg.msg_name = (struct sockaddr *)&address;
1723        msg.msg_namelen = addr_len;
1724    }
1725    if (sock->file->f_flags & O_NONBLOCK)
1726        flags |= MSG_DONTWAIT;
1727    msg.msg_flags = flags;
1728    err = sock_sendmsg(sock, &msg, len);
1729
1730out_put:
1731    fput_light(sock->file, fput_needed);
1732out:
1733    return err;
1734}
1735
1736/*
1737 * Send a datagram down a socket.
1738 */
1739
1740SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1741        unsigned, flags)
1742{
1743    return sys_sendto(fd, buff, len, flags, NULL, 0);
1744}
1745
1746/*
1747 * Receive a frame from the socket and optionally record the address of the
1748 * sender. We verify the buffers are writable and if needed move the
1749 * sender address from kernel to user space.
1750 */
1751
1752SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1753        unsigned, flags, struct sockaddr __user *, addr,
1754        int __user *, addr_len)
1755{
1756    struct socket *sock;
1757    struct iovec iov;
1758    struct msghdr msg;
1759    struct sockaddr_storage address;
1760    int err, err2;
1761    int fput_needed;
1762
1763    if (size > INT_MAX)
1764        size = INT_MAX;
1765    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1766    if (!sock)
1767        goto out;
1768
1769    msg.msg_control = NULL;
1770    msg.msg_controllen = 0;
1771    msg.msg_iovlen = 1;
1772    msg.msg_iov = &iov;
1773    iov.iov_len = size;
1774    iov.iov_base = ubuf;
1775    msg.msg_name = (struct sockaddr *)&address;
1776    msg.msg_namelen = sizeof(address);
1777    if (sock->file->f_flags & O_NONBLOCK)
1778        flags |= MSG_DONTWAIT;
1779    err = sock_recvmsg(sock, &msg, size, flags);
1780
1781    if (err >= 0 && addr != NULL) {
1782        err2 = move_addr_to_user(&address,
1783                     msg.msg_namelen, addr, addr_len);
1784        if (err2 < 0)
1785            err = err2;
1786    }
1787
1788    fput_light(sock->file, fput_needed);
1789out:
1790    return err;
1791}
1792
1793/*
1794 * Receive a datagram from a socket.
1795 */
1796
1797asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1798             unsigned flags)
1799{
1800    return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1801}
1802
1803/*
1804 * Set a socket option. Because we don't know the option lengths we have
1805 * to pass the user mode parameter for the protocols to sort out.
1806 */
1807
1808SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1809        char __user *, optval, int, optlen)
1810{
1811    int err, fput_needed;
1812    struct socket *sock;
1813
1814    if (optlen < 0)
1815        return -EINVAL;
1816
1817    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1818    if (sock != NULL) {
1819        err = security_socket_setsockopt(sock, level, optname);
1820        if (err)
1821            goto out_put;
1822
1823        if (level == SOL_SOCKET)
1824            err =
1825                sock_setsockopt(sock, level, optname, optval,
1826                        optlen);
1827        else
1828            err =
1829                sock->ops->setsockopt(sock, level, optname, optval,
1830                          optlen);
1831out_put:
1832        fput_light(sock->file, fput_needed);
1833    }
1834    return err;
1835}
1836
1837/*
1838 * Get a socket option. Because we don't know the option lengths we have
1839 * to pass a user mode parameter for the protocols to sort out.
1840 */
1841
1842SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1843        char __user *, optval, int __user *, optlen)
1844{
1845    int err, fput_needed;
1846    struct socket *sock;
1847
1848    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1849    if (sock != NULL) {
1850        err = security_socket_getsockopt(sock, level, optname);
1851        if (err)
1852            goto out_put;
1853
1854        if (level == SOL_SOCKET)
1855            err =
1856                sock_getsockopt(sock, level, optname, optval,
1857                        optlen);
1858        else
1859            err =
1860                sock->ops->getsockopt(sock, level, optname, optval,
1861                          optlen);
1862out_put:
1863        fput_light(sock->file, fput_needed);
1864    }
1865    return err;
1866}
1867
1868/*
1869 * Shutdown a socket.
1870 */
1871
1872SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1873{
1874    int err, fput_needed;
1875    struct socket *sock;
1876
1877    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1878    if (sock != NULL) {
1879        err = security_socket_shutdown(sock, how);
1880        if (!err)
1881            err = sock->ops->shutdown(sock, how);
1882        fput_light(sock->file, fput_needed);
1883    }
1884    return err;
1885}
1886
1887/* A couple of helpful macros for getting the address of the 32/64 bit
1888 * fields which are the same type (int / unsigned) on our platforms.
1889 */
1890#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1891#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1892#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1893
1894struct used_address {
1895    struct sockaddr_storage name;
1896    unsigned int name_len;
1897};
1898
1899static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1900             struct msghdr *msg_sys, unsigned flags,
1901             struct used_address *used_address)
1902{
1903    struct compat_msghdr __user *msg_compat =
1904        (struct compat_msghdr __user *)msg;
1905    struct sockaddr_storage address;
1906    struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1907    unsigned char ctl[sizeof(struct cmsghdr) + 20]
1908        __attribute__ ((aligned(sizeof(__kernel_size_t))));
1909    /* 20 is size of ipv6_pktinfo */
1910    unsigned char *ctl_buf = ctl;
1911    int err, ctl_len, iov_size, total_len;
1912
1913    err = -EFAULT;
1914    if (MSG_CMSG_COMPAT & flags) {
1915        if (get_compat_msghdr(msg_sys, msg_compat))
1916            return -EFAULT;
1917    } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1918        return -EFAULT;
1919
1920    /* do not move before msg_sys is valid */
1921    err = -EMSGSIZE;
1922    if (msg_sys->msg_iovlen > UIO_MAXIOV)
1923        goto out;
1924
1925    /* Check whether to allocate the iovec area */
1926    err = -ENOMEM;
1927    iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1928    if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1929        iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1930        if (!iov)
1931            goto out;
1932    }
1933
1934    /* This will also move the address data into kernel space */
1935    if (MSG_CMSG_COMPAT & flags) {
1936        err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
1937    } else
1938        err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
1939    if (err < 0)
1940        goto out_freeiov;
1941    total_len = err;
1942
1943    err = -ENOBUFS;
1944
1945    if (msg_sys->msg_controllen > INT_MAX)
1946        goto out_freeiov;
1947    ctl_len = msg_sys->msg_controllen;
1948    if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1949        err =
1950            cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1951                             sizeof(ctl));
1952        if (err)
1953            goto out_freeiov;
1954        ctl_buf = msg_sys->msg_control;
1955        ctl_len = msg_sys->msg_controllen;
1956    } else if (ctl_len) {
1957        if (ctl_len > sizeof(ctl)) {
1958            ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1959            if (ctl_buf == NULL)
1960                goto out_freeiov;
1961        }
1962        err = -EFAULT;
1963        /*
1964         * Careful! Before this, msg_sys->msg_control contains a user pointer.
