Version:  2.0.40 2.2.26 2.4.37 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19

Linux/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 #include <linux/errqueue.h>
110 
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
114 #endif
115 
116 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
117                          unsigned long nr_segs, loff_t pos);
118 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
119                           unsigned long nr_segs, loff_t pos);
120 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121 
122 static int sock_close(struct inode *inode, struct file *file);
123 static unsigned int sock_poll(struct file *file,
124                               struct poll_table_struct *wait);
125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 #ifdef CONFIG_COMPAT
127 static long compat_sock_ioctl(struct file *file,
128                               unsigned int cmd, unsigned long arg);
129 #endif
130 static int sock_fasync(int fd, struct file *filp, int on);
131 static ssize_t sock_sendpage(struct file *file, struct page *page,
132                              int offset, size_t size, loff_t *ppos, int more);
133 static 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 
142 static 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         .release =      sock_close,
154         .fasync =       sock_fasync,
155         .sendpage =     sock_sendpage,
156         .splice_write = generic_splice_sendpage,
157         .splice_read =  sock_splice_read,
158 };
159 
160 /*
161  *      The protocol list. Each protocol is registered in here.
162  */
163 
164 static DEFINE_SPINLOCK(net_family_lock);
165 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
166 
167 /*
168  *      Statistics counters of the socket lists
169  */
170 
171 static DEFINE_PER_CPU(int, sockets_in_use);
172 
173 /*
174  * Support routines.
175  * Move socket addresses back and forth across the kernel/user
176  * divide and look after the messy bits.
177  */
178 
179 /**
180  *      move_addr_to_kernel     -       copy a socket address into kernel space
181  *      @uaddr: Address in user space
182  *      @kaddr: Address in kernel space
183  *      @ulen: Length in user space
184  *
185  *      The address is copied into kernel space. If the provided address is
186  *      too long an error code of -EINVAL is returned. If the copy gives
187  *      invalid addresses -EFAULT is returned. On a success 0 is returned.
188  */
189 
190 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
191 {
192         if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
193                 return -EINVAL;
194         if (ulen == 0)
195                 return 0;
196         if (copy_from_user(kaddr, uaddr, ulen))
197                 return -EFAULT;
198         return audit_sockaddr(ulen, kaddr);
199 }
200 
201 /**
202  *      move_addr_to_user       -       copy an address to user space
203  *      @kaddr: kernel space address
204  *      @klen: length of address in kernel
205  *      @uaddr: user space address
206  *      @ulen: pointer to user length field
207  *
208  *      The value pointed to by ulen on entry is the buffer length available.
209  *      This is overwritten with the buffer space used. -EINVAL is returned
210  *      if an overlong buffer is specified or a negative buffer size. -EFAULT
211  *      is returned if either the buffer or the length field are not
212  *      accessible.
213  *      After copying the data up to the limit the user specifies, the true
214  *      length of the data is written over the length limit the user
215  *      specified. Zero is returned for a success.
216  */
217 
218 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
219                              void __user *uaddr, int __user *ulen)
220 {
221         int err;
222         int len;
223 
224         BUG_ON(klen > sizeof(struct sockaddr_storage));
225         err = get_user(len, ulen);
226         if (err)
227                 return err;
228         if (len > klen)
229                 len = klen;
230         if (len < 0)
231                 return -EINVAL;
232         if (len) {
233                 if (audit_sockaddr(klen, kaddr))
234                         return -ENOMEM;
235                 if (copy_to_user(uaddr, kaddr, len))
236                         return -EFAULT;
237         }
238         /*
239          *      "fromlen shall refer to the value before truncation.."
240          *                      1003.1g
241          */
242         return __put_user(klen, ulen);
243 }
244 
245 static struct kmem_cache *sock_inode_cachep __read_mostly;
246 
247 static struct inode *sock_alloc_inode(struct super_block *sb)
248 {
249         struct socket_alloc *ei;
250         struct socket_wq *wq;
251 
252         ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
253         if (!ei)
254                 return NULL;
255         wq = kmalloc(sizeof(*wq), GFP_KERNEL);
256         if (!wq) {
257                 kmem_cache_free(sock_inode_cachep, ei);
258                 return NULL;
259         }
260         init_waitqueue_head(&wq->wait);
261         wq->fasync_list = NULL;
262         RCU_INIT_POINTER(ei->socket.wq, wq);
263 
264         ei->socket.state = SS_UNCONNECTED;
265         ei->socket.flags = 0;
266         ei->socket.ops = NULL;
267         ei->socket.sk = NULL;
268         ei->socket.file = NULL;
269 
270         return &ei->vfs_inode;
271 }
272 
273 static void sock_destroy_inode(struct inode *inode)
274 {
275         struct socket_alloc *ei;
276         struct socket_wq *wq;
277 
278         ei = container_of(inode, struct socket_alloc, vfs_inode);
279         wq = rcu_dereference_protected(ei->socket.wq, 1);
280         kfree_rcu(wq, rcu);
281         kmem_cache_free(sock_inode_cachep, ei);
282 }
283 
284 static void init_once(void *foo)
285 {
286         struct socket_alloc *ei = (struct socket_alloc *)foo;
287 
288         inode_init_once(&ei->vfs_inode);
289 }
290 
291 static int init_inodecache(void)
292 {
293         sock_inode_cachep = kmem_cache_create("sock_inode_cache",
294                                               sizeof(struct socket_alloc),
295                                               0,
296                                               (SLAB_HWCACHE_ALIGN |
297                                                SLAB_RECLAIM_ACCOUNT |
298                                                SLAB_MEM_SPREAD),
299                                               init_once);
300         if (sock_inode_cachep == NULL)
301                 return -ENOMEM;
302         return 0;
303 }
304 
305 static const struct super_operations sockfs_ops = {
306         .alloc_inode    = sock_alloc_inode,
307         .destroy_inode  = sock_destroy_inode,
308         .statfs         = simple_statfs,
309 };
310 
311 /*
312  * sockfs_dname() is called from d_path().
313  */
314 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
315 {
316         return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317                                 dentry->d_inode->i_ino);
318 }
319 
320 static const struct dentry_operations sockfs_dentry_operations = {
321         .d_dname  = sockfs_dname,
322 };
323 
324 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325                          int flags, const char *dev_name, void *data)
326 {
327         return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328                 &sockfs_dentry_operations, SOCKFS_MAGIC);
329 }
330 
331 static struct vfsmount *sock_mnt __read_mostly;
332 
333 static struct file_system_type sock_fs_type = {
334         .name =         "sockfs",
335         .mount =        sockfs_mount,
336         .kill_sb =      kill_anon_super,
337 };
338 
339 /*
340  *      Obtains the first available file descriptor and sets it up for use.
341  *
342  *      These functions create file structures and maps them to fd space
343  *      of the current process. On success it returns file descriptor
344  *      and file struct implicitly stored in sock->file.
345  *      Note that another thread may close file descriptor before we return
346  *      from this function. We use the fact that now we do not refer
347  *      to socket after mapping. If one day we will need it, this
348  *      function will increment ref. count on file by 1.
349  *
350  *      In any case returned fd MAY BE not valid!
351  *      This race condition is unavoidable
352  *      with shared fd spaces, we cannot solve it inside kernel,
353  *      but we take care of internal coherence yet.
354  */
355 
356 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
357 {
358         struct qstr name = { .name = "" };
359         struct path path;
360         struct file *file;
361 
362         if (dname) {
363                 name.name = dname;
364                 name.len = strlen(name.name);
365         } else if (sock->sk) {
366                 name.name = sock->sk->sk_prot_creator->name;
367                 name.len = strlen(name.name);
368         }
369         path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
370         if (unlikely(!path.dentry))
371                 return ERR_PTR(-ENOMEM);
372         path.mnt = mntget(sock_mnt);
373 
374         d_instantiate(path.dentry, SOCK_INODE(sock));
375 
376         file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
377                   &socket_file_ops);
378         if (unlikely(IS_ERR(file))) {
379                 /* drop dentry, keep inode */
380                 ihold(path.dentry->d_inode);
381                 path_put(&path);
382                 return file;
383         }
384 
385         sock->file = file;
386         file->f_flags = O_RDWR | (flags & O_NONBLOCK);
387         file->private_data = sock;
388         return file;
389 }
390 EXPORT_SYMBOL(sock_alloc_file);
391 
392 static int sock_map_fd(struct socket *sock, int flags)
393 {
394         struct file *newfile;
395         int fd = get_unused_fd_flags(flags);
396         if (unlikely(fd < 0))
397                 return fd;
398 
399         newfile = sock_alloc_file(sock, flags, NULL);
400         if (likely(!IS_ERR(newfile))) {
401                 fd_install(fd, newfile);
402                 return fd;
403         }
404 
405         put_unused_fd(fd);
406         return PTR_ERR(newfile);
407 }
408 
409 struct socket *sock_from_file(struct file *file, int *err)
410 {
411         if (file->f_op == &socket_file_ops)
412                 return file->private_data;      /* set in sock_map_fd */
413 
414         *err = -ENOTSOCK;
415         return NULL;
416 }
417 EXPORT_SYMBOL(sock_from_file);
418 
419 /**
420  *      sockfd_lookup - Go from a file number to its socket slot
421  *      @fd: file handle
422  *      @err: pointer to an error code return
423  *
424  *      The file handle passed in is locked and the socket it is bound
425  *      too is returned. If an error occurs the err pointer is overwritten
426  *      with a negative errno code and NULL is returned. The function checks
427  *      for both invalid handles and passing a handle which is not a socket.
428  *
429  *      On a success the socket object pointer is returned.
430  */
431 
432 struct socket *sockfd_lookup(int fd, int *err)
433 {
434         struct file *file;
435         struct socket *sock;
436 
437         file = fget(fd);
438         if (!file) {
439                 *err = -EBADF;
440                 return NULL;
441         }
442 
443         sock = sock_from_file(file, err);
444         if (!sock)
445                 fput(file);
446         return sock;
447 }
448 EXPORT_SYMBOL(sockfd_lookup);
449 
450 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
451 {
452         struct fd f = fdget(fd);
453         struct socket *sock;
454 
455         *err = -EBADF;
456         if (f.file) {
457                 sock = sock_from_file(f.file, err);
458                 if (likely(sock)) {
459                         *fput_needed = f.flags;
460                         return sock;
461                 }
462                 fdput(f);
463         }
464         return NULL;
465 }
466 
467 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470 static ssize_t sockfs_getxattr(struct dentry *dentry,
471                                const char *name, void *value, size_t size)
472 {
473         const char *proto_name;
474         size_t proto_size;
475         int error;
476 
477         error = -ENODATA;
478         if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479                 proto_name = dentry->d_name.name;
480                 proto_size = strlen(proto_name);
481 
482                 if (value) {
483                         error = -ERANGE;
484                         if (proto_size + 1 > size)
485                                 goto out;
486 
487                         strncpy(value, proto_name, proto_size + 1);
488                 }
489                 error = proto_size + 1;
490         }
491 
492 out:
493         return error;
494 }
495 
496 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
497                                 size_t size)
498 {
499         ssize_t len;
500         ssize_t used = 0;
501 
502         len = security_inode_listsecurity(dentry->d_inode, buffer, size);
503         if (len < 0)
504                 return len;
505         used += len;
506         if (buffer) {
507                 if (size < used)
508                         return -ERANGE;
509                 buffer += len;
510         }
511 
512         len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
513         used += len;
514         if (buffer) {
515                 if (size < used)
516                         return -ERANGE;
517                 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
518                 buffer += len;
519         }
520 
521         return used;
522 }
523 
524 static const struct inode_operations sockfs_inode_ops = {
525         .getxattr = sockfs_getxattr,
526         .listxattr = sockfs_listxattr,
527 };
528 
529 /**
530  *      sock_alloc      -       allocate a socket
531  *
532  *      Allocate a new inode and socket object. The two are bound together
533  *      and initialised. The socket is then returned. If we are out of inodes
534  *      NULL is returned.
