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/if_bridge.h>
73#include <linux/if_frad.h>
74#include <linux/if_vlan.h>
75#include <linux/ptp_classify.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#include <linux/xattr.h>
92
93#include <asm/uaccess.h>
94#include <asm/unistd.h>
95
96#include <net/compat.h>
97#include <net/wext.h>
98#include <net/cls_cgroup.h>
99
100#include <net/sock.h>
101#include <linux/netfilter.h>
102
103#include <linux/if_tun.h>
104#include <linux/ipv6_route.h>
105#include <linux/route.h>
106#include <linux/sockios.h>
107#include <linux/atalk.h>
108#include <net/busy_poll.h>
109
110#ifdef CONFIG_NET_RX_BUSY_POLL
111unsigned int sysctl_net_busy_read __read_mostly;
112unsigned int sysctl_net_busy_poll __read_mostly;
113#endif
114
115static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
116static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
117             unsigned long nr_segs, loff_t pos);
118static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
119              unsigned long nr_segs, loff_t pos);
120static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121
122static int sock_close(struct inode *inode, struct file *file);
123static unsigned int sock_poll(struct file *file,
124                  struct poll_table_struct *wait);
125static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126#ifdef CONFIG_COMPAT
127static long compat_sock_ioctl(struct file *file,
128                  unsigned int cmd, unsigned long arg);
129#endif
130static int sock_fasync(int fd, struct file *filp, int on);
131static ssize_t sock_sendpage(struct file *file, struct page *page,
132                 int offset, size_t size, loff_t *ppos, int more);
133static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
134                struct pipe_inode_info *pipe, size_t len,
135                unsigned int flags);
136
137/*
138 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
139 * in the operation structures but are done directly via the socketcall() multiplexor.
140 */
141
142static const struct file_operations socket_file_ops = {
143    .owner = THIS_MODULE,
144    .llseek = no_llseek,
145    .aio_read = sock_aio_read,
146    .aio_write = sock_aio_write,
147    .poll = sock_poll,
148    .unlocked_ioctl = sock_ioctl,
149#ifdef CONFIG_COMPAT
150    .compat_ioctl = compat_sock_ioctl,
151#endif
152    .mmap = sock_mmap,
153    .open = sock_no_open, /* special open code to disallow open via /proc */
154    .release = sock_close,
155    .fasync = sock_fasync,
156    .sendpage = sock_sendpage,
157    .splice_write = generic_splice_sendpage,
158    .splice_read = sock_splice_read,
159};
160
161/*
162 * The protocol list. Each protocol is registered in here.
163 */
164
165static DEFINE_SPINLOCK(net_family_lock);
166static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
167
168/*
169 * Statistics counters of the socket lists
170 */
171
172static DEFINE_PER_CPU(int, sockets_in_use);
173
174/*
175 * Support routines.
176 * Move socket addresses back and forth across the kernel/user
177 * divide and look after the messy bits.
178 */
179
180/**
181 * move_addr_to_kernel - copy a socket address into kernel space
182 * @uaddr: Address in user space
183 * @kaddr: Address in kernel space
184 * @ulen: Length in user space
185 *
186 * The address is copied into kernel space. If the provided address is
187 * too long an error code of -EINVAL is returned. If the copy gives
188 * invalid addresses -EFAULT is returned. On a success 0 is returned.
189 */
190
191int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
192{
193    if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
194        return -EINVAL;
195    if (ulen == 0)
196        return 0;
197    if (copy_from_user(kaddr, uaddr, ulen))
198        return -EFAULT;
199    return audit_sockaddr(ulen, kaddr);
200}
201
202/**
203 * move_addr_to_user - copy an address to user space
204 * @kaddr: kernel space address
205 * @klen: length of address in kernel
206 * @uaddr: user space address
207 * @ulen: pointer to user length field
208 *
209 * The value pointed to by ulen on entry is the buffer length available.
210 * This is overwritten with the buffer space used. -EINVAL is returned
211 * if an overlong buffer is specified or a negative buffer size. -EFAULT
212 * is returned if either the buffer or the length field are not
213 * accessible.
214 * After copying the data up to the limit the user specifies, the true
215 * length of the data is written over the length limit the user
216 * specified. Zero is returned for a success.
217 */
218
219static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
220                 void __user *uaddr, int __user *ulen)
221{
222    int err;
223    int len;
224
225    BUG_ON(klen > sizeof(struct sockaddr_storage));
226    err = get_user(len, ulen);
227    if (err)
228        return err;
229    if (len > klen)
230        len = klen;
231    if (len < 0)
232        return -EINVAL;
233    if (len) {
234        if (audit_sockaddr(klen, kaddr))
235            return -ENOMEM;
236        if (copy_to_user(uaddr, kaddr, len))
237            return -EFAULT;
238    }
239    /*
240     * "fromlen shall refer to the value before truncation.."
241     * 1003.1g
242     */
243    return __put_user(klen, ulen);
244}
245
246static struct kmem_cache *sock_inode_cachep __read_mostly;
247
248static struct inode *sock_alloc_inode(struct super_block *sb)
249{
250    struct socket_alloc *ei;
251    struct socket_wq *wq;
252
253    ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
254    if (!ei)
255        return NULL;
256    wq = kmalloc(sizeof(*wq), GFP_KERNEL);
257    if (!wq) {
258        kmem_cache_free(sock_inode_cachep, ei);
259        return NULL;
260    }
261    init_waitqueue_head(&wq->wait);
262    wq->fasync_list = NULL;
263    RCU_INIT_POINTER(ei->socket.wq, wq);
264
265    ei->socket.state = SS_UNCONNECTED;
266    ei->socket.flags = 0;
267    ei->socket.ops = NULL;
268    ei->socket.sk = NULL;
269    ei->socket.file = NULL;
270
271    return &ei->vfs_inode;
272}
273
274static void sock_destroy_inode(struct inode *inode)
275{
276    struct socket_alloc *ei;
277    struct socket_wq *wq;
278
279    ei = container_of(inode, struct socket_alloc, vfs_inode);
280    wq = rcu_dereference_protected(ei->socket.wq, 1);
281    kfree_rcu(wq, rcu);
282    kmem_cache_free(sock_inode_cachep, ei);
283}
284
285static void init_once(void *foo)
286{
287    struct socket_alloc *ei = (struct socket_alloc *)foo;
288
289    inode_init_once(&ei->vfs_inode);
290}
291
292static int init_inodecache(void)
293{
294    sock_inode_cachep = kmem_cache_create("sock_inode_cache",
295                          sizeof(struct socket_alloc),
296                          0,
297                          (SLAB_HWCACHE_ALIGN |
298                           SLAB_RECLAIM_ACCOUNT |
299                           SLAB_MEM_SPREAD),
300                          init_once);
301    if (sock_inode_cachep == NULL)
302        return -ENOMEM;
303    return 0;
304}
305
306static const struct super_operations sockfs_ops = {
307    .alloc_inode = sock_alloc_inode,
308    .destroy_inode = sock_destroy_inode,
309    .statfs = simple_statfs,
310};
311
312/*
313 * sockfs_dname() is called from d_path().
314 */
315static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
316{
317    return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
318                dentry->d_inode->i_ino);
319}
320
321static const struct dentry_operations sockfs_dentry_operations = {
322    .d_dname = sockfs_dname,
323};
324
325static struct dentry *sockfs_mount(struct file_system_type *fs_type,
326             int flags, const char *dev_name, void *data)
327{
328    return mount_pseudo(fs_type, "socket:", &sockfs_ops,
329        &sockfs_dentry_operations, SOCKFS_MAGIC);
330}
331
332static struct vfsmount *sock_mnt __read_mostly;
333
334static struct file_system_type sock_fs_type = {
335    .name = "sockfs",
336    .mount = sockfs_mount,
337    .kill_sb = kill_anon_super,
338};
339
340/*
341 * Obtains the first available file descriptor and sets it up for use.
342 *
343 * These functions create file structures and maps them to fd space
344 * of the current process. On success it returns file descriptor
345 * and file struct implicitly stored in sock->file.
346 * Note that another thread may close file descriptor before we return
347 * from this function. We use the fact that now we do not refer
348 * to socket after mapping. If one day we will need it, this
349 * function will increment ref. count on file by 1.
350 *
351 * In any case returned fd MAY BE not valid!
352 * This race condition is unavoidable
353 * with shared fd spaces, we cannot solve it inside kernel,
354 * but we take care of internal coherence yet.
355 */
356
357struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
358{
359    struct qstr name = { .name = "" };
360    struct path path;
361    struct file *file;
362
363    if (dname) {
364        name.name = dname;
365        name.len = strlen(name.name);
366    } else if (sock->sk) {
367        name.name = sock->sk->sk_prot_creator->name;
368        name.len = strlen(name.name);
369    }
370    path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
371    if (unlikely(!path.dentry))
372        return ERR_PTR(-ENOMEM);
373    path.mnt = mntget(sock_mnt);
374
375    d_instantiate(path.dentry, SOCK_INODE(sock));
376    SOCK_INODE(sock)->i_fop = &socket_file_ops;
377
378    file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
379          &socket_file_ops);
380    if (unlikely(IS_ERR(file))) {
381        /* drop dentry, keep inode */
382        ihold(path.dentry->d_inode);
383        path_put(&path);
384        return file;
385    }
386
387    sock->file = file;
388    file->f_flags = O_RDWR | (flags & O_NONBLOCK);
389    file->private_data = sock;
390    return file;
391}
392EXPORT_SYMBOL(sock_alloc_file);
393
394static int sock_map_fd(struct socket *sock, int flags)
395{
396    struct file *newfile;
397    int fd = get_unused_fd_flags(flags);
398    if (unlikely(fd < 0))
399        return fd;
400
401    newfile = sock_alloc_file(sock, flags, NULL);
402    if (likely(!IS_ERR(newfile))) {
403        fd_install(fd, newfile);
404        return fd;
405    }
406
407    put_unused_fd(fd);
408    return PTR_ERR(newfile);
409}
410
411struct socket *sock_from_file(struct file *file, int *err)
412{
413    if (file->f_op == &socket_file_ops)
414        return file->private_data; /* set in sock_map_fd */
415
416    *err = -ENOTSOCK;
417    return NULL;
418}
419EXPORT_SYMBOL(sock_from_file);
420
421/**
422 * sockfd_lookup - Go from a file number to its socket slot
423 * @fd: file handle
424 * @err: pointer to an error code return
425 *
426 * The file handle passed in is locked and the socket it is bound
427 * too is returned. If an error occurs the err pointer is overwritten
428 * with a negative errno code and NULL is returned. The function checks
429 * for both invalid handles and passing a handle which is not a socket.
430 *
431 * On a success the socket object pointer is returned.
432 */
433
434struct socket *sockfd_lookup(int fd, int *err)
435{
436    struct file *file;
437    struct socket *sock;
438
439    file = fget(fd);
440    if (!file) {
441        *err = -EBADF;
442        return NULL;
443    }
444
445    sock = sock_from_file(file, err);
446    if (!sock)
447        fput(file);
448    return sock;
449}
450EXPORT_SYMBOL(sockfd_lookup);
451
452static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
453{
454    struct fd f = fdget(fd);
455    struct socket *sock;
456
457    *err = -EBADF;
458    if (f.file) {
459        sock = sock_from_file(f.file, err);
460        if (likely(sock)) {
461            *fput_needed = f.flags;
462            return sock;
463        }
464        fdput(f);
465    }
466    return NULL;
467}
468
469#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
470#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
471#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
472static ssize_t sockfs_getxattr(struct dentry *dentry,
473                   const char *name, void *value, size_t size)
474{
475    const char *proto_name;
476    size_t proto_size;
477    int error;
478
479    error = -ENODATA;
480    if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
481        proto_name = dentry->d_name.name;
482        proto_size = strlen(proto_name);
483
484        if (value) {
485            error = -ERANGE;
486            if (proto_size + 1 > size)
487                goto out;
488
489            strncpy(value, proto_name, proto_size + 1);
490        }
491        error = proto_size + 1;
492    }
493
494out:
495    return error;
496}
497
498static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
499                size_t size)
500{
501    ssize_t len;
502    ssize_t used = 0;
503
504    len = security_inode_listsecurity(dentry->d_inode, buffer, size);
505    if (len < 0)
506        return len;
507    used += len;
508    if (buffer) {
509        if (size < used)
510            return -ERANGE;
511        buffer += len;
512    }
513
514    len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
515    used += len;
516    if (buffer) {
517        if (size < used)
518            return -ERANGE;
519        memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
520        buffer += len;
521    }
522
523    return used;
524}
525
526static const struct inode_operations sockfs_inode_ops = {
527    .getxattr = sockfs_getxattr,
528    .listxattr = sockfs_listxattr,
529};
530
531/**
532 * sock_alloc - allocate a socket
533 *
534 * Allocate a new inode and socket object. The two are bound together
535 * and initialised. The socket is then returned. If we are out of inodes
536 * NULL is returned.