1965         * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1966         * checking falls down on this.
1967         */
1968        if (copy_from_user(ctl_buf,
1969                   (void __user __force *)msg_sys->msg_control,
1970                   ctl_len))
1971            goto out_freectl;
1972        msg_sys->msg_control = ctl_buf;
1973    }
1974    msg_sys->msg_flags = flags;
1975
1976    if (sock->file->f_flags & O_NONBLOCK)
1977        msg_sys->msg_flags |= MSG_DONTWAIT;
1978    /*
1979     * If this is sendmmsg() and current destination address is same as
1980     * previously succeeded address, omit asking LSM's decision.
1981     * used_address->name_len is initialized to UINT_MAX so that the first
1982     * destination address never matches.
1983     */
1984    if (used_address && msg_sys->msg_name &&
1985        used_address->name_len == msg_sys->msg_namelen &&
1986        !memcmp(&used_address->name, msg_sys->msg_name,
1987            used_address->name_len)) {
1988        err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1989        goto out_freectl;
1990    }
1991    err = sock_sendmsg(sock, msg_sys, total_len);
1992    /*
1993     * If this is sendmmsg() and sending to current destination address was
1994     * successful, remember it.
1995     */
1996    if (used_address && err >= 0) {
1997        used_address->name_len = msg_sys->msg_namelen;
1998        if (msg_sys->msg_name)
1999            memcpy(&used_address->name, msg_sys->msg_name,
2000                   used_address->name_len);
2001    }
2002
2003out_freectl:
2004    if (ctl_buf != ctl)
2005        sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2006out_freeiov:
2007    if (iov != iovstack)
2008        sock_kfree_s(sock->sk, iov, iov_size);
2009out:
2010    return err;
2011}
2012
2013/*
2014 * BSD sendmsg interface
2015 */
2016
2017SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
2018{
2019    int fput_needed, err;
2020    struct msghdr msg_sys;
2021    struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2022
2023    if (!sock)
2024        goto out;
2025
2026    err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2027
2028    fput_light(sock->file, fput_needed);
2029out:
2030    return err;
2031}
2032
2033/*
2034 * Linux sendmmsg interface
2035 */
2036
2037int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2038           unsigned int flags)
2039{
2040    int fput_needed, err, datagrams;
2041    struct socket *sock;
2042    struct mmsghdr __user *entry;
2043    struct compat_mmsghdr __user *compat_entry;
2044    struct msghdr msg_sys;
2045    struct used_address used_address;
2046
2047    if (vlen > UIO_MAXIOV)
2048        vlen = UIO_MAXIOV;
2049
2050    datagrams = 0;
2051
2052    sock = sockfd_lookup_light(fd, &err, &fput_needed);
2053    if (!sock)
2054        return err;
2055
2056    used_address.name_len = UINT_MAX;
2057    entry = mmsg;
2058    compat_entry = (struct compat_mmsghdr __user *)mmsg;
2059    err = 0;
2060
2061    while (datagrams < vlen) {
2062        if (MSG_CMSG_COMPAT & flags) {
2063            err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2064                        &msg_sys, flags, &used_address);
2065            if (err < 0)
2066                break;
2067            err = __put_user(err, &compat_entry->msg_len);
2068            ++compat_entry;
2069        } else {
2070            err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
2071                        &msg_sys, flags, &used_address);
2072            if (err < 0)
2073                break;
2074            err = put_user(err, &entry->msg_len);
2075            ++entry;
2076        }
2077
2078        if (err)
2079            break;
2080        ++datagrams;
2081    }
2082
2083    fput_light(sock->file, fput_needed);
2084
2085    /* We only return an error if no datagrams were able to be sent */
2086    if (datagrams != 0)
2087        return datagrams;
2088
2089    return err;
2090}
2091
2092SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2093        unsigned int, vlen, unsigned int, flags)
2094{
2095    return __sys_sendmmsg(fd, mmsg, vlen, flags);
2096}
2097
2098static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2099             struct msghdr *msg_sys, unsigned flags, int nosec)
2100{
2101    struct compat_msghdr __user *msg_compat =
2102        (struct compat_msghdr __user *)msg;
2103    struct iovec iovstack[UIO_FASTIOV];
2104    struct iovec *iov = iovstack;
2105    unsigned long cmsg_ptr;
2106    int err, iov_size, total_len, len;
2107
2108    /* kernel mode address */
2109    struct sockaddr_storage addr;
2110
2111    /* user mode address pointers */
2112    struct sockaddr __user *uaddr;
2113    int __user *uaddr_len;
2114
2115    if (MSG_CMSG_COMPAT & flags) {
2116        if (get_compat_msghdr(msg_sys, msg_compat))
2117            return -EFAULT;
2118    } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
2119        return -EFAULT;
2120
2121    err = -EMSGSIZE;
2122    if (msg_sys->msg_iovlen > UIO_MAXIOV)
2123        goto out;
2124
2125    /* Check whether to allocate the iovec area */
2126    err = -ENOMEM;
2127    iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2128    if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2129        iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2130        if (!iov)
2131            goto out;
2132    }
2133
2134    /*
2135     * Save the user-mode address (verify_iovec will change the
2136     * kernel msghdr to use the kernel address space)
2137     */
2138
2139    uaddr = (__force void __user *)msg_sys->msg_name;
2140    uaddr_len = COMPAT_NAMELEN(msg);
2141    if (MSG_CMSG_COMPAT & flags) {
2142        err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2143    } else
2144        err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2145    if (err < 0)
2146        goto out_freeiov;
2147    total_len = err;
2148
2149    cmsg_ptr = (unsigned long)msg_sys->msg_control;
2150    msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2151
2152    if (sock->file->f_flags & O_NONBLOCK)
2153        flags |= MSG_DONTWAIT;
