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

TOMOYO Linux Cross Reference
Linux/net/socket.c

Version: ~ [ linux-5.15-rc7 ] ~ [ linux-5.14.14 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.75 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.155 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.213 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.252 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.287 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.289 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ 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/ptp_classify.h>
 76 #include <linux/init.h>
 77 #include <linux/poll.h>
 78 #include <linux/cache.h>
 79 #include <linux/module.h>
 80 #include <linux/highmem.h>
 81 #include <linux/mount.h>
 82 #include <linux/security.h>
 83 #include <linux/syscalls.h>
 84 #include <linux/compat.h>
 85 #include <linux/kmod.h>
 86 #include <linux/audit.h>
 87 #include <linux/wireless.h>
 88 #include <linux/nsproxy.h>
 89 #include <linux/magic.h>
 90 #include <linux/slab.h>
 91 #include <linux/xattr.h>
 92 #include <linux/nospec.h>
 93 
 94 #include <asm/uaccess.h>
 95 #include <asm/unistd.h>
 96 
 97 #include <net/compat.h>
 98 #include <net/wext.h>
 99 #include <net/cls_cgroup.h>
100 
101 #include <net/sock.h>
102 #include <linux/netfilter.h>
103 
104 #include <linux/if_tun.h>
105 #include <linux/ipv6_route.h>
106 #include <linux/route.h>
107 #include <linux/sockios.h>
108 #include <linux/atalk.h>
109 #include <net/busy_poll.h>
110 
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
114 #endif
115 
116 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
117 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
118                          unsigned long nr_segs, loff_t pos);
119 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
120                           unsigned long nr_segs, loff_t pos);
121 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
122 
123 static int sock_close(struct inode *inode, struct file *file);
124 static unsigned int sock_poll(struct file *file,
125                               struct poll_table_struct *wait);
126 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
127 #ifdef CONFIG_COMPAT
128 static long compat_sock_ioctl(struct file *file,
129                               unsigned int cmd, unsigned long arg);
130 #endif
131 static int sock_fasync(int fd, struct file *filp, int on);
132 static ssize_t sock_sendpage(struct file *file, struct page *page,
133                              int offset, size_t size, loff_t *ppos, int more);
134 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
135                                 struct pipe_inode_info *pipe, size_t len,
136                                 unsigned int flags);
137 
138 /*
139  *      Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
140  *      in the operation structures but are done directly via the socketcall() multiplexor.
141  */
142 
143 static const struct file_operations socket_file_ops = {
144         .owner =        THIS_MODULE,
145         .llseek =       no_llseek,
146         .aio_read =     sock_aio_read,
147         .aio_write =    sock_aio_write,
148         .poll =         sock_poll,
149         .unlocked_ioctl = sock_ioctl,
150 #ifdef CONFIG_COMPAT
151         .compat_ioctl = compat_sock_ioctl,
152 #endif
153         .mmap =         sock_mmap,
154         .open =         sock_no_open,   /* special open code to disallow open via /proc */
155         .release =      sock_close,
156         .fasync =       sock_fasync,
157         .sendpage =     sock_sendpage,
158         .splice_write = generic_splice_sendpage,
159         .splice_read =  sock_splice_read,
160 };
161 
162 /*
163  *      The protocol list. Each protocol is registered in here.
164  */
165 
166 static DEFINE_SPINLOCK(net_family_lock);
167 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
168 
169 /*
170  *      Statistics counters of the socket lists
171  */
172 
173 static DEFINE_PER_CPU(int, sockets_in_use);
174 
175 /*
176  * Support routines.
177  * Move socket addresses back and forth across the kernel/user
178  * divide and look after the messy bits.
179  */
180 
181 /**
182  *      move_addr_to_kernel     -       copy a socket address into kernel space
183  *      @uaddr: Address in user space
184  *      @kaddr: Address in kernel space
185  *      @ulen: Length in user space
186  *
187  *      The address is copied into kernel space. If the provided address is
188  *      too long an error code of -EINVAL is returned. If the copy gives
189  *      invalid addresses -EFAULT is returned. On a success 0 is returned.
190  */
191 
192 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
193 {
194         if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
195                 return -EINVAL;
196         if (ulen == 0)
197                 return 0;
198         if (copy_from_user(kaddr, uaddr, ulen))
199                 return -EFAULT;
200         return audit_sockaddr(ulen, kaddr);
201 }
202 
203 /**
204  *      move_addr_to_user       -       copy an address to user space
205  *      @kaddr: kernel space address
206  *      @klen: length of address in kernel
207  *      @uaddr: user space address
208  *      @ulen: pointer to user length field
209  *
210  *      The value pointed to by ulen on entry is the buffer length available.
211  *      This is overwritten with the buffer space used. -EINVAL is returned
212  *      if an overlong buffer is specified or a negative buffer size. -EFAULT
213  *      is returned if either the buffer or the length field are not
214  *      accessible.
215  *      After copying the data up to the limit the user specifies, the true
216  *      length of the data is written over the length limit the user
217  *      specified. Zero is returned for a success.
218  */
219 
220 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
221                              void __user *uaddr, int __user *ulen)
222 {
223         int err;
224         int len;
225 
226         BUG_ON(klen > sizeof(struct sockaddr_storage));
227         err = get_user(len, ulen);
228         if (err)
229                 return err;
230         if (len > klen)
231                 len = klen;
232         if (len < 0)
233                 return -EINVAL;
234         if (len) {
235                 if (audit_sockaddr(klen, kaddr))
236                         return -ENOMEM;
237                 if (copy_to_user(uaddr, kaddr, len))
238                         return -EFAULT;
239         }
240         /*
241          *      "fromlen shall refer to the value before truncation.."
242          *                      1003.1g
243          */
244         return __put_user(klen, ulen);
245 }
246 
247 static struct kmem_cache *sock_inode_cachep __read_mostly;
248 
249 static struct inode *sock_alloc_inode(struct super_block *sb)
250 {
251         struct socket_alloc *ei;
252         struct socket_wq *wq;
253 
254         ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
255         if (!ei)
256                 return NULL;
257         wq = kmalloc(sizeof(*wq), GFP_KERNEL);
258         if (!wq) {
259                 kmem_cache_free(sock_inode_cachep, ei);
260                 return NULL;
261         }
262         init_waitqueue_head(&wq->wait);
263         wq->fasync_list = NULL;
264         RCU_INIT_POINTER(ei->socket.wq, wq);
265 
266         ei->socket.state = SS_UNCONNECTED;
267         ei->socket.flags = 0;
268         ei->socket.ops = NULL;
269         ei->socket.sk = NULL;
270         ei->socket.file = NULL;
271 
272         return &ei->vfs_inode;
273 }
274 
275 static void sock_destroy_inode(struct inode *inode)
276 {
277         struct socket_alloc *ei;
278         struct socket_wq *wq;
279 
280         ei = container_of(inode, struct socket_alloc, vfs_inode);
281         wq = rcu_dereference_protected(ei->socket.wq, 1);
282         kfree_rcu(wq, rcu);
283         kmem_cache_free(sock_inode_cachep, ei);
284 }
285 
286 static void init_once(void *foo)
287 {
288         struct socket_alloc *ei = (struct socket_alloc *)foo;
289 
290         inode_init_once(&ei->vfs_inode);
291 }
292 
293 static int init_inodecache(void)
294 {
295         sock_inode_cachep = kmem_cache_create("sock_inode_cache",
296                                               sizeof(struct socket_alloc),
297                                               0,
298                                               (SLAB_HWCACHE_ALIGN |
299                                                SLAB_RECLAIM_ACCOUNT |
300                                                SLAB_MEM_SPREAD),
301                                               init_once);
302         if (sock_inode_cachep == NULL)
303                 return -ENOMEM;
304         return 0;
305 }
306 
307 static const struct super_operations sockfs_ops = {
308         .alloc_inode    = sock_alloc_inode,
309         .destroy_inode  = sock_destroy_inode,
310         .statfs         = simple_statfs,
311 };
312 
313 /*
314  * sockfs_dname() is called from d_path().
315  */
316 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
317 {
318         return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
319                                 dentry->d_inode->i_ino);
320 }
321 
322 static const struct dentry_operations sockfs_dentry_operations = {
323         .d_dname  = sockfs_dname,
324 };
325 
326 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
327                          int flags, const char *dev_name, void *data)
328 {
329         return mount_pseudo(fs_type, "socket:", &sockfs_ops,
330                 &sockfs_dentry_operations, SOCKFS_MAGIC);
331 }
332 
333 static struct vfsmount *sock_mnt __read_mostly;
334 
335 static struct file_system_type sock_fs_type = {
336         .name =         "sockfs",
337         .mount =        sockfs_mount,
338         .kill_sb =      kill_anon_super,
339 };
340 
341 /*
342  *      Obtains the first available file descriptor and sets it up for use.
343  *
344  *      These functions create file structures and maps them to fd space
345  *      of the current process. On success it returns file descriptor
346  *      and file struct implicitly stored in sock->file.
347  *      Note that another thread may close file descriptor before we return
348  *      from this function. We use the fact that now we do not refer
349  *      to socket after mapping. If one day we will need it, this
350  *      function will increment ref. count on file by 1.
351  *
352  *      In any case returned fd MAY BE not valid!
353  *      This race condition is unavoidable
354  *      with shared fd spaces, we cannot solve it inside kernel,
355  *      but we take care of internal coherence yet.
356  */
357 
358 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
359 {
360         struct qstr name = { .name = "" };
361         struct path path;
362         struct file *file;
363 
364         if (dname) {
365                 name.name = dname;
366                 name.len = strlen(name.name);
367         } else if (sock->sk) {
368                 name.name = sock->sk->sk_prot_creator->name;
369                 name.len = strlen(name.name);
370         }
371         path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
372         if (unlikely(!path.dentry))
373                 return ERR_PTR(-ENOMEM);
374         path.mnt = mntget(sock_mnt);
375 
376         d_instantiate(path.dentry, SOCK_INODE(sock));
377         SOCK_INODE(sock)->i_fop = &socket_file_ops;
378 
379         file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
380                   &socket_file_ops);
381         if (unlikely(IS_ERR(file))) {
382                 /* drop dentry, keep inode */
383                 ihold(path.dentry->d_inode);
384                 path_put(&path);
385                 return file;
386         }
387 
388         sock->file = file;
389         file->f_flags = O_RDWR | (flags & O_NONBLOCK);
390         file->private_data = sock;
391         return file;
392 }
393 EXPORT_SYMBOL(sock_alloc_file);
394 
395 static int sock_map_fd(struct socket *sock, int flags)
396 {
397         struct file *newfile;
398         int fd = get_unused_fd_flags(flags);
399         if (unlikely(fd < 0))
400                 return fd;
401 
402         newfile = sock_alloc_file(sock, flags, NULL);
403         if (likely(!IS_ERR(newfile))) {
404                 fd_install(fd, newfile);
405                 return fd;
406         }
407 
408         put_unused_fd(fd);
409         return PTR_ERR(newfile);
410 }
411 
412 struct socket *sock_from_file(struct file *file, int *err)
413 {
414         if (file->f_op == &socket_file_ops)
415                 return file->private_data;      /* set in sock_map_fd */
416 
417         *err = -ENOTSOCK;
418         return NULL;
419 }
420 EXPORT_SYMBOL(sock_from_file);
421 
422 /**
423  *      sockfd_lookup - Go from a file number to its socket slot
424  *      @fd: file handle
425  *      @err: pointer to an error code return
426  *
427  *      The file handle passed in is locked and the socket it is bound
428  *      too is returned. If an error occurs the err pointer is overwritten
429  *      with a negative errno code and NULL is returned. The function checks
430  *      for both invalid handles and passing a handle which is not a socket.
431  *
432  *      On a success the socket object pointer is returned.