535  */
536 
537 static struct socket *sock_alloc(void)
538 {
539         struct inode *inode;
540         struct socket *sock;
541 
542         inode = new_inode_pseudo(sock_mnt->mnt_sb);
543         if (!inode)
544                 return NULL;
545 
546         sock = SOCKET_I(inode);
547 
548         kmemcheck_annotate_bitfield(sock, type);
549         inode->i_ino = get_next_ino();
550         inode->i_mode = S_IFSOCK | S_IRWXUGO;
551         inode->i_uid = current_fsuid();
552         inode->i_gid = current_fsgid();
553         inode->i_op = &sockfs_inode_ops;
554 
555         this_cpu_add(sockets_in_use, 1);
556         return sock;
557 }
558 
559 /**
560  *      sock_release    -       close a socket
561  *      @sock: socket to close
562  *
563  *      The socket is released from the protocol stack if it has a release
564  *      callback, and the inode is then released if the socket is bound to
565  *      an inode not a file.
566  */
567 
568 void sock_release(struct socket *sock)
569 {
570         if (sock->ops) {
571                 struct module *owner = sock->ops->owner;
572 
573                 sock->ops->release(sock);
574                 sock->ops = NULL;
575                 module_put(owner);
576         }
577 
578         if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
579                 pr_err("%s: fasync list not empty!\n", __func__);
580 
581         if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
582                 return;
583 
584         this_cpu_sub(sockets_in_use, 1);
585         if (!sock->file) {
586                 iput(SOCK_INODE(sock));
587                 return;
588         }
589         sock->file = NULL;
590 }
591 EXPORT_SYMBOL(sock_release);
592 
593 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
594 {
595         u8 flags = *tx_flags;
596 
597         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
598                 flags |= SKBTX_HW_TSTAMP;
599 
600         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
601                 flags |= SKBTX_SW_TSTAMP;
602 
603         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
604                 flags |= SKBTX_SCHED_TSTAMP;
605 
606         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
607                 flags |= SKBTX_ACK_TSTAMP;
608 
609         *tx_flags = flags;
610 }
611 EXPORT_SYMBOL(__sock_tx_timestamp);
612 
613 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
614                                        struct msghdr *msg, size_t size)
615 {
616         struct sock_iocb *si = kiocb_to_siocb(iocb);
617 
618         si->sock = sock;
619         si->scm = NULL;
620         si->msg = msg;
621         si->size = size;
622 
623         return sock->ops->sendmsg(iocb, sock, msg, size);
624 }
625 
626 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
627                                  struct msghdr *msg, size_t size)
628 {
629         int err = security_socket_sendmsg(sock, msg, size);
630 
631         return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
632 }
633 
634 static int do_sock_sendmsg(struct socket *sock, struct msghdr *msg,
635                            size_t size, bool nosec)
636 {
637         struct kiocb iocb;
638         struct sock_iocb siocb;
639         int ret;
640 
641         init_sync_kiocb(&iocb, NULL);
642         iocb.private = &siocb;
643         ret = nosec ? __sock_sendmsg_nosec(&iocb, sock, msg, size) :
644                       __sock_sendmsg(&iocb, sock, msg, size);
645         if (-EIOCBQUEUED == ret)
646                 ret = wait_on_sync_kiocb(&iocb);
647         return ret;
648 }
649 
650 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
651 {
652         return do_sock_sendmsg(sock, msg, size, false);
653 }
654 EXPORT_SYMBOL(sock_sendmsg);
655 
656 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
657 {
658         return do_sock_sendmsg(sock, msg, size, true);
659 }
660 
661 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
662                    struct kvec *vec, size_t num, size_t size)
663 {
664         mm_segment_t oldfs = get_fs();
665         int result;
666 
667         set_fs(KERNEL_DS);
668         /*
669          * the following is safe, since for compiler definitions of kvec and
670          * iovec are identical, yielding the same in-core layout and alignment
671          */
672         iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
673         result = sock_sendmsg(sock, msg, size);
674         set_fs(oldfs);
675         return result;
676 }
677 EXPORT_SYMBOL(kernel_sendmsg);
678 
679 /*
680  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
681  */
682 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
683         struct sk_buff *skb)
684 {
685         int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
686         struct scm_timestamping tss;
687         int empty = 1;
688         struct skb_shared_hwtstamps *shhwtstamps =
689                 skb_hwtstamps(skb);
690 
691         /* Race occurred between timestamp enabling and packet
692            receiving.  Fill in the current time for now. */
693         if (need_software_tstamp && skb->tstamp.tv64 == 0)
694                 __net_timestamp(skb);
695 
696         if (need_software_tstamp) {
697                 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
698                         struct timeval tv;
699                         skb_get_timestamp(skb, &tv);
700                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
701                                  sizeof(tv), &tv);
702                 } else {
703                         struct timespec ts;
704                         skb_get_timestampns(skb, &ts);
705                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
706                                  sizeof(ts), &ts);
707                 }
708         }
709 
710         memset(&tss, 0, sizeof(tss));
711         if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
712             ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
713                 empty = 0;
714         if (shhwtstamps &&
715             (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
716             ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
717                 empty = 0;
718         if (!empty)
719                 put_cmsg(msg, SOL_SOCKET,
720                          SCM_TIMESTAMPING, sizeof(tss), &tss);
721 }
722 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
723 
724 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
725         struct sk_buff *skb)
726 {
727         int ack;
728 
729         if (!sock_flag(sk, SOCK_WIFI_STATUS))
730                 return;
731         if (!skb->wifi_acked_valid)
732                 return;
733 
734         ack = skb->wifi_acked;
735 
736         put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
737 }
738 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
739 
740 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
741                                    struct sk_buff *skb)
742 {
743         if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
744                 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
745                         sizeof(__u32), &skb->dropcount);
746 }
747 
748 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
749         struct sk_buff *skb)
750 {
751         sock_recv_timestamp(msg, sk, skb);
752         sock_recv_drops(msg, sk, skb);
753 }
754 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
755 
756 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
757                                        struct msghdr *msg, size_t size, int flags)
758 {
759         struct sock_iocb *si = kiocb_to_siocb(iocb);
760 
761         si->sock = sock;
762         si->scm = NULL;
763         si->msg = msg;
764         si->size = size;
765         si->flags = flags;
766 
767         return sock->ops->recvmsg(iocb, sock, msg, size, flags);
768 }
769 
770 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
771                                  struct msghdr *msg, size_t size, int flags)
772 {
773         int err = security_socket_recvmsg(sock, msg, size, flags);
774 
775         return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
776 }
777 
778 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
779                  size_t size, int flags)
780 {
781         struct kiocb iocb;
782         struct sock_iocb siocb;
783         int ret;
784 
785         init_sync_kiocb(&iocb, NULL);
786         iocb.private = &siocb;
787         ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
788         if (-EIOCBQUEUED == ret)
789                 ret = wait_on_sync_kiocb(&iocb);
790         return ret;
791 }
792 EXPORT_SYMBOL(sock_recvmsg);
793 
794 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
795                               size_t size, int flags)
796 {
797         struct kiocb iocb;
798         struct sock_iocb siocb;
799         int ret;
800 
801         init_sync_kiocb(&iocb, NULL);
802         iocb.private = &siocb;
803         ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
804         if (-EIOCBQUEUED == ret)
805                 ret = wait_on_sync_kiocb(&iocb);
806         return ret;
807 }
808 
809 /**
810  * kernel_recvmsg - Receive a message from a socket (kernel space)
811  * @sock:       The socket to receive the message from
812  * @msg:        Received message
813  * @vec:        Input s/g array for message data
814  * @num:        Size of input s/g array
815  * @size:       Number of bytes to read
816  * @flags:      Message flags (MSG_DONTWAIT, etc...)
817  *
818  * On return the msg structure contains the scatter/gather array passed in the
819  * vec argument. The array is modified so that it consists of the unfilled
820  * portion of the original array.
821  *
822  * The returned value is the total number of bytes received, or an error.
823  */
824 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
825                    struct kvec *vec, size_t num, size_t size, int flags)
826 {
827         mm_segment_t oldfs = get_fs();
828         int result;
829 
830         set_fs(KERNEL_DS);
831         /*
832          * the following is safe, since for compiler definitions of kvec and
833          * iovec are identical, yielding the same in-core layout and alignment
834          */
835         iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
836         result = sock_recvmsg(sock, msg, size, flags);
837         set_fs(oldfs);
838         return result;
839 }
840 EXPORT_SYMBOL(kernel_recvmsg);
841 
842 static ssize_t sock_sendpage(struct file *file, struct page *page,
843                              int offset, size_t size, loff_t *ppos, int more)
844 {
845         struct socket *sock;
846         int flags;
847 
848         sock = file->private_data;
849 
850         flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
851         /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
852         flags |= more;
853 
854         return kernel_sendpage(sock, page, offset, size, flags);
855 }
856 
857 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
858                                 struct pipe_inode_info *pipe, size_t len,
859                                 unsigned int flags)
860 {
861         struct socket *sock = file->private_data;
862 
863         if (unlikely(!sock->ops->splice_read))
864                 return -EINVAL;
865 
866         return sock->ops->splice_read(sock, ppos, pipe, len, flags);
867 }
868 
869 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
870                                          struct sock_iocb *siocb)
871 {
872         siocb->kiocb = iocb;
873         iocb->private = siocb;
874         return siocb;
875 }
876 
877 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
878                 struct file *file, const struct iovec *iov,
879                 unsigned long nr_segs)
880 {
881         struct socket *sock = file->private_data;
882         size_t size = 0;
883         int i;
884 
885         for (i = 0; i < nr_segs; i++)
886                 size += iov[i].iov_len;
887 
888         msg->msg_name = NULL;
889         msg->msg_namelen = 0;
890         msg->msg_control = NULL;
891         msg->msg_controllen = 0;
892         iov_iter_init(&msg->msg_iter, READ, iov, nr_segs, size);
893         msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
894 
895         return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
896 }
897 
898 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
899                                 unsigned long nr_segs, loff_t pos)
900 {
901         struct sock_iocb siocb, *x;
902 
903         if (pos != 0)
904                 return -ESPIPE;
905 
906         if (iocb->ki_nbytes == 0)       /* Match SYS5 behaviour */
907                 return 0;
908 
909 
910         x = alloc_sock_iocb(iocb, &siocb);
911         if (!x)
912                 return -ENOMEM;
913         return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
914 }
915 
916 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
917                         struct file *file, const struct iovec *iov,
918                         unsigned long nr_segs)
919 {
920         struct socket *sock = file->private_data;
921         size_t size = 0;
922         int i;
923 
924         for (i = 0; i < nr_segs; i++)
925                 size += iov[i].iov_len;
926 
927         msg->msg_name = NULL;
928         msg->msg_namelen = 0;
929         msg->msg_control = NULL;
930         msg->msg_controllen = 0;
931         iov_iter_init(&msg->msg_iter, WRITE, iov, nr_segs, size);
932         msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
933         if (sock->type == SOCK_SEQPACKET)
934                 msg->msg_flags |= MSG_EOR;
935 
936         return __sock_sendmsg(iocb, sock, msg, size);
937 }
938 
939 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
940                           unsigned long nr_segs, loff_t pos)
941 {
942         struct sock_iocb siocb, *x;
943 
944         if (pos != 0)
945                 return -ESPIPE;
946 
947         x = alloc_sock_iocb(iocb, &siocb);
948         if (!x)
949                 return -ENOMEM;
950 
951         return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
952 }
953 
954 /*
955  * Atomic setting of ioctl hooks to avoid race
956  * with module unload.
957  */
958 
959 static DEFINE_MUTEX(br_ioctl_mutex);
960 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
961 
962 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
963 {
964         mutex_lock(&br_ioctl_mutex);
965         br_ioctl_hook = hook;
966         mutex_unlock(&br_ioctl_mutex);
967 }
968 EXPORT_SYMBOL(brioctl_set);
969 
970 static DEFINE_MUTEX(vlan_ioctl_mutex);
971 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
972 
973 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
974 {
975         mutex_lock(&vlan_ioctl_mutex);
976         vlan_ioctl_hook = hook;
977         mutex_unlock(&vlan_ioctl_mutex);
978 }
979 EXPORT_SYMBOL(vlan_ioctl_set);
980 
981 static DEFINE_MUTEX(dlci_ioctl_mutex);
982 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
983 
984 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
985 {
986         mutex_lock(&dlci_ioctl_mutex);
987         dlci_ioctl_hook = hook;
988         mutex_unlock(&dlci_ioctl_mutex);
989 }
990 EXPORT_SYMBOL(dlci_ioctl_set);
991 
992 static long sock_do_ioctl(struct net *net, struct socket *sock,
993                                  unsigned int cmd, unsigned long arg)
994 {
995         int err;
996         void __user *argp = (void __user *)arg;
997 
998         err = sock->ops->ioctl(sock, cmd, arg);
999 
1000         /*
1001          * If this ioctl is unknown try to hand it down
1002          * to the NIC driver.
