~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/socket.c

Version: ~ [ linux-5.11-rc3 ] ~ [ linux-5.10.7 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.89 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.167 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.215 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.251 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.251 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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/init.h>
 76 #include <linux/poll.h>
 77 #include <linux/cache.h>
 78 #include <linux/module.h>
 79 #include <linux/highmem.h>
 80 #include <linux/mount.h>
 81 #include <linux/security.h>
 82 #include <linux/syscalls.h>
 83 #include <linux/compat.h>
 84 #include <linux/kmod.h>
 85 #include <linux/audit.h>
 86 #include <linux/wireless.h>
 87 #include <linux/nsproxy.h>
 88 #include <linux/magic.h>
 89 #include <linux/slab.h>
 90 #include <linux/xattr.h>
 91 
 92 #include <asm/uaccess.h>
 93 #include <asm/unistd.h>
 94 
 95 #include <net/compat.h>
 96 #include <net/wext.h>
 97 #include <net/cls_cgroup.h>
 98 
 99 #include <net/sock.h>
100 #include <linux/netfilter.h>
101 
102 #include <linux/if_tun.h>
103 #include <linux/ipv6_route.h>
104 #include <linux/route.h>
105 #include <linux/sockios.h>
106 #include <linux/atalk.h>
107 #include <net/busy_poll.h>
108 
109 #ifdef CONFIG_NET_RX_BUSY_POLL
110 unsigned int sysctl_net_busy_read __read_mostly;
111 unsigned int sysctl_net_busy_poll __read_mostly;
112 #endif
113 
114 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
115 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
116                          unsigned long nr_segs, loff_t pos);
117 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
118                           unsigned long nr_segs, loff_t pos);
119 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
120 
121 static int sock_close(struct inode *inode, struct file *file);
122 static unsigned int sock_poll(struct file *file,
123                               struct poll_table_struct *wait);
124 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
125 #ifdef CONFIG_COMPAT
126 static long compat_sock_ioctl(struct file *file,
127                               unsigned int cmd, unsigned long arg);
128 #endif
129 static int sock_fasync(int fd, struct file *filp, int on);
130 static ssize_t sock_sendpage(struct file *file, struct page *page,
131                              int offset, size_t size, loff_t *ppos, int more);
132 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
133                                 struct pipe_inode_info *pipe, size_t len,
134                                 unsigned int flags);
135 
136 /*
137  *      Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
138  *      in the operation structures but are done directly via the socketcall() multiplexor.
139  */
140 
141 static const struct file_operations socket_file_ops = {
142         .owner =        THIS_MODULE,
143         .llseek =       no_llseek,
144         .aio_read =     sock_aio_read,
145         .aio_write =    sock_aio_write,
146         .poll =         sock_poll,
147         .unlocked_ioctl = sock_ioctl,
148 #ifdef CONFIG_COMPAT
149         .compat_ioctl = compat_sock_ioctl,
150 #endif
151         .mmap =         sock_mmap,
152         .open =         sock_no_open,   /* special open code to disallow open via /proc */
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         err = get_user(len, ulen);
225         if (err)
226                 return err;
227         if (len > klen)
228                 len = klen;
229         if (len < 0 || len > sizeof(struct sockaddr_storage))
230                 return -EINVAL;
231         if (len) {
232                 if (audit_sockaddr(klen, kaddr))
233                         return -ENOMEM;
234                 if (copy_to_user(uaddr, kaddr, len))
235                         return -EFAULT;
236         }
237         /*
238          *      "fromlen shall refer to the value before truncation.."
239          *                      1003.1g
240          */
241         return __put_user(klen, ulen);
242 }
243 
244 static struct kmem_cache *sock_inode_cachep __read_mostly;
245 
246 static struct inode *sock_alloc_inode(struct super_block *sb)
247 {
248         struct socket_alloc *ei;
249         struct socket_wq *wq;
250 
251         ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
252         if (!ei)
253                 return NULL;
254         wq = kmalloc(sizeof(*wq), GFP_KERNEL);
255         if (!wq) {
256                 kmem_cache_free(sock_inode_cachep, ei);
257                 return NULL;
258         }
259         init_waitqueue_head(&wq->wait);
260         wq->fasync_list = NULL;
261         RCU_INIT_POINTER(ei->socket.wq, wq);
262 
263         ei->socket.state = SS_UNCONNECTED;
264         ei->socket.flags = 0;
265         ei->socket.ops = NULL;
266         ei->socket.sk = NULL;
267         ei->socket.file = NULL;
268 
269         return &ei->vfs_inode;
270 }
271 
272 static void sock_destroy_inode(struct inode *inode)
273 {
274         struct socket_alloc *ei;
275         struct socket_wq *wq;
276 
277         ei = container_of(inode, struct socket_alloc, vfs_inode);
278         wq = rcu_dereference_protected(ei->socket.wq, 1);
279         kfree_rcu(wq, rcu);
280         kmem_cache_free(sock_inode_cachep, ei);
281 }
282 
283 static void init_once(void *foo)
284 {
285         struct socket_alloc *ei = (struct socket_alloc *)foo;
286 
287         inode_init_once(&ei->vfs_inode);
288 }
289 
290 static int init_inodecache(void)
291 {
292         sock_inode_cachep = kmem_cache_create("sock_inode_cache",
293                                               sizeof(struct socket_alloc),
294                                               0,
295                                               (SLAB_HWCACHE_ALIGN |
296                                                SLAB_RECLAIM_ACCOUNT |
297                                                SLAB_MEM_SPREAD),
298                                               init_once);
299         if (sock_inode_cachep == NULL)
300                 return -ENOMEM;
301         return 0;
302 }
303 
304 static const struct super_operations sockfs_ops = {
305         .alloc_inode    = sock_alloc_inode,
306         .destroy_inode  = sock_destroy_inode,
307         .statfs         = simple_statfs,
308 };
309 
310 /*
311  * sockfs_dname() is called from d_path().
312  */
313 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314 {
315         return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
316                                 dentry->d_inode->i_ino);
317 }
318 
319 static const struct dentry_operations sockfs_dentry_operations = {
320         .d_dname  = sockfs_dname,
321 };
322 
323 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
324                          int flags, const char *dev_name, void *data)
325 {
326         return mount_pseudo(fs_type, "socket:", &sockfs_ops,
327                 &sockfs_dentry_operations, SOCKFS_MAGIC);
328 }
329 
330 static struct vfsmount *sock_mnt __read_mostly;
331 
332 static struct file_system_type sock_fs_type = {
333         .name =         "sockfs",
334         .mount =        sockfs_mount,
335         .kill_sb =      kill_anon_super,
336 };
337 
338 /*
339  *      Obtains the first available file descriptor and sets it up for use.
340  *
341  *      These functions create file structures and maps them to fd space
342  *      of the current process. On success it returns file descriptor
343  *      and file struct implicitly stored in sock->file.
344  *      Note that another thread may close file descriptor before we return
345  *      from this function. We use the fact that now we do not refer
346  *      to socket after mapping. If one day we will need it, this
347  *      function will increment ref. count on file by 1.
348  *
349  *      In any case returned fd MAY BE not valid!
350  *      This race condition is unavoidable
351  *      with shared fd spaces, we cannot solve it inside kernel,
352  *      but we take care of internal coherence yet.
353  */
354 
355 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
356 {
357         struct qstr name = { .name = "" };
358         struct path path;
359         struct file *file;
360 
361         if (dname) {
362                 name.name = dname;
363                 name.len = strlen(name.name);
364         } else if (sock->sk) {
365                 name.name = sock->sk->sk_prot_creator->name;
366                 name.len = strlen(name.name);
367         }
368         path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
369         if (unlikely(!path.dentry))
370                 return ERR_PTR(-ENOMEM);
371         path.mnt = mntget(sock_mnt);
372 
373         d_instantiate(path.dentry, SOCK_INODE(sock));
374         SOCK_INODE(sock)->i_fop = &socket_file_ops;
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 file *file;
453         struct socket *sock;
454 
455         *err = -EBADF;
456         file = fget_light(fd, fput_needed);
457         if (file) {
458                 sock = sock_from_file(file, err);
459                 if (sock)
460                         return sock;
461                 fput_light(file, *fput_needed);
462         }
463         return NULL;
464 }
465 
466 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469 static ssize_t sockfs_getxattr(struct dentry *dentry,
470                                const char *name, void *value, size_t size)
471 {
472         const char *proto_name;
473         size_t proto_size;
474         int error;
475 
476         error = -ENODATA;
477         if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
478                 proto_name = dentry->d_name.name;
479                 proto_size = strlen(proto_name);
480 
481                 if (value) {
482                         error = -ERANGE;
483                         if (proto_size + 1 > size)
484                                 goto out;
485 
486                         strncpy(value, proto_name, proto_size + 1);
487                 }
488                 error = proto_size + 1;
489         }
490 
491 out:
492         return error;
493 }
494 
495 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
496                                 size_t size)
497 {
498         ssize_t len;
499         ssize_t used = 0;
500 
501         len = security_inode_listsecurity(dentry->d_inode, buffer, size);
502         if (len < 0)
503                 return len;
504         used += len;
505         if (buffer) {
506                 if (size < used)
507                         return -ERANGE;
508                 buffer += len;
509         }
510 
511         len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
512         used += len;
513         if (buffer) {
514                 if (size < used)
515                         return -ERANGE;
516                 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
517                 buffer += len;
518         }
519 
520         return used;
521 }
522 
523 static const struct inode_operations sockfs_inode_ops = {
524         .getxattr = sockfs_getxattr,
525         .listxattr = sockfs_listxattr,
526 };
527 
528 /**
529  *      sock_alloc      -       allocate a socket
530  *
531  *      Allocate a new inode and socket object. The two are bound together
532  *      and initialised. The socket is then returned. If we are out of inodes
533  *      NULL is returned.
534  */
535 
536 static struct socket *sock_alloc(void)
537 {
538         struct inode *inode;
539         struct socket *sock;
540 
541         inode = new_inode_pseudo(sock_mnt->mnt_sb);
542         if (!inode)
543                 return NULL;
544 
545         sock = SOCKET_I(inode);
546 
547         kmemcheck_annotate_bitfield(sock, type);
548         inode->i_ino = get_next_ino();
549         inode->i_mode = S_IFSOCK | S_IRWXUGO;
550         inode->i_uid = current_fsuid();
551         inode->i_gid = current_fsgid();
552         inode->i_op = &sockfs_inode_ops;
553 
554         this_cpu_add(sockets_in_use, 1);
555         return sock;
556 }
557 
558 /*
559  *      In theory you can't get an open on this inode, but /proc provides
560  *      a back door. Remember to keep it shut otherwise you'll let the
561  *      creepy crawlies in.
562  */
563 
564 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
565 {
566         return -ENXIO;
567 }
568 
569 const struct file_operations bad_sock_fops = {
570         .owner = THIS_MODULE,
571         .open = sock_no_open,
572         .llseek = noop_llseek,
573 };
574 
575 /**
576  *      sock_release    -       close a socket
577  *      @sock: socket to close
578  *
579  *      The socket is released from the protocol stack if it has a release
580  *      callback, and the inode is then released if the socket is bound to
581  *      an inode not a file.