537 */
538
539static struct socket *sock_alloc(void)
540{
541    struct inode *inode;
542    struct socket *sock;
543
544    inode = new_inode_pseudo(sock_mnt->mnt_sb);
545    if (!inode)
546        return NULL;
547
548    sock = SOCKET_I(inode);
549
550    kmemcheck_annotate_bitfield(sock, type);
551    inode->i_ino = get_next_ino();
552    inode->i_mode = S_IFSOCK | S_IRWXUGO;
553    inode->i_uid = current_fsuid();
554    inode->i_gid = current_fsgid();
555    inode->i_op = &sockfs_inode_ops;
556
557    this_cpu_add(sockets_in_use, 1);
558    return sock;
559}
560
561/*
562 * In theory you can't get an open on this inode, but /proc provides
563 * a back door. Remember to keep it shut otherwise you'll let the
564 * creepy crawlies in.
565 */
566
567static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
568{
569    return -ENXIO;
570}
571
572const struct file_operations bad_sock_fops = {
573    .owner = THIS_MODULE,
574    .open = sock_no_open,
575    .llseek = noop_llseek,
576};
577
578/**
579 * sock_release - close a socket
580 * @sock: socket to close
581 *
582 * The socket is released from the protocol stack if it has a release
583 * callback, and the inode is then released if the socket is bound to
584 * an inode not a file.
585 */
586
587void sock_release(struct socket *sock)
588{
589    if (sock->ops) {
590        struct module *owner = sock->ops->owner;
591
592        sock->ops->release(sock);
593        sock->ops = NULL;
594        module_put(owner);
595    }
596
597    if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
598        pr_err("%s: fasync list not empty!\n", __func__);
599
600    if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
601        return;
602
603    this_cpu_sub(sockets_in_use, 1);
604    if (!sock->file) {
605        iput(SOCK_INODE(sock));
606        return;
607    }
608    sock->file = NULL;
609}
610EXPORT_SYMBOL(sock_release);
611
612void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
613{
614    *tx_flags = 0;
615    if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
616        *tx_flags |= SKBTX_HW_TSTAMP;
617    if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
618        *tx_flags |= SKBTX_SW_TSTAMP;
619    if (sock_flag(sk, SOCK_WIFI_STATUS))
620        *tx_flags |= SKBTX_WIFI_STATUS;
621}
622EXPORT_SYMBOL(sock_tx_timestamp);
623
624static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
625                       struct msghdr *msg, size_t size)
626{
627    struct sock_iocb *si = kiocb_to_siocb(iocb);
628
629    si->sock = sock;
630    si->scm = NULL;
631    si->msg = msg;
632    si->size = size;
633
634    return sock->ops->sendmsg(iocb, sock, msg, size);
635}
636
637static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
638                 struct msghdr *msg, size_t size)
639{
640    int err = security_socket_sendmsg(sock, msg, size);
641
642    return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
643}
644
645int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
646{
647    struct kiocb iocb;
648    struct sock_iocb siocb;
649    int ret;
650
651    init_sync_kiocb(&iocb, NULL);
652    iocb.private = &siocb;
653    ret = __sock_sendmsg(&iocb, sock, msg, size);
654    if (-EIOCBQUEUED == ret)
655        ret = wait_on_sync_kiocb(&iocb);
656    return ret;
657}
658EXPORT_SYMBOL(sock_sendmsg);
659
660static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
661{
662    struct kiocb iocb;
663    struct sock_iocb siocb;
664    int ret;
665
666    init_sync_kiocb(&iocb, NULL);
667    iocb.private = &siocb;
668    ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
669    if (-EIOCBQUEUED == ret)
670        ret = wait_on_sync_kiocb(&iocb);
671    return ret;
672}
673
674int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
675           struct kvec *vec, size_t num, size_t size)
676{
677    mm_segment_t oldfs = get_fs();
678    int result;
679
680    set_fs(KERNEL_DS);
681    /*
682     * the following is safe, since for compiler definitions of kvec and
683     * iovec are identical, yielding the same in-core layout and alignment
684     */
685    msg->msg_iov = (struct iovec *)vec;
686    msg->msg_iovlen = num;
687    result = sock_sendmsg(sock, msg, size);
688    set_fs(oldfs);
689    return result;
690}
691EXPORT_SYMBOL(kernel_sendmsg);
692
693/*
694 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
695 */
696void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
697    struct sk_buff *skb)
698{
699    int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
700    struct timespec ts[3];
701    int empty = 1;
702    struct skb_shared_hwtstamps *shhwtstamps =
703        skb_hwtstamps(skb);
704
705    /* Race occurred between timestamp enabling and packet
706       receiving. Fill in the current time for now. */
707    if (need_software_tstamp && skb->tstamp.tv64 == 0)
708        __net_timestamp(skb);
709
710    if (need_software_tstamp) {
711        if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
712            struct timeval tv;
713            skb_get_timestamp(skb, &tv);
714            put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
715                 sizeof(tv), &tv);
716        } else {
717            skb_get_timestampns(skb, &ts[0]);
718            put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
719                 sizeof(ts[0]), &ts[0]);
720        }
721    }
722
723
724    memset(ts, 0, sizeof(ts));
725    if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
726        ktime_to_timespec_cond(skb->tstamp, ts + 0))
727        empty = 0;
728    if (shhwtstamps) {
729        if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
730            ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
731            empty = 0;
732        if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
733            ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
734            empty = 0;
735    }
736    if (!empty)
737        put_cmsg(msg, SOL_SOCKET,
738             SCM_TIMESTAMPING, sizeof(ts), &ts);
739}
740EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
741
742void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
743    struct sk_buff *skb)
744{
745    int ack;
746
747    if (!sock_flag(sk, SOCK_WIFI_STATUS))
748        return;
749    if (!skb->wifi_acked_valid)
750        return;
751
752    ack = skb->wifi_acked;
753
754    put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
755}
756EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
757
758static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
759                   struct sk_buff *skb)
760{
761    if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
762        put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
763            sizeof(__u32), &skb->dropcount);
764}
765
766void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
767    struct sk_buff *skb)
768{
769    sock_recv_timestamp(msg, sk, skb);
770    sock_recv_drops(msg, sk, skb);
771}
772EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
773
774static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
775                       struct msghdr *msg, size_t size, int flags)
776{
777    struct sock_iocb *si = kiocb_to_siocb(iocb);
778
779    si->sock = sock;
780    si->scm = NULL;
781    si->msg = msg;
782    si->size = size;
783    si->flags = flags;
784
785    return sock->ops->recvmsg(iocb, sock, msg, size, flags);
786}
787
788static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
789                 struct msghdr *msg, size_t size, int flags)
790{
791    int err = security_socket_recvmsg(sock, msg, size, flags);
792
793    return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
794}
795
796int sock_recvmsg(struct socket *sock, struct msghdr *msg,
797         size_t size, int flags)
798{
799    struct kiocb iocb;
800    struct sock_iocb siocb;
801    int ret;
802
803    init_sync_kiocb(&iocb, NULL);
804    iocb.private = &siocb;
805    ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
806    if (-EIOCBQUEUED == ret)
807        ret = wait_on_sync_kiocb(&iocb);
808    return ret;
809}
810EXPORT_SYMBOL(sock_recvmsg);
811
812static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
813                  size_t size, int flags)
814{
815    struct kiocb iocb;
816    struct sock_iocb siocb;
817    int ret;
818
819    init_sync_kiocb(&iocb, NULL);
820    iocb.private = &siocb;
821    ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
822    if (-EIOCBQUEUED == ret)
823        ret = wait_on_sync_kiocb(&iocb);
824    return ret;
825}
826
827/**
828 * kernel_recvmsg - Receive a message from a socket (kernel space)
829 * @sock: The socket to receive the message from
830 * @msg: Received message
831 * @vec: Input s/g array for message data
832 * @num: Size of input s/g array
833 * @size: Number of bytes to read
834 * @flags: Message flags (MSG_DONTWAIT, etc...)
835 *
836 * On return the msg structure contains the scatter/gather array passed in the
837 * vec argument. The array is modified so that it consists of the unfilled
838 * portion of the original array.
839 *
840 * The returned value is the total number of bytes received, or an error.
841 */
842int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
843           struct kvec *vec, size_t num, size_t size, int flags)
844{
845    mm_segment_t oldfs = get_fs();
846    int result;
847
848    set_fs(KERNEL_DS);
849    /*
850     * the following is safe, since for compiler definitions of kvec and
851     * iovec are identical, yielding the same in-core layout and alignment
852     */
853    msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
854    result = sock_recvmsg(sock, msg, size, flags);
855    set_fs(oldfs);
856    return result;
857}
858EXPORT_SYMBOL(kernel_recvmsg);
859
860static ssize_t sock_sendpage(struct file *file, struct page *page,
861                 int offset, size_t size, loff_t *ppos, int more)
862{
863    struct socket *sock;
864    int flags;
865
866    sock = file->private_data;
867
868    flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
869    /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
870    flags |= more;
871
872    return kernel_sendpage(sock, page, offset, size, flags);
873}
874
875static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
876                struct pipe_inode_info *pipe, size_t len,
877                unsigned int flags)
878{
879    struct socket *sock = file->private_data;
880
881    if (unlikely(!sock->ops->splice_read))
882        return -EINVAL;
883
884    return sock->ops->splice_read(sock, ppos, pipe, len, flags);
885}
886
887static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
888                     struct sock_iocb *siocb)
889{
890    if (!is_sync_kiocb(iocb))
891        BUG();
892
893    siocb->kiocb = iocb;
894    iocb->private = siocb;
895    return siocb;
896}
897
898static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
899        struct file *file, const struct iovec *iov,
900        unsigned long nr_segs)
901{
902    struct socket *sock = file->private_data;
903    size_t size = 0;
904    int i;
905
906    for (i = 0; i < nr_segs; i++)
907        size += iov[i].iov_len;
908
909    msg->msg_name = NULL;
910    msg->msg_namelen = 0;
911    msg->msg_control = NULL;
912    msg->msg_controllen = 0;
913    msg->msg_iov = (struct iovec *)iov;
914    msg->msg_iovlen = nr_segs;
915    msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
916
917    return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
918}
919
920static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
921                unsigned long nr_segs, loff_t pos)
922{
923    struct sock_iocb siocb, *x;
924
925    if (pos != 0)
926        return -ESPIPE;
927
928    if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
929        return 0;
930
931
932    x = alloc_sock_iocb(iocb, &siocb);
933    if (!x)
934        return -ENOMEM;
935    return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
936}
937
938static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
939            struct file *file, const struct iovec *iov,
940            unsigned long nr_segs)
941{
942    struct socket *sock = file->private_data;
943    size_t size = 0;
944    int i;
945
946    for (i = 0; i < nr_segs; i++)
947        size += iov[i].iov_len;
948
949    msg->msg_name = NULL;
950    msg->msg_namelen = 0;
951    msg->msg_control = NULL;
952    msg->msg_controllen = 0;
953    msg->msg_iov = (struct iovec *)iov;
954    msg->msg_iovlen = nr_segs;
955    msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
956    if (sock->type == SOCK_SEQPACKET)
957        msg->msg_flags |= MSG_EOR;
958
959    return __sock_sendmsg(iocb, sock, msg, size);
960}
961
962static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
963              unsigned long nr_segs, loff_t pos)
964{
965    struct sock_iocb siocb, *x;
966
967    if (pos != 0)
968        return -ESPIPE;
969
970    x = alloc_sock_iocb(iocb, &siocb);
971    if (!x)
972        return -ENOMEM;
973
974    return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
975}
976
977/*
978 * Atomic setting of ioctl hooks to avoid race
979 * with module unload.