2154    err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2155                              total_len, flags);
2156    if (err < 0)
2157        goto out_freeiov;
2158    len = err;
2159
2160    if (uaddr != NULL) {
2161        err = move_addr_to_user(&addr,
2162                    msg_sys->msg_namelen, uaddr,
2163                    uaddr_len);
2164        if (err < 0)
2165            goto out_freeiov;
2166    }
2167    err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2168             COMPAT_FLAGS(msg));
2169    if (err)
2170        goto out_freeiov;
2171    if (MSG_CMSG_COMPAT & flags)
2172        err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2173                 &msg_compat->msg_controllen);
2174    else
2175        err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2176                 &msg->msg_controllen);
2177    if (err)
2178        goto out_freeiov;
2179    err = len;
2180
2181out_freeiov:
2182    if (iov != iovstack)
2183        sock_kfree_s(sock->sk, iov, iov_size);
2184out:
2185    return err;
2186}
2187
2188/*
2189 * BSD recvmsg interface
2190 */
2191
2192SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2193        unsigned int, flags)
2194{
2195    int fput_needed, err;
2196    struct msghdr msg_sys;
2197    struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2198
2199    if (!sock)
2200        goto out;
2201
2202    err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2203
2204    fput_light(sock->file, fput_needed);
2205out:
2206    return err;
2207}
2208
2209/*
2210 * Linux recvmmsg interface
2211 */
2212
2213int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2214           unsigned int flags, struct timespec *timeout)
2215{
2216    int fput_needed, err, datagrams;
2217    struct socket *sock;
2218    struct mmsghdr __user *entry;
2219    struct compat_mmsghdr __user *compat_entry;
2220    struct msghdr msg_sys;
2221    struct timespec end_time;
2222
2223    if (timeout &&
2224        poll_select_set_timeout(&end_time, timeout->tv_sec,
2225                    timeout->tv_nsec))
2226        return -EINVAL;
2227
2228    datagrams = 0;
2229
2230    sock = sockfd_lookup_light(fd, &err, &fput_needed);
2231    if (!sock)
2232        return err;
2233
2234    err = sock_error(sock->sk);
2235    if (err)
2236        goto out_put;
2237
2238    entry = mmsg;
2239    compat_entry = (struct compat_mmsghdr __user *)mmsg;
2240
2241    while (datagrams < vlen) {
2242        /*
2243         * No need to ask LSM for more than the first datagram.
2244         */
2245        if (MSG_CMSG_COMPAT & flags) {
2246            err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2247                        &msg_sys, flags & ~MSG_WAITFORONE,
2248                        datagrams);
2249            if (err < 0)
2250                break;
2251            err = __put_user(err, &compat_entry->msg_len);
2252            ++compat_entry;
2253        } else {
2254            err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2255                        &msg_sys, flags & ~MSG_WAITFORONE,
2256                        datagrams);
2257            if (err < 0)
2258                break;
2259            err = put_user(err, &entry->msg_len);
2260            ++entry;
2261        }
2262
2263        if (err)
2264            break;
2265        ++datagrams;
2266
2267        /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2268        if (flags & MSG_WAITFORONE)
2269            flags |= MSG_DONTWAIT;
2270
2271        if (timeout) {
2272            ktime_get_ts(timeout);
2273            *timeout = timespec_sub(end_time, *timeout);
2274            if (timeout->tv_sec < 0) {
2275                timeout->tv_sec = timeout->tv_nsec = 0;
2276                break;
2277            }
2278
2279            /* Timeout, return less than vlen datagrams */
2280            if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2281                break;
2282        }
2283
2284        /* Out of band data, return right away */
2285        if (msg_sys.msg_flags & MSG_OOB)
2286            break;
2287    }
2288
2289out_put:
2290    fput_light(sock->file, fput_needed);
2291
2292    if (err == 0)
2293        return datagrams;
2294
2295    if (datagrams != 0) {
2296        /*
2297         * We may return less entries than requested (vlen) if the
2298         * sock is non block and there aren't enough datagrams...
2299         */
2300        if (err != -EAGAIN) {
2301            /*
2302             * ... or if recvmsg returns an error after we
2303             * received some datagrams, where we record the
2304             * error to return on the next call or if the
2305             * app asks about it using getsockopt(SO_ERROR).
2306             */
2307            sock->sk->sk_err = -err;
2308        }
2309
2310        return datagrams;
2311    }
2312
2313    return err;
2314}
2315
2316SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2317        unsigned int, vlen, unsigned int, flags,
2318        struct timespec __user *, timeout)
2319{
2320    int datagrams;
2321    struct timespec timeout_sys;
2322
2323    if (!timeout)
2324        return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2325
2326    if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2327        return -EFAULT;
2328
2329    datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2330
2331    if (datagrams > 0 &&
2332        copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2333        datagrams = -EFAULT;
2334
2335    return datagrams;
2336}
2337
2338#ifdef __ARCH_WANT_SYS_SOCKETCALL
2339/* Argument list sizes for sys_socketcall */
2340#define AL(x) ((x) * sizeof(unsigned long))
2341static const unsigned char nargs[21] = {
2342    AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2343    AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2344    AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2345    AL(4), AL(5), AL(4)
2346};
2347
2348#undef AL
2349
2350/*
2351 * System call vectors.
2352 *
2353 * Argument checking cleaned up. Saved 20% in size.
2354 * This function doesn't need to set the kernel lock because
2355 * it is set by the callees.