582  */
583 
584 void sock_release(struct socket *sock)
585 {
586         if (sock->ops) {
587                 struct module *owner = sock->ops->owner;
588 
589                 sock->ops->release(sock);
590                 sock->ops = NULL;
591                 module_put(owner);
592         }
593 
594         if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
595                 printk(KERN_ERR "sock_release: fasync list not empty!\n");
596 
597         if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
598                 return;
599 
600         this_cpu_sub(sockets_in_use, 1);
601         if (!sock->file) {
602                 iput(SOCK_INODE(sock));
603                 return;
604         }
605         sock->file = NULL;
606 }
607 EXPORT_SYMBOL(sock_release);
608 
609 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
610 {
611         *tx_flags = 0;
612         if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
613                 *tx_flags |= SKBTX_HW_TSTAMP;
614         if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
615                 *tx_flags |= SKBTX_SW_TSTAMP;
616         if (sock_flag(sk, SOCK_WIFI_STATUS))
617                 *tx_flags |= SKBTX_WIFI_STATUS;
618 }
619 EXPORT_SYMBOL(sock_tx_timestamp);
620 
621 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
622                                        struct msghdr *msg, size_t size)
623 {
624         struct sock_iocb *si = kiocb_to_siocb(iocb);
625 
626         si->sock = sock;
627         si->scm = NULL;
628         si->msg = msg;
629         si->size = size;
630 
631         return sock->ops->sendmsg(iocb, sock, msg, size);
632 }
633 
634 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
635                                  struct msghdr *msg, size_t size)
636 {
637         int err = security_socket_sendmsg(sock, msg, size);
638 
639         return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
640 }
641 
642 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
643 {
644         struct kiocb iocb;
645         struct sock_iocb siocb;
646         int ret;
647 
648         init_sync_kiocb(&iocb, NULL);
649         iocb.private = &siocb;
650         ret = __sock_sendmsg(&iocb, sock, msg, size);
651         if (-EIOCBQUEUED == ret)
652                 ret = wait_on_sync_kiocb(&iocb);
653         return ret;
654 }
655 EXPORT_SYMBOL(sock_sendmsg);
656 
657 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
658 {
659         struct kiocb iocb;
660         struct sock_iocb siocb;
661         int ret;
662 
663         init_sync_kiocb(&iocb, NULL);
664         iocb.private = &siocb;
665         ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
666         if (-EIOCBQUEUED == ret)
667                 ret = wait_on_sync_kiocb(&iocb);
668         return ret;
669 }
670 
671 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
672                    struct kvec *vec, size_t num, size_t size)
673 {
674         mm_segment_t oldfs = get_fs();
675         int result;
676 
677         set_fs(KERNEL_DS);
678         /*
679          * the following is safe, since for compiler definitions of kvec and
680          * iovec are identical, yielding the same in-core layout and alignment
681          */
682         msg->msg_iov = (struct iovec *)vec;
683         msg->msg_iovlen = num;
684         result = sock_sendmsg(sock, msg, size);
685         set_fs(oldfs);
686         return result;
687 }
688 EXPORT_SYMBOL(kernel_sendmsg);
689 
690 /*
691  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
692  */
693 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
694         struct sk_buff *skb)
695 {
696         int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
697         struct timespec ts[3];
698         int empty = 1;
699         struct skb_shared_hwtstamps *shhwtstamps =
700                 skb_hwtstamps(skb);
701 
702         /* Race occurred between timestamp enabling and packet
703            receiving.  Fill in the current time for now. */
704         if (need_software_tstamp && skb->tstamp.tv64 == 0)
705                 __net_timestamp(skb);
706 
707         if (need_software_tstamp) {
708                 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
709                         struct timeval tv;
710                         skb_get_timestamp(skb, &tv);
711                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
712                                  sizeof(tv), &tv);
713                 } else {
714                         skb_get_timestampns(skb, &ts[0]);
715                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
716                                  sizeof(ts[0]), &ts[0]);
717                 }
718         }
719 
720 
721         memset(ts, 0, sizeof(ts));
722         if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
723             ktime_to_timespec_cond(skb->tstamp, ts + 0))
724                 empty = 0;
725         if (shhwtstamps) {
726                 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
727                     ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
728                         empty = 0;
729                 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
730                     ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
731                         empty = 0;
732         }
733         if (!empty)
734                 put_cmsg(msg, SOL_SOCKET,
735                          SCM_TIMESTAMPING, sizeof(ts), &ts);
736 }
737 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
738 
739 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
740         struct sk_buff *skb)
741 {
742         int ack;
743 
744         if (!sock_flag(sk, SOCK_WIFI_STATUS))
745                 return;
746         if (!skb->wifi_acked_valid)
747                 return;
748 
749         ack = skb->wifi_acked;
750 
751         put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
752 }
753 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
754 
755 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
756                                    struct sk_buff *skb)
757 {
758         if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
759                 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
760                         sizeof(__u32), &skb->dropcount);
761 }
762 
763 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
764         struct sk_buff *skb)
765 {
766         sock_recv_timestamp(msg, sk, skb);
767         sock_recv_drops(msg, sk, skb);
768 }
769 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
770 
771 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
772                                        struct msghdr *msg, size_t size, int flags)
773 {
774         struct sock_iocb *si = kiocb_to_siocb(iocb);
775 
776         si->sock = sock;
777         si->scm = NULL;
778         si->msg = msg;
779         si->size = size;
780         si->flags = flags;
781 
782         return sock->ops->recvmsg(iocb, sock, msg, size, flags);
783 }
784 
785 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
786                                  struct msghdr *msg, size_t size, int flags)
787 {
788         int err = security_socket_recvmsg(sock, msg, size, flags);
789 
790         return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
791 }
792 
793 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
794                  size_t size, int flags)
795 {
796         struct kiocb iocb;
797         struct sock_iocb siocb;
798         int ret;
799 
800         init_sync_kiocb(&iocb, NULL);
801         iocb.private = &siocb;
802         ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
803         if (-EIOCBQUEUED == ret)
804                 ret = wait_on_sync_kiocb(&iocb);
805         return ret;
806 }
807 EXPORT_SYMBOL(sock_recvmsg);
808 
809 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
810                               size_t size, int flags)
811 {
812         struct kiocb iocb;
813         struct sock_iocb siocb;
814         int ret;
815 
816         init_sync_kiocb(&iocb, NULL);
817         iocb.private = &siocb;
818         ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
819         if (-EIOCBQUEUED == ret)
820                 ret = wait_on_sync_kiocb(&iocb);
821         return ret;
822 }
823 
824 /**
825  * kernel_recvmsg - Receive a message from a socket (kernel space)
826  * @sock:       The socket to receive the message from
827  * @msg:        Received message
828  * @vec:        Input s/g array for message data
829  * @num:        Size of input s/g array
830  * @size:       Number of bytes to read
831  * @flags:      Message flags (MSG_DONTWAIT, etc...)
832  *
833  * On return the msg structure contains the scatter/gather array passed in the
834  * vec argument. The array is modified so that it consists of the unfilled
835  * portion of the original array.
836  *
837  * The returned value is the total number of bytes received, or an error.
838  */
839 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
840                    struct kvec *vec, size_t num, size_t size, int flags)
841 {
842         mm_segment_t oldfs = get_fs();
843         int result;
844 
845         set_fs(KERNEL_DS);
846         /*
847          * the following is safe, since for compiler definitions of kvec and
848          * iovec are identical, yielding the same in-core layout and alignment
849          */
850         msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
851         result = sock_recvmsg(sock, msg, size, flags);
852         set_fs(oldfs);
853         return result;
854 }
855 EXPORT_SYMBOL(kernel_recvmsg);
856 
857 static void sock_aio_dtor(struct kiocb *iocb)
858 {
859         kfree(iocb->private);
860 }
861 
862 static ssize_t sock_sendpage(struct file *file, struct page *page,
863                              int offset, size_t size, loff_t *ppos, int more)
864 {
865         struct socket *sock;
866         int flags;
867 
868         sock = file->private_data;
869 
870         flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
871         /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
872         flags |= more;
873 
874         return kernel_sendpage(sock, page, offset, size, flags);
875 }
876 
877 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
878                                 struct pipe_inode_info *pipe, size_t len,
879                                 unsigned int flags)
880 {
881         struct socket *sock = file->private_data;
882 
883         if (unlikely(!sock->ops->splice_read))
884                 return -EINVAL;
885 
886         return sock->ops->splice_read(sock, ppos, pipe, len, flags);
887 }
888 
889 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
890                                          struct sock_iocb *siocb)
891 {
892         if (!is_sync_kiocb(iocb)) {
893                 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
894                 if (!siocb)
895                         return NULL;
896                 iocb->ki_dtor = sock_aio_dtor;
897         }
898 
899         siocb->kiocb = iocb;
900         iocb->private = siocb;
901         return siocb;
902 }
903 
904 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
905                 struct file *file, const struct iovec *iov,
906                 unsigned long nr_segs)
907 {
908         struct socket *sock = file->private_data;
909         size_t size = 0;
910         int i;
911 
912         for (i = 0; i < nr_segs; i++)
913                 size += iov[i].iov_len;
914 
915         msg->msg_name = NULL;
916         msg->msg_namelen = 0;
917         msg->msg_control = NULL;
918         msg->msg_controllen = 0;
919         msg->msg_iov = (struct iovec *)iov;
920         msg->msg_iovlen = nr_segs;
921         msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
922 
923         return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
924 }
925 
926 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
927                                 unsigned long nr_segs, loff_t pos)
928 {
929         struct sock_iocb siocb, *x;
930 
931         if (pos != 0)
932                 return -ESPIPE;
933 
934         if (iocb->ki_left == 0) /* Match SYS5 behaviour */
935                 return 0;
936 
937 
938         x = alloc_sock_iocb(iocb, &siocb);
939         if (!x)
940                 return -ENOMEM;
941         return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
942 }
943 
944 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
945                         struct file *file, const struct iovec *iov,
946                         unsigned long nr_segs)
947 {
948         struct socket *sock = file->private_data;
949         size_t size = 0;
950         int i;
951 
952         for (i = 0; i < nr_segs; i++)
953                 size += iov[i].iov_len;
954 
955         msg->msg_name = NULL;
956         msg->msg_namelen = 0;
957         msg->msg_control = NULL;
958         msg->msg_controllen = 0;
959         msg->msg_iov = (struct iovec *)iov;
960         msg->msg_iovlen = nr_segs;
961         msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
962         if (sock->type == SOCK_SEQPACKET)
963                 msg->msg_flags |= MSG_EOR;
964 
965         return __sock_sendmsg(iocb, sock, msg, size);
966 }
967 
968 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
969                           unsigned long nr_segs, loff_t pos)
970 {
971         struct sock_iocb siocb, *x;
972 
973         if (pos != 0)
974                 return -ESPIPE;
975 
976         x = alloc_sock_iocb(iocb, &siocb);
977         if (!x)
978                 return -ENOMEM;
979 
980         return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
981 }
982 
983 /*
984  * Atomic setting of ioctl hooks to avoid race
985  * with module unload.
986  */
987 
988 static DEFINE_MUTEX(br_ioctl_mutex);
989 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
990 
991 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
992 {
993         mutex_lock(&br_ioctl_mutex);
994         br_ioctl_hook = hook;
995         mutex_unlock(&br_ioctl_mutex);
996 }
997 EXPORT_SYMBOL(brioctl_set);
998 
999 static DEFINE_MUTEX(vlan_ioctl_mutex);
1000 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1001 
1002 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1003 {
1004         mutex_lock(&vlan_ioctl_mutex);
1005         vlan_ioctl_hook = hook;
1006         mutex_unlock(&vlan_ioctl_mutex);
1007 }
1008 EXPORT_SYMBOL(vlan_ioctl_set);
1009 
1010 static DEFINE_MUTEX(dlci_ioctl_mutex);
1011 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1012 
1013 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1014 {
1015         mutex_lock(&dlci_ioctl_mutex);
1016         dlci_ioctl_hook = hook;
1017         mutex_unlock(&dlci_ioctl_mutex);
1018 }
1019 EXPORT_SYMBOL(dlci_ioctl_set);
1020 
1021 static long sock_do_ioctl(struct net *net, struct socket *sock,
1022                                  unsigned int cmd, unsigned long arg)
1023 {
1024         int err;
1025         void __user *argp = (void __user *)arg;
1026 
1027         err = sock->ops->ioctl(sock, cmd, arg);
1028 
1029         /*
1030          * If this ioctl is unknown try to hand it down
1031          * to the NIC driver.
1032          */
1033         if (err == -ENOIOCTLCMD)
1034                 err = dev_ioctl(net, cmd, argp);
1035 
1036         return err;
1037 }
1038 
1039 /*
1040  *      With an ioctl, arg may well be a user mode pointer, but we don't know
1041  *      what to do with it - that's up to the protocol still.