980 */
981
982static DEFINE_MUTEX(br_ioctl_mutex);
983static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
984
985void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
986{
987    mutex_lock(&br_ioctl_mutex);
988    br_ioctl_hook = hook;
989    mutex_unlock(&br_ioctl_mutex);
990}
991EXPORT_SYMBOL(brioctl_set);
992
993static DEFINE_MUTEX(vlan_ioctl_mutex);
994static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
995
996void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
997{
998    mutex_lock(&vlan_ioctl_mutex);
999    vlan_ioctl_hook = hook;
1000    mutex_unlock(&vlan_ioctl_mutex);
1001}
1002EXPORT_SYMBOL(vlan_ioctl_set);
1003
1004static DEFINE_MUTEX(dlci_ioctl_mutex);
1005static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1006
1007void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1008{
1009    mutex_lock(&dlci_ioctl_mutex);
1010    dlci_ioctl_hook = hook;
1011    mutex_unlock(&dlci_ioctl_mutex);
1012}
1013EXPORT_SYMBOL(dlci_ioctl_set);
1014
1015static long sock_do_ioctl(struct net *net, struct socket *sock,
1016                 unsigned int cmd, unsigned long arg)
1017{
1018    int err;
1019    void __user *argp = (void __user *)arg;
1020
1021    err = sock->ops->ioctl(sock, cmd, arg);
1022
1023    /*
1024     * If this ioctl is unknown try to hand it down
1025     * to the NIC driver.
1026     */
1027    if (err == -ENOIOCTLCMD)
1028        err = dev_ioctl(net, cmd, argp);
1029
1030    return err;
1031}
1032
1033/*
1034 * With an ioctl, arg may well be a user mode pointer, but we don't know
1035 * what to do with it - that's up to the protocol still.
1036 */
1037
1038static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1039{
1040    struct socket *sock;
1041    struct sock *sk;
1042    void __user *argp = (void __user *)arg;
1043    int pid, err;
1044    struct net *net;
1045
1046    sock = file->private_data;
1047    sk = sock->sk;
1048    net = sock_net(sk);
1049    if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1050        err = dev_ioctl(net, cmd, argp);
1051    } else
1052#ifdef CONFIG_WEXT_CORE
1053    if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1054        err = dev_ioctl(net, cmd, argp);
1055    } else
1056#endif
1057        switch (cmd) {
1058        case FIOSETOWN:
1059        case SIOCSPGRP:
1060            err = -EFAULT;
1061            if (get_user(pid, (int __user *)argp))
1062                break;
1063            err = f_setown(sock->file, pid, 1);
1064            break;
1065        case FIOGETOWN:
1066        case SIOCGPGRP:
1067            err = put_user(f_getown(sock->file),
1068                       (int __user *)argp);
1069            break;
1070        case SIOCGIFBR:
1071        case SIOCSIFBR:
1072        case SIOCBRADDBR:
1073        case SIOCBRDELBR:
1074            err = -ENOPKG;
1075            if (!br_ioctl_hook)
1076                request_module("bridge");
1077
1078            mutex_lock(&br_ioctl_mutex);
1079            if (br_ioctl_hook)
1080                err = br_ioctl_hook(net, cmd, argp);
1081            mutex_unlock(&br_ioctl_mutex);
1082            break;
1083        case SIOCGIFVLAN:
1084        case SIOCSIFVLAN:
1085            err = -ENOPKG;
1086            if (!vlan_ioctl_hook)
1087                request_module("8021q");
1088
1089            mutex_lock(&vlan_ioctl_mutex);
1090            if (vlan_ioctl_hook)
1091                err = vlan_ioctl_hook(net, argp);
1092            mutex_unlock(&vlan_ioctl_mutex);
1093            break;
1094        case SIOCADDDLCI:
1095        case SIOCDELDLCI:
1096            err = -ENOPKG;
1097            if (!dlci_ioctl_hook)
1098                request_module("dlci");
1099
1100            mutex_lock(&dlci_ioctl_mutex);
1101            if (dlci_ioctl_hook)
1102                err = dlci_ioctl_hook(cmd, argp);
1103            mutex_unlock(&dlci_ioctl_mutex);
1104            break;
1105        default:
1106            err = sock_do_ioctl(net, sock, cmd, arg);
1107            break;
1108        }
1109    return err;
1110}
1111
1112int sock_create_lite(int family, int type, int protocol, struct socket **res)
1113{
1114    int err;
1115    struct socket *sock = NULL;
1116
1117    err = security_socket_create(family, type, protocol, 1);
1118    if (err)
1119        goto out;
1120
1121    sock = sock_alloc();
1122    if (!sock) {
1123        err = -ENOMEM;
1124        goto out;
1125    }
1126
1127    sock->type = type;
1128    err = security_socket_post_create(sock, family, type, protocol, 1);
1129    if (err)
1130        goto out_release;
1131
1132out:
1133    *res = sock;
1134    return err;
1135out_release:
1136    sock_release(sock);
1137    sock = NULL;
1138    goto out;
1139}
1140EXPORT_SYMBOL(sock_create_lite);
1141
1142/* No kernel lock held - perfect */
1143static unsigned int sock_poll(struct file *file, poll_table *wait)
1144{
1145    unsigned int busy_flag = 0;
1146    struct socket *sock;
1147
1148    /*
1149     * We can't return errors to poll, so it's either yes or no.
1150     */
1151    sock = file->private_data;
1152
1153    if (sk_can_busy_loop(sock->sk)) {
1154        /* this socket can poll_ll so tell the system call */
1155        busy_flag = POLL_BUSY_LOOP;
1156
1157        /* once, only if requested by syscall */
1158        if (wait && (wait->_key & POLL_BUSY_LOOP))
1159            sk_busy_loop(sock->sk, 1);
1160    }
1161
1162    return busy_flag | sock->ops->poll(file, sock, wait);
1163}
1164
1165static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1166{
1167    struct socket *sock = file->private_data;
1168
1169    return sock->ops->mmap(file, sock, vma);
1170}
1171
1172static int sock_close(struct inode *inode, struct file *filp)
1173{
1174    sock_release(SOCKET_I(inode));
1175    return 0;
1176}
1177
1178/*
1179 * Update the socket async list
1180 *
1181 * Fasync_list locking strategy.
1182 *
1183 * 1. fasync_list is modified only under process context socket lock
1184 * i.e. under semaphore.
1185 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1186 * or under socket lock
1187 */
1188
1189static int sock_fasync(int fd, struct file *filp, int on)
1190{
1191    struct socket *sock = filp->private_data;
1192    struct sock *sk = sock->sk;
1193    struct socket_wq *wq;
1194
1195    if (sk == NULL)
1196        return -EINVAL;
1197
1198    lock_sock(sk);
1199    wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1200    fasync_helper(fd, filp, on, &wq->fasync_list);
1201
1202    if (!wq->fasync_list)
1203        sock_reset_flag(sk, SOCK_FASYNC);
1204    else
1205        sock_set_flag(sk, SOCK_FASYNC);
1206
1207    release_sock(sk);
1208    return 0;
1209}
1210
1211/* This function may be called only under socket lock or callback_lock or rcu_lock */
1212
1213int sock_wake_async(struct socket *sock, int how, int band)
1214{
1215    struct socket_wq *wq;
1216
1217    if (!sock)
1218        return -1;
1219    rcu_read_lock();
1220    wq = rcu_dereference(sock->wq);
1221    if (!wq || !wq->fasync_list) {
1222        rcu_read_unlock();
1223        return -1;
1224    }
1225    switch (how) {
1226    case SOCK_WAKE_WAITD:
1227        if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1228            break;
1229        goto call_kill;
1230    case SOCK_WAKE_SPACE:
1231        if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1232            break;
1233        /* fall through */
1234    case SOCK_WAKE_IO:
1235call_kill:
1236        kill_fasync(&wq->fasync_list, SIGIO, band);
1237        break;
1238    case SOCK_WAKE_URG:
1239        kill_fasync(&wq->fasync_list, SIGURG, band);
1240    }
1241    rcu_read_unlock();
1242    return 0;
1243}
1244EXPORT_SYMBOL(sock_wake_async);
1245
1246int __sock_create(struct net *net, int family, int type, int protocol,
1247             struct socket **res, int kern)
1248{
1249    int err;
1250    struct socket *sock;
1251    const struct net_proto_family *pf;
1252
1253    /*
1254     * Check protocol is in range
1255     */
1256    if (family < 0 || family >= NPROTO)
1257        return -EAFNOSUPPORT;
1258    if (type < 0 || type >= SOCK_MAX)
1259        return -EINVAL;
1260
1261    /* Compatibility.
1262
1263       This uglymoron is moved from INET layer to here to avoid
1264       deadlock in module load.
1265     */
1266    if (family == PF_INET && type == SOCK_PACKET) {
1267        static int warned;
1268        if (!warned) {
1269            warned = 1;
1270            pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1271                current->comm);
1272        }
1273        family = PF_PACKET;
1274    }
1275
1276    err = security_socket_create(family, type, protocol, kern);
1277    if (err)
1278        return err;
1279
1280    /*
1281     * Allocate the socket and allow the family to set things up. if
1282     * the protocol is 0, the family is instructed to select an appropriate
1283     * default.
1284     */
1285    sock = sock_alloc();
1286    if (!sock) {
1287        net_warn_ratelimited("socket: no more sockets\n");
1288        return -ENFILE; /* Not exactly a match, but its the
1289                   closest posix thing */
1290    }
1291
1292    sock->type = type;
1293
1294#ifdef CONFIG_MODULES
1295    /* Attempt to load a protocol module if the find failed.
1296     *
1297     * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1298     * requested real, full-featured networking support upon configuration.
1299     * Otherwise module support will break!
1300     */
1301    if (rcu_access_pointer(net_families[family]) == NULL)
1302        request_module("net-pf-%d", family);
1303#endif
1304
1305    rcu_read_lock();
1306    pf = rcu_dereference(net_families[family]);
1307    err = -EAFNOSUPPORT;
1308    if (!pf)
1309        goto out_release;
1310
1311    /*
1312     * We will call the ->create function, that possibly is in a loadable
1313     * module, so we have to bump that loadable module refcnt first.
1314     */
1315    if (!try_module_get(pf->owner))
1316        goto out_release;
1317
1318    /* Now protected by module ref count */
1319    rcu_read_unlock();
1320
1321    err = pf->create(net, sock, protocol, kern);
1322    if (err < 0)
1323        goto out_module_put;
1324
1325    /*
1326     * Now to bump the refcnt of the [loadable] module that owns this
1327     * socket at sock_release time we decrement its refcnt.
1328     */
1329    if (!try_module_get(sock->ops->owner))
1330        goto out_module_busy;
1331
1332    /*
1333     * Now that we're done with the ->create function, the [loadable]
1334     * module can have its refcnt decremented
1335     */
1336    module_put(pf->owner);
1337    err = security_socket_post_create(sock, family, type, protocol, kern);
1338    if (err)
1339        goto out_sock_release;
1340    *res = sock;
1341
1342    return 0;
1343
1344out_module_busy:
1345    err = -EAFNOSUPPORT;
1346out_module_put:
1347    sock->ops = NULL;
1348    module_put(pf->owner);
1349out_sock_release:
1350    sock_release(sock);
1351    return err;
1352
1353out_release:
1354    rcu_read_unlock();
1355    goto out_sock_release;
1356}
1357EXPORT_SYMBOL(__sock_create);
1358
1359int sock_create(int family, int type, int protocol, struct socket **res)
1360{
1361    return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1362}
1363EXPORT_SYMBOL(sock_create);
1364
1365int sock_create_kern(int family, int type, int protocol, struct socket **res)
1366{
1367    return __sock_create(&init_net, family, type, protocol, res, 1);
1368}
1369EXPORT_SYMBOL(sock_create_kern);
1370
1371SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1372{
1373    int retval;
1374    struct socket *sock;
1375    int flags;
1376
1377    /* Check the SOCK_* constants for consistency. */
1378    BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1379    BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1380    BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1381    BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1382
1383    flags = type & ~SOCK_TYPE_MASK;
1384    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1385        return -EINVAL;
1386    type &= SOCK_TYPE_MASK;
1387
1388    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1389        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1390
1391    retval = sock_create(family, type, protocol, &sock);
1392    if (retval < 0)
1393        goto out;
1394
1395    retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1396    if (retval < 0)
1397        goto out_release;
1398
1399out:
1400    /* It may be already another descriptor 8) Not kernel problem. */
1401    return retval;
1402
1403out_release:
1404    sock_release(sock);
1405    return retval;
1406}
1407
1408/*
1409 * Create a pair of connected sockets.
