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

TOMOYO Linux Cross Reference
Linux/net/socket.c

Version: ~ [ linux-6.1-rc7 ] ~ [ linux-6.0.10 ] ~ [ linux-5.19.17 ] ~ [ linux-5.18.19 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.80 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.156 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.225 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.267 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.300 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.334 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.302 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.9 ] ~ [ 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 
 93 #include <linux/uaccess.h>
 94 #include <asm/unistd.h>
 95 
 96 #include <net/compat.h>
 97 #include <net/wext.h>
 98 #include <net/cls_cgroup.h>
 99 
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
102 
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
114 #endif
115 
116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119 
120 static int sock_close(struct inode *inode, struct file *file);
121 static unsigned int sock_poll(struct file *file,
122                               struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124 #ifdef CONFIG_COMPAT
125 static long compat_sock_ioctl(struct file *file,
126                               unsigned int cmd, unsigned long arg);
127 #endif
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130                              int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132                                 struct pipe_inode_info *pipe, size_t len,
133                                 unsigned int flags);
134 
135 /*
136  *      Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137  *      in the operation structures but are done directly via the socketcall() multiplexor.
138  */
139 
140 static const struct file_operations socket_file_ops = {
141         .owner =        THIS_MODULE,
142         .llseek =       no_llseek,
143         .read_iter =    sock_read_iter,
144         .write_iter =   sock_write_iter,
145         .poll =         sock_poll,
146         .unlocked_ioctl = sock_ioctl,
147 #ifdef CONFIG_COMPAT
148         .compat_ioctl = compat_sock_ioctl,
149 #endif
150         .mmap =         sock_mmap,
151         .release =      sock_close,
152         .fasync =       sock_fasync,
153         .sendpage =     sock_sendpage,
154         .splice_write = generic_splice_sendpage,
155         .splice_read =  sock_splice_read,
156 };
157 
158 /*
159  *      The protocol list. Each protocol is registered in here.
160  */
161 
162 static DEFINE_SPINLOCK(net_family_lock);
163 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164 
165 /*
166  *      Statistics counters of the socket lists
167  */
168 
169 static DEFINE_PER_CPU(int, sockets_in_use);
170 
171 /*
172  * Support routines.
173  * Move socket addresses back and forth across the kernel/user
174  * divide and look after the messy bits.
175  */
176 
177 /**
178  *      move_addr_to_kernel     -       copy a socket address into kernel space
179  *      @uaddr: Address in user space
180  *      @kaddr: Address in kernel space
181  *      @ulen: Length in user space
182  *
183  *      The address is copied into kernel space. If the provided address is
184  *      too long an error code of -EINVAL is returned. If the copy gives
185  *      invalid addresses -EFAULT is returned. On a success 0 is returned.
186  */
187 
188 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
189 {
190         if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
191                 return -EINVAL;
192         if (ulen == 0)
193                 return 0;
194         if (copy_from_user(kaddr, uaddr, ulen))
195                 return -EFAULT;
196         return audit_sockaddr(ulen, kaddr);
197 }
198 
199 /**
200  *      move_addr_to_user       -       copy an address to user space
201  *      @kaddr: kernel space address
202  *      @klen: length of address in kernel
203  *      @uaddr: user space address
204  *      @ulen: pointer to user length field
205  *
206  *      The value pointed to by ulen on entry is the buffer length available.
207  *      This is overwritten with the buffer space used. -EINVAL is returned
208  *      if an overlong buffer is specified or a negative buffer size. -EFAULT
209  *      is returned if either the buffer or the length field are not
210  *      accessible.
211  *      After copying the data up to the limit the user specifies, the true
212  *      length of the data is written over the length limit the user
213  *      specified. Zero is returned for a success.
214  */
215 
216 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
217                              void __user *uaddr, int __user *ulen)
218 {
219         int err;
220         int len;
221 
222         BUG_ON(klen > sizeof(struct sockaddr_storage));
223         err = get_user(len, ulen);
224         if (err)
225                 return err;
226         if (len > klen)
227                 len = klen;
228         if (len < 0)
229                 return -EINVAL;
230         if (len) {
231                 if (audit_sockaddr(klen, kaddr))
232                         return -ENOMEM;
233                 if (copy_to_user(uaddr, kaddr, len))
234                         return -EFAULT;
235         }
236         /*
237          *      "fromlen shall refer to the value before truncation.."
238          *                      1003.1g
239          */
240         return __put_user(klen, ulen);
241 }
242 
243 static struct kmem_cache *sock_inode_cachep __read_mostly;
244 
245 static struct inode *sock_alloc_inode(struct super_block *sb)
246 {
247         struct socket_alloc *ei;
248         struct socket_wq *wq;
249 
250         ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
251         if (!ei)
252                 return NULL;
253         wq = kmalloc(sizeof(*wq), GFP_KERNEL);
254         if (!wq) {
255                 kmem_cache_free(sock_inode_cachep, ei);
256                 return NULL;
257         }
258         init_waitqueue_head(&wq->wait);
259         wq->fasync_list = NULL;
260         wq->flags = 0;
261         RCU_INIT_POINTER(ei->socket.wq, wq);
262 
263         ei->socket.state = SS_UNCONNECTED;
264         ei->socket.flags = 0;
265         ei->socket.ops = NULL;
266         ei->socket.sk = NULL;
267         ei->socket.file = NULL;
268 
269         return &ei->vfs_inode;
270 }
271 
272 static void sock_destroy_inode(struct inode *inode)
273 {
274         struct socket_alloc *ei;
275         struct socket_wq *wq;
276 
277         ei = container_of(inode, struct socket_alloc, vfs_inode);
278         wq = rcu_dereference_protected(ei->socket.wq, 1);
279         kfree_rcu(wq, rcu);
280         kmem_cache_free(sock_inode_cachep, ei);
281 }
282 
283 static void init_once(void *foo)
284 {
285         struct socket_alloc *ei = (struct socket_alloc *)foo;
286 
287         inode_init_once(&ei->vfs_inode);
288 }
289 
290 static void init_inodecache(void)
291 {
292         sock_inode_cachep = kmem_cache_create("sock_inode_cache",
293                                               sizeof(struct socket_alloc),
294                                               0,
295                                               (SLAB_HWCACHE_ALIGN |
296                                                SLAB_RECLAIM_ACCOUNT |
297                                                SLAB_MEM_SPREAD | SLAB_ACCOUNT),
298                                               init_once);
299         BUG_ON(sock_inode_cachep == NULL);
300 }
301 
302 static const struct super_operations sockfs_ops = {
303         .alloc_inode    = sock_alloc_inode,
304         .destroy_inode  = sock_destroy_inode,
305         .statfs         = simple_statfs,
306 };
307 
308 /*
309  * sockfs_dname() is called from d_path().
310  */
311 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
312 {
313         return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
314                                 d_inode(dentry)->i_ino);
315 }
316 
317 static const struct dentry_operations sockfs_dentry_operations = {
318         .d_dname  = sockfs_dname,
319 };
320 
321 static int sockfs_xattr_get(const struct xattr_handler *handler,
322                             struct dentry *dentry, struct inode *inode,
323                             const char *suffix, void *value, size_t size)
324 {
325         if (value) {
326                 if (dentry->d_name.len + 1 > size)
327                         return -ERANGE;
328                 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
329         }
330         return dentry->d_name.len + 1;
331 }
332 
333 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
334 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
335 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
336 
337 static const struct xattr_handler sockfs_xattr_handler = {
338         .name = XATTR_NAME_SOCKPROTONAME,
339         .get = sockfs_xattr_get,
340 };
341 
342 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
343                                      struct dentry *dentry, struct inode *inode,
344                                      const char *suffix, const void *value,
345                                      size_t size, int flags)
346 {
347         /* Handled by LSM. */
348         return -EAGAIN;
349 }
350 
351 static const struct xattr_handler sockfs_security_xattr_handler = {
352         .prefix = XATTR_SECURITY_PREFIX,
353         .set = sockfs_security_xattr_set,
354 };
355 
356 static const struct xattr_handler *sockfs_xattr_handlers[] = {
357         &sockfs_xattr_handler,
358         &sockfs_security_xattr_handler,
359         NULL
360 };
361 
362 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
363                          int flags, const char *dev_name, void *data)
364 {
365         return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
366                                   sockfs_xattr_handlers,
367                                   &sockfs_dentry_operations, SOCKFS_MAGIC);
368 }
369 
370 static struct vfsmount *sock_mnt __read_mostly;
371 
372 static struct file_system_type sock_fs_type = {
373         .name =         "sockfs",
374         .mount =        sockfs_mount,
375         .kill_sb =      kill_anon_super,
376 };
377 
378 /*
379  *      Obtains the first available file descriptor and sets it up for use.
380  *
381  *      These functions create file structures and maps them to fd space
382  *      of the current process. On success it returns file descriptor
383  *      and file struct implicitly stored in sock->file.
384  *      Note that another thread may close file descriptor before we return
385  *      from this function. We use the fact that now we do not refer
386  *      to socket after mapping. If one day we will need it, this
387  *      function will increment ref. count on file by 1.
388  *
389  *      In any case returned fd MAY BE not valid!
390  *      This race condition is unavoidable
391  *      with shared fd spaces, we cannot solve it inside kernel,
392  *      but we take care of internal coherence yet.
393  */
394 
395 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
396 {
397         struct qstr name = { .name = "" };
398         struct path path;
399         struct file *file;
400 
401         if (dname) {
402                 name.name = dname;
403                 name.len = strlen(name.name);
404         } else if (sock->sk) {
405                 name.name = sock->sk->sk_prot_creator->name;
406                 name.len = strlen(name.name);
407         }
408         path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
409         if (unlikely(!path.dentry))
410                 return ERR_PTR(-ENOMEM);
411         path.mnt = mntget(sock_mnt);
412 
413         d_instantiate(path.dentry, SOCK_INODE(sock));
414 
415         file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
416                   &socket_file_ops);
417         if (IS_ERR(file)) {
418                 /* drop dentry, keep inode */
419                 ihold(d_inode(path.dentry));
420                 path_put(&path);
421                 return file;
422         }
423 
424         sock->file = file;
425         file->f_flags = O_RDWR | (flags & O_NONBLOCK);
426         file->private_data = sock;
427         return file;
428 }
429 EXPORT_SYMBOL(sock_alloc_file);
430 
431 static int sock_map_fd(struct socket *sock, int flags)
432 {
433         struct file *newfile;
434         int fd = get_unused_fd_flags(flags);
435         if (unlikely(fd < 0))
436                 return fd;
437 
438         newfile = sock_alloc_file(sock, flags, NULL);
439         if (likely(!IS_ERR(newfile))) {
440                 fd_install(fd, newfile);
441                 return fd;
442         }
443 
444         put_unused_fd(fd);
445         return PTR_ERR(newfile);
446 }
447 
448 struct socket *sock_from_file(struct file *file, int *err)
449 {
450         if (file->f_op == &socket_file_ops)
451                 return file->private_data;      /* set in sock_map_fd */
452 
453         *err = -ENOTSOCK;
454         return NULL;
455 }
456 EXPORT_SYMBOL(sock_from_file);
457 
458 /**
459  *      sockfd_lookup - Go from a file number to its socket slot
460  *      @fd: file handle
461  *      @err: pointer to an error code return
462  *
463  *      The file handle passed in is locked and the socket it is bound
464  *      too is returned. If an error occurs the err pointer is overwritten
465  *      with a negative errno code and NULL is returned. The function checks
466  *      for both invalid handles and passing a handle which is not a socket.
467  *
468  *      On a success the socket object pointer is returned.