1003          */
1004         if (err == -ENOIOCTLCMD)
1005                 err = dev_ioctl(net, cmd, argp);
1006 
1007         return err;
1008 }
1009 
1010 /*
1011  *      With an ioctl, arg may well be a user mode pointer, but we don't know
1012  *      what to do with it - that's up to the protocol still.
1013  */
1014 
1015 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1016 {
1017         struct socket *sock;
1018         struct sock *sk;
1019         void __user *argp = (void __user *)arg;
1020         int pid, err;
1021         struct net *net;
1022 
1023         sock = file->private_data;
1024         sk = sock->sk;
1025         net = sock_net(sk);
1026         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1027                 err = dev_ioctl(net, cmd, argp);
1028         } else
1029 #ifdef CONFIG_WEXT_CORE
1030         if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1031                 err = dev_ioctl(net, cmd, argp);
1032         } else
1033 #endif
1034                 switch (cmd) {
1035                 case FIOSETOWN:
1036                 case SIOCSPGRP:
1037                         err = -EFAULT;
1038                         if (get_user(pid, (int __user *)argp))
1039                                 break;
1040                         f_setown(sock->file, pid, 1);
1041                         err = 0;
1042                         break;
1043                 case FIOGETOWN:
1044                 case SIOCGPGRP:
1045                         err = put_user(f_getown(sock->file),
1046                                        (int __user *)argp);
1047                         break;
1048                 case SIOCGIFBR:
1049                 case SIOCSIFBR:
1050                 case SIOCBRADDBR:
1051                 case SIOCBRDELBR:
1052                         err = -ENOPKG;
1053                         if (!br_ioctl_hook)
1054                                 request_module("bridge");
1055 
1056                         mutex_lock(&br_ioctl_mutex);
1057                         if (br_ioctl_hook)
1058                                 err = br_ioctl_hook(net, cmd, argp);
1059                         mutex_unlock(&br_ioctl_mutex);
1060                         break;
1061                 case SIOCGIFVLAN:
1062                 case SIOCSIFVLAN:
1063                         err = -ENOPKG;
1064                         if (!vlan_ioctl_hook)
1065                                 request_module("8021q");
1066 
1067                         mutex_lock(&vlan_ioctl_mutex);
1068                         if (vlan_ioctl_hook)
1069                                 err = vlan_ioctl_hook(net, argp);
1070                         mutex_unlock(&vlan_ioctl_mutex);
1071                         break;
1072                 case SIOCADDDLCI:
1073                 case SIOCDELDLCI:
1074                         err = -ENOPKG;
1075                         if (!dlci_ioctl_hook)
1076                                 request_module("dlci");
1077 
1078                         mutex_lock(&dlci_ioctl_mutex);
1079                         if (dlci_ioctl_hook)
1080                                 err = dlci_ioctl_hook(cmd, argp);
1081                         mutex_unlock(&dlci_ioctl_mutex);
1082                         break;
1083                 default:
1084                         err = sock_do_ioctl(net, sock, cmd, arg);
1085                         break;
1086                 }
1087         return err;
1088 }
1089 
1090 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1091 {
1092         int err;
1093         struct socket *sock = NULL;
1094 
1095         err = security_socket_create(family, type, protocol, 1);
1096         if (err)
1097                 goto out;
1098 
1099         sock = sock_alloc();
1100         if (!sock) {
1101                 err = -ENOMEM;
1102                 goto out;
1103         }
1104 
1105         sock->type = type;
1106         err = security_socket_post_create(sock, family, type, protocol, 1);
1107         if (err)
1108                 goto out_release;
1109 
1110 out:
1111         *res = sock;
1112         return err;
1113 out_release:
1114         sock_release(sock);
1115         sock = NULL;
1116         goto out;
1117 }
1118 EXPORT_SYMBOL(sock_create_lite);
1119 
1120 /* No kernel lock held - perfect */
1121 static unsigned int sock_poll(struct file *file, poll_table *wait)
1122 {
1123         unsigned int busy_flag = 0;
1124         struct socket *sock;
1125 
1126         /*
1127          *      We can't return errors to poll, so it's either yes or no.
1128          */
1129         sock = file->private_data;
1130 
1131         if (sk_can_busy_loop(sock->sk)) {
1132                 /* this socket can poll_ll so tell the system call */
1133                 busy_flag = POLL_BUSY_LOOP;
1134 
1135                 /* once, only if requested by syscall */
1136                 if (wait && (wait->_key & POLL_BUSY_LOOP))
1137                         sk_busy_loop(sock->sk, 1);
1138         }
1139 
1140         return busy_flag | sock->ops->poll(file, sock, wait);
1141 }
1142 
1143 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1144 {
1145         struct socket *sock = file->private_data;
1146 
1147         return sock->ops->mmap(file, sock, vma);
1148 }
1149 
1150 static int sock_close(struct inode *inode, struct file *filp)
1151 {
1152         sock_release(SOCKET_I(inode));
1153         return 0;
1154 }
1155 
1156 /*
1157  *      Update the socket async list
1158  *
1159  *      Fasync_list locking strategy.
1160  *
1161  *      1. fasync_list is modified only under process context socket lock
1162  *         i.e. under semaphore.
1163  *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1164  *         or under socket lock
1165  */
1166 
1167 static int sock_fasync(int fd, struct file *filp, int on)
1168 {
1169         struct socket *sock = filp->private_data;
1170         struct sock *sk = sock->sk;
1171         struct socket_wq *wq;
1172 
1173         if (sk == NULL)
1174                 return -EINVAL;
1175 
1176         lock_sock(sk);
1177         wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1178         fasync_helper(fd, filp, on, &wq->fasync_list);
1179 
1180         if (!wq->fasync_list)
1181                 sock_reset_flag(sk, SOCK_FASYNC);
1182         else
1183                 sock_set_flag(sk, SOCK_FASYNC);
1184 
1185         release_sock(sk);
1186         return 0;
1187 }
1188 
1189 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1190 
1191 int sock_wake_async(struct socket *sock, int how, int band)
1192 {
1193         struct socket_wq *wq;
1194 
1195         if (!sock)
1196                 return -1;
1197         rcu_read_lock();
1198         wq = rcu_dereference(sock->wq);
1199         if (!wq || !wq->fasync_list) {
1200                 rcu_read_unlock();
1201                 return -1;
1202         }
1203         switch (how) {
1204         case SOCK_WAKE_WAITD:
1205                 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1206                         break;
1207                 goto call_kill;
1208         case SOCK_WAKE_SPACE:
1209                 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1210                         break;
1211                 /* fall through */
1212         case SOCK_WAKE_IO:
1213 call_kill:
1214                 kill_fasync(&wq->fasync_list, SIGIO, band);
1215                 break;
1216         case SOCK_WAKE_URG:
1217                 kill_fasync(&wq->fasync_list, SIGURG, band);
1218         }
1219         rcu_read_unlock();
1220         return 0;
1221 }
1222 EXPORT_SYMBOL(sock_wake_async);
1223 
1224 int __sock_create(struct net *net, int family, int type, int protocol,
1225                          struct socket **res, int kern)
1226 {
1227         int err;
1228         struct socket *sock;
1229         const struct net_proto_family *pf;
1230 
1231         /*
1232          *      Check protocol is in range
1233          */
1234         if (family < 0 || family >= NPROTO)
1235                 return -EAFNOSUPPORT;
1236         if (type < 0 || type >= SOCK_MAX)
1237                 return -EINVAL;
1238 
1239         /* Compatibility.
1240 
1241            This uglymoron is moved from INET layer to here to avoid
1242            deadlock in module load.
1243          */
1244         if (family == PF_INET && type == SOCK_PACKET) {
1245                 static int warned;
1246                 if (!warned) {
1247                         warned = 1;
1248                         pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1249                                 current->comm);
1250                 }
1251                 family = PF_PACKET;
1252         }
1253 
1254         err = security_socket_create(family, type, protocol, kern);
1255         if (err)
1256                 return err;
1257 
1258         /*
1259          *      Allocate the socket and allow the family to set things up. if
1260          *      the protocol is 0, the family is instructed to select an appropriate
1261          *      default.
1262          */
1263         sock = sock_alloc();
1264         if (!sock) {
1265                 net_warn_ratelimited("socket: no more sockets\n");
1266                 return -ENFILE; /* Not exactly a match, but its the
1267                                    closest posix thing */
1268         }
1269 
1270         sock->type = type;
1271 
1272 #ifdef CONFIG_MODULES
1273         /* Attempt to load a protocol module if the find failed.
1274          *
1275          * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1276          * requested real, full-featured networking support upon configuration.
1277          * Otherwise module support will break!
1278          */
1279         if (rcu_access_pointer(net_families[family]) == NULL)
1280                 request_module("net-pf-%d", family);
1281 #endif
1282 
1283         rcu_read_lock();
1284         pf = rcu_dereference(net_families[family]);
1285         err = -EAFNOSUPPORT;
1286         if (!pf)
1287                 goto out_release;
1288 
1289         /*
1290          * We will call the ->create function, that possibly is in a loadable
1291          * module, so we have to bump that loadable module refcnt first.
1292          */
1293         if (!try_module_get(pf->owner))
1294                 goto out_release;
1295 
1296         /* Now protected by module ref count */
1297         rcu_read_unlock();
1298 
1299         err = pf->create(net, sock, protocol, kern);
1300         if (err < 0)
1301                 goto out_module_put;
1302 
1303         /*
1304          * Now to bump the refcnt of the [loadable] module that owns this
1305          * socket at sock_release time we decrement its refcnt.
1306          */
1307         if (!try_module_get(sock->ops->owner))
1308                 goto out_module_busy;
1309 
1310         /*
1311          * Now that we're done with the ->create function, the [loadable]
1312          * module can have its refcnt decremented
1313          */
1314         module_put(pf->owner);
1315         err = security_socket_post_create(sock, family, type, protocol, kern);
1316         if (err)
1317                 goto out_sock_release;
1318         *res = sock;
1319 
1320         return 0;
1321 
1322 out_module_busy:
1323         err = -EAFNOSUPPORT;
1324 out_module_put:
1325         sock->ops = NULL;
1326         module_put(pf->owner);
1327 out_sock_release:
1328         sock_release(sock);
1329         return err;
1330 
1331 out_release:
1332         rcu_read_unlock();
1333         goto out_sock_release;
1334 }
1335 EXPORT_SYMBOL(__sock_create);
1336 
1337 int sock_create(int family, int type, int protocol, struct socket **res)
1338 {
1339         return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1340 }
1341 EXPORT_SYMBOL(sock_create);
1342 
1343 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1344 {
1345         return __sock_create(&init_net, family, type, protocol, res, 1);
1346 }
1347 EXPORT_SYMBOL(sock_create_kern);
1348 
1349 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1350 {
1351         int retval;
1352         struct socket *sock;
1353         int flags;
1354 
1355         /* Check the SOCK_* constants for consistency.  */
1356         BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1357         BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1358         BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1359         BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1360 
1361         flags = type & ~SOCK_TYPE_MASK;
1362         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1363                 return -EINVAL;
1364         type &= SOCK_TYPE_MASK;
1365 
1366         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1367                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1368 
1369         retval = sock_create(family, type, protocol, &sock);
1370         if (retval < 0)
1371                 goto out;
1372 
1373         retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1374         if (retval < 0)
1375                 goto out_release;
1376 
1377 out:
1378         /* It may be already another descriptor 8) Not kernel problem. */
1379         return retval;
1380 
1381 out_release:
1382         sock_release(sock);
1383         return retval;
1384 }
1385 
1386 /*
1387  *      Create a pair of connected sockets.