1410 */
1411
1412SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1413        int __user *, usockvec)
1414{
1415    struct socket *sock1, *sock2;
1416    int fd1, fd2, err;
1417    struct file *newfile1, *newfile2;
1418    int flags;
1419
1420    flags = type & ~SOCK_TYPE_MASK;
1421    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1422        return -EINVAL;
1423    type &= SOCK_TYPE_MASK;
1424
1425    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1426        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1427
1428    /*
1429     * Obtain the first socket and check if the underlying protocol
1430     * supports the socketpair call.
1431     */
1432
1433    err = sock_create(family, type, protocol, &sock1);
1434    if (err < 0)
1435        goto out;
1436
1437    err = sock_create(family, type, protocol, &sock2);
1438    if (err < 0)
1439        goto out_release_1;
1440
1441    err = sock1->ops->socketpair(sock1, sock2);
1442    if (err < 0)
1443        goto out_release_both;
1444
1445    fd1 = get_unused_fd_flags(flags);
1446    if (unlikely(fd1 < 0)) {
1447        err = fd1;
1448        goto out_release_both;
1449    }
1450
1451    fd2 = get_unused_fd_flags(flags);
1452    if (unlikely(fd2 < 0)) {
1453        err = fd2;
1454        goto out_put_unused_1;
1455    }
1456
1457    newfile1 = sock_alloc_file(sock1, flags, NULL);
1458    if (unlikely(IS_ERR(newfile1))) {
1459        err = PTR_ERR(newfile1);
1460        goto out_put_unused_both;
1461    }
1462
1463    newfile2 = sock_alloc_file(sock2, flags, NULL);
1464    if (IS_ERR(newfile2)) {
1465        err = PTR_ERR(newfile2);
1466        goto out_fput_1;
1467    }
1468
1469    err = put_user(fd1, &usockvec[0]);
1470    if (err)
1471        goto out_fput_both;
1472
1473    err = put_user(fd2, &usockvec[1]);
1474    if (err)
1475        goto out_fput_both;
1476
1477    audit_fd_pair(fd1, fd2);
1478
1479    fd_install(fd1, newfile1);
1480    fd_install(fd2, newfile2);
1481    /* fd1 and fd2 may be already another descriptors.
1482     * Not kernel problem.
1483     */
1484
1485    return 0;
1486
1487out_fput_both:
1488    fput(newfile2);
1489    fput(newfile1);
1490    put_unused_fd(fd2);
1491    put_unused_fd(fd1);
1492    goto out;
1493
1494out_fput_1:
1495    fput(newfile1);
1496    put_unused_fd(fd2);
1497    put_unused_fd(fd1);
1498    sock_release(sock2);
1499    goto out;
1500
1501out_put_unused_both:
1502    put_unused_fd(fd2);
1503out_put_unused_1:
1504    put_unused_fd(fd1);
1505out_release_both:
1506    sock_release(sock2);
1507out_release_1:
1508    sock_release(sock1);
1509out:
1510    return err;
1511}
1512
1513/*
1514 * Bind a name to a socket. Nothing much to do here since it's
1515 * the protocol's responsibility to handle the local address.
1516 *
1517 * We move the socket address to kernel space before we call
1518 * the protocol layer (having also checked the address is ok).
1519 */
1520
1521SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1522{
1523    struct socket *sock;
1524    struct sockaddr_storage address;
1525    int err, fput_needed;
1526
1527    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1528    if (sock) {
1529        err = move_addr_to_kernel(umyaddr, addrlen, &address);
1530        if (err >= 0) {
1531            err = security_socket_bind(sock,
1532                           (struct sockaddr *)&address,
1533                           addrlen);
1534            if (!err)
1535                err = sock->ops->bind(sock,
1536                              (struct sockaddr *)
1537                              &address, addrlen);
1538        }
1539        fput_light(sock->file, fput_needed);
1540    }
1541    return err;
1542}
1543
1544/*
1545 * Perform a listen. Basically, we allow the protocol to do anything
1546 * necessary for a listen, and if that works, we mark the socket as
1547 * ready for listening.
1548 */
1549
1550SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1551{
1552    struct socket *sock;
1553    int err, fput_needed;
1554    int somaxconn;
1555
1556    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1557    if (sock) {
1558        somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1559        if ((unsigned int)backlog > somaxconn)
1560            backlog = somaxconn;
1561
1562        err = security_socket_listen(sock, backlog);
1563        if (!err)
1564            err = sock->ops->listen(sock, backlog);
1565
1566        fput_light(sock->file, fput_needed);
1567    }
1568    return err;
1569}
1570
1571/*
1572 * For accept, we attempt to create a new socket, set up the link
1573 * with the client, wake up the client, then return the new
1574 * connected fd. We collect the address of the connector in kernel
1575 * space and move it to user at the very end. This is unclean because
1576 * we open the socket then return an error.
1577 *
1578 * 1003.1g adds the ability to recvmsg() to query connection pending
1579 * status to recvmsg. We need to add that support in a way thats
1580 * clean when we restucture accept also.
1581 */
1582
1583SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1584        int __user *, upeer_addrlen, int, flags)
1585{
1586    struct socket *sock, *newsock;
1587    struct file *newfile;
1588    int err, len, newfd, fput_needed;
1589    struct sockaddr_storage address;
1590
1591    if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1592        return -EINVAL;
1593
1594    if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1595        flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1596
1597    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1598    if (!sock)
1599        goto out;
1600
1601    err = -ENFILE;
1602    newsock = sock_alloc();
1603    if (!newsock)
1604        goto out_put;
1605
1606    newsock->type = sock->type;
1607    newsock->ops = sock->ops;
1608
1609    /*
1610     * We don't need try_module_get here, as the listening socket (sock)
1611     * has the protocol module (sock->ops->owner) held.
1612     */
1613    __module_get(newsock->ops->owner);
1614
1615    newfd = get_unused_fd_flags(flags);
1616    if (unlikely(newfd < 0)) {
1617        err = newfd;
1618        sock_release(newsock);
1619        goto out_put;
1620    }
1621    newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1622    if (unlikely(IS_ERR(newfile))) {
1623        err = PTR_ERR(newfile);
1624        put_unused_fd(newfd);
1625        sock_release(newsock);
1626        goto out_put;
1627    }
1628
1629    err = security_socket_accept(sock, newsock);
1630    if (err)
1631        goto out_fd;
1632
1633    err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1634    if (err < 0)
1635        goto out_fd;
1636
1637    if (upeer_sockaddr) {
1638        if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1639                      &len, 2) < 0) {
1640            err = -ECONNABORTED;
1641            goto out_fd;
1642        }
1643        err = move_addr_to_user(&address,
1644                    len, upeer_sockaddr, upeer_addrlen);
1645        if (err < 0)
1646            goto out_fd;
1647    }
1648
1649    /* File flags are not inherited via accept() unlike another OSes. */
1650
1651    fd_install(newfd, newfile);
1652    err = newfd;
1653
1654out_put:
1655    fput_light(sock->file, fput_needed);
1656out:
1657    return err;
1658out_fd:
1659    fput(newfile);
1660    put_unused_fd(newfd);
1661    goto out_put;
1662}
1663
1664SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1665        int __user *, upeer_addrlen)
1666{
1667    return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1668}
1669
1670/*
1671 * Attempt to connect to a socket with the server address. The address
1672 * is in user space so we verify it is OK and move it to kernel space.
1673 *
1674 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1675 * break bindings
1676 *
1677 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1678 * other SEQPACKET protocols that take time to connect() as it doesn't
1679 * include the -EINPROGRESS status for such sockets.
1680 */
1681
1682SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1683        int, addrlen)
1684{
1685    struct socket *sock;
1686    struct sockaddr_storage address;
1687    int err, fput_needed;
1688
1689    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1690    if (!sock)
1691        goto out;
1692    err = move_addr_to_kernel(uservaddr, addrlen, &address);
1693    if (err < 0)
1694        goto out_put;
1695
1696    err =
1697        security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1698    if (err)
1699        goto out_put;
1700
1701    err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1702                 sock->file->f_flags);
1703out_put:
1704    fput_light(sock->file, fput_needed);
1705out:
1706    return err;
1707}
1708
1709/*
1710 * Get the local address ('name') of a socket object. Move the obtained
1711 * name to user space.
1712 */
1713
1714SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1715        int __user *, usockaddr_len)
1716{
1717    struct socket *sock;
1718    struct sockaddr_storage address;
1719    int len, err, fput_needed;
1720
1721    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1722    if (!sock)
1723        goto out;
1724
1725    err = security_socket_getsockname(sock);
1726    if (err)
1727        goto out_put;
1728
1729    err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1730    if (err)
1731        goto out_put;
1732    err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1733
1734out_put:
1735    fput_light(sock->file, fput_needed);
1736out:
1737    return err;
1738}
1739
1740/*
1741 * Get the remote address ('name') of a socket object. Move the obtained
1742 * name to user space.
1743 */
1744
1745SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1746        int __user *, usockaddr_len)
1747{
1748    struct socket *sock;
1749    struct sockaddr_storage address;
1750    int len, err, fput_needed;
1751
1752    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1753    if (sock != NULL) {
1754        err = security_socket_getpeername(sock);
1755        if (err) {
1756            fput_light(sock->file, fput_needed);
1757            return err;
1758        }
1759
1760        err =
1761            sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1762                       1);
1763        if (!err)
1764            err = move_addr_to_user(&address, len, usockaddr,
1765                        usockaddr_len);
1766        fput_light(sock->file, fput_needed);
1767    }
1768    return err;
1769}
1770
1771/*
1772 * Send a datagram to a given address. We move the address into kernel
1773 * space and check the user space data area is readable before invoking
1774 * the protocol.
1775 */
1776
1777SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1778        unsigned int, flags, struct sockaddr __user *, addr,
1779        int, addr_len)
1780{
1781    struct socket *sock;
1782    struct sockaddr_storage address;
1783    int err;
1784    struct msghdr msg;
1785    struct iovec iov;
1786    int fput_needed;
1787
1788    if (len > INT_MAX)
1789        len = INT_MAX;
1790    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1791    if (!sock)
1792        goto out;
1793
1794    iov.iov_base = buff;
1795    iov.iov_len = len;
1796    msg.msg_name = NULL;
1797    msg.msg_iov = &iov;
1798    msg.msg_iovlen = 1;
1799    msg.msg_control = NULL;
1800    msg.msg_controllen = 0;
1801    msg.msg_namelen = 0;
1802    if (addr) {
1803        err = move_addr_to_kernel(addr, addr_len, &address);
1804        if (err < 0)
1805            goto out_put;
1806        msg.msg_name = (struct sockaddr *)&address;
1807        msg.msg_namelen = addr_len;
1808    }
1809    if (sock->file->f_flags & O_NONBLOCK)
1810        flags |= MSG_DONTWAIT;
1811    msg.msg_flags = flags;
1812    err = sock_sendmsg(sock, &msg, len);
1813
1814out_put:
1815    fput_light(sock->file, fput_needed);
1816out:
1817    return err;
1818}
1819
1820/*
1821 * Send a datagram down a socket.