1042  */
1043 
1044 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1045 {
1046         struct socket *sock;
1047         struct sock *sk;
1048         void __user *argp = (void __user *)arg;
1049         int pid, err;
1050         struct net *net;
1051 
1052         sock = file->private_data;
1053         sk = sock->sk;
1054         net = sock_net(sk);
1055         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1056                 err = dev_ioctl(net, cmd, argp);
1057         } else
1058 #ifdef CONFIG_WEXT_CORE
1059         if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1060                 err = dev_ioctl(net, cmd, argp);
1061         } else
1062 #endif
1063                 switch (cmd) {
1064                 case FIOSETOWN:
1065                 case SIOCSPGRP:
1066                         err = -EFAULT;
1067                         if (get_user(pid, (int __user *)argp))
1068                                 break;
1069                         err = f_setown(sock->file, pid, 1);
1070                         break;
1071                 case FIOGETOWN:
1072                 case SIOCGPGRP:
1073                         err = put_user(f_getown(sock->file),
1074                                        (int __user *)argp);
1075                         break;
1076                 case SIOCGIFBR:
1077                 case SIOCSIFBR:
1078                 case SIOCBRADDBR:
1079                 case SIOCBRDELBR:
1080                         err = -ENOPKG;
1081                         if (!br_ioctl_hook)
1082                                 request_module("bridge");
1083 
1084                         mutex_lock(&br_ioctl_mutex);
1085                         if (br_ioctl_hook)
1086                                 err = br_ioctl_hook(net, cmd, argp);
1087                         mutex_unlock(&br_ioctl_mutex);
1088                         break;
1089                 case SIOCGIFVLAN:
1090                 case SIOCSIFVLAN:
1091                         err = -ENOPKG;
1092                         if (!vlan_ioctl_hook)
1093                                 request_module("8021q");
1094 
1095                         mutex_lock(&vlan_ioctl_mutex);
1096                         if (vlan_ioctl_hook)
1097                                 err = vlan_ioctl_hook(net, argp);
1098                         mutex_unlock(&vlan_ioctl_mutex);
1099                         break;
1100                 case SIOCADDDLCI:
1101                 case SIOCDELDLCI:
1102                         err = -ENOPKG;
1103                         if (!dlci_ioctl_hook)
1104                                 request_module("dlci");
1105 
1106                         mutex_lock(&dlci_ioctl_mutex);
1107                         if (dlci_ioctl_hook)
1108                                 err = dlci_ioctl_hook(cmd, argp);
1109                         mutex_unlock(&dlci_ioctl_mutex);
1110                         break;
1111                 default:
1112                         err = sock_do_ioctl(net, sock, cmd, arg);
1113                         break;
1114                 }
1115         return err;
1116 }
1117 
1118 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1119 {
1120         int err;
1121         struct socket *sock = NULL;
1122 
1123         err = security_socket_create(family, type, protocol, 1);
1124         if (err)
1125                 goto out;
1126 
1127         sock = sock_alloc();
1128         if (!sock) {
1129                 err = -ENOMEM;
1130                 goto out;
1131         }
1132 
1133         sock->type = type;
1134         err = security_socket_post_create(sock, family, type, protocol, 1);
1135         if (err)
1136                 goto out_release;
1137 
1138 out:
1139         *res = sock;
1140         return err;
1141 out_release:
1142         sock_release(sock);
1143         sock = NULL;
1144         goto out;
1145 }
1146 EXPORT_SYMBOL(sock_create_lite);
1147 
1148 /* No kernel lock held - perfect */
1149 static unsigned int sock_poll(struct file *file, poll_table *wait)
1150 {
1151         unsigned int busy_flag = 0;
1152         struct socket *sock;
1153 
1154         /*
1155          *      We can't return errors to poll, so it's either yes or no.
1156          */
1157         sock = file->private_data;
1158 
1159         if (sk_can_busy_loop(sock->sk)) {
1160                 /* this socket can poll_ll so tell the system call */
1161                 busy_flag = POLL_BUSY_LOOP;
1162 
1163                 /* once, only if requested by syscall */
1164                 if (wait && (wait->_key & POLL_BUSY_LOOP))
1165                         sk_busy_loop(sock->sk, 1);
1166         }
1167 
1168         return busy_flag | sock->ops->poll(file, sock, wait);
1169 }
1170 
1171 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1172 {
1173         struct socket *sock = file->private_data;
1174 
1175         return sock->ops->mmap(file, sock, vma);
1176 }
1177 
1178 static int sock_close(struct inode *inode, struct file *filp)
1179 {
1180         sock_release(SOCKET_I(inode));
1181         return 0;
1182 }
1183 
1184 /*
1185  *      Update the socket async list
1186  *
1187  *      Fasync_list locking strategy.
1188  *
1189  *      1. fasync_list is modified only under process context socket lock
1190  *         i.e. under semaphore.
1191  *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1192  *         or under socket lock
1193  */
1194 
1195 static int sock_fasync(int fd, struct file *filp, int on)
1196 {
1197         struct socket *sock = filp->private_data;
1198         struct sock *sk = sock->sk;
1199         struct socket_wq *wq;
1200 
1201         if (sk == NULL)
1202                 return -EINVAL;
1203 
1204         lock_sock(sk);
1205         wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1206         fasync_helper(fd, filp, on, &wq->fasync_list);
1207 
1208         if (!wq->fasync_list)
1209                 sock_reset_flag(sk, SOCK_FASYNC);
1210         else
1211                 sock_set_flag(sk, SOCK_FASYNC);
1212 
1213         release_sock(sk);
1214         return 0;
1215 }
1216 
1217 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1218 
1219 int sock_wake_async(struct socket *sock, int how, int band)
1220 {
1221         struct socket_wq *wq;
1222 
1223         if (!sock)
1224                 return -1;
1225         rcu_read_lock();
1226         wq = rcu_dereference(sock->wq);
1227         if (!wq || !wq->fasync_list) {
1228                 rcu_read_unlock();
1229                 return -1;
1230         }
1231         switch (how) {
1232         case SOCK_WAKE_WAITD:
1233                 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1234                         break;
1235                 goto call_kill;
1236         case SOCK_WAKE_SPACE:
1237                 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1238                         break;
1239                 /* fall through */
1240         case SOCK_WAKE_IO:
1241 call_kill:
1242                 kill_fasync(&wq->fasync_list, SIGIO, band);
1243                 break;
1244         case SOCK_WAKE_URG:
1245                 kill_fasync(&wq->fasync_list, SIGURG, band);
1246         }
1247         rcu_read_unlock();
1248         return 0;
1249 }
1250 EXPORT_SYMBOL(sock_wake_async);
1251 
1252 int __sock_create(struct net *net, int family, int type, int protocol,
1253                          struct socket **res, int kern)
1254 {
1255         int err;
1256         struct socket *sock;
1257         const struct net_proto_family *pf;
1258 
1259         /*
1260          *      Check protocol is in range
1261          */
1262         if (family < 0 || family >= NPROTO)
1263                 return -EAFNOSUPPORT;
1264         if (type < 0 || type >= SOCK_MAX)
1265                 return -EINVAL;
1266 
1267         /* Compatibility.
1268 
1269            This uglymoron is moved from INET layer to here to avoid
1270            deadlock in module load.
1271          */
1272         if (family == PF_INET && type == SOCK_PACKET) {
1273                 static int warned;
1274                 if (!warned) {
1275                         warned = 1;
1276                         printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1277                                current->comm);
1278                 }
1279                 family = PF_PACKET;
1280         }
1281 
1282         err = security_socket_create(family, type, protocol, kern);
1283         if (err)
1284                 return err;
1285 
1286         /*
1287          *      Allocate the socket and allow the family to set things up. if
1288          *      the protocol is 0, the family is instructed to select an appropriate
1289          *      default.
1290          */
1291         sock = sock_alloc();
1292         if (!sock) {
1293                 net_warn_ratelimited("socket: no more sockets\n");
1294                 return -ENFILE; /* Not exactly a match, but its the
1295                                    closest posix thing */
1296         }
1297 
1298         sock->type = type;
1299 
1300 #ifdef CONFIG_MODULES
1301         /* Attempt to load a protocol module if the find failed.
1302          *
1303          * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1304          * requested real, full-featured networking support upon configuration.
1305          * Otherwise module support will break!
1306          */
1307         if (rcu_access_pointer(net_families[family]) == NULL)
1308                 request_module("net-pf-%d", family);
1309 #endif
1310 
1311         rcu_read_lock();
1312         pf = rcu_dereference(net_families[family]);
1313         err = -EAFNOSUPPORT;
1314         if (!pf)
1315                 goto out_release;
1316 
1317         /*
1318          * We will call the ->create function, that possibly is in a loadable
1319          * module, so we have to bump that loadable module refcnt first.
1320          */
1321         if (!try_module_get(pf->owner))
1322                 goto out_release;
1323 
1324         /* Now protected by module ref count */
1325         rcu_read_unlock();
1326 
1327         err = pf->create(net, sock, protocol, kern);
1328         if (err < 0)
1329                 goto out_module_put;
1330 
1331         /*
1332          * Now to bump the refcnt of the [loadable] module that owns this
1333          * socket at sock_release time we decrement its refcnt.
1334          */
1335         if (!try_module_get(sock->ops->owner))
1336                 goto out_module_busy;
1337 
1338         /*
1339          * Now that we're done with the ->create function, the [loadable]
1340          * module can have its refcnt decremented
1341          */
1342         module_put(pf->owner);
1343         err = security_socket_post_create(sock, family, type, protocol, kern);
1344         if (err)
1345                 goto out_sock_release;
1346         *res = sock;
1347 
1348         return 0;
1349 
1350 out_module_busy:
1351         err = -EAFNOSUPPORT;
1352 out_module_put:
1353         sock->ops = NULL;
1354         module_put(pf->owner);
1355 out_sock_release:
1356         sock_release(sock);
1357         return err;
1358 
1359 out_release:
1360         rcu_read_unlock();
1361         goto out_sock_release;
1362 }
1363 EXPORT_SYMBOL(__sock_create);
1364 
1365 int sock_create(int family, int type, int protocol, struct socket **res)
1366 {
1367         return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1368 }
1369 EXPORT_SYMBOL(sock_create);
1370 
1371 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1372 {
1373         return __sock_create(&init_net, family, type, protocol, res, 1);
1374 }
1375 EXPORT_SYMBOL(sock_create_kern);
1376 
1377 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1378 {
1379         int retval;
1380         struct socket *sock;
1381         int flags;
1382 
1383         /* Check the SOCK_* constants for consistency.  */
1384         BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1385         BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1386         BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1387         BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1388 
1389         flags = type & ~SOCK_TYPE_MASK;
1390         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1391                 return -EINVAL;
1392         type &= SOCK_TYPE_MASK;
1393 
1394         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1395                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1396 
1397         retval = sock_create(family, type, protocol, &sock);
1398         if (retval < 0)
1399                 goto out;
1400 
1401         retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1402         if (retval < 0)
1403                 goto out_release;
1404 
1405 out:
1406         /* It may be already another descriptor 8) Not kernel problem. */
1407         return retval;
1408 
1409 out_release:
1410         sock_release(sock);
1411         return retval;
1412 }
1413 
1414 /*
1415  *      Create a pair of connected sockets.
1416  */
1417 
1418 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1419                 int __user *, usockvec)
1420 {
1421         struct socket *sock1, *sock2;
1422         int fd1, fd2, err;
1423         struct file *newfile1, *newfile2;
1424         int flags;
1425 
1426         flags = type & ~SOCK_TYPE_MASK;
1427         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1428                 return -EINVAL;
1429         type &= SOCK_TYPE_MASK;
1430 
1431         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1432                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1433 
1434         /*
1435          * Obtain the first socket and check if the underlying protocol
1436          * supports the socketpair call.