1388  */
1389 
1390 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1391                 int __user *, usockvec)
1392 {
1393         struct socket *sock1, *sock2;
1394         int fd1, fd2, err;
1395         struct file *newfile1, *newfile2;
1396         int flags;
1397 
1398         flags = type & ~SOCK_TYPE_MASK;
1399         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1400                 return -EINVAL;
1401         type &= SOCK_TYPE_MASK;
1402 
1403         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1404                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1405 
1406         /*
1407          * Obtain the first socket and check if the underlying protocol
1408          * supports the socketpair call.
1409          */
1410 
1411         err = sock_create(family, type, protocol, &sock1);
1412         if (err < 0)
1413                 goto out;
1414 
1415         err = sock_create(family, type, protocol, &sock2);
1416         if (err < 0)
1417                 goto out_release_1;
1418 
1419         err = sock1->ops->socketpair(sock1, sock2);
1420         if (err < 0)
1421                 goto out_release_both;
1422 
1423         fd1 = get_unused_fd_flags(flags);
1424         if (unlikely(fd1 < 0)) {
1425                 err = fd1;
1426                 goto out_release_both;
1427         }
1428 
1429         fd2 = get_unused_fd_flags(flags);
1430         if (unlikely(fd2 < 0)) {
1431                 err = fd2;
1432                 goto out_put_unused_1;
1433         }
1434 
1435         newfile1 = sock_alloc_file(sock1, flags, NULL);
1436         if (unlikely(IS_ERR(newfile1))) {
1437                 err = PTR_ERR(newfile1);
1438                 goto out_put_unused_both;
1439         }
1440 
1441         newfile2 = sock_alloc_file(sock2, flags, NULL);
1442         if (IS_ERR(newfile2)) {
1443                 err = PTR_ERR(newfile2);
1444                 goto out_fput_1;
1445         }
1446 
1447         err = put_user(fd1, &usockvec[0]);
1448         if (err)
1449                 goto out_fput_both;
1450 
1451         err = put_user(fd2, &usockvec[1]);
1452         if (err)
1453                 goto out_fput_both;
1454 
1455         audit_fd_pair(fd1, fd2);
1456 
1457         fd_install(fd1, newfile1);
1458         fd_install(fd2, newfile2);
1459         /* fd1 and fd2 may be already another descriptors.
1460          * Not kernel problem.
1461          */
1462 
1463         return 0;
1464 
1465 out_fput_both:
1466         fput(newfile2);
1467         fput(newfile1);
1468         put_unused_fd(fd2);
1469         put_unused_fd(fd1);
1470         goto out;
1471 
1472 out_fput_1:
1473         fput(newfile1);
1474         put_unused_fd(fd2);
1475         put_unused_fd(fd1);
1476         sock_release(sock2);
1477         goto out;
1478 
1479 out_put_unused_both:
1480         put_unused_fd(fd2);
1481 out_put_unused_1:
1482         put_unused_fd(fd1);
1483 out_release_both:
1484         sock_release(sock2);
1485 out_release_1:
1486         sock_release(sock1);
1487 out:
1488         return err;
1489 }
1490 
1491 /*
1492  *      Bind a name to a socket. Nothing much to do here since it's
1493  *      the protocol's responsibility to handle the local address.
1494  *
1495  *      We move the socket address to kernel space before we call
1496  *      the protocol layer (having also checked the address is ok).
1497  */
1498 
1499 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1500 {
1501         struct socket *sock;
1502         struct sockaddr_storage address;
1503         int err, fput_needed;
1504 
1505         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1506         if (sock) {
1507                 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1508                 if (err >= 0) {
1509                         err = security_socket_bind(sock,
1510                                                    (struct sockaddr *)&address,
1511                                                    addrlen);
1512                         if (!err)
1513                                 err = sock->ops->bind(sock,
1514                                                       (struct sockaddr *)
1515                                                       &address, addrlen);
1516                 }
1517                 fput_light(sock->file, fput_needed);
1518         }
1519         return err;
1520 }
1521 
1522 /*
1523  *      Perform a listen. Basically, we allow the protocol to do anything
1524  *      necessary for a listen, and if that works, we mark the socket as
1525  *      ready for listening.
1526  */
1527 
1528 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1529 {
1530         struct socket *sock;
1531         int err, fput_needed;
1532         int somaxconn;
1533 
1534         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1535         if (sock) {
1536                 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1537                 if ((unsigned int)backlog > somaxconn)
1538                         backlog = somaxconn;
1539 
1540                 err = security_socket_listen(sock, backlog);
1541                 if (!err)
1542                         err = sock->ops->listen(sock, backlog);
1543 
1544                 fput_light(sock->file, fput_needed);
1545         }
1546         return err;
1547 }
1548 
1549 /*
1550  *      For accept, we attempt to create a new socket, set up the link
1551  *      with the client, wake up the client, then return the new
1552  *      connected fd. We collect the address of the connector in kernel
1553  *      space and move it to user at the very end. This is unclean because
1554  *      we open the socket then return an error.
1555  *
1556  *      1003.1g adds the ability to recvmsg() to query connection pending
1557  *      status to recvmsg. We need to add that support in a way thats
1558  *      clean when we restucture accept also.
1559  */
1560 
1561 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1562                 int __user *, upeer_addrlen, int, flags)
1563 {
1564         struct socket *sock, *newsock;
1565         struct file *newfile;
1566         int err, len, newfd, fput_needed;
1567         struct sockaddr_storage address;
1568 
1569         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1570                 return -EINVAL;
1571 
1572         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1573                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1574 
1575         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1576         if (!sock)
1577                 goto out;
1578 
1579         err = -ENFILE;
1580         newsock = sock_alloc();
1581         if (!newsock)
1582                 goto out_put;
1583 
1584         newsock->type = sock->type;
1585         newsock->ops = sock->ops;
1586 
1587         /*
1588          * We don't need try_module_get here, as the listening socket (sock)
1589          * has the protocol module (sock->ops->owner) held.
1590          */
1591         __module_get(newsock->ops->owner);
1592 
1593         newfd = get_unused_fd_flags(flags);
1594         if (unlikely(newfd < 0)) {
1595                 err = newfd;
1596                 sock_release(newsock);
1597                 goto out_put;
1598         }
1599         newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1600         if (unlikely(IS_ERR(newfile))) {
1601                 err = PTR_ERR(newfile);
1602                 put_unused_fd(newfd);
1603                 sock_release(newsock);
1604                 goto out_put;
1605         }
1606 
1607         err = security_socket_accept(sock, newsock);
1608         if (err)
1609                 goto out_fd;
1610 
1611         err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1612         if (err < 0)
1613                 goto out_fd;
1614 
1615         if (upeer_sockaddr) {
1616                 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1617                                           &len, 2) < 0) {
1618                         err = -ECONNABORTED;
1619                         goto out_fd;
1620                 }
1621                 err = move_addr_to_user(&address,
1622                                         len, upeer_sockaddr, upeer_addrlen);
1623                 if (err < 0)
1624                         goto out_fd;
1625         }
1626 
1627         /* File flags are not inherited via accept() unlike another OSes. */
1628 
1629         fd_install(newfd, newfile);
1630         err = newfd;
1631 
1632 out_put:
1633         fput_light(sock->file, fput_needed);
1634 out:
1635         return err;
1636 out_fd:
1637         fput(newfile);
1638         put_unused_fd(newfd);
1639         goto out_put;
1640 }
1641 
1642 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1643                 int __user *, upeer_addrlen)
1644 {
1645         return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1646 }
1647 
1648 /*
1649  *      Attempt to connect to a socket with the server address.  The address
1650  *      is in user space so we verify it is OK and move it to kernel space.
1651  *
1652  *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1653  *      break bindings
1654  *
1655  *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1656  *      other SEQPACKET protocols that take time to connect() as it doesn't
1657  *      include the -EINPROGRESS status for such sockets.
1658  */
1659 
1660 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1661                 int, addrlen)
1662 {
1663         struct socket *sock;
1664         struct sockaddr_storage address;
1665         int err, fput_needed;
1666 
1667         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1668         if (!sock)
1669                 goto out;
1670         err = move_addr_to_kernel(uservaddr, addrlen, &address);
1671         if (err < 0)
1672                 goto out_put;
1673 
1674         err =
1675             security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1676         if (err)
1677                 goto out_put;
1678 
1679         err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1680                                  sock->file->f_flags);
1681 out_put:
1682         fput_light(sock->file, fput_needed);
1683 out:
1684         return err;
1685 }
1686 
1687 /*
1688  *      Get the local address ('name') of a socket object. Move the obtained
1689  *      name to user space.
1690  */
1691 
1692 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1693                 int __user *, usockaddr_len)
1694 {
1695         struct socket *sock;
1696         struct sockaddr_storage address;
1697         int len, err, fput_needed;
1698 
1699         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1700         if (!sock)
1701                 goto out;
1702 
1703         err = security_socket_getsockname(sock);
1704         if (err)
1705                 goto out_put;
1706 
1707         err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1708         if (err)
1709                 goto out_put;
1710         err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1711 
1712 out_put:
1713         fput_light(sock->file, fput_needed);
1714 out:
1715         return err;
1716 }
1717 
1718 /*
1719  *      Get the remote address ('name') of a socket object. Move the obtained
1720  *      name to user space.
1721  */
1722 
1723 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1724                 int __user *, usockaddr_len)
1725 {
1726         struct socket *sock;
1727         struct sockaddr_storage address;
1728         int len, err, fput_needed;
1729 
1730         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1731         if (sock != NULL) {
1732                 err = security_socket_getpeername(sock);
1733                 if (err) {
1734                         fput_light(sock->file, fput_needed);
1735                         return err;
1736                 }
1737 
1738                 err =
1739                     sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1740                                        1);
1741                 if (!err)
1742                         err = move_addr_to_user(&address, len, usockaddr,
1743                                                 usockaddr_len);
1744                 fput_light(sock->file, fput_needed);
1745         }
1746         return err;
1747 }
1748 
1749 /*
1750  *      Send a datagram to a given address. We move the address into kernel
1751  *      space and check the user space data area is readable before invoking
1752  *      the protocol.
1753  */
1754 
1755 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1756                 unsigned int, flags, struct sockaddr __user *, addr,
1757                 int, addr_len)
1758 {
1759         struct socket *sock;
1760         struct sockaddr_storage address;
1761         int err;
1762         struct msghdr msg;
1763         struct iovec iov;
1764         int fput_needed;
1765 
1766         if (len > INT_MAX)
1767                 len = INT_MAX;
1768         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769         if (!sock)
1770                 goto out;
1771 
1772         iov.iov_base = buff;
1773         iov.iov_len = len;
1774         msg.msg_name = NULL;
1775         iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
1776         msg.msg_control = NULL;
1777         msg.msg_controllen = 0;
1778         msg.msg_namelen = 0;
1779         if (addr) {
1780                 err = move_addr_to_kernel(addr, addr_len, &address);
1781                 if (err < 0)
1782                         goto out_put;
1783                 msg.msg_name = (struct sockaddr *)&address;
1784                 msg.msg_namelen = addr_len;
1785         }
1786         if (sock->file->f_flags & O_NONBLOCK)
1787                 flags |= MSG_DONTWAIT;
1788         msg.msg_flags = flags;
1789         err = sock_sendmsg(sock, &msg, len);
1790 
1791 out_put:
1792         fput_light(sock->file, fput_needed);
1793 out:
1794         return err;
1795 }
1796 
1797 /*
1798  *      Send a datagram down a socket.
1799  */
1800 
1801 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1802                 unsigned int, flags)
1803 {
1804         return sys_sendto(fd, buff, len, flags, NULL, 0);
1805 }
1806 
1807 /*
1808  *      Receive a frame from the socket and optionally record the address of the
1809  *      sender. We verify the buffers are writable and if needed move the
1810  *      sender address from kernel to user space.
1811  */
1812 
1813 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1814                 unsigned int, flags, struct sockaddr __user *, addr,
1815                 int __user *, addr_len)
1816 {
1817         struct socket *sock;
1818         struct iovec iov;
1819         struct msghdr msg;
1820         struct sockaddr_storage address;
1821         int err, err2;
1822         int fput_needed;
1823 
1824         if (size > INT_MAX)
1825                 size = INT_MAX;
1826         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1827         if (!sock)
1828                 goto out;
1829 
1830         msg.msg_control = NULL;
1831         msg.msg_controllen = 0;
1832         iov.iov_len = size;
1833         iov.iov_base = ubuf;
1834         iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
1835         /* Save some cycles and don't copy the address if not needed */
1836         msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1837         /* We assume all kernel code knows the size of sockaddr_storage */
1838         msg.msg_namelen = 0;
1839         if (sock->file->f_flags & O_NONBLOCK)
1840                 flags |= MSG_DONTWAIT;
1841         err = sock_recvmsg(sock, &msg, size, flags);
1842 
1843         if (err >= 0 && addr != NULL) {
1844                 err2 = move_addr_to_user(&address,
1845                                          msg.msg_namelen, addr, addr_len);
1846                 if (err2 < 0)
1847                         err = err2;
1848         }
1849 
1850         fput_light(sock->file, fput_needed);
1851 out:
1852         return err;
1853 }
1854 
1855 /*
1856  *      Receive a datagram from a socket.
1857  */
1858 
1859 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1860                 unsigned int, flags)
1861 {
1862         return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1863 }
1864 
1865 /*
1866  *      Set a socket option. Because we don't know the option lengths we have
1867  *      to pass the user mode parameter for the protocols to sort out.
1868  */
1869 
1870 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1871                 char __user *, optval, int, optlen)
1872 {
1873         int err, fput_needed;
1874         struct socket *sock;
1875 
1876         if (optlen < 0)
1877                 return -EINVAL;
1878 
1879         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1880         if (sock != NULL) {
1881                 err = security_socket_setsockopt(sock, level, optname);
1882                 if (err)
1883                         goto out_put;
1884 
1885                 if (level == SOL_SOCKET)
1886                         err =
1887                             sock_setsockopt(sock, level, optname, optval,
1888                                             optlen);
1889                 else
1890                         err =
1891                             sock->ops->setsockopt(sock, level, optname, optval,
1892                                                   optlen);
1893 out_put:
1894                 fput_light(sock->file, fput_needed);
1895         }
1896         return err;
1897 }
1898 
1899 /*
1900  *      Get a socket option. Because we don't know the option lengths we have
1901  *      to pass a user mode parameter for the protocols to sort out.
1902  */
1903 
1904 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1905                 char __user *, optval, int __user *, optlen)
1906 {
1907         int err, fput_needed;
1908         struct socket *sock;
1909 
1910         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1911         if (sock != NULL) {
1912                 err = security_socket_getsockopt(sock, level, optname);
1913                 if (err)
1914                         goto out_put;
1915 
1916                 if (level == SOL_SOCKET)
1917                         err =
1918                             sock_getsockopt(sock, level, optname, optval,
1919                                             optlen);
1920                 else
1921                         err =
1922                             sock->ops->getsockopt(sock, level, optname, optval,
1923                                                   optlen);
1924 out_put:
1925                 fput_light(sock->file, fput_needed);
1926         }
1927         return err;
1928 }
1929 
1930 /*
1931  *      Shutdown a socket.
1932  */
1933 
1934 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1935 {
1936         int err, fput_needed;
1937         struct socket *sock;
1938 
1939         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1940         if (sock != NULL) {
1941                 err = security_socket_shutdown(sock, how);
1942                 if (!err)
1943                         err = sock->ops->shutdown(sock, how);
1944                 fput_light(sock->file, fput_needed);
1945         }
1946         return err;
1947 }
1948 
1949 /* A couple of helpful macros for getting the address of the 32/64 bit
1950  * fields which are the same type (int / unsigned) on our platforms.
1951  */
1952 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1953 #define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1954 #define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1955 
1956 struct used_address {
1957         struct sockaddr_storage name;
1958         unsigned int name_len;
1959 };
1960 
1961 static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1962                                      struct user_msghdr __user *umsg,
1963                                      struct sockaddr __user **save_addr,
1964                                      struct iovec **iov)
1965 {
1966         struct sockaddr __user *uaddr;
1967         struct iovec __user *uiov;
1968         size_t nr_segs;
1969         ssize_t err;
1970 
1971         if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1972             __get_user(uaddr, &umsg->msg_name) ||
1973             __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1974             __get_user(uiov, &umsg->msg_iov) ||
1975             __get_user(nr_segs, &umsg->msg_iovlen) ||
1976             __get_user(kmsg->msg_control, &umsg->msg_control) ||
1977             __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1978             __get_user(kmsg->msg_flags, &umsg->msg_flags))
1979                 return -EFAULT;
1980 
1981         if (!uaddr)
1982                 kmsg->msg_namelen = 0;
1983 
1984         if (kmsg->msg_namelen < 0)
1985                 return -EINVAL;
1986 
1987         if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1988                 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1989 
1990         if (save_addr)
1991                 *save_addr = uaddr;
1992 
1993         if (uaddr && kmsg->msg_namelen) {
1994                 if (!save_addr) {
1995                         err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1996                                                   kmsg->msg_name);
1997                         if (err < 0)
1998                                 return err;
1999                 }
2000         } else {
2001                 kmsg->msg_name = NULL;
2002                 kmsg->msg_namelen = 0;
2003         }
2004 
2005         if (nr_segs > UIO_MAXIOV)
2006                 return -EMSGSIZE;
2007 
2008         err = rw_copy_check_uvector(save_addr ? READ : WRITE,
2009                                     uiov, nr_segs,
2010                                     UIO_FASTIOV, *iov, iov);
2011         if (err >= 0)
2012                 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
2013                               *iov, nr_segs, err);
2014         return err;
2015 }
2016 
2017 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2018                          struct msghdr *msg_sys, unsigned int flags,
2019                          struct used_address *used_address)
2020 {
2021         struct compat_msghdr __user *msg_compat =
2022             (struct compat_msghdr __user *)msg;
2023         struct sockaddr_storage address;
2024         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2025         unsigned char ctl[sizeof(struct cmsghdr) + 20]
2026             __attribute__ ((aligned(sizeof(__kernel_size_t))));
2027         /* 20 is size of ipv6_pktinfo */
2028         unsigned char *ctl_buf = ctl;
2029         int ctl_len, total_len;
2030         ssize_t err;
2031 
2032         msg_sys->msg_name = &address;
2033 
2034         if (MSG_CMSG_COMPAT & flags)
2035                 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2036         else
2037                 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2038         if (err < 0)
2039                 goto out_freeiov;
2040         total_len = err;
2041 
2042         err = -ENOBUFS;
2043 
2044         if (msg_sys->msg_controllen > INT_MAX)
2045                 goto out_freeiov;
2046         ctl_len = msg_sys->msg_controllen;
2047         if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2048                 err =
2049                     cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2050                                                      sizeof(ctl));
2051                 if (err)
2052                         goto out_freeiov;
2053                 ctl_buf = msg_sys->msg_control;
2054                 ctl_len = msg_sys->msg_controllen;
2055         } else if (ctl_len) {
2056                 if (ctl_len > sizeof(ctl)) {
2057                         ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2058                         if (ctl_buf == NULL)
2059                                 goto out_freeiov;
2060                 }
2061                 err = -EFAULT;
2062                 /*
2063                  * Careful! Before this, msg_sys->msg_control contains a user pointer.
2064                  * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2065                  * checking falls down on this.
2066                  */
2067                 if (copy_from_user(ctl_buf,
2068                                    (void __user __force *)msg_sys->msg_control,
2069                                    ctl_len))
2070                         goto out_freectl;
2071                 msg_sys->msg_control = ctl_buf;
2072         }
2073         msg_sys->msg_flags = flags;
2074 
2075         if (sock->file->f_flags & O_NONBLOCK)
2076                 msg_sys->msg_flags |= MSG_DONTWAIT;
2077         /*
2078          * If this is sendmmsg() and current destination address is same as
2079          * previously succeeded address, omit asking LSM's decision.
2080          * used_address->name_len is initialized to UINT_MAX so that the first
2081          * destination address never matches.
2082          */
2083         if (used_address && msg_sys->msg_name &&
2084             used_address->name_len == msg_sys->msg_namelen &&
2085             !memcmp(&used_address->name, msg_sys->msg_name,
2086                     used_address->name_len)) {
2087                 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2088                 goto out_freectl;
2089         }
2090         err = sock_sendmsg(sock, msg_sys, total_len);
2091         /*
2092          * If this is sendmmsg() and sending to current destination address was
2093          * successful, remember it.
2094          */
2095         if (used_address && err >= 0) {
2096                 used_address->name_len = msg_sys->msg_namelen;
2097                 if (msg_sys->msg_name)
2098                         memcpy(&used_address->name, msg_sys->msg_name,
2099                                used_address->name_len);
2100         }
2101 
2102 out_freectl:
2103         if (ctl_buf != ctl)
2104                 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2105 out_freeiov:
2106         if (iov != iovstack)
2107                 kfree(iov);
2108         return err;
2109 }
2110 
2111 /*
2112  *      BSD sendmsg interface
2113  */
2114 
2115 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2116 {
2117         int fput_needed, err;
2118         struct msghdr msg_sys;
2119         struct socket *sock;
2120 
2121         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2122         if (!sock)
2123                 goto out;
2124 
2125         err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2126 
2127         fput_light(sock->file, fput_needed);
2128 out:
2129         return err;
2130 }
2131 
2132 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2133 {
2134         if (flags & MSG_CMSG_COMPAT)
2135                 return -EINVAL;
2136         return __sys_sendmsg(fd, msg, flags);
2137 }
2138 
2139 /*
2140  *      Linux sendmmsg interface
2141  */
2142 
2143 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2144                    unsigned int flags)
2145 {
2146         int fput_needed, err, datagrams;
2147         struct socket *sock;
2148         struct mmsghdr __user *entry;
2149         struct compat_mmsghdr __user *compat_entry;
2150         struct msghdr msg_sys;
2151         struct used_address used_address;
2152 
2153         if (vlen > UIO_MAXIOV)
2154                 vlen = UIO_MAXIOV;
2155 
2156         datagrams = 0;
2157 
2158         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2159         if (!sock)
2160                 return err;
2161 
2162         used_address.name_len = UINT_MAX;
2163         entry = mmsg;
2164         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2165         err = 0;
2166 
2167         while (datagrams < vlen) {
2168                 if (MSG_CMSG_COMPAT & flags) {
2169                         err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2170                                              &msg_sys, flags, &used_address);
2171                         if (err < 0)
2172                                 break;
2173                         err = __put_user(err, &compat_entry->msg_len);
2174                         ++compat_entry;
2175                 } else {
2176                         err = ___sys_sendmsg(sock,
2177                                              (struct user_msghdr __user *)entry,
2178                                              &msg_sys, flags, &used_address);
2179                         if (err < 0)
2180                                 break;
2181                         err = put_user(err, &entry->msg_len);
2182                         ++entry;
2183                 }
2184 
2185                 if (err)
2186                         break;
2187                 ++datagrams;
2188         }
2189 
2190         fput_light(sock->file, fput_needed);
2191 
2192         /* We only return an error if no datagrams were able to be sent */
2193         if (datagrams != 0)
2194                 return datagrams;
2195 
2196         return err;
2197 }
2198 
2199 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2200                 unsigned int, vlen, unsigned int, flags)
2201 {
2202         if (flags & MSG_CMSG_COMPAT)
2203                 return -EINVAL;
2204         return __sys_sendmmsg(fd, mmsg, vlen, flags);
2205 }
2206 
2207 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2208                          struct msghdr *msg_sys, unsigned int flags, int nosec)
2209 {
2210         struct compat_msghdr __user *msg_compat =
2211             (struct compat_msghdr __user *)msg;
2212         struct iovec iovstack[UIO_FASTIOV];
2213         struct iovec *iov = iovstack;
2214         unsigned long cmsg_ptr;
2215         int total_len, len;
2216         ssize_t err;
2217 
2218         /* kernel mode address */
2219         struct sockaddr_storage addr;
2220 
2221         /* user mode address pointers */
2222         struct sockaddr __user *uaddr;
2223         int __user *uaddr_len = COMPAT_NAMELEN(msg);
2224 
2225         msg_sys->msg_name = &addr;
2226 
2227         if (MSG_CMSG_COMPAT & flags)
2228                 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2229         else
2230                 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2231         if (err < 0)
2232                 goto out_freeiov;
2233         total_len = err;
2234 
2235         cmsg_ptr = (unsigned long)msg_sys->msg_control;
2236         msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2237 
2238         /* We assume all kernel code knows the size of sockaddr_storage */
2239         msg_sys->msg_namelen = 0;
2240 
2241         if (sock->file->f_flags & O_NONBLOCK)
2242                 flags |= MSG_DONTWAIT;
2243         err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2244                                                           total_len, flags);
2245         if (err < 0)
2246                 goto out_freeiov;
2247         len = err;
2248 
2249         if (uaddr != NULL) {
2250                 err = move_addr_to_user(&addr,
2251                                         msg_sys->msg_namelen, uaddr,
2252                                         uaddr_len);
2253                 if (err < 0)
2254                         goto out_freeiov;
2255         }
2256         err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2257                          COMPAT_FLAGS(msg));
2258         if (err)
2259                 goto out_freeiov;
2260         if (MSG_CMSG_COMPAT & flags)
2261                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2262                                  &msg_compat->msg_controllen);
2263         else
2264                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2265                                  &msg->msg_controllen);
2266         if (err)
2267                 goto out_freeiov;
2268         err = len;
2269 
2270 out_freeiov:
2271         if (iov != iovstack)
2272                 kfree(iov);
2273         return err;
2274 }
2275 
2276 /*
2277  *      BSD recvmsg interface
2278  */
2279 
2280 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2281 {
2282         int fput_needed, err;
2283         struct msghdr msg_sys;
2284         struct socket *sock;
2285 
2286         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2287         if (!sock)
2288                 goto out;
2289 
2290         err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2291 
2292         fput_light(sock->file, fput_needed);
2293 out:
2294         return err;
2295 }
2296 
2297 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2298                 unsigned int, flags)
2299 {
2300         if (flags & MSG_CMSG_COMPAT)
2301                 return -EINVAL;
2302         return __sys_recvmsg(fd, msg, flags);
2303 }
2304 
2305 /*
2306  *     Linux recvmmsg interface
2307  */
2308 
2309 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2310                    unsigned int flags, struct timespec *timeout)
2311 {
2312         int fput_needed, err, datagrams;
2313         struct socket *sock;
2314         struct mmsghdr __user *entry;
2315         struct compat_mmsghdr __user *compat_entry;
2316         struct msghdr msg_sys;
2317         struct timespec end_time;
2318 
2319         if (timeout &&
2320             poll_select_set_timeout(&end_time, timeout->tv_sec,
2321                                     timeout->tv_nsec))
2322                 return -EINVAL;
2323 
2324         datagrams = 0;
2325 
2326         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2327         if (!sock)
2328                 return err;
2329 
2330         err = sock_error(sock->sk);
2331         if (err)
2332                 goto out_put;
2333 
2334         entry = mmsg;
2335         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2336 
2337         while (datagrams < vlen) {
2338                 /*
2339                  * No need to ask LSM for more than the first datagram.