1822 */
1823
1824SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1825        unsigned int, flags)
1826{
1827    return sys_sendto(fd, buff, len, flags, NULL, 0);
1828}
1829
1830/*
1831 * Receive a frame from the socket and optionally record the address of the
1832 * sender. We verify the buffers are writable and if needed move the
1833 * sender address from kernel to user space.
1834 */
1835
1836SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1837        unsigned int, flags, struct sockaddr __user *, addr,
1838        int __user *, addr_len)
1839{
1840    struct socket *sock;
1841    struct iovec iov;
1842    struct msghdr msg;
1843    struct sockaddr_storage address;
1844    int err, err2;
1845    int fput_needed;
1846
1847    if (size > INT_MAX)
1848        size = INT_MAX;
1849    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1850    if (!sock)
1851        goto out;
1852
1853    msg.msg_control = NULL;
1854    msg.msg_controllen = 0;
1855    msg.msg_iovlen = 1;
1856    msg.msg_iov = &iov;
1857    iov.iov_len = size;
1858    iov.iov_base = ubuf;
1859    /* Save some cycles and don't copy the address if not needed */
1860    msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1861    /* We assume all kernel code knows the size of sockaddr_storage */
1862    msg.msg_namelen = 0;
1863    if (sock->file->f_flags & O_NONBLOCK)
1864        flags |= MSG_DONTWAIT;
1865    err = sock_recvmsg(sock, &msg, size, flags);
1866
1867    if (err >= 0 && addr != NULL) {
1868        err2 = move_addr_to_user(&address,
1869                     msg.msg_namelen, addr, addr_len);
1870        if (err2 < 0)
1871            err = err2;
1872    }
1873
1874    fput_light(sock->file, fput_needed);
1875out:
1876    return err;
1877}
1878
1879/*
1880 * Receive a datagram from a socket.
1881 */
1882
1883SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1884        unsigned int, flags)
1885{
1886    return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1887}
1888
1889/*
1890 * Set a socket option. Because we don't know the option lengths we have
1891 * to pass the user mode parameter for the protocols to sort out.
1892 */
1893
1894SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1895        char __user *, optval, int, optlen)
1896{
1897    int err, fput_needed;
1898    struct socket *sock;
1899
1900    if (optlen < 0)
1901        return -EINVAL;
1902
1903    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1904    if (sock != NULL) {
1905        err = security_socket_setsockopt(sock, level, optname);
1906        if (err)
1907            goto out_put;
1908
1909        if (level == SOL_SOCKET)
1910            err =
1911                sock_setsockopt(sock, level, optname, optval,
1912                        optlen);
1913        else
1914            err =
1915                sock->ops->setsockopt(sock, level, optname, optval,
1916                          optlen);
1917out_put:
1918        fput_light(sock->file, fput_needed);
1919    }
1920    return err;
1921}
1922
1923/*
1924 * Get a socket option. Because we don't know the option lengths we have
1925 * to pass a user mode parameter for the protocols to sort out.
1926 */
1927
1928SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1929        char __user *, optval, int __user *, optlen)
1930{
1931    int err, fput_needed;
1932    struct socket *sock;
1933
1934    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1935    if (sock != NULL) {
1936        err = security_socket_getsockopt(sock, level, optname);
1937        if (err)
1938            goto out_put;
1939
1940        if (level == SOL_SOCKET)
1941            err =
1942                sock_getsockopt(sock, level, optname, optval,
1943                        optlen);
1944        else
1945            err =
1946                sock->ops->getsockopt(sock, level, optname, optval,
1947                          optlen);
1948out_put:
1949        fput_light(sock->file, fput_needed);
1950    }
1951    return err;
1952}
1953
1954/*
1955 * Shutdown a socket.
1956 */
1957
1958SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1959{
1960    int err, fput_needed;
1961    struct socket *sock;
1962
1963    sock = sockfd_lookup_light(fd, &err, &fput_needed);
1964    if (sock != NULL) {
1965        err = security_socket_shutdown(sock, how);
1966        if (!err)
1967            err = sock->ops->shutdown(sock, how);
1968        fput_light(sock->file, fput_needed);
1969    }
1970    return err;
1971}
1972
1973/* A couple of helpful macros for getting the address of the 32/64 bit
1974 * fields which are the same type (int / unsigned) on our platforms.
1975 */
1976#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1977#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1978#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1979
1980struct used_address {
1981    struct sockaddr_storage name;
1982    unsigned int name_len;
1983};
1984
1985static int copy_msghdr_from_user(struct msghdr *kmsg,
1986                 struct msghdr __user *umsg)
1987{
1988    if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1989        return -EFAULT;
1990
1991    if (kmsg->msg_namelen < 0)
1992        return -EINVAL;
1993
1994    if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1995        kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1996    return 0;
1997}
1998
1999static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
2000             struct msghdr *msg_sys, unsigned int flags,
2001             struct used_address *used_address)
2002{
2003    struct compat_msghdr __user *msg_compat =
2004        (struct compat_msghdr __user *)msg;
2005    struct sockaddr_storage address;
2006    struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2007    unsigned char ctl[sizeof(struct cmsghdr) + 20]
2008        __attribute__ ((aligned(sizeof(__kernel_size_t))));
2009    /* 20 is size of ipv6_pktinfo */
2010    unsigned char *ctl_buf = ctl;
2011    int err, ctl_len, total_len;
2012
2013    err = -EFAULT;
2014    if (MSG_CMSG_COMPAT & flags) {
2015        if (get_compat_msghdr(msg_sys, msg_compat))
2016            return -EFAULT;
2017    } else {
2018        err = copy_msghdr_from_user(msg_sys, msg);
2019        if (err)
2020            return err;
2021    }
2022
2023    if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2024        err = -EMSGSIZE;
2025        if (msg_sys->msg_iovlen > UIO_MAXIOV)
2026            goto out;
2027        err = -ENOMEM;
2028        iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2029                  GFP_KERNEL);
2030        if (!iov)
2031            goto out;
2032    }
2033
2034    /* This will also move the address data into kernel space */
2035    if (MSG_CMSG_COMPAT & flags) {
2036        err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2037    } else
2038        err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2039    if (err < 0)
2040        goto out_freeiov;
2041    total_len = err;
2042
2043    err = -ENOBUFS;
2044
2045    if (msg_sys->msg_controllen > INT_MAX)
2046        goto out_freeiov;
2047    ctl_len = msg_sys->msg_controllen;
2048    if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2049        err =
2050            cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2051                             sizeof(ctl));
2052        if (err)
2053            goto out_freeiov;
2054        ctl_buf = msg_sys->msg_control;
2055        ctl_len = msg_sys->msg_controllen;
2056    } else if (ctl_len) {
2057        if (ctl_len > sizeof(ctl)) {
2058            ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2059            if (ctl_buf == NULL)
2060                goto out_freeiov;
2061        }
2062        err = -EFAULT;
2063        /*
2064         * Careful! Before this, msg_sys->msg_control contains a user pointer.
2065         * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2066         * checking falls down on this.
2067         */
2068        if (copy_from_user(ctl_buf,
2069                   (void __user __force *)msg_sys->msg_control,
2070                   ctl_len))
2071            goto out_freectl;
2072        msg_sys->msg_control = ctl_buf;
2073    }
2074    msg_sys->msg_flags = flags;
2075
2076    if (sock->file->f_flags & O_NONBLOCK)
2077        msg_sys->msg_flags |= MSG_DONTWAIT;
2078    /*
2079     * If this is sendmmsg() and current destination address is same as
2080     * previously succeeded address, omit asking LSM's decision.
2081     * used_address->name_len is initialized to UINT_MAX so that the first
2082     * destination address never matches.
2083     */
2084    if (used_address && msg_sys->msg_name &&
2085        used_address->name_len == msg_sys->msg_namelen &&
2086        !memcmp(&used_address->name, msg_sys->msg_name,
2087            used_address->name_len)) {
2088        err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2089        goto out_freectl;
2090    }
2091    err = sock_sendmsg(sock, msg_sys, total_len);
2092    /*
2093     * If this is sendmmsg() and sending to current destination address was
2094     * successful, remember it.
2095     */
2096    if (used_address && err >= 0) {
2097        used_address->name_len = msg_sys->msg_namelen;
2098        if (msg_sys->msg_name)
2099            memcpy(&used_address->name, msg_sys->msg_name,
2100                   used_address->name_len);
2101    }
2102
2103out_freectl:
2104    if (ctl_buf != ctl)
2105        sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2106out_freeiov:
2107    if (iov != iovstack)
2108        kfree(iov);
2109out:
2110    return err;
2111}
2112
2113/*
2114 * BSD sendmsg interface
2115 */
2116
2117long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2118{
2119    int fput_needed, err;
2120    struct msghdr msg_sys;
2121    struct socket *sock;
2122
2123    sock = sockfd_lookup_light(fd, &err, &fput_needed);
2124    if (!sock)
2125        goto out;
2126
2127    err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2128
2129    fput_light(sock->file, fput_needed);
2130out:
2131    return err;
2132}
2133
2134SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2135{
2136    if (flags & MSG_CMSG_COMPAT)
2137        return -EINVAL;
2138    return __sys_sendmsg(fd, msg, flags);
2139}
2140
2141/*
2142 * Linux sendmmsg interface
2143 */
2144
2145int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2146           unsigned int flags)
2147{
2148    int fput_needed, err, datagrams;
2149    struct socket *sock;
2150    struct mmsghdr __user *entry;
2151    struct compat_mmsghdr __user *compat_entry;
2152    struct msghdr msg_sys;
2153    struct used_address used_address;
2154
2155    if (vlen > UIO_MAXIOV)
2156        vlen = UIO_MAXIOV;
2157
2158    datagrams = 0;
2159
2160    sock = sockfd_lookup_light(fd, &err, &fput_needed);
2161    if (!sock)
2162        return err;
2163
2164    used_address.name_len = UINT_MAX;
2165    entry = mmsg;
2166    compat_entry = (struct compat_mmsghdr __user *)mmsg;
2167    err = 0;
2168
2169    while (datagrams < vlen) {
2170        if (MSG_CMSG_COMPAT & flags) {
2171            err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2172                         &msg_sys, flags, &used_address);
2173            if (err < 0)
2174                break;
2175            err = __put_user(err, &compat_entry->msg_len);
2176            ++compat_entry;
2177        } else {
2178            err = ___sys_sendmsg(sock,
2179                         (struct msghdr __user *)entry,
2180                         &msg_sys, flags, &used_address);
2181            if (err < 0)
2182                break;
2183            err = put_user(err, &entry->msg_len);
2184            ++entry;
2185        }
2186
2187        if (err)
2188            break;
2189        ++datagrams;
2190    }
2191
2192    fput_light(sock->file, fput_needed);
2193
2194    /* We only return an error if no datagrams were able to be sent */
2195    if (datagrams != 0)
2196        return datagrams;
2197
2198    return err;
2199}
2200
2201SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2202        unsigned int, vlen, unsigned int, flags)
2203{
2204    if (flags & MSG_CMSG_COMPAT)
2205        return -EINVAL;
2206    return __sys_sendmmsg(fd, mmsg, vlen, flags);
2207}
2208
2209static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2210             struct msghdr *msg_sys, unsigned int flags, int nosec)
2211{
2212    struct compat_msghdr __user *msg_compat =
2213        (struct compat_msghdr __user *)msg;
2214    struct iovec iovstack[UIO_FASTIOV];
2215    struct iovec *iov = iovstack;
2216    unsigned long cmsg_ptr;
2217    int err, total_len, len;
2218
2219    /* kernel mode address */
2220    struct sockaddr_storage addr;
2221
2222    /* user mode address pointers */
2223    struct sockaddr __user *uaddr;
2224    int __user *uaddr_len;
2225
2226    if (MSG_CMSG_COMPAT & flags) {
2227        if (get_compat_msghdr(msg_sys, msg_compat))
2228            return -EFAULT;
2229    } else {
2230        err = copy_msghdr_from_user(msg_sys, msg);
2231        if (err)
2232            return err;
2233    }
2234
2235    if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2236        err = -EMSGSIZE;
2237        if (msg_sys->msg_iovlen > UIO_MAXIOV)
2238            goto out;
2239        err = -ENOMEM;
2240        iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2241                  GFP_KERNEL);
2242        if (!iov)
2243            goto out;
2244    }
2245
2246    /* Save the user-mode address (verify_iovec will change the
2247     * kernel msghdr to use the kernel address space)
2248     */
2249    uaddr = (__force void __user *)msg_sys->msg_name;
2250    uaddr_len = COMPAT_NAMELEN(msg);
2251    if (MSG_CMSG_COMPAT & flags)
2252        err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2253    else
2254        err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2255    if (err < 0)
2256        goto out_freeiov;
2257    total_len = err;
2258
2259    cmsg_ptr = (unsigned long)msg_sys->msg_control;
2260    msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2261
2262    /* We assume all kernel code knows the size of sockaddr_storage */
2263    msg_sys->msg_namelen = 0;
2264
2265    if (sock->file->f_flags & O_NONBLOCK)
2266        flags |= MSG_DONTWAIT;
2267    err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2268                              total_len, flags);
2269    if (err < 0)
2270        goto out_freeiov;
2271    len = err;
2272
2273    if (uaddr != NULL) {
2274        err = move_addr_to_user(&addr,
2275                    msg_sys->msg_namelen, uaddr,
2276                    uaddr_len);
2277        if (err < 0)
2278            goto out_freeiov;
2279    }
2280    err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2281             COMPAT_FLAGS(msg));
2282    if (err)
2283        goto out_freeiov;
2284    if (MSG_CMSG_COMPAT & flags)
2285        err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2286                 &msg_compat->msg_controllen);
2287    else
2288        err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2289                 &msg->msg_controllen);
2290    if (err)
2291        goto out_freeiov;
2292    err = len;
2293
2294out_freeiov:
2295    if (iov != iovstack)
2296        kfree(iov);
2297out:
2298    return err;
2299}
2300
2301/*
2302 * BSD recvmsg interface
2303 */
2304
2305long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2306{
2307    int fput_needed, err;
2308    struct msghdr msg_sys;
2309    struct socket *sock;
2310
2311    sock = sockfd_lookup_light(fd, &err, &fput_needed);
2312    if (!sock)
2313        goto out;
2314
2315    err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2316
2317    fput_light(sock->file, fput_needed);
2318out:
2319    return err;
2320}
2321
2322SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2323        unsigned int, flags)
2324{
2325    if (flags & MSG_CMSG_COMPAT)
2326        return -EINVAL;
2327    return __sys_recvmsg(fd, msg, flags);
2328}
2329
2330/*
2331 * Linux recvmmsg interface
2332 */
2333
2334int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2335           unsigned int flags, struct timespec *timeout)
2336{
2337    int fput_needed, err, datagrams;
2338    struct socket *sock;
2339    struct mmsghdr __user *entry;
2340    struct compat_mmsghdr __user *compat_entry;
2341    struct msghdr msg_sys;
2342    struct timespec end_time;
2343
2344    if (timeout &&
2345        poll_select_set_timeout(&end_time, timeout->tv_sec,
2346                    timeout->tv_nsec))
2347        return -EINVAL;
2348
2349    datagrams = 0;
2350
2351    sock = sockfd_lookup_light(fd, &err, &fput_needed);
2352    if (!sock)
2353        return err;
2354
2355    err = sock_error(sock->sk);
2356    if (err)
2357        goto out_put;
2358
2359    entry = mmsg;
2360    compat_entry = (struct compat_mmsghdr __user *)mmsg;
2361
2362    while (datagrams < vlen) {
2363        /*
2364         * No need to ask LSM for more than the first datagram.
2365         */
2366        if (MSG_CMSG_COMPAT & flags) {
2367            err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2368                         &msg_sys, flags & ~MSG_WAITFORONE,
2369                         datagrams);
2370            if (err < 0)
2371                break;
2372            err = __put_user(err, &compat_entry->msg_len);
2373            ++compat_entry;
2374        } else {
2375            err = ___sys_recvmsg(sock,
2376                         (struct msghdr __user *)entry,
2377                         &msg_sys, flags & ~MSG_WAITFORONE,
2378                         datagrams);
2379            if (err < 0)
2380                break;
2381            err = put_user(err, &entry->msg_len);
2382            ++entry;
2383        }
2384
2385        if (err)
2386            break;
2387        ++datagrams;
2388
2389        /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2390        if (flags & MSG_WAITFORONE)
2391            flags |= MSG_DONTWAIT;
2392
2393        if (timeout) {
2394            ktime_get_ts(timeout);
2395            *timeout = timespec_sub(end_time, *timeout);
2396            if (timeout->tv_sec < 0) {
2397                timeout->tv_sec = timeout->tv_nsec = 0;
2398                break;
2399            }
2400
2401            /* Timeout, return less than vlen datagrams */
2402            if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2403                break;
2404        }
2405
2406        /* Out of band data, return right away */
2407        if (msg_sys.msg_flags & MSG_OOB)
2408            break;
2409    }
2410
2411out_put:
2412    fput_light(sock->file, fput_needed);
2413
2414    if (err == 0)
2415        return datagrams;
2416
2417    if (datagrams != 0) {
2418        /*
2419         * We may return less entries than requested (vlen) if the
2420         * sock is non block and there aren't enough datagrams...
2421         */
2422        if (err != -EAGAIN) {
2423            /*
2424             * ... or if recvmsg returns an error after we
2425             * received some datagrams, where we record the
2426             * error to return on the next call or if the
2427             * app asks about it using getsockopt(SO_ERROR).
2428             */
2429            sock->sk->sk_err = -err;
2430        }
2431
2432        return datagrams;
2433    }
2434
2435    return err;
2436}
2437
2438SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2439        unsigned int, vlen, unsigned int, flags,
2440        struct timespec __user *, timeout)
2441{
2442    int datagrams;
2443    struct timespec timeout_sys;
2444
2445    if (flags & MSG_CMSG_COMPAT)
2446        return -EINVAL;
2447
2448    if (!timeout)
2449        return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2450
2451    if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2452        return -EFAULT;
2453
2454    datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2455
2456    if (datagrams > 0 &&
2457        copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2458        datagrams = -EFAULT;
2459
2460    return datagrams;
2461}
2462
2463#ifdef __ARCH_WANT_SYS_SOCKETCALL
2464/* Argument list sizes for sys_socketcall */
2465#define AL(x) ((x) * sizeof(unsigned long))
2466static const unsigned char nargs[21] = {
2467    AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2468    AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2469    AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2470    AL(4), AL(5), AL(4)
2471};
2472
2473#undef AL
2474
2475/*
2476 * System call vectors.
2477 *
2478 * Argument checking cleaned up. Saved 20% in size.
2479 * This function doesn't need to set the kernel lock because
2480 * it is set by the callees.
2481 */
2482
2483SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2484{
2485    unsigned long a[AUDITSC_ARGS];
2486    unsigned long a0, a1;
2487    int err;
2488    unsigned int len;
2489
2490    if (call < 1 || call > SYS_SENDMMSG)
2491        return -EINVAL;
2492
2493    len = nargs[call];
2494    if (len > sizeof(a))
2495        return -EINVAL;
2496
2497    /* copy_from_user should be SMP safe. */
2498    if (copy_from_user(a, args, len))
2499        return -EFAULT;
2500
2501    err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2502    if (err)
2503        return err;
2504
2505    a0 = a[0];
2506    a1 = a[1];
2507
2508    switch (call) {
2509    case SYS_SOCKET:
2510        err = sys_socket(a0, a1, a[2]);
2511        break;
2512    case SYS_BIND:
2513        err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2514        break;
2515    case SYS_CONNECT:
2516        err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2517        break;
2518    case SYS_LISTEN:
2519        err = sys_listen(a0, a1);
2520        break;
2521    case SYS_ACCEPT:
2522        err = sys_accept4(a0, (struct sockaddr __user *)a1,
2523                  (int __user *)a[2], 0);
2524        break;
2525    case SYS_GETSOCKNAME:
2526        err =
2527            sys_getsockname(a0, (struct sockaddr __user *)a1,
2528                    (int __user *)a[2]);
2529        break;
2530    case SYS_GETPEERNAME:
2531        err =
2532            sys_getpeername(a0, (struct sockaddr __user *)a1,
2533                    (int __user *)a[2]);
2534        break;
2535    case SYS_SOCKETPAIR:
2536        err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2537        break;
2538    case SYS_SEND:
2539        err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2540        break;
2541    case SYS_SENDTO:
2542        err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2543                 (struct sockaddr __user *)a[4], a[5]);
2544        break;
2545    case SYS_RECV:
2546        err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2547        break;
2548    case SYS_RECVFROM:
2549        err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2550                   (struct sockaddr __user *)a[4],
2551                   (int __user *)a[5]);
2552        break;
2553    case SYS_SHUTDOWN:
2554        err = sys_shutdown(a0, a1);
2555        break;
2556    case SYS_SETSOCKOPT:
2557        err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2558        break;
2559    case SYS_GETSOCKOPT:
2560        err =
2561            sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2562                   (int __user *)a[4]);
2563        break;
2564    case SYS_SENDMSG:
2565        err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2566        break;
2567    case SYS_SENDMMSG:
2568        err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2569        break;
2570    case SYS_RECVMSG:
2571        err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2572        break;
2573    case SYS_RECVMMSG:
2574        err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2575                   (struct timespec __user *)a[4]);
2576        break;
2577    case SYS_ACCEPT4:
2578        err = sys_accept4(a0, (struct sockaddr __user *)a1,
2579                  (int __user *)a[2], a[3]);
2580        break;
2581    default:
2582        err = -EINVAL;
2583        break;
2584    }
2585    return err;
2586}
2587
2588#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2589
2590/**
2591 * sock_register - add a socket protocol handler
2592 * @ops: description of protocol
2593 *
2594 * This function is called by a protocol handler that wants to
2595 * advertise its address family, and have it linked into the
2596 * socket interface. The value ops->family coresponds to the
2597 * socket system call protocol family.
2598 */
2599int sock_register(const struct net_proto_family *ops)
2600{
2601    int err;
2602
2603    if (ops->family >= NPROTO) {
2604        pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2605        return -ENOBUFS;
2606    }
2607
2608    spin_lock(&net_family_lock);
2609    if (rcu_dereference_protected(net_families[ops->family],
2610                      lockdep_is_held(&net_family_lock)))
2611        err = -EEXIST;
2612    else {
2613        rcu_assign_pointer(net_families[ops->family], ops);
2614        err = 0;
2615    }
2616    spin_unlock(&net_family_lock);
2617
2618    pr_info("NET: Registered protocol family %d\n", ops->family);
2619    return err;
2620}
2621EXPORT_SYMBOL(sock_register);
2622
2623/**
2624 * sock_unregister - remove a protocol handler
2625 * @family: protocol family to remove
2626 *
2627 * This function is called by a protocol handler that wants to
2628 * remove its address family, and have it unlinked from the
2629 * new socket creation.
2630 *
2631 * If protocol handler is a module, then it can use module reference
2632 * counts to protect against new references. If protocol handler is not
2633 * a module then it needs to provide its own protection in
2634 * the ops->create routine.
2635 */
2636void sock_unregister(int family)
2637{
2638    BUG_ON(family < 0 || family >= NPROTO);
2639
2640    spin_lock(&net_family_lock);
2641    RCU_INIT_POINTER(net_families[family], NULL);
2642    spin_unlock(&net_family_lock);
2643
2644    synchronize_rcu();
2645
2646    pr_info("NET: Unregistered protocol family %d\n", family);
2647}
2648EXPORT_SYMBOL(sock_unregister);
2649
2650static int __init sock_init(void)
2651{
2652    int err;
2653    /*
2654     * Initialize the network sysctl infrastructure.
2655     */
2656    err = net_sysctl_init();
2657    if (err)
2658        goto out;
2659
2660    /*
2661     * Initialize skbuff SLAB cache
2662     */
2663    skb_init();
2664
2665    /*
2666     * Initialize the protocols module.
2667     */
2668
2669    init_inodecache();
2670
2671    err = register_filesystem(&sock_fs_type);
2672    if (err)
2673        goto out_fs;
2674    sock_mnt = kern_mount(&sock_fs_type);
2675    if (IS_ERR(sock_mnt)) {
2676        err = PTR_ERR(sock_mnt);
2677        goto out_mount;
2678    }
2679
2680    /* The real protocol initialization is performed in later initcalls.
2681     */
2682
2683#ifdef CONFIG_NETFILTER
2684    err = netfilter_init();
2685    if (err)
2686        goto out;
2687#endif
2688
2689    ptp_classifier_init();
2690
2691out:
2692    return err;
2693
2694out_mount:
2695    unregister_filesystem(&sock_fs_type);
2696out_fs:
2697    goto out;
2698}
2699
2700core_initcall(sock_init); /* early initcall */
2701
2702#ifdef CONFIG_PROC_FS
2703void socket_seq_show(struct seq_file *seq)
2704{
2705    int cpu;
2706    int counter = 0;
2707
2708    for_each_possible_cpu(cpu)
2709        counter += per_cpu(sockets_in_use, cpu);
2710
2711    /* It can be negative, by the way. 8) */
2712    if (counter < 0)
2713        counter = 0;
2714
2715    seq_printf(seq, "sockets: used %d\n", counter);
2716}
2717#endif /* CONFIG_PROC_FS */
2718
2719#ifdef CONFIG_COMPAT
2720static int do_siocgstamp(struct net *net, struct socket *sock,
2721             unsigned int cmd, void __user *up)
2722{
2723    mm_segment_t old_fs = get_fs();
2724    struct timeval ktv;
2725    int err;
2726
2727    set_fs(KERNEL_DS);
2728    err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2729    set_fs(old_fs);
2730    if (!err)
2731        err = compat_put_timeval(&ktv, up);
2732
2733    return err;
2734}
2735
2736static int do_siocgstampns(struct net *net, struct socket *sock,
2737               unsigned int cmd, void __user *up)
2738{
2739    mm_segment_t old_fs = get_fs();
2740    struct timespec kts;
2741    int err;
2742
2743    set_fs(KERNEL_DS);
2744    err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2745    set_fs(old_fs);
2746    if (!err)
2747        err = compat_put_timespec(&kts, up);
2748
2749    return err;
2750}
2751
2752static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2753{
2754    struct ifreq __user *uifr;
2755    int err;
2756
2757    uifr = compat_alloc_user_space(sizeof(struct ifreq));
2758    if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2759        return -EFAULT;
2760
2761    err = dev_ioctl(net, SIOCGIFNAME, uifr);
2762    if (err)
2763        return err;
2764
2765    if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2766        return -EFAULT;
2767
2768    return 0;
2769}
2770
2771static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2772{
2773    struct compat_ifconf ifc32;
2774    struct ifconf ifc;
2775    struct ifconf __user *uifc;
2776    struct compat_ifreq __user *ifr32;
2777    struct ifreq __user *ifr;
2778    unsigned int i, j;
2779    int err;
2780
2781    if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2782        return -EFAULT;
2783
2784    memset(&ifc, 0, sizeof(ifc));
2785    if (ifc32.ifcbuf == 0) {
2786        ifc32.ifc_len = 0;
2787        ifc.ifc_len = 0;
2788        ifc.ifc_req = NULL;
2789        uifc = compat_alloc_user_space(sizeof(struct ifconf));
2790    } else {
2791        size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2792            sizeof(struct ifreq);
2793        uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2794        ifc.ifc_len = len;
2795        ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2796        ifr32 = compat_ptr(ifc32.ifcbuf);
2797        for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2798            if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2799                return -EFAULT;
2800            ifr++;
2801            ifr32++;
2802        }
2803    }
2804    if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2805        return -EFAULT;
2806
2807    err = dev_ioctl(net, SIOCGIFCONF, uifc);
2808    if (err)
2809        return err;
2810
2811    if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2812        return -EFAULT;
2813
2814    ifr = ifc.ifc_req;
2815    ifr32 = compat_ptr(ifc32.ifcbuf);
2816    for (i = 0, j = 0;
2817         i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2818         i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2819        if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2820            return -EFAULT;
2821        ifr32++;
2822        ifr++;
2823    }
2824
2825    if (ifc32.ifcbuf == 0) {
2826        /* Translate from 64-bit structure multiple to
2827         * a 32-bit one.
2828         */
2829        i = ifc.ifc_len;
2830        i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2831        ifc32.ifc_len = i;
2832    } else {
2833        ifc32.ifc_len = i;
2834    }
2835    if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2836        return -EFAULT;
2837
2838    return 0;
2839}
2840
2841static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2842{
2843    struct compat_ethtool_rxnfc __user *compat_rxnfc;
2844    bool convert_in = false, convert_out = false;
2845    size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2846    struct ethtool_rxnfc __user *rxnfc;
2847    struct ifreq __user *ifr;
2848    u32 rule_cnt = 0, actual_rule_cnt;
2849    u32 ethcmd;
2850    u32 data;
2851    int ret;
2852
2853    if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2854        return -EFAULT;
2855
2856    compat_rxnfc = compat_ptr(data);
2857
2858    if (get_user(ethcmd, &compat_rxnfc->cmd))
2859        return -EFAULT;
2860
2861    /* Most ethtool structures are defined without padding.
2862     * Unfortunately struct ethtool_rxnfc is an exception.
2863     */
2864    switch (ethcmd) {
2865    default:
2866        break;
2867    case ETHTOOL_GRXCLSRLALL:
2868        /* Buffer size is variable */
2869        if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2870            return -EFAULT;
2871        if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2872            return -ENOMEM;
2873        buf_size += rule_cnt * sizeof(u32);
2874        /* fall through */
2875    case ETHTOOL_GRXRINGS:
2876    case ETHTOOL_GRXCLSRLCNT:
2877    case ETHTOOL_GRXCLSRULE:
2878    case ETHTOOL_SRXCLSRLINS:
2879        convert_out = true;
2880        /* fall through */
2881    case ETHTOOL_SRXCLSRLDEL:
2882        buf_size += sizeof(struct ethtool_rxnfc);
2883        convert_in = true;
2884        break;
2885    }
2886
2887    ifr = compat_alloc_user_space(buf_size);
2888    rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2889
2890    if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2891        return -EFAULT;
2892
2893    if (put_user(convert_in ? rxnfc : compat_ptr(data),
2894             &ifr->ifr_ifru.ifru_data))
2895        return -EFAULT;
2896
2897    if (convert_in) {
2898        /* We expect there to be holes between fs.m_ext and
2899         * fs.ring_cookie and at the end of fs, but nowhere else.
2900         */
2901        BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2902                 sizeof(compat_rxnfc->fs.m_ext) !=
2903                 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2904                 sizeof(rxnfc->fs.m_ext));
2905        BUILD_BUG_ON(
2906            offsetof(struct compat_ethtool_rxnfc, fs.location) -
2907            offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2908            offsetof(struct ethtool_rxnfc, fs.location) -
2909            offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2910
2911        if (copy_in_user(rxnfc, compat_rxnfc,
2912                 (void __user *)(&rxnfc->fs.m_ext + 1) -
2913                 (void __user *)rxnfc) ||
2914            copy_in_user(&rxnfc->fs.ring_cookie,
2915                 &compat_rxnfc->fs.ring_cookie,
2916                 (void __user *)(&rxnfc->fs.location + 1) -
2917                 (void __user *)&rxnfc->fs.ring_cookie) ||
2918            copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2919                 sizeof(rxnfc->rule_cnt)))
2920            return -EFAULT;
2921    }
2922
2923    ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2924    if (ret)
2925        return ret;
2926
2927    if (convert_out) {
2928        if (copy_in_user(compat_rxnfc, rxnfc,
2929                 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2930                 (const void __user *)rxnfc) ||
2931            copy_in_user(&compat_rxnfc->fs.ring_cookie,
2932                 &rxnfc->fs.ring_cookie,
2933                 (const void __user *)(&rxnfc->fs.location + 1) -
2934                 (const void __user *)&rxnfc->fs.ring_cookie) ||
2935            copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2936                 sizeof(rxnfc->rule_cnt)))
2937            return -EFAULT;
2938
2939        if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2940            /* As an optimisation, we only copy the actual
2941             * number of rules that the underlying
2942             * function returned. Since Mallory might
2943             * change the rule count in user memory, we
2944             * check that it is less than the rule count
2945             * originally given (as the user buffer size),
2946             * which has been range-checked.
2947             */
2948            if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2949                return -EFAULT;
2950            if (actual_rule_cnt < rule_cnt)
2951                rule_cnt = actual_rule_cnt;
2952            if (copy_in_user(&compat_rxnfc->rule_locs[0],
2953                     &rxnfc->rule_locs[0],
2954                     rule_cnt * sizeof(u32)))
2955                return -EFAULT;
2956        }
2957    }
2958
2959    return 0;
2960}
2961
2962static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2963{
2964    void __user *uptr;
2965    compat_uptr_t uptr32;
2966    struct ifreq __user *uifr;
2967
2968    uifr = compat_alloc_user_space(sizeof(*uifr));
2969    if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2970        return -EFAULT;
2971
2972    if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2973        return -EFAULT;
2974
2975    uptr = compat_ptr(uptr32);
2976
2977    if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2978        return -EFAULT;
2979
2980    return dev_ioctl(net, SIOCWANDEV, uifr);
2981}
2982
2983static int bond_ioctl(struct net *net, unsigned int cmd,
2984             struct compat_ifreq __user *ifr32)
2985{
2986    struct ifreq kifr;
2987    mm_segment_t old_fs;
2988    int err;
2989
2990    switch (cmd) {
2991    case SIOCBONDENSLAVE:
2992    case SIOCBONDRELEASE:
2993    case SIOCBONDSETHWADDR:
2994    case SIOCBONDCHANGEACTIVE:
2995        if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2996            return -EFAULT;
2997
2998        old_fs = get_fs();
2999        set_fs(KERNEL_DS);
3000        err = dev_ioctl(net, cmd,
3001                (struct ifreq __user __force *) &kifr);
3002        set_fs(old_fs);
3003
3004        return err;
3005    default:
3006        return -ENOIOCTLCMD;
3007    }
3008}
3009
3010/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3011static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3012                 struct compat_ifreq __user *u_ifreq32)
3013{
3014    struct ifreq __user *u_ifreq64;
3015    char tmp_buf[IFNAMSIZ];
3016    void __user *data64;
3017    u32 data32;
3018
3019    if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3020               IFNAMSIZ))
3021        return -EFAULT;
3022    if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3023        return -EFAULT;
3024    data64 = compat_ptr(data32);
3025
3026    u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3027
3028    if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3029             IFNAMSIZ))
3030        return -EFAULT;
3031    if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3032        return -EFAULT;
3033
3034    return dev_ioctl(net, cmd, u_ifreq64);
3035}
3036
3037static int dev_ifsioc(struct net *net, struct socket *sock,
3038             unsigned int cmd, struct compat_ifreq __user *uifr32)
3039{
3040    struct ifreq __user *uifr;
3041    int err;
3042
3043    uifr = compat_alloc_user_space(sizeof(*uifr));
3044    if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3045        return -EFAULT;
3046
3047    err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3048
3049    if (!err) {
3050        switch (cmd) {
3051        case SIOCGIFFLAGS:
3052        case SIOCGIFMETRIC:
3053        case SIOCGIFMTU:
3054        case SIOCGIFMEM:
3055        case SIOCGIFHWADDR:
3056        case SIOCGIFINDEX:
3057        case SIOCGIFADDR:
3058        case SIOCGIFBRDADDR:
3059        case SIOCGIFDSTADDR:
3060        case SIOCGIFNETMASK:
3061        case SIOCGIFPFLAGS:
3062        case SIOCGIFTXQLEN:
3063        case SIOCGMIIPHY:
3064        case SIOCGMIIREG:
3065            if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3066                err = -EFAULT;
3067            break;
3068        }
3069    }
3070    return err;
3071}
3072
3073static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3074            struct compat_ifreq __user *uifr32)
3075{
3076    struct ifreq ifr;
3077    struct compat_ifmap __user *uifmap32;
3078    mm_segment_t old_fs;
3079    int err;
3080
3081    uifmap32 = &uifr32->ifr_ifru.ifru_map;
3082    err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3083    err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3084    err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3085    err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3086    err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3087    err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3088    err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3089    if (err)
3090        return -EFAULT;
3091
3092    old_fs = get_fs();
3093    set_fs(KERNEL_DS);
3094    err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3095    set_fs(old_fs);
3096
3097    if (cmd == SIOCGIFMAP && !err) {
3098        err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3099        err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3100        err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3101        err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3102        err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3103        err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3104        err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3105        if (err)
3106            err = -EFAULT;
3107    }
3108    return err;
3109}
3110
3111struct rtentry32 {
3112    u32 rt_pad1;
3113    struct sockaddr rt_dst; /* target address */
3114    struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3115    struct sockaddr rt_genmask; /* target network mask (IP) */
3116    unsigned short rt_flags;
3117    short rt_pad2;
3118    u32 rt_pad3;
3119    unsigned char rt_tos;
3120    unsigned char rt_class;
3121    short rt_pad4;
3122    short rt_metric; /* +1 for binary compatibility! */
3123    /* char * */ u32 rt_dev; /* forcing the device at add */
3124    u32 rt_mtu; /* per route MTU/Window */
3125    u32 rt_window; /* Window clamping */
3126    unsigned short rt_irtt; /* Initial RTT */
3127};
3128
3129struct in6_rtmsg32 {
3130    struct in6_addr rtmsg_dst;
3131    struct in6_addr rtmsg_src;
3132    struct in6_addr rtmsg_gateway;
3133    u32 rtmsg_type;
3134    u16 rtmsg_dst_len;
3135    u16 rtmsg_src_len;
3136    u32 rtmsg_metric;
3137    u32 rtmsg_info;
3138    u32 rtmsg_flags;
3139    s32 rtmsg_ifindex;
3140};
3141
3142static int routing_ioctl(struct net *net, struct socket *sock,
3143             unsigned int cmd, void __user *argp)
3144{
3145    int ret;
3146    void *r = NULL;
3147    struct in6_rtmsg r6;
3148    struct rtentry r4;
3149    char devname[16];
3150    u32 rtdev;
3151    mm_segment_t old_fs = get_fs();
3152
3153    if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3154        struct in6_rtmsg32 __user *ur6 = argp;
3155        ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3156            3 * sizeof(struct in6_addr));
3157        ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3158        ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3159        ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3160        ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3161        ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3162        ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3163        ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3164
3165        r = (void *) &r6;
3166    } else { /* ipv4 */
3167        struct rtentry32 __user *ur4 = argp;
3168        ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3169                    3 * sizeof(struct sockaddr));
3170        ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3171        ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3172        ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3173        ret |= get_user(r4.rt_window, &(ur4->rt_window));
3174        ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3175        ret |= get_user(rtdev, &(ur4->rt_dev));
3176        if (rtdev) {
3177            ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3178            r4.rt_dev = (char __user __force *)devname;
3179            devname[15] = 0;
3180        } else
3181            r4.rt_dev = NULL;
3182
3183        r = (void *) &r4;
3184    }
3185
3186    if (ret) {
3187        ret = -EFAULT;
3188        goto out;
3189    }
3190
3191    set_fs(KERNEL_DS);
3192    ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3193    set_fs(old_fs);
3194
3195out:
3196    return ret;
3197}
3198
3199/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3200 * for some operations; this forces use of the newer bridge-utils that
3201 * use compatible ioctls
3202 */
3203static int old_bridge_ioctl(compat_ulong_t __user *argp)
3204{
3205    compat_ulong_t tmp;
3206
3207    if (get_user(tmp, argp))
3208        return -EFAULT;
3209    if (tmp == BRCTL_GET_VERSION)
3210        return BRCTL_VERSION + 1;
3211    return -EINVAL;
3212}
3213
3214static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3215             unsigned int cmd, unsigned long arg)
3216{
3217    void __user *argp = compat_ptr(arg);
3218    struct sock *sk = sock->sk;
3219    struct net *net = sock_net(sk);
3220
3221    if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3222        return compat_ifr_data_ioctl(net, cmd, argp);
3223
3224    switch (cmd) {
3225    case SIOCSIFBR:
3226    case SIOCGIFBR:
3227        return old_bridge_ioctl(argp);
3228    case SIOCGIFNAME:
3229        return dev_ifname32(net, argp);
3230    case SIOCGIFCONF:
3231        return dev_ifconf(net, argp);
3232    case SIOCETHTOOL:
3233        return ethtool_ioctl(net, argp);
3234    case SIOCWANDEV:
3235        return compat_siocwandev(net, argp);
3236    case SIOCGIFMAP:
3237    case SIOCSIFMAP:
3238        return compat_sioc_ifmap(net, cmd, argp);
3239    case SIOCBONDENSLAVE:
3240    case SIOCBONDRELEASE:
3241    case SIOCBONDSETHWADDR:
3242    case SIOCBONDCHANGEACTIVE:
3243        return bond_ioctl(net, cmd, argp);
3244    case SIOCADDRT:
3245    case SIOCDELRT:
3246        return routing_ioctl(net, sock, cmd, argp);
3247    case SIOCGSTAMP:
3248        return do_siocgstamp(net, sock, cmd, argp);
3249    case SIOCGSTAMPNS:
3250        return do_siocgstampns(net, sock, cmd, argp);
3251    case SIOCBONDSLAVEINFOQUERY:
3252    case SIOCBONDINFOQUERY:
3253    case SIOCSHWTSTAMP:
3254    case SIOCGHWTSTAMP:
3255        return compat_ifr_data_ioctl(net, cmd, argp);
3256
3257    case FIOSETOWN:
3258    case SIOCSPGRP:
3259    case FIOGETOWN:
3260    case SIOCGPGRP:
3261    case SIOCBRADDBR:
3262    case SIOCBRDELBR:
3263    case SIOCGIFVLAN:
3264    case SIOCSIFVLAN:
3265    case SIOCADDDLCI:
3266    case SIOCDELDLCI:
3267        return sock_ioctl(file, cmd, arg);
3268
3269    case SIOCGIFFLAGS:
3270    case SIOCSIFFLAGS:
3271    case SIOCGIFMETRIC:
3272    case SIOCSIFMETRIC:
3273    case SIOCGIFMTU:
3274    case SIOCSIFMTU:
3275    case SIOCGIFMEM:
3276    case SIOCSIFMEM:
3277    case SIOCGIFHWADDR:
3278    case SIOCSIFHWADDR:
3279    case SIOCADDMULTI:
3280    case SIOCDELMULTI:
3281    case SIOCGIFINDEX:
3282    case SIOCGIFADDR:
3283    case SIOCSIFADDR:
3284    case SIOCSIFHWBROADCAST:
3285    case SIOCDIFADDR:
3286    case SIOCGIFBRDADDR:
3287    case SIOCSIFBRDADDR:
3288    case SIOCGIFDSTADDR:
3289    case SIOCSIFDSTADDR:
3290    case SIOCGIFNETMASK:
3291    case SIOCSIFNETMASK:
3292    case SIOCSIFPFLAGS:
3293    case SIOCGIFPFLAGS:
3294    case SIOCGIFTXQLEN:
3295    case SIOCSIFTXQLEN:
3296    case SIOCBRADDIF:
3297    case SIOCBRDELIF:
3298    case SIOCSIFNAME:
3299    case SIOCGMIIPHY:
3300    case SIOCGMIIREG:
3301    case SIOCSMIIREG:
3302        return dev_ifsioc(net, sock, cmd, argp);
3303
3304    case SIOCSARP:
3305    case SIOCGARP:
3306    case SIOCDARP:
3307    case SIOCATMARK:
3308        return sock_do_ioctl(net, sock, cmd, arg);
3309    }
3310
3311    return -ENOIOCTLCMD;
3312}
3313
3314static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3315                  unsigned long arg)
3316{
3317    struct socket *sock = file->private_data;
3318    int ret = -ENOIOCTLCMD;
3319    struct sock *sk;
3320    struct net *net;
3321
3322    sk = sock->sk;
3323    net = sock_net(sk);
3324
3325    if (sock->ops->compat_ioctl)
3326        ret = sock->ops->compat_ioctl(sock, cmd, arg);
3327
3328    if (ret == -ENOIOCTLCMD &&
3329        (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3330        ret = compat_wext_handle_ioctl(net, cmd, arg);
3331
3332    if (ret == -ENOIOCTLCMD)
3333        ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3334
3335    return ret;
3336}
3337#endif
3338
3339int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3340{
3341    return sock->ops->bind(sock, addr, addrlen);
3342}
3343EXPORT_SYMBOL(kernel_bind);
3344
3345int kernel_listen(struct socket *sock, int backlog)
3346{
3347    return sock->ops->listen(sock, backlog);
3348}
3349EXPORT_SYMBOL(kernel_listen);
3350
3351int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3352{
3353    struct sock *sk = sock->sk;
3354    int err;
3355
3356    err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3357                   newsock);
3358    if (err < 0)
3359        goto done;
3360
3361    err = sock->ops->accept(sock, *newsock, flags);
3362    if (err < 0) {
3363        sock_release(*newsock);
3364        *newsock = NULL;
3365        goto done;
3366    }
3367
3368    (*newsock)->ops = sock->ops;
3369    __module_get((*newsock)->ops->owner);
3370
3371done:
3372    return err;
3373}
3374EXPORT_SYMBOL(kernel_accept);
3375
3376int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3377           int flags)
3378{
3379    return sock->ops->connect(sock, addr, addrlen, flags);
3380}
3381EXPORT_SYMBOL(kernel_connect);
3382
3383int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3384             int *addrlen)
3385{
3386    return sock->ops->getname(sock, addr, addrlen, 0);
3387}
3388EXPORT_SYMBOL(kernel_getsockname);
3389
3390int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3391             int *addrlen)
3392{
3393    return sock->ops->getname(sock, addr, addrlen, 1);
3394}
3395EXPORT_SYMBOL(kernel_getpeername);
3396
3397int kernel_getsockopt(struct socket *sock, int level, int optname,
3398            char *optval, int *optlen)
3399{
3400    mm_segment_t oldfs = get_fs();
3401    char __user *uoptval;
3402    int __user *uoptlen;
3403    int err;
3404
3405    uoptval = (char __user __force *) optval;
3406    uoptlen = (int __user __force *) optlen;
3407
3408    set_fs(KERNEL_DS);
3409    if (level == SOL_SOCKET)
3410        err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3411    else
3412        err = sock->ops->getsockopt(sock, level, optname, uoptval,
3413                        uoptlen);
3414    set_fs(oldfs);
3415    return err;
3416}
3417EXPORT_SYMBOL(kernel_getsockopt);
3418
3419int kernel_setsockopt(struct socket *sock, int level, int optname,
3420            char *optval, unsigned int optlen)
3421{
3422    mm_segment_t oldfs = get_fs();
3423    char __user *uoptval;
3424    int err;
3425
3426    uoptval = (char __user __force *) optval;
3427
3428    set_fs(KERNEL_DS);
3429    if (level == SOL_SOCKET)
3430        err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3431    else
3432        err = sock->ops->setsockopt(sock, level, optname, uoptval,
3433                        optlen);
3434    set_fs(oldfs);
3435    return err;
3436}
3437EXPORT_SYMBOL(kernel_setsockopt);
3438
3439int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3440            size_t size, int flags)
3441{
3442    if (sock->ops->sendpage)
3443        return sock->ops->sendpage(sock, page, offset, size, flags);
3444
3445    return sock_no_sendpage(sock, page, offset, size, flags);
3446}
3447EXPORT_SYMBOL(kernel_sendpage);
3448
3449int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3450{
3451    mm_segment_t oldfs = get_fs();
3452    int err;
3453
3454    set_fs(KERNEL_DS);
3455    err = sock->ops->ioctl(sock, cmd, arg);
3456    set_fs(oldfs);
3457
3458    return err;
3459}
3460EXPORT_SYMBOL(kernel_sock_ioctl);
3461
3462int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3463{
3464    return sock->ops->shutdown(sock, how);
3465}
3466EXPORT_SYMBOL(kernel_sock_shutdown);
3467

Archive Download this file



interactive