469  */
470 
471 struct socket *sockfd_lookup(int fd, int *err)
472 {
473         struct file *file;
474         struct socket *sock;
475 
476         file = fget(fd);
477         if (!file) {
478                 *err = -EBADF;
479                 return NULL;
480         }
481 
482         sock = sock_from_file(file, err);
483         if (!sock)
484                 fput(file);
485         return sock;
486 }
487 EXPORT_SYMBOL(sockfd_lookup);
488 
489 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
490 {
491         struct fd f = fdget(fd);
492         struct socket *sock;
493 
494         *err = -EBADF;
495         if (f.file) {
496                 sock = sock_from_file(f.file, err);
497                 if (likely(sock)) {
498                         *fput_needed = f.flags;
499                         return sock;
500                 }
501                 fdput(f);
502         }
503         return NULL;
504 }
505 
506 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
507                                 size_t size)
508 {
509         ssize_t len;
510         ssize_t used = 0;
511 
512         len = security_inode_listsecurity(d_inode(dentry), buffer, size);
513         if (len < 0)
514                 return len;
515         used += len;
516         if (buffer) {
517                 if (size < used)
518                         return -ERANGE;
519                 buffer += len;
520         }
521 
522         len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
523         used += len;
524         if (buffer) {
525                 if (size < used)
526                         return -ERANGE;
527                 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
528                 buffer += len;
529         }
530 
531         return used;
532 }
533 
534 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
535 {
536         int err = simple_setattr(dentry, iattr);
537 
538         if (!err && (iattr->ia_valid & ATTR_UID)) {
539                 struct socket *sock = SOCKET_I(d_inode(dentry));
540 
541                 sock->sk->sk_uid = iattr->ia_uid;
542         }
543 
544         return err;
545 }
546 
547 static const struct inode_operations sockfs_inode_ops = {
548         .listxattr = sockfs_listxattr,
549         .setattr = sockfs_setattr,
550 };
551 
552 /**
553  *      sock_alloc      -       allocate a socket
554  *
555  *      Allocate a new inode and socket object. The two are bound together
556  *      and initialised. The socket is then returned. If we are out of inodes
557  *      NULL is returned.
558  */
559 
560 struct socket *sock_alloc(void)
561 {
562         struct inode *inode;
563         struct socket *sock;
564 
565         inode = new_inode_pseudo(sock_mnt->mnt_sb);
566         if (!inode)
567                 return NULL;
568 
569         sock = SOCKET_I(inode);
570 
571         kmemcheck_annotate_bitfield(sock, type);
572         inode->i_ino = get_next_ino();
573         inode->i_mode = S_IFSOCK | S_IRWXUGO;
574         inode->i_uid = current_fsuid();
575         inode->i_gid = current_fsgid();
576         inode->i_op = &sockfs_inode_ops;
577 
578         this_cpu_add(sockets_in_use, 1);
579         return sock;
580 }
581 EXPORT_SYMBOL(sock_alloc);
582 
583 /**
584  *      sock_release    -       close a socket
585  *      @sock: socket to close
586  *
587  *      The socket is released from the protocol stack if it has a release
588  *      callback, and the inode is then released if the socket is bound to
589  *      an inode not a file.
590  */
591 
592 void sock_release(struct socket *sock)
593 {
594         if (sock->ops) {
595                 struct module *owner = sock->ops->owner;
596 
597                 sock->ops->release(sock);
598                 sock->ops = NULL;
599                 module_put(owner);
600         }
601 
602         if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
603                 pr_err("%s: fasync list not empty!\n", __func__);
604 
605         this_cpu_sub(sockets_in_use, 1);
606         if (!sock->file) {
607                 iput(SOCK_INODE(sock));
608                 return;
609         }
610         sock->file = NULL;
611 }
612 EXPORT_SYMBOL(sock_release);
613 
614 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
615 {
616         u8 flags = *tx_flags;
617 
618         if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
619                 flags |= SKBTX_HW_TSTAMP;
620 
621         if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
622                 flags |= SKBTX_SW_TSTAMP;
623 
624         if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
625                 flags |= SKBTX_SCHED_TSTAMP;
626 
627         *tx_flags = flags;
628 }
629 EXPORT_SYMBOL(__sock_tx_timestamp);
630 
631 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
632 {
633         int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
634         BUG_ON(ret == -EIOCBQUEUED);
635         return ret;
636 }
637 
638 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
639 {
640         int err = security_socket_sendmsg(sock, msg,
641                                           msg_data_left(msg));
642 
643         return err ?: sock_sendmsg_nosec(sock, msg);
644 }
645 EXPORT_SYMBOL(sock_sendmsg);
646 
647 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
648                    struct kvec *vec, size_t num, size_t size)
649 {
650         iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
651         return sock_sendmsg(sock, msg);
652 }
653 EXPORT_SYMBOL(kernel_sendmsg);
654 
655 static bool skb_is_err_queue(const struct sk_buff *skb)
656 {
657         /* pkt_type of skbs enqueued on the error queue are set to
658          * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
659          * in recvmsg, since skbs received on a local socket will never
660          * have a pkt_type of PACKET_OUTGOING.
661          */
662         return skb->pkt_type == PACKET_OUTGOING;
663 }
664 
665 /*
666  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
667  */
668 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
669         struct sk_buff *skb)
670 {
671         int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
672         struct scm_timestamping tss;
673         int empty = 1;
674         struct skb_shared_hwtstamps *shhwtstamps =
675                 skb_hwtstamps(skb);
676 
677         /* Race occurred between timestamp enabling and packet
678            receiving.  Fill in the current time for now. */
679         if (need_software_tstamp && skb->tstamp == 0)
680                 __net_timestamp(skb);
681 
682         if (need_software_tstamp) {
683                 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
684                         struct timeval tv;
685                         skb_get_timestamp(skb, &tv);
686                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
687                                  sizeof(tv), &tv);
688                 } else {
689                         struct timespec ts;
690                         skb_get_timestampns(skb, &ts);
691                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
692                                  sizeof(ts), &ts);
693                 }
694         }
695 
696         memset(&tss, 0, sizeof(tss));
697         if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
698             ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
699                 empty = 0;
700         if (shhwtstamps &&
701             (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
702             ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
703                 empty = 0;
704         if (!empty) {
705                 put_cmsg(msg, SOL_SOCKET,
706                          SCM_TIMESTAMPING, sizeof(tss), &tss);
707 
708                 if (skb_is_err_queue(skb) && skb->len &&
709                     SKB_EXT_ERR(skb)->opt_stats)
710                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
711                                  skb->len, skb->data);
712         }
713 }
714 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
715 
716 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
717         struct sk_buff *skb)
718 {
719         int ack;
720 
721         if (!sock_flag(sk, SOCK_WIFI_STATUS))
722                 return;
723         if (!skb->wifi_acked_valid)
724                 return;
725 
726         ack = skb->wifi_acked;
727 
728         put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
729 }
730 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
731 
732 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
733                                    struct sk_buff *skb)
734 {
735         if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
736                 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
737                         sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
738 }
739 
740 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
741         struct sk_buff *skb)
742 {
743         sock_recv_timestamp(msg, sk, skb);
744         sock_recv_drops(msg, sk, skb);
745 }
746 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
747 
748 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
749                                      int flags)
750 {
751         return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
752 }
753 
754 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
755 {
756         int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
757 
758         return err ?: sock_recvmsg_nosec(sock, msg, flags);
759 }
760 EXPORT_SYMBOL(sock_recvmsg);
761 
762 /**
763  * kernel_recvmsg - Receive a message from a socket (kernel space)
764  * @sock:       The socket to receive the message from
765  * @msg:        Received message
766  * @vec:        Input s/g array for message data
767  * @num:        Size of input s/g array
768  * @size:       Number of bytes to read
769  * @flags:      Message flags (MSG_DONTWAIT, etc...)
770  *
771  * On return the msg structure contains the scatter/gather array passed in the
772  * vec argument. The array is modified so that it consists of the unfilled
773  * portion of the original array.
774  *
775  * The returned value is the total number of bytes received, or an error.
776  */
777 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
778                    struct kvec *vec, size_t num, size_t size, int flags)
779 {
780         mm_segment_t oldfs = get_fs();
781         int result;
782 
783         iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
784         set_fs(KERNEL_DS);
785         result = sock_recvmsg(sock, msg, flags);
786         set_fs(oldfs);
787         return result;
788 }
789 EXPORT_SYMBOL(kernel_recvmsg);
790 
791 static ssize_t sock_sendpage(struct file *file, struct page *page,
792                              int offset, size_t size, loff_t *ppos, int more)
793 {
794         struct socket *sock;
795         int flags;
796 
797         sock = file->private_data;
798 
799         flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
800         /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
801         flags |= more;
802 
803         return kernel_sendpage(sock, page, offset, size, flags);
804 }
805 
806 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
807                                 struct pipe_inode_info *pipe, size_t len,
808                                 unsigned int flags)
809 {
810         struct socket *sock = file->private_data;
811 
812         if (unlikely(!sock->ops->splice_read))
813                 return -EINVAL;
814 
815         return sock->ops->splice_read(sock, ppos, pipe, len, flags);
816 }
817 
818 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
819 {
820         struct file *file = iocb->ki_filp;
821         struct socket *sock = file->private_data;
822         struct msghdr msg = {.msg_iter = *to,
823                              .msg_iocb = iocb};
824         ssize_t res;
825 
826         if (file->f_flags & O_NONBLOCK)
827                 msg.msg_flags = MSG_DONTWAIT;
828 
829         if (iocb->ki_pos != 0)
830                 return -ESPIPE;
831 
832         if (!iov_iter_count(to))        /* Match SYS5 behaviour */
833                 return 0;
834 
835         res = sock_recvmsg(sock, &msg, msg.msg_flags);
836         *to = msg.msg_iter;
837         return res;
838 }
839 
840 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
841 {
842         struct file *file = iocb->ki_filp;
843         struct socket *sock = file->private_data;
844         struct msghdr msg = {.msg_iter = *from,
845                              .msg_iocb = iocb};
846         ssize_t res;
847 
848         if (iocb->ki_pos != 0)
849                 return -ESPIPE;
850 
851         if (file->f_flags & O_NONBLOCK)
852                 msg.msg_flags = MSG_DONTWAIT;
853 
854         if (sock->type == SOCK_SEQPACKET)
855                 msg.msg_flags |= MSG_EOR;
856 
857         res = sock_sendmsg(sock, &msg);
858         *from = msg.msg_iter;
859         return res;
860 }
861 
862 /*
863  * Atomic setting of ioctl hooks to avoid race
864  * with module unload.
865  */
866 
867 static DEFINE_MUTEX(br_ioctl_mutex);
868 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
869 
870 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
871 {
872         mutex_lock(&br_ioctl_mutex);
873         br_ioctl_hook = hook;
874         mutex_unlock(&br_ioctl_mutex);
875 }
876 EXPORT_SYMBOL(brioctl_set);
877 
878 static DEFINE_MUTEX(vlan_ioctl_mutex);
879 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
880 
881 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
882 {
883         mutex_lock(&vlan_ioctl_mutex);
884         vlan_ioctl_hook = hook;
885         mutex_unlock(&vlan_ioctl_mutex);
886 }
887 EXPORT_SYMBOL(vlan_ioctl_set);
888 
889 static DEFINE_MUTEX(dlci_ioctl_mutex);
890 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
891 
892 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
893 {
894         mutex_lock(&dlci_ioctl_mutex);
895         dlci_ioctl_hook = hook;
896         mutex_unlock(&dlci_ioctl_mutex);
897 }
898 EXPORT_SYMBOL(dlci_ioctl_set);
899 
900 static long sock_do_ioctl(struct net *net, struct socket *sock,
901                                  unsigned int cmd, unsigned long arg)
902 {
903         int err;
904         void __user *argp = (void __user *)arg;
905 
906         err = sock->ops->ioctl(sock, cmd, arg);
907 
908         /*
909          * If this ioctl is unknown try to hand it down
910          * to the NIC driver.
911          */
912         if (err == -ENOIOCTLCMD)
913                 err = dev_ioctl(net, cmd, argp);
914 
915         return err;
916 }
917 
918 /*
919  *      With an ioctl, arg may well be a user mode pointer, but we don't know
920  *      what to do with it - that's up to the protocol still.