2340                  */
2341                 if (MSG_CMSG_COMPAT & flags) {
2342                         err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2343                                              &msg_sys, flags & ~MSG_WAITFORONE,
2344                                              datagrams);
2345                         if (err < 0)
2346                                 break;
2347                         err = __put_user(err, &compat_entry->msg_len);
2348                         ++compat_entry;
2349                 } else {
2350                         err = ___sys_recvmsg(sock,
2351                                              (struct user_msghdr __user *)entry,
2352                                              &msg_sys, flags & ~MSG_WAITFORONE,
2353                                              datagrams);
2354                         if (err < 0)
2355                                 break;
2356                         err = put_user(err, &entry->msg_len);
2357                         ++entry;
2358                 }
2359 
2360                 if (err)
2361                         break;
2362                 ++datagrams;
2363 
2364                 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2365                 if (flags & MSG_WAITFORONE)
2366                         flags |= MSG_DONTWAIT;
2367 
2368                 if (timeout) {
2369                         ktime_get_ts(timeout);
2370                         *timeout = timespec_sub(end_time, *timeout);
2371                         if (timeout->tv_sec < 0) {
2372                                 timeout->tv_sec = timeout->tv_nsec = 0;
2373                                 break;
2374                         }
2375 
2376                         /* Timeout, return less than vlen datagrams */
2377                         if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2378                                 break;
2379                 }
2380 
2381                 /* Out of band data, return right away */
2382                 if (msg_sys.msg_flags & MSG_OOB)
2383                         break;
2384         }
2385 
2386 out_put:
2387         fput_light(sock->file, fput_needed);
2388 
2389         if (err == 0)
2390                 return datagrams;
2391 
2392         if (datagrams != 0) {
2393                 /*
2394                  * We may return less entries than requested (vlen) if the
2395                  * sock is non block and there aren't enough datagrams...
2396                  */
2397                 if (err != -EAGAIN) {
2398                         /*
2399                          * ... or  if recvmsg returns an error after we
2400                          * received some datagrams, where we record the
2401                          * error to return on the next call or if the
2402                          * app asks about it using getsockopt(SO_ERROR).
2403                          */
2404                         sock->sk->sk_err = -err;
2405                 }
2406 
2407                 return datagrams;
2408         }
2409 
2410         return err;
2411 }
2412 
2413 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2414                 unsigned int, vlen, unsigned int, flags,
2415                 struct timespec __user *, timeout)
2416 {
2417         int datagrams;
2418         struct timespec timeout_sys;
2419 
2420         if (flags & MSG_CMSG_COMPAT)
2421                 return -EINVAL;
2422 
2423         if (!timeout)
2424                 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2425 
2426         if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2427                 return -EFAULT;
2428 
2429         datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2430 
2431         if (datagrams > 0 &&
2432             copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2433                 datagrams = -EFAULT;
2434 
2435         return datagrams;
2436 }
2437 
2438 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2439 /* Argument list sizes for sys_socketcall */
2440 #define AL(x) ((x) * sizeof(unsigned long))
2441 static const unsigned char nargs[21] = {
2442         AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2443         AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2444         AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2445         AL(4), AL(5), AL(4)
2446 };
2447 
2448 #undef AL
2449 
2450 /*
2451  *      System call vectors.
2452  *
2453  *      Argument checking cleaned up. Saved 20% in size.
2454  *  This function doesn't need to set the kernel lock because
2455  *  it is set by the callees.
2456  */
2457 
2458 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2459 {
2460         unsigned long a[AUDITSC_ARGS];
2461         unsigned long a0, a1;
2462         int err;
2463         unsigned int len;
2464 
2465         if (call < 1 || call > SYS_SENDMMSG)
2466                 return -EINVAL;
2467 
2468         len = nargs[call];
2469         if (len > sizeof(a))
2470                 return -EINVAL;
2471 
2472         /* copy_from_user should be SMP safe. */
2473         if (copy_from_user(a, args, len))
2474                 return -EFAULT;
2475 
2476         err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2477         if (err)
2478                 return err;
2479 
2480         a0 = a[0];
2481         a1 = a[1];
2482 
2483         switch (call) {
2484         case SYS_SOCKET:
2485                 err = sys_socket(a0, a1, a[2]);
2486                 break;
2487         case SYS_BIND:
2488                 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2489                 break;
2490         case SYS_CONNECT:
2491                 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2492                 break;
2493         case SYS_LISTEN:
2494                 err = sys_listen(a0, a1);
2495                 break;
2496         case SYS_ACCEPT:
2497                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2498                                   (int __user *)a[2], 0);
2499                 break;
2500         case SYS_GETSOCKNAME:
2501                 err =
2502                     sys_getsockname(a0, (struct sockaddr __user *)a1,
2503                                     (int __user *)a[2]);
2504                 break;
2505         case SYS_GETPEERNAME:
2506                 err =
2507                     sys_getpeername(a0, (struct sockaddr __user *)a1,
2508                                     (int __user *)a[2]);
2509                 break;
2510         case SYS_SOCKETPAIR:
2511                 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2512                 break;
2513         case SYS_SEND:
2514                 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2515                 break;
2516         case SYS_SENDTO:
2517                 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2518                                  (struct sockaddr __user *)a[4], a[5]);
2519                 break;
2520         case SYS_RECV:
2521                 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2522                 break;
2523         case SYS_RECVFROM:
2524                 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2525                                    (struct sockaddr __user *)a[4],
2526                                    (int __user *)a[5]);
2527                 break;
2528         case SYS_SHUTDOWN:
2529                 err = sys_shutdown(a0, a1);
2530                 break;
2531         case SYS_SETSOCKOPT:
2532                 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2533                 break;
2534         case SYS_GETSOCKOPT:
2535                 err =
2536                     sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2537                                    (int __user *)a[4]);
2538                 break;
2539         case SYS_SENDMSG:
2540                 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2541                 break;
2542         case SYS_SENDMMSG:
2543                 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2544                 break;
2545         case SYS_RECVMSG:
2546                 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2547                 break;
2548         case SYS_RECVMMSG:
2549                 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2550                                    (struct timespec __user *)a[4]);
2551                 break;
2552         case SYS_ACCEPT4:
2553                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2554                                   (int __user *)a[2], a[3]);
2555                 break;
2556         default:
2557                 err = -EINVAL;
2558                 break;
2559         }
2560         return err;
2561 }
2562 
2563 #endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2564 
2565 /**
2566  *      sock_register - add a socket protocol handler
2567  *      @ops: description of protocol
2568  *
2569  *      This function is called by a protocol handler that wants to
2570  *      advertise its address family, and have it linked into the
2571  *      socket interface. The value ops->family corresponds to the
2572  *      socket system call protocol family.
2573  */
2574 int sock_register(const struct net_proto_family *ops)
2575 {
2576         int err;
2577 
2578         if (ops->family >= NPROTO) {
2579                 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2580                 return -ENOBUFS;
2581         }
2582 
2583         spin_lock(&net_family_lock);
2584         if (rcu_dereference_protected(net_families[ops->family],
2585                                       lockdep_is_held(&net_family_lock)))
2586                 err = -EEXIST;
2587         else {
2588                 rcu_assign_pointer(net_families[ops->family], ops);
2589                 err = 0;
2590         }
2591         spin_unlock(&net_family_lock);
2592 
2593         pr_info("NET: Registered protocol family %d\n", ops->family);
2594         return err;
2595 }
2596 EXPORT_SYMBOL(sock_register);
2597 
2598 /**
2599  *      sock_unregister - remove a protocol handler
2600  *      @family: protocol family to remove
2601  *
2602  *      This function is called by a protocol handler that wants to
2603  *      remove its address family, and have it unlinked from the
2604  *      new socket creation.
2605  *
2606  *      If protocol handler is a module, then it can use module reference
2607  *      counts to protect against new references. If protocol handler is not
2608  *      a module then it needs to provide its own protection in
2609  *      the ops->create routine.
2610  */
2611 void sock_unregister(int family)
2612 {
2613         BUG_ON(family < 0 || family >= NPROTO);
2614 
2615         spin_lock(&net_family_lock);
2616         RCU_INIT_POINTER(net_families[family], NULL);
2617         spin_unlock(&net_family_lock);
2618 
2619         synchronize_rcu();
2620 
2621         pr_info("NET: Unregistered protocol family %d\n", family);
2622 }
2623 EXPORT_SYMBOL(sock_unregister);
2624 
2625 static int __init sock_init(void)
2626 {
2627         int err;
2628         /*
2629          *      Initialize the network sysctl infrastructure.
2630          */
2631         err = net_sysctl_init();
2632         if (err)
2633                 goto out;
2634 
2635         /*
2636          *      Initialize skbuff SLAB cache
2637          */
2638         skb_init();
2639 
2640         /*
2641          *      Initialize the protocols module.
2642          */
2643 
2644         init_inodecache();
2645 
2646         err = register_filesystem(&sock_fs_type);
2647         if (err)
2648                 goto out_fs;
2649         sock_mnt = kern_mount(&sock_fs_type);
2650         if (IS_ERR(sock_mnt)) {
2651                 err = PTR_ERR(sock_mnt);
2652                 goto out_mount;
2653         }
2654 
2655         /* The real protocol initialization is performed in later initcalls.