433  */
434 
435 struct socket *sockfd_lookup(int fd, int *err)
436 {
437         struct file *file;
438         struct socket *sock;
439 
440         file = fget(fd);
441         if (!file) {
442                 *err = -EBADF;
443                 return NULL;
444         }
445 
446         sock = sock_from_file(file, err);
447         if (!sock)
448                 fput(file);
449         return sock;
450 }
451 EXPORT_SYMBOL(sockfd_lookup);
452 
453 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
454 {
455         struct fd f = fdget(fd);
456         struct socket *sock;
457 
458         *err = -EBADF;
459         if (f.file) {
460                 sock = sock_from_file(f.file, err);
461                 if (likely(sock)) {
462                         *fput_needed = f.flags;
463                         return sock;
464                 }
465                 fdput(f);
466         }
467         return NULL;
468 }
469 
470 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
471 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
472 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
473 static ssize_t sockfs_getxattr(struct dentry *dentry,
474                                const char *name, void *value, size_t size)
475 {
476         const char *proto_name;
477         size_t proto_size;
478         int error;
479 
480         error = -ENODATA;
481         if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
482                 proto_name = dentry->d_name.name;
483                 proto_size = strlen(proto_name);
484 
485                 if (value) {
486                         error = -ERANGE;
487                         if (proto_size + 1 > size)
488                                 goto out;
489 
490                         strncpy(value, proto_name, proto_size + 1);
491                 }
492                 error = proto_size + 1;
493         }
494 
495 out:
496         return error;
497 }
498 
499 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
500                                 size_t size)
501 {
502         ssize_t len;
503         ssize_t used = 0;
504 
505         len = security_inode_listsecurity(dentry->d_inode, buffer, size);
506         if (len < 0)
507                 return len;
508         used += len;
509         if (buffer) {
510                 if (size < used)
511                         return -ERANGE;
512                 buffer += len;
513         }
514 
515         len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
516         used += len;
517         if (buffer) {
518                 if (size < used)
519                         return -ERANGE;
520                 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
521                 buffer += len;
522         }
523 
524         return used;
525 }
526 
527 static const struct inode_operations sockfs_inode_ops = {
528         .getxattr = sockfs_getxattr,
529         .listxattr = sockfs_listxattr,
530 };
531 
532 /**
533  *      sock_alloc      -       allocate a socket
534  *
535  *      Allocate a new inode and socket object. The two are bound together
536  *      and initialised. The socket is then returned. If we are out of inodes
537  *      NULL is returned.
538  */
539 
540 static struct socket *sock_alloc(void)
541 {
542         struct inode *inode;
543         struct socket *sock;
544 
545         inode = new_inode_pseudo(sock_mnt->mnt_sb);
546         if (!inode)
547                 return NULL;
548 
549         sock = SOCKET_I(inode);
550 
551         kmemcheck_annotate_bitfield(sock, type);
552         inode->i_ino = get_next_ino();
553         inode->i_mode = S_IFSOCK | S_IRWXUGO;
554         inode->i_uid = current_fsuid();
555         inode->i_gid = current_fsgid();
556         inode->i_op = &sockfs_inode_ops;
557 
558         this_cpu_add(sockets_in_use, 1);
559         return sock;
560 }
561 
562 /*
563  *      In theory you can't get an open on this inode, but /proc provides
564  *      a back door. Remember to keep it shut otherwise you'll let the
565  *      creepy crawlies in.
566  */
567 
568 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
569 {
570         return -ENXIO;
571 }
572 
573 const struct file_operations bad_sock_fops = {
574         .owner = THIS_MODULE,
575         .open = sock_no_open,
576         .llseek = noop_llseek,
577 };
578 
579 /**
580  *      sock_release    -       close a socket
581  *      @sock: socket to close
582  *
583  *      The socket is released from the protocol stack if it has a release
584  *      callback, and the inode is then released if the socket is bound to
585  *      an inode not a file.
586  */
587 
588 void sock_release(struct socket *sock)
589 {
590         if (sock->ops) {
591                 struct module *owner = sock->ops->owner;
592 
593                 sock->ops->release(sock);
594                 sock->ops = NULL;
595                 module_put(owner);
596         }
597 
598         if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
599                 pr_err("%s: fasync list not empty!\n", __func__);
600 
601         if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
602                 return;
603 
604         this_cpu_sub(sockets_in_use, 1);
605         if (!sock->file) {
606                 iput(SOCK_INODE(sock));
607                 return;
608         }
609         sock->file = NULL;
610 }
611 EXPORT_SYMBOL(sock_release);
612 
613 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
614 {
615         *tx_flags = 0;
616         if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
617                 *tx_flags |= SKBTX_HW_TSTAMP;
618         if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
619                 *tx_flags |= SKBTX_SW_TSTAMP;
620         if (sock_flag(sk, SOCK_WIFI_STATUS))
621                 *tx_flags |= SKBTX_WIFI_STATUS;
622 }
623 EXPORT_SYMBOL(sock_tx_timestamp);
624 
625 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
626                                        struct msghdr *msg, size_t size)
627 {
628         struct sock_iocb *si = kiocb_to_siocb(iocb);
629 
630         si->sock = sock;
631         si->scm = NULL;
632         si->msg = msg;
633         si->size = size;
634 
635         return sock->ops->sendmsg(iocb, sock, msg, size);
636 }
637 
638 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
639                                  struct msghdr *msg, size_t size)
640 {
641         int err = security_socket_sendmsg(sock, msg, size);
642 
643         return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
644 }
645 
646 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
647 {
648         struct kiocb iocb;
649         struct sock_iocb siocb;
650         int ret;
651 
652         init_sync_kiocb(&iocb, NULL);
653         iocb.private = &siocb;
654         ret = __sock_sendmsg(&iocb, sock, msg, size);
655         if (-EIOCBQUEUED == ret)
656                 ret = wait_on_sync_kiocb(&iocb);
657         return ret;
658 }
659 EXPORT_SYMBOL(sock_sendmsg);
660 
661 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
662 {
663         struct kiocb iocb;
664         struct sock_iocb siocb;
665         int ret;
666 
667         init_sync_kiocb(&iocb, NULL);
668         iocb.private = &siocb;
669         ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
670         if (-EIOCBQUEUED == ret)
671                 ret = wait_on_sync_kiocb(&iocb);
672         return ret;
673 }
674 
675 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
676                    struct kvec *vec, size_t num, size_t size)
677 {
678         mm_segment_t oldfs = get_fs();
679         int result;
680 
681         set_fs(KERNEL_DS);
682         /*
683          * the following is safe, since for compiler definitions of kvec and
684          * iovec are identical, yielding the same in-core layout and alignment
685          */
686         msg->msg_iov = (struct iovec *)vec;
687         msg->msg_iovlen = num;
688         result = sock_sendmsg(sock, msg, size);
689         set_fs(oldfs);
690         return result;
691 }
692 EXPORT_SYMBOL(kernel_sendmsg);
693 
694 /*
695  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
696  */
697 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
698         struct sk_buff *skb)
699 {
700         int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
701         struct timespec ts[3];
702         int empty = 1;
703         struct skb_shared_hwtstamps *shhwtstamps =
704                 skb_hwtstamps(skb);
705 
706         /* Race occurred between timestamp enabling and packet
707            receiving.  Fill in the current time for now. */
708         if (need_software_tstamp && skb->tstamp.tv64 == 0)
709                 __net_timestamp(skb);
710 
711         if (need_software_tstamp) {
712                 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
713                         struct timeval tv;
714                         skb_get_timestamp(skb, &tv);
715                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
716                                  sizeof(tv), &tv);
717                 } else {
718                         skb_get_timestampns(skb, &ts[0]);
719                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
720                                  sizeof(ts[0]), &ts[0]);
721                 }
722         }
723 
724 
725         memset(ts, 0, sizeof(ts));
726         if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
727             ktime_to_timespec_cond(skb->tstamp, ts + 0))
728                 empty = 0;
729         if (shhwtstamps) {
730                 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
731                     ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
732                         empty = 0;
733                 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
734                     ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
735                         empty = 0;
736         }
737         if (!empty)
738                 put_cmsg(msg, SOL_SOCKET,
739                          SCM_TIMESTAMPING, sizeof(ts), &ts);
740 }
741 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
742 
743 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
744         struct sk_buff *skb)
745 {
746         int ack;
747 
748         if (!sock_flag(sk, SOCK_WIFI_STATUS))
749                 return;
750         if (!skb->wifi_acked_valid)
751                 return;
752 
753         ack = skb->wifi_acked;
754 
755         put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
756 }
757 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
758 
759 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
760                                    struct sk_buff *skb)
761 {
762         if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
763                 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
764                         sizeof(__u32), &skb->dropcount);
765 }
766 
767 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
768         struct sk_buff *skb)
769 {
770         sock_recv_timestamp(msg, sk, skb);
771         sock_recv_drops(msg, sk, skb);
772 }
773 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
774 
775 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
776                                        struct msghdr *msg, size_t size, int flags)
777 {
778         struct sock_iocb *si = kiocb_to_siocb(iocb);
779 
780         si->sock = sock;
781         si->scm = NULL;
782         si->msg = msg;
783         si->size = size;
784         si->flags = flags;
785 
786         return sock->ops->recvmsg(iocb, sock, msg, size, flags);
787 }
788 
789 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
790                                  struct msghdr *msg, size_t size, int flags)
791 {
792         int err = security_socket_recvmsg(sock, msg, size, flags);
793 
794         return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
795 }
796 
797 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
798                  size_t size, int flags)
799 {
800         struct kiocb iocb;
801         struct sock_iocb siocb;
802         int ret;
803 
804         init_sync_kiocb(&iocb, NULL);
805         iocb.private = &siocb;
806         ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
807         if (-EIOCBQUEUED == ret)
808                 ret = wait_on_sync_kiocb(&iocb);
809         return ret;
810 }
811 EXPORT_SYMBOL(sock_recvmsg);
812 
813 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
814                               size_t size, int flags)
815 {
816         struct kiocb iocb;
817         struct sock_iocb siocb;
818         int ret;
819 
820         init_sync_kiocb(&iocb, NULL);
821         iocb.private = &siocb;
822         ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
823         if (-EIOCBQUEUED == ret)
824                 ret = wait_on_sync_kiocb(&iocb);
825         return ret;
826 }
827 
828 /**
829  * kernel_recvmsg - Receive a message from a socket (kernel space)
830  * @sock:       The socket to receive the message from
831  * @msg:        Received message
832  * @vec:        Input s/g array for message data
833  * @num:        Size of input s/g array
834  * @size:       Number of bytes to read
835  * @flags:      Message flags (MSG_DONTWAIT, etc...)
836  *
837  * On return the msg structure contains the scatter/gather array passed in the
838  * vec argument. The array is modified so that it consists of the unfilled
839  * portion of the original array.
840  *
841  * The returned value is the total number of bytes received, or an error.
842  */
843 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
844                    struct kvec *vec, size_t num, size_t size, int flags)
845 {
846         mm_segment_t oldfs = get_fs();
847         int result;
848 
849         set_fs(KERNEL_DS);
850         /*
851          * the following is safe, since for compiler definitions of kvec and
852          * iovec are identical, yielding the same in-core layout and alignment
853          */
854         msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
855         result = sock_recvmsg(sock, msg, size, flags);
856         set_fs(oldfs);
857         return result;
858 }
859 EXPORT_SYMBOL(kernel_recvmsg);
860 
861 static ssize_t sock_sendpage(struct file *file, struct page *page,
862                              int offset, size_t size, loff_t *ppos, int more)
863 {
864         struct socket *sock;
865         int flags;
866 
867         sock = file->private_data;
868 
869         flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
870         /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
871         flags |= more;
872 
873         return kernel_sendpage(sock, page, offset, size, flags);
874 }
875 
876 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
877                                 struct pipe_inode_info *pipe, size_t len,
878                                 unsigned int flags)
879 {
880         struct socket *sock = file->private_data;
881 
882         if (unlikely(!sock->ops->splice_read))
883                 return -EINVAL;
884 
885         return sock->ops->splice_read(sock, ppos, pipe, len, flags);
886 }
887 
888 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
889                                          struct sock_iocb *siocb)
890 {
891         siocb->kiocb = iocb;
892         iocb->private = siocb;
893         return siocb;
894 }
895 
896 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
897                 struct file *file, const struct iovec *iov,
898                 unsigned long nr_segs)
899 {
900         struct socket *sock = file->private_data;
901         size_t size = 0;
902         int i;
903 
904         for (i = 0; i < nr_segs; i++)
905                 size += iov[i].iov_len;
906 
907         msg->msg_name = NULL;
908         msg->msg_namelen = 0;
909         msg->msg_control = NULL;
910         msg->msg_controllen = 0;
911         msg->msg_iov = (struct iovec *)iov;
912         msg->msg_iovlen = nr_segs;
913         msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
914 
915         return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
916 }
917 
918 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
919                                 unsigned long nr_segs, loff_t pos)
920 {
921         struct sock_iocb siocb, *x;
922 
923         if (pos != 0)
924                 return -ESPIPE;
925 
926         if (iocb->ki_nbytes == 0)       /* Match SYS5 behaviour */
927                 return 0;
928 
929 
930         x = alloc_sock_iocb(iocb, &siocb);
931         if (!x)
932                 return -ENOMEM;
933         return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
934 }
935 
936 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
937                         struct file *file, const struct iovec *iov,
938                         unsigned long nr_segs)
939 {
940         struct socket *sock = file->private_data;
941         size_t size = 0;
942         int i;
943 
944         for (i = 0; i < nr_segs; i++)
945                 size += iov[i].iov_len;
946 
947         msg->msg_name = NULL;
948         msg->msg_namelen = 0;
949         msg->msg_control = NULL;
950         msg->msg_controllen = 0;
951         msg->msg_iov = (struct iovec *)iov;
952         msg->msg_iovlen = nr_segs;
953         msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
954         if (sock->type == SOCK_SEQPACKET)
955                 msg->msg_flags |= MSG_EOR;
956 
957         return __sock_sendmsg(iocb, sock, msg, size);
958 }
959 
960 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
961                           unsigned long nr_segs, loff_t pos)
962 {
963         struct sock_iocb siocb, *x;
964 
965         if (pos != 0)
966                 return -ESPIPE;
967 
968         x = alloc_sock_iocb(iocb, &siocb);
969         if (!x)
970                 return -ENOMEM;
971 
972         return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
973 }
974 
975 /*
976  * Atomic setting of ioctl hooks to avoid race
977  * with module unload.
978  */
979 
980 static DEFINE_MUTEX(br_ioctl_mutex);
981 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
982 
983 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
984 {
985         mutex_lock(&br_ioctl_mutex);
986         br_ioctl_hook = hook;
987         mutex_unlock(&br_ioctl_mutex);
988 }
989 EXPORT_SYMBOL(brioctl_set);
990 
991 static DEFINE_MUTEX(vlan_ioctl_mutex);
992 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
993 
994 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
995 {
996         mutex_lock(&vlan_ioctl_mutex);
997         vlan_ioctl_hook = hook;
998         mutex_unlock(&vlan_ioctl_mutex);
999 }
1000 EXPORT_SYMBOL(vlan_ioctl_set);
1001 
1002 static DEFINE_MUTEX(dlci_ioctl_mutex);
1003 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1004 
1005 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1006 {
1007         mutex_lock(&dlci_ioctl_mutex);
1008         dlci_ioctl_hook = hook;
1009         mutex_unlock(&dlci_ioctl_mutex);
1010 }
1011 EXPORT_SYMBOL(dlci_ioctl_set);
1012 
1013 static long sock_do_ioctl(struct net *net, struct socket *sock,
1014                                  unsigned int cmd, unsigned long arg)
1015 {
1016         int err;
1017         void __user *argp = (void __user *)arg;
1018 
1019         err = sock->ops->ioctl(sock, cmd, arg);
1020 
1021         /*
1022          * If this ioctl is unknown try to hand it down
1023          * to the NIC driver.
1024          */
1025         if (err == -ENOIOCTLCMD)
1026                 err = dev_ioctl(net, cmd, argp);
1027 
1028         return err;
1029 }
1030 
1031 /*
1032  *      With an ioctl, arg may well be a user mode pointer, but we don't know
1033  *      what to do with it - that's up to the protocol still.
1034  */
1035 
1036 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1037 {
1038         struct socket *sock;
1039         struct sock *sk;
1040         void __user *argp = (void __user *)arg;
1041         int pid, err;
1042         struct net *net;
1043 
1044         sock = file->private_data;
1045         sk = sock->sk;
1046         net = sock_net(sk);
1047         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1048                 err = dev_ioctl(net, cmd, argp);
1049         } else
1050 #ifdef CONFIG_WEXT_CORE
1051         if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1052                 err = dev_ioctl(net, cmd, argp);
1053         } else
1054 #endif
1055                 switch (cmd) {
1056                 case FIOSETOWN:
1057                 case SIOCSPGRP:
1058                         err = -EFAULT;
1059                         if (get_user(pid, (int __user *)argp))
1060                                 break;
1061                         err = f_setown(sock->file, pid, 1);
1062                         break;
1063                 case FIOGETOWN:
1064                 case SIOCGPGRP:
1065                         err = put_user(f_getown(sock->file),
1066                                        (int __user *)argp);
1067                         break;
1068                 case SIOCGIFBR:
1069                 case SIOCSIFBR:
1070                 case SIOCBRADDBR:
1071                 case SIOCBRDELBR:
1072                         err = -ENOPKG;
1073                         if (!br_ioctl_hook)
1074                                 request_module("bridge");
1075 
1076                         mutex_lock(&br_ioctl_mutex);
1077                         if (br_ioctl_hook)
1078                                 err = br_ioctl_hook(net, cmd, argp);
1079                         mutex_unlock(&br_ioctl_mutex);
1080                         break;
1081                 case SIOCGIFVLAN:
1082                 case SIOCSIFVLAN:
1083                         err = -ENOPKG;
1084                         if (!vlan_ioctl_hook)
1085                                 request_module("8021q");
1086 
1087                         mutex_lock(&vlan_ioctl_mutex);
1088                         if (vlan_ioctl_hook)
1089                                 err = vlan_ioctl_hook(net, argp);
1090                         mutex_unlock(&vlan_ioctl_mutex);
1091                         break;
1092                 case SIOCADDDLCI:
1093                 case SIOCDELDLCI:
1094                         err = -ENOPKG;
1095                         if (!dlci_ioctl_hook)
1096                                 request_module("dlci");
1097 
1098                         mutex_lock(&dlci_ioctl_mutex);
1099                         if (dlci_ioctl_hook)
1100                                 err = dlci_ioctl_hook(cmd, argp);
1101                         mutex_unlock(&dlci_ioctl_mutex);
1102                         break;
1103                 default:
1104                         err = sock_do_ioctl(net, sock, cmd, arg);
1105                         break;
1106                 }
1107         return err;
1108 }
1109 
1110 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1111 {
1112         int err;
1113         struct socket *sock = NULL;
1114 
1115         err = security_socket_create(family, type, protocol, 1);
1116         if (err)
1117                 goto out;
1118 
1119         sock = sock_alloc();
1120         if (!sock) {
1121                 err = -ENOMEM;
1122                 goto out;
1123         }
1124 
1125         sock->type = type;
1126         err = security_socket_post_create(sock, family, type, protocol, 1);
1127         if (err)
1128                 goto out_release;
1129 
1130 out:
1131         *res = sock;
1132         return err;
1133 out_release:
1134         sock_release(sock);
1135         sock = NULL;
1136         goto out;
1137 }
1138 EXPORT_SYMBOL(sock_create_lite);
1139 
1140 /* No kernel lock held - perfect */
1141 static unsigned int sock_poll(struct file *file, poll_table *wait)
1142 {
1143         unsigned int busy_flag = 0;
1144         struct socket *sock;
1145 
1146         /*
1147          *      We can't return errors to poll, so it's either yes or no.
1148          */
1149         sock = file->private_data;
1150 
1151         if (sk_can_busy_loop(sock->sk)) {
1152                 /* this socket can poll_ll so tell the system call */
1153                 busy_flag = POLL_BUSY_LOOP;
1154 
1155                 /* once, only if requested by syscall */
1156                 if (wait && (wait->_key & POLL_BUSY_LOOP))
1157                         sk_busy_loop(sock->sk, 1);
1158         }
1159 
1160         return busy_flag | sock->ops->poll(file, sock, wait);
1161 }
1162 
1163 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1164 {
1165         struct socket *sock = file->private_data;
1166 
1167         return sock->ops->mmap(file, sock, vma);
1168 }
1169 
1170 static int sock_close(struct inode *inode, struct file *filp)
1171 {
1172         sock_release(SOCKET_I(inode));
1173         return 0;
1174 }
1175 
1176 /*
1177  *      Update the socket async list
1178  *
1179  *      Fasync_list locking strategy.
1180  *
1181  *      1. fasync_list is modified only under process context socket lock
1182  *         i.e. under semaphore.
1183  *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1184  *         or under socket lock
1185  */
1186 
1187 static int sock_fasync(int fd, struct file *filp, int on)
1188 {
1189         struct socket *sock = filp->private_data;
1190         struct sock *sk = sock->sk;
1191         struct socket_wq *wq;
1192 
1193         if (sk == NULL)
1194                 return -EINVAL;
1195 
1196         lock_sock(sk);
1197         wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1198         fasync_helper(fd, filp, on, &wq->fasync_list);
1199 
1200         if (!wq->fasync_list)
1201                 sock_reset_flag(sk, SOCK_FASYNC);
1202         else
1203                 sock_set_flag(sk, SOCK_FASYNC);
1204 
1205         release_sock(sk);
1206         return 0;
1207 }
1208 
1209 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1210 
1211 int sock_wake_async(struct socket *sock, int how, int band)
1212 {
1213         struct socket_wq *wq;
1214 
1215         if (!sock)
1216                 return -1;
1217         rcu_read_lock();
1218         wq = rcu_dereference(sock->wq);
1219         if (!wq || !wq->fasync_list) {
1220                 rcu_read_unlock();
1221                 return -1;
1222         }
1223         switch (how) {
1224         case SOCK_WAKE_WAITD:
1225                 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1226                         break;
1227                 goto call_kill;
1228         case SOCK_WAKE_SPACE:
1229                 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1230                         break;
1231                 /* fall through */
1232         case SOCK_WAKE_IO:
1233 call_kill:
1234                 kill_fasync(&wq->fasync_list, SIGIO, band);
1235                 break;
1236         case SOCK_WAKE_URG:
1237                 kill_fasync(&wq->fasync_list, SIGURG, band);
1238         }
1239         rcu_read_unlock();
1240         return 0;
1241 }
1242 EXPORT_SYMBOL(sock_wake_async);
1243 
1244 int __sock_create(struct net *net, int family, int type, int protocol,
1245                          struct socket **res, int kern)
1246 {
1247         int err;
1248         struct socket *sock;
1249         const struct net_proto_family *pf;
1250 
1251         /*
1252          *      Check protocol is in range
1253          */
1254         if (family < 0 || family >= NPROTO)
1255                 return -EAFNOSUPPORT;
1256         if (type < 0 || type >= SOCK_MAX)
1257                 return -EINVAL;
1258 
1259         /* Compatibility.
1260 
1261            This uglymoron is moved from INET layer to here to avoid
1262            deadlock in module load.
1263          */
1264         if (family == PF_INET && type == SOCK_PACKET) {
1265                 static int warned;
1266                 if (!warned) {
1267                         warned = 1;
1268                         pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1269                                 current->comm);
1270                 }
1271                 family = PF_PACKET;
1272         }
1273 
1274         err = security_socket_create(family, type, protocol, kern);
1275         if (err)
1276                 return err;
1277 
1278         /*
1279          *      Allocate the socket and allow the family to set things up. if
1280          *      the protocol is 0, the family is instructed to select an appropriate
1281          *      default.
1282          */
1283         sock = sock_alloc();
1284         if (!sock) {
1285                 net_warn_ratelimited("socket: no more sockets\n");
1286                 return -ENFILE; /* Not exactly a match, but its the
1287                                    closest posix thing */
1288         }
1289 
1290         sock->type = type;
1291 
1292 #ifdef CONFIG_MODULES
1293         /* Attempt to load a protocol module if the find failed.
1294          *
1295          * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1296          * requested real, full-featured networking support upon configuration.
1297          * Otherwise module support will break!
1298          */
1299         if (rcu_access_pointer(net_families[family]) == NULL)
1300                 request_module("net-pf-%d", family);
1301 #endif
1302 
1303         rcu_read_lock();
1304         pf = rcu_dereference(net_families[family]);
1305         err = -EAFNOSUPPORT;
1306         if (!pf)
1307                 goto out_release;
1308 
1309         /*
1310          * We will call the ->create function, that possibly is in a loadable
1311          * module, so we have to bump that loadable module refcnt first.
1312          */
1313         if (!try_module_get(pf->owner))
1314                 goto out_release;
1315 
1316         /* Now protected by module ref count */
1317         rcu_read_unlock();
1318 
1319         err = pf->create(net, sock, protocol, kern);
1320         if (err < 0)
1321                 goto out_module_put;
1322 
1323         /*
1324          * Now to bump the refcnt of the [loadable] module that owns this
1325          * socket at sock_release time we decrement its refcnt.
1326          */
1327         if (!try_module_get(sock->ops->owner))
1328                 goto out_module_busy;
1329 
1330         /*
1331          * Now that we're done with the ->create function, the [loadable]
1332          * module can have its refcnt decremented
1333          */
1334         module_put(pf->owner);
1335         err = security_socket_post_create(sock, family, type, protocol, kern);
1336         if (err)
1337                 goto out_sock_release;
1338         *res = sock;
1339 
1340         return 0;
1341 
1342 out_module_busy:
1343         err = -EAFNOSUPPORT;
1344 out_module_put:
1345         sock->ops = NULL;
1346         module_put(pf->owner);
1347 out_sock_release:
1348         sock_release(sock);
1349         return err;
1350 
1351 out_release:
1352         rcu_read_unlock();
1353         goto out_sock_release;
1354 }
1355 EXPORT_SYMBOL(__sock_create);
1356 
1357 int sock_create(int family, int type, int protocol, struct socket **res)
1358 {
1359         return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1360 }
1361 EXPORT_SYMBOL(sock_create);
1362 
1363 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1364 {
1365         return __sock_create(&init_net, family, type, protocol, res, 1);
1366 }
1367 EXPORT_SYMBOL(sock_create_kern);
1368 
1369 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1370 {
1371         int retval;
1372         struct socket *sock;
1373         int flags;
1374 
1375         /* Check the SOCK_* constants for consistency.  */
1376         BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1377         BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1378         BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1379         BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1380 
1381         flags = type & ~SOCK_TYPE_MASK;
1382         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1383                 return -EINVAL;
1384         type &= SOCK_TYPE_MASK;
1385 
1386         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1387                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1388 
1389         retval = sock_create(family, type, protocol, &sock);
1390         if (retval < 0)
1391                 goto out;
1392 
1393         retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1394         if (retval < 0)
1395                 goto out_release;
1396 
1397 out:
1398         /* It may be already another descriptor 8) Not kernel problem. */
1399         return retval;
1400 
1401 out_release:
1402         sock_release(sock);
1403         return retval;
1404 }
1405 
1406 /*
1407  *      Create a pair of connected sockets.
1408  */
1409 
1410 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1411                 int __user *, usockvec)
1412 {
1413         struct socket *sock1, *sock2;
1414         int fd1, fd2, err;
1415         struct file *newfile1, *newfile2;
1416         int flags;
1417 
1418         flags = type & ~SOCK_TYPE_MASK;
1419         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1420                 return -EINVAL;
1421         type &= SOCK_TYPE_MASK;
1422 
1423         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1424                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1425 
1426         /*
1427          * Obtain the first socket and check if the underlying protocol
1428          * supports the socketpair call.
1429          */
1430 
1431         err = sock_create(family, type, protocol, &sock1);
1432         if (err < 0)
1433                 goto out;
1434 
1435         err = sock_create(family, type, protocol, &sock2);
1436         if (err < 0)
1437                 goto out_release_1;
1438 
1439         err = sock1->ops->socketpair(sock1, sock2);
1440         if (err < 0)
1441                 goto out_release_both;
1442 
1443         fd1 = get_unused_fd_flags(flags);
1444         if (unlikely(fd1 < 0)) {
1445                 err = fd1;
1446                 goto out_release_both;
1447         }
1448 
1449         fd2 = get_unused_fd_flags(flags);
1450         if (unlikely(fd2 < 0)) {
1451                 err = fd2;
1452                 goto out_put_unused_1;
1453         }
1454 
1455         newfile1 = sock_alloc_file(sock1, flags, NULL);
1456         if (unlikely(IS_ERR(newfile1))) {
1457                 err = PTR_ERR(newfile1);
1458                 goto out_put_unused_both;
1459         }
1460 
1461         newfile2 = sock_alloc_file(sock2, flags, NULL);
1462         if (IS_ERR(newfile2)) {
1463                 err = PTR_ERR(newfile2);
1464                 goto out_fput_1;
1465         }
1466 
1467         err = put_user(fd1, &usockvec[0]);
1468         if (err)
1469                 goto out_fput_both;
1470 
1471         err = put_user(fd2, &usockvec[1]);
1472         if (err)
1473                 goto out_fput_both;
1474 
1475         audit_fd_pair(fd1, fd2);
1476 
1477         fd_install(fd1, newfile1);
1478         fd_install(fd2, newfile2);
1479         /* fd1 and fd2 may be already another descriptors.
1480          * Not kernel problem.
1481          */
1482 
1483         return 0;
1484 
1485 out_fput_both:
1486         fput(newfile2);
1487         fput(newfile1);
1488         put_unused_fd(fd2);
1489         put_unused_fd(fd1);
1490         goto out;
1491 
1492 out_fput_1:
1493         fput(newfile1);
1494         put_unused_fd(fd2);
1495         put_unused_fd(fd1);
1496         sock_release(sock2);
1497         goto out;
1498 
1499 out_put_unused_both:
1500         put_unused_fd(fd2);
1501 out_put_unused_1:
1502         put_unused_fd(fd1);
1503 out_release_both:
1504         sock_release(sock2);
1505 out_release_1:
1506         sock_release(sock1);
1507 out:
1508         return err;
1509 }
1510 
1511 /*
1512  *      Bind a name to a socket. Nothing much to do here since it's
1513  *      the protocol's responsibility to handle the local address.
1514  *
1515  *      We move the socket address to kernel space before we call
1516  *      the protocol layer (having also checked the address is ok).
1517  */
1518 
1519 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1520 {
1521         struct socket *sock;
1522         struct sockaddr_storage address;
1523         int err, fput_needed;
1524 
1525         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1526         if (sock) {
1527                 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1528                 if (err >= 0) {
1529                         err = security_socket_bind(sock,
1530                                                    (struct sockaddr *)&address,
1531                                                    addrlen);
1532                         if (!err)
1533                                 err = sock->ops->bind(sock,
1534                                                       (struct sockaddr *)
1535                                                       &address, addrlen);
1536                 }
1537                 fput_light(sock->file, fput_needed);
1538         }
1539         return err;
1540 }
1541 
1542 /*
1543  *      Perform a listen. Basically, we allow the protocol to do anything
1544  *      necessary for a listen, and if that works, we mark the socket as
1545  *      ready for listening.
1546  */
1547 
1548 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1549 {
1550         struct socket *sock;
1551         int err, fput_needed;
1552         int somaxconn;
1553 
1554         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1555         if (sock) {
1556                 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1557                 if ((unsigned int)backlog > somaxconn)
1558                         backlog = somaxconn;
1559 
1560                 err = security_socket_listen(sock, backlog);
1561                 if (!err)
1562                         err = sock->ops->listen(sock, backlog);
1563 
1564                 fput_light(sock->file, fput_needed);
1565         }
1566         return err;
1567 }
1568 
1569 /*
1570  *      For accept, we attempt to create a new socket, set up the link
1571  *      with the client, wake up the client, then return the new
1572  *      connected fd. We collect the address of the connector in kernel
1573  *      space and move it to user at the very end. This is unclean because
1574  *      we open the socket then return an error.
1575  *
1576  *      1003.1g adds the ability to recvmsg() to query connection pending
1577  *      status to recvmsg. We need to add that support in a way thats
1578  *      clean when we restucture accept also.
1579  */
1580 
1581 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1582                 int __user *, upeer_addrlen, int, flags)
1583 {
1584         struct socket *sock, *newsock;
1585         struct file *newfile;
1586         int err, len, newfd, fput_needed;
1587         struct sockaddr_storage address;
1588 
1589         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1590                 return -EINVAL;
1591 
1592         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1593                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1594 
1595         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1596         if (!sock)
1597                 goto out;
1598 
1599         err = -ENFILE;
1600         newsock = sock_alloc();
1601         if (!newsock)
1602                 goto out_put;
1603 
1604         newsock->type = sock->type;
1605         newsock->ops = sock->ops;
1606 
1607         /*
1608          * We don't need try_module_get here, as the listening socket (sock)
1609          * has the protocol module (sock->ops->owner) held.
1610          */
1611         __module_get(newsock->ops->owner);
1612 
1613         newfd = get_unused_fd_flags(flags);
1614         if (unlikely(newfd < 0)) {
1615                 err = newfd;
1616                 sock_release(newsock);
1617                 goto out_put;
1618         }
1619         newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1620         if (unlikely(IS_ERR(newfile))) {
1621                 err = PTR_ERR(newfile);
1622                 put_unused_fd(newfd);
1623                 sock_release(newsock);
1624                 goto out_put;
1625         }
1626 
1627         err = security_socket_accept(sock, newsock);
1628         if (err)
1629                 goto out_fd;
1630 
1631         err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1632         if (err < 0)
1633                 goto out_fd;
1634 
1635         if (ccs_socket_post_accept_permission(sock, newsock)) {
1636                 err = -EAGAIN; /* Hope less harmful than -EPERM. */
1637                 goto out_fd;
1638         }
1639         if (upeer_sockaddr) {
1640                 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1641                                           &len, 2) < 0) {
1642                         err = -ECONNABORTED;
1643                         goto out_fd;
1644                 }
1645                 err = move_addr_to_user(&address,
1646                                         len, upeer_sockaddr, upeer_addrlen);
1647                 if (err < 0)
1648                         goto out_fd;
1649         }
1650 
1651         /* File flags are not inherited via accept() unlike another OSes. */
1652 
1653         fd_install(newfd, newfile);
1654         err = newfd;
1655 
1656 out_put:
1657         fput_light(sock->file, fput_needed);
1658 out:
1659         return err;
1660 out_fd:
1661         fput(newfile);
1662         put_unused_fd(newfd);
1663         goto out_put;
1664 }
1665 
1666 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1667                 int __user *, upeer_addrlen)
1668 {
1669         return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1670 }
1671 
1672 /*
1673  *      Attempt to connect to a socket with the server address.  The address
1674  *      is in user space so we verify it is OK and move it to kernel space.
1675  *
1676  *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1677  *      break bindings
1678  *
1679  *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1680  *      other SEQPACKET protocols that take time to connect() as it doesn't
1681  *      include the -EINPROGRESS status for such sockets.
1682  */
1683 
1684 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1685                 int, addrlen)
1686 {
1687         struct socket *sock;
1688         struct sockaddr_storage address;
1689         int err, fput_needed;
1690 
1691         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1692         if (!sock)
1693                 goto out;
1694         err = move_addr_to_kernel(uservaddr, addrlen, &address);
1695         if (err < 0)
1696                 goto out_put;
1697 
1698         err =
1699             security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1700         if (err)
1701                 goto out_put;
1702 
1703         err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1704                                  sock->file->f_flags);
1705 out_put:
1706         fput_light(sock->file, fput_needed);
1707 out:
1708         return err;
1709 }
1710 
1711 /*
1712  *      Get the local address ('name') of a socket object. Move the obtained
1713  *      name to user space.
1714  */
1715 
1716 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1717                 int __user *, usockaddr_len)
1718 {
1719         struct socket *sock;
1720         struct sockaddr_storage address;
1721         int len, err, fput_needed;
1722 
1723         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1724         if (!sock)
1725                 goto out;
1726 
1727         err = security_socket_getsockname(sock);
1728         if (err)
1729                 goto out_put;
1730 
1731         err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1732         if (err)
1733                 goto out_put;
1734         err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1735 
1736 out_put:
1737         fput_light(sock->file, fput_needed);
1738 out:
1739         return err;
1740 }
1741 
1742 /*
1743  *      Get the remote address ('name') of a socket object. Move the obtained
1744  *      name to user space.
1745  */
1746 
1747 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1748                 int __user *, usockaddr_len)
1749 {
1750         struct socket *sock;
1751         struct sockaddr_storage address;
1752         int len, err, fput_needed;
1753 
1754         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1755         if (sock != NULL) {
1756                 err = security_socket_getpeername(sock);
1757                 if (err) {
1758                         fput_light(sock->file, fput_needed);
1759                         return err;
1760                 }
1761 
1762                 err =
1763                     sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1764                                        1);
1765                 if (!err)
1766                         err = move_addr_to_user(&address, len, usockaddr,
1767                                                 usockaddr_len);
1768                 fput_light(sock->file, fput_needed);
1769         }
1770         return err;
1771 }
1772 
1773 /*
1774  *      Send a datagram to a given address. We move the address into kernel
1775  *      space and check the user space data area is readable before invoking
1776  *      the protocol.
1777  */
1778 
1779 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1780                 unsigned int, flags, struct sockaddr __user *, addr,
1781                 int, addr_len)
1782 {
1783         struct socket *sock;
1784         struct sockaddr_storage address;
1785         int err;
1786         struct msghdr msg;
1787         struct iovec iov;
1788         int fput_needed;
1789 
1790         if (len > INT_MAX)
1791                 len = INT_MAX;
1792         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1793         if (!sock)
1794                 goto out;
1795 
1796         iov.iov_base = buff;
1797         iov.iov_len = len;
1798         msg.msg_name = NULL;
1799         msg.msg_iov = &iov;
1800         msg.msg_iovlen = 1;
1801         msg.msg_control = NULL;
1802         msg.msg_controllen = 0;
1803         msg.msg_namelen = 0;
1804         if (addr) {
1805                 err = move_addr_to_kernel(addr, addr_len, &address);
1806                 if (err < 0)
1807                         goto out_put;
1808                 msg.msg_name = (struct sockaddr *)&address;
1809                 msg.msg_namelen = addr_len;
1810         }
1811         if (sock->file->f_flags & O_NONBLOCK)
1812                 flags |= MSG_DONTWAIT;
1813         msg.msg_flags = flags;
1814         err = sock_sendmsg(sock, &msg, len);
1815 
1816 out_put:
1817         fput_light(sock->file, fput_needed);
1818 out:
1819         return err;
1820 }
1821 
1822 /*
1823  *      Send a datagram down a socket.
1824  */
1825 
1826 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1827                 unsigned int, flags)
1828 {
1829         return sys_sendto(fd, buff, len, flags, NULL, 0);
1830 }
1831 
1832 /*
1833  *      Receive a frame from the socket and optionally record the address of the
1834  *      sender. We verify the buffers are writable and if needed move the
1835  *      sender address from kernel to user space.
1836  */
1837 
1838 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1839                 unsigned int, flags, struct sockaddr __user *, addr,
1840                 int __user *, addr_len)
1841 {
1842         struct socket *sock;
1843         struct iovec iov;
1844         struct msghdr msg;
1845         struct sockaddr_storage address;
1846         int err, err2;
1847         int fput_needed;
1848 
1849         if (size > INT_MAX)
1850                 size = INT_MAX;
1851         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1852         if (!sock)
1853                 goto out;
1854 
1855         msg.msg_control = NULL;
1856         msg.msg_controllen = 0;
1857         msg.msg_iovlen = 1;
1858         msg.msg_iov = &iov;
1859         iov.iov_len = size;
1860         iov.iov_base = ubuf;
1861         /* Save some cycles and don't copy the address if not needed */
1862         msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1863         /* We assume all kernel code knows the size of sockaddr_storage */
1864         msg.msg_namelen = 0;
1865         if (sock->file->f_flags & O_NONBLOCK)
1866                 flags |= MSG_DONTWAIT;
1867         err = sock_recvmsg(sock, &msg, size, flags);
1868 
1869         if (err >= 0 && addr != NULL) {
1870                 err2 = move_addr_to_user(&address,
1871                                          msg.msg_namelen, addr, addr_len);
1872                 if (err2 < 0)
1873                         err = err2;
1874         }
1875 
1876         fput_light(sock->file, fput_needed);
1877 out:
1878         return err;
1879 }
1880 
1881 /*
1882  *      Receive a datagram from a socket.
1883  */
1884 
1885 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1886                 unsigned int, flags)
1887 {
1888         return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1889 }
1890 
1891 /*
1892  *      Set a socket option. Because we don't know the option lengths we have
1893  *      to pass the user mode parameter for the protocols to sort out.
1894  */
1895 
1896 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1897                 char __user *, optval, int, optlen)
1898 {
1899         int err, fput_needed;
1900         struct socket *sock;
1901 
1902         if (optlen < 0)
1903                 return -EINVAL;
1904 
1905         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1906         if (sock != NULL) {
1907                 err = security_socket_setsockopt(sock, level, optname);
1908                 if (err)
1909                         goto out_put;
1910 
1911                 if (level == SOL_SOCKET)
1912                         err =
1913                             sock_setsockopt(sock, level, optname, optval,
1914                                             optlen);
1915                 else
1916                         err =
1917                             sock->ops->setsockopt(sock, level, optname, optval,
1918                                                   optlen);
1919 out_put:
1920                 fput_light(sock->file, fput_needed);
1921         }
1922         return err;
1923 }
1924 
1925 /*
1926  *      Get a socket option. Because we don't know the option lengths we have
1927  *      to pass a user mode parameter for the protocols to sort out.
1928  */
1929 
1930 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1931                 char __user *, optval, int __user *, optlen)
1932 {
1933         int err, fput_needed;
1934         struct socket *sock;
1935 
1936         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1937         if (sock != NULL) {
1938                 err = security_socket_getsockopt(sock, level, optname);
1939                 if (err)
1940                         goto out_put;
1941 
1942                 if (level == SOL_SOCKET)
1943                         err =
1944                             sock_getsockopt(sock, level, optname, optval,
1945                                             optlen);
1946                 else
1947                         err =
1948                             sock->ops->getsockopt(sock, level, optname, optval,
1949                                                   optlen);
1950 out_put:
1951                 fput_light(sock->file, fput_needed);
1952         }
1953         return err;
1954 }
1955 
1956 /*
1957  *      Shutdown a socket.
1958  */
1959 
1960 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1961 {
1962         int err, fput_needed;
1963         struct socket *sock;
1964 
1965         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1966         if (sock != NULL) {
1967                 err = security_socket_shutdown(sock, how);
1968                 if (!err)
1969                         err = sock->ops->shutdown(sock, how);
1970                 fput_light(sock->file, fput_needed);
1971         }
1972         return err;
1973 }
1974 
1975 /* A couple of helpful macros for getting the address of the 32/64 bit
1976  * fields which are the same type (int / unsigned) on our platforms.
1977  */
1978 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1979 #define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1980 #define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1981 
1982 struct used_address {
1983         struct sockaddr_storage name;
1984         unsigned int name_len;
1985 };
1986 
1987 static int copy_msghdr_from_user(struct msghdr *kmsg,
1988                                  struct msghdr __user *umsg)
1989 {
1990         if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1991                 return -EFAULT;
1992 
1993         if (kmsg->msg_name == NULL)
1994                 kmsg->msg_namelen = 0;
1995 
1996         if (kmsg->msg_namelen < 0)
1997                 return -EINVAL;
1998 
1999         if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2000                 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2001         return 0;
2002 }
2003 
2004 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
2005                          struct msghdr *msg_sys, unsigned int flags,
2006                          struct used_address *used_address, int *residue)
2007 {
2008         struct compat_msghdr __user *msg_compat =
2009             (struct compat_msghdr __user *)msg;
2010         struct sockaddr_storage address;
2011         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2012         unsigned char ctl[sizeof(struct cmsghdr) + 20]
2013             __attribute__ ((aligned(sizeof(__kernel_size_t))));
2014         /* 20 is size of ipv6_pktinfo */
2015         unsigned char *ctl_buf = ctl;
2016         int err, ctl_len, total_len;
2017 
2018         err = -EFAULT;
2019         if (MSG_CMSG_COMPAT & flags)
2020                 err = get_compat_msghdr(msg_sys, msg_compat);
2021         else
2022                 err = copy_msghdr_from_user(msg_sys, msg);
2023         if (err)
2024                 return err;
2025 
2026         if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2027                 err = -EMSGSIZE;
2028                 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2029                         goto out;
2030                 err = -ENOMEM;
2031                 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2032                               GFP_KERNEL);
2033                 if (!iov)
2034                         goto out;
2035         }
2036 
2037         /* This will also move the address data into kernel space */
2038         if (MSG_CMSG_COMPAT & flags) {
2039                 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2040         } else
2041                 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2042         if (err < 0)
2043                 goto out_freeiov;
2044         total_len = err;
2045 
2046         err = -ENOBUFS;
2047 
2048         if (msg_sys->msg_controllen > INT_MAX)
2049                 goto out_freeiov;
2050         ctl_len = msg_sys->msg_controllen;
2051         if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2052                 err =
2053                     cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2054                                                      sizeof(ctl));
2055                 if (err)
2056                         goto out_freeiov;
2057                 ctl_buf = msg_sys->msg_control;
2058                 ctl_len = msg_sys->msg_controllen;
2059         } else if (ctl_len) {
2060                 if (ctl_len > sizeof(ctl)) {
2061                         ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2062                         if (ctl_buf == NULL)
2063                                 goto out_freeiov;
2064                 }
2065                 err = -EFAULT;
2066                 /*
2067                  * Careful! Before this, msg_sys->msg_control contains a user pointer.
2068                  * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2069                  * checking falls down on this.
2070                  */
2071                 if (copy_from_user(ctl_buf,
2072                                    (void __user __force *)msg_sys->msg_control,
2073                                    ctl_len))
2074                         goto out_freectl;
2075                 msg_sys->msg_control = ctl_buf;
2076         }
2077         msg_sys->msg_flags = flags;
2078 
2079         if (sock->file->f_flags & O_NONBLOCK)
2080                 msg_sys->msg_flags |= MSG_DONTWAIT;
2081         /*
2082          * If this is sendmmsg() and current destination address is same as
2083          * previously succeeded address, omit asking LSM's decision.
2084          * used_address->name_len is initialized to UINT_MAX so that the first
2085          * destination address never matches.
2086          */
2087         if (used_address && msg_sys->msg_name &&
2088             used_address->name_len == msg_sys->msg_namelen &&
2089             !memcmp(&used_address->name, msg_sys->msg_name,
2090                     used_address->name_len)) {
2091                 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2092                 goto out_freectl;
2093         }
2094         err = sock_sendmsg(sock, msg_sys, total_len);
2095         /*
2096          * If this is sendmmsg() and sending to current destination address was
2097          * successful, remember it.
2098          */
2099         if (used_address && err >= 0) {
2100                 used_address->name_len = msg_sys->msg_namelen;
2101                 if (msg_sys->msg_name)
2102                         memcpy(&used_address->name, msg_sys->msg_name,
2103                                used_address->name_len);
2104         }
2105         if (residue && err >= 0)
2106                 *residue = total_len - err;
2107 
2108 out_freectl:
2109         if (ctl_buf != ctl)
2110                 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2111 out_freeiov:
2112         if (iov != iovstack)
2113                 kfree(iov);
2114 out:
2115         return err;
2116 }
2117 
2118 /*
2119  *      BSD sendmsg interface
2120  */
2121 
2122 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2123 {
2124         int fput_needed, err;
2125         struct msghdr msg_sys;
2126         struct socket *sock;
2127 
2128         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2129         if (!sock)
2130                 goto out;
2131 
2132         err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, NULL);
2133 
2134         fput_light(sock->file, fput_needed);
2135 out:
2136         return err;
2137 }
2138 
2139 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2140 {
2141         if (flags & MSG_CMSG_COMPAT)
2142                 return -EINVAL;
2143         return __sys_sendmsg(fd, msg, flags);
2144 }
2145 
2146 /*
2147  *      Linux sendmmsg interface
2148  */
2149 
2150 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2151                    unsigned int flags)
2152 {
2153         int fput_needed, err, datagrams;
2154         struct socket *sock;
2155         struct mmsghdr __user *entry;
2156         struct compat_mmsghdr __user *compat_entry;
2157         struct msghdr msg_sys;
2158         struct used_address used_address;
2159         int residue;
2160 
2161         if (vlen > UIO_MAXIOV)
2162                 vlen = UIO_MAXIOV;
2163 
2164         datagrams = 0;
2165 
2166         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2167         if (!sock)
2168                 return err;
2169 
2170         used_address.name_len = UINT_MAX;
2171         entry = mmsg;
2172         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2173         err = 0;
2174 
2175         while (datagrams < vlen) {
2176                 if (MSG_CMSG_COMPAT & flags) {
2177                         err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2178                                              &msg_sys, flags, &used_address,
2179                                              &residue);
2180                         if (err < 0)
2181                                 break;
2182                         err = __put_user(err, &compat_entry->msg_len);
2183                         ++compat_entry;
2184                 } else {
2185                         err = ___sys_sendmsg(sock,
2186                                              (struct msghdr __user *)entry,
2187                                              &msg_sys, flags, &used_address,
2188                                              &residue);
2189                         if (err < 0)
2190                                 break;
2191                         err = put_user(err, &entry->msg_len);
2192                         ++entry;
2193                 }
2194 
2195                 if (err)
2196                         break;
2197                 ++datagrams;
2198                 if (residue)
2199                         break;
2200         }
2201 
2202         fput_light(sock->file, fput_needed);
2203 
2204         /* We only return an error if no datagrams were able to be sent */
2205         if (datagrams != 0)
2206                 return datagrams;
2207 
2208         return err;
2209 }
2210 
2211 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2212                 unsigned int, vlen, unsigned int, flags)
2213 {
2214         if (flags & MSG_CMSG_COMPAT)
2215                 return -EINVAL;
2216         return __sys_sendmmsg(fd, mmsg, vlen, flags);
2217 }
2218 
2219 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2220                          struct msghdr *msg_sys, unsigned int flags, int nosec)
2221 {
2222         struct compat_msghdr __user *msg_compat =
2223             (struct compat_msghdr __user *)msg;
2224         struct iovec iovstack[UIO_FASTIOV];
2225         struct iovec *iov = iovstack;
2226         unsigned long cmsg_ptr;
2227         int err, total_len, len;
2228 
2229         /* kernel mode address */
2230         struct sockaddr_storage addr;
2231 
2232         /* user mode address pointers */
2233         struct sockaddr __user *uaddr;
2234         int __user *uaddr_len;
2235 
2236         if (MSG_CMSG_COMPAT & flags)
2237                 err = get_compat_msghdr(msg_sys, msg_compat);
2238         else
2239                 err = copy_msghdr_from_user(msg_sys, msg);
2240         if (err)
2241                 return err;
2242 
2243         if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2244                 err = -EMSGSIZE;
2245                 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2246                         goto out;
2247                 err = -ENOMEM;
2248                 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2249                               GFP_KERNEL);
2250                 if (!iov)
2251                         goto out;
2252         }
2253 
2254         /* Save the user-mode address (verify_iovec will change the
2255          * kernel msghdr to use the kernel address space)
2256          */
2257         uaddr = (__force void __user *)msg_sys->msg_name;
2258         uaddr_len = COMPAT_NAMELEN(msg);
2259         if (MSG_CMSG_COMPAT & flags)
2260                 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2261         else
2262                 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2263         if (err < 0)
2264                 goto out_freeiov;
2265         total_len = err;
2266 
2267         cmsg_ptr = (unsigned long)msg_sys->msg_control;
2268         msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2269 
2270         /* We assume all kernel code knows the size of sockaddr_storage */
2271         msg_sys->msg_namelen = 0;
2272 
2273         if (sock->file->f_flags & O_NONBLOCK)
2274                 flags |= MSG_DONTWAIT;
2275         err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2276                                                           total_len, flags);
2277         if (err < 0)
2278                 goto out_freeiov;
2279         len = err;
2280 
2281         if (uaddr != NULL) {
2282                 err = move_addr_to_user(&addr,
2283                                         msg_sys->msg_namelen, uaddr,
2284                                         uaddr_len);
2285                 if (err < 0)
2286                         goto out_freeiov;
2287         }
2288         err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2289                          COMPAT_FLAGS(msg));
2290         if (err)
2291                 goto out_freeiov;
2292         if (MSG_CMSG_COMPAT & flags)
2293                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2294                                  &msg_compat->msg_controllen);
2295         else
2296                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2297                                  &msg->msg_controllen);
2298         if (err)
2299                 goto out_freeiov;
2300         err = len;
2301 
2302 out_freeiov:
2303         if (iov != iovstack)
2304                 kfree(iov);
2305 out:
2306         return err;
2307 }
2308 
2309 /*
2310  *      BSD recvmsg interface
2311  */
2312 
2313 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2314 {
2315         int fput_needed, err;
2316         struct msghdr msg_sys;
2317         struct socket *sock;
2318 
2319         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2320         if (!sock)
2321                 goto out;
2322 
2323         err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2324 
2325         fput_light(sock->file, fput_needed);
2326 out:
2327         return err;
2328 }
2329 
2330 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2331                 unsigned int, flags)
2332 {
2333         if (flags & MSG_CMSG_COMPAT)
2334                 return -EINVAL;
2335         return __sys_recvmsg(fd, msg, flags);
2336 }
2337 
2338 /*
2339  *     Linux recvmmsg interface
2340  */
2341 
2342 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2343                    unsigned int flags, struct timespec *timeout)
2344 {
2345         int fput_needed, err, datagrams;
2346         struct socket *sock;
2347         struct mmsghdr __user *entry;
2348         struct compat_mmsghdr __user *compat_entry;
2349         struct msghdr msg_sys;
2350         struct timespec end_time;
2351 
2352         if (timeout &&
2353             poll_select_set_timeout(&end_time, timeout->tv_sec,
2354                                     timeout->tv_nsec))
2355                 return -EINVAL;
2356 
2357         datagrams = 0;
2358 
2359         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2360         if (!sock)
2361                 return err;
2362 
2363         err = sock_error(sock->sk);
2364         if (err) {
2365                 datagrams = err;
2366                 goto out_put;
2367         }
2368 
2369         entry = mmsg;
2370         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2371 
2372         while (datagrams < vlen) {
2373                 /*
2374                  * No need to ask LSM for more than the first datagram.
2375                  */
2376                 if (MSG_CMSG_COMPAT & flags) {
2377                         err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2378                                              &msg_sys, flags & ~MSG_WAITFORONE,
2379                                              datagrams);
2380                         if (err < 0)
2381                                 break;
2382                         err = __put_user(err, &compat_entry->msg_len);
2383                         ++compat_entry;
2384                 } else {
2385                         err = ___sys_recvmsg(sock,
2386                                              (struct msghdr __user *)entry,
2387                                              &msg_sys, flags & ~MSG_WAITFORONE,
2388                                              datagrams);
2389                         if (err < 0)
2390                                 break;
2391                         err = put_user(err, &entry->msg_len);
2392                         ++entry;
2393                 }
2394 
2395                 if (err)
2396                         break;
2397                 ++datagrams;
2398 
2399                 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2400                 if (flags & MSG_WAITFORONE)
2401                         flags |= MSG_DONTWAIT;
2402 
2403                 if (timeout) {
2404                         ktime_get_ts(timeout);
2405                         *timeout = timespec_sub(end_time, *timeout);
2406                         if (timeout->tv_sec < 0) {
2407                                 timeout->tv_sec = timeout->tv_nsec = 0;
2408                                 break;
2409                         }
2410 
2411                         /* Timeout, return less than vlen datagrams */
2412                         if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2413                                 break;
2414                 }
2415 
2416                 /* Out of band data, return right away */
2417                 if (msg_sys.msg_flags & MSG_OOB)
2418                         break;
2419         }
2420 
2421         if (err == 0)
2422                 goto out_put;
2423 
2424         if (datagrams == 0) {
2425                 datagrams = err;
2426                 goto out_put;
2427         }
2428 
2429         /*
2430          * We may return less entries than requested (vlen) if the
2431          * sock is non block and there aren't enough datagrams...
2432          */
2433         if (err != -EAGAIN) {
2434                 /*
2435                  * ... or  if recvmsg returns an error after we
2436                  * received some datagrams, where we record the
2437                  * error to return on the next call or if the
2438                  * app asks about it using getsockopt(SO_ERROR).
2439                  */
2440                 sock->sk->sk_err = -err;
2441         }
2442 out_put:
2443         fput_light(sock->file, fput_needed);
2444 
2445         return datagrams;
2446 }
2447 
2448 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2449                 unsigned int, vlen, unsigned int, flags,
2450                 struct timespec __user *, timeout)
2451 {
2452         int datagrams;
2453         struct timespec timeout_sys;
2454 
2455         if (flags & MSG_CMSG_COMPAT)
2456                 return -EINVAL;
2457 
2458         if (!timeout)
2459                 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2460 
2461         if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2462                 return -EFAULT;
2463 
2464         datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2465 
2466         if (datagrams > 0 &&
2467             copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2468                 datagrams = -EFAULT;
2469 
2470         return datagrams;
2471 }
2472 
2473 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2474 /* Argument list sizes for sys_socketcall */
2475 #define AL(x) ((x) * sizeof(unsigned long))
2476 static const unsigned char nargs[21] = {
2477         AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2478         AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2479         AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2480         AL(4), AL(5), AL(4)
2481 };
2482 
2483 #undef AL
2484 
2485 /*
2486  *      System call vectors.
2487  *
2488  *      Argument checking cleaned up. Saved 20% in size.
2489  *  This function doesn't need to set the kernel lock because
2490  *  it is set by the callees.
2491  */
2492 
2493 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2494 {
2495         unsigned long a[AUDITSC_ARGS];
2496         unsigned long a0, a1;
2497         int err;
2498         unsigned int len;
2499 
2500         if (call < 1 || call > SYS_SENDMMSG)
2501                 return -EINVAL;
2502         call = array_index_nospec(call, SYS_SENDMMSG + 1);
2503 
2504         len = nargs[call];
2505         if (len > sizeof(a))
2506                 return -EINVAL;
2507 
2508         /* copy_from_user should be SMP safe. */
2509         if (copy_from_user(a, args, len))
2510                 return -EFAULT;
2511 
2512         err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2513         if (err)
2514                 return err;
2515 
2516         a0 = a[0];
2517         a1 = a[1];
2518 
2519         switch (call) {
2520         case SYS_SOCKET:
2521                 err = sys_socket(a0, a1, a[2]);
2522                 break;
2523         case SYS_BIND:
2524                 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2525                 break;
2526         case SYS_CONNECT:
2527                 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2528                 break;
2529         case SYS_LISTEN:
2530                 err = sys_listen(a0, a1);
2531                 break;
2532         case SYS_ACCEPT:
2533                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2534                                   (int __user *)a[2], 0);
2535                 break;
2536         case SYS_GETSOCKNAME:
2537                 err =
2538                     sys_getsockname(a0, (struct sockaddr __user *)a1,
2539                                     (int __user *)a[2]);
2540                 break;
2541         case SYS_GETPEERNAME:
2542                 err =
2543                     sys_getpeername(a0, (struct sockaddr __user *)a1,
2544                                     (int __user *)a[2]);
2545                 break;
2546         case SYS_SOCKETPAIR:
2547                 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2548                 break;
2549         case SYS_SEND:
2550                 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2551                 break;
2552         case SYS_SENDTO:
2553                 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2554                                  (struct sockaddr __user *)a[4], a[5]);
2555                 break;
2556         case SYS_RECV:
2557                 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2558                 break;
2559         case SYS_RECVFROM:
2560                 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2561                                    (struct sockaddr __user *)a[4],
2562                                    (int __user *)a[5]);
2563                 break;
2564         case SYS_SHUTDOWN:
2565                 err = sys_shutdown(a0, a1);
2566                 break;
2567         case SYS_SETSOCKOPT:
2568                 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2569                 break;
2570         case SYS_GETSOCKOPT:
2571                 err =
2572                     sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2573                                    (int __user *)a[4]);
2574                 break;
2575         case SYS_SENDMSG:
2576                 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2577                 break;
2578         case SYS_SENDMMSG:
2579                 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2580                 break;
2581         case SYS_RECVMSG:
2582                 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2583                 break;
2584         case SYS_RECVMMSG:
2585                 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2586                                    (struct timespec __user *)a[4]);
2587                 break;
2588         case SYS_ACCEPT4:
2589                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2590                                   (int __user *)a[2], a[3]);
2591                 break;
2592         default:
2593                 err = -EINVAL;
2594                 break;
2595         }
2596         return err;
2597 }
2598 
2599 #endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2600 
2601 /**
2602  *      sock_register - add a socket protocol handler
2603  *      @ops: description of protocol
2604  *
2605  *      This function is called by a protocol handler that wants to
2606  *      advertise its address family, and have it linked into the
2607  *      socket interface. The value ops->family coresponds to the
2608  *      socket system call protocol family.
2609  */
2610 int sock_register(const struct net_proto_family *ops)
2611 {
2612         int err;
2613 
2614         if (ops->family >= NPROTO) {
2615                 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2616                 return -ENOBUFS;
2617         }
2618 
2619         spin_lock(&net_family_lock);
2620         if (rcu_dereference_protected(net_families[ops->family],
2621                                       lockdep_is_held(&net_family_lock)))
2622                 err = -EEXIST;
2623         else {
2624                 rcu_assign_pointer(net_families[ops->family], ops);
2625                 err = 0;
2626         }
2627         spin_unlock(&net_family_lock);
2628 
2629         pr_info("NET: Registered protocol family %d\n", ops->family);
2630         return err;
2631 }
2632 EXPORT_SYMBOL(sock_register);
2633 
2634 /**
2635  *      sock_unregister - remove a protocol handler
2636  *      @family: protocol family to remove
2637  *
2638  *      This function is called by a protocol handler that wants to
2639  *      remove its address family, and have it unlinked from the
2640  *      new socket creation.
2641  *
2642  *      If protocol handler is a module, then it can use module reference
2643  *      counts to protect against new references. If protocol handler is not
2644  *      a module then it needs to provide its own protection in
2645  *      the ops->create routine.
2646  */
2647 void sock_unregister(int family)
2648 {
2649         BUG_ON(family < 0 || family >= NPROTO);
2650 
2651         spin_lock(&net_family_lock);
2652         RCU_INIT_POINTER(net_families[family], NULL);
2653         spin_unlock(&net_family_lock);
2654 
2655         synchronize_rcu();
2656 
2657         pr_info("NET: Unregistered protocol family %d\n", family);
2658 }
2659 EXPORT_SYMBOL(sock_unregister);
2660 
2661 static int __init sock_init(void)
2662 {
2663         int err;
2664         /*
2665          *      Initialize the network sysctl infrastructure.
2666          */
2667         err = net_sysctl_init();
2668         if (err)
2669                 goto out;
2670 
2671         /*
2672          *      Initialize skbuff SLAB cache
2673          */
2674         skb_init();
2675 
2676         /*
2677          *      Initialize the protocols module.