2356 */
2357
2358SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2359{
2360    unsigned long a[6];
2361    unsigned long a0, a1;
2362    int err;
2363    unsigned int len;
2364
2365    if (call < 1 || call > SYS_SENDMMSG)
2366        return -EINVAL;
2367
2368    len = nargs[call];
2369    if (len > sizeof(a))
2370        return -EINVAL;
2371
2372    /* copy_from_user should be SMP safe. */
2373    if (copy_from_user(a, args, len))
2374        return -EFAULT;
2375
2376    audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2377
2378    a0 = a[0];
2379    a1 = a[1];
2380
2381    switch (call) {
2382    case SYS_SOCKET:
2383        err = sys_socket(a0, a1, a[2]);
2384        break;
2385    case SYS_BIND:
2386        err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2387        break;
2388    case SYS_CONNECT:
2389        err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2390        break;
2391    case SYS_LISTEN:
2392        err = sys_listen(a0, a1);
2393        break;
2394    case SYS_ACCEPT:
2395        err = sys_accept4(a0, (struct sockaddr __user *)a1,
2396                  (int __user *)a[2], 0);
2397        break;
2398    case SYS_GETSOCKNAME:
2399        err =
2400            sys_getsockname(a0, (struct sockaddr __user *)a1,
2401                    (int __user *)a[2]);
2402        break;
2403    case SYS_GETPEERNAME:
2404        err =
2405            sys_getpeername(a0, (struct sockaddr __user *)a1,
2406                    (int __user *)a[2]);
2407        break;
2408    case SYS_SOCKETPAIR:
2409        err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2410        break;
2411    case SYS_SEND:
2412        err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2413        break;
2414    case SYS_SENDTO:
2415        err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2416                 (struct sockaddr __user *)a[4], a[5]);
2417        break;
2418    case SYS_RECV:
2419        err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2420        break;
2421    case SYS_RECVFROM:
2422        err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2423                   (struct sockaddr __user *)a[4],
2424                   (int __user *)a[5]);
2425        break;
2426    case SYS_SHUTDOWN:
2427        err = sys_shutdown(a0, a1);
2428        break;
2429    case SYS_SETSOCKOPT:
2430        err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2431        break;
2432    case SYS_GETSOCKOPT:
2433        err =
2434            sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2435                   (int __user *)a[4]);
2436        break;
2437    case SYS_SENDMSG:
2438        err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2439        break;
2440    case SYS_SENDMMSG:
2441        err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2442        break;
2443    case SYS_RECVMSG:
2444        err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2445        break;
2446    case SYS_RECVMMSG:
2447        err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2448                   (struct timespec __user *)a[4]);
2449        break;
2450    case SYS_ACCEPT4:
2451        err = sys_accept4(a0, (struct sockaddr __user *)a1,
2452                  (int __user *)a[2], a[3]);
2453        break;
2454    default:
2455        err = -EINVAL;
2456        break;
2457    }
2458    return err;
2459}
2460
2461#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2462
2463/**
2464 * sock_register - add a socket protocol handler
2465 * @ops: description of protocol
2466 *
2467 * This function is called by a protocol handler that wants to
2468 * advertise its address family, and have it linked into the
2469 * socket interface. The value ops->family coresponds to the
2470 * socket system call protocol family.
2471 */
2472int sock_register(const struct net_proto_family *ops)
2473{
2474    int err;
2475
2476    if (ops->family >= NPROTO) {
2477        printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2478               NPROTO);
2479        return -ENOBUFS;
2480    }
2481
2482    spin_lock(&net_family_lock);
2483    if (rcu_dereference_protected(net_families[ops->family],
2484                      lockdep_is_held(&net_family_lock)))
2485        err = -EEXIST;
2486    else {
2487        rcu_assign_pointer(net_families[ops->family], ops);
2488        err = 0;
2489    }
2490    spin_unlock(&net_family_lock);
2491
2492    printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2493    return err;
2494}
2495EXPORT_SYMBOL(sock_register);
2496
2497/**
2498 * sock_unregister - remove a protocol handler
2499 * @family: protocol family to remove
2500 *
2501 * This function is called by a protocol handler that wants to
2502 * remove its address family, and have it unlinked from the
2503 * new socket creation.
2504 *
2505 * If protocol handler is a module, then it can use module reference
2506 * counts to protect against new references. If protocol handler is not
2507 * a module then it needs to provide its own protection in
2508 * the ops->create routine.
2509 */
2510void sock_unregister(int family)
2511{
2512    BUG_ON(family < 0 || family >= NPROTO);
2513
2514    spin_lock(&net_family_lock);
2515    RCU_INIT_POINTER(net_families[family], NULL);
2516    spin_unlock(&net_family_lock);
2517
2518    synchronize_rcu();
2519
2520    printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2521}
2522EXPORT_SYMBOL(sock_unregister);
2523
2524static int __init sock_init(void)
2525{
2526    int err;
2527
2528    /*
2529     * Initialize sock SLAB cache.
2530     */
2531
2532    sk_init();
2533
2534    /*
2535     * Initialize skbuff SLAB cache
2536     */
2537    skb_init();
2538
2539    /*
2540     * Initialize the protocols module.
2541     */
2542
2543    init_inodecache();
2544
2545    err = register_filesystem(&sock_fs_type);
2546    if (err)
2547        goto out_fs;
2548    sock_mnt = kern_mount(&sock_fs_type);
2549    if (IS_ERR(sock_mnt)) {
2550        err = PTR_ERR(sock_mnt);
2551        goto out_mount;
2552    }
2553
2554    /* The real protocol initialization is performed in later initcalls.
