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

TOMOYO Linux Cross Reference
Linux/net/socket.c

Version: ~ [ linux-5.12-rc7 ] ~ [ linux-5.11.13 ] ~ [ linux-5.10.29 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.111 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.186 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.230 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.266 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.266 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * NET          An implementation of the SOCKET network access protocol.
  3  *
  4  * Version:     @(#)socket.c    1.1.93  18/02/95
  5  *
  6  * Authors:     Orest Zborowski, <obz@Kodak.COM>
  7  *              Ross Biro
  8  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  9  *
 10  * Fixes:
 11  *              Anonymous       :       NOTSOCK/BADF cleanup. Error fix in
 12  *                                      shutdown()
 13  *              Alan Cox        :       verify_area() fixes
 14  *              Alan Cox        :       Removed DDI
 15  *              Jonathan Kamens :       SOCK_DGRAM reconnect bug
 16  *              Alan Cox        :       Moved a load of checks to the very
 17  *                                      top level.
 18  *              Alan Cox        :       Move address structures to/from user
 19  *                                      mode above the protocol layers.
 20  *              Rob Janssen     :       Allow 0 length sends.
 21  *              Alan Cox        :       Asynchronous I/O support (cribbed from the
 22  *                                      tty drivers).
 23  *              Niibe Yutaka    :       Asynchronous I/O for writes (4.4BSD style)
 24  *              Jeff Uphoff     :       Made max number of sockets command-line
 25  *                                      configurable.
 26  *              Matti Aarnio    :       Made the number of sockets dynamic,
 27  *                                      to be allocated when needed, and mr.
 28  *                                      Uphoff's max is used as max to be
 29  *                                      allowed to allocate.
 30  *              Linus           :       Argh. removed all the socket allocation
 31  *                                      altogether: it's in the inode now.
 32  *              Alan Cox        :       Made sock_alloc()/sock_release() public
 33  *                                      for NetROM and future kernel nfsd type
 34  *                                      stuff.
 35  *              Alan Cox        :       sendmsg/recvmsg basics.
 36  *              Tom Dyas        :       Export net symbols.
 37  *              Marcin Dalecki  :       Fixed problems with CONFIG_NET="n".
 38  *              Alan Cox        :       Added thread locking to sys_* calls
 39  *                                      for sockets. May have errors at the
 40  *                                      moment.
 41  *              Kevin Buhr      :       Fixed the dumb errors in the above.
 42  *              Andi Kleen      :       Some small cleanups, optimizations,
 43  *                                      and fixed a copy_from_user() bug.
 44  *              Tigran Aivazian :       sys_send(args) calls sys_sendto(args, NULL, 0)
 45  *              Tigran Aivazian :       Made listen(2) backlog sanity checks
 46  *                                      protocol-independent
 47  *
 48  *
 49  *              This program is free software; you can redistribute it and/or
 50  *              modify it under the terms of the GNU General Public License
 51  *              as published by the Free Software Foundation; either version
 52  *              2 of the License, or (at your option) any later version.
 53  *
 54  *
 55  *      This module is effectively the top level interface to the BSD socket
 56  *      paradigm.
 57  *
 58  *      Based upon Swansea University Computer Society NET3.039
 59  */
 60 
 61 #include <linux/mm.h>
 62 #include <linux/socket.h>
 63 #include <linux/file.h>
 64 #include <linux/net.h>
 65 #include <linux/interrupt.h>
 66 #include <linux/thread_info.h>
 67 #include <linux/rcupdate.h>
 68 #include <linux/netdevice.h>
 69 #include <linux/proc_fs.h>
 70 #include <linux/seq_file.h>
 71 #include <linux/mutex.h>
 72 #include <linux/if_bridge.h>
 73 #include <linux/if_frad.h>
 74 #include <linux/if_vlan.h>
 75 #include <linux/ptp_classify.h>
 76 #include <linux/init.h>
 77 #include <linux/poll.h>
 78 #include <linux/cache.h>
 79 #include <linux/module.h>
 80 #include <linux/highmem.h>
 81 #include <linux/mount.h>
 82 #include <linux/security.h>
 83 #include <linux/syscalls.h>
 84 #include <linux/compat.h>
 85 #include <linux/kmod.h>
 86 #include <linux/audit.h>
 87 #include <linux/wireless.h>
 88 #include <linux/nsproxy.h>
 89 #include <linux/magic.h>
 90 #include <linux/slab.h>
 91 #include <linux/xattr.h>
 92 
 93 #include <asm/uaccess.h>
 94 #include <asm/unistd.h>
 95 
 96 #include <net/compat.h>
 97 #include <net/wext.h>
 98 #include <net/cls_cgroup.h>
 99 
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
102 
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
114 #endif
115 
116 static ssize_t sock_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 =         new_sync_read,
144         .write =        new_sync_write,
145         .read_iter =    sock_read_iter,
146         .write_iter =   sock_write_iter,
147         .poll =         sock_poll,
148         .unlocked_ioctl = sock_ioctl,
149 #ifdef CONFIG_COMPAT
150         .compat_ioctl = compat_sock_ioctl,
151 #endif
152         .mmap =         sock_mmap,
153         .release =      sock_close,
154         .fasync =       sock_fasync,
155         .sendpage =     sock_sendpage,
156         .splice_write = generic_splice_sendpage,
157         .splice_read =  sock_splice_read,
158 };
159 
160 /*
161  *      The protocol list. Each protocol is registered in here.
162  */
163 
164 static DEFINE_SPINLOCK(net_family_lock);
165 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
166 
167 /*
168  *      Statistics counters of the socket lists
169  */
170 
171 static DEFINE_PER_CPU(int, sockets_in_use);
172 
173 /*
174  * Support routines.
175  * Move socket addresses back and forth across the kernel/user
176  * divide and look after the messy bits.
177  */
178 
179 /**
180  *      move_addr_to_kernel     -       copy a socket address into kernel space
181  *      @uaddr: Address in user space
182  *      @kaddr: Address in kernel space
183  *      @ulen: Length in user space
184  *
185  *      The address is copied into kernel space. If the provided address is
186  *      too long an error code of -EINVAL is returned. If the copy gives
187  *      invalid addresses -EFAULT is returned. On a success 0 is returned.
188  */
189 
190 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
191 {
192         if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
193                 return -EINVAL;
194         if (ulen == 0)
195                 return 0;
196         if (copy_from_user(kaddr, uaddr, ulen))
197                 return -EFAULT;
198         return audit_sockaddr(ulen, kaddr);
199 }
200 
201 /**
202  *      move_addr_to_user       -       copy an address to user space
203  *      @kaddr: kernel space address
204  *      @klen: length of address in kernel
205  *      @uaddr: user space address
206  *      @ulen: pointer to user length field
207  *
208  *      The value pointed to by ulen on entry is the buffer length available.
209  *      This is overwritten with the buffer space used. -EINVAL is returned
210  *      if an overlong buffer is specified or a negative buffer size. -EFAULT
211  *      is returned if either the buffer or the length field are not
212  *      accessible.
213  *      After copying the data up to the limit the user specifies, the true
214  *      length of the data is written over the length limit the user
215  *      specified. Zero is returned for a success.
216  */
217 
218 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
219                              void __user *uaddr, int __user *ulen)
220 {
221         int err;
222         int len;
223 
224         BUG_ON(klen > sizeof(struct sockaddr_storage));
225         err = get_user(len, ulen);
226         if (err)
227                 return err;
228         if (len > klen)
229                 len = klen;
230         if (len < 0)
231                 return -EINVAL;
232         if (len) {
233                 if (audit_sockaddr(klen, kaddr))
234                         return -ENOMEM;
235                 if (copy_to_user(uaddr, kaddr, len))
236                         return -EFAULT;
237         }
238         /*
239          *      "fromlen shall refer to the value before truncation.."
240          *                      1003.1g
241          */
242         return __put_user(klen, ulen);
243 }
244 
245 static struct kmem_cache *sock_inode_cachep __read_mostly;
246 
247 static struct inode *sock_alloc_inode(struct super_block *sb)
248 {
249         struct socket_alloc *ei;
250         struct socket_wq *wq;
251 
252         ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
253         if (!ei)
254                 return NULL;
255         wq = kmalloc(sizeof(*wq), GFP_KERNEL);
256         if (!wq) {
257                 kmem_cache_free(sock_inode_cachep, ei);
258                 return NULL;
259         }
260         init_waitqueue_head(&wq->wait);
261         wq->fasync_list = NULL;
262         RCU_INIT_POINTER(ei->socket.wq, wq);
263 
264         ei->socket.state = SS_UNCONNECTED;
265         ei->socket.flags = 0;
266         ei->socket.ops = NULL;
267         ei->socket.sk = NULL;
268         ei->socket.file = NULL;
269 
270         return &ei->vfs_inode;
271 }
272 
273 static void sock_destroy_inode(struct inode *inode)
274 {
275         struct socket_alloc *ei;
276         struct socket_wq *wq;
277 
278         ei = container_of(inode, struct socket_alloc, vfs_inode);
279         wq = rcu_dereference_protected(ei->socket.wq, 1);
280         kfree_rcu(wq, rcu);
281         kmem_cache_free(sock_inode_cachep, ei);
282 }
283 
284 static void init_once(void *foo)
285 {
286         struct socket_alloc *ei = (struct socket_alloc *)foo;
287 
288         inode_init_once(&ei->vfs_inode);
289 }
290 
291 static int init_inodecache(void)
292 {
293         sock_inode_cachep = kmem_cache_create("sock_inode_cache",
294                                               sizeof(struct socket_alloc),
295                                               0,
296                                               (SLAB_HWCACHE_ALIGN |
297                                                SLAB_RECLAIM_ACCOUNT |
298                                                SLAB_MEM_SPREAD),
299                                               init_once);
300         if (sock_inode_cachep == NULL)
301                 return -ENOMEM;
302         return 0;
303 }
304 
305 static const struct super_operations sockfs_ops = {
306         .alloc_inode    = sock_alloc_inode,
307         .destroy_inode  = sock_destroy_inode,
308         .statfs         = simple_statfs,
309 };
310 
311 /*
312  * sockfs_dname() is called from d_path().
313  */
314 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
315 {
316         return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317                                 dentry->d_inode->i_ino);
318 }
319 
320 static const struct dentry_operations sockfs_dentry_operations = {
321         .d_dname  = sockfs_dname,
322 };
323 
324 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325                          int flags, const char *dev_name, void *data)
326 {
327         return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328                 &sockfs_dentry_operations, SOCKFS_MAGIC);
329 }
330 
331 static struct vfsmount *sock_mnt __read_mostly;
332 
333 static struct file_system_type sock_fs_type = {
334         .name =         "sockfs",
335         .mount =        sockfs_mount,
336         .kill_sb =      kill_anon_super,
337 };
338 
339 /*
340  *      Obtains the first available file descriptor and sets it up for use.
341  *
342  *      These functions create file structures and maps them to fd space
343  *      of the current process. On success it returns file descriptor
344  *      and file struct implicitly stored in sock->file.
345  *      Note that another thread may close file descriptor before we return
346  *      from this function. We use the fact that now we do not refer
347  *      to socket after mapping. If one day we will need it, this
348  *      function will increment ref. count on file by 1.
349  *
350  *      In any case returned fd MAY BE not valid!
351  *      This race condition is unavoidable
352  *      with shared fd spaces, we cannot solve it inside kernel,
353  *      but we take care of internal coherence yet.
354  */
355 
356 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
357 {
358         struct qstr name = { .name = "" };
359         struct path path;
360         struct file *file;
361 
362         if (dname) {
363                 name.name = dname;
364                 name.len = strlen(name.name);
365         } else if (sock->sk) {
366                 name.name = sock->sk->sk_prot_creator->name;
367                 name.len = strlen(name.name);
368         }
369         path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
370         if (unlikely(!path.dentry))
371                 return ERR_PTR(-ENOMEM);
372         path.mnt = mntget(sock_mnt);
373 
374         d_instantiate(path.dentry, SOCK_INODE(sock));
375 
376         file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
377                   &socket_file_ops);
378         if (unlikely(IS_ERR(file))) {
379                 /* drop dentry, keep inode */
380                 ihold(path.dentry->d_inode);
381                 path_put(&path);
382                 return file;
383         }
384 
385         sock->file = file;
386         file->f_flags = O_RDWR | (flags & O_NONBLOCK);
387         file->private_data = sock;
388         return file;
389 }
390 EXPORT_SYMBOL(sock_alloc_file);
391 
392 static int sock_map_fd(struct socket *sock, int flags)
393 {
394         struct file *newfile;
395         int fd = get_unused_fd_flags(flags);
396         if (unlikely(fd < 0))
397                 return fd;
398 
399         newfile = sock_alloc_file(sock, flags, NULL);
400         if (likely(!IS_ERR(newfile))) {
401                 fd_install(fd, newfile);
402                 return fd;
403         }
404 
405         put_unused_fd(fd);
406         return PTR_ERR(newfile);
407 }
408 
409 struct socket *sock_from_file(struct file *file, int *err)
410 {
411         if (file->f_op == &socket_file_ops)
412                 return file->private_data;      /* set in sock_map_fd */
413 
414         *err = -ENOTSOCK;
415         return NULL;
416 }
417 EXPORT_SYMBOL(sock_from_file);
418 
419 /**
420  *      sockfd_lookup - Go from a file number to its socket slot
421  *      @fd: file handle
422  *      @err: pointer to an error code return
423  *
424  *      The file handle passed in is locked and the socket it is bound
425  *      too is returned. If an error occurs the err pointer is overwritten
426  *      with a negative errno code and NULL is returned. The function checks
427  *      for both invalid handles and passing a handle which is not a socket.
428  *
429  *      On a success the socket object pointer is returned.
430  */
431 
432 struct socket *sockfd_lookup(int fd, int *err)
433 {
434         struct file *file;
435         struct socket *sock;
436 
437         file = fget(fd);
438         if (!file) {
439                 *err = -EBADF;
440                 return NULL;
441         }
442 
443         sock = sock_from_file(file, err);
444         if (!sock)
445                 fput(file);
446         return sock;
447 }
448 EXPORT_SYMBOL(sockfd_lookup);
449 
450 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
451 {
452         struct fd f = fdget(fd);
453         struct socket *sock;
454 
455         *err = -EBADF;
456         if (f.file) {
457                 sock = sock_from_file(f.file, err);
458                 if (likely(sock)) {
459                         *fput_needed = f.flags;
460                         return sock;
461                 }
462                 fdput(f);
463         }
464         return NULL;
465 }
466 
467 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470 static ssize_t sockfs_getxattr(struct dentry *dentry,
471                                const char *name, void *value, size_t size)
472 {
473         const char *proto_name;
474         size_t proto_size;
475         int error;
476 
477         error = -ENODATA;
478         if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479                 proto_name = dentry->d_name.name;
480                 proto_size = strlen(proto_name);
481 
482                 if (value) {
483                         error = -ERANGE;
484                         if (proto_size + 1 > size)
485                                 goto out;
486 
487                         strncpy(value, proto_name, proto_size + 1);
488                 }
489                 error = proto_size + 1;
490         }
491 
492 out:
493         return error;
494 }
495 
496 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
497                                 size_t size)
498 {
499         ssize_t len;
500         ssize_t used = 0;
501 
502         len = security_inode_listsecurity(dentry->d_inode, buffer, size);
503         if (len < 0)
504                 return len;
505         used += len;
506         if (buffer) {
507                 if (size < used)
508                         return -ERANGE;
509                 buffer += len;
510         }
511 
512         len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
513         used += len;
514         if (buffer) {
515                 if (size < used)
516                         return -ERANGE;
517                 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
518                 buffer += len;
519         }
520 
521         return used;
522 }
523 
524 static const struct inode_operations sockfs_inode_ops = {
525         .getxattr = sockfs_getxattr,
526         .listxattr = sockfs_listxattr,
527 };
528 
529 /**
530  *      sock_alloc      -       allocate a socket
531  *
532  *      Allocate a new inode and socket object. The two are bound together
533  *      and initialised. The socket is then returned. If we are out of inodes
534  *      NULL is returned.
535  */
536 
537 static struct socket *sock_alloc(void)
538 {
539         struct inode *inode;
540         struct socket *sock;
541 
542         inode = new_inode_pseudo(sock_mnt->mnt_sb);
543         if (!inode)
544                 return NULL;
545 
546         sock = SOCKET_I(inode);
547 
548         kmemcheck_annotate_bitfield(sock, type);
549         inode->i_ino = get_next_ino();
550         inode->i_mode = S_IFSOCK | S_IRWXUGO;
551         inode->i_uid = current_fsuid();
552         inode->i_gid = current_fsgid();
553         inode->i_op = &sockfs_inode_ops;
554 
555         this_cpu_add(sockets_in_use, 1);
556         return sock;
557 }
558 
559 /**
560  *      sock_release    -       close a socket
561  *      @sock: socket to close
562  *
563  *      The socket is released from the protocol stack if it has a release
564  *      callback, and the inode is then released if the socket is bound to
565  *      an inode not a file.
566  */
567 
568 void sock_release(struct socket *sock)
569 {
570         if (sock->ops) {
571                 struct module *owner = sock->ops->owner;
572 
573                 sock->ops->release(sock);
574                 sock->ops = NULL;
575                 module_put(owner);
576         }
577 
578         if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
579                 pr_err("%s: fasync list not empty!\n", __func__);
580 
581         if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
582                 return;
583 
584         this_cpu_sub(sockets_in_use, 1);
585         if (!sock->file) {
586                 iput(SOCK_INODE(sock));
587                 return;
588         }
589         sock->file = NULL;
590 }
591 EXPORT_SYMBOL(sock_release);
592 
593 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
594 {
595         u8 flags = *tx_flags;
596 
597         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
598                 flags |= SKBTX_HW_TSTAMP;
599 
600         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
601                 flags |= SKBTX_SW_TSTAMP;
602 
603         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
604                 flags |= SKBTX_SCHED_TSTAMP;
605 
606         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
607                 flags |= SKBTX_ACK_TSTAMP;
608 
609         *tx_flags = flags;
610 }
611 EXPORT_SYMBOL(__sock_tx_timestamp);
612 
613 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
614                                        struct msghdr *msg, size_t size)
615 {
616         return sock->ops->sendmsg(iocb, sock, msg, size);
617 }
618 
619 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
620                                  struct msghdr *msg, size_t size)
621 {
622         int err = security_socket_sendmsg(sock, msg, size);
623 
624         return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
625 }
626 
627 static int do_sock_sendmsg(struct socket *sock, struct msghdr *msg,
628                            size_t size, bool nosec)
629 {
630         struct kiocb iocb;
631         int ret;
632 
633         init_sync_kiocb(&iocb, NULL);
634         ret = nosec ? __sock_sendmsg_nosec(&iocb, sock, msg, size) :
635                       __sock_sendmsg(&iocb, sock, msg, size);
636         if (-EIOCBQUEUED == ret)
637                 ret = wait_on_sync_kiocb(&iocb);
638         return ret;
639 }
640 
641 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
642 {
643         return do_sock_sendmsg(sock, msg, size, false);
644 }
645 EXPORT_SYMBOL(sock_sendmsg);
646 
647 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
648 {
649         return do_sock_sendmsg(sock, msg, size, true);
650 }
651 
652 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
653                    struct kvec *vec, size_t num, size_t size)
654 {
655         mm_segment_t oldfs = get_fs();
656         int result;
657 
658         set_fs(KERNEL_DS);
659         /*
660          * the following is safe, since for compiler definitions of kvec and
661          * iovec are identical, yielding the same in-core layout and alignment
662          */
663         iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
664         result = sock_sendmsg(sock, msg, size);
665         set_fs(oldfs);
666         return result;
667 }
668 EXPORT_SYMBOL(kernel_sendmsg);
669 
670 /*
671  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
672  */
673 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
674         struct sk_buff *skb)
675 {
676         int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
677         struct scm_timestamping tss;
678         int empty = 1;
679         struct skb_shared_hwtstamps *shhwtstamps =
680                 skb_hwtstamps(skb);
681 
682         /* Race occurred between timestamp enabling and packet
683            receiving.  Fill in the current time for now. */
684         if (need_software_tstamp && skb->tstamp.tv64 == 0)
685                 __net_timestamp(skb);
686 
687         if (need_software_tstamp) {
688                 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
689                         struct timeval tv;
690                         skb_get_timestamp(skb, &tv);
691                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
692                                  sizeof(tv), &tv);
693                 } else {
694                         struct timespec ts;
695                         skb_get_timestampns(skb, &ts);
696                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
697                                  sizeof(ts), &ts);
698                 }
699         }
700 
701         memset(&tss, 0, sizeof(tss));
702         if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
703             ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
704                 empty = 0;
705         if (shhwtstamps &&
706             (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
707             ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
708                 empty = 0;
709         if (!empty)
710                 put_cmsg(msg, SOL_SOCKET,
711                          SCM_TIMESTAMPING, sizeof(tss), &tss);
712 }
713 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
714 
715 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
716         struct sk_buff *skb)
717 {
718         int ack;
719 
720         if (!sock_flag(sk, SOCK_WIFI_STATUS))
721                 return;
722         if (!skb->wifi_acked_valid)
723                 return;
724 
725         ack = skb->wifi_acked;
726 
727         put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
728 }
729 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
730 
731 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
732                                    struct sk_buff *skb)
733 {
734         if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
735                 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
736                         sizeof(__u32), &skb->dropcount);
737 }
738 
739 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
740         struct sk_buff *skb)
741 {
742         sock_recv_timestamp(msg, sk, skb);
743         sock_recv_drops(msg, sk, skb);
744 }
745 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
746 
747 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
748                                        struct msghdr *msg, size_t size, int flags)
749 {
750         return sock->ops->recvmsg(iocb, sock, msg, size, flags);
751 }
752 
753 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
754                                  struct msghdr *msg, size_t size, int flags)
755 {
756         int err = security_socket_recvmsg(sock, msg, size, flags);
757 
758         return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
759 }
760 
761 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
762                  size_t size, int flags)
763 {
764         struct kiocb iocb;
765         int ret;
766 
767         init_sync_kiocb(&iocb, NULL);
768         ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
769         if (-EIOCBQUEUED == ret)
770                 ret = wait_on_sync_kiocb(&iocb);
771         return ret;
772 }
773 EXPORT_SYMBOL(sock_recvmsg);
774 
775 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
776                               size_t size, int flags)
777 {
778         struct kiocb iocb;
779         int ret;
780 
781         init_sync_kiocb(&iocb, NULL);
782         ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
783         if (-EIOCBQUEUED == ret)
784                 ret = wait_on_sync_kiocb(&iocb);
785         return ret;
786 }
787 
788 /**
789  * kernel_recvmsg - Receive a message from a socket (kernel space)
790  * @sock:       The socket to receive the message from
791  * @msg:        Received message
792  * @vec:        Input s/g array for message data
793  * @num:        Size of input s/g array
794  * @size:       Number of bytes to read
795  * @flags:      Message flags (MSG_DONTWAIT, etc...)
796  *
797  * On return the msg structure contains the scatter/gather array passed in the
798  * vec argument. The array is modified so that it consists of the unfilled
799  * portion of the original array.
800  *
801  * The returned value is the total number of bytes received, or an error.
802  */
803 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
804                    struct kvec *vec, size_t num, size_t size, int flags)
805 {
806         mm_segment_t oldfs = get_fs();
807         int result;
808 
809         set_fs(KERNEL_DS);
810         /*
811          * the following is safe, since for compiler definitions of kvec and
812          * iovec are identical, yielding the same in-core layout and alignment
813          */
814         iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
815         result = sock_recvmsg(sock, msg, size, flags);
816         set_fs(oldfs);
817         return result;
818 }
819 EXPORT_SYMBOL(kernel_recvmsg);
820 
821 static ssize_t sock_sendpage(struct file *file, struct page *page,
822                              int offset, size_t size, loff_t *ppos, int more)
823 {
824         struct socket *sock;
825         int flags;
826 
827         sock = file->private_data;
828 
829         flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
830         /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
831         flags |= more;
832 
833         return kernel_sendpage(sock, page, offset, size, flags);
834 }
835 
836 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
837                                 struct pipe_inode_info *pipe, size_t len,
838                                 unsigned int flags)
839 {
840         struct socket *sock = file->private_data;
841 
842         if (unlikely(!sock->ops->splice_read))
843                 return -EINVAL;
844 
845         return sock->ops->splice_read(sock, ppos, pipe, len, flags);
846 }
847 
848 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
849 {
850         struct file *file = iocb->ki_filp;
851         struct socket *sock = file->private_data;
852         struct msghdr msg = {.msg_iter = *to};
853         ssize_t res;
854 
855         if (file->f_flags & O_NONBLOCK)
856                 msg.msg_flags = MSG_DONTWAIT;
857 
858         if (iocb->ki_pos != 0)
859                 return -ESPIPE;
860 
861         if (iocb->ki_nbytes == 0)       /* Match SYS5 behaviour */
862                 return 0;
863 
864         res = __sock_recvmsg(iocb, sock, &msg,
865                              iocb->ki_nbytes, msg.msg_flags);
866         *to = msg.msg_iter;
867         return res;
868 }
869 
870 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
871 {
872         struct file *file = iocb->ki_filp;
873         struct socket *sock = file->private_data;
874         struct msghdr msg = {.msg_iter = *from};
875         ssize_t res;
876 
877         if (iocb->ki_pos != 0)
878                 return -ESPIPE;
879 
880         if (file->f_flags & O_NONBLOCK)
881                 msg.msg_flags = MSG_DONTWAIT;
882 
883         if (sock->type == SOCK_SEQPACKET)
884                 msg.msg_flags |= MSG_EOR;
885 
886         res = __sock_sendmsg(iocb, sock, &msg, iocb->ki_nbytes);
887         *from = msg.msg_iter;
888         return res;
889 }
890 
891 /*
892  * Atomic setting of ioctl hooks to avoid race
893  * with module unload.
894  */
895 
896 static DEFINE_MUTEX(br_ioctl_mutex);
897 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
898 
899 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
900 {
901         mutex_lock(&br_ioctl_mutex);
902         br_ioctl_hook = hook;
903         mutex_unlock(&br_ioctl_mutex);
904 }
905 EXPORT_SYMBOL(brioctl_set);
906 
907 static DEFINE_MUTEX(vlan_ioctl_mutex);
908 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
909 
910 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
911 {
912         mutex_lock(&vlan_ioctl_mutex);
913         vlan_ioctl_hook = hook;
914         mutex_unlock(&vlan_ioctl_mutex);
915 }
916 EXPORT_SYMBOL(vlan_ioctl_set);
917 
918 static DEFINE_MUTEX(dlci_ioctl_mutex);
919 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
920 
921 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
922 {
923         mutex_lock(&dlci_ioctl_mutex);
924         dlci_ioctl_hook = hook;
925         mutex_unlock(&dlci_ioctl_mutex);
926 }
927 EXPORT_SYMBOL(dlci_ioctl_set);
928 
929 static long sock_do_ioctl(struct net *net, struct socket *sock,
930                                  unsigned int cmd, unsigned long arg)
931 {
932         int err;
933         void __user *argp = (void __user *)arg;
934 
935         err = sock->ops->ioctl(sock, cmd, arg);
936 
937         /*
938          * If this ioctl is unknown try to hand it down
939          * to the NIC driver.
940          */
941         if (err == -ENOIOCTLCMD)
942                 err = dev_ioctl(net, cmd, argp);
943 
944         return err;
945 }
946 
947 /*
948  *      With an ioctl, arg may well be a user mode pointer, but we don't know
949  *      what to do with it - that's up to the protocol still.
950  */
951 
952 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
953 {
954         struct socket *sock;
955         struct sock *sk;
956         void __user *argp = (void __user *)arg;
957         int pid, err;
958         struct net *net;
959 
960         sock = file->private_data;
961         sk = sock->sk;
962         net = sock_net(sk);
963         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
964                 err = dev_ioctl(net, cmd, argp);
965         } else
966 #ifdef CONFIG_WEXT_CORE
967         if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
968                 err = dev_ioctl(net, cmd, argp);
969         } else
970 #endif
971                 switch (cmd) {
972                 case FIOSETOWN:
973                 case SIOCSPGRP:
974                         err = -EFAULT;
975                         if (get_user(pid, (int __user *)argp))
976                                 break;
977                         f_setown(sock->file, pid, 1);
978                         err = 0;
979                         break;
980                 case FIOGETOWN:
981                 case SIOCGPGRP:
982                         err = put_user(f_getown(sock->file),
983                                        (int __user *)argp);
984                         break;
985                 case SIOCGIFBR:
986                 case SIOCSIFBR:
987                 case SIOCBRADDBR:
988                 case SIOCBRDELBR:
989                         err = -ENOPKG;
990                         if (!br_ioctl_hook)
991                                 request_module("bridge");
992 
993                         mutex_lock(&br_ioctl_mutex);
994                         if (br_ioctl_hook)
995                                 err = br_ioctl_hook(net, cmd, argp);
996                         mutex_unlock(&br_ioctl_mutex);
997                         break;
998                 case SIOCGIFVLAN:
999                 case SIOCSIFVLAN:
1000                         err = -ENOPKG;
1001                         if (!vlan_ioctl_hook)
1002                                 request_module("8021q");
1003 
1004                         mutex_lock(&vlan_ioctl_mutex);
1005                         if (vlan_ioctl_hook)
1006                                 err = vlan_ioctl_hook(net, argp);
1007                         mutex_unlock(&vlan_ioctl_mutex);
1008                         break;
1009                 case SIOCADDDLCI:
1010                 case SIOCDELDLCI:
1011                         err = -ENOPKG;
1012                         if (!dlci_ioctl_hook)
1013                                 request_module("dlci");
1014 
1015                         mutex_lock(&dlci_ioctl_mutex);
1016                         if (dlci_ioctl_hook)
1017                                 err = dlci_ioctl_hook(cmd, argp);
1018                         mutex_unlock(&dlci_ioctl_mutex);
1019                         break;
1020                 default:
1021                         err = sock_do_ioctl(net, sock, cmd, arg);
1022                         break;
1023                 }
1024         return err;
1025 }
1026 
1027 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1028 {
1029         int err;
1030         struct socket *sock = NULL;
1031 
1032         err = security_socket_create(family, type, protocol, 1);
1033         if (err)
1034                 goto out;
1035 
1036         sock = sock_alloc();
1037         if (!sock) {
1038                 err = -ENOMEM;
1039                 goto out;
1040         }
1041 
1042         sock->type = type;
1043         err = security_socket_post_create(sock, family, type, protocol, 1);
1044         if (err)
1045                 goto out_release;
1046 
1047 out:
1048         *res = sock;
1049         return err;
1050 out_release:
1051         sock_release(sock);
1052         sock = NULL;
1053         goto out;
1054 }
1055 EXPORT_SYMBOL(sock_create_lite);
1056 
1057 /* No kernel lock held - perfect */
1058 static unsigned int sock_poll(struct file *file, poll_table *wait)
1059 {
1060         unsigned int busy_flag = 0;
1061         struct socket *sock;
1062 
1063         /*
1064          *      We can't return errors to poll, so it's either yes or no.
1065          */
1066         sock = file->private_data;
1067 
1068         if (sk_can_busy_loop(sock->sk)) {
1069                 /* this socket can poll_ll so tell the system call */
1070                 busy_flag = POLL_BUSY_LOOP;
1071 
1072                 /* once, only if requested by syscall */
1073                 if (wait && (wait->_key & POLL_BUSY_LOOP))
1074                         sk_busy_loop(sock->sk, 1);
1075         }
1076 
1077         return busy_flag | sock->ops->poll(file, sock, wait);
1078 }
1079 
1080 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1081 {
1082         struct socket *sock = file->private_data;
1083 
1084         return sock->ops->mmap(file, sock, vma);
1085 }
1086 
1087 static int sock_close(struct inode *inode, struct file *filp)
1088 {
1089         sock_release(SOCKET_I(inode));
1090         return 0;
1091 }
1092 
1093 /*
1094  *      Update the socket async list
1095  *
1096  *      Fasync_list locking strategy.
1097  *
1098  *      1. fasync_list is modified only under process context socket lock
1099  *         i.e. under semaphore.
1100  *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1101  *         or under socket lock
1102  */
1103 
1104 static int sock_fasync(int fd, struct file *filp, int on)
1105 {
1106         struct socket *sock = filp->private_data;
1107         struct sock *sk = sock->sk;
1108         struct socket_wq *wq;
1109 
1110         if (sk == NULL)
1111                 return -EINVAL;
1112 
1113         lock_sock(sk);
1114         wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1115         fasync_helper(fd, filp, on, &wq->fasync_list);
1116 
1117         if (!wq->fasync_list)
1118                 sock_reset_flag(sk, SOCK_FASYNC);
1119         else
1120                 sock_set_flag(sk, SOCK_FASYNC);
1121 
1122         release_sock(sk);
1123         return 0;
1124 }
1125 
1126 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1127 
1128 int sock_wake_async(struct socket *sock, int how, int band)
1129 {
1130         struct socket_wq *wq;
1131 
1132         if (!sock)
1133                 return -1;
1134         rcu_read_lock();
1135         wq = rcu_dereference(sock->wq);
1136         if (!wq || !wq->fasync_list) {
1137                 rcu_read_unlock();
1138                 return -1;
1139         }
1140         switch (how) {
1141         case SOCK_WAKE_WAITD:
1142                 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1143                         break;
1144                 goto call_kill;
1145         case SOCK_WAKE_SPACE:
1146                 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1147                         break;
1148                 /* fall through */
1149         case SOCK_WAKE_IO:
1150 call_kill:
1151                 kill_fasync(&wq->fasync_list, SIGIO, band);
1152                 break;
1153         case SOCK_WAKE_URG:
1154                 kill_fasync(&wq->fasync_list, SIGURG, band);
1155         }
1156         rcu_read_unlock();
1157         return 0;
1158 }
1159 EXPORT_SYMBOL(sock_wake_async);
1160 
1161 int __sock_create(struct net *net, int family, int type, int protocol,
1162                          struct socket **res, int kern)
1163 {
1164         int err;
1165         struct socket *sock;
1166         const struct net_proto_family *pf;
1167 
1168         /*
1169          *      Check protocol is in range
1170          */
1171         if (family < 0 || family >= NPROTO)
1172                 return -EAFNOSUPPORT;
1173         if (type < 0 || type >= SOCK_MAX)
1174                 return -EINVAL;
1175 
1176         /* Compatibility.
1177 
1178            This uglymoron is moved from INET layer to here to avoid
1179            deadlock in module load.
1180          */
1181         if (family == PF_INET && type == SOCK_PACKET) {
1182                 static int warned;
1183                 if (!warned) {
1184                         warned = 1;
1185                         pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1186                                 current->comm);
1187                 }
1188                 family = PF_PACKET;
1189         }
1190 
1191         err = security_socket_create(family, type, protocol, kern);
1192         if (err)
1193                 return err;
1194 
1195         /*
1196          *      Allocate the socket and allow the family to set things up. if
1197          *      the protocol is 0, the family is instructed to select an appropriate
1198          *      default.
1199          */
1200         sock = sock_alloc();
1201         if (!sock) {
1202                 net_warn_ratelimited("socket: no more sockets\n");
1203                 return -ENFILE; /* Not exactly a match, but its the
1204                                    closest posix thing */
1205         }
1206 
1207         sock->type = type;
1208 
1209 #ifdef CONFIG_MODULES
1210         /* Attempt to load a protocol module if the find failed.
1211          *
1212          * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1213          * requested real, full-featured networking support upon configuration.
1214          * Otherwise module support will break!
1215          */
1216         if (rcu_access_pointer(net_families[family]) == NULL)
1217                 request_module("net-pf-%d", family);
1218 #endif
1219 
1220         rcu_read_lock();
1221         pf = rcu_dereference(net_families[family]);
1222         err = -EAFNOSUPPORT;
1223         if (!pf)
1224                 goto out_release;
1225 
1226         /*
1227          * We will call the ->create function, that possibly is in a loadable
1228          * module, so we have to bump that loadable module refcnt first.
1229          */
1230         if (!try_module_get(pf->owner))
1231                 goto out_release;
1232 
1233         /* Now protected by module ref count */
1234         rcu_read_unlock();
1235 
1236         err = pf->create(net, sock, protocol, kern);
1237         if (err < 0)
1238                 goto out_module_put;
1239 
1240         /*
1241          * Now to bump the refcnt of the [loadable] module that owns this
1242          * socket at sock_release time we decrement its refcnt.
1243          */
1244         if (!try_module_get(sock->ops->owner))
1245                 goto out_module_busy;
1246 
1247         /*
1248          * Now that we're done with the ->create function, the [loadable]
1249          * module can have its refcnt decremented
1250          */
1251         module_put(pf->owner);
1252         err = security_socket_post_create(sock, family, type, protocol, kern);
1253         if (err)
1254                 goto out_sock_release;
1255         *res = sock;
1256 
1257         return 0;
1258 
1259 out_module_busy:
1260         err = -EAFNOSUPPORT;
1261 out_module_put:
1262         sock->ops = NULL;
1263         module_put(pf->owner);
1264 out_sock_release:
1265         sock_release(sock);
1266         return err;
1267 
1268 out_release:
1269         rcu_read_unlock();
1270         goto out_sock_release;
1271 }
1272 EXPORT_SYMBOL(__sock_create);
1273 
1274 int sock_create(int family, int type, int protocol, struct socket **res)
1275 {
1276         return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1277 }
1278 EXPORT_SYMBOL(sock_create);
1279 
1280 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1281 {
1282         return __sock_create(&init_net, family, type, protocol, res, 1);
1283 }
1284 EXPORT_SYMBOL(sock_create_kern);
1285 
1286 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1287 {
1288         int retval;
1289         struct socket *sock;
1290         int flags;
1291 
1292         /* Check the SOCK_* constants for consistency.  */
1293         BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1294         BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1295         BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1296         BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1297 
1298         flags = type & ~SOCK_TYPE_MASK;
1299         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1300                 return -EINVAL;
1301         type &= SOCK_TYPE_MASK;
1302 
1303         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1304                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1305 
1306         retval = sock_create(family, type, protocol, &sock);
1307         if (retval < 0)
1308                 goto out;
1309 
1310         retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1311         if (retval < 0)
1312                 goto out_release;
1313 
1314 out:
1315         /* It may be already another descriptor 8) Not kernel problem. */
1316         return retval;
1317 
1318 out_release:
1319         sock_release(sock);
1320         return retval;
1321 }
1322 
1323 /*
1324  *      Create a pair of connected sockets.
1325  */
1326 
1327 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1328                 int __user *, usockvec)
1329 {
1330         struct socket *sock1, *sock2;
1331         int fd1, fd2, err;
1332         struct file *newfile1, *newfile2;
1333         int flags;
1334 
1335         flags = type & ~SOCK_TYPE_MASK;
1336         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1337                 return -EINVAL;
1338         type &= SOCK_TYPE_MASK;
1339 
1340         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1341                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1342 
1343         /*
1344          * Obtain the first socket and check if the underlying protocol
1345          * supports the socketpair call.
1346          */
1347 
1348         err = sock_create(family, type, protocol, &sock1);
1349         if (err < 0)
1350                 goto out;
1351 
1352         err = sock_create(family, type, protocol, &sock2);
1353         if (err < 0)
1354                 goto out_release_1;
1355 
1356         err = sock1->ops->socketpair(sock1, sock2);
1357         if (err < 0)
1358                 goto out_release_both;
1359 
1360         fd1 = get_unused_fd_flags(flags);
1361         if (unlikely(fd1 < 0)) {
1362                 err = fd1;
1363                 goto out_release_both;
1364         }
1365 
1366         fd2 = get_unused_fd_flags(flags);
1367         if (unlikely(fd2 < 0)) {
1368                 err = fd2;
1369                 goto out_put_unused_1;
1370         }
1371 
1372         newfile1 = sock_alloc_file(sock1, flags, NULL);
1373         if (unlikely(IS_ERR(newfile1))) {
1374                 err = PTR_ERR(newfile1);
1375                 goto out_put_unused_both;
1376         }
1377 
1378         newfile2 = sock_alloc_file(sock2, flags, NULL);
1379         if (IS_ERR(newfile2)) {
1380                 err = PTR_ERR(newfile2);
1381                 goto out_fput_1;
1382         }
1383 
1384         err = put_user(fd1, &usockvec[0]);
1385         if (err)
1386                 goto out_fput_both;
1387 
1388         err = put_user(fd2, &usockvec[1]);
1389         if (err)
1390                 goto out_fput_both;
1391 
1392         audit_fd_pair(fd1, fd2);
1393 
1394         fd_install(fd1, newfile1);
1395         fd_install(fd2, newfile2);
1396         /* fd1 and fd2 may be already another descriptors.
1397          * Not kernel problem.
1398          */
1399 
1400         return 0;
1401 
1402 out_fput_both:
1403         fput(newfile2);
1404         fput(newfile1);
1405         put_unused_fd(fd2);
1406         put_unused_fd(fd1);
1407         goto out;
1408 
1409 out_fput_1:
1410         fput(newfile1);
1411         put_unused_fd(fd2);
1412         put_unused_fd(fd1);
1413         sock_release(sock2);
1414         goto out;
1415 
1416 out_put_unused_both:
1417         put_unused_fd(fd2);
1418 out_put_unused_1:
1419         put_unused_fd(fd1);
1420 out_release_both:
1421         sock_release(sock2);
1422 out_release_1:
1423         sock_release(sock1);
1424 out:
1425         return err;
1426 }
1427 
1428 /*
1429  *      Bind a name to a socket. Nothing much to do here since it's
1430  *      the protocol's responsibility to handle the local address.
1431  *
1432  *      We move the socket address to kernel space before we call
1433  *      the protocol layer (having also checked the address is ok).
1434  */
1435 
1436 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1437 {
1438         struct socket *sock;
1439         struct sockaddr_storage address;
1440         int err, fput_needed;
1441 
1442         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1443         if (sock) {
1444                 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1445                 if (err >= 0) {
1446                         err = security_socket_bind(sock,
1447                                                    (struct sockaddr *)&address,
1448                                                    addrlen);
1449                         if (!err)
1450                                 err = sock->ops->bind(sock,
1451                                                       (struct sockaddr *)
1452                                                       &address, addrlen);
1453                 }
1454                 fput_light(sock->file, fput_needed);
1455         }
1456         return err;
1457 }
1458 
1459 /*
1460  *      Perform a listen. Basically, we allow the protocol to do anything
1461  *      necessary for a listen, and if that works, we mark the socket as
1462  *      ready for listening.
1463  */
1464 
1465 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1466 {
1467         struct socket *sock;
1468         int err, fput_needed;
1469         int somaxconn;
1470 
1471         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1472         if (sock) {
1473                 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1474                 if ((unsigned int)backlog > somaxconn)
1475                         backlog = somaxconn;
1476 
1477                 err = security_socket_listen(sock, backlog);
1478                 if (!err)
1479                         err = sock->ops->listen(sock, backlog);
1480 
1481                 fput_light(sock->file, fput_needed);
1482         }
1483         return err;
1484 }
1485 
1486 /*
1487  *      For accept, we attempt to create a new socket, set up the link
1488  *      with the client, wake up the client, then return the new
1489  *      connected fd. We collect the address of the connector in kernel
1490  *      space and move it to user at the very end. This is unclean because
1491  *      we open the socket then return an error.
1492  *
1493  *      1003.1g adds the ability to recvmsg() to query connection pending
1494  *      status to recvmsg. We need to add that support in a way thats
1495  *      clean when we restucture accept also.
1496  */
1497 
1498 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1499                 int __user *, upeer_addrlen, int, flags)
1500 {
1501         struct socket *sock, *newsock;
1502         struct file *newfile;
1503         int err, len, newfd, fput_needed;
1504         struct sockaddr_storage address;
1505 
1506         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1507                 return -EINVAL;
1508 
1509         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1510                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1511 
1512         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1513         if (!sock)
1514                 goto out;
1515 
1516         err = -ENFILE;
1517         newsock = sock_alloc();
1518         if (!newsock)
1519                 goto out_put;
1520 
1521         newsock->type = sock->type;
1522         newsock->ops = sock->ops;
1523 
1524         /*
1525          * We don't need try_module_get here, as the listening socket (sock)
1526          * has the protocol module (sock->ops->owner) held.
1527          */
1528         __module_get(newsock->ops->owner);
1529 
1530         newfd = get_unused_fd_flags(flags);
1531         if (unlikely(newfd < 0)) {
1532                 err = newfd;
1533                 sock_release(newsock);
1534                 goto out_put;
1535         }
1536         newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1537         if (unlikely(IS_ERR(newfile))) {
1538                 err = PTR_ERR(newfile);
1539                 put_unused_fd(newfd);
1540                 sock_release(newsock);
1541                 goto out_put;
1542         }
1543 
1544         err = security_socket_accept(sock, newsock);
1545         if (err)
1546                 goto out_fd;
1547 
1548         err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1549         if (err < 0)
1550                 goto out_fd;
1551 
1552         if (ccs_socket_post_accept_permission(sock, newsock)) {
1553                 err = -EAGAIN; /* Hope less harmful than -EPERM. */
1554                 goto out_fd;
1555         }
1556         if (upeer_sockaddr) {
1557                 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1558                                           &len, 2) < 0) {
1559                         err = -ECONNABORTED;
1560                         goto out_fd;
1561                 }
1562                 err = move_addr_to_user(&address,
1563                                         len, upeer_sockaddr, upeer_addrlen);
1564                 if (err < 0)
1565                         goto out_fd;
1566         }
1567 
1568         /* File flags are not inherited via accept() unlike another OSes. */
1569 
1570         fd_install(newfd, newfile);
1571         err = newfd;
1572 
1573 out_put:
1574         fput_light(sock->file, fput_needed);
1575 out:
1576         return err;
1577 out_fd:
1578         fput(newfile);
1579         put_unused_fd(newfd);
1580         goto out_put;
1581 }
1582 
1583 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1584                 int __user *, upeer_addrlen)
1585 {
1586         return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1587 }
1588 
1589 /*
1590  *      Attempt to connect to a socket with the server address.  The address
1591  *      is in user space so we verify it is OK and move it to kernel space.
1592  *
1593  *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1594  *      break bindings
1595  *
1596  *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1597  *      other SEQPACKET protocols that take time to connect() as it doesn't
1598  *      include the -EINPROGRESS status for such sockets.
1599  */
1600 
1601 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1602                 int, addrlen)
1603 {
1604         struct socket *sock;
1605         struct sockaddr_storage address;
1606         int err, fput_needed;
1607 
1608         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1609         if (!sock)
1610                 goto out;
1611         err = move_addr_to_kernel(uservaddr, addrlen, &address);
1612         if (err < 0)
1613                 goto out_put;
1614 
1615         err =
1616             security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1617         if (err)
1618                 goto out_put;
1619 
1620         err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1621                                  sock->file->f_flags);
1622 out_put:
1623         fput_light(sock->file, fput_needed);
1624 out:
1625         return err;
1626 }
1627 
1628 /*
1629  *      Get the local address ('name') of a socket object. Move the obtained
1630  *      name to user space.
1631  */
1632 
1633 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1634                 int __user *, usockaddr_len)
1635 {
1636         struct socket *sock;
1637         struct sockaddr_storage address;
1638         int len, err, fput_needed;
1639 
1640         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1641         if (!sock)
1642                 goto out;
1643 
1644         err = security_socket_getsockname(sock);
1645         if (err)
1646                 goto out_put;
1647 
1648         err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1649         if (err)
1650                 goto out_put;
1651         err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1652 
1653 out_put:
1654         fput_light(sock->file, fput_needed);
1655 out:
1656         return err;
1657 }
1658 
1659 /*
1660  *      Get the remote address ('name') of a socket object. Move the obtained
1661  *      name to user space.
1662  */
1663 
1664 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1665                 int __user *, usockaddr_len)
1666 {
1667         struct socket *sock;
1668         struct sockaddr_storage address;
1669         int len, err, fput_needed;
1670 
1671         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1672         if (sock != NULL) {
1673                 err = security_socket_getpeername(sock);
1674                 if (err) {
1675                         fput_light(sock->file, fput_needed);
1676                         return err;
1677                 }
1678 
1679                 err =
1680                     sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1681                                        1);
1682                 if (!err)
1683                         err = move_addr_to_user(&address, len, usockaddr,
1684                                                 usockaddr_len);
1685                 fput_light(sock->file, fput_needed);
1686         }
1687         return err;
1688 }
1689 
1690 /*
1691  *      Send a datagram to a given address. We move the address into kernel
1692  *      space and check the user space data area is readable before invoking
1693  *      the protocol.
1694  */
1695 
1696 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1697                 unsigned int, flags, struct sockaddr __user *, addr,
1698                 int, addr_len)
1699 {
1700         struct socket *sock;
1701         struct sockaddr_storage address;
1702         int err;
1703         struct msghdr msg;
1704         struct iovec iov;
1705         int fput_needed;
1706 
1707         if (len > INT_MAX)
1708                 len = INT_MAX;
1709         if (unlikely(!access_ok(VERIFY_READ, buff, len)))
1710                 return -EFAULT;
1711         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1712         if (!sock)
1713                 goto out;
1714 
1715         iov.iov_base = buff;
1716         iov.iov_len = len;
1717         msg.msg_name = NULL;
1718         iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
1719         msg.msg_control = NULL;
1720         msg.msg_controllen = 0;
1721         msg.msg_namelen = 0;
1722         if (addr) {
1723                 err = move_addr_to_kernel(addr, addr_len, &address);
1724                 if (err < 0)
1725                         goto out_put;
1726                 msg.msg_name = (struct sockaddr *)&address;
1727                 msg.msg_namelen = addr_len;
1728         }
1729         if (sock->file->f_flags & O_NONBLOCK)
1730                 flags |= MSG_DONTWAIT;
1731         msg.msg_flags = flags;
1732         err = sock_sendmsg(sock, &msg, len);
1733 
1734 out_put:
1735         fput_light(sock->file, fput_needed);
1736 out:
1737         return err;
1738 }
1739 
1740 /*
1741  *      Send a datagram down a socket.
1742  */
1743 
1744 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1745                 unsigned int, flags)
1746 {
1747         return sys_sendto(fd, buff, len, flags, NULL, 0);
1748 }
1749 
1750 /*
1751  *      Receive a frame from the socket and optionally record the address of the
1752  *      sender. We verify the buffers are writable and if needed move the
1753  *      sender address from kernel to user space.
1754  */
1755 
1756 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1757                 unsigned int, flags, struct sockaddr __user *, addr,
1758                 int __user *, addr_len)
1759 {
1760         struct socket *sock;
1761         struct iovec iov;
1762         struct msghdr msg;
1763         struct sockaddr_storage address;
1764         int err, err2;
1765         int fput_needed;
1766 
1767         if (size > INT_MAX)
1768                 size = INT_MAX;
1769         if (unlikely(!access_ok(VERIFY_WRITE, ubuf, size)))
1770                 return -EFAULT;
1771         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1772         if (!sock)
1773                 goto out;
1774 
1775         msg.msg_control = NULL;
1776         msg.msg_controllen = 0;
1777         iov.iov_len = size;
1778         iov.iov_base = ubuf;
1779         iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
1780         /* Save some cycles and don't copy the address if not needed */
1781         msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1782         /* We assume all kernel code knows the size of sockaddr_storage */
1783         msg.msg_namelen = 0;
1784         if (sock->file->f_flags & O_NONBLOCK)
1785                 flags |= MSG_DONTWAIT;
1786         err = sock_recvmsg(sock, &msg, size, flags);
1787 
1788         if (err >= 0 && addr != NULL) {
1789                 err2 = move_addr_to_user(&address,
1790                                          msg.msg_namelen, addr, addr_len);
1791                 if (err2 < 0)
1792                         err = err2;
1793         }
1794 
1795         fput_light(sock->file, fput_needed);
1796 out:
1797         return err;
1798 }
1799 
1800 /*
1801  *      Receive a datagram from a socket.
1802  */
1803 
1804 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1805                 unsigned int, flags)
1806 {
1807         return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1808 }
1809 
1810 /*
1811  *      Set a socket option. Because we don't know the option lengths we have
1812  *      to pass the user mode parameter for the protocols to sort out.
1813  */
1814 
1815 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1816                 char __user *, optval, int, optlen)
1817 {
1818         int err, fput_needed;
1819         struct socket *sock;
1820 
1821         if (optlen < 0)
1822                 return -EINVAL;
1823 
1824         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1825         if (sock != NULL) {
1826                 err = security_socket_setsockopt(sock, level, optname);
1827                 if (err)
1828                         goto out_put;
1829 
1830                 if (level == SOL_SOCKET)
1831                         err =
1832                             sock_setsockopt(sock, level, optname, optval,
1833                                             optlen);
1834                 else
1835                         err =
1836                             sock->ops->setsockopt(sock, level, optname, optval,
1837                                                   optlen);
1838 out_put:
1839                 fput_light(sock->file, fput_needed);
1840         }
1841         return err;
1842 }
1843 
1844 /*
1845  *      Get a socket option. Because we don't know the option lengths we have
1846  *      to pass a user mode parameter for the protocols to sort out.
1847  */
1848 
1849 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1850                 char __user *, optval, int __user *, optlen)
1851 {
1852         int err, fput_needed;
1853         struct socket *sock;
1854 
1855         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1856         if (sock != NULL) {
1857                 err = security_socket_getsockopt(sock, level, optname);
1858                 if (err)
1859                         goto out_put;
1860 
1861                 if (level == SOL_SOCKET)
1862                         err =
1863                             sock_getsockopt(sock, level, optname, optval,
1864                                             optlen);
1865                 else
1866                         err =
1867                             sock->ops->getsockopt(sock, level, optname, optval,
1868                                                   optlen);
1869 out_put:
1870                 fput_light(sock->file, fput_needed);
1871         }
1872         return err;
1873 }
1874 
1875 /*
1876  *      Shutdown a socket.
1877  */
1878 
1879 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1880 {
1881         int err, fput_needed;
1882         struct socket *sock;
1883 
1884         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1885         if (sock != NULL) {
1886                 err = security_socket_shutdown(sock, how);
1887                 if (!err)
1888                         err = sock->ops->shutdown(sock, how);
1889                 fput_light(sock->file, fput_needed);
1890         }
1891         return err;
1892 }
1893 
1894 /* A couple of helpful macros for getting the address of the 32/64 bit
1895  * fields which are the same type (int / unsigned) on our platforms.
1896  */
1897 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1898 #define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1899 #define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1900 
1901 struct used_address {
1902         struct sockaddr_storage name;
1903         unsigned int name_len;
1904 };
1905 
1906 static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1907                                      struct user_msghdr __user *umsg,
1908                                      struct sockaddr __user **save_addr,
1909                                      struct iovec **iov)
1910 {
1911         struct sockaddr __user *uaddr;
1912         struct iovec __user *uiov;
1913         size_t nr_segs;
1914         ssize_t err;
1915 
1916         if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1917             __get_user(uaddr, &umsg->msg_name) ||
1918             __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1919             __get_user(uiov, &umsg->msg_iov) ||
1920             __get_user(nr_segs, &umsg->msg_iovlen) ||
1921             __get_user(kmsg->msg_control, &umsg->msg_control) ||
1922             __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1923             __get_user(kmsg->msg_flags, &umsg->msg_flags))
1924                 return -EFAULT;
1925 
1926         if (!uaddr)
1927                 kmsg->msg_namelen = 0;
1928 
1929         if (kmsg->msg_namelen < 0)
1930                 return -EINVAL;
1931 
1932         if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1933                 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1934 
1935         if (save_addr)
1936                 *save_addr = uaddr;
1937 
1938         if (uaddr && kmsg->msg_namelen) {
1939                 if (!save_addr) {
1940                         err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1941                                                   kmsg->msg_name);
1942                         if (err < 0)
1943                                 return err;
1944                 }
1945         } else {
1946                 kmsg->msg_name = NULL;
1947                 kmsg->msg_namelen = 0;
1948         }
1949 
1950         if (nr_segs > UIO_MAXIOV)
1951                 return -EMSGSIZE;
1952 
1953         err = rw_copy_check_uvector(save_addr ? READ : WRITE,
1954                                     uiov, nr_segs,
1955                                     UIO_FASTIOV, *iov, iov);
1956         if (err >= 0)
1957                 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
1958                               *iov, nr_segs, err);
1959         return err;
1960 }
1961 
1962 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1963                          struct msghdr *msg_sys, unsigned int flags,
1964                          struct used_address *used_address)
1965 {
1966         struct compat_msghdr __user *msg_compat =
1967             (struct compat_msghdr __user *)msg;
1968         struct sockaddr_storage address;
1969         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1970         unsigned char ctl[sizeof(struct cmsghdr) + 20]
1971             __attribute__ ((aligned(sizeof(__kernel_size_t))));
1972         /* 20 is size of ipv6_pktinfo */
1973         unsigned char *ctl_buf = ctl;
1974         int ctl_len, total_len;
1975         ssize_t err;
1976 
1977         msg_sys->msg_name = &address;
1978 
1979         if (MSG_CMSG_COMPAT & flags)
1980                 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1981         else
1982                 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1983         if (err < 0)
1984                 goto out_freeiov;
1985         total_len = err;
1986 
1987         err = -ENOBUFS;
1988 
1989         if (msg_sys->msg_controllen > INT_MAX)
1990                 goto out_freeiov;
1991         ctl_len = msg_sys->msg_controllen;
1992         if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1993                 err =
1994                     cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1995                                                      sizeof(ctl));
1996                 if (err)
1997                         goto out_freeiov;
1998                 ctl_buf = msg_sys->msg_control;
1999                 ctl_len = msg_sys->msg_controllen;
2000         } else if (ctl_len) {
2001                 if (ctl_len > sizeof(ctl)) {
2002                         ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2003                         if (ctl_buf == NULL)
2004                                 goto out_freeiov;
2005                 }
2006                 err = -EFAULT;
2007                 /*
2008                  * Careful! Before this, msg_sys->msg_control contains a user pointer.
2009                  * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2010                  * checking falls down on this.
2011                  */
2012                 if (copy_from_user(ctl_buf,
2013                                    (void __user __force *)msg_sys->msg_control,
2014                                    ctl_len))
2015                         goto out_freectl;
2016                 msg_sys->msg_control = ctl_buf;
2017         }
2018         msg_sys->msg_flags = flags;
2019 
2020         if (sock->file->f_flags & O_NONBLOCK)
2021                 msg_sys->msg_flags |= MSG_DONTWAIT;
2022         /*
2023          * If this is sendmmsg() and current destination address is same as
2024          * previously succeeded address, omit asking LSM's decision.
2025          * used_address->name_len is initialized to UINT_MAX so that the first
2026          * destination address never matches.
2027          */
2028         if (used_address && msg_sys->msg_name &&
2029             used_address->name_len == msg_sys->msg_namelen &&
2030             !memcmp(&used_address->name, msg_sys->msg_name,
2031                     used_address->name_len)) {
2032                 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2033                 goto out_freectl;
2034         }
2035         err = sock_sendmsg(sock, msg_sys, total_len);
2036         /*
2037          * If this is sendmmsg() and sending to current destination address was
2038          * successful, remember it.
2039          */
2040         if (used_address && err >= 0) {
2041                 used_address->name_len = msg_sys->msg_namelen;
2042                 if (msg_sys->msg_name)
2043                         memcpy(&used_address->name, msg_sys->msg_name,
2044                                used_address->name_len);
2045         }
2046 
2047 out_freectl:
2048         if (ctl_buf != ctl)
2049                 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2050 out_freeiov:
2051         if (iov != iovstack)
2052                 kfree(iov);
2053         return err;
2054 }
2055 
2056 /*
2057  *      BSD sendmsg interface
2058  */
2059 
2060 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2061 {
2062         int fput_needed, err;
2063         struct msghdr msg_sys;
2064         struct socket *sock;
2065 
2066         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2067         if (!sock)
2068                 goto out;
2069 
2070         err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2071 
2072         fput_light(sock->file, fput_needed);
2073 out:
2074         return err;
2075 }
2076 
2077 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2078 {
2079         if (flags & MSG_CMSG_COMPAT)
2080                 return -EINVAL;
2081         return __sys_sendmsg(fd, msg, flags);
2082 }
2083 
2084 /*
2085  *      Linux sendmmsg interface
2086  */
2087 
2088 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2089                    unsigned int flags)
2090 {
2091         int fput_needed, err, datagrams;
2092         struct socket *sock;
2093         struct mmsghdr __user *entry;
2094         struct compat_mmsghdr __user *compat_entry;
2095         struct msghdr msg_sys;
2096         struct used_address used_address;
2097 
2098         if (vlen > UIO_MAXIOV)
2099                 vlen = UIO_MAXIOV;
2100 
2101         datagrams = 0;
2102 
2103         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2104         if (!sock)
2105                 return err;
2106 
2107         used_address.name_len = UINT_MAX;
2108         entry = mmsg;
2109         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2110         err = 0;
2111 
2112         while (datagrams < vlen) {
2113                 if (MSG_CMSG_COMPAT & flags) {
2114                         err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2115                                              &msg_sys, flags, &used_address);
2116                         if (err < 0)
2117                                 break;
2118                         err = __put_user(err, &compat_entry->msg_len);
2119                         ++compat_entry;
2120                 } else {
2121                         err = ___sys_sendmsg(sock,
2122                                              (struct user_msghdr __user *)entry,
2123                                              &msg_sys, flags, &used_address);
2124                         if (err < 0)
2125                                 break;
2126                         err = put_user(err, &entry->msg_len);
2127                         ++entry;
2128                 }
2129 
2130                 if (err)
2131                         break;
2132                 ++datagrams;
2133         }
2134 
2135         fput_light(sock->file, fput_needed);
2136 
2137         /* We only return an error if no datagrams were able to be sent */
2138         if (datagrams != 0)
2139                 return datagrams;
2140 
2141         return err;
2142 }
2143 
2144 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2145                 unsigned int, vlen, unsigned int, flags)
2146 {
2147         if (flags & MSG_CMSG_COMPAT)
2148                 return -EINVAL;
2149         return __sys_sendmmsg(fd, mmsg, vlen, flags);
2150 }
2151 
2152 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2153                          struct msghdr *msg_sys, unsigned int flags, int nosec)
2154 {
2155         struct compat_msghdr __user *msg_compat =
2156             (struct compat_msghdr __user *)msg;
2157         struct iovec iovstack[UIO_FASTIOV];
2158         struct iovec *iov = iovstack;
2159         unsigned long cmsg_ptr;
2160         int total_len, len;
2161         ssize_t err;
2162 
2163         /* kernel mode address */
2164         struct sockaddr_storage addr;
2165 
2166         /* user mode address pointers */
2167         struct sockaddr __user *uaddr;
2168         int __user *uaddr_len = COMPAT_NAMELEN(msg);
2169 
2170         msg_sys->msg_name = &addr;
2171 
2172         if (MSG_CMSG_COMPAT & flags)
2173                 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2174         else
2175                 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2176         if (err < 0)
2177                 goto out_freeiov;
2178         total_len = err;
2179 
2180         cmsg_ptr = (unsigned long)msg_sys->msg_control;
2181         msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2182 
2183         /* We assume all kernel code knows the size of sockaddr_storage */
2184         msg_sys->msg_namelen = 0;
2185 
2186         if (sock->file->f_flags & O_NONBLOCK)
2187                 flags |= MSG_DONTWAIT;
2188         err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2189                                                           total_len, flags);
2190         if (err < 0)
2191                 goto out_freeiov;
2192         len = err;
2193 
2194         if (uaddr != NULL) {
2195                 err = move_addr_to_user(&addr,
2196                                         msg_sys->msg_namelen, uaddr,
2197                                         uaddr_len);
2198                 if (err < 0)
2199                         goto out_freeiov;
2200         }
2201         err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2202                          COMPAT_FLAGS(msg));
2203         if (err)
2204                 goto out_freeiov;
2205         if (MSG_CMSG_COMPAT & flags)
2206                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2207                                  &msg_compat->msg_controllen);
2208         else
2209                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2210                                  &msg->msg_controllen);
2211         if (err)
2212                 goto out_freeiov;
2213         err = len;
2214 
2215 out_freeiov:
2216         if (iov != iovstack)
2217                 kfree(iov);
2218         return err;
2219 }
2220 
2221 /*
2222  *      BSD recvmsg interface
2223  */
2224 
2225 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2226 {
2227         int fput_needed, err;
2228         struct msghdr msg_sys;
2229         struct socket *sock;
2230 
2231         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2232         if (!sock)
2233                 goto out;
2234 
2235         err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2236 
2237         fput_light(sock->file, fput_needed);
2238 out:
2239         return err;
2240 }
2241 
2242 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2243                 unsigned int, flags)
2244 {
2245         if (flags & MSG_CMSG_COMPAT)
2246                 return -EINVAL;
2247         return __sys_recvmsg(fd, msg, flags);
2248 }
2249 
2250 /*
2251  *     Linux recvmmsg interface
2252  */
2253 
2254 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2255                    unsigned int flags, struct timespec *timeout)
2256 {
2257         int fput_needed, err, datagrams;
2258         struct socket *sock;
2259         struct mmsghdr __user *entry;
2260         struct compat_mmsghdr __user *compat_entry;
2261         struct msghdr msg_sys;
2262         struct timespec end_time;
2263 
2264         if (timeout &&
2265             poll_select_set_timeout(&end_time, timeout->tv_sec,
2266                                     timeout->tv_nsec))
2267                 return -EINVAL;
2268 
2269         datagrams = 0;
2270 
2271         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2272         if (!sock)
2273                 return err;
2274 
2275         err = sock_error(sock->sk);
2276         if (err)
2277                 goto out_put;
2278 
2279         entry = mmsg;
2280         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2281 
2282         while (datagrams < vlen) {
2283                 /*
2284                  * No need to ask LSM for more than the first datagram.
2285                  */
2286                 if (MSG_CMSG_COMPAT & flags) {
2287                         err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2288                                              &msg_sys, flags & ~MSG_WAITFORONE,
2289                                              datagrams);
2290                         if (err < 0)
2291                                 break;
2292                         err = __put_user(err, &compat_entry->msg_len);
2293                         ++compat_entry;
2294                 } else {
2295                         err = ___sys_recvmsg(sock,
2296                                              (struct user_msghdr __user *)entry,
2297                                              &msg_sys, flags & ~MSG_WAITFORONE,
2298                                              datagrams);
2299                         if (err < 0)
2300                                 break;
2301                         err = put_user(err, &entry->msg_len);
2302                         ++entry;
2303                 }
2304 
2305                 if (err)
2306                         break;
2307                 ++datagrams;
2308 
2309                 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2310                 if (flags & MSG_WAITFORONE)
2311                         flags |= MSG_DONTWAIT;
2312 
2313                 if (timeout) {
2314                         ktime_get_ts(timeout);
2315                         *timeout = timespec_sub(end_time, *timeout);
2316                         if (timeout->tv_sec < 0) {
2317                                 timeout->tv_sec = timeout->tv_nsec = 0;
2318                                 break;
2319                         }
2320 
2321                         /* Timeout, return less than vlen datagrams */
2322                         if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2323                                 break;
2324                 }
2325 
2326                 /* Out of band data, return right away */
2327                 if (msg_sys.msg_flags & MSG_OOB)
2328                         break;
2329         }
2330 
2331 out_put:
2332         fput_light(sock->file, fput_needed);
2333 
2334         if (err == 0)
2335                 return datagrams;
2336 
2337         if (datagrams != 0) {
2338                 /*
2339                  * We may return less entries than requested (vlen) if the
2340                  * sock is non block and there aren't enough datagrams...
2341                  */
2342                 if (err != -EAGAIN) {
2343                         /*
2344                          * ... or  if recvmsg returns an error after we
2345                          * received some datagrams, where we record the
2346                          * error to return on the next call or if the
2347                          * app asks about it using getsockopt(SO_ERROR).
2348                          */
2349                         sock->sk->sk_err = -err;
2350                 }
2351 
2352                 return datagrams;
2353         }
2354 
2355         return err;
2356 }
2357 
2358 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2359                 unsigned int, vlen, unsigned int, flags,
2360                 struct timespec __user *, timeout)
2361 {
2362         int datagrams;
2363         struct timespec timeout_sys;
2364 
2365         if (flags & MSG_CMSG_COMPAT)
2366                 return -EINVAL;
2367 
2368         if (!timeout)
2369                 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2370 
2371         if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2372                 return -EFAULT;
2373 
2374         datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2375 
2376         if (datagrams > 0 &&
2377             copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2378                 datagrams = -EFAULT;
2379 
2380         return datagrams;
2381 }
2382 
2383 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2384 /* Argument list sizes for sys_socketcall */
2385 #define AL(x) ((x) * sizeof(unsigned long))
2386 static const unsigned char nargs[21] = {
2387         AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2388         AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2389         AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2390         AL(4), AL(5), AL(4)
2391 };
2392 
2393 #undef AL
2394 
2395 /*
2396  *      System call vectors.
2397  *
2398  *      Argument checking cleaned up. Saved 20% in size.
2399  *  This function doesn't need to set the kernel lock because
2400  *  it is set by the callees.
2401  */
2402 
2403 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2404 {
2405         unsigned long a[AUDITSC_ARGS];
2406         unsigned long a0, a1;
2407         int err;
2408         unsigned int len;
2409 
2410         if (call < 1 || call > SYS_SENDMMSG)
2411                 return -EINVAL;
2412 
2413         len = nargs[call];
2414         if (len > sizeof(a))
2415                 return -EINVAL;
2416 
2417         /* copy_from_user should be SMP safe. */
2418         if (copy_from_user(a, args, len))
2419                 return -EFAULT;
2420 
2421         err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2422         if (err)
2423                 return err;
2424 
2425         a0 = a[0];
2426         a1 = a[1];
2427 
2428         switch (call) {
2429         case SYS_SOCKET:
2430                 err = sys_socket(a0, a1, a[2]);
2431                 break;
2432         case SYS_BIND:
2433                 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2434                 break;
2435         case SYS_CONNECT:
2436                 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2437                 break;
2438         case SYS_LISTEN:
2439                 err = sys_listen(a0, a1);
2440                 break;
2441         case SYS_ACCEPT:
2442                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2443                                   (int __user *)a[2], 0);
2444                 break;
2445         case SYS_GETSOCKNAME:
2446                 err =
2447                     sys_getsockname(a0, (struct sockaddr __user *)a1,
2448                                     (int __user *)a[2]);
2449                 break;
2450         case SYS_GETPEERNAME:
2451                 err =
2452                     sys_getpeername(a0, (struct sockaddr __user *)a1,
2453                                     (int __user *)a[2]);
2454                 break;
2455         case SYS_SOCKETPAIR:
2456                 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2457                 break;
2458         case SYS_SEND:
2459                 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2460                 break;
2461         case SYS_SENDTO:
2462                 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2463                                  (struct sockaddr __user *)a[4], a[5]);
2464                 break;
2465         case SYS_RECV:
2466                 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2467                 break;
2468         case SYS_RECVFROM:
2469                 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2470                                    (struct sockaddr __user *)a[4],
2471                                    (int __user *)a[5]);
2472                 break;
2473         case SYS_SHUTDOWN:
2474                 err = sys_shutdown(a0, a1);
2475                 break;
2476         case SYS_SETSOCKOPT:
2477                 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2478                 break;
2479         case SYS_GETSOCKOPT:
2480                 err =
2481                     sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2482                                    (int __user *)a[4]);
2483                 break;
2484         case SYS_SENDMSG:
2485                 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2486                 break;
2487         case SYS_SENDMMSG:
2488                 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2489                 break;
2490         case SYS_RECVMSG:
2491                 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2492                 break;
2493         case SYS_RECVMMSG:
2494                 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2495                                    (struct timespec __user *)a[4]);
2496                 break;
2497         case SYS_ACCEPT4:
2498                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2499                                   (int __user *)a[2], a[3]);
2500                 break;
2501         default:
2502                 err = -EINVAL;
2503                 break;
2504         }
2505         return err;
2506 }
2507 
2508 #endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2509 
2510 /**
2511  *      sock_register - add a socket protocol handler
2512  *      @ops: description of protocol
2513  *
2514  *      This function is called by a protocol handler that wants to
2515  *      advertise its address family, and have it linked into the
2516  *      socket interface. The value ops->family corresponds to the
2517  *      socket system call protocol family.
2518  */
2519 int sock_register(const struct net_proto_family *ops)
2520 {
2521         int err;
2522 
2523         if (ops->family >= NPROTO) {
2524                 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2525                 return -ENOBUFS;
2526         }
2527 
2528         spin_lock(&net_family_lock);
2529         if (rcu_dereference_protected(net_families[ops->family],
2530                                       lockdep_is_held(&net_family_lock)))
2531                 err = -EEXIST;
2532         else {
2533                 rcu_assign_pointer(net_families[ops->family], ops);
2534                 err = 0;
2535         }
2536         spin_unlock(&net_family_lock);
2537 
2538         pr_info("NET: Registered protocol family %d\n", ops->family);
2539         return err;
2540 }
2541 EXPORT_SYMBOL(sock_register);
2542 
2543 /**
2544  *      sock_unregister - remove a protocol handler
2545  *      @family: protocol family to remove
2546  *
2547  *      This function is called by a protocol handler that wants to
2548  *      remove its address family, and have it unlinked from the
2549  *      new socket creation.
2550  *
2551  *      If protocol handler is a module, then it can use module reference
2552  *      counts to protect against new references. If protocol handler is not
2553  *      a module then it needs to provide its own protection in
2554  *      the ops->create routine.
2555  */
2556 void sock_unregister(int family)
2557 {
2558         BUG_ON(family < 0 || family >= NPROTO);
2559 
2560         spin_lock(&net_family_lock);
2561         RCU_INIT_POINTER(net_families[family], NULL);
2562         spin_unlock(&net_family_lock);
2563 
2564         synchronize_rcu();
2565 
2566         pr_info("NET: Unregistered protocol family %d\n", family);
2567 }
2568 EXPORT_SYMBOL(sock_unregister);
2569 
2570 static int __init sock_init(void)
2571 {
2572         int err;
2573         /*
2574          *      Initialize the network sysctl infrastructure.
2575          */
2576         err = net_sysctl_init();
2577         if (err)
2578                 goto out;
2579 
2580         /*
2581          *      Initialize skbuff SLAB cache
2582          */
2583         skb_init();
2584 
2585         /*
2586          *      Initialize the protocols module.
2587          */
2588 
2589         init_inodecache();
2590 
2591         err = register_filesystem(&sock_fs_type);
2592         if (err)
2593                 goto out_fs;
2594         sock_mnt = kern_mount(&sock_fs_type);
2595         if (IS_ERR(sock_mnt)) {
2596                 err = PTR_ERR(sock_mnt);
2597                 goto out_mount;
2598         }
2599 
2600         /* The real protocol initialization is performed in later initcalls.
2601          */
2602 
2603 #ifdef CONFIG_NETFILTER
2604         err = netfilter_init();
2605         if (err)
2606                 goto out;
2607 #endif
2608 
2609         ptp_classifier_init();
2610 
2611 out:
2612         return err;
2613 
2614 out_mount:
2615         unregister_filesystem(&sock_fs_type);
2616 out_fs:
2617         goto out;
2618 }
2619 
2620 core_initcall(sock_init);       /* early initcall */
2621 
2622 #ifdef CONFIG_PROC_FS
2623 void socket_seq_show(struct seq_file *seq)
2624 {
2625         int cpu;
2626         int counter = 0;
2627 
2628         for_each_possible_cpu(cpu)
2629             counter += per_cpu(sockets_in_use, cpu);
2630 
2631         /* It can be negative, by the way. 8) */
2632         if (counter < 0)
2633                 counter = 0;
2634 
2635         seq_printf(seq, "sockets: used %d\n", counter);
2636 }
2637 #endif                          /* CONFIG_PROC_FS */
2638 
2639 #ifdef CONFIG_COMPAT
2640 static int do_siocgstamp(struct net *net, struct socket *sock,
2641                          unsigned int cmd, void __user *up)
2642 {
2643         mm_segment_t old_fs = get_fs();
2644         struct timeval ktv;
2645         int err;
2646 
2647         set_fs(KERNEL_DS);
2648         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2649         set_fs(old_fs);
2650         if (!err)
2651                 err = compat_put_timeval(&ktv, up);
2652 
2653         return err;
2654 }
2655 
2656 static int do_siocgstampns(struct net *net, struct socket *sock,
2657                            unsigned int cmd, void __user *up)
2658 {
2659         mm_segment_t old_fs = get_fs();
2660         struct timespec kts;
2661         int err;
2662 
2663         set_fs(KERNEL_DS);
2664         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2665         set_fs(old_fs);
2666         if (!err)
2667                 err = compat_put_timespec(&kts, up);
2668 
2669         return err;
2670 }
2671 
2672 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2673 {
2674         struct ifreq __user *uifr;
2675         int err;
2676 
2677         uifr = compat_alloc_user_space(sizeof(struct ifreq));
2678         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2679                 return -EFAULT;
2680 
2681         err = dev_ioctl(net, SIOCGIFNAME, uifr);
2682         if (err)
2683                 return err;
2684 
2685         if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2686                 return -EFAULT;
2687 
2688         return 0;
2689 }
2690 
2691 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2692 {
2693         struct compat_ifconf ifc32;
2694         struct ifconf ifc;
2695         struct ifconf __user *uifc;
2696         struct compat_ifreq __user *ifr32;
2697         struct ifreq __user *ifr;
2698         unsigned int i, j;
2699         int err;
2700 
2701         if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2702                 return -EFAULT;
2703 
2704         memset(&ifc, 0, sizeof(ifc));
2705         if (ifc32.ifcbuf == 0) {
2706                 ifc32.ifc_len = 0;
2707                 ifc.ifc_len = 0;
2708                 ifc.ifc_req = NULL;
2709                 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2710         } else {
2711                 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2712                         sizeof(struct ifreq);
2713                 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2714                 ifc.ifc_len = len;
2715                 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2716                 ifr32 = compat_ptr(ifc32.ifcbuf);
2717                 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2718                         if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2719                                 return -EFAULT;
2720                         ifr++;
2721                         ifr32++;
2722                 }
2723         }
2724         if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2725                 return -EFAULT;
2726 
2727         err = dev_ioctl(net, SIOCGIFCONF, uifc);
2728         if (err)
2729                 return err;
2730 
2731         if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2732                 return -EFAULT;
2733 
2734         ifr = ifc.ifc_req;
2735         ifr32 = compat_ptr(ifc32.ifcbuf);
2736         for (i = 0, j = 0;
2737              i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2738              i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2739                 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2740                         return -EFAULT;
2741                 ifr32++;
2742                 ifr++;
2743         }
2744 
2745         if (ifc32.ifcbuf == 0) {
2746                 /* Translate from 64-bit structure multiple to
2747                  * a 32-bit one.
2748                  */
2749                 i = ifc.ifc_len;
2750                 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2751                 ifc32.ifc_len = i;
2752         } else {
2753                 ifc32.ifc_len = i;
2754         }
2755         if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2756                 return -EFAULT;
2757 
2758         return 0;
2759 }
2760 
2761 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2762 {
2763         struct compat_ethtool_rxnfc __user *compat_rxnfc;
2764         bool convert_in = false, convert_out = false;
2765         size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2766         struct ethtool_rxnfc __user *rxnfc;
2767         struct ifreq __user *ifr;
2768         u32 rule_cnt = 0, actual_rule_cnt;
2769         u32 ethcmd;
2770         u32 data;
2771         int ret;
2772 
2773         if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2774                 return -EFAULT;
2775 
2776         compat_rxnfc = compat_ptr(data);
2777 
2778         if (get_user(ethcmd, &compat_rxnfc->cmd))
2779                 return -EFAULT;
2780 
2781         /* Most ethtool structures are defined without padding.
2782          * Unfortunately struct ethtool_rxnfc is an exception.
2783          */
2784         switch (ethcmd) {
2785         default:
2786                 break;
2787         case ETHTOOL_GRXCLSRLALL:
2788                 /* Buffer size is variable */
2789                 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2790                         return -EFAULT;
2791                 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2792                         return -ENOMEM;
2793                 buf_size += rule_cnt * sizeof(u32);
2794                 /* fall through */
2795         case ETHTOOL_GRXRINGS:
2796         case ETHTOOL_GRXCLSRLCNT:
2797         case ETHTOOL_GRXCLSRULE:
2798         case ETHTOOL_SRXCLSRLINS:
2799                 convert_out = true;
2800                 /* fall through */
2801         case ETHTOOL_SRXCLSRLDEL:
2802                 buf_size += sizeof(struct ethtool_rxnfc);
2803                 convert_in = true;
2804                 break;
2805         }
2806 
2807         ifr = compat_alloc_user_space(buf_size);
2808         rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2809 
2810         if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2811                 return -EFAULT;
2812 
2813         if (put_user(convert_in ? rxnfc : compat_ptr(data),
2814                      &ifr->ifr_ifru.ifru_data))
2815                 return -EFAULT;
2816 
2817         if (convert_in) {
2818                 /* We expect there to be holes between fs.m_ext and
2819                  * fs.ring_cookie and at the end of fs, but nowhere else.
2820                  */
2821                 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2822                              sizeof(compat_rxnfc->fs.m_ext) !=
2823                              offsetof(struct ethtool_rxnfc, fs.m_ext) +
2824                              sizeof(rxnfc->fs.m_ext));
2825                 BUILD_BUG_ON(
2826                         offsetof(struct compat_ethtool_rxnfc, fs.location) -
2827                         offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2828                         offsetof(struct ethtool_rxnfc, fs.location) -
2829                         offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2830 
2831                 if (copy_in_user(rxnfc, compat_rxnfc,
2832                                  (void __user *)(&rxnfc->fs.m_ext + 1) -
2833                                  (void __user *)rxnfc) ||
2834                     copy_in_user(&rxnfc->fs.ring_cookie,
2835                                  &compat_rxnfc->fs.ring_cookie,
2836                                  (void __user *)(&rxnfc->fs.location + 1) -
2837                                  (void __user *)&rxnfc->fs.ring_cookie) ||
2838                     copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2839                                  sizeof(rxnfc->rule_cnt)))
2840                         return -EFAULT;
2841         }
2842 
2843         ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2844         if (ret)
2845                 return ret;
2846 
2847         if (convert_out) {
2848                 if (copy_in_user(compat_rxnfc, rxnfc,
2849                                  (const void __user *)(&rxnfc->fs.m_ext + 1) -
2850                                  (const void __user *)rxnfc) ||
2851                     copy_in_user(&compat_rxnfc->fs.ring_cookie,
2852                                  &rxnfc->fs.ring_cookie,
2853                                  (const void __user *)(&rxnfc->fs.location + 1) -
2854                                  (const void __user *)&rxnfc->fs.ring_cookie) ||
2855                     copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2856                                  sizeof(rxnfc->rule_cnt)))
2857                         return -EFAULT;
2858 
2859                 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2860                         /* As an optimisation, we only copy the actual
2861                          * number of rules that the underlying
2862                          * function returned.  Since Mallory might
2863                          * change the rule count in user memory, we
2864                          * check that it is less than the rule count
2865                          * originally given (as the user buffer size),
2866                          * which has been range-checked.
2867                          */
2868                         if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2869                                 return -EFAULT;
2870                         if (actual_rule_cnt < rule_cnt)
2871                                 rule_cnt = actual_rule_cnt;
2872                         if (copy_in_user(&compat_rxnfc->rule_locs[0],
2873                                          &rxnfc->rule_locs[0],
2874                                          rule_cnt * sizeof(u32)))
2875                                 return -EFAULT;
2876                 }
2877         }
2878 
2879         return 0;
2880 }
2881 
2882 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2883 {
2884         void __user *uptr;
2885         compat_uptr_t uptr32;
2886         struct ifreq __user *uifr;
2887 
2888         uifr = compat_alloc_user_space(sizeof(*uifr));
2889         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2890                 return -EFAULT;
2891 
2892         if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2893                 return -EFAULT;
2894 
2895         uptr = compat_ptr(uptr32);
2896 
2897         if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2898                 return -EFAULT;
2899 
2900         return dev_ioctl(net, SIOCWANDEV, uifr);
2901 }
2902 
2903 static int bond_ioctl(struct net *net, unsigned int cmd,
2904                          struct compat_ifreq __user *ifr32)
2905 {
2906         struct ifreq kifr;
2907         mm_segment_t old_fs;
2908         int err;
2909 
2910         switch (cmd) {
2911         case SIOCBONDENSLAVE:
2912         case SIOCBONDRELEASE:
2913         case SIOCBONDSETHWADDR:
2914         case SIOCBONDCHANGEACTIVE:
2915                 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2916                         return -EFAULT;
2917 
2918                 old_fs = get_fs();
2919                 set_fs(KERNEL_DS);
2920                 err = dev_ioctl(net, cmd,
2921                                 (struct ifreq __user __force *) &kifr);
2922                 set_fs(old_fs);
2923 
2924                 return err;
2925         default:
2926                 return -ENOIOCTLCMD;
2927         }
2928 }
2929 
2930 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2931 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2932                                  struct compat_ifreq __user *u_ifreq32)
2933 {
2934         struct ifreq __user *u_ifreq64;
2935         char tmp_buf[IFNAMSIZ];
2936         void __user *data64;
2937         u32 data32;
2938 
2939         if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2940                            IFNAMSIZ))
2941                 return -EFAULT;
2942         if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2943                 return -EFAULT;
2944         data64 = compat_ptr(data32);
2945 
2946         u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2947 
2948         if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2949                          IFNAMSIZ))
2950                 return -EFAULT;
2951         if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2952                 return -EFAULT;
2953 
2954         return dev_ioctl(net, cmd, u_ifreq64);
2955 }
2956 
2957 static int dev_ifsioc(struct net *net, struct socket *sock,
2958                          unsigned int cmd, struct compat_ifreq __user *uifr32)
2959 {
2960         struct ifreq __user *uifr;
2961         int err;
2962 
2963         uifr = compat_alloc_user_space(sizeof(*uifr));
2964         if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2965                 return -EFAULT;
2966 
2967         err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2968 
2969         if (!err) {
2970                 switch (cmd) {
2971                 case SIOCGIFFLAGS:
2972                 case SIOCGIFMETRIC:
2973                 case SIOCGIFMTU:
2974                 case SIOCGIFMEM:
2975                 case SIOCGIFHWADDR:
2976                 case SIOCGIFINDEX:
2977                 case SIOCGIFADDR:
2978                 case SIOCGIFBRDADDR:
2979                 case SIOCGIFDSTADDR:
2980                 case SIOCGIFNETMASK:
2981                 case SIOCGIFPFLAGS:
2982                 case SIOCGIFTXQLEN:
2983                 case SIOCGMIIPHY:
2984                 case SIOCGMIIREG:
2985                         if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2986                                 err = -EFAULT;
2987                         break;
2988                 }
2989         }
2990         return err;
2991 }
2992 
2993 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2994                         struct compat_ifreq __user *uifr32)
2995 {
2996         struct ifreq ifr;
2997         struct compat_ifmap __user *uifmap32;
2998         mm_segment_t old_fs;
2999         int err;
3000 
3001         uifmap32 = &uifr32->ifr_ifru.ifru_map;
3002         err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3003         err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3004         err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3005         err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3006         err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3007         err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3008         err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3009         if (err)
3010                 return -EFAULT;
3011 
3012         old_fs = get_fs();
3013         set_fs(KERNEL_DS);
3014         err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3015         set_fs(old_fs);
3016 
3017         if (cmd == SIOCGIFMAP && !err) {
3018                 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3019                 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3020                 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3021                 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3022                 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3023                 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3024                 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3025                 if (err)
3026                         err = -EFAULT;
3027         }
3028         return err;
3029 }
3030 
3031 struct rtentry32 {
3032         u32             rt_pad1;
3033         struct sockaddr rt_dst;         /* target address               */
3034         struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3035         struct sockaddr rt_genmask;     /* target network mask (IP)     */
3036         unsigned short  rt_flags;
3037         short           rt_pad2;
3038         u32             rt_pad3;
3039         unsigned char   rt_tos;
3040         unsigned char   rt_class;
3041         short           rt_pad4;
3042         short           rt_metric;      /* +1 for binary compatibility! */
3043         /* char * */ u32 rt_dev;        /* forcing the device at add    */
3044         u32             rt_mtu;         /* per route MTU/Window         */
3045         u32             rt_window;      /* Window clamping              */
3046         unsigned short  rt_irtt;        /* Initial RTT                  */
3047 };
3048 
3049 struct in6_rtmsg32 {
3050         struct in6_addr         rtmsg_dst;
3051         struct in6_addr         rtmsg_src;
3052         struct in6_addr         rtmsg_gateway;
3053         u32                     rtmsg_type;
3054         u16                     rtmsg_dst_len;
3055         u16                     rtmsg_src_len;
3056         u32                     rtmsg_metric;
3057         u32                     rtmsg_info;
3058         u32                     rtmsg_flags;
3059         s32                     rtmsg_ifindex;
3060 };
3061 
3062 static int routing_ioctl(struct net *net, struct socket *sock,
3063                          unsigned int cmd, void __user *argp)
3064 {
3065         int ret;
3066         void *r = NULL;
3067         struct in6_rtmsg r6;
3068         struct rtentry r4;
3069         char devname[16];
3070         u32 rtdev;
3071         mm_segment_t old_fs = get_fs();
3072 
3073         if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3074                 struct in6_rtmsg32 __user *ur6 = argp;
3075                 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3076                         3 * sizeof(struct in6_addr));
3077                 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3078                 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3079                 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3080                 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3081                 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3082                 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3083                 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3084 
3085                 r = (void *) &r6;
3086         } else { /* ipv4 */
3087                 struct rtentry32 __user *ur4 = argp;
3088                 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3089                                         3 * sizeof(struct sockaddr));
3090                 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3091                 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3092                 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3093                 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3094                 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3095                 ret |= get_user(rtdev, &(ur4->rt_dev));
3096                 if (rtdev) {
3097                         ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3098                         r4.rt_dev = (char __user __force *)devname;
3099                         devname[15] = 0;
3100                 } else
3101                         r4.rt_dev = NULL;
3102 
3103                 r = (void *) &r4;
3104         }
3105 
3106         if (ret) {
3107                 ret = -EFAULT;
3108                 goto out;
3109         }
3110 
3111         set_fs(KERNEL_DS);
3112         ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3113         set_fs(old_fs);
3114 
3115 out:
3116         return ret;
3117 }
3118 
3119 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3120  * for some operations; this forces use of the newer bridge-utils that
3121  * use compatible ioctls
3122  */
3123 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3124 {
3125         compat_ulong_t tmp;
3126 
3127         if (get_user(tmp, argp))
3128                 return -EFAULT;
3129         if (tmp == BRCTL_GET_VERSION)
3130                 return BRCTL_VERSION + 1;
3131         return -EINVAL;
3132 }
3133 
3134 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3135                          unsigned int cmd, unsigned long arg)
3136 {
3137         void __user *argp = compat_ptr(arg);
3138         struct sock *sk = sock->sk;
3139         struct net *net = sock_net(sk);
3140 
3141         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3142                 return compat_ifr_data_ioctl(net, cmd, argp);
3143 
3144         switch (cmd) {
3145         case SIOCSIFBR:
3146         case SIOCGIFBR:
3147                 return old_bridge_ioctl(argp);
3148         case SIOCGIFNAME:
3149                 return dev_ifname32(net, argp);
3150         case SIOCGIFCONF:
3151                 return dev_ifconf(net, argp);
3152         case SIOCETHTOOL:
3153                 return ethtool_ioctl(net, argp);
3154         case SIOCWANDEV:
3155                 return compat_siocwandev(net, argp);
3156         case SIOCGIFMAP:
3157         case SIOCSIFMAP:
3158                 return compat_sioc_ifmap(net, cmd, argp);
3159         case SIOCBONDENSLAVE:
3160         case SIOCBONDRELEASE:
3161         case SIOCBONDSETHWADDR:
3162         case SIOCBONDCHANGEACTIVE:
3163                 return bond_ioctl(net, cmd, argp);
3164         case SIOCADDRT:
3165         case SIOCDELRT:
3166                 return routing_ioctl(net, sock, cmd, argp);
3167         case SIOCGSTAMP:
3168                 return do_siocgstamp(net, sock, cmd, argp);
3169         case SIOCGSTAMPNS:
3170                 return do_siocgstampns(net, sock, cmd, argp);
3171         case SIOCBONDSLAVEINFOQUERY:
3172         case SIOCBONDINFOQUERY:
3173         case SIOCSHWTSTAMP:
3174         case SIOCGHWTSTAMP:
3175                 return compat_ifr_data_ioctl(net, cmd, argp);
3176 
3177         case FIOSETOWN:
3178         case SIOCSPGRP:
3179         case FIOGETOWN:
3180         case SIOCGPGRP:
3181         case SIOCBRADDBR:
3182         case SIOCBRDELBR:
3183         case SIOCGIFVLAN:
3184         case SIOCSIFVLAN:
3185         case SIOCADDDLCI:
3186         case SIOCDELDLCI:
3187                 return sock_ioctl(file, cmd, arg);
3188 
3189         case SIOCGIFFLAGS:
3190         case SIOCSIFFLAGS:
3191         case SIOCGIFMETRIC:
3192         case SIOCSIFMETRIC:
3193         case SIOCGIFMTU:
3194         case SIOCSIFMTU:
3195         case SIOCGIFMEM:
3196         case SIOCSIFMEM:
3197         case SIOCGIFHWADDR:
3198         case SIOCSIFHWADDR:
3199         case SIOCADDMULTI:
3200         case SIOCDELMULTI:
3201         case SIOCGIFINDEX:
3202         case SIOCGIFADDR:
3203         case SIOCSIFADDR:
3204         case SIOCSIFHWBROADCAST:
3205         case SIOCDIFADDR:
3206         case SIOCGIFBRDADDR:
3207         case SIOCSIFBRDADDR:
3208         case SIOCGIFDSTADDR:
3209         case SIOCSIFDSTADDR:
3210         case SIOCGIFNETMASK:
3211         case SIOCSIFNETMASK:
3212         case SIOCSIFPFLAGS:
3213         case SIOCGIFPFLAGS:
3214         case SIOCGIFTXQLEN:
3215         case SIOCSIFTXQLEN:
3216         case SIOCBRADDIF:
3217         case SIOCBRDELIF:
3218         case SIOCSIFNAME:
3219         case SIOCGMIIPHY:
3220         case SIOCGMIIREG:
3221         case SIOCSMIIREG:
3222                 return dev_ifsioc(net, sock, cmd, argp);
3223 
3224         case SIOCSARP:
3225         case SIOCGARP:
3226         case SIOCDARP:
3227         case SIOCATMARK:
3228                 return sock_do_ioctl(net, sock, cmd, arg);
3229         }
3230 
3231         return -ENOIOCTLCMD;
3232 }
3233 
3234 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3235                               unsigned long arg)
3236 {
3237         struct socket *sock = file->private_data;
3238         int ret = -ENOIOCTLCMD;
3239         struct sock *sk;
3240         struct net *net;
3241 
3242         sk = sock->sk;
3243         net = sock_net(sk);
3244 
3245         if (sock->ops->compat_ioctl)
3246                 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3247 
3248         if (ret == -ENOIOCTLCMD &&
3249             (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3250                 ret = compat_wext_handle_ioctl(net, cmd, arg);
3251 
3252         if (ret == -ENOIOCTLCMD)
3253                 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3254 
3255         return ret;
3256 }
3257 #endif
3258 
3259 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3260 {
3261         return sock->ops->bind(sock, addr, addrlen);
3262 }
3263 EXPORT_SYMBOL(kernel_bind);
3264 
3265 int kernel_listen(struct socket *sock, int backlog)
3266 {
3267         return sock->ops->listen(sock, backlog);
3268 }
3269 EXPORT_SYMBOL(kernel_listen);
3270 
3271 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3272 {
3273         struct sock *sk = sock->sk;
3274         int err;
3275 
3276         err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3277                                newsock);
3278         if (err < 0)
3279                 goto done;
3280 
3281         err = sock->ops->accept(sock, *newsock, flags);
3282         if (err < 0) {
3283                 sock_release(*newsock);
3284                 *newsock = NULL;
3285                 goto done;
3286         }
3287 
3288         (*newsock)->ops = sock->ops;
3289         __module_get((*newsock)->ops->owner);
3290 
3291 done:
3292         return err;
3293 }
3294 EXPORT_SYMBOL(kernel_accept);
3295 
3296 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3297                    int flags)
3298 {
3299         return sock->ops->connect(sock, addr, addrlen, flags);
3300 }
3301 EXPORT_SYMBOL(kernel_connect);
3302 
3303 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3304                          int *addrlen)
3305 {
3306         return sock->ops->getname(sock, addr, addrlen, 0);
3307 }
3308 EXPORT_SYMBOL(kernel_getsockname);
3309 
3310 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3311                          int *addrlen)
3312 {
3313         return sock->ops->getname(sock, addr, addrlen, 1);
3314 }
3315 EXPORT_SYMBOL(kernel_getpeername);
3316 
3317 int kernel_getsockopt(struct socket *sock, int level, int optname,
3318                         char *optval, int *optlen)
3319 {
3320         mm_segment_t oldfs = get_fs();
3321         char __user *uoptval;
3322         int __user *uoptlen;
3323         int err;
3324 
3325         uoptval = (char __user __force *) optval;
3326         uoptlen = (int __user __force *) optlen;
3327 
3328         set_fs(KERNEL_DS);
3329         if (level == SOL_SOCKET)
3330                 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3331         else
3332                 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3333                                             uoptlen);
3334         set_fs(oldfs);
3335         return err;
3336 }
3337 EXPORT_SYMBOL(kernel_getsockopt);
3338 
3339 int kernel_setsockopt(struct socket *sock, int level, int optname,
3340                         char *optval, unsigned int optlen)
3341 {
3342         mm_segment_t oldfs = get_fs();
3343         char __user *uoptval;
3344         int err;
3345 
3346         uoptval = (char __user __force *) optval;
3347 
3348         set_fs(KERNEL_DS);
3349         if (level == SOL_SOCKET)
3350                 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3351         else
3352                 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3353                                             optlen);
3354         set_fs(oldfs);
3355         return err;
3356 }
3357 EXPORT_SYMBOL(kernel_setsockopt);
3358 
3359 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3360                     size_t size, int flags)
3361 {
3362         if (sock->ops->sendpage)
3363                 return sock->ops->sendpage(sock, page, offset, size, flags);
3364 
3365         return sock_no_sendpage(sock, page, offset, size, flags);
3366 }
3367 EXPORT_SYMBOL(kernel_sendpage);
3368 
3369 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3370 {
3371         mm_segment_t oldfs = get_fs();
3372         int err;
3373 
3374         set_fs(KERNEL_DS);
3375         err = sock->ops->ioctl(sock, cmd, arg);
3376         set_fs(oldfs);
3377 
3378         return err;
3379 }
3380 EXPORT_SYMBOL(kernel_sock_ioctl);
3381 
3382 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3383 {
3384         return sock->ops->shutdown(sock, how);
3385 }
3386 EXPORT_SYMBOL(kernel_sock_shutdown);
3387 

~ [ 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