2678          */
2679 
2680         init_inodecache();
2681 
2682         err = register_filesystem(&sock_fs_type);
2683         if (err)
2684                 goto out_fs;
2685         sock_mnt = kern_mount(&sock_fs_type);
2686         if (IS_ERR(sock_mnt)) {
2687                 err = PTR_ERR(sock_mnt);
2688                 goto out_mount;
2689         }
2690 
2691         /* The real protocol initialization is performed in later initcalls.
2692          */
2693 
2694 #ifdef CONFIG_NETFILTER
2695         err = netfilter_init();
2696         if (err)
2697                 goto out;
2698 #endif
2699 
2700         ptp_classifier_init();
2701 
2702 out:
2703         return err;
2704 
2705 out_mount:
2706         unregister_filesystem(&sock_fs_type);
2707 out_fs:
2708         goto out;
2709 }
2710 
2711 core_initcall(sock_init);       /* early initcall */
2712 
2713 #ifdef CONFIG_PROC_FS
2714 void socket_seq_show(struct seq_file *seq)
2715 {
2716         int cpu;
2717         int counter = 0;
2718 
2719         for_each_possible_cpu(cpu)
2720             counter += per_cpu(sockets_in_use, cpu);
2721 
2722         /* It can be negative, by the way. 8) */
2723         if (counter < 0)
2724                 counter = 0;
2725 
2726         seq_printf(seq, "sockets: used %d\n", counter);
2727 }
2728 #endif                          /* CONFIG_PROC_FS */
2729 
2730 #ifdef CONFIG_COMPAT
2731 static int do_siocgstamp(struct net *net, struct socket *sock,
2732                          unsigned int cmd, void __user *up)
2733 {
2734         mm_segment_t old_fs = get_fs();
2735         struct timeval ktv;
2736         int err;
2737 
2738         set_fs(KERNEL_DS);
2739         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2740         set_fs(old_fs);
2741         if (!err)
2742                 err = compat_put_timeval(&ktv, up);
2743 
2744         return err;
2745 }
2746 
2747 static int do_siocgstampns(struct net *net, struct socket *sock,
2748                            unsigned int cmd, void __user *up)
2749 {
2750         mm_segment_t old_fs = get_fs();
2751         struct timespec kts;
2752         int err;
2753 
2754         set_fs(KERNEL_DS);
2755         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2756         set_fs(old_fs);
2757         if (!err)
2758                 err = compat_put_timespec(&kts, up);
2759 
2760         return err;
2761 }
2762 
2763 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2764 {
2765         struct ifreq __user *uifr;
2766         int err;
2767 
2768         uifr = compat_alloc_user_space(sizeof(struct ifreq));
2769         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2770                 return -EFAULT;
2771 
2772         err = dev_ioctl(net, SIOCGIFNAME, uifr);
2773         if (err)
2774                 return err;
2775 
2776         if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2777                 return -EFAULT;
2778 
2779         return 0;
2780 }
2781 
2782 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2783 {
2784         struct compat_ifconf ifc32;
2785         struct ifconf ifc;
2786         struct ifconf __user *uifc;
2787         struct compat_ifreq __user *ifr32;
2788         struct ifreq __user *ifr;
2789         unsigned int i, j;
2790         int err;
2791 
2792         if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2793                 return -EFAULT;
2794 
2795         memset(&ifc, 0, sizeof(ifc));
2796         if (ifc32.ifcbuf == 0) {
2797                 ifc32.ifc_len = 0;
2798                 ifc.ifc_len = 0;
2799                 ifc.ifc_req = NULL;
2800                 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2801         } else {
2802                 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2803                         sizeof(struct ifreq);
2804                 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2805                 ifc.ifc_len = len;
2806                 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2807                 ifr32 = compat_ptr(ifc32.ifcbuf);
2808                 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2809                         if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2810                                 return -EFAULT;
2811                         ifr++;
2812                         ifr32++;
2813                 }
2814         }
2815         if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2816                 return -EFAULT;
2817 
2818         err = dev_ioctl(net, SIOCGIFCONF, uifc);
2819         if (err)
2820                 return err;
2821 
2822         if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2823                 return -EFAULT;
2824 
2825         ifr = ifc.ifc_req;
2826         ifr32 = compat_ptr(ifc32.ifcbuf);
2827         for (i = 0, j = 0;
2828              i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2829              i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2830                 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2831                         return -EFAULT;
2832                 ifr32++;
2833                 ifr++;
2834         }
2835 
2836         if (ifc32.ifcbuf == 0) {
2837                 /* Translate from 64-bit structure multiple to
2838                  * a 32-bit one.
2839                  */
2840                 i = ifc.ifc_len;
2841                 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2842                 ifc32.ifc_len = i;
2843         } else {
2844                 ifc32.ifc_len = i;
2845         }
2846         if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2847                 return -EFAULT;
2848 
2849         return 0;
2850 }
2851 
2852 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2853 {
2854         struct compat_ethtool_rxnfc __user *compat_rxnfc;
2855         bool convert_in = false, convert_out = false;
2856         size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2857         struct ethtool_rxnfc __user *rxnfc;
2858         struct ifreq __user *ifr;
2859         u32 rule_cnt = 0, actual_rule_cnt;
2860         u32 ethcmd;
2861         u32 data;
2862         int ret;
2863 
2864         if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2865                 return -EFAULT;
2866 
2867         compat_rxnfc = compat_ptr(data);
2868 
2869         if (get_user(ethcmd, &compat_rxnfc->cmd))
2870                 return -EFAULT;
2871 
2872         /* Most ethtool structures are defined without padding.
2873          * Unfortunately struct ethtool_rxnfc is an exception.
2874          */
2875         switch (ethcmd) {
2876         default:
2877                 break;
2878         case ETHTOOL_GRXCLSRLALL:
2879                 /* Buffer size is variable */
2880                 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2881                         return -EFAULT;
2882                 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2883                         return -ENOMEM;
2884                 buf_size += rule_cnt * sizeof(u32);
2885                 /* fall through */
2886         case ETHTOOL_GRXRINGS:
2887         case ETHTOOL_GRXCLSRLCNT:
2888         case ETHTOOL_GRXCLSRULE:
2889         case ETHTOOL_SRXCLSRLINS:
2890                 convert_out = true;
2891                 /* fall through */
2892         case ETHTOOL_SRXCLSRLDEL:
2893                 buf_size += sizeof(struct ethtool_rxnfc);
2894                 convert_in = true;
2895                 break;
2896         }
2897 
2898         ifr = compat_alloc_user_space(buf_size);
2899         rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2900 
2901         if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2902                 return -EFAULT;
2903 
2904         if (put_user(convert_in ? rxnfc : compat_ptr(data),
2905                      &ifr->ifr_ifru.ifru_data))
2906                 return -EFAULT;
2907 
2908         if (convert_in) {
2909                 /* We expect there to be holes between fs.m_ext and
2910                  * fs.ring_cookie and at the end of fs, but nowhere else.
2911                  */
2912                 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2913                              sizeof(compat_rxnfc->fs.m_ext) !=
2914                              offsetof(struct ethtool_rxnfc, fs.m_ext) +
2915                              sizeof(rxnfc->fs.m_ext));
2916                 BUILD_BUG_ON(
2917                         offsetof(struct compat_ethtool_rxnfc, fs.location) -
2918                         offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2919                         offsetof(struct ethtool_rxnfc, fs.location) -
2920                         offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2921 
2922                 if (copy_in_user(rxnfc, compat_rxnfc,
2923                                  (void __user *)(&rxnfc->fs.m_ext + 1) -
2924                                  (void __user *)rxnfc) ||
2925                     copy_in_user(&rxnfc->fs.ring_cookie,
2926                                  &compat_rxnfc->fs.ring_cookie,
2927                                  (void __user *)(&rxnfc->fs.location + 1) -
2928                                  (void __user *)&rxnfc->fs.ring_cookie) ||
2929                     copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2930                                  sizeof(rxnfc->rule_cnt)))
2931                         return -EFAULT;
2932         }
2933 
2934         ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2935         if (ret)
2936                 return ret;
2937 
2938         if (convert_out) {
2939                 if (copy_in_user(compat_rxnfc, rxnfc,
2940                                  (const void __user *)(&rxnfc->fs.m_ext + 1) -
2941                                  (const void __user *)rxnfc) ||
2942                     copy_in_user(&compat_rxnfc->fs.ring_cookie,
2943                                  &rxnfc->fs.ring_cookie,
2944                                  (const void __user *)(&rxnfc->fs.location + 1) -
2945                                  (const void __user *)&rxnfc->fs.ring_cookie) ||
2946                     copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2947                                  sizeof(rxnfc->rule_cnt)))
2948                         return -EFAULT;
2949 
2950                 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2951                         /* As an optimisation, we only copy the actual
2952                          * number of rules that the underlying
2953                          * function returned.  Since Mallory might
2954                          * change the rule count in user memory, we
2955                          * check that it is less than the rule count
2956                          * originally given (as the user buffer size),
2957                          * which has been range-checked.