1437          */
1438 
1439         err = sock_create(family, type, protocol, &sock1);
1440         if (err < 0)
1441                 goto out;
1442 
1443         err = sock_create(family, type, protocol, &sock2);
1444         if (err < 0)
1445                 goto out_release_1;
1446 
1447         err = sock1->ops->socketpair(sock1, sock2);
1448         if (err < 0)
1449                 goto out_release_both;
1450 
1451         fd1 = get_unused_fd_flags(flags);
1452         if (unlikely(fd1 < 0)) {
1453                 err = fd1;
1454                 goto out_release_both;
1455         }
1456         fd2 = get_unused_fd_flags(flags);
1457         if (unlikely(fd2 < 0)) {
1458                 err = fd2;
1459                 put_unused_fd(fd1);
1460                 goto out_release_both;
1461         }
1462 
1463         newfile1 = sock_alloc_file(sock1, flags, NULL);
1464         if (unlikely(IS_ERR(newfile1))) {
1465                 err = PTR_ERR(newfile1);
1466                 put_unused_fd(fd1);
1467                 put_unused_fd(fd2);
1468                 goto out_release_both;
1469         }
1470 
1471         newfile2 = sock_alloc_file(sock2, flags, NULL);
1472         if (IS_ERR(newfile2)) {
1473                 err = PTR_ERR(newfile2);
1474                 fput(newfile1);
1475                 put_unused_fd(fd1);
1476                 put_unused_fd(fd2);
1477                 sock_release(sock2);
1478                 goto out;
1479         }
1480 
1481         audit_fd_pair(fd1, fd2);
1482         fd_install(fd1, newfile1);
1483         fd_install(fd2, newfile2);
1484         /* fd1 and fd2 may be already another descriptors.
1485          * Not kernel problem.
1486          */
1487 
1488         err = put_user(fd1, &usockvec[0]);
1489         if (!err)
1490                 err = put_user(fd2, &usockvec[1]);
1491         if (!err)
1492                 return 0;
1493 
1494         sys_close(fd2);
1495         sys_close(fd1);
1496         return err;
1497 
1498 out_release_both:
1499         sock_release(sock2);
1500 out_release_1:
1501         sock_release(sock1);
1502 out:
1503         return err;
1504 }
1505 
1506 /*
1507  *      Bind a name to a socket. Nothing much to do here since it's
1508  *      the protocol's responsibility to handle the local address.
1509  *
1510  *      We move the socket address to kernel space before we call
1511  *      the protocol layer (having also checked the address is ok).
1512  */
1513 
1514 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1515 {
1516         struct socket *sock;
1517         struct sockaddr_storage address;
1518         int err, fput_needed;
1519 
1520         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1521         if (sock) {
1522                 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1523                 if (err >= 0) {
1524                         err = security_socket_bind(sock,
1525                                                    (struct sockaddr *)&address,
1526                                                    addrlen);
1527                         if (!err)
1528                                 err = sock->ops->bind(sock,
1529                                                       (struct sockaddr *)
1530                                                       &address, addrlen);
1531                 }
1532                 fput_light(sock->file, fput_needed);
1533         }
1534         return err;
1535 }
1536 
1537 /*
1538  *      Perform a listen. Basically, we allow the protocol to do anything
1539  *      necessary for a listen, and if that works, we mark the socket as
1540  *      ready for listening.
1541  */
1542 
1543 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1544 {
1545         struct socket *sock;
1546         int err, fput_needed;
1547         int somaxconn;
1548 
1549         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1550         if (sock) {
1551                 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1552                 if ((unsigned int)backlog > somaxconn)
1553                         backlog = somaxconn;
1554 
1555                 err = security_socket_listen(sock, backlog);
1556                 if (!err)
1557                         err = sock->ops->listen(sock, backlog);
1558 
1559                 fput_light(sock->file, fput_needed);
1560         }
1561         return err;
1562 }
1563 
1564 /*
1565  *      For accept, we attempt to create a new socket, set up the link
1566  *      with the client, wake up the client, then return the new
1567  *      connected fd. We collect the address of the connector in kernel
1568  *      space and move it to user at the very end. This is unclean because
1569  *      we open the socket then return an error.
1570  *
1571  *      1003.1g adds the ability to recvmsg() to query connection pending
1572  *      status to recvmsg. We need to add that support in a way thats
1573  *      clean when we restucture accept also.
1574  */
1575 
1576 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1577                 int __user *, upeer_addrlen, int, flags)
1578 {
1579         struct socket *sock, *newsock;
1580         struct file *newfile;
1581         int err, len, newfd, fput_needed;
1582         struct sockaddr_storage address;
1583 
1584         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1585                 return -EINVAL;
1586 
1587         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1588                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1589 
1590         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1591         if (!sock)
1592                 goto out;
1593 
1594         err = -ENFILE;
1595         newsock = sock_alloc();
1596         if (!newsock)
1597                 goto out_put;
1598 
1599         newsock->type = sock->type;
1600         newsock->ops = sock->ops;
1601 
1602         /*
1603          * We don't need try_module_get here, as the listening socket (sock)
1604          * has the protocol module (sock->ops->owner) held.
1605          */
1606         __module_get(newsock->ops->owner);
1607 
1608         newfd = get_unused_fd_flags(flags);
1609         if (unlikely(newfd < 0)) {
1610                 err = newfd;
1611                 sock_release(newsock);
1612                 goto out_put;
1613         }
1614         newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1615         if (unlikely(IS_ERR(newfile))) {
1616                 err = PTR_ERR(newfile);
1617                 put_unused_fd(newfd);
1618                 sock_release(newsock);
1619                 goto out_put;
1620         }
1621 
1622         err = security_socket_accept(sock, newsock);
1623         if (err)
1624                 goto out_fd;
1625 
1626         err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1627         if (err < 0)
1628                 goto out_fd;
1629 
1630         if (ccs_socket_post_accept_permission(sock, newsock)) {
1631                 err = -EAGAIN; /* Hope less harmful than -EPERM. */
1632                 goto out_fd;
1633         }
1634         if (upeer_sockaddr) {
1635                 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1636                                           &len, 2) < 0) {
1637                         err = -ECONNABORTED;
1638                         goto out_fd;
1639                 }
1640                 err = move_addr_to_user(&address,
1641                                         len, upeer_sockaddr, upeer_addrlen);
1642                 if (err < 0)
1643                         goto out_fd;
1644         }
1645 
1646         /* File flags are not inherited via accept() unlike another OSes. */
1647 
1648         fd_install(newfd, newfile);
1649         err = newfd;
1650 
1651 out_put:
1652         fput_light(sock->file, fput_needed);
1653 out:
1654         return err;
1655 out_fd:
1656         fput(newfile);
1657         put_unused_fd(newfd);
1658         goto out_put;
1659 }
1660 
1661 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1662                 int __user *, upeer_addrlen)
1663 {
1664         return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1665 }
1666 
1667 /*
1668  *      Attempt to connect to a socket with the server address.  The address
1669  *      is in user space so we verify it is OK and move it to kernel space.
1670  *
1671  *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1672  *      break bindings
1673  *
1674  *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1675  *      other SEQPACKET protocols that take time to connect() as it doesn't
1676  *      include the -EINPROGRESS status for such sockets.
1677  */
1678 
1679 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1680                 int, addrlen)
1681 {
1682         struct socket *sock;
1683         struct sockaddr_storage address;
1684         int err, fput_needed;
1685 
1686         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1687         if (!sock)
1688                 goto out;
1689         err = move_addr_to_kernel(uservaddr, addrlen, &address);
1690         if (err < 0)
1691                 goto out_put;
1692 
1693         err =
1694             security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1695         if (err)
1696                 goto out_put;
1697 
1698         err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1699                                  sock->file->f_flags);
1700 out_put:
1701         fput_light(sock->file, fput_needed);
1702 out:
1703         return err;
1704 }
1705 
1706 /*
1707  *      Get the local address ('name') of a socket object. Move the obtained
1708  *      name to user space.
1709  */
1710 
1711 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1712                 int __user *, usockaddr_len)
1713 {
1714         struct socket *sock;
1715         struct sockaddr_storage address;
1716         int len, err, fput_needed;
1717 
1718         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1719         if (!sock)
1720                 goto out;
1721 
1722         err = security_socket_getsockname(sock);
1723         if (err)
1724                 goto out_put;
1725 
1726         err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1727         if (err)
1728                 goto out_put;
1729         err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1730 
1731 out_put:
1732         fput_light(sock->file, fput_needed);
1733 out:
1734         return err;
1735 }
1736 
1737 /*
1738  *      Get the remote address ('name') of a socket object. Move the obtained
1739  *      name to user space.
1740  */
1741 
1742 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1743                 int __user *, usockaddr_len)
1744 {
1745         struct socket *sock;
1746         struct sockaddr_storage address;
1747         int len, err, fput_needed;
1748 
1749         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1750         if (sock != NULL) {
1751                 err = security_socket_getpeername(sock);
1752                 if (err) {
1753                         fput_light(sock->file, fput_needed);
1754                         return err;
1755                 }
1756 
1757                 err =
1758                     sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1759                                        1);
1760                 if (!err)
1761                         err = move_addr_to_user(&address, len, usockaddr,
1762                                                 usockaddr_len);
1763                 fput_light(sock->file, fput_needed);
1764         }
1765         return err;
1766 }
1767 
1768 /*
1769  *      Send a datagram to a given address. We move the address into kernel
1770  *      space and check the user space data area is readable before invoking
1771  *      the protocol.
1772  */
1773 
1774 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1775                 unsigned int, flags, struct sockaddr __user *, addr,
1776                 int, addr_len)
1777 {
1778         struct socket *sock;
1779         struct sockaddr_storage address;
1780         int err;
1781         struct msghdr msg;
1782         struct iovec iov;
1783         int fput_needed;
1784 
1785         if (len > INT_MAX)
1786                 len = INT_MAX;
1787         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1788         if (!sock)
1789                 goto out;
1790 
1791         iov.iov_base = buff;
1792         iov.iov_len = len;
1793         msg.msg_name = NULL;
1794         msg.msg_iov = &iov;
1795         msg.msg_iovlen = 1;
1796         msg.msg_control = NULL;
1797         msg.msg_controllen = 0;
1798         msg.msg_namelen = 0;
1799         if (addr) {
1800                 err = move_addr_to_kernel(addr, addr_len, &address);
1801                 if (err < 0)
1802                         goto out_put;
1803                 msg.msg_name = (struct sockaddr *)&address;
1804                 msg.msg_namelen = addr_len;
1805         }
1806         if (sock->file->f_flags & O_NONBLOCK)
1807                 flags |= MSG_DONTWAIT;
1808         msg.msg_flags = flags;
1809         err = sock_sendmsg(sock, &msg, len);
1810 
1811 out_put:
1812         fput_light(sock->file, fput_needed);
1813 out:
1814         return err;
1815 }
1816 
1817 /*
1818  *      Send a datagram down a socket.
1819  */
1820 
1821 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1822                 unsigned int, flags)
1823 {
1824         return sys_sendto(fd, buff, len, flags, NULL, 0);
1825 }
1826 
1827 /*
1828  *      Receive a frame from the socket and optionally record the address of the
1829  *      sender. We verify the buffers are writable and if needed move the
1830  *      sender address from kernel to user space.
1831  */
1832 
1833 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1834                 unsigned int, flags, struct sockaddr __user *, addr,
1835                 int __user *, addr_len)
1836 {
1837         struct socket *sock;
1838         struct iovec iov;
1839         struct msghdr msg;
1840         struct sockaddr_storage address;
1841         int err, err2;
1842         int fput_needed;
1843 
1844         if (size > INT_MAX)
1845                 size = INT_MAX;
1846         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1847         if (!sock)
1848                 goto out;
1849 
1850         msg.msg_control = NULL;
1851         msg.msg_controllen = 0;
1852         msg.msg_iovlen = 1;
1853         msg.msg_iov = &iov;
1854         iov.iov_len = size;
1855         iov.iov_base = ubuf;
1856         msg.msg_name = (struct sockaddr *)&address;
1857         msg.msg_namelen = sizeof(address);
1858         if (sock->file->f_flags & O_NONBLOCK)
1859                 flags |= MSG_DONTWAIT;
1860         err = sock_recvmsg(sock, &msg, size, flags);
1861 
1862         if (err >= 0 && addr != NULL) {
1863                 err2 = move_addr_to_user(&address,
1864                                          msg.msg_namelen, addr, addr_len);
1865                 if (err2 < 0)
1866                         err = err2;
1867         }
1868 
1869         fput_light(sock->file, fput_needed);
1870 out:
1871         return err;
1872 }
1873 
1874 /*
1875  *      Receive a datagram from a socket.
1876  */
1877 
1878 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1879                          unsigned int flags)
1880 {
1881         return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1882 }
1883 
1884 /*
1885  *      Set a socket option. Because we don't know the option lengths we have
1886  *      to pass the user mode parameter for the protocols to sort out.