2555     */
2556
2557#ifdef CONFIG_NETFILTER
2558    netfilter_init();
2559#endif
2560
2561#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2562    skb_timestamping_init();
2563#endif
2564
2565out:
2566    return err;
2567
2568out_mount:
2569    unregister_filesystem(&sock_fs_type);
2570out_fs:
2571    goto out;
2572}
2573
2574core_initcall(sock_init); /* early initcall */
2575
2576#ifdef CONFIG_PROC_FS
2577void socket_seq_show(struct seq_file *seq)
2578{
2579    int cpu;
2580    int counter = 0;
2581
2582    for_each_possible_cpu(cpu)
2583        counter += per_cpu(sockets_in_use, cpu);
2584
2585    /* It can be negative, by the way. 8) */
2586    if (counter < 0)
2587        counter = 0;
2588
2589    seq_printf(seq, "sockets: used %d\n", counter);
2590}
2591#endif /* CONFIG_PROC_FS */
2592
2593#ifdef CONFIG_COMPAT
2594static int do_siocgstamp(struct net *net, struct socket *sock,
2595             unsigned int cmd, void __user *up)
2596{
2597    mm_segment_t old_fs = get_fs();
2598    struct timeval ktv;
2599    int err;
2600
2601    set_fs(KERNEL_DS);
2602    err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2603    set_fs(old_fs);
2604    if (!err)
2605        err = compat_put_timeval(up, &ktv);
2606
2607    return err;
2608}
2609
2610static int do_siocgstampns(struct net *net, struct socket *sock,
2611               unsigned int cmd, void __user *up)
2612{
2613    mm_segment_t old_fs = get_fs();
2614    struct timespec kts;
2615    int err;
2616
2617    set_fs(KERNEL_DS);
2618    err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2619    set_fs(old_fs);
2620    if (!err)
2621        err = compat_put_timespec(up, &kts);
2622
2623    return err;
2624}
2625
2626static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2627{
2628    struct ifreq __user *uifr;
2629    int err;
2630
2631    uifr = compat_alloc_user_space(sizeof(struct ifreq));
2632    if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2633        return -EFAULT;
2634
2635    err = dev_ioctl(net, SIOCGIFNAME, uifr);
2636    if (err)
2637        return err;
2638
2639    if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2640        return -EFAULT;
2641
2642    return 0;
2643}
2644
2645static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2646{
2647    struct compat_ifconf ifc32;
2648    struct ifconf ifc;
2649    struct ifconf __user *uifc;
2650    struct compat_ifreq __user *ifr32;
2651    struct ifreq __user *ifr;
2652    unsigned int i, j;
2653    int err;
2654
2655    if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2656        return -EFAULT;
2657
2658    if (ifc32.ifcbuf == 0) {
2659        ifc32.ifc_len = 0;
2660        ifc.ifc_len = 0;
2661        ifc.ifc_req = NULL;
2662        uifc = compat_alloc_user_space(sizeof(struct ifconf));
2663    } else {
2664        size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2665            sizeof(struct ifreq);
2666        uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2667        ifc.ifc_len = len;
2668        ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2669        ifr32 = compat_ptr(ifc32.ifcbuf);
2670        for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2671            if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2672                return -EFAULT;
2673            ifr++;
2674            ifr32++;
2675        }
2676    }
2677    if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2678        return -EFAULT;
2679
2680    err = dev_ioctl(net, SIOCGIFCONF, uifc);
2681    if (err)
2682        return err;
2683
2684    if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2685        return -EFAULT;
2686
2687    ifr = ifc.ifc_req;
2688    ifr32 = compat_ptr(ifc32.ifcbuf);
2689    for (i = 0, j = 0;
2690         i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2691         i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2692        if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2693            return -EFAULT;
2694        ifr32++;
2695        ifr++;
2696    }
2697
2698    if (ifc32.ifcbuf == 0) {
2699        /* Translate from 64-bit structure multiple to
2700         * a 32-bit one.
2701         */
2702        i = ifc.ifc_len;
2703        i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2704        ifc32.ifc_len = i;
2705    } else {
2706        ifc32.ifc_len = i;
2707    }
2708    if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2709        return -EFAULT;
2710
2711    return 0;
2712}
2713
2714static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2715{
2716    struct compat_ethtool_rxnfc __user *compat_rxnfc;
2717    bool convert_in = false, convert_out = false;
2718    size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2719    struct ethtool_rxnfc __user *rxnfc;
2720    struct ifreq __user *ifr;
2721    u32 rule_cnt = 0, actual_rule_cnt;
2722    u32 ethcmd;
2723    u32 data;
2724    int ret;
2725
2726    if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2727        return -EFAULT;
2728
2729    compat_rxnfc = compat_ptr(data);
2730
2731    if (get_user(ethcmd, &compat_rxnfc->cmd))
2732        return -EFAULT;
2733
2734    /* Most ethtool structures are defined without padding.
2735     * Unfortunately struct ethtool_rxnfc is an exception.
2736     */
2737    switch (ethcmd) {
2738    default:
2739        break;
2740    case ETHTOOL_GRXCLSRLALL:
2741        /* Buffer size is variable */
2742        if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2743            return -EFAULT;
2744        if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2745            return -ENOMEM;
2746        buf_size += rule_cnt * sizeof(u32);
2747        /* fall through */
2748    case ETHTOOL_GRXRINGS:
2749    case ETHTOOL_GRXCLSRLCNT:
2750    case ETHTOOL_GRXCLSRULE:
2751    case ETHTOOL_SRXCLSRLINS:
2752        convert_out = true;
2753        /* fall through */
2754    case ETHTOOL_SRXCLSRLDEL:
2755        buf_size += sizeof(struct ethtool_rxnfc);
2756        convert_in = true;
2757        break;
2758    }
2759
2760    ifr = compat_alloc_user_space(buf_size);
2761    rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2762
2763    if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2764        return -EFAULT;
2765
2766    if (put_user(convert_in ? rxnfc : compat_ptr(data),
2767             &ifr->ifr_ifru.ifru_data))
2768        return -EFAULT;
2769
2770    if (convert_in) {
2771        /* We expect there to be holes between fs.m_ext and
2772         * fs.ring_cookie and at the end of fs, but nowhere else.