921  */
922 
923 static struct ns_common *get_net_ns(struct ns_common *ns)
924 {
925         return &get_net(container_of(ns, struct net, ns))->ns;
926 }
927 
928 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
929 {
930         struct socket *sock;
931         struct sock *sk;
932         void __user *argp = (void __user *)arg;
933         int pid, err;
934         struct net *net;
935 
936         sock = file->private_data;
937         sk = sock->sk;
938         net = sock_net(sk);
939         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
940                 err = dev_ioctl(net, cmd, argp);
941         } else
942 #ifdef CONFIG_WEXT_CORE
943         if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
944                 err = dev_ioctl(net, cmd, argp);
945         } else
946 #endif
947                 switch (cmd) {
948                 case FIOSETOWN:
949                 case SIOCSPGRP:
950                         err = -EFAULT;
951                         if (get_user(pid, (int __user *)argp))
952                                 break;
953                         f_setown(sock->file, pid, 1);
954                         err = 0;
955                         break;
956                 case FIOGETOWN:
957                 case SIOCGPGRP:
958                         err = put_user(f_getown(sock->file),
959                                        (int __user *)argp);
960                         break;
961                 case SIOCGIFBR:
962                 case SIOCSIFBR:
963                 case SIOCBRADDBR:
964                 case SIOCBRDELBR:
965                         err = -ENOPKG;
966                         if (!br_ioctl_hook)
967                                 request_module("bridge");
968 
969                         mutex_lock(&br_ioctl_mutex);
970                         if (br_ioctl_hook)
971                                 err = br_ioctl_hook(net, cmd, argp);
972                         mutex_unlock(&br_ioctl_mutex);
973                         break;
974                 case SIOCGIFVLAN:
975                 case SIOCSIFVLAN:
976                         err = -ENOPKG;
977                         if (!vlan_ioctl_hook)
978                                 request_module("8021q");
979 
980                         mutex_lock(&vlan_ioctl_mutex);
981                         if (vlan_ioctl_hook)
982                                 err = vlan_ioctl_hook(net, argp);
983                         mutex_unlock(&vlan_ioctl_mutex);
984                         break;
985                 case SIOCADDDLCI:
986                 case SIOCDELDLCI:
987                         err = -ENOPKG;
988                         if (!dlci_ioctl_hook)
989                                 request_module("dlci");
990 
991                         mutex_lock(&dlci_ioctl_mutex);
992                         if (dlci_ioctl_hook)
993                                 err = dlci_ioctl_hook(cmd, argp);
994                         mutex_unlock(&dlci_ioctl_mutex);
995                         break;
996                 case SIOCGSKNS:
997                         err = -EPERM;
998                         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
999                                 break;
1000 
1001                         err = open_related_ns(&net->ns, get_net_ns);
1002                         break;
1003                 default:
1004                         err = sock_do_ioctl(net, sock, cmd, arg);
1005                         break;
1006                 }
1007         return err;
1008 }
1009 
1010 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1011 {
1012         int err;
1013         struct socket *sock = NULL;
1014 
1015         err = security_socket_create(family, type, protocol, 1);
1016         if (err)
1017                 goto out;
1018 
1019         sock = sock_alloc();
1020         if (!sock) {
1021                 err = -ENOMEM;
1022                 goto out;
1023         }
1024 
1025         sock->type = type;
1026         err = security_socket_post_create(sock, family, type, protocol, 1);
1027         if (err)
1028                 goto out_release;
1029 
1030 out:
1031         *res = sock;
1032         return err;
1033 out_release:
1034         sock_release(sock);
1035         sock = NULL;
1036         goto out;
1037 }
1038 EXPORT_SYMBOL(sock_create_lite);
1039 
1040 /* No kernel lock held - perfect */
1041 static unsigned int sock_poll(struct file *file, poll_table *wait)
1042 {
1043         unsigned int busy_flag = 0;
1044         struct socket *sock;
1045 
1046         /*
1047          *      We can't return errors to poll, so it's either yes or no.
1048          */
1049         sock = file->private_data;
1050 
1051         if (sk_can_busy_loop(sock->sk)) {
1052                 /* this socket can poll_ll so tell the system call */
1053                 busy_flag = POLL_BUSY_LOOP;
1054 
1055                 /* once, only if requested by syscall */
1056                 if (wait && (wait->_key & POLL_BUSY_LOOP))
1057                         sk_busy_loop(sock->sk, 1);
1058         }
1059 
1060         return busy_flag | sock->ops->poll(file, sock, wait);
1061 }
1062 
1063 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1064 {
1065         struct socket *sock = file->private_data;
1066 
1067         return sock->ops->mmap(file, sock, vma);
1068 }
1069 
1070 static int sock_close(struct inode *inode, struct file *filp)
1071 {
1072         sock_release(SOCKET_I(inode));
1073         return 0;
1074 }
1075 
1076 /*
1077  *      Update the socket async list
1078  *
1079  *      Fasync_list locking strategy.
1080  *
1081  *      1. fasync_list is modified only under process context socket lock
1082  *         i.e. under semaphore.
1083  *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1084  *         or under socket lock
1085  */
1086 
1087 static int sock_fasync(int fd, struct file *filp, int on)
1088 {
1089         struct socket *sock = filp->private_data;
1090         struct sock *sk = sock->sk;
1091         struct socket_wq *wq;
1092 
1093         if (sk == NULL)
1094                 return -EINVAL;
1095 
1096         lock_sock(sk);
1097         wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1098         fasync_helper(fd, filp, on, &wq->fasync_list);
1099 
1100         if (!wq->fasync_list)
1101                 sock_reset_flag(sk, SOCK_FASYNC);
1102         else
1103                 sock_set_flag(sk, SOCK_FASYNC);
1104 
1105         release_sock(sk);
1106         return 0;
1107 }
1108 
1109 /* This function may be called only under rcu_lock */
1110 
1111 int sock_wake_async(struct socket_wq *wq, int how, int band)
1112 {
1113         if (!wq || !wq->fasync_list)
1114                 return -1;
1115 
1116         switch (how) {
1117         case SOCK_WAKE_WAITD:
1118                 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1119                         break;
1120                 goto call_kill;
1121         case SOCK_WAKE_SPACE:
1122                 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1123                         break;
1124                 /* fall through */
1125         case SOCK_WAKE_IO:
1126 call_kill:
1127                 kill_fasync(&wq->fasync_list, SIGIO, band);
1128                 break;
1129         case SOCK_WAKE_URG:
1130                 kill_fasync(&wq->fasync_list, SIGURG, band);
1131         }
1132 
1133         return 0;
1134 }
1135 EXPORT_SYMBOL(sock_wake_async);
1136 
1137 int __sock_create(struct net *net, int family, int type, int protocol,
1138                          struct socket **res, int kern)
1139 {
1140         int err;
1141         struct socket *sock;
1142         const struct net_proto_family *pf;
1143 
1144         /*
1145          *      Check protocol is in range
1146          */
1147         if (family < 0 || family >= NPROTO)
1148                 return -EAFNOSUPPORT;
1149         if (type < 0 || type >= SOCK_MAX)
1150                 return -EINVAL;
1151 
1152         /* Compatibility.
1153 
1154            This uglymoron is moved from INET layer to here to avoid
1155            deadlock in module load.
1156          */
1157         if (family == PF_INET && type == SOCK_PACKET) {
1158                 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1159                              current->comm);
1160                 family = PF_PACKET;
1161         }
1162 
1163         err = security_socket_create(family, type, protocol, kern);
1164         if (err)
1165                 return err;
1166 
1167         /*
1168          *      Allocate the socket and allow the family to set things up. if
1169          *      the protocol is 0, the family is instructed to select an appropriate
1170          *      default.
1171          */
1172         sock = sock_alloc();
1173         if (!sock) {
1174                 net_warn_ratelimited("socket: no more sockets\n");
1175                 return -ENFILE; /* Not exactly a match, but its the
1176                                    closest posix thing */
1177         }
1178 
1179         sock->type = type;
1180 
1181 #ifdef CONFIG_MODULES
1182         /* Attempt to load a protocol module if the find failed.
1183          *
1184          * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1185          * requested real, full-featured networking support upon configuration.
1186          * Otherwise module support will break!
1187          */
1188         if (rcu_access_pointer(net_families[family]) == NULL)
1189                 request_module("net-pf-%d", family);
1190 #endif
1191 
1192         rcu_read_lock();
1193         pf = rcu_dereference(net_families[family]);
1194         err = -EAFNOSUPPORT;
1195         if (!pf)
1196                 goto out_release;
1197 
1198         /*
1199          * We will call the ->create function, that possibly is in a loadable
1200          * module, so we have to bump that loadable module refcnt first.
1201          */
1202         if (!try_module_get(pf->owner))
1203                 goto out_release;
1204 
1205         /* Now protected by module ref count */
1206         rcu_read_unlock();
1207 
1208         err = pf->create(net, sock, protocol, kern);
1209         if (err < 0)
1210                 goto out_module_put;
1211 
1212         /*
1213          * Now to bump the refcnt of the [loadable] module that owns this
1214          * socket at sock_release time we decrement its refcnt.
1215          */
1216         if (!try_module_get(sock->ops->owner))
1217                 goto out_module_busy;
1218 
1219         /*
1220          * Now that we're done with the ->create function, the [loadable]
1221          * module can have its refcnt decremented
1222          */
1223         module_put(pf->owner);
1224         err = security_socket_post_create(sock, family, type, protocol, kern);
1225         if (err)
1226                 goto out_sock_release;
1227         *res = sock;
1228 
1229         return 0;
1230 
1231 out_module_busy:
1232         err = -EAFNOSUPPORT;
1233 out_module_put:
1234         sock->ops = NULL;
1235         module_put(pf->owner);
1236 out_sock_release:
1237         sock_release(sock);
1238         return err;
1239 
1240 out_release:
1241         rcu_read_unlock();
1242         goto out_sock_release;
1243 }
1244 EXPORT_SYMBOL(__sock_create);
1245 
1246 int sock_create(int family, int type, int protocol, struct socket **res)
1247 {
1248         return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1249 }
1250 EXPORT_SYMBOL(sock_create);
1251 
1252 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1253 {
1254         return __sock_create(net, family, type, protocol, res, 1);
1255 }
1256 EXPORT_SYMBOL(sock_create_kern);
1257 
1258 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1259 {
1260         int retval;
1261         struct socket *sock;
1262         int flags;
1263 
1264         /* Check the SOCK_* constants for consistency.  */
1265         BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1266         BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1267         BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1268         BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1269 
1270         flags = type & ~SOCK_TYPE_MASK;
1271         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1272                 return -EINVAL;
1273         type &= SOCK_TYPE_MASK;
1274 
1275         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1276                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1277 
1278         retval = sock_create(family, type, protocol, &sock);
1279         if (retval < 0)
1280                 goto out;
1281 
1282         retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1283         if (retval < 0)
1284                 goto out_release;
1285 
1286 out:
1287         /* It may be already another descriptor 8) Not kernel problem. */
1288         return retval;
1289 
1290 out_release:
1291         sock_release(sock);
1292         return retval;
1293 }
1294 
1295 /*
1296  *      Create a pair of connected sockets.
1297  */
1298 
1299 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1300                 int __user *, usockvec)
1301 {
1302         struct socket *sock1, *sock2;
1303         int fd1, fd2, err;
1304         struct file *newfile1, *newfile2;
1305         int flags;
1306 
1307         flags = type & ~SOCK_TYPE_MASK;
1308         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1309                 return -EINVAL;
1310         type &= SOCK_TYPE_MASK;
1311 
1312         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1313                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1314 
1315         /*
1316          * Obtain the first socket and check if the underlying protocol
1317          * supports the socketpair call.
1318          */
1319 
1320         err = sock_create(family, type, protocol, &sock1);
1321         if (err < 0)
1322                 goto out;
1323 
1324         err = sock_create(family, type, protocol, &sock2);
1325         if (err < 0)
1326                 goto out_release_1;
1327 
1328         err = sock1->ops->socketpair(sock1, sock2);
1329         if (err < 0)
1330                 goto out_release_both;
1331 
1332         fd1 = get_unused_fd_flags(flags);
1333         if (unlikely(fd1 < 0)) {
1334                 err = fd1;
1335                 goto out_release_both;
1336         }
1337 
1338         fd2 = get_unused_fd_flags(flags);
1339         if (unlikely(fd2 < 0)) {
1340                 err = fd2;
1341                 goto out_put_unused_1;
1342         }
1343 
1344         newfile1 = sock_alloc_file(sock1, flags, NULL);
1345         if (IS_ERR(newfile1)) {
1346                 err = PTR_ERR(newfile1);
1347                 goto out_put_unused_both;
1348         }
1349 
1350         newfile2 = sock_alloc_file(sock2, flags, NULL);
1351         if (IS_ERR(newfile2)) {
1352                 err = PTR_ERR(newfile2);
1353                 goto out_fput_1;
1354         }
1355 
1356         err = put_user(fd1, &usockvec[0]);
1357         if (err)
1358                 goto out_fput_both;
1359 
1360         err = put_user(fd2, &usockvec[1]);
1361         if (err)
1362                 goto out_fput_both;
1363 
1364         audit_fd_pair(fd1, fd2);
1365 
1366         fd_install(fd1, newfile1);
1367         fd_install(fd2, newfile2);
1368         /* fd1 and fd2 may be already another descriptors.