1887  */
1888 
1889 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1890                 char __user *, optval, int, optlen)
1891 {
1892         int err, fput_needed;
1893         struct socket *sock;
1894 
1895         if (optlen < 0)
1896                 return -EINVAL;
1897 
1898         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1899         if (sock != NULL) {
1900                 err = security_socket_setsockopt(sock, level, optname);
1901                 if (err)
1902                         goto out_put;
1903 
1904                 if (level == SOL_SOCKET)
1905                         err =
1906                             sock_setsockopt(sock, level, optname, optval,
1907                                             optlen);
1908                 else
1909                         err =
1910                             sock->ops->setsockopt(sock, level, optname, optval,
1911                                                   optlen);
1912 out_put:
1913                 fput_light(sock->file, fput_needed);
1914         }
1915         return err;
1916 }
1917 
1918 /*
1919  *      Get a socket option. Because we don't know the option lengths we have
1920  *      to pass a user mode parameter for the protocols to sort out.
1921  */
1922 
1923 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1924                 char __user *, optval, int __user *, optlen)
1925 {
1926         int err, fput_needed;
1927         struct socket *sock;
1928 
1929         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1930         if (sock != NULL) {
1931                 err = security_socket_getsockopt(sock, level, optname);
1932                 if (err)
1933                         goto out_put;
1934 
1935                 if (level == SOL_SOCKET)
1936                         err =
1937                             sock_getsockopt(sock, level, optname, optval,
1938                                             optlen);
1939                 else
1940                         err =
1941                             sock->ops->getsockopt(sock, level, optname, optval,
1942                                                   optlen);
1943 out_put:
1944                 fput_light(sock->file, fput_needed);
1945         }
1946         return err;
1947 }
1948 
1949 /*
1950  *      Shutdown a socket.
1951  */
1952 
1953 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1954 {
1955         int err, fput_needed;
1956         struct socket *sock;
1957 
1958         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1959         if (sock != NULL) {
1960                 err = security_socket_shutdown(sock, how);
1961                 if (!err)
1962                         err = sock->ops->shutdown(sock, how);
1963                 fput_light(sock->file, fput_needed);
1964         }
1965         return err;
1966 }
1967 
1968 /* A couple of helpful macros for getting the address of the 32/64 bit
1969  * fields which are the same type (int / unsigned) on our platforms.
1970  */
1971 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1972 #define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1973 #define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1974 
1975 struct used_address {
1976         struct sockaddr_storage name;
1977         unsigned int name_len;
1978 };
1979 
1980 static int copy_msghdr_from_user(struct msghdr *kmsg,
1981                                  struct msghdr __user *umsg)
1982 {
1983         if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1984                 return -EFAULT;
1985         if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1986                 return -EINVAL;
1987         return 0;
1988 }
1989 
1990 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1991                          struct msghdr *msg_sys, unsigned int flags,
1992                          struct used_address *used_address)
1993 {
1994         struct compat_msghdr __user *msg_compat =
1995             (struct compat_msghdr __user *)msg;
1996         struct sockaddr_storage address;
1997         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1998         unsigned char ctl[sizeof(struct cmsghdr) + 20]
1999             __attribute__ ((aligned(sizeof(__kernel_size_t))));
2000         /* 20 is size of ipv6_pktinfo */
2001         unsigned char *ctl_buf = ctl;
2002         int err, ctl_len, total_len;
2003 
2004         err = -EFAULT;
2005         if (MSG_CMSG_COMPAT & flags) {
2006                 if (get_compat_msghdr(msg_sys, msg_compat))
2007                         return -EFAULT;
2008         } else {
2009                 err = copy_msghdr_from_user(msg_sys, msg);
2010                 if (err)
2011                         return err;
2012         }
2013 
2014         if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2015                 err = -EMSGSIZE;
2016                 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2017                         goto out;
2018                 err = -ENOMEM;
2019                 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2020                               GFP_KERNEL);
2021                 if (!iov)
2022                         goto out;
2023         }
2024 
2025         /* This will also move the address data into kernel space */
2026         if (MSG_CMSG_COMPAT & flags) {
2027                 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2028         } else
2029                 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2030         if (err < 0)
2031                 goto out_freeiov;
2032         total_len = err;
2033 
2034         err = -ENOBUFS;
2035 
2036         if (msg_sys->msg_controllen > INT_MAX)
2037                 goto out_freeiov;
2038         ctl_len = msg_sys->msg_controllen;
2039         if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2040                 err =
2041                     cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2042                                                      sizeof(ctl));
2043                 if (err)
2044                         goto out_freeiov;
2045                 ctl_buf = msg_sys->msg_control;
2046                 ctl_len = msg_sys->msg_controllen;
2047         } else if (ctl_len) {
2048                 if (ctl_len > sizeof(ctl)) {
2049                         ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2050                         if (ctl_buf == NULL)
2051                                 goto out_freeiov;
2052                 }
2053                 err = -EFAULT;
2054                 /*
2055                  * Careful! Before this, msg_sys->msg_control contains a user pointer.
2056                  * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2057                  * checking falls down on this.
2058                  */
2059                 if (copy_from_user(ctl_buf,
2060                                    (void __user __force *)msg_sys->msg_control,
2061                                    ctl_len))
2062                         goto out_freectl;
2063                 msg_sys->msg_control = ctl_buf;
2064         }
2065         msg_sys->msg_flags = flags;
2066 
2067         if (sock->file->f_flags & O_NONBLOCK)
2068                 msg_sys->msg_flags |= MSG_DONTWAIT;
2069         /*
2070          * If this is sendmmsg() and current destination address is same as
2071          * previously succeeded address, omit asking LSM's decision.
2072          * used_address->name_len is initialized to UINT_MAX so that the first
2073          * destination address never matches.
2074          */
2075         if (used_address && msg_sys->msg_name &&
2076             used_address->name_len == msg_sys->msg_namelen &&
2077             !memcmp(&used_address->name, msg_sys->msg_name,
2078                     used_address->name_len)) {
2079                 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2080                 goto out_freectl;
2081         }
2082         err = sock_sendmsg(sock, msg_sys, total_len);
2083         /*
2084          * If this is sendmmsg() and sending to current destination address was
2085          * successful, remember it.
2086          */
2087         if (used_address && err >= 0) {
2088                 used_address->name_len = msg_sys->msg_namelen;
2089                 if (msg_sys->msg_name)
2090                         memcpy(&used_address->name, msg_sys->msg_name,
2091                                used_address->name_len);
2092         }
2093 
2094 out_freectl:
2095         if (ctl_buf != ctl)
2096                 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2097 out_freeiov:
2098         if (iov != iovstack)
2099                 kfree(iov);
2100 out:
2101         return err;
2102 }
2103 
2104 /*
2105  *      BSD sendmsg interface
2106  */
2107 
2108 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2109 {
2110         int fput_needed, err;
2111         struct msghdr msg_sys;
2112         struct socket *sock;
2113 
2114         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2115         if (!sock)
2116                 goto out;
2117 
2118         err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2119 
2120         fput_light(sock->file, fput_needed);
2121 out:
2122         return err;
2123 }
2124 
2125 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2126 {
2127         if (flags & MSG_CMSG_COMPAT)
2128                 return -EINVAL;
2129         return __sys_sendmsg(fd, msg, flags);
2130 }
2131 
2132 /*
2133  *      Linux sendmmsg interface
2134  */
2135 
2136 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2137                    unsigned int flags)
2138 {
2139         int fput_needed, err, datagrams;
2140         struct socket *sock;
2141         struct mmsghdr __user *entry;
2142         struct compat_mmsghdr __user *compat_entry;
2143         struct msghdr msg_sys;
2144         struct used_address used_address;
2145 
2146         if (vlen > UIO_MAXIOV)
2147                 vlen = UIO_MAXIOV;
2148 
2149         datagrams = 0;
2150 
2151         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2152         if (!sock)
2153                 return err;
2154 
2155         used_address.name_len = UINT_MAX;
2156         entry = mmsg;
2157         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2158         err = 0;
2159 
2160         while (datagrams < vlen) {
2161                 if (MSG_CMSG_COMPAT & flags) {
2162                         err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2163                                              &msg_sys, flags, &used_address);
2164                         if (err < 0)
2165                                 break;
2166                         err = __put_user(err, &compat_entry->msg_len);
2167                         ++compat_entry;
2168                 } else {
2169                         err = ___sys_sendmsg(sock,
2170                                              (struct msghdr __user *)entry,
2171                                              &msg_sys, flags, &used_address);
2172                         if (err < 0)
2173                                 break;
2174                         err = put_user(err, &entry->msg_len);
2175                         ++entry;
2176                 }
2177 
2178                 if (err)
2179                         break;
2180                 ++datagrams;
2181         }
2182 
2183         fput_light(sock->file, fput_needed);
2184 
2185         /* We only return an error if no datagrams were able to be sent */
2186         if (datagrams != 0)
2187                 return datagrams;
2188 
2189         return err;
2190 }
2191 
2192 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2193                 unsigned int, vlen, unsigned int, flags)
2194 {
2195         if (flags & MSG_CMSG_COMPAT)
2196                 return -EINVAL;
2197         return __sys_sendmmsg(fd, mmsg, vlen, flags);
2198 }
2199 
2200 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2201                          struct msghdr *msg_sys, unsigned int flags, int nosec)
2202 {
2203         struct compat_msghdr __user *msg_compat =
2204             (struct compat_msghdr __user *)msg;
2205         struct iovec iovstack[UIO_FASTIOV];
2206         struct iovec *iov = iovstack;
2207         unsigned long cmsg_ptr;
2208         int err, total_len, len;
2209 
2210         /* kernel mode address */
2211         struct sockaddr_storage addr;
2212 
2213         /* user mode address pointers */
2214         struct sockaddr __user *uaddr;
2215         int __user *uaddr_len;
2216 
2217         if (MSG_CMSG_COMPAT & flags) {
2218                 if (get_compat_msghdr(msg_sys, msg_compat))
2219                         return -EFAULT;
2220         } else {
2221                 err = copy_msghdr_from_user(msg_sys, msg);
2222                 if (err)
2223                         return err;
2224         }
2225 
2226         if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2227                 err = -EMSGSIZE;
2228                 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2229                         goto out;
2230                 err = -ENOMEM;
2231                 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2232                               GFP_KERNEL);
2233                 if (!iov)
2234                         goto out;
2235         }
2236 
2237         /*
2238          *      Save the user-mode address (verify_iovec will change the
2239          *      kernel msghdr to use the kernel address space)
2240          */
2241 
2242         uaddr = (__force void __user *)msg_sys->msg_name;
2243         uaddr_len = COMPAT_NAMELEN(msg);
2244         if (MSG_CMSG_COMPAT & flags) {
2245                 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2246         } else
2247                 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2248         if (err < 0)
2249                 goto out_freeiov;
2250         total_len = err;
2251 
2252         cmsg_ptr = (unsigned long)msg_sys->msg_control;
2253         msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2254 
2255         if (sock->file->f_flags & O_NONBLOCK)
2256                 flags |= MSG_DONTWAIT;
2257         err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2258                                                           total_len, flags);
2259         if (err < 0)
2260                 goto out_freeiov;
2261         len = err;
2262 
2263         if (uaddr != NULL) {
2264                 err = move_addr_to_user(&addr,
2265                                         msg_sys->msg_namelen, uaddr,
2266                                         uaddr_len);
2267                 if (err < 0)
2268                         goto out_freeiov;
2269         }
2270         err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2271                          COMPAT_FLAGS(msg));
2272         if (err)
2273                 goto out_freeiov;
2274         if (MSG_CMSG_COMPAT & flags)
2275                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2276                                  &msg_compat->msg_controllen);
2277         else
2278                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2279                                  &msg->msg_controllen);
2280         if (err)
2281                 goto out_freeiov;
2282         err = len;
2283 
2284 out_freeiov:
2285         if (iov != iovstack)
2286                 kfree(iov);
2287 out:
2288         return err;
2289 }
2290 
2291 /*
2292  *      BSD recvmsg interface
2293  */
2294 
2295 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2296 {
2297         int fput_needed, err;
2298         struct msghdr msg_sys;
2299         struct socket *sock;
2300 
2301         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2302         if (!sock)
2303                 goto out;
2304 
2305         err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2306 
2307         fput_light(sock->file, fput_needed);
2308 out:
2309         return err;
2310 }
2311 
2312 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2313                 unsigned int, flags)
2314 {
2315         if (flags & MSG_CMSG_COMPAT)
2316                 return -EINVAL;
2317         return __sys_recvmsg(fd, msg, flags);
2318 }
2319 
2320 /*
2321  *     Linux recvmmsg interface
2322  */
2323 
2324 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2325                    unsigned int flags, struct timespec *timeout)
2326 {
2327         int fput_needed, err, datagrams;
2328         struct socket *sock;
2329         struct mmsghdr __user *entry;
2330         struct compat_mmsghdr __user *compat_entry;
2331         struct msghdr msg_sys;
2332         struct timespec end_time;
2333 
2334         if (timeout &&
2335             poll_select_set_timeout(&end_time, timeout->tv_sec,
2336                                     timeout->tv_nsec))
2337                 return -EINVAL;
2338 
2339         datagrams = 0;
2340 
2341         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2342         if (!sock)
2343                 return err;
2344 
2345         err = sock_error(sock->sk);
2346         if (err)
2347                 goto out_put;
2348 
2349         entry = mmsg;
2350         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2351 
2352         while (datagrams < vlen) {
2353                 /*
2354                  * No need to ask LSM for more than the first datagram.