2958                          */
2959                         if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2960                                 return -EFAULT;
2961                         if (actual_rule_cnt < rule_cnt)
2962                                 rule_cnt = actual_rule_cnt;
2963                         if (copy_in_user(&compat_rxnfc->rule_locs[0],
2964                                          &rxnfc->rule_locs[0],
2965                                          rule_cnt * sizeof(u32)))
2966                                 return -EFAULT;
2967                 }
2968         }
2969 
2970         return 0;
2971 }
2972 
2973 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2974 {
2975         void __user *uptr;
2976         compat_uptr_t uptr32;
2977         struct ifreq __user *uifr;
2978 
2979         uifr = compat_alloc_user_space(sizeof(*uifr));
2980         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2981                 return -EFAULT;
2982 
2983         if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2984                 return -EFAULT;
2985 
2986         uptr = compat_ptr(uptr32);
2987 
2988         if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2989                 return -EFAULT;
2990 
2991         return dev_ioctl(net, SIOCWANDEV, uifr);
2992 }
2993 
2994 static int bond_ioctl(struct net *net, unsigned int cmd,
2995                          struct compat_ifreq __user *ifr32)
2996 {
2997         struct ifreq kifr;
2998         mm_segment_t old_fs;
2999         int err;
3000 
3001         switch (cmd) {
3002         case SIOCBONDENSLAVE:
3003         case SIOCBONDRELEASE:
3004         case SIOCBONDSETHWADDR:
3005         case SIOCBONDCHANGEACTIVE:
3006                 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
3007                         return -EFAULT;
3008 
3009                 old_fs = get_fs();
3010                 set_fs(KERNEL_DS);
3011                 err = dev_ioctl(net, cmd,
3012                                 (struct ifreq __user __force *) &kifr);
3013                 set_fs(old_fs);
3014 
3015                 return err;
3016         default:
3017                 return -ENOIOCTLCMD;
3018         }
3019 }
3020 
3021 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3022 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3023                                  struct compat_ifreq __user *u_ifreq32)
3024 {
3025         struct ifreq __user *u_ifreq64;
3026         char tmp_buf[IFNAMSIZ];
3027         void __user *data64;
3028         u32 data32;
3029 
3030         if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3031                            IFNAMSIZ))
3032                 return -EFAULT;
3033         if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3034                 return -EFAULT;
3035         data64 = compat_ptr(data32);
3036 
3037         u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3038 
3039         if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3040                          IFNAMSIZ))
3041                 return -EFAULT;
3042         if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3043                 return -EFAULT;
3044 
3045         return dev_ioctl(net, cmd, u_ifreq64);
3046 }
3047 
3048 static int dev_ifsioc(struct net *net, struct socket *sock,
3049                          unsigned int cmd, struct compat_ifreq __user *uifr32)
3050 {
3051         struct ifreq __user *uifr;
3052         int err;
3053 
3054         uifr = compat_alloc_user_space(sizeof(*uifr));
3055         if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3056                 return -EFAULT;
3057 
3058         err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3059 
3060         if (!err) {
3061                 switch (cmd) {
3062                 case SIOCGIFFLAGS:
3063                 case SIOCGIFMETRIC:
3064                 case SIOCGIFMTU:
3065                 case SIOCGIFMEM:
3066                 case SIOCGIFHWADDR:
3067                 case SIOCGIFINDEX:
3068                 case SIOCGIFADDR:
3069                 case SIOCGIFBRDADDR:
3070                 case SIOCGIFDSTADDR:
3071                 case SIOCGIFNETMASK:
3072                 case SIOCGIFPFLAGS:
3073                 case SIOCGIFTXQLEN:
3074                 case SIOCGMIIPHY:
3075                 case SIOCGMIIREG:
3076                         if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3077                                 err = -EFAULT;
3078                         break;
3079                 }
3080         }
3081         return err;
3082 }
3083 
3084 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3085                         struct compat_ifreq __user *uifr32)
3086 {
3087         struct ifreq ifr;
3088         struct compat_ifmap __user *uifmap32;
3089         mm_segment_t old_fs;
3090         int err;
3091 
3092         uifmap32 = &uifr32->ifr_ifru.ifru_map;
3093         err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3094         err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3095         err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3096         err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3097         err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3098         err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3099         err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3100         if (err)
3101                 return -EFAULT;
3102 
3103         old_fs = get_fs();
3104         set_fs(KERNEL_DS);
3105         err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3106         set_fs(old_fs);
3107 
3108         if (cmd == SIOCGIFMAP && !err) {
3109                 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3110                 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3111                 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3112                 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3113                 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3114                 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3115                 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3116                 if (err)
3117                         err = -EFAULT;
3118         }
3119         return err;
3120 }
3121 
3122 struct rtentry32 {
3123         u32             rt_pad1;
3124         struct sockaddr rt_dst;         /* target address               */
3125         struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3126         struct sockaddr rt_genmask;     /* target network mask (IP)     */
3127         unsigned short  rt_flags;
3128         short           rt_pad2;
3129         u32             rt_pad3;
3130         unsigned char   rt_tos;
3131         unsigned char   rt_class;
3132         short           rt_pad4;
3133         short           rt_metric;      /* +1 for binary compatibility! */
3134         /* char * */ u32 rt_dev;        /* forcing the device at add    */
3135         u32             rt_mtu;         /* per route MTU/Window         */
3136         u32             rt_window;      /* Window clamping              */
3137         unsigned short  rt_irtt;        /* Initial RTT                  */
3138 };
3139 
3140 struct in6_rtmsg32 {
3141         struct in6_addr         rtmsg_dst;
3142         struct in6_addr         rtmsg_src;
3143         struct in6_addr         rtmsg_gateway;
3144         u32                     rtmsg_type;
3145         u16                     rtmsg_dst_len;
3146         u16                     rtmsg_src_len;
3147         u32                     rtmsg_metric;
3148         u32                     rtmsg_info;
3149         u32                     rtmsg_flags;
3150         s32                     rtmsg_ifindex;
3151 };
3152 
3153 static int routing_ioctl(struct net *net, struct socket *sock,
3154                          unsigned int cmd, void __user *argp)
3155 {
3156         int ret;
3157         void *r = NULL;
3158         struct in6_rtmsg r6;
3159         struct rtentry r4;
3160         char devname[16];
3161         u32 rtdev;
3162         mm_segment_t old_fs = get_fs();
3163 
3164         if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3165                 struct in6_rtmsg32 __user *ur6 = argp;
3166                 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3167                         3 * sizeof(struct in6_addr));
3168                 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3169                 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3170                 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3171                 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3172                 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3173                 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3174                 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3175 
3176                 r = (void *) &r6;
3177         } else { /* ipv4 */
3178                 struct rtentry32 __user *ur4 = argp;
3179                 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3180                                         3 * sizeof(struct sockaddr));
3181                 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3182                 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3183                 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3184                 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3185                 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3186                 ret |= get_user(rtdev, &(ur4->rt_dev));
3187                 if (rtdev) {
3188                         ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3189                         r4.rt_dev = (char __user __force *)devname;
3190                         devname[15] = 0;
3191                 } else
3192                         r4.rt_dev = NULL;
3193 
3194                 r = (void *) &r4;
3195         }
3196 
3197         if (ret) {
3198                 ret = -EFAULT;
3199                 goto out;
3200         }
3201 
3202         set_fs(KERNEL_DS);
3203         ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3204         set_fs(old_fs);
3205 
3206 out:
3207         return ret;
3208 }
3209 
3210 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3211  * for some operations; this forces use of the newer bridge-utils that
3212  * use compatible ioctls
3213  */
3214 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3215 {
3216         compat_ulong_t tmp;
3217 
3218         if (get_user(tmp, argp))
3219                 return -EFAULT;
3220         if (tmp == BRCTL_GET_VERSION)
3221                 return BRCTL_VERSION + 1;
3222         return -EINVAL;
3223 }
3224 
3225 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3226                          unsigned int cmd, unsigned long arg)
3227 {
3228         void __user *argp = compat_ptr(arg);
3229         struct sock *sk = sock->sk;
3230         struct net *net = sock_net(sk);
3231 
3232         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3233                 return compat_ifr_data_ioctl(net, cmd, argp);
3234 
3235         switch (cmd) {
3236         case SIOCSIFBR:
3237         case SIOCGIFBR:
3238                 return old_bridge_ioctl(argp);
3239         case SIOCGIFNAME:
3240                 return dev_ifname32(net, argp);
3241         case SIOCGIFCONF:
3242                 return dev_ifconf(net, argp);
3243         case SIOCETHTOOL:
3244                 return ethtool_ioctl(net, argp);
3245         case SIOCWANDEV:
3246                 return compat_siocwandev(net, argp);
3247         case SIOCGIFMAP:
3248         case SIOCSIFMAP:
3249                 return compat_sioc_ifmap(net, cmd, argp);
3250         case SIOCBONDENSLAVE:
3251         case SIOCBONDRELEASE:
3252         case SIOCBONDSETHWADDR:
3253         case SIOCBONDCHANGEACTIVE:
3254                 return bond_ioctl(net, cmd, argp);
3255         case SIOCADDRT:
3256         case SIOCDELRT:
3257                 return routing_ioctl(net, sock, cmd, argp);
3258         case SIOCGSTAMP:
3259                 return do_siocgstamp(net, sock, cmd, argp);
3260         case SIOCGSTAMPNS:
3261                 return do_siocgstampns(net, sock, cmd, argp);
3262         case SIOCBONDSLAVEINFOQUERY:
3263         case SIOCBONDINFOQUERY:
3264         case SIOCSHWTSTAMP:
3265         case SIOCGHWTSTAMP:
3266                 return compat_ifr_data_ioctl(net, cmd, argp);
3267 
3268         case FIOSETOWN:
3269         case SIOCSPGRP:
3270         case FIOGETOWN:
3271         case SIOCGPGRP:
3272         case SIOCBRADDBR:
3273         case SIOCBRDELBR:
3274         case SIOCGIFVLAN:
3275         case SIOCSIFVLAN:
3276         case SIOCADDDLCI:
3277         case SIOCDELDLCI:
3278                 return sock_ioctl(file, cmd, arg);
3279 
3280         case SIOCGIFFLAGS:
3281         case SIOCSIFFLAGS:
3282         case SIOCGIFMETRIC:
3283         case SIOCSIFMETRIC:
3284         case SIOCGIFMTU:
3285         case SIOCSIFMTU:
3286         case SIOCGIFMEM:
3287         case SIOCSIFMEM:
3288         case SIOCGIFHWADDR:
3289         case SIOCSIFHWADDR:
3290         case SIOCADDMULTI:
3291         case SIOCDELMULTI:
3292         case SIOCGIFINDEX:
3293         case SIOCGIFADDR:
3294         case SIOCSIFADDR:
3295         case SIOCSIFHWBROADCAST:
3296         case SIOCDIFADDR:
3297         case SIOCGIFBRDADDR:
3298         case SIOCSIFBRDADDR:
3299         case SIOCGIFDSTADDR:
3300         case SIOCSIFDSTADDR:
3301         case SIOCGIFNETMASK:
3302         case SIOCSIFNETMASK:
3303         case SIOCSIFPFLAGS:
3304         case SIOCGIFPFLAGS:
3305         case SIOCGIFTXQLEN:
3306         case SIOCSIFTXQLEN:
3307         case SIOCBRADDIF:
3308         case SIOCBRDELIF:
3309         case SIOCSIFNAME:
3310         case SIOCGMIIPHY:
3311         case SIOCGMIIREG:
3312         case SIOCSMIIREG:
3313                 return dev_ifsioc(net, sock, cmd, argp);
3314 
3315         case SIOCSARP:
3316         case SIOCGARP:
3317         case SIOCDARP:
3318         case SIOCOUTQNSD:
3319         case SIOCATMARK:
3320                 return sock_do_ioctl(net, sock, cmd, arg);
3321         }
3322 
3323         return -ENOIOCTLCMD;
3324 }
3325 
3326 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3327                               unsigned long arg)
3328 {
3329         struct socket *sock = file->private_data;
3330         int ret = -ENOIOCTLCMD;
3331         struct sock *sk;
3332         struct net *net;
3333 
3334         sk = sock->sk;
3335         net = sock_net(sk);
3336 
3337         if (sock->ops->compat_ioctl)
3338                 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3339 
3340         if (ret == -ENOIOCTLCMD &&
3341             (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3342                 ret = compat_wext_handle_ioctl(net, cmd, arg);
3343 
3344         if (ret == -ENOIOCTLCMD)
3345                 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3346 
3347         return ret;
3348 }
3349 #endif
3350 
3351 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3352 {
3353         return sock->ops->bind(sock, addr, addrlen);
3354 }
3355 EXPORT_SYMBOL(kernel_bind);
3356 
3357 int kernel_listen(struct socket *sock, int backlog)
3358 {
3359         return sock->ops->listen(sock, backlog);
3360 }
3361 EXPORT_SYMBOL(kernel_listen);
3362 
3363 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3364 {
3365         struct sock *sk = sock->sk;
3366         int err;
3367 
3368         err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3369                                newsock);
3370         if (err < 0)
3371                 goto done;
3372 
3373         err = sock->ops->accept(sock, *newsock, flags);
3374         if (err < 0) {
3375                 sock_release(*newsock);
3376                 *newsock = NULL;
3377                 goto done;
3378         }
3379 
3380         (*newsock)->ops = sock->ops;
3381         __module_get((*newsock)->ops->owner);
3382 
3383 done:
3384         return err;
3385 }
3386 EXPORT_SYMBOL(kernel_accept);
3387 
3388 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3389                    int flags)
3390 {
3391         return sock->ops->connect(sock, addr, addrlen, flags);
3392 }
3393 EXPORT_SYMBOL(kernel_connect);
3394 
3395 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3396                          int *addrlen)
3397 {
3398         return sock->ops->getname(sock, addr, addrlen, 0);
3399 }
3400 EXPORT_SYMBOL(kernel_getsockname);
3401 
3402 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3403                          int *addrlen)
3404 {
3405         return sock->ops->getname(sock, addr, addrlen, 1);
3406 }
3407 EXPORT_SYMBOL(kernel_getpeername);
3408 
3409 int kernel_getsockopt(struct socket *sock, int level, int optname,
3410                         char *optval, int *optlen)
3411 {
3412         mm_segment_t oldfs = get_fs();
3413         char __user *uoptval;
3414         int __user *uoptlen;
3415         int err;
3416 
3417         uoptval = (char __user __force *) optval;
3418         uoptlen = (int __user __force *) optlen;
3419 
3420         set_fs(KERNEL_DS);
3421         if (level == SOL_SOCKET)
3422                 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3423         else
3424                 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3425                                             uoptlen);
3426         set_fs(oldfs);
3427         return err;
3428 }
3429 EXPORT_SYMBOL(kernel_getsockopt);
3430 
3431 int kernel_setsockopt(struct socket *sock, int level, int optname,
3432                         char *optval, unsigned int optlen)
3433 {
3434         mm_segment_t oldfs = get_fs();
3435         char __user *uoptval;
3436         int err;
3437 
3438         uoptval = (char __user __force *) optval;
3439 
3440         set_fs(KERNEL_DS);
3441         if (level == SOL_SOCKET)
3442                 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3443         else
3444                 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3445                                             optlen);
3446         set_fs(oldfs);
3447         return err;
3448 }
3449 EXPORT_SYMBOL(kernel_setsockopt);
3450 
3451 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3452                     size_t size, int flags)
3453 {
3454         if (sock->ops->sendpage)
3455                 return sock->ops->sendpage(sock, page, offset, size, flags);
3456 
3457         return sock_no_sendpage(sock, page, offset, size, flags);
3458 }
3459 EXPORT_SYMBOL(kernel_sendpage);
3460 
3461 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3462 {
3463         mm_segment_t oldfs = get_fs();
3464         int err;
3465 
3466         set_fs(KERNEL_DS);
3467         err = sock->ops->ioctl(sock, cmd, arg);
3468         set_fs(oldfs);
3469 
3470         return err;
3471 }
3472 EXPORT_SYMBOL(kernel_sock_ioctl);
3473 
3474 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3475 {
3476         return sock->ops->shutdown(sock, how);
3477 }
3478 EXPORT_SYMBOL(kernel_sock_shutdown);
3479 

~ [ 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