1369          * Not kernel problem.
1370          */
1371 
1372         return 0;
1373 
1374 out_fput_both:
1375         fput(newfile2);
1376         fput(newfile1);
1377         put_unused_fd(fd2);
1378         put_unused_fd(fd1);
1379         goto out;
1380 
1381 out_fput_1:
1382         fput(newfile1);
1383         put_unused_fd(fd2);
1384         put_unused_fd(fd1);
1385         sock_release(sock2);
1386         goto out;
1387 
1388 out_put_unused_both:
1389         put_unused_fd(fd2);
1390 out_put_unused_1:
1391         put_unused_fd(fd1);
1392 out_release_both:
1393         sock_release(sock2);
1394 out_release_1:
1395         sock_release(sock1);
1396 out:
1397         return err;
1398 }
1399 
1400 /*
1401  *      Bind a name to a socket. Nothing much to do here since it's
1402  *      the protocol's responsibility to handle the local address.
1403  *
1404  *      We move the socket address to kernel space before we call
1405  *      the protocol layer (having also checked the address is ok).
1406  */
1407 
1408 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1409 {
1410         struct socket *sock;
1411         struct sockaddr_storage address;
1412         int err, fput_needed;
1413 
1414         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1415         if (sock) {
1416                 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1417                 if (err >= 0) {
1418                         err = security_socket_bind(sock,
1419                                                    (struct sockaddr *)&address,
1420                                                    addrlen);
1421                         if (!err)
1422                                 err = sock->ops->bind(sock,
1423                                                       (struct sockaddr *)
1424                                                       &address, addrlen);
1425                 }
1426                 fput_light(sock->file, fput_needed);
1427         }
1428         return err;
1429 }
1430 
1431 /*
1432  *      Perform a listen. Basically, we allow the protocol to do anything
1433  *      necessary for a listen, and if that works, we mark the socket as
1434  *      ready for listening.
1435  */
1436 
1437 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1438 {
1439         struct socket *sock;
1440         int err, fput_needed;
1441         int somaxconn;
1442 
1443         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1444         if (sock) {
1445                 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1446                 if ((unsigned int)backlog > somaxconn)
1447                         backlog = somaxconn;
1448 
1449                 err = security_socket_listen(sock, backlog);
1450                 if (!err)
1451                         err = sock->ops->listen(sock, backlog);
1452 
1453                 fput_light(sock->file, fput_needed);
1454         }
1455         return err;
1456 }
1457 
1458 /*
1459  *      For accept, we attempt to create a new socket, set up the link
1460  *      with the client, wake up the client, then return the new
1461  *      connected fd. We collect the address of the connector in kernel
1462  *      space and move it to user at the very end. This is unclean because
1463  *      we open the socket then return an error.
1464  *
1465  *      1003.1g adds the ability to recvmsg() to query connection pending
1466  *      status to recvmsg. We need to add that support in a way thats
1467  *      clean when we restucture accept also.
1468  */
1469 
1470 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1471                 int __user *, upeer_addrlen, int, flags)
1472 {
1473         struct socket *sock, *newsock;
1474         struct file *newfile;
1475         int err, len, newfd, fput_needed;
1476         struct sockaddr_storage address;
1477 
1478         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1479                 return -EINVAL;
1480 
1481         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1482                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1483 
1484         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1485         if (!sock)
1486                 goto out;
1487 
1488         err = -ENFILE;
1489         newsock = sock_alloc();
1490         if (!newsock)
1491                 goto out_put;
1492 
1493         newsock->type = sock->type;
1494         newsock->ops = sock->ops;
1495 
1496         /*
1497          * We don't need try_module_get here, as the listening socket (sock)
1498          * has the protocol module (sock->ops->owner) held.
1499          */
1500         __module_get(newsock->ops->owner);
1501 
1502         newfd = get_unused_fd_flags(flags);
1503         if (unlikely(newfd < 0)) {
1504                 err = newfd;
1505                 sock_release(newsock);
1506                 goto out_put;
1507         }
1508         newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1509         if (IS_ERR(newfile)) {
1510                 err = PTR_ERR(newfile);
1511                 put_unused_fd(newfd);
1512                 sock_release(newsock);
1513                 goto out_put;
1514         }
1515 
1516         err = security_socket_accept(sock, newsock);
1517         if (err)
1518                 goto out_fd;
1519 
1520         err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1521         if (err < 0)
1522                 goto out_fd;
1523 
1524         if (ccs_socket_post_accept_permission(sock, newsock)) {
1525                 err = -EAGAIN; /* Hope less harmful than -EPERM. */
1526                 goto out_fd;
1527         }
1528         if (upeer_sockaddr) {
1529                 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1530                                           &len, 2) < 0) {
1531                         err = -ECONNABORTED;
1532                         goto out_fd;
1533                 }
1534                 err = move_addr_to_user(&address,
1535                                         len, upeer_sockaddr, upeer_addrlen);
1536                 if (err < 0)
1537                         goto out_fd;
1538         }
1539 
1540         /* File flags are not inherited via accept() unlike another OSes. */
1541 
1542         fd_install(newfd, newfile);
1543         err = newfd;
1544 
1545 out_put:
1546         fput_light(sock->file, fput_needed);
1547 out:
1548         return err;
1549 out_fd:
1550         fput(newfile);
1551         put_unused_fd(newfd);
1552         goto out_put;
1553 }
1554 
1555 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1556                 int __user *, upeer_addrlen)
1557 {
1558         return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1559 }
1560 
1561 /*
1562  *      Attempt to connect to a socket with the server address.  The address
1563  *      is in user space so we verify it is OK and move it to kernel space.
1564  *
1565  *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1566  *      break bindings
1567  *
1568  *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1569  *      other SEQPACKET protocols that take time to connect() as it doesn't
1570  *      include the -EINPROGRESS status for such sockets.
1571  */
1572 
1573 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1574                 int, addrlen)
1575 {
1576         struct socket *sock;
1577         struct sockaddr_storage address;
1578         int err, fput_needed;
1579 
1580         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1581         if (!sock)
1582                 goto out;
1583         err = move_addr_to_kernel(uservaddr, addrlen, &address);
1584         if (err < 0)
1585                 goto out_put;
1586 
1587         err =
1588             security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1589         if (err)
1590                 goto out_put;
1591 
1592         err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1593                                  sock->file->f_flags);
1594 out_put:
1595         fput_light(sock->file, fput_needed);
1596 out:
1597         return err;
1598 }
1599 
1600 /*
1601  *      Get the local address ('name') of a socket object. Move the obtained
1602  *      name to user space.
1603  */
1604 
1605 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1606                 int __user *, usockaddr_len)
1607 {
1608         struct socket *sock;
1609         struct sockaddr_storage address;
1610         int len, err, fput_needed;
1611 
1612         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1613         if (!sock)
1614                 goto out;
1615 
1616         err = security_socket_getsockname(sock);
1617         if (err)
1618                 goto out_put;
1619 
1620         err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1621         if (err)
1622                 goto out_put;
1623         err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1624 
1625 out_put:
1626         fput_light(sock->file, fput_needed);
1627 out:
1628         return err;
1629 }
1630 
1631 /*
1632  *      Get the remote address ('name') of a socket object. Move the obtained
1633  *      name to user space.
1634  */
1635 
1636 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1637                 int __user *, usockaddr_len)
1638 {
1639         struct socket *sock;
1640         struct sockaddr_storage address;
1641         int len, err, fput_needed;
1642 
1643         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1644         if (sock != NULL) {
1645                 err = security_socket_getpeername(sock);
1646                 if (err) {
1647                         fput_light(sock->file, fput_needed);
1648                         return err;
1649                 }
1650 
1651                 err =
1652                     sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1653                                        1);
1654                 if (!err)
1655                         err = move_addr_to_user(&address, len, usockaddr,
1656                                                 usockaddr_len);
1657                 fput_light(sock->file, fput_needed);
1658         }
1659         return err;
1660 }
1661 
1662 /*
1663  *      Send a datagram to a given address. We move the address into kernel
1664  *      space and check the user space data area is readable before invoking
1665  *      the protocol.
1666  */
1667 
1668 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1669                 unsigned int, flags, struct sockaddr __user *, addr,
1670                 int, addr_len)
1671 {
1672         struct socket *sock;
1673         struct sockaddr_storage address;
1674         int err;
1675         struct msghdr msg;
1676         struct iovec iov;
1677         int fput_needed;
1678 
1679         err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1680         if (unlikely(err))
1681                 return err;
1682         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1683         if (!sock)
1684                 goto out;
1685 
1686         msg.msg_name = NULL;
1687         msg.msg_control = NULL;
1688         msg.msg_controllen = 0;
1689         msg.msg_namelen = 0;
1690         if (addr) {
1691                 err = move_addr_to_kernel(addr, addr_len, &address);
1692                 if (err < 0)
1693                         goto out_put;
1694                 msg.msg_name = (struct sockaddr *)&address;
1695                 msg.msg_namelen = addr_len;
1696         }
1697         if (sock->file->f_flags & O_NONBLOCK)
1698                 flags |= MSG_DONTWAIT;
1699         msg.msg_flags = flags;
1700         err = sock_sendmsg(sock, &msg);
1701 
1702 out_put:
1703         fput_light(sock->file, fput_needed);
1704 out:
1705         return err;
1706 }
1707 
1708 /*
1709  *      Send a datagram down a socket.
1710  */
1711 
1712 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1713                 unsigned int, flags)
1714 {
1715         return sys_sendto(fd, buff, len, flags, NULL, 0);
1716 }
1717 
1718 /*
1719  *      Receive a frame from the socket and optionally record the address of the
1720  *      sender. We verify the buffers are writable and if needed move the
1721  *      sender address from kernel to user space.
1722  */
1723 
1724 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1725                 unsigned int, flags, struct sockaddr __user *, addr,
1726                 int __user *, addr_len)
1727 {
1728         struct socket *sock;
1729         struct iovec iov;
1730         struct msghdr msg;
1731         struct sockaddr_storage address;
1732         int err, err2;
1733         int fput_needed;
1734 
1735         err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1736         if (unlikely(err))
1737                 return err;
1738         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1739         if (!sock)
1740                 goto out;
1741 
1742         msg.msg_control = NULL;
1743         msg.msg_controllen = 0;
1744         /* Save some cycles and don't copy the address if not needed */
1745         msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1746         /* We assume all kernel code knows the size of sockaddr_storage */
1747         msg.msg_namelen = 0;
1748         msg.msg_iocb = NULL;
1749         msg.msg_flags = 0;
1750         if (sock->file->f_flags & O_NONBLOCK)
1751                 flags |= MSG_DONTWAIT;
1752         err = sock_recvmsg(sock, &msg, flags);
1753 
1754         if (err >= 0 && addr != NULL) {
1755                 err2 = move_addr_to_user(&address,
1756                                          msg.msg_namelen, addr, addr_len);
1757                 if (err2 < 0)
1758                         err = err2;
1759         }
1760 
1761         fput_light(sock->file, fput_needed);
1762 out:
1763         return err;
1764 }
1765 
1766 /*
1767  *      Receive a datagram from a socket.
1768  */
1769 
1770 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1771                 unsigned int, flags)
1772 {
1773         return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1774 }
1775 
1776 /*
1777  *      Set a socket option. Because we don't know the option lengths we have
1778  *      to pass the user mode parameter for the protocols to sort out.
1779  */
1780 
1781 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1782                 char __user *, optval, int, optlen)
1783 {
1784         int err, fput_needed;
1785         struct socket *sock;
1786 
1787         if (optlen < 0)
1788                 return -EINVAL;
1789 
1790         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1791         if (sock != NULL) {
1792                 err = security_socket_setsockopt(sock, level, optname);
1793                 if (err)
1794                         goto out_put;
1795 
1796                 if (level == SOL_SOCKET)
1797                         err =
1798                             sock_setsockopt(sock, level, optname, optval,
1799                                             optlen);
1800                 else
1801                         err =
1802                             sock->ops->setsockopt(sock, level, optname, optval,
1803                                                   optlen);
1804 out_put:
1805                 fput_light(sock->file, fput_needed);
1806         }
1807         return err;
1808 }
1809 
1810 /*
1811  *      Get a socket option. Because we don't know the option lengths we have
1812  *      to pass a user mode parameter for the protocols to sort out.
1813  */
1814 
1815 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1816                 char __user *, optval, int __user *, optlen)
1817 {
1818         int err, fput_needed;
1819         struct socket *sock;
1820 
1821         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1822         if (sock != NULL) {
1823                 err = security_socket_getsockopt(sock, level, optname);
1824                 if (err)
1825                         goto out_put;
1826 
1827                 if (level == SOL_SOCKET)
1828                         err =
1829                             sock_getsockopt(sock, level, optname, optval,
1830                                             optlen);
1831                 else
1832                         err =
1833                             sock->ops->getsockopt(sock, level, optname, optval,
1834                                                   optlen);
1835 out_put:
1836                 fput_light(sock->file, fput_needed);
1837         }
1838         return err;
1839 }
1840 
1841 /*
1842  *      Shutdown a socket.
1843  */
1844 
1845 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1846 {
1847         int err, fput_needed;
1848         struct socket *sock;
1849 
1850         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1851         if (sock != NULL) {
1852                 err = security_socket_shutdown(sock, how);
1853                 if (!err)
1854                         err = sock->ops->shutdown(sock, how);
1855                 fput_light(sock->file, fput_needed);
1856         }
1857         return err;
1858 }
1859 
1860 /* A couple of helpful macros for getting the address of the 32/64 bit
1861  * fields which are the same type (int / unsigned) on our platforms.
1862  */
1863 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1864 #define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1865 #define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1866 
1867 struct used_address {
1868         struct sockaddr_storage name;
1869         unsigned int name_len;
1870 };
1871 
1872 static int copy_msghdr_from_user(struct msghdr *kmsg,
1873                                  struct user_msghdr __user *umsg,
1874                                  struct sockaddr __user **save_addr,
1875                                  struct iovec **iov)
1876 {
1877         struct sockaddr __user *uaddr;
1878         struct iovec __user *uiov;
1879         size_t nr_segs;
1880         ssize_t err;
1881 
1882         if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1883             __get_user(uaddr, &umsg->msg_name) ||
1884             __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1885             __get_user(uiov, &umsg->msg_iov) ||
1886             __get_user(nr_segs, &umsg->msg_iovlen) ||
1887             __get_user(kmsg->msg_control, &umsg->msg_control) ||
1888             __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1889             __get_user(kmsg->msg_flags, &umsg->msg_flags))
1890                 return -EFAULT;
1891 
1892         if (!uaddr)
1893                 kmsg->msg_namelen = 0;
1894 
1895         if (kmsg->msg_namelen < 0)
1896                 return -EINVAL;
1897 
1898         if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1899                 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1900 
1901         if (save_addr)
1902                 *save_addr = uaddr;
1903 
1904         if (uaddr && kmsg->msg_namelen) {
1905                 if (!save_addr) {
1906                         err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1907                                                   kmsg->msg_name);
1908                         if (err < 0)
1909                                 return err;
1910                 }
1911         } else {
1912                 kmsg->msg_name = NULL;
1913                 kmsg->msg_namelen = 0;
1914         }
1915 
1916         if (nr_segs > UIO_MAXIOV)
1917                 return -EMSGSIZE;
1918 
1919         kmsg->msg_iocb = NULL;
1920 
1921         return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1922                             UIO_FASTIOV, iov, &kmsg->msg_iter);
1923 }
1924 
1925 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1926                          struct msghdr *msg_sys, unsigned int flags,
1927                          struct used_address *used_address,
1928                          unsigned int allowed_msghdr_flags)
1929 {
1930         struct compat_msghdr __user *msg_compat =
1931             (struct compat_msghdr __user *)msg;
1932         struct sockaddr_storage address;
1933         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1934         unsigned char ctl[sizeof(struct cmsghdr) + 20]
1935                                 __aligned(sizeof(__kernel_size_t));
1936         /* 20 is size of ipv6_pktinfo */
1937         unsigned char *ctl_buf = ctl;
1938         int ctl_len;
1939         ssize_t err;
1940 
1941         msg_sys->msg_name = &address;
1942 
1943         if (MSG_CMSG_COMPAT & flags)
1944                 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1945         else
1946                 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1947         if (err < 0)
1948                 return err;
1949 
1950         err = -ENOBUFS;
1951 
1952         if (msg_sys->msg_controllen > INT_MAX)
1953                 goto out_freeiov;
1954         flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1955         ctl_len = msg_sys->msg_controllen;
1956         if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1957                 err =
1958                     cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1959                                                      sizeof(ctl));
1960                 if (err)
1961                         goto out_freeiov;
1962                 ctl_buf = msg_sys->msg_control;
1963                 ctl_len = msg_sys->msg_controllen;
1964         } else if (ctl_len) {
1965                 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
1966                              CMSG_ALIGN(sizeof(struct cmsghdr)));
1967                 if (ctl_len > sizeof(ctl)) {
1968                         ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1969                         if (ctl_buf == NULL)
1970                                 goto out_freeiov;
1971                 }
1972                 err = -EFAULT;
1973                 /*
1974                  * Careful! Before this, msg_sys->msg_control contains a user pointer.
1975                  * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1976                  * checking falls down on this.
1977                  */
1978                 if (copy_from_user(ctl_buf,
1979                                    (void __user __force *)msg_sys->msg_control,
1980                                    ctl_len))
1981                         goto out_freectl;
1982                 msg_sys->msg_control = ctl_buf;
1983         }
1984         msg_sys->msg_flags = flags;
1985 
1986         if (sock->file->f_flags & O_NONBLOCK)
1987                 msg_sys->msg_flags |= MSG_DONTWAIT;
1988         /*
1989          * If this is sendmmsg() and current destination address is same as
1990          * previously succeeded address, omit asking LSM's decision.
1991          * used_address->name_len is initialized to UINT_MAX so that the first
1992          * destination address never matches.
1993          */
1994         if (used_address && msg_sys->msg_name &&
1995             used_address->name_len == msg_sys->msg_namelen &&
1996             !memcmp(&used_address->name, msg_sys->msg_name,
1997                     used_address->name_len)) {
1998                 err = sock_sendmsg_nosec(sock, msg_sys);
1999                 goto out_freectl;
2000         }
2001         err = sock_sendmsg(sock, msg_sys);
2002         /*
2003          * If this is sendmmsg() and sending to current destination address was
2004          * successful, remember it.
2005          */
2006         if (used_address && err >= 0) {
2007                 used_address->name_len = msg_sys->msg_namelen;
2008                 if (msg_sys->msg_name)
2009                         memcpy(&used_address->name, msg_sys->msg_name,
2010                                used_address->name_len);
2011         }
2012 
2013 out_freectl:
2014         if (ctl_buf != ctl)
2015                 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2016 out_freeiov:
2017         kfree(iov);
2018         return err;
2019 }
2020 
2021 /*
2022  *      BSD sendmsg interface
2023  */
2024 
2025 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2026 {
2027         int fput_needed, err;
2028         struct msghdr msg_sys;
2029         struct socket *sock;
2030 
2031         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2032         if (!sock)
2033                 goto out;
2034 
2035         err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2036 
2037         fput_light(sock->file, fput_needed);
2038 out:
2039         return err;
2040 }
2041 
2042 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2043 {
2044         if (flags & MSG_CMSG_COMPAT)
2045                 return -EINVAL;
2046         return __sys_sendmsg(fd, msg, flags);
2047 }
2048 
2049 /*
2050  *      Linux sendmmsg interface
2051  */
2052 
2053 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2054                    unsigned int flags)
2055 {
2056         int fput_needed, err, datagrams;
2057         struct socket *sock;
2058         struct mmsghdr __user *entry;
2059         struct compat_mmsghdr __user *compat_entry;
2060         struct msghdr msg_sys;
2061         struct used_address used_address;
2062         unsigned int oflags = flags;
2063 
2064         if (vlen > UIO_MAXIOV)
2065                 vlen = UIO_MAXIOV;
2066 
2067         datagrams = 0;
2068 
2069         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2070         if (!sock)
2071                 return err;
2072 
2073         used_address.name_len = UINT_MAX;
2074         entry = mmsg;
2075         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2076         err = 0;
2077         flags |= MSG_BATCH;
2078 
2079         while (datagrams < vlen) {
2080                 if (datagrams == vlen - 1)
2081                         flags = oflags;
2082 
2083                 if (MSG_CMSG_COMPAT & flags) {
2084                         err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2085                                              &msg_sys, flags, &used_address, MSG_EOR);
2086                         if (err < 0)
2087                                 break;
2088                         err = __put_user(err, &compat_entry->msg_len);
2089                         ++compat_entry;
2090                 } else {
2091                         err = ___sys_sendmsg(sock,
2092                                              (struct user_msghdr __user *)entry,
2093                                              &msg_sys, flags, &used_address, MSG_EOR);
2094                         if (err < 0)
2095                                 break;
2096                         err = put_user(err, &entry->msg_len);
2097                         ++entry;
2098                 }
2099 
2100                 if (err)
2101                         break;
2102                 ++datagrams;
2103                 if (msg_data_left(&msg_sys))
2104                         break;
2105                 cond_resched();
2106         }
2107 
2108         fput_light(sock->file, fput_needed);
2109 
2110         /* We only return an error if no datagrams were able to be sent */
2111         if (datagrams != 0)
2112                 return datagrams;
2113 
2114         return err;
2115 }
2116 
2117 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2118                 unsigned int, vlen, unsigned int, flags)
2119 {
2120         if (flags & MSG_CMSG_COMPAT)
2121                 return -EINVAL;
2122         return __sys_sendmmsg(fd, mmsg, vlen, flags);
2123 }
2124 
2125 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2126                          struct msghdr *msg_sys, unsigned int flags, int nosec)
2127 {
2128         struct compat_msghdr __user *msg_compat =
2129             (struct compat_msghdr __user *)msg;
2130         struct iovec iovstack[UIO_FASTIOV];
2131         struct iovec *iov = iovstack;
2132         unsigned long cmsg_ptr;
2133         int len;
2134         ssize_t err;
2135 
2136         /* kernel mode address */
2137         struct sockaddr_storage addr;
2138 
2139         /* user mode address pointers */
2140         struct sockaddr __user *uaddr;
2141         int __user *uaddr_len = COMPAT_NAMELEN(msg);
2142 
2143         msg_sys->msg_name = &addr;
2144 
2145         if (MSG_CMSG_COMPAT & flags)
2146                 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2147         else
2148                 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2149         if (err < 0)
2150                 return err;
2151 
2152         cmsg_ptr = (unsigned long)msg_sys->msg_control;
2153         msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2154 
2155         /* We assume all kernel code knows the size of sockaddr_storage */
2156         msg_sys->msg_namelen = 0;
2157 
2158         if (sock->file->f_flags & O_NONBLOCK)
2159                 flags |= MSG_DONTWAIT;
2160         err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2161         if (err < 0)
2162                 goto out_freeiov;
2163         len = err;
2164 
2165         if (uaddr != NULL) {
2166                 err = move_addr_to_user(&addr,
2167                                         msg_sys->msg_namelen, uaddr,
2168                                         uaddr_len);
2169                 if (err < 0)
2170                         goto out_freeiov;
2171         }
2172         err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2173                          COMPAT_FLAGS(msg));
2174         if (err)
2175                 goto out_freeiov;
2176         if (MSG_CMSG_COMPAT & flags)
2177                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2178                                  &msg_compat->msg_controllen);
2179         else
2180                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2181                                  &msg->msg_controllen);
2182         if (err)
2183                 goto out_freeiov;
2184         err = len;
2185 
2186 out_freeiov:
2187         kfree(iov);
2188         return err;
2189 }
2190 
2191 /*
2192  *      BSD recvmsg interface
2193  */
2194 
2195 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2196 {
2197         int fput_needed, err;
2198         struct msghdr msg_sys;
2199         struct socket *sock;
2200 
2201         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2202         if (!sock)
2203                 goto out;
2204 
2205         err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2206 
2207         fput_light(sock->file, fput_needed);
2208 out:
2209         return err;
2210 }
2211 
2212 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2213                 unsigned int, flags)
2214 {
2215         if (flags & MSG_CMSG_COMPAT)
2216                 return -EINVAL;
2217         return __sys_recvmsg(fd, msg, flags);
2218 }
2219 
2220 /*
2221  *     Linux recvmmsg interface
2222  */
2223 
2224 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2225                    unsigned int flags, struct timespec *timeout)
2226 {
2227         int fput_needed, err, datagrams;
2228         struct socket *sock;
2229         struct mmsghdr __user *entry;
2230         struct compat_mmsghdr __user *compat_entry;
2231         struct msghdr msg_sys;
2232         struct timespec64 end_time;
2233         struct timespec64 timeout64;
2234 
2235         if (timeout &&
2236             poll_select_set_timeout(&end_time, timeout->tv_sec,
2237                                     timeout->tv_nsec))
2238                 return -EINVAL;
2239 
2240         datagrams = 0;
2241 
2242         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2243         if (!sock)
2244                 return err;
2245 
2246         err = sock_error(sock->sk);
2247         if (err) {
2248                 datagrams = err;
2249                 goto out_put;
2250         }
2251 
2252         entry = mmsg;
2253         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2254 
2255         while (datagrams < vlen) {
2256                 /*
2257                  * No need to ask LSM for more than the first datagram.
2258                  */
2259                 if (MSG_CMSG_COMPAT & flags) {
2260                         err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2261                                              &msg_sys, flags & ~MSG_WAITFORONE,
2262                                              datagrams);
2263                         if (err < 0)
2264                                 break;
2265                         err = __put_user(err, &compat_entry->msg_len);
2266                         ++compat_entry;
2267                 } else {
2268                         err = ___sys_recvmsg(sock,
2269                                              (struct user_msghdr __user *)entry,
2270                                              &msg_sys, flags & ~MSG_WAITFORONE,
2271                                              datagrams);
2272                         if (err < 0)
2273                                 break;
2274                         err = put_user(err, &entry->msg_len);
2275                         ++entry;
2276                 }
2277 
2278                 if (err)
2279                         break;
2280                 ++datagrams;
2281 
2282                 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2283                 if (flags & MSG_WAITFORONE)
2284                         flags |= MSG_DONTWAIT;
2285 
2286                 if (timeout) {
2287                         ktime_get_ts64(&timeout64);
2288                         *timeout = timespec64_to_timespec(
2289                                         timespec64_sub(end_time, timeout64));
2290                         if (timeout->tv_sec < 0) {
2291                                 timeout->tv_sec = timeout->tv_nsec = 0;
2292                                 break;
2293                         }
2294 
2295                         /* Timeout, return less than vlen datagrams */
2296                         if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2297                                 break;
2298                 }
2299 
2300                 /* Out of band data, return right away */
2301                 if (msg_sys.msg_flags & MSG_OOB)
2302                         break;
2303                 cond_resched();
2304         }
2305 
2306         if (err == 0)
2307                 goto out_put;
2308 
2309         if (datagrams == 0) {
2310                 datagrams = err;
2311                 goto out_put;
2312         }
2313 
2314         /*
2315          * We may return less entries than requested (vlen) if the
2316          * sock is non block and there aren't enough datagrams...
2317          */
2318         if (err != -EAGAIN) {
2319                 /*
2320                  * ... or  if recvmsg returns an error after we
2321                  * received some datagrams, where we record the
2322                  * error to return on the next call or if the
2323                  * app asks about it using getsockopt(SO_ERROR).
2324                  */
2325                 sock->sk->sk_err = -err;
2326         }
2327 out_put:
2328         fput_light(sock->file, fput_needed);
2329 
2330         return datagrams;
2331 }
2332 
2333 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2334                 unsigned int, vlen, unsigned int, flags,
2335                 struct timespec __user *, timeout)
2336 {
2337         int datagrams;
2338         struct timespec timeout_sys;
2339 
2340         if (flags & MSG_CMSG_COMPAT)
2341                 return -EINVAL;
2342 
2343         if (!timeout)
2344                 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2345 
2346         if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2347                 return -EFAULT;
2348 
2349         datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2350 
2351         if (datagrams > 0 &&
2352             copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2353                 datagrams = -EFAULT;
2354 
2355         return datagrams;
2356 }
2357 
2358 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2359 /* Argument list sizes for sys_socketcall */
2360 #define AL(x) ((x) * sizeof(unsigned long))
2361 static const unsigned char nargs[21] = {
2362         AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2363         AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2364         AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2365         AL(4), AL(5), AL(4)
2366 };
2367 
2368 #undef AL
2369 
2370 /*
2371  *      System call vectors.
2372  *
2373  *      Argument checking cleaned up. Saved 20% in size.
2374  *  This function doesn't need to set the kernel lock because
2375  *  it is set by the callees.
2376  */
2377 
2378 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2379 {
2380         unsigned long a[AUDITSC_ARGS];
2381         unsigned long a0, a1;
2382         int err;
2383         unsigned int len;
2384 
2385         if (call < 1 || call > SYS_SENDMMSG)
2386                 return -EINVAL;
2387 
2388         len = nargs[call];
2389         if (len > sizeof(a))
2390                 return -EINVAL;
2391 
2392         /* copy_from_user should be SMP safe. */
2393         if (copy_from_user(a, args, len))
2394                 return -EFAULT;
2395 
2396         err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2397         if (err)
2398                 return err;
2399 
2400         a0 = a[0];
2401         a1 = a[1];
2402 
2403         switch (call) {
2404         case SYS_SOCKET:
2405                 err = sys_socket(a0, a1, a[2]);
2406                 break;
2407         case SYS_BIND:
2408                 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2409                 break;
2410         case SYS_CONNECT:
2411                 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2412                 break;
2413         case SYS_LISTEN:
2414                 err = sys_listen(a0, a1);
2415                 break;
2416         case SYS_ACCEPT:
2417                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2418                                   (int __user *)a[2], 0);
2419                 break;
2420         case SYS_GETSOCKNAME:
2421                 err =
2422                     sys_getsockname(a0, (struct sockaddr __user *)a1,
2423                                     (int __user *)a[2]);
2424                 break;
2425         case SYS_GETPEERNAME:
2426                 err =
2427                     sys_getpeername(a0, (struct sockaddr __user *)a1,
2428                                     (int __user *)a[2]);
2429                 break;
2430         case SYS_SOCKETPAIR:
2431                 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2432                 break;
2433         case SYS_SEND:
2434                 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2435                 break;
2436         case SYS_SENDTO:
2437                 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2438                                  (struct sockaddr __user *)a[4], a[5]);
2439                 break;
2440         case SYS_RECV:
2441                 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2442                 break;
2443         case SYS_RECVFROM:
2444                 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2445                                    (struct sockaddr __user *)a[4],
2446                                    (int __user *)a[5]);
2447                 break;
2448         case SYS_SHUTDOWN:
2449                 err = sys_shutdown(a0, a1);
2450                 break;
2451         case SYS_SETSOCKOPT:
2452                 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2453                 break;
2454         case SYS_GETSOCKOPT:
2455                 err =
2456                     sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2457                                    (int __user *)a[4]);
2458                 break;
2459         case SYS_SENDMSG:
2460                 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2461                 break;
2462         case SYS_SENDMMSG:
2463                 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2464                 break;
2465         case SYS_RECVMSG:
2466                 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2467                 break;
2468         case SYS_RECVMMSG:
2469                 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2470                                    (struct timespec __user *)a[4]);
2471                 break;
2472         case SYS_ACCEPT4:
2473                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2474                                   (int __user *)a[2], a[3]);
2475                 break;
2476         default:
2477                 err = -EINVAL;
2478                 break;
2479         }
2480         return err;
2481 }
2482 
2483 #endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2484 
2485 /**
2486  *      sock_register - add a socket protocol handler
2487  *      @ops: description of protocol
2488  *
2489  *      This function is called by a protocol handler that wants to
2490  *      advertise its address family, and have it linked into the
2491  *      socket interface. The value ops->family corresponds to the
2492  *      socket system call protocol family.
2493  */
2494 int sock_register(const struct net_proto_family *ops)
2495 {
2496         int err;
2497 
2498         if (ops->family >= NPROTO) {
2499                 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2500                 return -ENOBUFS;
2501         }
2502 
2503         spin_lock(&net_family_lock);
2504         if (rcu_dereference_protected(net_families[ops->family],
2505                                       lockdep_is_held(&net_family_lock)))
2506                 err = -EEXIST;
2507         else {
2508                 rcu_assign_pointer(net_families[ops->family], ops);
2509                 err = 0;
2510         }
2511         spin_unlock(&net_family_lock);
2512 
2513         pr_info("NET: Registered protocol family %d\n", ops->family);
2514         return err;
2515 }
2516 EXPORT_SYMBOL(sock_register);
2517 
2518 /**
2519  *      sock_unregister - remove a protocol handler
2520  *      @family: protocol family to remove
2521  *
2522  *      This function is called by a protocol handler that wants to
2523  *      remove its address family, and have it unlinked from the
2524  *      new socket creation.
2525  *
2526  *      If protocol handler is a module, then it can use module reference
2527  *      counts to protect against new references. If protocol handler is not
2528  *      a module then it needs to provide its own protection in
2529  *      the ops->create routine.
2530  */
2531 void sock_unregister(int family)
2532 {
2533         BUG_ON(family < 0 || family >= NPROTO);
2534 
2535         spin_lock(&net_family_lock);
2536         RCU_INIT_POINTER(net_families[family], NULL);
2537         spin_unlock(&net_family_lock);
2538 
2539         synchronize_rcu();
2540 
2541         pr_info("NET: Unregistered protocol family %d\n", family);
2542 }
2543 EXPORT_SYMBOL(sock_unregister);
2544 
2545 static int __init sock_init(void)
2546 {
2547         int err;
2548         /*
2549          *      Initialize the network sysctl infrastructure.
2550          */
2551         err = net_sysctl_init();
2552         if (err)
2553                 goto out;
2554 
2555         /*
2556          *      Initialize skbuff SLAB cache
2557          */
2558         skb_init();
2559 
2560         /*
2561          *      Initialize the protocols module.
2562          */
2563 
2564         init_inodecache();
2565 
2566         err = register_filesystem(&sock_fs_type);
2567         if (err)
2568                 goto out_fs;
2569         sock_mnt = kern_mount(&sock_fs_type);
2570         if (IS_ERR(sock_mnt)) {
2571                 err = PTR_ERR(sock_mnt);
2572                 goto out_mount;
2573         }
2574 
2575         /* The real protocol initialization is performed in later initcalls.