2355                  */
2356                 if (MSG_CMSG_COMPAT & flags) {
2357                         err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2358                                              &msg_sys, flags & ~MSG_WAITFORONE,
2359                                              datagrams);
2360                         if (err < 0)
2361                                 break;
2362                         err = __put_user(err, &compat_entry->msg_len);
2363                         ++compat_entry;
2364                 } else {
2365                         err = ___sys_recvmsg(sock,
2366                                              (struct msghdr __user *)entry,
2367                                              &msg_sys, flags & ~MSG_WAITFORONE,
2368                                              datagrams);
2369                         if (err < 0)
2370                                 break;
2371                         err = put_user(err, &entry->msg_len);
2372                         ++entry;
2373                 }
2374 
2375                 if (err)
2376                         break;
2377                 ++datagrams;
2378 
2379                 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2380                 if (flags & MSG_WAITFORONE)
2381                         flags |= MSG_DONTWAIT;
2382 
2383                 if (timeout) {
2384                         ktime_get_ts(timeout);
2385                         *timeout = timespec_sub(end_time, *timeout);
2386                         if (timeout->tv_sec < 0) {
2387                                 timeout->tv_sec = timeout->tv_nsec = 0;
2388                                 break;
2389                         }
2390 
2391                         /* Timeout, return less than vlen datagrams */
2392                         if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2393                                 break;
2394                 }
2395 
2396                 /* Out of band data, return right away */
2397                 if (msg_sys.msg_flags & MSG_OOB)
2398                         break;
2399         }
2400 
2401 out_put:
2402         fput_light(sock->file, fput_needed);
2403 
2404         if (err == 0)
2405                 return datagrams;
2406 
2407         if (datagrams != 0) {
2408                 /*
2409                  * We may return less entries than requested (vlen) if the
2410                  * sock is non block and there aren't enough datagrams...
2411                  */
2412                 if (err != -EAGAIN) {
2413                         /*
2414                          * ... or  if recvmsg returns an error after we
2415                          * received some datagrams, where we record the
2416                          * error to return on the next call or if the
2417                          * app asks about it using getsockopt(SO_ERROR).
2418                          */
2419                         sock->sk->sk_err = -err;
2420                 }
2421 
2422                 return datagrams;
2423         }
2424 
2425         return err;
2426 }
2427 
2428 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2429                 unsigned int, vlen, unsigned int, flags,
2430                 struct timespec __user *, timeout)
2431 {
2432         int datagrams;
2433         struct timespec timeout_sys;
2434 
2435         if (flags & MSG_CMSG_COMPAT)
2436                 return -EINVAL;
2437 
2438         if (!timeout)
2439                 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2440 
2441         if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2442                 return -EFAULT;
2443 
2444         datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2445 
2446         if (datagrams > 0 &&
2447             copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2448                 datagrams = -EFAULT;
2449 
2450         return datagrams;
2451 }
2452 
2453 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2454 /* Argument list sizes for sys_socketcall */
2455 #define AL(x) ((x) * sizeof(unsigned long))
2456 static const unsigned char nargs[21] = {
2457         AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2458         AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2459         AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2460         AL(4), AL(5), AL(4)
2461 };
2462 
2463 #undef AL
2464 
2465 /*
2466  *      System call vectors.
2467  *
2468  *      Argument checking cleaned up. Saved 20% in size.
2469  *  This function doesn't need to set the kernel lock because
2470  *  it is set by the callees.
2471  */
2472 
2473 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2474 {
2475         unsigned long a[AUDITSC_ARGS];
2476         unsigned long a0, a1;
2477         int err;
2478         unsigned int len;
2479 
2480         if (call < 1 || call > SYS_SENDMMSG)
2481                 return -EINVAL;
2482 
2483         len = nargs[call];
2484         if (len > sizeof(a))
2485                 return -EINVAL;
2486 
2487         /* copy_from_user should be SMP safe. */
2488         if (copy_from_user(a, args, len))
2489                 return -EFAULT;
2490 
2491         err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2492         if (err)
2493                 return err;
2494 
2495         a0 = a[0];
2496         a1 = a[1];
2497 
2498         switch (call) {
2499         case SYS_SOCKET:
2500                 err = sys_socket(a0, a1, a[2]);
2501                 break;
2502         case SYS_BIND:
2503                 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2504                 break;
2505         case SYS_CONNECT:
2506                 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2507                 break;
2508         case SYS_LISTEN:
2509                 err = sys_listen(a0, a1);
2510                 break;
2511         case SYS_ACCEPT:
2512                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2513                                   (int __user *)a[2], 0);
2514                 break;
2515         case SYS_GETSOCKNAME:
2516                 err =
2517                     sys_getsockname(a0, (struct sockaddr __user *)a1,
2518                                     (int __user *)a[2]);
2519                 break;
2520         case SYS_GETPEERNAME:
2521                 err =
2522                     sys_getpeername(a0, (struct sockaddr __user *)a1,
2523                                     (int __user *)a[2]);
2524                 break;
2525         case SYS_SOCKETPAIR:
2526                 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2527                 break;
2528         case SYS_SEND:
2529                 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2530                 break;
2531         case SYS_SENDTO:
2532                 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2533                                  (struct sockaddr __user *)a[4], a[5]);
2534                 break;
2535         case SYS_RECV:
2536                 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2537                 break;
2538         case SYS_RECVFROM:
2539                 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2540                                    (struct sockaddr __user *)a[4],
2541                                    (int __user *)a[5]);
2542                 break;
2543         case SYS_SHUTDOWN:
2544                 err = sys_shutdown(a0, a1);
2545                 break;
2546         case SYS_SETSOCKOPT:
2547                 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2548                 break;
2549         case SYS_GETSOCKOPT:
2550                 err =
2551                     sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2552                                    (int __user *)a[4]);
2553                 break;
2554         case SYS_SENDMSG:
2555                 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2556                 break;
2557         case SYS_SENDMMSG:
2558                 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2559                 break;
2560         case SYS_RECVMSG:
2561                 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2562                 break;
2563         case SYS_RECVMMSG:
2564                 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2565                                    (struct timespec __user *)a[4]);
2566                 break;
2567         case SYS_ACCEPT4:
2568                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2569                                   (int __user *)a[2], a[3]);
2570                 break;
2571         default:
2572                 err = -EINVAL;
2573                 break;
2574         }
2575         return err;
2576 }
2577 
2578 #endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2579 
2580 /**
2581  *      sock_register - add a socket protocol handler
2582  *      @ops: description of protocol
2583  *
2584  *      This function is called by a protocol handler that wants to
2585  *      advertise its address family, and have it linked into the
2586  *      socket interface. The value ops->family coresponds to the
2587  *      socket system call protocol family.
2588  */
2589 int sock_register(const struct net_proto_family *ops)
2590 {
2591         int err;
2592 
2593         if (ops->family >= NPROTO) {
2594                 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2595                        NPROTO);
2596                 return -ENOBUFS;
2597         }
2598 
2599         spin_lock(&net_family_lock);
2600         if (rcu_dereference_protected(net_families[ops->family],
2601                                       lockdep_is_held(&net_family_lock)))
2602                 err = -EEXIST;
2603         else {
2604                 rcu_assign_pointer(net_families[ops->family], ops);
2605                 err = 0;
2606         }
2607         spin_unlock(&net_family_lock);
2608 
2609         printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2610         return err;
2611 }
2612 EXPORT_SYMBOL(sock_register);
2613 
2614 /**
2615  *      sock_unregister - remove a protocol handler
2616  *      @family: protocol family to remove
2617  *
2618  *      This function is called by a protocol handler that wants to
2619  *      remove its address family, and have it unlinked from the
2620  *      new socket creation.
2621  *
2622  *      If protocol handler is a module, then it can use module reference
2623  *      counts to protect against new references. If protocol handler is not
2624  *      a module then it needs to provide its own protection in
2625  *      the ops->create routine.
2626  */
2627 void sock_unregister(int family)
2628 {
2629         BUG_ON(family < 0 || family >= NPROTO);
2630 
2631         spin_lock(&net_family_lock);
2632         RCU_INIT_POINTER(net_families[family], NULL);
2633         spin_unlock(&net_family_lock);
2634 
2635         synchronize_rcu();
2636 
2637         printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2638 }
2639 EXPORT_SYMBOL(sock_unregister);
2640 
2641 static int __init sock_init(void)
2642 {
2643         int err;
2644         /*
2645          *      Initialize the network sysctl infrastructure.
2646          */
2647         err = net_sysctl_init();
2648         if (err)
2649                 goto out;
2650 
2651         /*
2652          *      Initialize skbuff SLAB cache
2653          */
2654         skb_init();
2655 
2656         /*
2657          *      Initialize the protocols module.
2658          */
2659 
2660         init_inodecache();
2661 
2662         err = register_filesystem(&sock_fs_type);
2663         if (err)
2664                 goto out_fs;
2665         sock_mnt = kern_mount(&sock_fs_type);
2666         if (IS_ERR(sock_mnt)) {
2667                 err = PTR_ERR(sock_mnt);
2668                 goto out_mount;
2669         }
2670 
2671         /* The real protocol initialization is performed in later initcalls.