2656          */
2657 
2658 #ifdef CONFIG_NETFILTER
2659         err = netfilter_init();
2660         if (err)
2661                 goto out;
2662 #endif
2663 
2664         ptp_classifier_init();
2665 
2666 out:
2667         return err;
2668 
2669 out_mount:
2670         unregister_filesystem(&sock_fs_type);
2671 out_fs:
2672         goto out;
2673 }
2674 
2675 core_initcall(sock_init);       /* early initcall */
2676 
2677 #ifdef CONFIG_PROC_FS
2678 void socket_seq_show(struct seq_file *seq)
2679 {
2680         int cpu;
2681         int counter = 0;
2682 
2683         for_each_possible_cpu(cpu)
2684             counter += per_cpu(sockets_in_use, cpu);
2685 
2686         /* It can be negative, by the way. 8) */
2687         if (counter < 0)
2688                 counter = 0;
2689 
2690         seq_printf(seq, "sockets: used %d\n", counter);
2691 }
2692 #endif                          /* CONFIG_PROC_FS */
2693 
2694 #ifdef CONFIG_COMPAT
2695 static int do_siocgstamp(struct net *net, struct socket *sock,
2696                          unsigned int cmd, void __user *up)
2697 {
2698         mm_segment_t old_fs = get_fs();
2699         struct timeval ktv;
2700         int err;
2701 
2702         set_fs(KERNEL_DS);
2703         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2704         set_fs(old_fs);
2705         if (!err)
2706                 err = compat_put_timeval(&ktv, up);
2707 
2708         return err;
2709 }
2710 
2711 static int do_siocgstampns(struct net *net, struct socket *sock,
2712                            unsigned int cmd, void __user *up)
2713 {
2714         mm_segment_t old_fs = get_fs();
2715         struct timespec kts;
2716         int err;
2717 
2718         set_fs(KERNEL_DS);
2719         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2720         set_fs(old_fs);
2721         if (!err)
2722                 err = compat_put_timespec(&kts, up);
2723 
2724         return err;
2725 }
2726 
2727 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2728 {
2729         struct ifreq __user *uifr;
2730         int err;
2731 
2732         uifr = compat_alloc_user_space(sizeof(struct ifreq));
2733         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2734                 return -EFAULT;
2735 
2736         err = dev_ioctl(net, SIOCGIFNAME, uifr);
2737         if (err)
2738                 return err;
2739 
2740         if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2741                 return -EFAULT;
2742 
2743         return 0;
2744 }
2745 
2746 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2747 {
2748         struct compat_ifconf ifc32;
2749         struct ifconf ifc;
2750         struct ifconf __user *uifc;
2751         struct compat_ifreq __user *ifr32;
2752         struct ifreq __user *ifr;
2753         unsigned int i, j;
2754         int err;
2755 
2756         if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2757                 return -EFAULT;
2758 
2759         memset(&ifc, 0, sizeof(ifc));
2760         if (ifc32.ifcbuf == 0) {
2761                 ifc32.ifc_len = 0;
2762                 ifc.ifc_len = 0;
2763                 ifc.ifc_req = NULL;
2764                 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2765         } else {
2766                 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2767                         sizeof(struct ifreq);
2768                 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2769                 ifc.ifc_len = len;
2770                 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2771                 ifr32 = compat_ptr(ifc32.ifcbuf);
2772                 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2773                         if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2774                                 return -EFAULT;
2775                         ifr++;
2776                         ifr32++;
2777                 }
2778         }
2779         if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2780                 return -EFAULT;
2781 
2782         err = dev_ioctl(net, SIOCGIFCONF, uifc);
2783         if (err)
2784                 return err;
2785 
2786         if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2787                 return -EFAULT;
2788 
2789         ifr = ifc.ifc_req;
2790         ifr32 = compat_ptr(ifc32.ifcbuf);
2791         for (i = 0, j = 0;
2792              i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2793              i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2794                 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2795                         return -EFAULT;
2796                 ifr32++;
2797                 ifr++;
2798         }
2799 
2800         if (ifc32.ifcbuf == 0) {
2801                 /* Translate from 64-bit structure multiple to
2802                  * a 32-bit one.
2803                  */
2804                 i = ifc.ifc_len;
2805                 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2806                 ifc32.ifc_len = i;
2807         } else {
2808                 ifc32.ifc_len = i;
2809         }
2810         if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2811                 return -EFAULT;
2812 
2813         return 0;
2814 }
2815 
2816 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2817 {
2818         struct compat_ethtool_rxnfc __user *compat_rxnfc;
2819         bool convert_in = false, convert_out = false;
2820         size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2821         struct ethtool_rxnfc __user *rxnfc;
2822         struct ifreq __user *ifr;
2823         u32 rule_cnt = 0, actual_rule_cnt;
2824         u32 ethcmd;
2825         u32 data;
2826         int ret;
2827 
2828         if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2829                 return -EFAULT;
2830 
2831         compat_rxnfc = compat_ptr(data);
2832 
2833         if (get_user(ethcmd, &compat_rxnfc->cmd))
2834                 return -EFAULT;
2835 
2836         /* Most ethtool structures are defined without padding.
2837          * Unfortunately struct ethtool_rxnfc is an exception.
2838          */
2839         switch (ethcmd) {
2840         default:
2841                 break;
2842         case ETHTOOL_GRXCLSRLALL:
2843                 /* Buffer size is variable */
2844                 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2845                         return -EFAULT;
2846                 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2847                         return -ENOMEM;
2848                 buf_size += rule_cnt * sizeof(u32);
2849                 /* fall through */
2850         case ETHTOOL_GRXRINGS:
2851         case ETHTOOL_GRXCLSRLCNT:
2852         case ETHTOOL_GRXCLSRULE:
2853         case ETHTOOL_SRXCLSRLINS:
2854                 convert_out = true;
2855                 /* fall through */
2856         case ETHTOOL_SRXCLSRLDEL:
2857                 buf_size += sizeof(struct ethtool_rxnfc);
2858                 convert_in = true;
2859                 break;
2860         }
2861 
2862         ifr = compat_alloc_user_space(buf_size);
2863         rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2864 
2865         if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2866                 return -EFAULT;
2867 
2868         if (put_user(convert_in ? rxnfc : compat_ptr(data),
2869                      &ifr->ifr_ifru.ifru_data))
2870                 return -EFAULT;
2871 
2872         if (convert_in) {
2873                 /* We expect there to be holes between fs.m_ext and
2874                  * fs.ring_cookie and at the end of fs, but nowhere else.
2875                  */
2876                 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2877                              sizeof(compat_rxnfc->fs.m_ext) !=
2878                              offsetof(struct ethtool_rxnfc, fs.m_ext) +
2879                              sizeof(rxnfc->fs.m_ext));
2880                 BUILD_BUG_ON(
2881                         offsetof(struct compat_ethtool_rxnfc, fs.location) -
2882                         offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2883                         offsetof(struct ethtool_rxnfc, fs.location) -
2884                         offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2885 
2886                 if (copy_in_user(rxnfc, compat_rxnfc,
2887                                  (void __user *)(&rxnfc->fs.m_ext + 1) -
2888                                  (void __user *)rxnfc) ||
2889                     copy_in_user(&rxnfc->fs.ring_cookie,
2890                                  &compat_rxnfc->fs.ring_cookie,
2891                                  (void __user *)(&rxnfc->fs.location + 1) -
2892                                  (void __user *)&rxnfc->fs.ring_cookie) ||
2893                     copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2894                                  sizeof(rxnfc->rule_cnt)))
2895                         return -EFAULT;
2896         }
2897 
2898         ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2899         if (ret)
2900                 return ret;
2901 
2902         if (convert_out) {
2903                 if (copy_in_user(compat_rxnfc, rxnfc,
2904                                  (const void __user *)(&rxnfc->fs.m_ext + 1) -
2905                                  (const void __user *)rxnfc) ||
2906                     copy_in_user(&compat_rxnfc->fs.ring_cookie,
2907                                  &rxnfc->fs.ring_cookie,
2908                                  (const void __user *)(&rxnfc->fs.location + 1) -
2909                                  (const void __user *)&rxnfc->fs.ring_cookie) ||
2910                     copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2911                                  sizeof(rxnfc->rule_cnt)))
2912                         return -EFAULT;
2913 
2914                 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2915                         /* As an optimisation, we only copy the actual
2916                          * number of rules that the underlying
2917                          * function returned.  Since Mallory might
2918                          * change the rule count in user memory, we
2919                          * check that it is less than the rule count
2920                          * originally given (as the user buffer size),
2921                          * which has been range-checked.
2922                          */
2923                         if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2924                                 return -EFAULT;
2925                         if (actual_rule_cnt < rule_cnt)
2926                                 rule_cnt = actual_rule_cnt;
2927                         if (copy_in_user(&compat_rxnfc->rule_locs[0],
2928                                          &rxnfc->rule_locs[0],
2929                                          rule_cnt * sizeof(u32)))
2930                                 return -EFAULT;
2931                 }
2932         }
2933 
2934         return 0;
2935 }
2936 
2937 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2938 {
2939         void __user *uptr;
2940         compat_uptr_t uptr32;
2941         struct ifreq __user *uifr;
2942 
2943         uifr = compat_alloc_user_space(sizeof(*uifr));
2944         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2945                 return -EFAULT;
2946 
2947         if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2948                 return -EFAULT;
2949 
2950         uptr = compat_ptr(uptr32);
2951 
2952         if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2953                 return -EFAULT;
2954 
2955         return dev_ioctl(net, SIOCWANDEV, uifr);
2956 }
2957 
2958 static int bond_ioctl(struct net *net, unsigned int cmd,
2959                          struct compat_ifreq __user *ifr32)
2960 {
2961         struct ifreq kifr;
2962         mm_segment_t old_fs;
2963         int err;
2964 
2965         switch (cmd) {
2966         case SIOCBONDENSLAVE:
2967         case SIOCBONDRELEASE:
2968         case SIOCBONDSETHWADDR:
2969         case SIOCBONDCHANGEACTIVE:
2970                 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2971                         return -EFAULT;
2972 
2973                 old_fs = get_fs();
2974                 set_fs(KERNEL_DS);
2975                 err = dev_ioctl(net, cmd,
2976                                 (struct ifreq __user __force *) &kifr);
2977                 set_fs(old_fs);
2978 
2979                 return err;
2980         default:
2981                 return -ENOIOCTLCMD;
2982         }
2983 }
2984 
2985 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2986 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2987                                  struct compat_ifreq __user *u_ifreq32)
2988 {
2989         struct ifreq __user *u_ifreq64;
2990         char tmp_buf[IFNAMSIZ];
2991         void __user *data64;
2992         u32 data32;
2993 
2994         if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2995                            IFNAMSIZ))
2996                 return -EFAULT;
2997         if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2998                 return -EFAULT;
2999         data64 = compat_ptr(data32);
3000 
3001         u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3002 
3003         if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3004                          IFNAMSIZ))
3005                 return -EFAULT;
3006         if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3007                 return -EFAULT;
3008 
3009         return dev_ioctl(net, cmd, u_ifreq64);
3010 }
3011 
3012 static int dev_ifsioc(struct net *net, struct socket *sock,
3013                          unsigned int cmd, struct compat_ifreq __user *uifr32)
3014 {
3015         struct ifreq __user *uifr;
3016         int err;
3017 
3018         uifr = compat_alloc_user_space(sizeof(*uifr));
3019         if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3020                 return -EFAULT;
3021 
3022         err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3023 
3024         if (!err) {
3025                 switch (cmd) {
3026                 case SIOCGIFFLAGS:
3027                 case SIOCGIFMETRIC:
3028                 case SIOCGIFMTU:
3029                 case SIOCGIFMEM:
3030                 case SIOCGIFHWADDR:
3031                 case SIOCGIFINDEX:
3032                 case SIOCGIFADDR:
3033                 case SIOCGIFBRDADDR:
3034                 case SIOCGIFDSTADDR:
3035                 case SIOCGIFNETMASK:
3036                 case SIOCGIFPFLAGS:
3037                 case SIOCGIFTXQLEN:
3038                 case SIOCGMIIPHY:
3039                 case SIOCGMIIREG:
3040                         if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3041                                 err = -EFAULT;
3042                         break;
3043                 }
3044         }
3045         return err;
3046 }
3047 
3048 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3049                         struct compat_ifreq __user *uifr32)
3050 {
3051         struct ifreq ifr;
3052         struct compat_ifmap __user *uifmap32;
3053         mm_segment_t old_fs;
3054         int err;
3055 
3056         uifmap32 = &uifr32->ifr_ifru.ifru_map;
3057         err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3058         err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3059         err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3060         err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3061         err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3062         err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3063         err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3064         if (err)
3065                 return -EFAULT;
3066 
3067         old_fs = get_fs();
3068         set_fs(KERNEL_DS);
3069         err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3070         set_fs(old_fs);
3071 
3072         if (cmd == SIOCGIFMAP && !err) {
3073                 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3074                 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3075                 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3076                 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3077                 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3078                 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3079                 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3080                 if (err)
3081                         err = -EFAULT;
3082         }
3083         return err;
3084 }
3085 
3086 struct rtentry32 {
3087         u32             rt_pad1;
3088         struct sockaddr rt_dst;         /* target address               */
3089         struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3090         struct sockaddr rt_genmask;     /* target network mask (IP)     */
3091         unsigned short  rt_flags;
3092         short           rt_pad2;
3093         u32             rt_pad3;
3094         unsigned char   rt_tos;
3095         unsigned char   rt_class;
3096         short           rt_pad4;
3097         short           rt_metric;      /* +1 for binary compatibility! */
3098         /* char * */ u32 rt_dev;        /* forcing the device at add    */
3099         u32             rt_mtu;         /* per route MTU/Window         */
3100         u32             rt_window;      /* Window clamping              */
3101         unsigned short  rt_irtt;        /* Initial RTT                  */
3102 };
3103 
3104 struct in6_rtmsg32 {
3105         struct in6_addr         rtmsg_dst;
3106         struct in6_addr         rtmsg_src;
3107         struct in6_addr         rtmsg_gateway;
3108         u32                     rtmsg_type;
3109         u16                     rtmsg_dst_len;
3110         u16                     rtmsg_src_len;
3111         u32                     rtmsg_metric;
3112         u32                     rtmsg_info;
3113         u32                     rtmsg_flags;
3114         s32                     rtmsg_ifindex;
3115 };
3116 
3117 static int routing_ioctl(struct net *net, struct socket *sock,
3118                          unsigned int cmd, void __user *argp)
3119 {
3120         int ret;
3121         void *r = NULL;
3122         struct in6_rtmsg r6;
3123         struct rtentry r4;
3124         char devname[16];
3125         u32 rtdev;
3126         mm_segment_t old_fs = get_fs();
3127 
3128         if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3129                 struct in6_rtmsg32 __user *ur6 = argp;
3130                 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3131                         3 * sizeof(struct in6_addr));
3132                 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3133                 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3134                 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3135                 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3136                 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3137                 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3138                 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3139 
3140                 r = (void *) &r6;
3141         } else { /* ipv4 */
3142                 struct rtentry32 __user *ur4 = argp;
3143                 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3144                                         3 * sizeof(struct sockaddr));
3145                 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3146                 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3147                 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3148                 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3149                 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3150                 ret |= get_user(rtdev, &(ur4->rt_dev));
3151                 if (rtdev) {
3152                         ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3153                         r4.rt_dev = (char __user __force *)devname;
3154                         devname[15] = 0;
3155                 } else
3156                         r4.rt_dev = NULL;
3157 
3158                 r = (void *) &r4;
3159         }
3160 
3161         if (ret) {
3162                 ret = -EFAULT;
3163                 goto out;
3164         }
3165 
3166         set_fs(KERNEL_DS);
3167         ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3168         set_fs(old_fs);
3169 
3170 out:
3171         return ret;
3172 }
3173 
3174 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3175  * for some operations; this forces use of the newer bridge-utils that
3176  * use compatible ioctls
3177  */
3178 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3179 {
3180         compat_ulong_t tmp;
3181 
3182         if (get_user(tmp, argp))
3183                 return -EFAULT;
3184         if (tmp == BRCTL_GET_VERSION)
3185                 return BRCTL_VERSION + 1;
3186         return -EINVAL;
3187 }
3188 
3189 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3190                          unsigned int cmd, unsigned long arg)
3191 {
3192         void __user *argp = compat_ptr(arg);
3193         struct sock *sk = sock->sk;
3194         struct net *net = sock_net(sk);
3195 
3196         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3197                 return compat_ifr_data_ioctl(net, cmd, argp);
3198 
3199         switch (cmd) {
3200         case SIOCSIFBR:
3201         case SIOCGIFBR:
3202                 return old_bridge_ioctl(argp);
3203         case SIOCGIFNAME:
3204                 return dev_ifname32(net, argp);
3205         case SIOCGIFCONF:
3206                 return dev_ifconf(net, argp);
3207         case SIOCETHTOOL:
3208                 return ethtool_ioctl(net, argp);
3209         case SIOCWANDEV:
3210                 return compat_siocwandev(net, argp);
3211         case SIOCGIFMAP:
3212         case SIOCSIFMAP:
3213                 return compat_sioc_ifmap(net, cmd, argp);
3214         case SIOCBONDENSLAVE:
3215         case SIOCBONDRELEASE:
3216         case SIOCBONDSETHWADDR:
3217         case SIOCBONDCHANGEACTIVE:
3218                 return bond_ioctl(net, cmd, argp);
3219         case SIOCADDRT:
3220         case SIOCDELRT:
3221                 return routing_ioctl(net, sock, cmd, argp);
3222         case SIOCGSTAMP:
3223                 return do_siocgstamp(net, sock, cmd, argp);
3224         case SIOCGSTAMPNS:
3225                 return do_siocgstampns(net, sock, cmd, argp);
3226         case SIOCBONDSLAVEINFOQUERY:
3227         case SIOCBONDINFOQUERY:
3228         case SIOCSHWTSTAMP:
3229         case SIOCGHWTSTAMP:
3230                 return compat_ifr_data_ioctl(net, cmd, argp);
3231 
3232         case FIOSETOWN:
3233         case SIOCSPGRP:
3234         case FIOGETOWN:
3235         case SIOCGPGRP:
3236         case SIOCBRADDBR:
3237         case SIOCBRDELBR:
3238         case SIOCGIFVLAN:
3239         case SIOCSIFVLAN:
3240         case SIOCADDDLCI:
3241         case SIOCDELDLCI:
3242                 return sock_ioctl(file, cmd, arg);
3243 
3244         case SIOCGIFFLAGS:
3245         case SIOCSIFFLAGS:
3246         case SIOCGIFMETRIC:
3247         case SIOCSIFMETRIC:
3248         case SIOCGIFMTU:
3249         case SIOCSIFMTU:
3250         case SIOCGIFMEM:
3251         case SIOCSIFMEM:
3252         case SIOCGIFHWADDR:
3253         case SIOCSIFHWADDR:
3254         case SIOCADDMULTI:
3255         case SIOCDELMULTI:
3256         case SIOCGIFINDEX:
3257         case SIOCGIFADDR:
3258         case SIOCSIFADDR:
3259         case SIOCSIFHWBROADCAST:
3260         case SIOCDIFADDR:
3261         case SIOCGIFBRDADDR:
3262         case SIOCSIFBRDADDR:
3263         case SIOCGIFDSTADDR:
3264         case SIOCSIFDSTADDR:
3265         case SIOCGIFNETMASK:
3266         case SIOCSIFNETMASK:
3267         case SIOCSIFPFLAGS:
3268         case SIOCGIFPFLAGS:
3269         case SIOCGIFTXQLEN:
3270         case SIOCSIFTXQLEN:
3271         case SIOCBRADDIF:
3272         case SIOCBRDELIF:
3273         case SIOCSIFNAME:
3274         case SIOCGMIIPHY:
3275         case SIOCGMIIREG:
3276         case SIOCSMIIREG:
3277                 return dev_ifsioc(net, sock, cmd, argp);
3278 
3279         case SIOCSARP:
3280         case SIOCGARP:
3281         case SIOCDARP:
3282         case SIOCATMARK:
3283                 return sock_do_ioctl(net, sock, cmd, arg);
3284         }
3285 
3286         return -ENOIOCTLCMD;
3287 }
3288 
3289 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3290                               unsigned long arg)
3291 {
3292         struct socket *sock = file->private_data;
3293         int ret = -ENOIOCTLCMD;
3294         struct sock *sk;
3295         struct net *net;
3296 
3297         sk = sock->sk;
3298         net = sock_net(sk);
3299 
3300         if (sock->ops->compat_ioctl)
3301                 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3302 
3303         if (ret == -ENOIOCTLCMD &&
3304             (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3305                 ret = compat_wext_handle_ioctl(net, cmd, arg);
3306 
3307         if (ret == -ENOIOCTLCMD)
3308                 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3309 
3310         return ret;
3311 }
3312 #endif
3313 
3314 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3315 {
3316         return sock->ops->bind(sock, addr, addrlen);
3317 }
3318 EXPORT_SYMBOL(kernel_bind);
3319 
3320 int kernel_listen(struct socket *sock, int backlog)
3321 {
3322         return sock->ops->listen(sock, backlog);
3323 }
3324 EXPORT_SYMBOL(kernel_listen);
3325 
3326 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3327 {
3328         struct sock *sk = sock->sk;
3329         int err;
3330 
3331         err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3332                                newsock);
3333         if (err < 0)
3334                 goto done;
3335 
3336         err = sock->ops->accept(sock, *newsock, flags);
3337         if (err < 0) {
3338                 sock_release(*newsock);
3339                 *newsock = NULL;
3340                 goto done;
3341         }
3342 
3343         (*newsock)->ops = sock->ops;
3344         __module_get((*newsock)->ops->owner);
3345 
3346 done:
3347         return err;
3348 }
3349 EXPORT_SYMBOL(kernel_accept);
3350 
3351 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3352                    int flags)
3353 {
3354         return sock->ops->connect(sock, addr, addrlen, flags);
3355 }
3356 EXPORT_SYMBOL(kernel_connect);
3357 
3358 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3359                          int *addrlen)
3360 {
3361         return sock->ops->getname(sock, addr, addrlen, 0);
3362 }
3363 EXPORT_SYMBOL(kernel_getsockname);
3364 
3365 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3366                          int *addrlen)
3367 {
3368         return sock->ops->getname(sock, addr, addrlen, 1);
3369 }
3370 EXPORT_SYMBOL(kernel_getpeername);
3371 
3372 int kernel_getsockopt(struct socket *sock, int level, int optname,
3373                         char *optval, int *optlen)
3374 {
3375         mm_segment_t oldfs = get_fs();
3376         char __user *uoptval;
3377         int __user *uoptlen;
3378         int err;
3379 
3380         uoptval = (char __user __force *) optval;
3381         uoptlen = (int __user __force *) optlen;
3382 
3383         set_fs(KERNEL_DS);
3384         if (level == SOL_SOCKET)
3385                 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3386         else
3387                 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3388                                             uoptlen);
3389         set_fs(oldfs);
3390         return err;
3391 }
3392 EXPORT_SYMBOL(kernel_getsockopt);
3393 
3394 int kernel_setsockopt(struct socket *sock, int level, int optname,
3395                         char *optval, unsigned int optlen)
3396 {
3397         mm_segment_t oldfs = get_fs();
3398         char __user *uoptval;
3399         int err;
3400 
3401         uoptval = (char __user __force *) optval;
3402 
3403         set_fs(KERNEL_DS);
3404         if (level == SOL_SOCKET)
3405                 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3406         else
3407                 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3408                                             optlen);
3409         set_fs(oldfs);
3410         return err;
3411 }
3412 EXPORT_SYMBOL(kernel_setsockopt);
3413 
3414 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3415                     size_t size, int flags)
3416 {
3417         if (sock->ops->sendpage)
3418                 return sock->ops->sendpage(sock, page, offset, size, flags);
3419 
3420         return sock_no_sendpage(sock, page, offset, size, flags);
3421 }
3422 EXPORT_SYMBOL(kernel_sendpage);
3423 
3424 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3425 {
3426         mm_segment_t oldfs = get_fs();
3427         int err;
3428 
3429         set_fs(KERNEL_DS);
3430         err = sock->ops->ioctl(sock, cmd, arg);
3431         set_fs(oldfs);
3432 
3433         return err;
3434 }
3435 EXPORT_SYMBOL(kernel_sock_ioctl);
3436 
3437 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3438 {
3439         return sock->ops->shutdown(sock, how);
3440 }
3441 EXPORT_SYMBOL(kernel_sock_shutdown);
3442 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us