2773         */
2774        BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2775                 sizeof(compat_rxnfc->fs.m_ext) !=
2776                 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2777                 sizeof(rxnfc->fs.m_ext));
2778        BUILD_BUG_ON(
2779            offsetof(struct compat_ethtool_rxnfc, fs.location) -
2780            offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2781            offsetof(struct ethtool_rxnfc, fs.location) -
2782            offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2783
2784        if (copy_in_user(rxnfc, compat_rxnfc,
2785                 (void *)(&rxnfc->fs.m_ext + 1) -
2786                 (void *)rxnfc) ||
2787            copy_in_user(&rxnfc->fs.ring_cookie,
2788                 &compat_rxnfc->fs.ring_cookie,
2789                 (void *)(&rxnfc->fs.location + 1) -
2790                 (void *)&rxnfc->fs.ring_cookie) ||
2791            copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2792                 sizeof(rxnfc->rule_cnt)))
2793            return -EFAULT;
2794    }
2795
2796    ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2797    if (ret)
2798        return ret;
2799
2800    if (convert_out) {
2801        if (copy_in_user(compat_rxnfc, rxnfc,
2802                 (const void *)(&rxnfc->fs.m_ext + 1) -
2803                 (const void *)rxnfc) ||
2804            copy_in_user(&compat_rxnfc->fs.ring_cookie,
2805                 &rxnfc->fs.ring_cookie,
2806                 (const void *)(&rxnfc->fs.location + 1) -
2807                 (const void *)&rxnfc->fs.ring_cookie) ||
2808            copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2809                 sizeof(rxnfc->rule_cnt)))
2810            return -EFAULT;
2811
2812        if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2813            /* As an optimisation, we only copy the actual
2814             * number of rules that the underlying
2815             * function returned. Since Mallory might
2816             * change the rule count in user memory, we
2817             * check that it is less than the rule count
2818             * originally given (as the user buffer size),
2819             * which has been range-checked.
2820             */
2821            if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2822                return -EFAULT;
2823            if (actual_rule_cnt < rule_cnt)
2824                rule_cnt = actual_rule_cnt;
2825            if (copy_in_user(&compat_rxnfc->rule_locs[0],
2826                     &rxnfc->rule_locs[0],
2827                     rule_cnt * sizeof(u32)))
2828                return -EFAULT;
2829        }
2830    }
2831
2832    return 0;
2833}
2834
2835static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2836{
2837    void __user *uptr;
2838    compat_uptr_t uptr32;
2839    struct ifreq __user *uifr;
2840
2841    uifr = compat_alloc_user_space(sizeof(*uifr));
2842    if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2843        return -EFAULT;
2844
2845    if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2846        return -EFAULT;
2847
2848    uptr = compat_ptr(uptr32);
2849
2850    if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2851        return -EFAULT;
2852
2853    return dev_ioctl(net, SIOCWANDEV, uifr);
2854}
2855
2856static int bond_ioctl(struct net *net, unsigned int cmd,
2857             struct compat_ifreq __user *ifr32)
2858{
2859    struct ifreq kifr;
2860    struct ifreq __user *uifr;
2861    mm_segment_t old_fs;
2862    int err;
2863    u32 data;
2864    void __user *datap;
2865
2866    switch (cmd) {
2867    case SIOCBONDENSLAVE:
2868    case SIOCBONDRELEASE:
2869    case SIOCBONDSETHWADDR:
2870    case SIOCBONDCHANGEACTIVE:
2871        if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2872            return -EFAULT;
2873
2874        old_fs = get_fs();
2875        set_fs(KERNEL_DS);
2876        err = dev_ioctl(net, cmd,
2877                (struct ifreq __user __force *) &kifr);
2878        set_fs(old_fs);
2879
2880        return err;
2881    case SIOCBONDSLAVEINFOQUERY:
2882    case SIOCBONDINFOQUERY:
2883        uifr = compat_alloc_user_space(sizeof(*uifr));
2884        if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2885            return -EFAULT;
2886
2887        if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2888            return -EFAULT;
2889
2890        datap = compat_ptr(data);
2891        if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2892            return -EFAULT;
2893
2894        return dev_ioctl(net, cmd, uifr);
2895    default:
2896        return -ENOIOCTLCMD;
2897    }
2898}
2899
2900static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2901                 struct compat_ifreq __user *u_ifreq32)
2902{
2903    struct ifreq __user *u_ifreq64;
2904    char tmp_buf[IFNAMSIZ];
2905    void __user *data64;
2906    u32 data32;
2907
2908    if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2909               IFNAMSIZ))
2910        return -EFAULT;
2911    if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2912        return -EFAULT;
2913    data64 = compat_ptr(data32);
2914
2915    u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2916
2917    /* Don't check these user accesses, just let that get trapped
2918     * in the ioctl handler instead.
2919     */
2920    if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2921             IFNAMSIZ))
2922        return -EFAULT;
2923    if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2924        return -EFAULT;
2925
2926    return dev_ioctl(net, cmd, u_ifreq64);
2927}
2928
2929static int dev_ifsioc(struct net *net, struct socket *sock,
2930             unsigned int cmd, struct compat_ifreq __user *uifr32)
2931{
2932    struct ifreq __user *uifr;
2933    int err;
2934
2935    uifr = compat_alloc_user_space(sizeof(*uifr));
2936    if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2937        return -EFAULT;
2938
2939    err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2940
2941    if (!err) {
2942        switch (cmd) {
2943        case SIOCGIFFLAGS:
2944        case SIOCGIFMETRIC:
2945        case SIOCGIFMTU:
2946        case SIOCGIFMEM:
2947        case SIOCGIFHWADDR:
2948        case SIOCGIFINDEX:
2949        case SIOCGIFADDR:
2950        case SIOCGIFBRDADDR:
2951        case SIOCGIFDSTADDR:
2952        case SIOCGIFNETMASK:
2953        case SIOCGIFPFLAGS:
2954        case SIOCGIFTXQLEN:
2955        case SIOCGMIIPHY:
2956        case SIOCGMIIREG:
2957            if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2958                err = -EFAULT;
2959            break;
2960        }
2961    }
2962    return err;
2963}
2964
2965static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2966            struct compat_ifreq __user *uifr32)
2967{
2968    struct ifreq ifr;
2969    struct compat_ifmap __user *uifmap32;
2970    mm_segment_t old_fs;
2971    int err;
2972
2973    uifmap32 = &uifr32->ifr_ifru.ifru_map;
2974    err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2975    err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2976    err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2977    err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2978    err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2979    err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2980    err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2981    if (err)
2982        return -EFAULT;
2983
2984    old_fs = get_fs();
2985    set_fs(KERNEL_DS);
2986    err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2987    set_fs(old_fs);
2988
2989    if (cmd == SIOCGIFMAP && !err) {
2990        err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2991        err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2992        err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2993        err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2994        err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2995        err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2996        err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2997        if (err)
2998            err = -EFAULT;
2999    }
3000    return err;
3001}
3002
3003static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3004{
3005    void __user *uptr;
3006    compat_uptr_t uptr32;
3007    struct ifreq __user *uifr;
3008
3009    uifr = compat_alloc_user_space(sizeof(*uifr));
3010    if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3011        return -EFAULT;
3012
3013    if (get_user(uptr32, &uifr32->ifr_data))
3014        return -EFAULT;
3015
3016    uptr = compat_ptr(uptr32);
3017
3018    if (put_user(uptr, &uifr->ifr_data))
3019        return -EFAULT;
3020
3021    return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3022}
3023
3024struct rtentry32 {
3025    u32 rt_pad1;
3026    struct sockaddr rt_dst; /* target address */
3027    struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3028    struct sockaddr rt_genmask; /* target network mask (IP) */
3029    unsigned short rt_flags;
3030    short rt_pad2;
3031    u32 rt_pad3;
3032    unsigned char rt_tos;
3033    unsigned char rt_class;
3034    short rt_pad4;
3035    short rt_metric; /* +1 for binary compatibility! */
3036    /* char * */ u32 rt_dev; /* forcing the device at add */
3037    u32 rt_mtu; /* per route MTU/Window */
3038    u32 rt_window; /* Window clamping */
3039    unsigned short rt_irtt; /* Initial RTT */
3040};
3041
3042struct in6_rtmsg32 {
3043    struct in6_addr rtmsg_dst;
3044    struct in6_addr rtmsg_src;
3045    struct in6_addr rtmsg_gateway;
3046    u32 rtmsg_type;
3047    u16 rtmsg_dst_len;
3048    u16 rtmsg_src_len;
3049    u32 rtmsg_metric;
3050    u32 rtmsg_info;
3051    u32 rtmsg_flags;
3052    s32 rtmsg_ifindex;
3053};
3054
3055static int routing_ioctl(struct net *net, struct socket *sock,
3056             unsigned int cmd, void __user *argp)
3057{
3058    int ret;
3059    void *r = NULL;
3060    struct in6_rtmsg r6;
3061    struct rtentry r4;
3062    char devname[16];
3063    u32 rtdev;
3064    mm_segment_t old_fs = get_fs();
3065
3066    if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3067        struct in6_rtmsg32 __user *ur6 = argp;
3068        ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3069            3 * sizeof(struct in6_addr));
3070        ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3071        ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3072        ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3073        ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3074        ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3075        ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3076        ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3077
3078        r = (void *) &r6;
3079    } else { /* ipv4 */
3080        struct rtentry32 __user *ur4 = argp;
3081        ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3082                    3 * sizeof(struct sockaddr));
3083        ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3084        ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3085        ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3086        ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3087        ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3088        ret |= __get_user(rtdev, &(ur4->rt_dev));
3089        if (rtdev) {
3090            ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3091            r4.rt_dev = (char __user __force *)devname;
3092            devname[15] = 0;
3093        } else
3094            r4.rt_dev = NULL;
3095
3096        r = (void *) &r4;
3097    }
3098
3099    if (ret) {
3100        ret = -EFAULT;
3101        goto out;
3102    }
3103
3104    set_fs(KERNEL_DS);
3105    ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3106    set_fs(old_fs);
3107
3108out:
3109    return ret;
3110}
3111
3112/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3113 * for some operations; this forces use of the newer bridge-utils that
3114 * use compatible ioctls
3115 */
3116static int old_bridge_ioctl(compat_ulong_t __user *argp)
3117{
3118    compat_ulong_t tmp;
3119
3120    if (get_user(tmp, argp))
3121        return -EFAULT;
3122    if (tmp == BRCTL_GET_VERSION)
3123        return BRCTL_VERSION + 1;
3124    return -EINVAL;
3125}
3126
3127static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3128             unsigned int cmd, unsigned long arg)
3129{
3130    void __user *argp = compat_ptr(arg);
3131    struct sock *sk = sock->sk;
3132    struct net *net = sock_net(sk);
3133
3134    if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3135        return siocdevprivate_ioctl(net, cmd, argp);
3136
3137    switch (cmd) {
3138    case SIOCSIFBR:
3139    case SIOCGIFBR:
3140        return old_bridge_ioctl(argp);
3141    case SIOCGIFNAME:
3142        return dev_ifname32(net, argp);
3143    case SIOCGIFCONF:
3144        return dev_ifconf(net, argp);
3145    case SIOCETHTOOL:
3146        return ethtool_ioctl(net, argp);
3147    case SIOCWANDEV:
3148        return compat_siocwandev(net, argp);
3149    case SIOCGIFMAP:
3150    case SIOCSIFMAP:
3151        return compat_sioc_ifmap(net, cmd, argp);
3152    case SIOCBONDENSLAVE:
3153    case SIOCBONDRELEASE:
3154    case SIOCBONDSETHWADDR:
3155    case SIOCBONDSLAVEINFOQUERY:
3156    case SIOCBONDINFOQUERY:
3157    case SIOCBONDCHANGEACTIVE:
3158        return bond_ioctl(net, cmd, argp);
3159    case SIOCADDRT:
3160    case SIOCDELRT:
3161        return routing_ioctl(net, sock, cmd, argp);
3162    case SIOCGSTAMP:
3163        return do_siocgstamp(net, sock, cmd, argp);
3164    case SIOCGSTAMPNS:
3165        return do_siocgstampns(net, sock, cmd, argp);
3166    case SIOCSHWTSTAMP:
3167        return compat_siocshwtstamp(net, argp);
3168
3169    case FIOSETOWN:
3170    case SIOCSPGRP:
3171    case FIOGETOWN:
3172    case SIOCGPGRP:
3173    case SIOCBRADDBR:
3174    case SIOCBRDELBR:
3175    case SIOCGIFVLAN:
3176    case SIOCSIFVLAN:
3177    case SIOCADDDLCI:
3178    case SIOCDELDLCI:
3179        return sock_ioctl(file, cmd, arg);
3180
3181    case SIOCGIFFLAGS:
3182    case SIOCSIFFLAGS:
3183    case SIOCGIFMETRIC:
3184    case SIOCSIFMETRIC:
3185    case SIOCGIFMTU:
3186    case SIOCSIFMTU:
3187    case SIOCGIFMEM:
3188    case SIOCSIFMEM:
3189    case SIOCGIFHWADDR:
3190    case SIOCSIFHWADDR:
3191    case SIOCADDMULTI:
3192    case SIOCDELMULTI:
3193    case SIOCGIFINDEX:
3194    case SIOCGIFADDR:
3195    case SIOCSIFADDR:
3196    case SIOCSIFHWBROADCAST:
3197    case SIOCDIFADDR:
3198    case SIOCGIFBRDADDR:
3199    case SIOCSIFBRDADDR:
3200    case SIOCGIFDSTADDR:
3201    case SIOCSIFDSTADDR:
3202    case SIOCGIFNETMASK:
3203    case SIOCSIFNETMASK:
3204    case SIOCSIFPFLAGS:
3205    case SIOCGIFPFLAGS:
3206    case SIOCGIFTXQLEN:
3207    case SIOCSIFTXQLEN:
3208    case SIOCBRADDIF:
3209    case SIOCBRDELIF:
3210    case SIOCSIFNAME:
3211    case SIOCGMIIPHY:
3212    case SIOCGMIIREG:
3213    case SIOCSMIIREG:
3214        return dev_ifsioc(net, sock, cmd, argp);
3215
3216    case SIOCSARP:
3217    case SIOCGARP:
3218    case SIOCDARP:
3219    case SIOCATMARK:
3220        return sock_do_ioctl(net, sock, cmd, arg);
3221    }
3222
3223    return -ENOIOCTLCMD;
3224}
3225
3226static long compat_sock_ioctl(struct file *file, unsigned cmd,
3227                  unsigned long arg)
3228{
3229    struct socket *sock = file->private_data;
3230    int ret = -ENOIOCTLCMD;
3231    struct sock *sk;
3232    struct net *net;
3233
3234    sk = sock->sk;
3235    net = sock_net(sk);
3236
3237    if (sock->ops->compat_ioctl)
3238        ret = sock->ops->compat_ioctl(sock, cmd, arg);
3239
3240    if (ret == -ENOIOCTLCMD &&
3241        (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3242        ret = compat_wext_handle_ioctl(net, cmd, arg);
3243
3244    if (ret == -ENOIOCTLCMD)
3245        ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3246
3247    return ret;
3248}
3249#endif
3250
3251int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3252{
3253    return sock->ops->bind(sock, addr, addrlen);
3254}
3255EXPORT_SYMBOL(kernel_bind);
3256
3257int kernel_listen(struct socket *sock, int backlog)
3258{
3259    return sock->ops->listen(sock, backlog);
3260}
3261EXPORT_SYMBOL(kernel_listen);
3262
3263int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3264{
3265    struct sock *sk = sock->sk;
3266    int err;
3267
3268    err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3269                   newsock);
3270    if (err < 0)
3271        goto done;
3272
3273    err = sock->ops->accept(sock, *newsock, flags);
3274    if (err < 0) {
3275        sock_release(*newsock);
3276        *newsock = NULL;
3277        goto done;
3278    }
3279
3280    (*newsock)->ops = sock->ops;
3281    __module_get((*newsock)->ops->owner);
3282
3283done:
3284    return err;
3285}
3286EXPORT_SYMBOL(kernel_accept);
3287
3288int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3289           int flags)
3290{
3291    return sock->ops->connect(sock, addr, addrlen, flags);
3292}
3293EXPORT_SYMBOL(kernel_connect);
3294
3295int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3296             int *addrlen)
3297{
3298    return sock->ops->getname(sock, addr, addrlen, 0);
3299}
3300EXPORT_SYMBOL(kernel_getsockname);
3301
3302int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3303             int *addrlen)
3304{
3305    return sock->ops->getname(sock, addr, addrlen, 1);
3306}
3307EXPORT_SYMBOL(kernel_getpeername);
3308
3309int kernel_getsockopt(struct socket *sock, int level, int optname,
3310            char *optval, int *optlen)
3311{
3312    mm_segment_t oldfs = get_fs();
3313    char __user *uoptval;
3314    int __user *uoptlen;
3315    int err;
3316
3317    uoptval = (char __user __force *) optval;
3318    uoptlen = (int __user __force *) optlen;
3319
3320    set_fs(KERNEL_DS);
3321    if (level == SOL_SOCKET)
3322        err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3323    else
3324        err = sock->ops->getsockopt(sock, level, optname, uoptval,
3325                        uoptlen);
3326    set_fs(oldfs);
3327    return err;
3328}
3329EXPORT_SYMBOL(kernel_getsockopt);
3330
3331int kernel_setsockopt(struct socket *sock, int level, int optname,
3332            char *optval, unsigned int optlen)
3333{
3334    mm_segment_t oldfs = get_fs();
3335    char __user *uoptval;
3336    int err;
3337
3338    uoptval = (char __user __force *) optval;
3339
3340    set_fs(KERNEL_DS);
3341    if (level == SOL_SOCKET)
3342        err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3343    else
3344        err = sock->ops->setsockopt(sock, level, optname, uoptval,
3345                        optlen);
3346    set_fs(oldfs);
3347    return err;
3348}
3349EXPORT_SYMBOL(kernel_setsockopt);
3350
3351int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3352            size_t size, int flags)
3353{
3354    sock_update_classid(sock->sk);
3355
3356    if (sock->ops->sendpage)
3357        return sock->ops->sendpage(sock, page, offset, size, flags);
3358
3359    return sock_no_sendpage(sock, page, offset, size, flags);
3360}
3361EXPORT_SYMBOL(kernel_sendpage);
3362
3363int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3364{
3365    mm_segment_t oldfs = get_fs();
3366    int err;
3367
3368    set_fs(KERNEL_DS);
3369    err = sock->ops->ioctl(sock, cmd, arg);
3370    set_fs(oldfs);
3371
3372    return err;
3373}
3374EXPORT_SYMBOL(kernel_sock_ioctl);
3375
3376int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3377{
3378    return sock->ops->shutdown(sock, how);
3379}
3380EXPORT_SYMBOL(kernel_sock_shutdown);
3381

Archive Download this file



interactive