2672          */
2673 
2674 #ifdef CONFIG_NETFILTER
2675         err = netfilter_init();
2676         if (err)
2677                 goto out;
2678 #endif
2679 
2680 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2681         skb_timestamping_init();
2682 #endif
2683 
2684 out:
2685         return err;
2686 
2687 out_mount:
2688         unregister_filesystem(&sock_fs_type);
2689 out_fs:
2690         goto out;
2691 }
2692 
2693 core_initcall(sock_init);       /* early initcall */
2694 
2695 #ifdef CONFIG_PROC_FS
2696 void socket_seq_show(struct seq_file *seq)
2697 {
2698         int cpu;
2699         int counter = 0;
2700 
2701         for_each_possible_cpu(cpu)
2702             counter += per_cpu(sockets_in_use, cpu);
2703 
2704         /* It can be negative, by the way. 8) */
2705         if (counter < 0)
2706                 counter = 0;
2707 
2708         seq_printf(seq, "sockets: used %d\n", counter);
2709 }
2710 #endif                          /* CONFIG_PROC_FS */
2711 
2712 #ifdef CONFIG_COMPAT
2713 static int do_siocgstamp(struct net *net, struct socket *sock,
2714                          unsigned int cmd, void __user *up)
2715 {
2716         mm_segment_t old_fs = get_fs();
2717         struct timeval ktv;
2718         int err;
2719 
2720         set_fs(KERNEL_DS);
2721         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2722         set_fs(old_fs);
2723         if (!err)
2724                 err = compat_put_timeval(&ktv, up);
2725 
2726         return err;
2727 }
2728 
2729 static int do_siocgstampns(struct net *net, struct socket *sock,
2730                            unsigned int cmd, void __user *up)
2731 {
2732         mm_segment_t old_fs = get_fs();
2733         struct timespec kts;
2734         int err;
2735 
2736         set_fs(KERNEL_DS);
2737         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2738         set_fs(old_fs);
2739         if (!err)
2740                 err = compat_put_timespec(&kts, up);
2741 
2742         return err;
2743 }
2744 
2745 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2746 {
2747         struct ifreq __user *uifr;
2748         int err;
2749 
2750         uifr = compat_alloc_user_space(sizeof(struct ifreq));
2751         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2752                 return -EFAULT;
2753 
2754         err = dev_ioctl(net, SIOCGIFNAME, uifr);
2755         if (err)
2756                 return err;
2757 
2758         if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2759                 return -EFAULT;
2760 
2761         return 0;
2762 }
2763 
2764 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2765 {
2766         struct compat_ifconf ifc32;
2767         struct ifconf ifc;
2768         struct ifconf __user *uifc;
2769         struct compat_ifreq __user *ifr32;
2770         struct ifreq __user *ifr;
2771         unsigned int i, j;
2772         int err;
2773 
2774         if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2775                 return -EFAULT;
2776 
2777         memset(&ifc, 0, sizeof(ifc));
2778         if (ifc32.ifcbuf == 0) {
2779                 ifc32.ifc_len = 0;
2780                 ifc.ifc_len = 0;
2781                 ifc.ifc_req = NULL;
2782                 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2783         } else {
2784                 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2785                         sizeof(struct ifreq);
2786                 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2787                 ifc.ifc_len = len;
2788                 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2789                 ifr32 = compat_ptr(ifc32.ifcbuf);
2790                 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2791                         if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2792                                 return -EFAULT;
2793                         ifr++;
2794                         ifr32++;
2795                 }
2796         }
2797         if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2798                 return -EFAULT;
2799 
2800         err = dev_ioctl(net, SIOCGIFCONF, uifc);
2801         if (err)
2802                 return err;
2803 
2804         if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2805                 return -EFAULT;
2806 
2807         ifr = ifc.ifc_req;
2808         ifr32 = compat_ptr(ifc32.ifcbuf);
2809         for (i = 0, j = 0;
2810              i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2811              i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2812                 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2813                         return -EFAULT;
2814                 ifr32++;
2815                 ifr++;
2816         }
2817 
2818         if (ifc32.ifcbuf == 0) {
2819                 /* Translate from 64-bit structure multiple to
2820                  * a 32-bit one.
2821                  */
2822                 i = ifc.ifc_len;
2823                 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2824                 ifc32.ifc_len = i;
2825         } else {
2826                 ifc32.ifc_len = i;
2827         }
2828         if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2829                 return -EFAULT;
2830 
2831         return 0;
2832 }
2833 
2834 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2835 {
2836         struct compat_ethtool_rxnfc __user *compat_rxnfc;
2837         bool convert_in = false, convert_out = false;
2838         size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2839         struct ethtool_rxnfc __user *rxnfc;
2840         struct ifreq __user *ifr;
2841         u32 rule_cnt = 0, actual_rule_cnt;
2842         u32 ethcmd;
2843         u32 data;
2844         int ret;
2845 
2846         if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2847                 return -EFAULT;
2848 
2849         compat_rxnfc = compat_ptr(data);
2850 
2851         if (get_user(ethcmd, &compat_rxnfc->cmd))
2852                 return -EFAULT;
2853 
2854         /* Most ethtool structures are defined without padding.
2855          * Unfortunately struct ethtool_rxnfc is an exception.
2856          */
2857         switch (ethcmd) {
2858         default:
2859                 break;
2860         case ETHTOOL_GRXCLSRLALL:
2861                 /* Buffer size is variable */
2862                 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2863                         return -EFAULT;
2864                 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2865                         return -ENOMEM;
2866                 buf_size += rule_cnt * sizeof(u32);
2867                 /* fall through */
2868         case ETHTOOL_GRXRINGS:
2869         case ETHTOOL_GRXCLSRLCNT:
2870         case ETHTOOL_GRXCLSRULE:
2871         case ETHTOOL_SRXCLSRLINS:
2872                 convert_out = true;
2873                 /* fall through */
2874         case ETHTOOL_SRXCLSRLDEL:
2875                 buf_size += sizeof(struct ethtool_rxnfc);
2876                 convert_in = true;
2877                 break;
2878         }
2879 
2880         ifr = compat_alloc_user_space(buf_size);
2881         rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2882 
2883         if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2884                 return -EFAULT;
2885 
2886         if (put_user(convert_in ? rxnfc : compat_ptr(data),
2887                      &ifr->ifr_ifru.ifru_data))
2888                 return -EFAULT;
2889 
2890         if (convert_in) {
2891                 /* We expect there to be holes between fs.m_ext and
2892                  * fs.ring_cookie and at the end of fs, but nowhere else.
2893                  */
2894                 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2895                              sizeof(compat_rxnfc->fs.m_ext) !=
2896                              offsetof(struct ethtool_rxnfc, fs.m_ext) +
2897                              sizeof(rxnfc->fs.m_ext));
2898                 BUILD_BUG_ON(
2899                         offsetof(struct compat_ethtool_rxnfc, fs.location) -
2900                         offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2901                         offsetof(struct ethtool_rxnfc, fs.location) -
2902                         offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2903 
2904                 if (copy_in_user(rxnfc, compat_rxnfc,
2905                                  (void __user *)(&rxnfc->fs.m_ext + 1) -
2906                                  (void __user *)rxnfc) ||
2907                     copy_in_user(&rxnfc->fs.ring_cookie,
2908                                  &compat_rxnfc->fs.ring_cookie,
2909                                  (void __user *)(&rxnfc->fs.location + 1) -
2910                                  (void __user *)&rxnfc->fs.ring_cookie) ||
2911                     copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2912                                  sizeof(rxnfc->rule_cnt)))
2913                         return -EFAULT;
2914         }
2915 
2916         ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2917         if (ret)
2918                 return ret;
2919 
2920         if (convert_out) {
2921                 if (copy_in_user(compat_rxnfc, rxnfc,
2922                                  (const void __user *)(&rxnfc->fs.m_ext + 1) -
2923                                  (const void __user *)rxnfc) ||
2924                     copy_in_user(&compat_rxnfc->fs.ring_cookie,
2925                                  &rxnfc->fs.ring_cookie,
2926                                  (const void __user *)(&rxnfc->fs.location + 1) -
2927                                  (const void __user *)&rxnfc->fs.ring_cookie) ||
2928                     copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2929                                  sizeof(rxnfc->rule_cnt)))
2930                         return -EFAULT;
2931 
2932                 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2933                         /* As an optimisation, we only copy the actual
2934                          * number of rules that the underlying
2935                          * function returned.  Since Mallory might
2936                          * change the rule count in user memory, we
2937                          * check that it is less than the rule count
2938                          * originally given (as the user buffer size),
2939                          * which has been range-checked.
2940                          */
2941                         if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2942                                 return -EFAULT;
2943                         if (actual_rule_cnt < rule_cnt)
2944                                 rule_cnt = actual_rule_cnt;
2945                         if (copy_in_user(&compat_rxnfc->rule_locs[0],
2946                                          &rxnfc->rule_locs[0],
2947                                          rule_cnt * sizeof(u32)))
2948                                 return -EFAULT;
2949                 }
2950         }
2951 
2952         return 0;
2953 }
2954 
2955 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2956 {
2957         void __user *uptr;
2958         compat_uptr_t uptr32;
2959         struct ifreq __user *uifr;
2960 
2961         uifr = compat_alloc_user_space(sizeof(*uifr));
2962         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2963                 return -EFAULT;
2964 
2965         if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2966                 return -EFAULT;
2967 
2968         uptr = compat_ptr(uptr32);
2969 
2970         if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2971                 return -EFAULT;
2972 
2973         return dev_ioctl(net, SIOCWANDEV, uifr);
2974 }
2975 
2976 static int bond_ioctl(struct net *net, unsigned int cmd,
2977                          struct compat_ifreq __user *ifr32)
2978 {
2979         struct ifreq kifr;
2980         struct ifreq __user *uifr;
2981         mm_segment_t old_fs;
2982         int err;
2983         u32 data;
2984         void __user *datap;
2985 
2986         switch (cmd) {
2987         case SIOCBONDENSLAVE:
2988         case SIOCBONDRELEASE:
2989         case SIOCBONDSETHWADDR:
2990         case SIOCBONDCHANGEACTIVE:
2991                 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2992                         return -EFAULT;
2993 
2994                 old_fs = get_fs();
2995                 set_fs(KERNEL_DS);
2996                 err = dev_ioctl(net, cmd,
2997                                 (struct ifreq __user __force *) &kifr);
2998                 set_fs(old_fs);
2999 
3000                 return err;
3001         case SIOCBONDSLAVEINFOQUERY:
3002         case SIOCBONDINFOQUERY:
3003                 uifr = compat_alloc_user_space(sizeof(*uifr));
3004                 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3005                         return -EFAULT;
3006 
3007                 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3008                         return -EFAULT;
3009 
3010                 datap = compat_ptr(data);
3011                 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
3012                         return -EFAULT;
3013 
3014                 return dev_ioctl(net, cmd, uifr);
3015         default:
3016                 return -ENOIOCTLCMD;
3017         }
3018 }
3019 
3020 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
3021                                  struct compat_ifreq __user *u_ifreq32)
3022 {
3023         struct ifreq __user *u_ifreq64;
3024         char tmp_buf[IFNAMSIZ];
3025         void __user *data64;
3026         u32 data32;
3027 
3028         if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3029                            IFNAMSIZ))
3030                 return -EFAULT;
3031         if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3032                 return -EFAULT;
3033         data64 = compat_ptr(data32);
3034 
3035         u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3036 
3037         /* Don't check these user accesses, just let that get trapped
3038          * in the ioctl handler instead.