2576          */
2577 
2578 #ifdef CONFIG_NETFILTER
2579         err = netfilter_init();
2580         if (err)
2581                 goto out;
2582 #endif
2583 
2584         ptp_classifier_init();
2585 
2586 out:
2587         return err;
2588 
2589 out_mount:
2590         unregister_filesystem(&sock_fs_type);
2591 out_fs:
2592         goto out;
2593 }
2594 
2595 core_initcall(sock_init);       /* early initcall */
2596 
2597 #ifdef CONFIG_PROC_FS
2598 void socket_seq_show(struct seq_file *seq)
2599 {
2600         int cpu;
2601         int counter = 0;
2602 
2603         for_each_possible_cpu(cpu)
2604             counter += per_cpu(sockets_in_use, cpu);
2605 
2606         /* It can be negative, by the way. 8) */
2607         if (counter < 0)
2608                 counter = 0;
2609 
2610         seq_printf(seq, "sockets: used %d\n", counter);
2611 }
2612 #endif                          /* CONFIG_PROC_FS */
2613 
2614 #ifdef CONFIG_COMPAT
2615 static int do_siocgstamp(struct net *net, struct socket *sock,
2616                          unsigned int cmd, void __user *up)
2617 {
2618         mm_segment_t old_fs = get_fs();
2619         struct timeval ktv;
2620         int err;
2621 
2622         set_fs(KERNEL_DS);
2623         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2624         set_fs(old_fs);
2625         if (!err)
2626                 err = compat_put_timeval(&ktv, up);
2627 
2628         return err;
2629 }
2630 
2631 static int do_siocgstampns(struct net *net, struct socket *sock,
2632                            unsigned int cmd, void __user *up)
2633 {
2634         mm_segment_t old_fs = get_fs();
2635         struct timespec kts;
2636         int err;
2637 
2638         set_fs(KERNEL_DS);
2639         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2640         set_fs(old_fs);
2641         if (!err)
2642                 err = compat_put_timespec(&kts, up);
2643 
2644         return err;
2645 }
2646 
2647 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2648 {
2649         struct ifreq __user *uifr;
2650         int err;
2651 
2652         uifr = compat_alloc_user_space(sizeof(struct ifreq));
2653         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2654                 return -EFAULT;
2655 
2656         err = dev_ioctl(net, SIOCGIFNAME, uifr);
2657         if (err)
2658                 return err;
2659 
2660         if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2661                 return -EFAULT;
2662 
2663         return 0;
2664 }
2665 
2666 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2667 {
2668         struct compat_ifconf ifc32;
2669         struct ifconf ifc;
2670         struct ifconf __user *uifc;
2671         struct compat_ifreq __user *ifr32;
2672         struct ifreq __user *ifr;
2673         unsigned int i, j;
2674         int err;
2675 
2676         if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2677                 return -EFAULT;
2678 
2679         memset(&ifc, 0, sizeof(ifc));
2680         if (ifc32.ifcbuf == 0) {
2681                 ifc32.ifc_len = 0;
2682                 ifc.ifc_len = 0;
2683                 ifc.ifc_req = NULL;
2684                 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2685         } else {
2686                 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2687                         sizeof(struct ifreq);
2688                 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2689                 ifc.ifc_len = len;
2690                 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2691                 ifr32 = compat_ptr(ifc32.ifcbuf);
2692                 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2693                         if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2694                                 return -EFAULT;
2695                         ifr++;
2696                         ifr32++;
2697                 }
2698         }
2699         if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2700                 return -EFAULT;
2701 
2702         err = dev_ioctl(net, SIOCGIFCONF, uifc);
2703         if (err)
2704                 return err;
2705 
2706         if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2707                 return -EFAULT;
2708 
2709         ifr = ifc.ifc_req;
2710         ifr32 = compat_ptr(ifc32.ifcbuf);
2711         for (i = 0, j = 0;
2712              i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2713              i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2714                 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2715                         return -EFAULT;
2716                 ifr32++;
2717                 ifr++;
2718         }
2719 
2720         if (ifc32.ifcbuf == 0) {
2721                 /* Translate from 64-bit structure multiple to
2722                  * a 32-bit one.
2723                  */
2724                 i = ifc.ifc_len;
2725                 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2726                 ifc32.ifc_len = i;
2727         } else {
2728                 ifc32.ifc_len = i;
2729         }
2730         if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2731                 return -EFAULT;
2732 
2733         return 0;
2734 }
2735 
2736 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2737 {
2738         struct compat_ethtool_rxnfc __user *compat_rxnfc;
2739         bool convert_in = false, convert_out = false;
2740         size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2741         struct ethtool_rxnfc __user *rxnfc;
2742         struct ifreq __user *ifr;
2743         u32 rule_cnt = 0, actual_rule_cnt;
2744         u32 ethcmd;
2745         u32 data;
2746         int ret;
2747 
2748         if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2749                 return -EFAULT;
2750 
2751         compat_rxnfc = compat_ptr(data);
2752 
2753         if (get_user(ethcmd, &compat_rxnfc->cmd))
2754                 return -EFAULT;
2755 
2756         /* Most ethtool structures are defined without padding.
2757          * Unfortunately struct ethtool_rxnfc is an exception.
2758          */
2759         switch (ethcmd) {
2760         default:
2761                 break;
2762         case ETHTOOL_GRXCLSRLALL:
2763                 /* Buffer size is variable */
2764                 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2765                         return -EFAULT;
2766                 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2767                         return -ENOMEM;
2768                 buf_size += rule_cnt * sizeof(u32);
2769                 /* fall through */
2770         case ETHTOOL_GRXRINGS:
2771         case ETHTOOL_GRXCLSRLCNT:
2772         case ETHTOOL_GRXCLSRULE:
2773         case ETHTOOL_SRXCLSRLINS:
2774                 convert_out = true;
2775                 /* fall through */
2776         case ETHTOOL_SRXCLSRLDEL:
2777                 buf_size += sizeof(struct ethtool_rxnfc);
2778                 convert_in = true;
2779                 break;
2780         }
2781 
2782         ifr = compat_alloc_user_space(buf_size);
2783         rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2784 
2785         if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2786                 return -EFAULT;
2787 
2788         if (put_user(convert_in ? rxnfc : compat_ptr(data),
2789                      &ifr->ifr_ifru.ifru_data))
2790                 return -EFAULT;
2791 
2792         if (convert_in) {
2793                 /* We expect there to be holes between fs.m_ext and
2794                  * fs.ring_cookie and at the end of fs, but nowhere else.
2795                  */
2796                 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2797                              sizeof(compat_rxnfc->fs.m_ext) !=
2798                              offsetof(struct ethtool_rxnfc, fs.m_ext) +
2799                              sizeof(rxnfc->fs.m_ext));
2800                 BUILD_BUG_ON(
2801                         offsetof(struct compat_ethtool_rxnfc, fs.location) -
2802                         offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2803                         offsetof(struct ethtool_rxnfc, fs.location) -
2804                         offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2805 
2806                 if (copy_in_user(rxnfc, compat_rxnfc,
2807                                  (void __user *)(&rxnfc->fs.m_ext + 1) -
2808                                  (void __user *)rxnfc) ||
2809                     copy_in_user(&rxnfc->fs.ring_cookie,
2810                                  &compat_rxnfc->fs.ring_cookie,
2811                                  (void __user *)(&rxnfc->fs.location + 1) -
2812                                  (void __user *)&rxnfc->fs.ring_cookie) ||
2813                     copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2814                                  sizeof(rxnfc->rule_cnt)))
2815                         return -EFAULT;
2816         }
2817 
2818         ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2819         if (ret)
2820                 return ret;
2821 
2822         if (convert_out) {
2823                 if (copy_in_user(compat_rxnfc, rxnfc,
2824                                  (const void __user *)(&rxnfc->fs.m_ext + 1) -
2825                                  (const void __user *)rxnfc) ||
2826                     copy_in_user(&compat_rxnfc->fs.ring_cookie,
2827                                  &rxnfc->fs.ring_cookie,
2828                                  (const void __user *)(&rxnfc->fs.location + 1) -
2829                                  (const void __user *)&rxnfc->fs.ring_cookie) ||
2830                     copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2831                                  sizeof(rxnfc->rule_cnt)))
2832                         return -EFAULT;
2833 
2834                 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2835                         /* As an optimisation, we only copy the actual
2836                          * number of rules that the underlying
2837                          * function returned.  Since Mallory might
2838                          * change the rule count in user memory, we
2839                          * check that it is less than the rule count
2840                          * originally given (as the user buffer size),
2841                          * which has been range-checked.
2842                          */
2843                         if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2844                                 return -EFAULT;
2845                         if (actual_rule_cnt < rule_cnt)
2846                                 rule_cnt = actual_rule_cnt;
2847                         if (copy_in_user(&compat_rxnfc->rule_locs[0],
2848                                          &rxnfc->rule_locs[0],
2849                                          rule_cnt * sizeof(u32)))
2850                                 return -EFAULT;
2851                 }
2852         }
2853 
2854         return 0;
2855 }
2856 
2857 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2858 {
2859         void __user *uptr;
2860         compat_uptr_t uptr32;
2861         struct ifreq __user *uifr;
2862 
2863         uifr = compat_alloc_user_space(sizeof(*uifr));
2864         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2865                 return -EFAULT;
2866 
2867         if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2868                 return -EFAULT;
2869 
2870         uptr = compat_ptr(uptr32);
2871 
2872         if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2873                 return -EFAULT;
2874 
2875         return dev_ioctl(net, SIOCWANDEV, uifr);
2876 }
2877 
2878 static int bond_ioctl(struct net *net, unsigned int cmd,
2879                          struct compat_ifreq __user *ifr32)
2880 {
2881         struct ifreq kifr;
2882         mm_segment_t old_fs;
2883         int err;
2884 
2885         switch (cmd) {
2886         case SIOCBONDENSLAVE:
2887         case SIOCBONDRELEASE:
2888         case SIOCBONDSETHWADDR:
2889         case SIOCBONDCHANGEACTIVE:
2890                 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2891                         return -EFAULT;
2892 
2893                 old_fs = get_fs();
2894                 set_fs(KERNEL_DS);
2895                 err = dev_ioctl(net, cmd,
2896                                 (struct ifreq __user __force *) &kifr);
2897                 set_fs(old_fs);
2898 
2899                 return err;
2900         default:
2901                 return -ENOIOCTLCMD;
2902         }
2903 }
2904 
2905 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2906 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2907                                  struct compat_ifreq __user *u_ifreq32)
2908 {
2909         struct ifreq __user *u_ifreq64;
2910         char tmp_buf[IFNAMSIZ];
2911         void __user *data64;
2912         u32 data32;
2913 
2914         if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2915                            IFNAMSIZ))
2916                 return -EFAULT;
2917         if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2918                 return -EFAULT;
2919         data64 = compat_ptr(data32);
2920 
2921         u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2922 
2923         if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2924                          IFNAMSIZ))
2925                 return -EFAULT;
2926         if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2927                 return -EFAULT;
2928 
2929         return dev_ioctl(net, cmd, u_ifreq64);
2930 }
2931 
2932 static int dev_ifsioc(struct net *net, struct socket *sock,
2933                          unsigned int cmd, struct compat_ifreq __user *uifr32)
2934 {
2935         struct ifreq __user *uifr;
2936         int err;
2937 
2938         uifr = compat_alloc_user_space(sizeof(*uifr));
2939         if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2940                 return -EFAULT;
2941 
2942         err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2943 
2944         if (!err) {
2945                 switch (cmd) {
2946                 case SIOCGIFFLAGS:
2947                 case SIOCGIFMETRIC:
2948                 case SIOCGIFMTU:
2949                 case SIOCGIFMEM:
2950                 case SIOCGIFHWADDR:
2951                 case SIOCGIFINDEX:
2952                 case SIOCGIFADDR:
2953                 case SIOCGIFBRDADDR:
2954                 case SIOCGIFDSTADDR:
2955                 case SIOCGIFNETMASK:
2956                 case SIOCGIFPFLAGS:
2957                 case SIOCGIFTXQLEN:
2958                 case SIOCGMIIPHY:
2959                 case SIOCGMIIREG:
2960                         if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2961                                 err = -EFAULT;
2962                         break;
2963                 }
2964         }
2965         return err;
2966 }
2967 
2968 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2969                         struct compat_ifreq __user *uifr32)
2970 {
2971         struct ifreq ifr;
2972         struct compat_ifmap __user *uifmap32;
2973         mm_segment_t old_fs;
2974         int err;
2975 
2976         uifmap32 = &uifr32->ifr_ifru.ifru_map;
2977         err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2978         err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2979         err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2980         err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2981         err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2982         err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2983         err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2984         if (err)
2985                 return -EFAULT;
2986 
2987         old_fs = get_fs();
2988         set_fs(KERNEL_DS);
2989         err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
2990         set_fs(old_fs);
2991 
2992         if (cmd == SIOCGIFMAP && !err) {
2993                 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2994                 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2995                 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2996                 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2997                 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2998                 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2999                 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3000                 if (err)
3001                         err = -EFAULT;
3002         }
3003         return err;
3004 }
3005 
3006 struct rtentry32 {
3007         u32             rt_pad1;
3008         struct sockaddr rt_dst;         /* target address               */
3009         struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3010         struct sockaddr rt_genmask;     /* target network mask (IP)     */
3011         unsigned short  rt_flags;
3012         short           rt_pad2;
3013         u32             rt_pad3;
3014         unsigned char   rt_tos;
3015         unsigned char   rt_class;
3016         short           rt_pad4;
3017         short           rt_metric;      /* +1 for binary compatibility! */
3018         /* char * */ u32 rt_dev;        /* forcing the device at add    */
3019         u32             rt_mtu;         /* per route MTU/Window         */
3020         u32             rt_window;      /* Window clamping              */
3021         unsigned short  rt_irtt;        /* Initial RTT                  */
3022 };
3023 
3024 struct in6_rtmsg32 {
3025         struct in6_addr         rtmsg_dst;
3026         struct in6_addr         rtmsg_src;
3027         struct in6_addr         rtmsg_gateway;
3028         u32                     rtmsg_type;
3029         u16                     rtmsg_dst_len;
3030         u16                     rtmsg_src_len;
3031         u32                     rtmsg_metric;
3032         u32                     rtmsg_info;
3033         u32                     rtmsg_flags;
3034         s32                     rtmsg_ifindex;
3035 };
3036 
3037 static int routing_ioctl(struct net *net, struct socket *sock,
3038                          unsigned int cmd, void __user *argp)
3039 {
3040         int ret;
3041         void *r = NULL;
3042         struct in6_rtmsg r6;
3043         struct rtentry r4;
3044         char devname[16];
3045         u32 rtdev;
3046         mm_segment_t old_fs = get_fs();
3047 
3048         if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3049                 struct in6_rtmsg32 __user *ur6 = argp;
3050                 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3051                         3 * sizeof(struct in6_addr));
3052                 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3053                 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3054                 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3055                 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3056                 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3057                 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3058                 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3059 
3060                 r = (void *) &r6;
3061         } else { /* ipv4 */
3062                 struct rtentry32 __user *ur4 = argp;
3063                 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3064                                         3 * sizeof(struct sockaddr));
3065                 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3066                 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3067                 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3068                 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3069                 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3070                 ret |= get_user(rtdev, &(ur4->rt_dev));
3071                 if (rtdev) {
3072                         ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3073                         r4.rt_dev = (char __user __force *)devname;
3074                         devname[15] = 0;
3075                 } else
3076                         r4.rt_dev = NULL;
3077 
3078                 r = (void *) &r4;
3079         }
3080 
3081         if (ret) {
3082                 ret = -EFAULT;
3083                 goto out;
3084         }
3085 
3086         set_fs(KERNEL_DS);
3087         ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3088         set_fs(old_fs);
3089 
3090 out:
3091         return ret;
3092 }
3093 
3094 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3095  * for some operations; this forces use of the newer bridge-utils that
3096  * use compatible ioctls
3097  */
3098 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3099 {
3100         compat_ulong_t tmp;
3101 
3102         if (get_user(tmp, argp))
3103                 return -EFAULT;
3104         if (tmp == BRCTL_GET_VERSION)
3105                 return BRCTL_VERSION + 1;
3106         return -EINVAL;
3107 }
3108 
3109 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3110                          unsigned int cmd, unsigned long arg)
3111 {
3112         void __user *argp = compat_ptr(arg);
3113         struct sock *sk = sock->sk;
3114         struct net *net = sock_net(sk);
3115 
3116         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3117                 return compat_ifr_data_ioctl(net, cmd, argp);
3118 
3119         switch (cmd) {
3120         case SIOCSIFBR:
3121         case SIOCGIFBR:
3122                 return old_bridge_ioctl(argp);
3123         case SIOCGIFNAME:
3124                 return dev_ifname32(net, argp);
3125         case SIOCGIFCONF:
3126                 return dev_ifconf(net, argp);
3127         case SIOCETHTOOL:
3128                 return ethtool_ioctl(net, argp);
3129         case SIOCWANDEV:
3130                 return compat_siocwandev(net, argp);
3131         case SIOCGIFMAP:
3132         case SIOCSIFMAP:
3133                 return compat_sioc_ifmap(net, cmd, argp);
3134         case SIOCBONDENSLAVE:
3135         case SIOCBONDRELEASE:
3136         case SIOCBONDSETHWADDR:
3137         case SIOCBONDCHANGEACTIVE:
3138                 return bond_ioctl(net, cmd, argp);
3139         case SIOCADDRT:
3140         case SIOCDELRT:
3141                 return routing_ioctl(net, sock, cmd, argp);
3142         case SIOCGSTAMP:
3143                 return do_siocgstamp(net, sock, cmd, argp);
3144         case SIOCGSTAMPNS:
3145                 return do_siocgstampns(net, sock, cmd, argp);
3146         case SIOCBONDSLAVEINFOQUERY:
3147         case SIOCBONDINFOQUERY:
3148         case SIOCSHWTSTAMP:
3149         case SIOCGHWTSTAMP:
3150                 return compat_ifr_data_ioctl(net, cmd, argp);
3151 
3152         case FIOSETOWN:
3153         case SIOCSPGRP:
3154         case FIOGETOWN:
3155         case SIOCGPGRP:
3156         case SIOCBRADDBR:
3157         case SIOCBRDELBR:
3158         case SIOCGIFVLAN:
3159         case SIOCSIFVLAN:
3160         case SIOCADDDLCI:
3161         case SIOCDELDLCI:
3162         case SIOCGSKNS:
3163                 return sock_ioctl(file, cmd, arg);
3164 
3165         case SIOCGIFFLAGS:
3166         case SIOCSIFFLAGS:
3167         case SIOCGIFMETRIC:
3168         case SIOCSIFMETRIC:
3169         case SIOCGIFMTU:
3170         case SIOCSIFMTU:
3171         case SIOCGIFMEM:
3172         case SIOCSIFMEM:
3173         case SIOCGIFHWADDR:
3174         case SIOCSIFHWADDR:
3175         case SIOCADDMULTI:
3176         case SIOCDELMULTI:
3177         case SIOCGIFINDEX:
3178         case SIOCGIFADDR:
3179         case SIOCSIFADDR:
3180         case SIOCSIFHWBROADCAST:
3181         case SIOCDIFADDR:
3182         case SIOCGIFBRDADDR:
3183         case SIOCSIFBRDADDR:
3184         case SIOCGIFDSTADDR:
3185         case SIOCSIFDSTADDR:
3186         case SIOCGIFNETMASK:
3187         case SIOCSIFNETMASK:
3188         case SIOCSIFPFLAGS:
3189         case SIOCGIFPFLAGS:
3190         case SIOCGIFTXQLEN:
3191         case SIOCSIFTXQLEN:
3192         case SIOCBRADDIF:
3193         case SIOCBRDELIF:
3194         case SIOCSIFNAME:
3195         case SIOCGMIIPHY:
3196         case SIOCGMIIREG:
3197         case SIOCSMIIREG:
3198                 return dev_ifsioc(net, sock, cmd, argp);
3199 
3200         case SIOCSARP:
3201         case SIOCGARP:
3202         case SIOCDARP:
3203         case SIOCATMARK:
3204                 return sock_do_ioctl(net, sock, cmd, arg);
3205         }
3206 
3207         return -ENOIOCTLCMD;
3208 }
3209 
3210 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3211                               unsigned long arg)
3212 {
3213         struct socket *sock = file->private_data;
3214         int ret = -ENOIOCTLCMD;
3215         struct sock *sk;
3216         struct net *net;
3217 
3218         sk = sock->sk;
3219         net = sock_net(sk);
3220 
3221         if (sock->ops->compat_ioctl)
3222                 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3223 
3224         if (ret == -ENOIOCTLCMD &&
3225             (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3226                 ret = compat_wext_handle_ioctl(net, cmd, arg);
3227 
3228         if (ret == -ENOIOCTLCMD)
3229                 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3230 
3231         return ret;
3232 }
3233 #endif
3234 
3235 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3236 {
3237         return sock->ops->bind(sock, addr, addrlen);
3238 }
3239 EXPORT_SYMBOL(kernel_bind);
3240 
3241 int kernel_listen(struct socket *sock, int backlog)
3242 {
3243         return sock->ops->listen(sock, backlog);
3244 }
3245 EXPORT_SYMBOL(kernel_listen);
3246 
3247 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3248 {
3249         struct sock *sk = sock->sk;
3250         int err;
3251 
3252         err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3253                                newsock);
3254         if (err < 0)
3255                 goto done;
3256 
3257         err = sock->ops->accept(sock, *newsock, flags, true);
3258         if (err < 0) {
3259                 sock_release(*newsock);
3260                 *newsock = NULL;
3261                 goto done;
3262         }
3263 
3264         (*newsock)->ops = sock->ops;
3265         __module_get((*newsock)->ops->owner);
3266 
3267 done:
3268         return err;
3269 }
3270 EXPORT_SYMBOL(kernel_accept);
3271 
3272 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3273                    int flags)
3274 {
3275         return sock->ops->connect(sock, addr, addrlen, flags);
3276 }
3277 EXPORT_SYMBOL(kernel_connect);
3278 
3279 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3280                          int *addrlen)
3281 {
3282         return sock->ops->getname(sock, addr, addrlen, 0);
3283 }
3284 EXPORT_SYMBOL(kernel_getsockname);
3285 
3286 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3287                          int *addrlen)
3288 {
3289         return sock->ops->getname(sock, addr, addrlen, 1);
3290 }
3291 EXPORT_SYMBOL(kernel_getpeername);
3292 
3293 int kernel_getsockopt(struct socket *sock, int level, int optname,
3294                         char *optval, int *optlen)
3295 {
3296         mm_segment_t oldfs = get_fs();
3297         char __user *uoptval;
3298         int __user *uoptlen;
3299         int err;
3300 
3301         uoptval = (char __user __force *) optval;
3302         uoptlen = (int __user __force *) optlen;
3303 
3304         set_fs(KERNEL_DS);
3305         if (level == SOL_SOCKET)
3306                 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3307         else
3308                 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3309                                             uoptlen);
3310         set_fs(oldfs);
3311         return err;
3312 }
3313 EXPORT_SYMBOL(kernel_getsockopt);
3314 
3315 int kernel_setsockopt(struct socket *sock, int level, int optname,
3316                         char *optval, unsigned int optlen)
3317 {
3318         mm_segment_t oldfs = get_fs();
3319         char __user *uoptval;
3320         int err;
3321 
3322         uoptval = (char __user __force *) optval;
3323 
3324         set_fs(KERNEL_DS);
3325         if (level == SOL_SOCKET)
3326                 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3327         else
3328                 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3329                                             optlen);
3330         set_fs(oldfs);
3331         return err;
3332 }
3333 EXPORT_SYMBOL(kernel_setsockopt);
3334 
3335 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3336                     size_t size, int flags)
3337 {
3338         if (sock->ops->sendpage)
3339                 return sock->ops->sendpage(sock, page, offset, size, flags);
3340 
3341         return sock_no_sendpage(sock, page, offset, size, flags);
3342 }
3343 EXPORT_SYMBOL(kernel_sendpage);
3344 
3345 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3346 {
3347         mm_segment_t oldfs = get_fs();
3348         int err;
3349 
3350         set_fs(KERNEL_DS);
3351         err = sock->ops->ioctl(sock, cmd, arg);
3352         set_fs(oldfs);
3353 
3354         return err;
3355 }
3356 EXPORT_SYMBOL(kernel_sock_ioctl);
3357 
3358 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3359 {
3360         return sock->ops->shutdown(sock, how);
3361 }
3362 EXPORT_SYMBOL(kernel_sock_shutdown);
3363 
3364 /* This routine returns the IP overhead imposed by a socket i.e.
3365  * the length of the underlying IP header, depending on whether
3366  * this is an IPv4 or IPv6 socket and the length from IP options turned
3367  * on at the socket. Assumes that the caller has a lock on the socket.
3368  */
3369 u32 kernel_sock_ip_overhead(struct sock *sk)
3370 {
3371         struct inet_sock *inet;
3372         struct ip_options_rcu *opt;
3373         u32 overhead = 0;
3374         bool owned_by_user;
3375 #if IS_ENABLED(CONFIG_IPV6)
3376         struct ipv6_pinfo *np;
3377         struct ipv6_txoptions *optv6 = NULL;
3378 #endif /* IS_ENABLED(CONFIG_IPV6) */
3379 
3380         if (!sk)
3381                 return overhead;
3382 
3383         owned_by_user = sock_owned_by_user(sk);
3384         switch (sk->sk_family) {
3385         case AF_INET:
3386                 inet = inet_sk(sk);
3387                 overhead += sizeof(struct iphdr);
3388                 opt = rcu_dereference_protected(inet->inet_opt,
3389                                                 owned_by_user);
3390                 if (opt)
3391                         overhead += opt->opt.optlen;
3392                 return overhead;
3393 #if IS_ENABLED(CONFIG_IPV6)
3394         case AF_INET6:
3395                 np = inet6_sk(sk);
3396                 overhead += sizeof(struct ipv6hdr);
3397                 if (np)
3398                         optv6 = rcu_dereference_protected(np->opt,
3399                                                           owned_by_user);
3400                 if (optv6)
3401                         overhead += (optv6->opt_flen + optv6->opt_nflen);
3402                 return overhead;
3403 #endif /* IS_ENABLED(CONFIG_IPV6) */
3404         default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3405                 return overhead;
3406         }
3407 }
3408 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3409 

~ [ 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