3039          */
3040         if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3041                          IFNAMSIZ))
3042                 return -EFAULT;
3043         if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3044                 return -EFAULT;
3045 
3046         return dev_ioctl(net, cmd, u_ifreq64);
3047 }
3048 
3049 static int dev_ifsioc(struct net *net, struct socket *sock,
3050                          unsigned int cmd, struct compat_ifreq __user *uifr32)
3051 {
3052         struct ifreq __user *uifr;
3053         int err;
3054 
3055         uifr = compat_alloc_user_space(sizeof(*uifr));
3056         if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3057                 return -EFAULT;
3058 
3059         err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3060 
3061         if (!err) {
3062                 switch (cmd) {
3063                 case SIOCGIFFLAGS:
3064                 case SIOCGIFMETRIC:
3065                 case SIOCGIFMTU:
3066                 case SIOCGIFMEM:
3067                 case SIOCGIFHWADDR:
3068                 case SIOCGIFINDEX:
3069                 case SIOCGIFADDR:
3070                 case SIOCGIFBRDADDR:
3071                 case SIOCGIFDSTADDR:
3072                 case SIOCGIFNETMASK:
3073                 case SIOCGIFPFLAGS:
3074                 case SIOCGIFTXQLEN:
3075                 case SIOCGMIIPHY:
3076                 case SIOCGMIIREG:
3077                         if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3078                                 err = -EFAULT;
3079                         break;
3080                 }
3081         }
3082         return err;
3083 }
3084 
3085 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3086                         struct compat_ifreq __user *uifr32)
3087 {
3088         struct ifreq ifr;
3089         struct compat_ifmap __user *uifmap32;
3090         mm_segment_t old_fs;
3091         int err;
3092 
3093         uifmap32 = &uifr32->ifr_ifru.ifru_map;
3094         err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3095         err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3096         err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3097         err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3098         err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3099         err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3100         err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3101         if (err)
3102                 return -EFAULT;
3103 
3104         old_fs = get_fs();
3105         set_fs(KERNEL_DS);
3106         err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3107         set_fs(old_fs);
3108 
3109         if (cmd == SIOCGIFMAP && !err) {
3110                 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3111                 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3112                 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3113                 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3114                 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3115                 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3116                 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3117                 if (err)
3118                         err = -EFAULT;
3119         }
3120         return err;
3121 }
3122 
3123 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3124 {
3125         void __user *uptr;
3126         compat_uptr_t uptr32;
3127         struct ifreq __user *uifr;
3128 
3129         uifr = compat_alloc_user_space(sizeof(*uifr));
3130         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3131                 return -EFAULT;
3132 
3133         if (get_user(uptr32, &uifr32->ifr_data))
3134                 return -EFAULT;
3135 
3136         uptr = compat_ptr(uptr32);
3137 
3138         if (put_user(uptr, &uifr->ifr_data))
3139                 return -EFAULT;
3140 
3141         return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3142 }
3143 
3144 struct rtentry32 {
3145         u32             rt_pad1;
3146         struct sockaddr rt_dst;         /* target address               */
3147         struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3148         struct sockaddr rt_genmask;     /* target network mask (IP)     */
3149         unsigned short  rt_flags;
3150         short           rt_pad2;
3151         u32             rt_pad3;
3152         unsigned char   rt_tos;
3153         unsigned char   rt_class;
3154         short           rt_pad4;
3155         short           rt_metric;      /* +1 for binary compatibility! */
3156         /* char * */ u32 rt_dev;        /* forcing the device at add    */
3157         u32             rt_mtu;         /* per route MTU/Window         */
3158         u32             rt_window;      /* Window clamping              */
3159         unsigned short  rt_irtt;        /* Initial RTT                  */
3160 };
3161 
3162 struct in6_rtmsg32 {
3163         struct in6_addr         rtmsg_dst;
3164         struct in6_addr         rtmsg_src;
3165         struct in6_addr         rtmsg_gateway;
3166         u32                     rtmsg_type;
3167         u16                     rtmsg_dst_len;
3168         u16                     rtmsg_src_len;
3169         u32                     rtmsg_metric;
3170         u32                     rtmsg_info;
3171         u32                     rtmsg_flags;
3172         s32                     rtmsg_ifindex;
3173 };
3174 
3175 static int routing_ioctl(struct net *net, struct socket *sock,
3176                          unsigned int cmd, void __user *argp)
3177 {
3178         int ret;
3179         void *r = NULL;
3180         struct in6_rtmsg r6;
3181         struct rtentry r4;
3182         char devname[16];
3183         u32 rtdev;
3184         mm_segment_t old_fs = get_fs();
3185 
3186         if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3187                 struct in6_rtmsg32 __user *ur6 = argp;
3188                 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3189                         3 * sizeof(struct in6_addr));
3190                 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3191                 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3192                 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3193                 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3194                 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3195                 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3196                 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3197 
3198                 r = (void *) &r6;
3199         } else { /* ipv4 */
3200                 struct rtentry32 __user *ur4 = argp;
3201                 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3202                                         3 * sizeof(struct sockaddr));
3203                 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3204                 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3205                 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3206                 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3207                 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3208                 ret |= __get_user(rtdev, &(ur4->rt_dev));
3209                 if (rtdev) {
3210                         ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3211                         r4.rt_dev = (char __user __force *)devname;
3212                         devname[15] = 0;
3213                 } else
3214                         r4.rt_dev = NULL;
3215 
3216                 r = (void *) &r4;
3217         }
3218 
3219         if (ret) {
3220                 ret = -EFAULT;
3221                 goto out;
3222         }
3223 
3224         set_fs(KERNEL_DS);
3225         ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3226         set_fs(old_fs);
3227 
3228 out:
3229         return ret;
3230 }
3231 
3232 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3233  * for some operations; this forces use of the newer bridge-utils that
3234  * use compatible ioctls
3235  */
3236 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3237 {
3238         compat_ulong_t tmp;
3239 
3240         if (get_user(tmp, argp))
3241                 return -EFAULT;
3242         if (tmp == BRCTL_GET_VERSION)
3243                 return BRCTL_VERSION + 1;
3244         return -EINVAL;
3245 }
3246 
3247 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3248                          unsigned int cmd, unsigned long arg)
3249 {
3250         void __user *argp = compat_ptr(arg);
3251         struct sock *sk = sock->sk;
3252         struct net *net = sock_net(sk);
3253 
3254         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3255                 return siocdevprivate_ioctl(net, cmd, argp);
3256 
3257         switch (cmd) {
3258         case SIOCSIFBR:
3259         case SIOCGIFBR:
3260                 return old_bridge_ioctl(argp);
3261         case SIOCGIFNAME:
3262                 return dev_ifname32(net, argp);
3263         case SIOCGIFCONF:
3264                 return dev_ifconf(net, argp);
3265         case SIOCETHTOOL:
3266                 return ethtool_ioctl(net, argp);
3267         case SIOCWANDEV:
3268                 return compat_siocwandev(net, argp);
3269         case SIOCGIFMAP:
3270         case SIOCSIFMAP:
3271                 return compat_sioc_ifmap(net, cmd, argp);
3272         case SIOCBONDENSLAVE:
3273         case SIOCBONDRELEASE:
3274         case SIOCBONDSETHWADDR:
3275         case SIOCBONDSLAVEINFOQUERY:
3276         case SIOCBONDINFOQUERY:
3277         case SIOCBONDCHANGEACTIVE:
3278                 return bond_ioctl(net, cmd, argp);
3279         case SIOCADDRT:
3280         case SIOCDELRT:
3281                 return routing_ioctl(net, sock, cmd, argp);
3282         case SIOCGSTAMP:
3283                 return do_siocgstamp(net, sock, cmd, argp);
3284         case SIOCGSTAMPNS:
3285                 return do_siocgstampns(net, sock, cmd, argp);
3286         case SIOCSHWTSTAMP:
3287                 return compat_siocshwtstamp(net, argp);
3288 
3289         case FIOSETOWN:
3290         case SIOCSPGRP:
3291         case FIOGETOWN:
3292         case SIOCGPGRP:
3293         case SIOCBRADDBR:
3294         case SIOCBRDELBR:
3295         case SIOCGIFVLAN:
3296         case SIOCSIFVLAN:
3297         case SIOCADDDLCI:
3298         case SIOCDELDLCI:
3299                 return sock_ioctl(file, cmd, arg);
3300 
3301         case SIOCGIFFLAGS:
3302         case SIOCSIFFLAGS:
3303         case SIOCGIFMETRIC:
3304         case SIOCSIFMETRIC:
3305         case SIOCGIFMTU:
3306         case SIOCSIFMTU:
3307         case SIOCGIFMEM:
3308         case SIOCSIFMEM:
3309         case SIOCGIFHWADDR:
3310         case SIOCSIFHWADDR:
3311         case SIOCADDMULTI:
3312         case SIOCDELMULTI:
3313         case SIOCGIFINDEX:
3314         case SIOCGIFADDR:
3315         case SIOCSIFADDR:
3316         case SIOCSIFHWBROADCAST:
3317         case SIOCDIFADDR:
3318         case SIOCGIFBRDADDR:
3319         case SIOCSIFBRDADDR:
3320         case SIOCGIFDSTADDR:
3321         case SIOCSIFDSTADDR:
3322         case SIOCGIFNETMASK:
3323         case SIOCSIFNETMASK:
3324         case SIOCSIFPFLAGS:
3325         case SIOCGIFPFLAGS:
3326         case SIOCGIFTXQLEN:
3327         case SIOCSIFTXQLEN:
3328         case SIOCBRADDIF:
3329         case SIOCBRDELIF:
3330         case SIOCSIFNAME:
3331         case SIOCGMIIPHY:
3332         case SIOCGMIIREG:
3333         case SIOCSMIIREG:
3334                 return dev_ifsioc(net, sock, cmd, argp);
3335 
3336         case SIOCSARP:
3337         case SIOCGARP:
3338         case SIOCDARP:
3339         case SIOCATMARK:
3340                 return sock_do_ioctl(net, sock, cmd, arg);
3341         }
3342 
3343         return -ENOIOCTLCMD;
3344 }
3345 
3346 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3347                               unsigned long arg)
3348 {
3349         struct socket *sock = file->private_data;
3350         int ret = -ENOIOCTLCMD;
3351         struct sock *sk;
3352         struct net *net;
3353 
3354         sk = sock->sk;
3355         net = sock_net(sk);
3356 
3357         if (sock->ops->compat_ioctl)
3358                 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3359 
3360         if (ret == -ENOIOCTLCMD &&
3361             (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3362                 ret = compat_wext_handle_ioctl(net, cmd, arg);
3363 
3364         if (ret == -ENOIOCTLCMD)
3365                 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3366 
3367         return ret;
3368 }
3369 #endif
3370 
3371 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3372 {
3373         return sock->ops->bind(sock, addr, addrlen);
3374 }
3375 EXPORT_SYMBOL(kernel_bind);
3376 
3377 int kernel_listen(struct socket *sock, int backlog)
3378 {
3379         return sock->ops->listen(sock, backlog);
3380 }
3381 EXPORT_SYMBOL(kernel_listen);
3382 
3383 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3384 {
3385         struct sock *sk = sock->sk;
3386         int err;
3387 
3388         err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3389                                newsock);
3390         if (err < 0)
3391                 goto done;
3392 
3393         err = sock->ops->accept(sock, *newsock, flags);
3394         if (err < 0) {
3395                 sock_release(*newsock);
3396                 *newsock = NULL;
3397                 goto done;
3398         }
3399 
3400         (*newsock)->ops = sock->ops;
3401         __module_get((*newsock)->ops->owner);
3402 
3403 done:
3404         return err;
3405 }
3406 EXPORT_SYMBOL(kernel_accept);
3407 
3408 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3409                    int flags)
3410 {
3411         return sock->ops->connect(sock, addr, addrlen, flags);
3412 }
3413 EXPORT_SYMBOL(kernel_connect);
3414 
3415 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3416                          int *addrlen)
3417 {
3418         return sock->ops->getname(sock, addr, addrlen, 0);
3419 }
3420 EXPORT_SYMBOL(kernel_getsockname);
3421 
3422 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3423                          int *addrlen)
3424 {
3425         return sock->ops->getname(sock, addr, addrlen, 1);
3426 }
3427 EXPORT_SYMBOL(kernel_getpeername);
3428 
3429 int kernel_getsockopt(struct socket *sock, int level, int optname,
3430                         char *optval, int *optlen)
3431 {
3432         mm_segment_t oldfs = get_fs();
3433         char __user *uoptval;
3434         int __user *uoptlen;
3435         int err;
3436 
3437         uoptval = (char __user __force *) optval;
3438         uoptlen = (int __user __force *) optlen;
3439 
3440         set_fs(KERNEL_DS);
3441         if (level == SOL_SOCKET)
3442                 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3443         else
3444                 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3445                                             uoptlen);
3446         set_fs(oldfs);
3447         return err;
3448 }
3449 EXPORT_SYMBOL(kernel_getsockopt);
3450 
3451 int kernel_setsockopt(struct socket *sock, int level, int optname,
3452                         char *optval, unsigned int optlen)
3453 {
3454         mm_segment_t oldfs = get_fs();
3455         char __user *uoptval;
3456         int err;
3457 
3458         uoptval = (char __user __force *) optval;
3459 
3460         set_fs(KERNEL_DS);
3461         if (level == SOL_SOCKET)
3462                 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3463         else
3464                 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3465                                             optlen);
3466         set_fs(oldfs);
3467         return err;
3468 }
3469 EXPORT_SYMBOL(kernel_setsockopt);
3470 
3471 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3472                     size_t size, int flags)
3473 {
3474         if (sock->ops->sendpage)
3475                 return sock->ops->sendpage(sock, page, offset, size, flags);
3476 
3477         return sock_no_sendpage(sock, page, offset, size, flags);
3478 }
3479 EXPORT_SYMBOL(kernel_sendpage);
3480 
3481 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3482 {
3483         mm_segment_t oldfs = get_fs();
3484         int err;
3485 
3486         set_fs(KERNEL_DS);
3487         err = sock->ops->ioctl(sock, cmd, arg);
3488         set_fs(oldfs);
3489 
3490         return err;
3491 }
3492 EXPORT_SYMBOL(kernel_sock_ioctl);
3493 
3494 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3495 {
3496         return sock->ops->shutdown(sock, how);
3497 }
3498 EXPORT_SYMBOL(kernel_sock_shutdown);
3499 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp