root/net/socket.c

DEFINITIONS

1. move_addr_to_kernel
2. move_addr_to_user
3. sock_alloc_inode
4. sock_free_inode
5. init_once
6. init_inodecache
7. sockfs_dname
8. sockfs_xattr_get
9. sockfs_security_xattr_set
10. sockfs_init_fs_context
11. sock_alloc_file
12. sock_map_fd
13. sock_from_file
14. sockfd_lookup
15. sockfd_lookup_light
16. sockfs_listxattr
17. sockfs_setattr
18. sock_alloc
19. __sock_release
20. sock_release
21. __sock_tx_timestamp
22. sock_sendmsg_nosec
23. sock_sendmsg
24. kernel_sendmsg
25. kernel_sendmsg_locked
26. skb_is_err_queue
27. skb_is_swtx_tstamp
28. put_ts_pktinfo
29. __sock_recv_timestamp
30. __sock_recv_wifi_status
31. sock_recv_drops
32. __sock_recv_ts_and_drops
33. sock_recvmsg_nosec
34. sock_recvmsg
35. kernel_recvmsg
36. sock_sendpage
37. sock_splice_read
38. sock_read_iter
39. sock_write_iter
40. brioctl_set
41. vlan_ioctl_set
42. dlci_ioctl_set
43. sock_do_ioctl
44. get_net_ns
45. sock_ioctl
46. sock_create_lite
47. sock_poll
48. sock_mmap
49. sock_close
50. sock_fasync
51. sock_wake_async
52. __sock_create
53. sock_create
54. sock_create_kern
55. __sys_socket
56. SYSCALL_DEFINE3
57. __sys_socketpair
58. SYSCALL_DEFINE4
59. __sys_bind
60. SYSCALL_DEFINE3
61. __sys_listen
62. SYSCALL_DEFINE2
63. __sys_accept4
64. SYSCALL_DEFINE4
65. SYSCALL_DEFINE3
66. __sys_connect
67. SYSCALL_DEFINE3
68. __sys_getsockname
69. SYSCALL_DEFINE3
70. __sys_getpeername
71. SYSCALL_DEFINE3
72. __sys_sendto
73. SYSCALL_DEFINE6
74. SYSCALL_DEFINE4
75. __sys_recvfrom
76. SYSCALL_DEFINE6
77. SYSCALL_DEFINE4
78. __sys_setsockopt
79. SYSCALL_DEFINE5
80. __sys_getsockopt
81. SYSCALL_DEFINE5
82. __sys_shutdown
83. SYSCALL_DEFINE2
84. copy_msghdr_from_user
85. ____sys_sendmsg
86. sendmsg_copy_msghdr
87. ___sys_sendmsg
88. __sys_sendmsg_sock
89. __sys_sendmsg
90. SYSCALL_DEFINE3
91. __sys_sendmmsg
92. SYSCALL_DEFINE4
93. recvmsg_copy_msghdr
94. ____sys_recvmsg
95. ___sys_recvmsg
96. __sys_recvmsg_sock
97. __sys_recvmsg
98. SYSCALL_DEFINE3
99. do_recvmmsg
100. __sys_recvmmsg
101. SYSCALL_DEFINE5
102. SYSCALL_DEFINE5
103. SYSCALL_DEFINE2
104. sock_register
105. sock_unregister
106. sock_is_registered
107. sock_init
108. socket_seq_show
109. compat_dev_ifconf
110. ethtool_ioctl
111. compat_siocwandev
112. compat_ifr_data_ioctl
113. compat_ifreq_ioctl
114. compat_sioc_ifmap
115. routing_ioctl
116. old_bridge_ioctl
117. compat_sock_ioctl_trans
118. compat_sock_ioctl
119. kernel_bind
120. kernel_listen
121. kernel_accept
122. kernel_connect
123. kernel_getsockname
124. kernel_getpeername
125. kernel_getsockopt
126. kernel_setsockopt
127. kernel_sendpage
128. kernel_sendpage_locked
129. kernel_sock_shutdown
130. kernel_sock_ip_overhead

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * NET          An implementation of the SOCKET network access protocol.
   4  *
   5  * Version:     @(#)socket.c    1.1.93  18/02/95
   6  *
   7  * Authors:     Orest Zborowski, <obz@Kodak.COM>
   8  *              Ross Biro
   9  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  10  *
  11  * Fixes:
  12  *              Anonymous       :       NOTSOCK/BADF cleanup. Error fix in
  13  *                                      shutdown()
  14  *              Alan Cox        :       verify_area() fixes
  15  *              Alan Cox        :       Removed DDI
  16  *              Jonathan Kamens :       SOCK_DGRAM reconnect bug
  17  *              Alan Cox        :       Moved a load of checks to the very
  18  *                                      top level.
  19  *              Alan Cox        :       Move address structures to/from user
  20  *                                      mode above the protocol layers.
  21  *              Rob Janssen     :       Allow 0 length sends.
  22  *              Alan Cox        :       Asynchronous I/O support (cribbed from the
  23  *                                      tty drivers).
  24  *              Niibe Yutaka    :       Asynchronous I/O for writes (4.4BSD style)
  25  *              Jeff Uphoff     :       Made max number of sockets command-line
  26  *                                      configurable.
  27  *              Matti Aarnio    :       Made the number of sockets dynamic,
  28  *                                      to be allocated when needed, and mr.
  29  *                                      Uphoff's max is used as max to be
  30  *                                      allowed to allocate.
  31  *              Linus           :       Argh. removed all the socket allocation
  32  *                                      altogether: it's in the inode now.
  33  *              Alan Cox        :       Made sock_alloc()/sock_release() public
  34  *                                      for NetROM and future kernel nfsd type
  35  *                                      stuff.
  36  *              Alan Cox        :       sendmsg/recvmsg basics.
  37  *              Tom Dyas        :       Export net symbols.
  38  *              Marcin Dalecki  :       Fixed problems with CONFIG_NET="n".
  39  *              Alan Cox        :       Added thread locking to sys_* calls
  40  *                                      for sockets. May have errors at the
  41  *                                      moment.
  42  *              Kevin Buhr      :       Fixed the dumb errors in the above.
  43  *              Andi Kleen      :       Some small cleanups, optimizations,
  44  *                                      and fixed a copy_from_user() bug.
  45  *              Tigran Aivazian :       sys_send(args) calls sys_sendto(args, NULL, 0)
  46  *              Tigran Aivazian :       Made listen(2) backlog sanity checks
  47  *                                      protocol-independent
  48  *
  49  *      This module is effectively the top level interface to the BSD socket
  50  *      paradigm.
  51  *
  52  *      Based upon Swansea University Computer Society NET3.039
  53  */
  54 
  55 #include <linux/mm.h>
  56 #include <linux/socket.h>
  57 #include <linux/file.h>
  58 #include <linux/net.h>
  59 #include <linux/interrupt.h>
  60 #include <linux/thread_info.h>
  61 #include <linux/rcupdate.h>
  62 #include <linux/netdevice.h>
  63 #include <linux/proc_fs.h>
  64 #include <linux/seq_file.h>
  65 #include <linux/mutex.h>
  66 #include <linux/if_bridge.h>
  67 #include <linux/if_frad.h>
  68 #include <linux/if_vlan.h>
  69 #include <linux/ptp_classify.h>
  70 #include <linux/init.h>
  71 #include <linux/poll.h>
  72 #include <linux/cache.h>
  73 #include <linux/module.h>
  74 #include <linux/highmem.h>
  75 #include <linux/mount.h>
  76 #include <linux/pseudo_fs.h>
  77 #include <linux/security.h>
  78 #include <linux/syscalls.h>
  79 #include <linux/compat.h>
  80 #include <linux/kmod.h>
  81 #include <linux/audit.h>
  82 #include <linux/wireless.h>
  83 #include <linux/nsproxy.h>
  84 #include <linux/magic.h>
  85 #include <linux/slab.h>
  86 #include <linux/xattr.h>
  87 #include <linux/nospec.h>
  88 #include <linux/indirect_call_wrapper.h>
  89 
  90 #include <linux/uaccess.h>
  91 #include <asm/unistd.h>
  92 
  93 #include <net/compat.h>
  94 #include <net/wext.h>
  95 #include <net/cls_cgroup.h>
  96 
  97 #include <net/sock.h>
  98 #include <linux/netfilter.h>
  99 
 100 #include <linux/if_tun.h>
 101 #include <linux/ipv6_route.h>
 102 #include <linux/route.h>
 103 #include <linux/sockios.h>
 104 #include <net/busy_poll.h>
 105 #include <linux/errqueue.h>
 106 
 107 #ifdef CONFIG_NET_RX_BUSY_POLL
 108 unsigned int sysctl_net_busy_read __read_mostly;
 109 unsigned int sysctl_net_busy_poll __read_mostly;
 110 #endif
 111 
 112 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
 113 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
 114 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
 115 
 116 static int sock_close(struct inode *inode, struct file *file);
 117 static __poll_t sock_poll(struct file *file,
 118                               struct poll_table_struct *wait);
 119 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 120 #ifdef CONFIG_COMPAT
 121 static long compat_sock_ioctl(struct file *file,
 122                               unsigned int cmd, unsigned long arg);
 123 #endif
 124 static int sock_fasync(int fd, struct file *filp, int on);
 125 static ssize_t sock_sendpage(struct file *file, struct page *page,
 126                              int offset, size_t size, loff_t *ppos, int more);
 127 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 128                                 struct pipe_inode_info *pipe, size_t len,
 129                                 unsigned int flags);
 130 
 131 /*
 132  *      Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
 133  *      in the operation structures but are done directly via the socketcall() multiplexor.
 134  */
 135 
 136 static const struct file_operations socket_file_ops = {
 137         .owner =        THIS_MODULE,
 138         .llseek =       no_llseek,
 139         .read_iter =    sock_read_iter,
 140         .write_iter =   sock_write_iter,
 141         .poll =         sock_poll,
 142         .unlocked_ioctl = sock_ioctl,
 143 #ifdef CONFIG_COMPAT
 144         .compat_ioctl = compat_sock_ioctl,
 145 #endif
 146         .mmap =         sock_mmap,
 147         .release =      sock_close,
 148         .fasync =       sock_fasync,
 149         .sendpage =     sock_sendpage,
 150         .splice_write = generic_splice_sendpage,
 151         .splice_read =  sock_splice_read,
 152 };
 153 
 154 /*
 155  *      The protocol list. Each protocol is registered in here.
 156  */
 157 
 158 static DEFINE_SPINLOCK(net_family_lock);
 159 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
 160 
 161 /*
 162  * Support routines.
 163  * Move socket addresses back and forth across the kernel/user
 164  * divide and look after the messy bits.
 165  */
 166 
 167 /**
 168  *      move_addr_to_kernel     -       copy a socket address into kernel space
 169  *      @uaddr: Address in user space
 170  *      @kaddr: Address in kernel space
 171  *      @ulen: Length in user space
 172  *
 173  *      The address is copied into kernel space. If the provided address is
 174  *      too long an error code of -EINVAL is returned. If the copy gives
 175  *      invalid addresses -EFAULT is returned. On a success 0 is returned.
 176  */
 177 
 178 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
 179 {
 180         if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
 181                 return -EINVAL;
 182         if (ulen == 0)
 183                 return 0;
 184         if (copy_from_user(kaddr, uaddr, ulen))
 185                 return -EFAULT;
 186         return audit_sockaddr(ulen, kaddr);
 187 }
 188 
 189 /**
 190  *      move_addr_to_user       -       copy an address to user space
 191  *      @kaddr: kernel space address
 192  *      @klen: length of address in kernel
 193  *      @uaddr: user space address
 194  *      @ulen: pointer to user length field
 195  *
 196  *      The value pointed to by ulen on entry is the buffer length available.
 197  *      This is overwritten with the buffer space used. -EINVAL is returned
 198  *      if an overlong buffer is specified or a negative buffer size. -EFAULT
 199  *      is returned if either the buffer or the length field are not
 200  *      accessible.
 201  *      After copying the data up to the limit the user specifies, the true
 202  *      length of the data is written over the length limit the user
 203  *      specified. Zero is returned for a success.
 204  */
 205 
 206 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
 207                              void __user *uaddr, int __user *ulen)
 208 {
 209         int err;
 210         int len;
 211 
 212         BUG_ON(klen > sizeof(struct sockaddr_storage));
 213         err = get_user(len, ulen);
 214         if (err)
 215                 return err;
 216         if (len > klen)
 217                 len = klen;
 218         if (len < 0)
 219                 return -EINVAL;
 220         if (len) {
 221                 if (audit_sockaddr(klen, kaddr))
 222                         return -ENOMEM;
 223                 if (copy_to_user(uaddr, kaddr, len))
 224                         return -EFAULT;
 225         }
 226         /*
 227          *      "fromlen shall refer to the value before truncation.."
 228          *                      1003.1g
 229          */
 230         return __put_user(klen, ulen);
 231 }
 232 
 233 static struct kmem_cache *sock_inode_cachep __ro_after_init;
 234 
 235 static struct inode *sock_alloc_inode(struct super_block *sb)
 236 {
 237         struct socket_alloc *ei;
 238 
 239         ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 240         if (!ei)
 241                 return NULL;
 242         init_waitqueue_head(&ei->socket.wq.wait);
 243         ei->socket.wq.fasync_list = NULL;
 244         ei->socket.wq.flags = 0;
 245 
 246         ei->socket.state = SS_UNCONNECTED;
 247         ei->socket.flags = 0;
 248         ei->socket.ops = NULL;
 249         ei->socket.sk = NULL;
 250         ei->socket.file = NULL;
 251 
 252         return &ei->vfs_inode;
 253 }
 254 
 255 static void sock_free_inode(struct inode *inode)
 256 {
 257         struct socket_alloc *ei;
 258 
 259         ei = container_of(inode, struct socket_alloc, vfs_inode);
 260         kmem_cache_free(sock_inode_cachep, ei);
 261 }
 262 
 263 static void init_once(void *foo)
 264 {
 265         struct socket_alloc *ei = (struct socket_alloc *)foo;
 266 
 267         inode_init_once(&ei->vfs_inode);
 268 }
 269 
 270 static void init_inodecache(void)
 271 {
 272         sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 273                                               sizeof(struct socket_alloc),
 274                                               0,
 275                                               (SLAB_HWCACHE_ALIGN |
 276                                                SLAB_RECLAIM_ACCOUNT |
 277                                                SLAB_MEM_SPREAD | SLAB_ACCOUNT),
 278                                               init_once);
 279         BUG_ON(sock_inode_cachep == NULL);
 280 }
 281 
 282 static const struct super_operations sockfs_ops = {
 283         .alloc_inode    = sock_alloc_inode,
 284         .free_inode     = sock_free_inode,
 285         .statfs         = simple_statfs,
 286 };
 287 
 288 /*
 289  * sockfs_dname() is called from d_path().
 290  */
 291 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 292 {
 293         return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 294                                 d_inode(dentry)->i_ino);
 295 }
 296 
 297 static const struct dentry_operations sockfs_dentry_operations = {
 298         .d_dname  = sockfs_dname,
 299 };
 300 
 301 static int sockfs_xattr_get(const struct xattr_handler *handler,
 302                             struct dentry *dentry, struct inode *inode,
 303                             const char *suffix, void *value, size_t size)
 304 {
 305         if (value) {
 306                 if (dentry->d_name.len + 1 > size)
 307                         return -ERANGE;
 308                 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
 309         }
 310         return dentry->d_name.len + 1;
 311 }
 312 
 313 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
 314 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
 315 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
 316 
 317 static const struct xattr_handler sockfs_xattr_handler = {
 318         .name = XATTR_NAME_SOCKPROTONAME,
 319         .get = sockfs_xattr_get,
 320 };
 321 
 322 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
 323                                      struct dentry *dentry, struct inode *inode,
 324                                      const char *suffix, const void *value,
 325                                      size_t size, int flags)
 326 {
 327         /* Handled by LSM. */
 328         return -EAGAIN;
 329 }
 330 
 331 static const struct xattr_handler sockfs_security_xattr_handler = {
 332         .prefix = XATTR_SECURITY_PREFIX,
 333         .set = sockfs_security_xattr_set,
 334 };
 335 
 336 static const struct xattr_handler *sockfs_xattr_handlers[] = {
 337         &sockfs_xattr_handler,
 338         &sockfs_security_xattr_handler,
 339         NULL
 340 };
 341 
 342 static int sockfs_init_fs_context(struct fs_context *fc)
 343 {
 344         struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
 345         if (!ctx)
 346                 return -ENOMEM;
 347         ctx->ops = &sockfs_ops;
 348         ctx->dops = &sockfs_dentry_operations;
 349         ctx->xattr = sockfs_xattr_handlers;
 350         return 0;
 351 }
 352 
 353 static struct vfsmount *sock_mnt __read_mostly;
 354 
 355 static struct file_system_type sock_fs_type = {
 356         .name =         "sockfs",
 357         .init_fs_context = sockfs_init_fs_context,
 358         .kill_sb =      kill_anon_super,
 359 };
 360 
 361 /*
 362  *      Obtains the first available file descriptor and sets it up for use.
 363  *
 364  *      These functions create file structures and maps them to fd space
 365  *      of the current process. On success it returns file descriptor
 366  *      and file struct implicitly stored in sock->file.
 367  *      Note that another thread may close file descriptor before we return
 368  *      from this function. We use the fact that now we do not refer
 369  *      to socket after mapping. If one day we will need it, this
 370  *      function will increment ref. count on file by 1.
 371  *
 372  *      In any case returned fd MAY BE not valid!
 373  *      This race condition is unavoidable
 374  *      with shared fd spaces, we cannot solve it inside kernel,
 375  *      but we take care of internal coherence yet.
 376  */
 377 
 378 /**
 379  *      sock_alloc_file - Bind a &socket to a &file
 380  *      @sock: socket
 381  *      @flags: file status flags
 382  *      @dname: protocol name
 383  *
 384  *      Returns the &file bound with @sock, implicitly storing it
 385  *      in sock->file. If dname is %NULL, sets to "".
 386  *      On failure the return is a ERR pointer (see linux/err.h).
 387  *      This function uses GFP_KERNEL internally.
 388  */
 389 
 390 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
 391 {
 392         struct file *file;
 393 
 394         if (!dname)
 395                 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
 396 
 397         file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
 398                                 O_RDWR | (flags & O_NONBLOCK),
 399                                 &socket_file_ops);
 400         if (IS_ERR(file)) {
 401                 sock_release(sock);
 402                 return file;
 403         }
 404 
 405         sock->file = file;
 406         file->private_data = sock;
 407         return file;
 408 }
 409 EXPORT_SYMBOL(sock_alloc_file);
 410 
 411 static int sock_map_fd(struct socket *sock, int flags)
 412 {
 413         struct file *newfile;
 414         int fd = get_unused_fd_flags(flags);
 415         if (unlikely(fd < 0)) {
 416                 sock_release(sock);
 417                 return fd;
 418         }
 419 
 420         newfile = sock_alloc_file(sock, flags, NULL);
 421         if (!IS_ERR(newfile)) {
 422                 fd_install(fd, newfile);
 423                 return fd;
 424         }
 425 
 426         put_unused_fd(fd);
 427         return PTR_ERR(newfile);
 428 }
 429 
 430 /**
 431  *      sock_from_file - Return the &socket bounded to @file.
 432  *      @file: file
 433  *      @err: pointer to an error code return
 434  *
 435  *      On failure returns %NULL and assigns -ENOTSOCK to @err.
 436  */
 437 
 438 struct socket *sock_from_file(struct file *file, int *err)
 439 {
 440         if (file->f_op == &socket_file_ops)
 441                 return file->private_data;      /* set in sock_map_fd */
 442 
 443         *err = -ENOTSOCK;
 444         return NULL;
 445 }
 446 EXPORT_SYMBOL(sock_from_file);
 447 
 448 /**
 449  *      sockfd_lookup - Go from a file number to its socket slot
 450  *      @fd: file handle
 451  *      @err: pointer to an error code return
 452  *
 453  *      The file handle passed in is locked and the socket it is bound
 454  *      to is returned. If an error occurs the err pointer is overwritten
 455  *      with a negative errno code and NULL is returned. The function checks
 456  *      for both invalid handles and passing a handle which is not a socket.
 457  *
 458  *      On a success the socket object pointer is returned.
 459  */
 460 
 461 struct socket *sockfd_lookup(int fd, int *err)
 462 {
 463         struct file *file;
 464         struct socket *sock;
 465 
 466         file = fget(fd);
 467         if (!file) {
 468                 *err = -EBADF;
 469                 return NULL;
 470         }
 471 
 472         sock = sock_from_file(file, err);
 473         if (!sock)
 474                 fput(file);
 475         return sock;
 476 }
 477 EXPORT_SYMBOL(sockfd_lookup);
 478 
 479 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 480 {
 481         struct fd f = fdget(fd);
 482         struct socket *sock;
 483 
 484         *err = -EBADF;
 485         if (f.file) {
 486                 sock = sock_from_file(f.file, err);
 487                 if (likely(sock)) {
 488                         *fput_needed = f.flags;
 489                         return sock;
 490                 }
 491                 fdput(f);
 492         }
 493         return NULL;
 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(d_inode(dentry), 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 int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
 525 {
 526         int err = simple_setattr(dentry, iattr);
 527 
 528         if (!err && (iattr->ia_valid & ATTR_UID)) {
 529                 struct socket *sock = SOCKET_I(d_inode(dentry));
 530 
 531                 if (sock->sk)
 532                         sock->sk->sk_uid = iattr->ia_uid;
 533                 else
 534                         err = -ENOENT;
 535         }
 536 
 537         return err;
 538 }
 539 
 540 static const struct inode_operations sockfs_inode_ops = {
 541         .listxattr = sockfs_listxattr,
 542         .setattr = sockfs_setattr,
 543 };
 544 
 545 /**
 546  *      sock_alloc - allocate a socket
 547  *
 548  *      Allocate a new inode and socket object. The two are bound together
 549  *      and initialised. The socket is then returned. If we are out of inodes
 550  *      NULL is returned. This functions uses GFP_KERNEL internally.
 551  */
 552 
 553 struct socket *sock_alloc(void)
 554 {
 555         struct inode *inode;
 556         struct socket *sock;
 557 
 558         inode = new_inode_pseudo(sock_mnt->mnt_sb);
 559         if (!inode)
 560                 return NULL;
 561 
 562         sock = SOCKET_I(inode);
 563 
 564         inode->i_ino = get_next_ino();
 565         inode->i_mode = S_IFSOCK | S_IRWXUGO;
 566         inode->i_uid = current_fsuid();
 567         inode->i_gid = current_fsgid();
 568         inode->i_op = &sockfs_inode_ops;
 569 
 570         return sock;
 571 }
 572 EXPORT_SYMBOL(sock_alloc);
 573 
 574 /**
 575  *      sock_release - close a socket
 576  *      @sock: socket to close
 577  *
 578  *      The socket is released from the protocol stack if it has a release
 579  *      callback, and the inode is then released if the socket is bound to
 580  *      an inode not a file.
 581  */
 582 
 583 static void __sock_release(struct socket *sock, struct inode *inode)
 584 {
 585         if (sock->ops) {
 586                 struct module *owner = sock->ops->owner;
 587 
 588                 if (inode)
 589                         inode_lock(inode);
 590                 sock->ops->release(sock);
 591                 sock->sk = NULL;
 592                 if (inode)
 593                         inode_unlock(inode);
 594                 sock->ops = NULL;
 595                 module_put(owner);
 596         }
 597 
 598         if (sock->wq.fasync_list)
 599                 pr_err("%s: fasync list not empty!\n", __func__);
 600 
 601         if (!sock->file) {
 602                 iput(SOCK_INODE(sock));
 603                 return;
 604         }
 605         sock->file = NULL;
 606 }
 607 
 608 void sock_release(struct socket *sock)
 609 {
 610         __sock_release(sock, NULL);
 611 }
 612 EXPORT_SYMBOL(sock_release);
 613 
 614 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
 615 {
 616         u8 flags = *tx_flags;
 617 
 618         if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
 619                 flags |= SKBTX_HW_TSTAMP;
 620 
 621         if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
 622                 flags |= SKBTX_SW_TSTAMP;
 623 
 624         if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
 625                 flags |= SKBTX_SCHED_TSTAMP;
 626 
 627         *tx_flags = flags;
 628 }
 629 EXPORT_SYMBOL(__sock_tx_timestamp);
 630 
 631 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
 632                                            size_t));
 633 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
 634                                             size_t));
 635 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
 636 {
 637         int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
 638                                      inet_sendmsg, sock, msg,
 639                                      msg_data_left(msg));
 640         BUG_ON(ret == -EIOCBQUEUED);
 641         return ret;
 642 }
 643 
 644 /**
 645  *      sock_sendmsg - send a message through @sock
 646  *      @sock: socket
 647  *      @msg: message to send
 648  *
 649  *      Sends @msg through @sock, passing through LSM.
 650  *      Returns the number of bytes sent, or an error code.
 651  */
 652 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
 653 {
 654         int err = security_socket_sendmsg(sock, msg,
 655                                           msg_data_left(msg));
 656 
 657         return err ?: sock_sendmsg_nosec(sock, msg);
 658 }
 659 EXPORT_SYMBOL(sock_sendmsg);
 660 
 661 /**
 662  *      kernel_sendmsg - send a message through @sock (kernel-space)
 663  *      @sock: socket
 664  *      @msg: message header
 665  *      @vec: kernel vec
 666  *      @num: vec array length
 667  *      @size: total message data size
 668  *
 669  *      Builds the message data with @vec and sends it through @sock.
 670  *      Returns the number of bytes sent, or an error code.
 671  */
 672 
 673 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 674                    struct kvec *vec, size_t num, size_t size)
 675 {
 676         iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
 677         return sock_sendmsg(sock, msg);
 678 }
 679 EXPORT_SYMBOL(kernel_sendmsg);
 680 
 681 /**
 682  *      kernel_sendmsg_locked - send a message through @sock (kernel-space)
 683  *      @sk: sock
 684  *      @msg: message header
 685  *      @vec: output s/g array
 686  *      @num: output s/g array length
 687  *      @size: total message data size
 688  *
 689  *      Builds the message data with @vec and sends it through @sock.
 690  *      Returns the number of bytes sent, or an error code.
 691  *      Caller must hold @sk.
 692  */
 693 
 694 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
 695                           struct kvec *vec, size_t num, size_t size)
 696 {
 697         struct socket *sock = sk->sk_socket;
 698 
 699         if (!sock->ops->sendmsg_locked)
 700                 return sock_no_sendmsg_locked(sk, msg, size);
 701 
 702         iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
 703 
 704         return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
 705 }
 706 EXPORT_SYMBOL(kernel_sendmsg_locked);
 707 
 708 static bool skb_is_err_queue(const struct sk_buff *skb)
 709 {
 710         /* pkt_type of skbs enqueued on the error queue are set to
 711          * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
 712          * in recvmsg, since skbs received on a local socket will never
 713          * have a pkt_type of PACKET_OUTGOING.
 714          */
 715         return skb->pkt_type == PACKET_OUTGOING;
 716 }
 717 
 718 /* On transmit, software and hardware timestamps are returned independently.
 719  * As the two skb clones share the hardware timestamp, which may be updated
 720  * before the software timestamp is received, a hardware TX timestamp may be
 721  * returned only if there is no software TX timestamp. Ignore false software
 722  * timestamps, which may be made in the __sock_recv_timestamp() call when the
 723  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
 724  * hardware timestamp.
 725  */
 726 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
 727 {
 728         return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
 729 }
 730 
 731 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
 732 {
 733         struct scm_ts_pktinfo ts_pktinfo;
 734         struct net_device *orig_dev;
 735 
 736         if (!skb_mac_header_was_set(skb))
 737                 return;
 738 
 739         memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
 740 
 741         rcu_read_lock();
 742         orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
 743         if (orig_dev)
 744                 ts_pktinfo.if_index = orig_dev->ifindex;
 745         rcu_read_unlock();
 746 
 747         ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
 748         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
 749                  sizeof(ts_pktinfo), &ts_pktinfo);
 750 }
 751 
 752 /*
 753  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 754  */
 755 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 756         struct sk_buff *skb)
 757 {
 758         int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 759         int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
 760         struct scm_timestamping_internal tss;
 761 
 762         int empty = 1, false_tstamp = 0;
 763         struct skb_shared_hwtstamps *shhwtstamps =
 764                 skb_hwtstamps(skb);
 765 
 766         /* Race occurred between timestamp enabling and packet
 767            receiving.  Fill in the current time for now. */
 768         if (need_software_tstamp && skb->tstamp == 0) {
 769                 __net_timestamp(skb);
 770                 false_tstamp = 1;
 771         }
 772 
 773         if (need_software_tstamp) {
 774                 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 775                         if (new_tstamp) {
 776                                 struct __kernel_sock_timeval tv;
 777 
 778                                 skb_get_new_timestamp(skb, &tv);
 779                                 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
 780                                          sizeof(tv), &tv);
 781                         } else {
 782                                 struct __kernel_old_timeval tv;
 783 
 784                                 skb_get_timestamp(skb, &tv);
 785                                 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
 786                                          sizeof(tv), &tv);
 787                         }
 788                 } else {
 789                         if (new_tstamp) {
 790                                 struct __kernel_timespec ts;
 791 
 792                                 skb_get_new_timestampns(skb, &ts);
 793                                 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
 794                                          sizeof(ts), &ts);
 795                         } else {
 796                                 struct timespec ts;
 797 
 798                                 skb_get_timestampns(skb, &ts);
 799                                 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
 800                                          sizeof(ts), &ts);
 801                         }
 802                 }
 803         }
 804 
 805         memset(&tss, 0, sizeof(tss));
 806         if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
 807             ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
 808                 empty = 0;
 809         if (shhwtstamps &&
 810             (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
 811             !skb_is_swtx_tstamp(skb, false_tstamp) &&
 812             ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
 813                 empty = 0;
 814                 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
 815                     !skb_is_err_queue(skb))
 816                         put_ts_pktinfo(msg, skb);
 817         }
 818         if (!empty) {
 819                 if (sock_flag(sk, SOCK_TSTAMP_NEW))
 820                         put_cmsg_scm_timestamping64(msg, &tss);
 821                 else
 822                         put_cmsg_scm_timestamping(msg, &tss);
 823 
 824                 if (skb_is_err_queue(skb) && skb->len &&
 825                     SKB_EXT_ERR(skb)->opt_stats)
 826                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
 827                                  skb->len, skb->data);
 828         }
 829 }
 830 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 831 
 832 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
 833         struct sk_buff *skb)
 834 {
 835         int ack;
 836 
 837         if (!sock_flag(sk, SOCK_WIFI_STATUS))
 838                 return;
 839         if (!skb->wifi_acked_valid)
 840                 return;
 841 
 842         ack = skb->wifi_acked;
 843 
 844         put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
 845 }
 846 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
 847 
 848 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 849                                    struct sk_buff *skb)
 850 {
 851         if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
 852                 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 853                         sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
 854 }
 855 
 856 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 857         struct sk_buff *skb)
 858 {
 859         sock_recv_timestamp(msg, sk, skb);
 860         sock_recv_drops(msg, sk, skb);
 861 }
 862 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 863 
 864 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
 865                                            size_t, int));
 866 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
 867                                             size_t, int));
 868 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 869                                      int flags)
 870 {
 871         return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
 872                                   inet_recvmsg, sock, msg, msg_data_left(msg),
 873                                   flags);
 874 }
 875 
 876 /**
 877  *      sock_recvmsg - receive a message from @sock
 878  *      @sock: socket
 879  *      @msg: message to receive
 880  *      @flags: message flags
 881  *
 882  *      Receives @msg from @sock, passing through LSM. Returns the total number
 883  *      of bytes received, or an error.
 884  */
 885 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
 886 {
 887         int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
 888 
 889         return err ?: sock_recvmsg_nosec(sock, msg, flags);
 890 }
 891 EXPORT_SYMBOL(sock_recvmsg);
 892 
 893 /**
 894  *      kernel_recvmsg - Receive a message from a socket (kernel space)
 895  *      @sock: The socket to receive the message from
 896  *      @msg: Received message
 897  *      @vec: Input s/g array for message data
 898  *      @num: Size of input s/g array
 899  *      @size: Number of bytes to read
 900  *      @flags: Message flags (MSG_DONTWAIT, etc...)
 901  *
 902  *      On return the msg structure contains the scatter/gather array passed in the
 903  *      vec argument. The array is modified so that it consists of the unfilled
 904  *      portion of the original array.
 905  *
 906  *      The returned value is the total number of bytes received, or an error.
 907  */
 908 
 909 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 910                    struct kvec *vec, size_t num, size_t size, int flags)
 911 {
 912         mm_segment_t oldfs = get_fs();
 913         int result;
 914 
 915         iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
 916         set_fs(KERNEL_DS);
 917         result = sock_recvmsg(sock, msg, flags);
 918         set_fs(oldfs);
 919         return result;
 920 }
 921 EXPORT_SYMBOL(kernel_recvmsg);
 922 
 923 static ssize_t sock_sendpage(struct file *file, struct page *page,
 924                              int offset, size_t size, loff_t *ppos, int more)
 925 {
 926         struct socket *sock;
 927         int flags;
 928 
 929         sock = file->private_data;
 930 
 931         flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 932         /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 933         flags |= more;
 934 
 935         return kernel_sendpage(sock, page, offset, size, flags);
 936 }
 937 
 938 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 939                                 struct pipe_inode_info *pipe, size_t len,
 940                                 unsigned int flags)
 941 {
 942         struct socket *sock = file->private_data;
 943 
 944         if (unlikely(!sock->ops->splice_read))
 945                 return generic_file_splice_read(file, ppos, pipe, len, flags);
 946 
 947         return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 948 }
 949 
 950 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
 951 {
 952         struct file *file = iocb->ki_filp;
 953         struct socket *sock = file->private_data;
 954         struct msghdr msg = {.msg_iter = *to,
 955                              .msg_iocb = iocb};
 956         ssize_t res;
 957 
 958         if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
 959                 msg.msg_flags = MSG_DONTWAIT;
 960 
 961         if (iocb->ki_pos != 0)
 962                 return -ESPIPE;
 963 
 964         if (!iov_iter_count(to))        /* Match SYS5 behaviour */
 965                 return 0;
 966 
 967         res = sock_recvmsg(sock, &msg, msg.msg_flags);
 968         *to = msg.msg_iter;
 969         return res;
 970 }
 971 
 972 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
 973 {
 974         struct file *file = iocb->ki_filp;
 975         struct socket *sock = file->private_data;
 976         struct msghdr msg = {.msg_iter = *from,
 977                              .msg_iocb = iocb};
 978         ssize_t res;
 979 
 980         if (iocb->ki_pos != 0)
 981                 return -ESPIPE;
 982 
 983         if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
 984                 msg.msg_flags = MSG_DONTWAIT;
 985 
 986         if (sock->type == SOCK_SEQPACKET)
 987                 msg.msg_flags |= MSG_EOR;
 988 
 989         res = sock_sendmsg(sock, &msg);
 990         *from = msg.msg_iter;
 991         return res;
 992 }
 993 
 994 /*
 995  * Atomic setting of ioctl hooks to avoid race
 996  * with module unload.
 997  */
 998 
 999 static DEFINE_MUTEX(br_ioctl_mutex);
1000 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1001 
1002 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1003 {
1004         mutex_lock(&br_ioctl_mutex);
1005         br_ioctl_hook = hook;
1006         mutex_unlock(&br_ioctl_mutex);
1007 }
1008 EXPORT_SYMBOL(brioctl_set);
1009 
1010 static DEFINE_MUTEX(vlan_ioctl_mutex);
1011 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1012 
1013 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1014 {
1015         mutex_lock(&vlan_ioctl_mutex);
1016         vlan_ioctl_hook = hook;
1017         mutex_unlock(&vlan_ioctl_mutex);
1018 }
1019 EXPORT_SYMBOL(vlan_ioctl_set);
1020 
1021 static DEFINE_MUTEX(dlci_ioctl_mutex);
1022 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1023 
1024 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1025 {
1026         mutex_lock(&dlci_ioctl_mutex);
1027         dlci_ioctl_hook = hook;
1028         mutex_unlock(&dlci_ioctl_mutex);
1029 }
1030 EXPORT_SYMBOL(dlci_ioctl_set);
1031 
1032 static long sock_do_ioctl(struct net *net, struct socket *sock,
1033                           unsigned int cmd, unsigned long arg)
1034 {
1035         int err;
1036         void __user *argp = (void __user *)arg;
1037 
1038         err = sock->ops->ioctl(sock, cmd, arg);
1039 
1040         /*
1041          * If this ioctl is unknown try to hand it down
1042          * to the NIC driver.
1043          */
1044         if (err != -ENOIOCTLCMD)
1045                 return err;
1046 
1047         if (cmd == SIOCGIFCONF) {
1048                 struct ifconf ifc;
1049                 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1050                         return -EFAULT;
1051                 rtnl_lock();
1052                 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1053                 rtnl_unlock();
1054                 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1055                         err = -EFAULT;
1056         } else {
1057                 struct ifreq ifr;
1058                 bool need_copyout;
1059                 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1060                         return -EFAULT;
1061                 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1062                 if (!err && need_copyout)
1063                         if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1064                                 return -EFAULT;
1065         }
1066         return err;
1067 }
1068 
1069 /*
1070  *      With an ioctl, arg may well be a user mode pointer, but we don't know
1071  *      what to do with it - that's up to the protocol still.
1072  */
1073 
1074 /**
1075  *      get_net_ns - increment the refcount of the network namespace
1076  *      @ns: common namespace (net)
1077  *
1078  *      Returns the net's common namespace.
1079  */
1080 
1081 struct ns_common *get_net_ns(struct ns_common *ns)
1082 {
1083         return &get_net(container_of(ns, struct net, ns))->ns;
1084 }
1085 EXPORT_SYMBOL_GPL(get_net_ns);
1086 
1087 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1088 {
1089         struct socket *sock;
1090         struct sock *sk;
1091         void __user *argp = (void __user *)arg;
1092         int pid, err;
1093         struct net *net;
1094 
1095         sock = file->private_data;
1096         sk = sock->sk;
1097         net = sock_net(sk);
1098         if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1099                 struct ifreq ifr;
1100                 bool need_copyout;
1101                 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1102                         return -EFAULT;
1103                 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1104                 if (!err && need_copyout)
1105                         if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1106                                 return -EFAULT;
1107         } else
1108 #ifdef CONFIG_WEXT_CORE
1109         if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1110                 err = wext_handle_ioctl(net, cmd, argp);
1111         } else
1112 #endif
1113                 switch (cmd) {
1114                 case FIOSETOWN:
1115                 case SIOCSPGRP:
1116                         err = -EFAULT;
1117                         if (get_user(pid, (int __user *)argp))
1118                                 break;
1119                         err = f_setown(sock->file, pid, 1);
1120                         break;
1121                 case FIOGETOWN:
1122                 case SIOCGPGRP:
1123                         err = put_user(f_getown(sock->file),
1124                                        (int __user *)argp);
1125                         break;
1126                 case SIOCGIFBR:
1127                 case SIOCSIFBR:
1128                 case SIOCBRADDBR:
1129                 case SIOCBRDELBR:
1130                         err = -ENOPKG;
1131                         if (!br_ioctl_hook)
1132                                 request_module("bridge");
1133 
1134                         mutex_lock(&br_ioctl_mutex);
1135                         if (br_ioctl_hook)
1136                                 err = br_ioctl_hook(net, cmd, argp);
1137                         mutex_unlock(&br_ioctl_mutex);
1138                         break;
1139                 case SIOCGIFVLAN:
1140                 case SIOCSIFVLAN:
1141                         err = -ENOPKG;
1142                         if (!vlan_ioctl_hook)
1143                                 request_module("8021q");
1144 
1145                         mutex_lock(&vlan_ioctl_mutex);
1146                         if (vlan_ioctl_hook)
1147                                 err = vlan_ioctl_hook(net, argp);
1148                         mutex_unlock(&vlan_ioctl_mutex);
1149                         break;
1150                 case SIOCADDDLCI:
1151                 case SIOCDELDLCI:
1152                         err = -ENOPKG;
1153                         if (!dlci_ioctl_hook)
1154                                 request_module("dlci");
1155 
1156                         mutex_lock(&dlci_ioctl_mutex);
1157                         if (dlci_ioctl_hook)
1158                                 err = dlci_ioctl_hook(cmd, argp);
1159                         mutex_unlock(&dlci_ioctl_mutex);
1160                         break;
1161                 case SIOCGSKNS:
1162                         err = -EPERM;
1163                         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1164                                 break;
1165 
1166                         err = open_related_ns(&net->ns, get_net_ns);
1167                         break;
1168                 case SIOCGSTAMP_OLD:
1169                 case SIOCGSTAMPNS_OLD:
1170                         if (!sock->ops->gettstamp) {
1171                                 err = -ENOIOCTLCMD;
1172                                 break;
1173                         }
1174                         err = sock->ops->gettstamp(sock, argp,
1175                                                    cmd == SIOCGSTAMP_OLD,
1176                                                    !IS_ENABLED(CONFIG_64BIT));
1177                         break;
1178                 case SIOCGSTAMP_NEW:
1179                 case SIOCGSTAMPNS_NEW:
1180                         if (!sock->ops->gettstamp) {
1181                                 err = -ENOIOCTLCMD;
1182                                 break;
1183                         }
1184                         err = sock->ops->gettstamp(sock, argp,
1185                                                    cmd == SIOCGSTAMP_NEW,
1186                                                    false);
1187                         break;
1188                 default:
1189                         err = sock_do_ioctl(net, sock, cmd, arg);
1190                         break;
1191                 }
1192         return err;
1193 }
1194 
1195 /**
1196  *      sock_create_lite - creates a socket
1197  *      @family: protocol family (AF_INET, ...)
1198  *      @type: communication type (SOCK_STREAM, ...)
1199  *      @protocol: protocol (0, ...)
1200  *      @res: new socket
1201  *
1202  *      Creates a new socket and assigns it to @res, passing through LSM.
1203  *      The new socket initialization is not complete, see kernel_accept().
1204  *      Returns 0 or an error. On failure @res is set to %NULL.
1205  *      This function internally uses GFP_KERNEL.
1206  */
1207 
1208 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1209 {
1210         int err;
1211         struct socket *sock = NULL;
1212 
1213         err = security_socket_create(family, type, protocol, 1);
1214         if (err)
1215                 goto out;
1216 
1217         sock = sock_alloc();
1218         if (!sock) {
1219                 err = -ENOMEM;
1220                 goto out;
1221         }
1222 
1223         sock->type = type;
1224         err = security_socket_post_create(sock, family, type, protocol, 1);
1225         if (err)
1226                 goto out_release;
1227 
1228 out:
1229         *res = sock;
1230         return err;
1231 out_release:
1232         sock_release(sock);
1233         sock = NULL;
1234         goto out;
1235 }
1236 EXPORT_SYMBOL(sock_create_lite);
1237 
1238 /* No kernel lock held - perfect */
1239 static __poll_t sock_poll(struct file *file, poll_table *wait)
1240 {
1241         struct socket *sock = file->private_data;
1242         __poll_t events = poll_requested_events(wait), flag = 0;
1243 
1244         if (!sock->ops->poll)
1245                 return 0;
1246 
1247         if (sk_can_busy_loop(sock->sk)) {
1248                 /* poll once if requested by the syscall */
1249                 if (events & POLL_BUSY_LOOP)
1250                         sk_busy_loop(sock->sk, 1);
1251 
1252                 /* if this socket can poll_ll, tell the system call */
1253                 flag = POLL_BUSY_LOOP;
1254         }
1255 
1256         return sock->ops->poll(file, sock, wait) | flag;
1257 }
1258 
1259 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1260 {
1261         struct socket *sock = file->private_data;
1262 
1263         return sock->ops->mmap(file, sock, vma);
1264 }
1265 
1266 static int sock_close(struct inode *inode, struct file *filp)
1267 {
1268         __sock_release(SOCKET_I(inode), inode);
1269         return 0;
1270 }
1271 
1272 /*
1273  *      Update the socket async list
1274  *
1275  *      Fasync_list locking strategy.
1276  *
1277  *      1. fasync_list is modified only under process context socket lock
1278  *         i.e. under semaphore.
1279  *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1280  *         or under socket lock
1281  */
1282 
1283 static int sock_fasync(int fd, struct file *filp, int on)
1284 {
1285         struct socket *sock = filp->private_data;
1286         struct sock *sk = sock->sk;
1287         struct socket_wq *wq = &sock->wq;
1288 
1289         if (sk == NULL)
1290                 return -EINVAL;
1291 
1292         lock_sock(sk);
1293         fasync_helper(fd, filp, on, &wq->fasync_list);
1294 
1295         if (!wq->fasync_list)
1296                 sock_reset_flag(sk, SOCK_FASYNC);
1297         else
1298                 sock_set_flag(sk, SOCK_FASYNC);
1299 
1300         release_sock(sk);
1301         return 0;
1302 }
1303 
1304 /* This function may be called only under rcu_lock */
1305 
1306 int sock_wake_async(struct socket_wq *wq, int how, int band)
1307 {
1308         if (!wq || !wq->fasync_list)
1309                 return -1;
1310 
1311         switch (how) {
1312         case SOCK_WAKE_WAITD:
1313                 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1314                         break;
1315                 goto call_kill;
1316         case SOCK_WAKE_SPACE:
1317                 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1318                         break;
1319                 /* fall through */
1320         case SOCK_WAKE_IO:
1321 call_kill:
1322                 kill_fasync(&wq->fasync_list, SIGIO, band);
1323                 break;
1324         case SOCK_WAKE_URG:
1325                 kill_fasync(&wq->fasync_list, SIGURG, band);
1326         }
1327 
1328         return 0;
1329 }
1330 EXPORT_SYMBOL(sock_wake_async);
1331 
1332 /**
1333  *      __sock_create - creates a socket
1334  *      @net: net namespace
1335  *      @family: protocol family (AF_INET, ...)
1336  *      @type: communication type (SOCK_STREAM, ...)
1337  *      @protocol: protocol (0, ...)
1338  *      @res: new socket
1339  *      @kern: boolean for kernel space sockets
1340  *
1341  *      Creates a new socket and assigns it to @res, passing through LSM.
1342  *      Returns 0 or an error. On failure @res is set to %NULL. @kern must
1343  *      be set to true if the socket resides in kernel space.
1344  *      This function internally uses GFP_KERNEL.
1345  */
1346 
1347 int __sock_create(struct net *net, int family, int type, int protocol,
1348                          struct socket **res, int kern)
1349 {
1350         int err;
1351         struct socket *sock;
1352         const struct net_proto_family *pf;
1353 
1354         /*
1355          *      Check protocol is in range
1356          */
1357         if (family < 0 || family >= NPROTO)
1358                 return -EAFNOSUPPORT;
1359         if (type < 0 || type >= SOCK_MAX)
1360                 return -EINVAL;
1361 
1362         /* Compatibility.
1363 
1364            This uglymoron is moved from INET layer to here to avoid
1365            deadlock in module load.
1366          */
1367         if (family == PF_INET && type == SOCK_PACKET) {
1368                 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1369                              current->comm);
1370                 family = PF_PACKET;
1371         }
1372 
1373         err = security_socket_create(family, type, protocol, kern);
1374         if (err)
1375                 return err;
1376 
1377         /*
1378          *      Allocate the socket and allow the family to set things up. if
1379          *      the protocol is 0, the family is instructed to select an appropriate
1380          *      default.
1381          */
1382         sock = sock_alloc();
1383         if (!sock) {
1384                 net_warn_ratelimited("socket: no more sockets\n");
1385                 return -ENFILE; /* Not exactly a match, but its the
1386                                    closest posix thing */
1387         }
1388 
1389         sock->type = type;
1390 
1391 #ifdef CONFIG_MODULES
1392         /* Attempt to load a protocol module if the find failed.
1393          *
1394          * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1395          * requested real, full-featured networking support upon configuration.
1396          * Otherwise module support will break!
1397          */
1398         if (rcu_access_pointer(net_families[family]) == NULL)
1399                 request_module("net-pf-%d", family);
1400 #endif
1401 
1402         rcu_read_lock();
1403         pf = rcu_dereference(net_families[family]);
1404         err = -EAFNOSUPPORT;
1405         if (!pf)
1406                 goto out_release;
1407 
1408         /*
1409          * We will call the ->create function, that possibly is in a loadable
1410          * module, so we have to bump that loadable module refcnt first.
1411          */
1412         if (!try_module_get(pf->owner))
1413                 goto out_release;
1414 
1415         /* Now protected by module ref count */
1416         rcu_read_unlock();
1417 
1418         err = pf->create(net, sock, protocol, kern);
1419         if (err < 0)
1420                 goto out_module_put;
1421 
1422         /*
1423          * Now to bump the refcnt of the [loadable] module that owns this
1424          * socket at sock_release time we decrement its refcnt.
1425          */
1426         if (!try_module_get(sock->ops->owner))
1427                 goto out_module_busy;
1428 
1429         /*
1430          * Now that we're done with the ->create function, the [loadable]
1431          * module can have its refcnt decremented
1432          */
1433         module_put(pf->owner);
1434         err = security_socket_post_create(sock, family, type, protocol, kern);
1435         if (err)
1436                 goto out_sock_release;
1437         *res = sock;
1438 
1439         return 0;
1440 
1441 out_module_busy:
1442         err = -EAFNOSUPPORT;
1443 out_module_put:
1444         sock->ops = NULL;
1445         module_put(pf->owner);
1446 out_sock_release:
1447         sock_release(sock);
1448         return err;
1449 
1450 out_release:
1451         rcu_read_unlock();
1452         goto out_sock_release;
1453 }
1454 EXPORT_SYMBOL(__sock_create);
1455 
1456 /**
1457  *      sock_create - creates a socket
1458  *      @family: protocol family (AF_INET, ...)
1459  *      @type: communication type (SOCK_STREAM, ...)
1460  *      @protocol: protocol (0, ...)
1461  *      @res: new socket
1462  *
1463  *      A wrapper around __sock_create().
1464  *      Returns 0 or an error. This function internally uses GFP_KERNEL.
1465  */
1466 
1467 int sock_create(int family, int type, int protocol, struct socket **res)
1468 {
1469         return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1470 }
1471 EXPORT_SYMBOL(sock_create);
1472 
1473 /**
1474  *      sock_create_kern - creates a socket (kernel space)
1475  *      @net: net namespace
1476  *      @family: protocol family (AF_INET, ...)
1477  *      @type: communication type (SOCK_STREAM, ...)
1478  *      @protocol: protocol (0, ...)
1479  *      @res: new socket
1480  *
1481  *      A wrapper around __sock_create().
1482  *      Returns 0 or an error. This function internally uses GFP_KERNEL.
1483  */
1484 
1485 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1486 {
1487         return __sock_create(net, family, type, protocol, res, 1);
1488 }
1489 EXPORT_SYMBOL(sock_create_kern);
1490 
1491 int __sys_socket(int family, int type, int protocol)
1492 {
1493         int retval;
1494         struct socket *sock;
1495         int flags;
1496 
1497         /* Check the SOCK_* constants for consistency.  */
1498         BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1499         BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1500         BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1501         BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1502 
1503         flags = type & ~SOCK_TYPE_MASK;
1504         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1505                 return -EINVAL;
1506         type &= SOCK_TYPE_MASK;
1507 
1508         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1509                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1510 
1511         retval = sock_create(family, type, protocol, &sock);
1512         if (retval < 0)
1513                 return retval;
1514 
1515         return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1516 }
1517 
1518 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1519 {
1520         return __sys_socket(family, type, protocol);
1521 }
1522 
1523 /*
1524  *      Create a pair of connected sockets.
1525  */
1526 
1527 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1528 {
1529         struct socket *sock1, *sock2;
1530         int fd1, fd2, err;
1531         struct file *newfile1, *newfile2;
1532         int flags;
1533 
1534         flags = type & ~SOCK_TYPE_MASK;
1535         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1536                 return -EINVAL;
1537         type &= SOCK_TYPE_MASK;
1538 
1539         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1540                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1541 
1542         /*
1543          * reserve descriptors and make sure we won't fail
1544          * to return them to userland.
1545          */
1546         fd1 = get_unused_fd_flags(flags);
1547         if (unlikely(fd1 < 0))
1548                 return fd1;
1549 
1550         fd2 = get_unused_fd_flags(flags);
1551         if (unlikely(fd2 < 0)) {
1552                 put_unused_fd(fd1);
1553                 return fd2;
1554         }
1555 
1556         err = put_user(fd1, &usockvec[0]);
1557         if (err)
1558                 goto out;
1559 
1560         err = put_user(fd2, &usockvec[1]);
1561         if (err)
1562                 goto out;
1563 
1564         /*
1565          * Obtain the first socket and check if the underlying protocol
1566          * supports the socketpair call.
1567          */
1568 
1569         err = sock_create(family, type, protocol, &sock1);
1570         if (unlikely(err < 0))
1571                 goto out;
1572 
1573         err = sock_create(family, type, protocol, &sock2);
1574         if (unlikely(err < 0)) {
1575                 sock_release(sock1);
1576                 goto out;
1577         }
1578 
1579         err = security_socket_socketpair(sock1, sock2);
1580         if (unlikely(err)) {
1581                 sock_release(sock2);
1582                 sock_release(sock1);
1583                 goto out;
1584         }
1585 
1586         err = sock1->ops->socketpair(sock1, sock2);
1587         if (unlikely(err < 0)) {
1588                 sock_release(sock2);
1589                 sock_release(sock1);
1590                 goto out;
1591         }
1592 
1593         newfile1 = sock_alloc_file(sock1, flags, NULL);
1594         if (IS_ERR(newfile1)) {
1595                 err = PTR_ERR(newfile1);
1596                 sock_release(sock2);
1597                 goto out;
1598         }
1599 
1600         newfile2 = sock_alloc_file(sock2, flags, NULL);
1601         if (IS_ERR(newfile2)) {
1602                 err = PTR_ERR(newfile2);
1603                 fput(newfile1);
1604                 goto out;
1605         }
1606 
1607         audit_fd_pair(fd1, fd2);
1608 
1609         fd_install(fd1, newfile1);
1610         fd_install(fd2, newfile2);
1611         return 0;
1612 
1613 out:
1614         put_unused_fd(fd2);
1615         put_unused_fd(fd1);
1616         return err;
1617 }
1618 
1619 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1620                 int __user *, usockvec)
1621 {
1622         return __sys_socketpair(family, type, protocol, usockvec);
1623 }
1624 
1625 /*
1626  *      Bind a name to a socket. Nothing much to do here since it's
1627  *      the protocol's responsibility to handle the local address.
1628  *
1629  *      We move the socket address to kernel space before we call
1630  *      the protocol layer (having also checked the address is ok).
1631  */
1632 
1633 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1634 {
1635         struct socket *sock;
1636         struct sockaddr_storage address;
1637         int err, fput_needed;
1638 
1639         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1640         if (sock) {
1641                 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1642                 if (!err) {
1643                         err = security_socket_bind(sock,
1644                                                    (struct sockaddr *)&address,
1645                                                    addrlen);
1646                         if (!err)
1647                                 err = sock->ops->bind(sock,
1648                                                       (struct sockaddr *)
1649                                                       &address, addrlen);
1650                 }
1651                 fput_light(sock->file, fput_needed);
1652         }
1653         return err;
1654 }
1655 
1656 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1657 {
1658         return __sys_bind(fd, umyaddr, addrlen);
1659 }
1660 
1661 /*
1662  *      Perform a listen. Basically, we allow the protocol to do anything
1663  *      necessary for a listen, and if that works, we mark the socket as
1664  *      ready for listening.
1665  */
1666 
1667 int __sys_listen(int fd, int backlog)
1668 {
1669         struct socket *sock;
1670         int err, fput_needed;
1671         int somaxconn;
1672 
1673         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1674         if (sock) {
1675                 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1676                 if ((unsigned int)backlog > somaxconn)
1677                         backlog = somaxconn;
1678 
1679                 err = security_socket_listen(sock, backlog);
1680                 if (!err)
1681                         err = sock->ops->listen(sock, backlog);
1682 
1683                 fput_light(sock->file, fput_needed);
1684         }
1685         return err;
1686 }
1687 
1688 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1689 {
1690         return __sys_listen(fd, backlog);
1691 }
1692 
1693 /*
1694  *      For accept, we attempt to create a new socket, set up the link
1695  *      with the client, wake up the client, then return the new
1696  *      connected fd. We collect the address of the connector in kernel
1697  *      space and move it to user at the very end. This is unclean because
1698  *      we open the socket then return an error.
1699  *
1700  *      1003.1g adds the ability to recvmsg() to query connection pending
1701  *      status to recvmsg. We need to add that support in a way thats
1702  *      clean when we restructure accept also.
1703  */
1704 
1705 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1706                   int __user *upeer_addrlen, int flags)
1707 {
1708         struct socket *sock, *newsock;
1709         struct file *newfile;
1710         int err, len, newfd, fput_needed;
1711         struct sockaddr_storage address;
1712 
1713         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1714                 return -EINVAL;
1715 
1716         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1717                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1718 
1719         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1720         if (!sock)
1721                 goto out;
1722 
1723         err = -ENFILE;
1724         newsock = sock_alloc();
1725         if (!newsock)
1726                 goto out_put;
1727 
1728         newsock->type = sock->type;
1729         newsock->ops = sock->ops;
1730 
1731         /*
1732          * We don't need try_module_get here, as the listening socket (sock)
1733          * has the protocol module (sock->ops->owner) held.
1734          */
1735         __module_get(newsock->ops->owner);
1736 
1737         newfd = get_unused_fd_flags(flags);
1738         if (unlikely(newfd < 0)) {
1739                 err = newfd;
1740                 sock_release(newsock);
1741                 goto out_put;
1742         }
1743         newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1744         if (IS_ERR(newfile)) {
1745                 err = PTR_ERR(newfile);
1746                 put_unused_fd(newfd);
1747                 goto out_put;
1748         }
1749 
1750         err = security_socket_accept(sock, newsock);
1751         if (err)
1752                 goto out_fd;
1753 
1754         err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1755         if (err < 0)
1756                 goto out_fd;
1757 
1758         if (upeer_sockaddr) {
1759                 len = newsock->ops->getname(newsock,
1760                                         (struct sockaddr *)&address, 2);
1761                 if (len < 0) {
1762                         err = -ECONNABORTED;
1763                         goto out_fd;
1764                 }
1765                 err = move_addr_to_user(&address,
1766                                         len, upeer_sockaddr, upeer_addrlen);
1767                 if (err < 0)
1768                         goto out_fd;
1769         }
1770 
1771         /* File flags are not inherited via accept() unlike another OSes. */
1772 
1773         fd_install(newfd, newfile);
1774         err = newfd;
1775 
1776 out_put:
1777         fput_light(sock->file, fput_needed);
1778 out:
1779         return err;
1780 out_fd:
1781         fput(newfile);
1782         put_unused_fd(newfd);
1783         goto out_put;
1784 }
1785 
1786 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1787                 int __user *, upeer_addrlen, int, flags)
1788 {
1789         return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1790 }
1791 
1792 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1793                 int __user *, upeer_addrlen)
1794 {
1795         return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1796 }
1797 
1798 /*
1799  *      Attempt to connect to a socket with the server address.  The address
1800  *      is in user space so we verify it is OK and move it to kernel space.
1801  *
1802  *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1803  *      break bindings
1804  *
1805  *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1806  *      other SEQPACKET protocols that take time to connect() as it doesn't
1807  *      include the -EINPROGRESS status for such sockets.
1808  */
1809 
1810 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1811 {
1812         struct socket *sock;
1813         struct sockaddr_storage address;
1814         int err, fput_needed;
1815 
1816         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1817         if (!sock)
1818                 goto out;
1819         err = move_addr_to_kernel(uservaddr, addrlen, &address);
1820         if (err < 0)
1821                 goto out_put;
1822 
1823         err =
1824             security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1825         if (err)
1826                 goto out_put;
1827 
1828         err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1829                                  sock->file->f_flags);
1830 out_put:
1831         fput_light(sock->file, fput_needed);
1832 out:
1833         return err;
1834 }
1835 
1836 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1837                 int, addrlen)
1838 {
1839         return __sys_connect(fd, uservaddr, addrlen);
1840 }
1841 
1842 /*
1843  *      Get the local address ('name') of a socket object. Move the obtained
1844  *      name to user space.
1845  */
1846 
1847 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1848                       int __user *usockaddr_len)
1849 {
1850         struct socket *sock;
1851         struct sockaddr_storage address;
1852         int err, fput_needed;
1853 
1854         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1855         if (!sock)
1856                 goto out;
1857 
1858         err = security_socket_getsockname(sock);
1859         if (err)
1860                 goto out_put;
1861 
1862         err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1863         if (err < 0)
1864                 goto out_put;
1865         /* "err" is actually length in this case */
1866         err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1867 
1868 out_put:
1869         fput_light(sock->file, fput_needed);
1870 out:
1871         return err;
1872 }
1873 
1874 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1875                 int __user *, usockaddr_len)
1876 {
1877         return __sys_getsockname(fd, usockaddr, usockaddr_len);
1878 }
1879 
1880 /*
1881  *      Get the remote address ('name') of a socket object. Move the obtained
1882  *      name to user space.
1883  */
1884 
1885 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1886                       int __user *usockaddr_len)
1887 {
1888         struct socket *sock;
1889         struct sockaddr_storage address;
1890         int err, fput_needed;
1891 
1892         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1893         if (sock != NULL) {
1894                 err = security_socket_getpeername(sock);
1895                 if (err) {
1896                         fput_light(sock->file, fput_needed);
1897                         return err;
1898                 }
1899 
1900                 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1901                 if (err >= 0)
1902                         /* "err" is actually length in this case */
1903                         err = move_addr_to_user(&address, err, usockaddr,
1904                                                 usockaddr_len);
1905                 fput_light(sock->file, fput_needed);
1906         }
1907         return err;
1908 }
1909 
1910 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1911                 int __user *, usockaddr_len)
1912 {
1913         return __sys_getpeername(fd, usockaddr, usockaddr_len);
1914 }
1915 
1916 /*
1917  *      Send a datagram to a given address. We move the address into kernel
1918  *      space and check the user space data area is readable before invoking
1919  *      the protocol.
1920  */
1921 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1922                  struct sockaddr __user *addr,  int addr_len)
1923 {
1924         struct socket *sock;
1925         struct sockaddr_storage address;
1926         int err;
1927         struct msghdr msg;
1928         struct iovec iov;
1929         int fput_needed;
1930 
1931         err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1932         if (unlikely(err))
1933                 return err;
1934         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1935         if (!sock)
1936                 goto out;
1937 
1938         msg.msg_name = NULL;
1939         msg.msg_control = NULL;
1940         msg.msg_controllen = 0;
1941         msg.msg_namelen = 0;
1942         if (addr) {
1943                 err = move_addr_to_kernel(addr, addr_len, &address);
1944                 if (err < 0)
1945                         goto out_put;
1946                 msg.msg_name = (struct sockaddr *)&address;
1947                 msg.msg_namelen = addr_len;
1948         }
1949         if (sock->file->f_flags & O_NONBLOCK)
1950                 flags |= MSG_DONTWAIT;
1951         msg.msg_flags = flags;
1952         err = sock_sendmsg(sock, &msg);
1953 
1954 out_put:
1955         fput_light(sock->file, fput_needed);
1956 out:
1957         return err;
1958 }
1959 
1960 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1961                 unsigned int, flags, struct sockaddr __user *, addr,
1962                 int, addr_len)
1963 {
1964         return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1965 }
1966 
1967 /*
1968  *      Send a datagram down a socket.
1969  */
1970 
1971 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1972                 unsigned int, flags)
1973 {
1974         return __sys_sendto(fd, buff, len, flags, NULL, 0);
1975 }
1976 
1977 /*
1978  *      Receive a frame from the socket and optionally record the address of the
1979  *      sender. We verify the buffers are writable and if needed move the
1980  *      sender address from kernel to user space.
1981  */
1982 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1983                    struct sockaddr __user *addr, int __user *addr_len)
1984 {
1985         struct socket *sock;
1986         struct iovec iov;
1987         struct msghdr msg;
1988         struct sockaddr_storage address;
1989         int err, err2;
1990         int fput_needed;
1991 
1992         err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1993         if (unlikely(err))
1994                 return err;
1995         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1996         if (!sock)
1997                 goto out;
1998 
1999         msg.msg_control = NULL;
2000         msg.msg_controllen = 0;
2001         /* Save some cycles and don't copy the address if not needed */
2002         msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2003         /* We assume all kernel code knows the size of sockaddr_storage */
2004         msg.msg_namelen = 0;
2005         msg.msg_iocb = NULL;
2006         msg.msg_flags = 0;
2007         if (sock->file->f_flags & O_NONBLOCK)
2008                 flags |= MSG_DONTWAIT;
2009         err = sock_recvmsg(sock, &msg, flags);
2010 
2011         if (err >= 0 && addr != NULL) {
2012                 err2 = move_addr_to_user(&address,
2013                                          msg.msg_namelen, addr, addr_len);
2014                 if (err2 < 0)
2015                         err = err2;
2016         }
2017 
2018         fput_light(sock->file, fput_needed);
2019 out:
2020         return err;
2021 }
2022 
2023 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2024                 unsigned int, flags, struct sockaddr __user *, addr,
2025                 int __user *, addr_len)
2026 {
2027         return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2028 }
2029 
2030 /*
2031  *      Receive a datagram from a socket.
2032  */
2033 
2034 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2035                 unsigned int, flags)
2036 {
2037         return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2038 }
2039 
2040 /*
2041  *      Set a socket option. Because we don't know the option lengths we have
2042  *      to pass the user mode parameter for the protocols to sort out.
2043  */
2044 
2045 static int __sys_setsockopt(int fd, int level, int optname,
2046                             char __user *optval, int optlen)
2047 {
2048         mm_segment_t oldfs = get_fs();
2049         char *kernel_optval = NULL;
2050         int err, fput_needed;
2051         struct socket *sock;
2052 
2053         if (optlen < 0)
2054                 return -EINVAL;
2055 
2056         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2057         if (sock != NULL) {
2058                 err = security_socket_setsockopt(sock, level, optname);
2059                 if (err)
2060                         goto out_put;
2061 
2062                 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level,
2063                                                      &optname, optval, &optlen,
2064                                                      &kernel_optval);
2065 
2066                 if (err < 0) {
2067                         goto out_put;
2068                 } else if (err > 0) {
2069                         err = 0;
2070                         goto out_put;
2071                 }
2072 
2073                 if (kernel_optval) {
2074                         set_fs(KERNEL_DS);
2075                         optval = (char __user __force *)kernel_optval;
2076                 }
2077 
2078                 if (level == SOL_SOCKET)
2079                         err =
2080                             sock_setsockopt(sock, level, optname, optval,
2081                                             optlen);
2082                 else
2083                         err =
2084                             sock->ops->setsockopt(sock, level, optname, optval,
2085                                                   optlen);
2086 
2087                 if (kernel_optval) {
2088                         set_fs(oldfs);
2089                         kfree(kernel_optval);
2090                 }
2091 out_put:
2092                 fput_light(sock->file, fput_needed);
2093         }
2094         return err;
2095 }
2096 
2097 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2098                 char __user *, optval, int, optlen)
2099 {
2100         return __sys_setsockopt(fd, level, optname, optval, optlen);
2101 }
2102 
2103 /*
2104  *      Get a socket option. Because we don't know the option lengths we have
2105  *      to pass a user mode parameter for the protocols to sort out.
2106  */
2107 
2108 static int __sys_getsockopt(int fd, int level, int optname,
2109                             char __user *optval, int __user *optlen)
2110 {
2111         int err, fput_needed;
2112         struct socket *sock;
2113         int max_optlen;
2114 
2115         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2116         if (sock != NULL) {
2117                 err = security_socket_getsockopt(sock, level, optname);
2118                 if (err)
2119                         goto out_put;
2120 
2121                 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2122 
2123                 if (level == SOL_SOCKET)
2124                         err =
2125                             sock_getsockopt(sock, level, optname, optval,
2126                                             optlen);
2127                 else
2128                         err =
2129                             sock->ops->getsockopt(sock, level, optname, optval,
2130                                                   optlen);
2131 
2132                 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2133                                                      optval, optlen,
2134                                                      max_optlen, err);
2135 out_put:
2136                 fput_light(sock->file, fput_needed);
2137         }
2138         return err;
2139 }
2140 
2141 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2142                 char __user *, optval, int __user *, optlen)
2143 {
2144         return __sys_getsockopt(fd, level, optname, optval, optlen);
2145 }
2146 
2147 /*
2148  *      Shutdown a socket.
2149  */
2150 
2151 int __sys_shutdown(int fd, int how)
2152 {
2153         int err, fput_needed;
2154         struct socket *sock;
2155 
2156         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2157         if (sock != NULL) {
2158                 err = security_socket_shutdown(sock, how);
2159                 if (!err)
2160                         err = sock->ops->shutdown(sock, how);
2161                 fput_light(sock->file, fput_needed);
2162         }
2163         return err;
2164 }
2165 
2166 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2167 {
2168         return __sys_shutdown(fd, how);
2169 }
2170 
2171 /* A couple of helpful macros for getting the address of the 32/64 bit
2172  * fields which are the same type (int / unsigned) on our platforms.
2173  */
2174 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2175 #define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
2176 #define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
2177 
2178 struct used_address {
2179         struct sockaddr_storage name;
2180         unsigned int name_len;
2181 };
2182 
2183 static int copy_msghdr_from_user(struct msghdr *kmsg,
2184                                  struct user_msghdr __user *umsg,
2185                                  struct sockaddr __user **save_addr,
2186                                  struct iovec **iov)
2187 {
2188         struct user_msghdr msg;
2189         ssize_t err;
2190 
2191         if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2192                 return -EFAULT;
2193 
2194         kmsg->msg_control = (void __force *)msg.msg_control;
2195         kmsg->msg_controllen = msg.msg_controllen;
2196         kmsg->msg_flags = msg.msg_flags;
2197 
2198         kmsg->msg_namelen = msg.msg_namelen;
2199         if (!msg.msg_name)
2200                 kmsg->msg_namelen = 0;
2201 
2202         if (kmsg->msg_namelen < 0)
2203                 return -EINVAL;
2204 
2205         if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2206                 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2207 
2208         if (save_addr)
2209                 *save_addr = msg.msg_name;
2210 
2211         if (msg.msg_name && kmsg->msg_namelen) {
2212                 if (!save_addr) {
2213                         err = move_addr_to_kernel(msg.msg_name,
2214                                                   kmsg->msg_namelen,
2215                                                   kmsg->msg_name);
2216                         if (err < 0)
2217                                 return err;
2218                 }
2219         } else {
2220                 kmsg->msg_name = NULL;
2221                 kmsg->msg_namelen = 0;
2222         }
2223 
2224         if (msg.msg_iovlen > UIO_MAXIOV)
2225                 return -EMSGSIZE;
2226 
2227         kmsg->msg_iocb = NULL;
2228 
2229         err = import_iovec(save_addr ? READ : WRITE,
2230                             msg.msg_iov, msg.msg_iovlen,
2231                             UIO_FASTIOV, iov, &kmsg->msg_iter);
2232         return err < 0 ? err : 0;
2233 }
2234 
2235 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2236                            unsigned int flags, struct used_address *used_address,
2237                            unsigned int allowed_msghdr_flags)
2238 {
2239         unsigned char ctl[sizeof(struct cmsghdr) + 20]
2240                                 __aligned(sizeof(__kernel_size_t));
2241         /* 20 is size of ipv6_pktinfo */
2242         unsigned char *ctl_buf = ctl;
2243         int ctl_len;
2244         ssize_t err;
2245 
2246         err = -ENOBUFS;
2247 
2248         if (msg_sys->msg_controllen > INT_MAX)
2249                 goto out;
2250         flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2251         ctl_len = msg_sys->msg_controllen;
2252         if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2253                 err =
2254                     cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2255                                                      sizeof(ctl));
2256                 if (err)
2257                         goto out;
2258                 ctl_buf = msg_sys->msg_control;
2259                 ctl_len = msg_sys->msg_controllen;
2260         } else if (ctl_len) {
2261                 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2262                              CMSG_ALIGN(sizeof(struct cmsghdr)));
2263                 if (ctl_len > sizeof(ctl)) {
2264                         ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2265                         if (ctl_buf == NULL)
2266                                 goto out;
2267                 }
2268                 err = -EFAULT;
2269                 /*
2270                  * Careful! Before this, msg_sys->msg_control contains a user pointer.
2271                  * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2272                  * checking falls down on this.
2273                  */
2274                 if (copy_from_user(ctl_buf,
2275                                    (void __user __force *)msg_sys->msg_control,
2276                                    ctl_len))
2277                         goto out_freectl;
2278                 msg_sys->msg_control = ctl_buf;
2279         }
2280         msg_sys->msg_flags = flags;
2281 
2282         if (sock->file->f_flags & O_NONBLOCK)
2283                 msg_sys->msg_flags |= MSG_DONTWAIT;
2284         /*
2285          * If this is sendmmsg() and current destination address is same as
2286          * previously succeeded address, omit asking LSM's decision.
2287          * used_address->name_len is initialized to UINT_MAX so that the first
2288          * destination address never matches.
2289          */
2290         if (used_address && msg_sys->msg_name &&
2291             used_address->name_len == msg_sys->msg_namelen &&
2292             !memcmp(&used_address->name, msg_sys->msg_name,
2293                     used_address->name_len)) {
2294                 err = sock_sendmsg_nosec(sock, msg_sys);
2295                 goto out_freectl;
2296         }
2297         err = sock_sendmsg(sock, msg_sys);
2298         /*
2299          * If this is sendmmsg() and sending to current destination address was
2300          * successful, remember it.
2301          */
2302         if (used_address && err >= 0) {
2303                 used_address->name_len = msg_sys->msg_namelen;
2304                 if (msg_sys->msg_name)
2305                         memcpy(&used_address->name, msg_sys->msg_name,
2306                                used_address->name_len);
2307         }
2308 
2309 out_freectl:
2310         if (ctl_buf != ctl)
2311                 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2312 out:
2313         return err;
2314 }
2315 
2316 static int sendmsg_copy_msghdr(struct msghdr *msg,
2317                                struct user_msghdr __user *umsg, unsigned flags,
2318                                struct iovec **iov)
2319 {
2320         int err;
2321 
2322         if (flags & MSG_CMSG_COMPAT) {
2323                 struct compat_msghdr __user *msg_compat;
2324 
2325                 msg_compat = (struct compat_msghdr __user *) umsg;
2326                 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2327         } else {
2328                 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2329         }
2330         if (err < 0)
2331                 return err;
2332 
2333         return 0;
2334 }
2335 
2336 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2337                          struct msghdr *msg_sys, unsigned int flags,
2338                          struct used_address *used_address,
2339                          unsigned int allowed_msghdr_flags)
2340 {
2341         struct sockaddr_storage address;
2342         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2343         ssize_t err;
2344 
2345         msg_sys->msg_name = &address;
2346 
2347         err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2348         if (err < 0)
2349                 return err;
2350 
2351         err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2352                                 allowed_msghdr_flags);
2353         kfree(iov);
2354         return err;
2355 }
2356 
2357 /*
2358  *      BSD sendmsg interface
2359  */
2360 long __sys_sendmsg_sock(struct socket *sock, struct user_msghdr __user *umsg,
2361                         unsigned int flags)
2362 {
2363         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2364         struct sockaddr_storage address;
2365         struct msghdr msg = { .msg_name = &address };
2366         ssize_t err;
2367 
2368         err = sendmsg_copy_msghdr(&msg, umsg, flags, &iov);
2369         if (err)
2370                 return err;
2371         /* disallow ancillary data requests from this path */
2372         if (msg.msg_control || msg.msg_controllen) {
2373                 err = -EINVAL;
2374                 goto out;
2375         }
2376 
2377         err = ____sys_sendmsg(sock, &msg, flags, NULL, 0);
2378 out:
2379         kfree(iov);
2380         return err;
2381 }
2382 
2383 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2384                    bool forbid_cmsg_compat)
2385 {
2386         int fput_needed, err;
2387         struct msghdr msg_sys;
2388         struct socket *sock;
2389 
2390         if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2391                 return -EINVAL;
2392 
2393         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2394         if (!sock)
2395                 goto out;
2396 
2397         err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2398 
2399         fput_light(sock->file, fput_needed);
2400 out:
2401         return err;
2402 }
2403 
2404 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2405 {
2406         return __sys_sendmsg(fd, msg, flags, true);
2407 }
2408 
2409 /*
2410  *      Linux sendmmsg interface
2411  */
2412 
2413 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2414                    unsigned int flags, bool forbid_cmsg_compat)
2415 {
2416         int fput_needed, err, datagrams;
2417         struct socket *sock;
2418         struct mmsghdr __user *entry;
2419         struct compat_mmsghdr __user *compat_entry;
2420         struct msghdr msg_sys;
2421         struct used_address used_address;
2422         unsigned int oflags = flags;
2423 
2424         if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2425                 return -EINVAL;
2426 
2427         if (vlen > UIO_MAXIOV)
2428                 vlen = UIO_MAXIOV;
2429 
2430         datagrams = 0;
2431 
2432         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2433         if (!sock)
2434                 return err;
2435 
2436         used_address.name_len = UINT_MAX;
2437         entry = mmsg;
2438         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2439         err = 0;
2440         flags |= MSG_BATCH;
2441 
2442         while (datagrams < vlen) {
2443                 if (datagrams == vlen - 1)
2444                         flags = oflags;
2445 
2446                 if (MSG_CMSG_COMPAT & flags) {
2447                         err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2448                                              &msg_sys, flags, &used_address, MSG_EOR);
2449                         if (err < 0)
2450                                 break;
2451                         err = __put_user(err, &compat_entry->msg_len);
2452                         ++compat_entry;
2453                 } else {
2454                         err = ___sys_sendmsg(sock,
2455                                              (struct user_msghdr __user *)entry,
2456                                              &msg_sys, flags, &used_address, MSG_EOR);
2457                         if (err < 0)
2458                                 break;
2459                         err = put_user(err, &entry->msg_len);
2460                         ++entry;
2461                 }
2462 
2463                 if (err)
2464                         break;
2465                 ++datagrams;
2466                 if (msg_data_left(&msg_sys))
2467                         break;
2468                 cond_resched();
2469         }
2470 
2471         fput_light(sock->file, fput_needed);
2472 
2473         /* We only return an error if no datagrams were able to be sent */
2474         if (datagrams != 0)
2475                 return datagrams;
2476 
2477         return err;
2478 }
2479 
2480 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2481                 unsigned int, vlen, unsigned int, flags)
2482 {
2483         return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2484 }
2485 
2486 static int recvmsg_copy_msghdr(struct msghdr *msg,
2487                                struct user_msghdr __user *umsg, unsigned flags,
2488                                struct sockaddr __user **uaddr,
2489                                struct iovec **iov)
2490 {
2491         ssize_t err;
2492 
2493         if (MSG_CMSG_COMPAT & flags) {
2494                 struct compat_msghdr __user *msg_compat;
2495 
2496                 msg_compat = (struct compat_msghdr __user *) umsg;
2497                 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2498         } else {
2499                 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2500         }
2501         if (err < 0)
2502                 return err;
2503 
2504         return 0;
2505 }
2506 
2507 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2508                            struct user_msghdr __user *msg,
2509                            struct sockaddr __user *uaddr,
2510                            unsigned int flags, int nosec)
2511 {
2512         struct compat_msghdr __user *msg_compat =
2513                                         (struct compat_msghdr __user *) msg;
2514         int __user *uaddr_len = COMPAT_NAMELEN(msg);
2515         struct sockaddr_storage addr;
2516         unsigned long cmsg_ptr;
2517         int len;
2518         ssize_t err;
2519 
2520         msg_sys->msg_name = &addr;
2521         cmsg_ptr = (unsigned long)msg_sys->msg_control;
2522         msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2523 
2524         /* We assume all kernel code knows the size of sockaddr_storage */
2525         msg_sys->msg_namelen = 0;
2526 
2527         if (sock->file->f_flags & O_NONBLOCK)
2528                 flags |= MSG_DONTWAIT;
2529         err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2530         if (err < 0)
2531                 goto out;
2532         len = err;
2533 
2534         if (uaddr != NULL) {
2535                 err = move_addr_to_user(&addr,
2536                                         msg_sys->msg_namelen, uaddr,
2537                                         uaddr_len);
2538                 if (err < 0)
2539                         goto out;
2540         }
2541         err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2542                          COMPAT_FLAGS(msg));
2543         if (err)
2544                 goto out;
2545         if (MSG_CMSG_COMPAT & flags)
2546                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2547                                  &msg_compat->msg_controllen);
2548         else
2549                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2550                                  &msg->msg_controllen);
2551         if (err)
2552                 goto out;
2553         err = len;
2554 out:
2555         return err;
2556 }
2557 
2558 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2559                          struct msghdr *msg_sys, unsigned int flags, int nosec)
2560 {
2561         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2562         /* user mode address pointers */
2563         struct sockaddr __user *uaddr;
2564         ssize_t err;
2565 
2566         err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2567         if (err < 0)
2568                 return err;
2569 
2570         err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2571         kfree(iov);
2572         return err;
2573 }
2574 
2575 /*
2576  *      BSD recvmsg interface
2577  */
2578 
2579 long __sys_recvmsg_sock(struct socket *sock, struct user_msghdr __user *umsg,
2580                         unsigned int flags)
2581 {
2582         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2583         struct sockaddr_storage address;
2584         struct msghdr msg = { .msg_name = &address };
2585         struct sockaddr __user *uaddr;
2586         ssize_t err;
2587 
2588         err = recvmsg_copy_msghdr(&msg, umsg, flags, &uaddr, &iov);
2589         if (err)
2590                 return err;
2591         /* disallow ancillary data requests from this path */
2592         if (msg.msg_control || msg.msg_controllen) {
2593                 err = -EINVAL;
2594                 goto out;
2595         }
2596 
2597         err = ____sys_recvmsg(sock, &msg, umsg, uaddr, flags, 0);
2598 out:
2599         kfree(iov);
2600         return err;
2601 }
2602 
2603 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2604                    bool forbid_cmsg_compat)
2605 {
2606         int fput_needed, err;
2607         struct msghdr msg_sys;
2608         struct socket *sock;
2609 
2610         if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2611                 return -EINVAL;
2612 
2613         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2614         if (!sock)
2615                 goto out;
2616 
2617         err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2618 
2619         fput_light(sock->file, fput_needed);
2620 out:
2621         return err;
2622 }
2623 
2624 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2625                 unsigned int, flags)
2626 {
2627         return __sys_recvmsg(fd, msg, flags, true);
2628 }
2629 
2630 /*
2631  *     Linux recvmmsg interface
2632  */
2633 
2634 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2635                           unsigned int vlen, unsigned int flags,
2636                           struct timespec64 *timeout)
2637 {
2638         int fput_needed, err, datagrams;
2639         struct socket *sock;
2640         struct mmsghdr __user *entry;
2641         struct compat_mmsghdr __user *compat_entry;
2642         struct msghdr msg_sys;
2643         struct timespec64 end_time;
2644         struct timespec64 timeout64;
2645 
2646         if (timeout &&
2647             poll_select_set_timeout(&end_time, timeout->tv_sec,
2648                                     timeout->tv_nsec))
2649                 return -EINVAL;
2650 
2651         datagrams = 0;
2652 
2653         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2654         if (!sock)
2655                 return err;
2656 
2657         if (likely(!(flags & MSG_ERRQUEUE))) {
2658                 err = sock_error(sock->sk);
2659                 if (err) {
2660                         datagrams = err;
2661                         goto out_put;
2662                 }
2663         }
2664 
2665         entry = mmsg;
2666         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2667 
2668         while (datagrams < vlen) {
2669                 /*
2670                  * No need to ask LSM for more than the first datagram.
2671                  */
2672                 if (MSG_CMSG_COMPAT & flags) {
2673                         err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2674                                              &msg_sys, flags & ~MSG_WAITFORONE,
2675                                              datagrams);
2676                         if (err < 0)
2677                                 break;
2678                         err = __put_user(err, &compat_entry->msg_len);
2679                         ++compat_entry;
2680                 } else {
2681                         err = ___sys_recvmsg(sock,
2682                                              (struct user_msghdr __user *)entry,
2683                                              &msg_sys, flags & ~MSG_WAITFORONE,
2684                                              datagrams);
2685                         if (err < 0)
2686                                 break;
2687                         err = put_user(err, &entry->msg_len);
2688                         ++entry;
2689                 }
2690 
2691                 if (err)
2692                         break;
2693                 ++datagrams;
2694 
2695                 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2696                 if (flags & MSG_WAITFORONE)
2697                         flags |= MSG_DONTWAIT;
2698 
2699                 if (timeout) {
2700                         ktime_get_ts64(&timeout64);
2701                         *timeout = timespec64_sub(end_time, timeout64);
2702                         if (timeout->tv_sec < 0) {
2703                                 timeout->tv_sec = timeout->tv_nsec = 0;
2704                                 break;
2705                         }
2706 
2707                         /* Timeout, return less than vlen datagrams */
2708                         if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2709                                 break;
2710                 }
2711 
2712                 /* Out of band data, return right away */
2713                 if (msg_sys.msg_flags & MSG_OOB)
2714                         break;
2715                 cond_resched();
2716         }
2717 
2718         if (err == 0)
2719                 goto out_put;
2720 
2721         if (datagrams == 0) {
2722                 datagrams = err;
2723                 goto out_put;
2724         }
2725 
2726         /*
2727          * We may return less entries than requested (vlen) if the
2728          * sock is non block and there aren't enough datagrams...
2729          */
2730         if (err != -EAGAIN) {
2731                 /*
2732                  * ... or  if recvmsg returns an error after we
2733                  * received some datagrams, where we record the
2734                  * error to return on the next call or if the
2735                  * app asks about it using getsockopt(SO_ERROR).
2736                  */
2737                 sock->sk->sk_err = -err;
2738         }
2739 out_put:
2740         fput_light(sock->file, fput_needed);
2741 
2742         return datagrams;
2743 }
2744 
2745 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2746                    unsigned int vlen, unsigned int flags,
2747                    struct __kernel_timespec __user *timeout,
2748                    struct old_timespec32 __user *timeout32)
2749 {
2750         int datagrams;
2751         struct timespec64 timeout_sys;
2752 
2753         if (timeout && get_timespec64(&timeout_sys, timeout))
2754                 return -EFAULT;
2755 
2756         if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2757                 return -EFAULT;
2758 
2759         if (!timeout && !timeout32)
2760                 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2761 
2762         datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2763 
2764         if (datagrams <= 0)
2765                 return datagrams;
2766 
2767         if (timeout && put_timespec64(&timeout_sys, timeout))
2768                 datagrams = -EFAULT;
2769 
2770         if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2771                 datagrams = -EFAULT;
2772 
2773         return datagrams;
2774 }
2775 
2776 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2777                 unsigned int, vlen, unsigned int, flags,
2778                 struct __kernel_timespec __user *, timeout)
2779 {
2780         if (flags & MSG_CMSG_COMPAT)
2781                 return -EINVAL;
2782 
2783         return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2784 }
2785 
2786 #ifdef CONFIG_COMPAT_32BIT_TIME
2787 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2788                 unsigned int, vlen, unsigned int, flags,
2789                 struct old_timespec32 __user *, timeout)
2790 {
2791         if (flags & MSG_CMSG_COMPAT)
2792                 return -EINVAL;
2793 
2794         return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2795 }
2796 #endif
2797 
2798 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2799 /* Argument list sizes for sys_socketcall */
2800 #define AL(x) ((x) * sizeof(unsigned long))
2801 static const unsigned char nargs[21] = {
2802         AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2803         AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2804         AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2805         AL(4), AL(5), AL(4)
2806 };
2807 
2808 #undef AL
2809 
2810 /*
2811  *      System call vectors.
2812  *
2813  *      Argument checking cleaned up. Saved 20% in size.
2814  *  This function doesn't need to set the kernel lock because
2815  *  it is set by the callees.
2816  */
2817 
2818 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2819 {
2820         unsigned long a[AUDITSC_ARGS];
2821         unsigned long a0, a1;
2822         int err;
2823         unsigned int len;
2824 
2825         if (call < 1 || call > SYS_SENDMMSG)
2826                 return -EINVAL;
2827         call = array_index_nospec(call, SYS_SENDMMSG + 1);
2828 
2829         len = nargs[call];
2830         if (len > sizeof(a))
2831                 return -EINVAL;
2832 
2833         /* copy_from_user should be SMP safe. */
2834         if (copy_from_user(a, args, len))
2835                 return -EFAULT;
2836 
2837         err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2838         if (err)
2839                 return err;
2840 
2841         a0 = a[0];
2842         a1 = a[1];
2843 
2844         switch (call) {
2845         case SYS_SOCKET:
2846                 err = __sys_socket(a0, a1, a[2]);
2847                 break;
2848         case SYS_BIND:
2849                 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2850                 break;
2851         case SYS_CONNECT:
2852                 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2853                 break;
2854         case SYS_LISTEN:
2855                 err = __sys_listen(a0, a1);
2856                 break;
2857         case SYS_ACCEPT:
2858                 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2859                                     (int __user *)a[2], 0);
2860                 break;
2861         case SYS_GETSOCKNAME:
2862                 err =
2863                     __sys_getsockname(a0, (struct sockaddr __user *)a1,
2864                                       (int __user *)a[2]);
2865                 break;
2866         case SYS_GETPEERNAME:
2867                 err =
2868                     __sys_getpeername(a0, (struct sockaddr __user *)a1,
2869                                       (int __user *)a[2]);
2870                 break;
2871         case SYS_SOCKETPAIR:
2872                 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2873                 break;
2874         case SYS_SEND:
2875                 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2876                                    NULL, 0);
2877                 break;
2878         case SYS_SENDTO:
2879                 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2880                                    (struct sockaddr __user *)a[4], a[5]);
2881                 break;
2882         case SYS_RECV:
2883                 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2884                                      NULL, NULL);
2885                 break;
2886         case SYS_RECVFROM:
2887                 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2888                                      (struct sockaddr __user *)a[4],
2889                                      (int __user *)a[5]);
2890                 break;
2891         case SYS_SHUTDOWN:
2892                 err = __sys_shutdown(a0, a1);
2893                 break;
2894         case SYS_SETSOCKOPT:
2895                 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2896                                        a[4]);
2897                 break;
2898         case SYS_GETSOCKOPT:
2899                 err =
2900                     __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2901                                      (int __user *)a[4]);
2902                 break;
2903         case SYS_SENDMSG:
2904                 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2905                                     a[2], true);
2906                 break;
2907         case SYS_SENDMMSG:
2908                 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2909                                      a[3], true);
2910                 break;
2911         case SYS_RECVMSG:
2912                 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2913                                     a[2], true);
2914                 break;
2915         case SYS_RECVMMSG:
2916                 if (IS_ENABLED(CONFIG_64BIT) || !IS_ENABLED(CONFIG_64BIT_TIME))
2917                         err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2918                                              a[2], a[3],
2919                                              (struct __kernel_timespec __user *)a[4],
2920                                              NULL);
2921                 else
2922                         err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2923                                              a[2], a[3], NULL,
2924                                              (struct old_timespec32 __user *)a[4]);
2925                 break;
2926         case SYS_ACCEPT4:
2927                 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2928                                     (int __user *)a[2], a[3]);
2929                 break;
2930         default:
2931                 err = -EINVAL;
2932                 break;
2933         }
2934         return err;
2935 }
2936 
2937 #endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2938 
2939 /**
2940  *      sock_register - add a socket protocol handler
2941  *      @ops: description of protocol
2942  *
2943  *      This function is called by a protocol handler that wants to
2944  *      advertise its address family, and have it linked into the
2945  *      socket interface. The value ops->family corresponds to the
2946  *      socket system call protocol family.
2947  */
2948 int sock_register(const struct net_proto_family *ops)
2949 {
2950         int err;
2951 
2952         if (ops->family >= NPROTO) {
2953                 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2954                 return -ENOBUFS;
2955         }
2956 
2957         spin_lock(&net_family_lock);
2958         if (rcu_dereference_protected(net_families[ops->family],
2959                                       lockdep_is_held(&net_family_lock)))
2960                 err = -EEXIST;
2961         else {
2962                 rcu_assign_pointer(net_families[ops->family], ops);
2963                 err = 0;
2964         }
2965         spin_unlock(&net_family_lock);
2966 
2967         pr_info("NET: Registered protocol family %d\n", ops->family);
2968         return err;
2969 }
2970 EXPORT_SYMBOL(sock_register);
2971 
2972 /**
2973  *      sock_unregister - remove a protocol handler
2974  *      @family: protocol family to remove
2975  *
2976  *      This function is called by a protocol handler that wants to
2977  *      remove its address family, and have it unlinked from the
2978  *      new socket creation.
2979  *
2980  *      If protocol handler is a module, then it can use module reference
2981  *      counts to protect against new references. If protocol handler is not
2982  *      a module then it needs to provide its own protection in
2983  *      the ops->create routine.
2984  */
2985 void sock_unregister(int family)
2986 {
2987         BUG_ON(family < 0 || family >= NPROTO);
2988 
2989         spin_lock(&net_family_lock);
2990         RCU_INIT_POINTER(net_families[family], NULL);
2991         spin_unlock(&net_family_lock);
2992 
2993         synchronize_rcu();
2994 
2995         pr_info("NET: Unregistered protocol family %d\n", family);
2996 }
2997 EXPORT_SYMBOL(sock_unregister);
2998 
2999 bool sock_is_registered(int family)
3000 {
3001         return family < NPROTO && rcu_access_pointer(net_families[family]);
3002 }
3003 
3004 static int __init sock_init(void)
3005 {
3006         int err;
3007         /*
3008          *      Initialize the network sysctl infrastructure.
3009          */
3010         err = net_sysctl_init();
3011         if (err)
3012                 goto out;
3013 
3014         /*
3015          *      Initialize skbuff SLAB cache
3016          */
3017         skb_init();
3018 
3019         /*
3020          *      Initialize the protocols module.
3021          */
3022 
3023         init_inodecache();
3024 
3025         err = register_filesystem(&sock_fs_type);
3026         if (err)
3027                 goto out_fs;
3028         sock_mnt = kern_mount(&sock_fs_type);
3029         if (IS_ERR(sock_mnt)) {
3030                 err = PTR_ERR(sock_mnt);
3031                 goto out_mount;
3032         }
3033 
3034         /* The real protocol initialization is performed in later initcalls.
3035          */
3036 
3037 #ifdef CONFIG_NETFILTER
3038         err = netfilter_init();
3039         if (err)
3040                 goto out;
3041 #endif
3042 
3043         ptp_classifier_init();
3044 
3045 out:
3046         return err;
3047 
3048 out_mount:
3049         unregister_filesystem(&sock_fs_type);
3050 out_fs:
3051         goto out;
3052 }
3053 
3054 core_initcall(sock_init);       /* early initcall */
3055 
3056 #ifdef CONFIG_PROC_FS
3057 void socket_seq_show(struct seq_file *seq)
3058 {
3059         seq_printf(seq, "sockets: used %d\n",
3060                    sock_inuse_get(seq->private));
3061 }
3062 #endif                          /* CONFIG_PROC_FS */
3063 
3064 #ifdef CONFIG_COMPAT
3065 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3066 {
3067         struct compat_ifconf ifc32;
3068         struct ifconf ifc;
3069         int err;
3070 
3071         if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3072                 return -EFAULT;
3073 
3074         ifc.ifc_len = ifc32.ifc_len;
3075         ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3076 
3077         rtnl_lock();
3078         err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3079         rtnl_unlock();
3080         if (err)
3081                 return err;
3082 
3083         ifc32.ifc_len = ifc.ifc_len;
3084         if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3085                 return -EFAULT;
3086 
3087         return 0;
3088 }
3089 
3090 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3091 {
3092         struct compat_ethtool_rxnfc __user *compat_rxnfc;
3093         bool convert_in = false, convert_out = false;
3094         size_t buf_size = 0;
3095         struct ethtool_rxnfc __user *rxnfc = NULL;
3096         struct ifreq ifr;
3097         u32 rule_cnt = 0, actual_rule_cnt;
3098         u32 ethcmd;
3099         u32 data;
3100         int ret;
3101 
3102         if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3103                 return -EFAULT;
3104 
3105         compat_rxnfc = compat_ptr(data);
3106 
3107         if (get_user(ethcmd, &compat_rxnfc->cmd))
3108                 return -EFAULT;
3109 
3110         /* Most ethtool structures are defined without padding.
3111          * Unfortunately struct ethtool_rxnfc is an exception.
3112          */
3113         switch (ethcmd) {
3114         default:
3115                 break;
3116         case ETHTOOL_GRXCLSRLALL:
3117                 /* Buffer size is variable */
3118                 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3119                         return -EFAULT;
3120                 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3121                         return -ENOMEM;
3122                 buf_size += rule_cnt * sizeof(u32);
3123                 /* fall through */
3124         case ETHTOOL_GRXRINGS:
3125         case ETHTOOL_GRXCLSRLCNT:
3126         case ETHTOOL_GRXCLSRULE:
3127         case ETHTOOL_SRXCLSRLINS:
3128                 convert_out = true;
3129                 /* fall through */
3130         case ETHTOOL_SRXCLSRLDEL:
3131                 buf_size += sizeof(struct ethtool_rxnfc);
3132                 convert_in = true;
3133                 rxnfc = compat_alloc_user_space(buf_size);
3134                 break;
3135         }
3136 
3137         if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3138                 return -EFAULT;
3139 
3140         ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3141 
3142         if (convert_in) {
3143                 /* We expect there to be holes between fs.m_ext and
3144                  * fs.ring_cookie and at the end of fs, but nowhere else.
3145                  */
3146                 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3147                              sizeof(compat_rxnfc->fs.m_ext) !=
3148                              offsetof(struct ethtool_rxnfc, fs.m_ext) +
3149                              sizeof(rxnfc->fs.m_ext));
3150                 BUILD_BUG_ON(
3151                         offsetof(struct compat_ethtool_rxnfc, fs.location) -
3152                         offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3153                         offsetof(struct ethtool_rxnfc, fs.location) -
3154                         offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3155 
3156                 if (copy_in_user(rxnfc, compat_rxnfc,
3157                                  (void __user *)(&rxnfc->fs.m_ext + 1) -
3158                                  (void __user *)rxnfc) ||
3159                     copy_in_user(&rxnfc->fs.ring_cookie,
3160                                  &compat_rxnfc->fs.ring_cookie,
3161                                  (void __user *)(&rxnfc->fs.location + 1) -
3162                                  (void __user *)&rxnfc->fs.ring_cookie))
3163                         return -EFAULT;
3164                 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3165                         if (put_user(rule_cnt, &rxnfc->rule_cnt))
3166                                 return -EFAULT;
3167                 } else if (copy_in_user(&rxnfc->rule_cnt,
3168                                         &compat_rxnfc->rule_cnt,
3169                                         sizeof(rxnfc->rule_cnt)))
3170                         return -EFAULT;
3171         }
3172 
3173         ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3174         if (ret)
3175                 return ret;
3176 
3177         if (convert_out) {
3178                 if (copy_in_user(compat_rxnfc, rxnfc,
3179                                  (const void __user *)(&rxnfc->fs.m_ext + 1) -
3180                                  (const void __user *)rxnfc) ||
3181                     copy_in_user(&compat_rxnfc->fs.ring_cookie,
3182                                  &rxnfc->fs.ring_cookie,
3183                                  (const void __user *)(&rxnfc->fs.location + 1) -
3184                                  (const void __user *)&rxnfc->fs.ring_cookie) ||
3185                     copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3186                                  sizeof(rxnfc->rule_cnt)))
3187                         return -EFAULT;
3188 
3189                 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3190                         /* As an optimisation, we only copy the actual
3191                          * number of rules that the underlying
3192                          * function returned.  Since Mallory might
3193                          * change the rule count in user memory, we
3194                          * check that it is less than the rule count
3195                          * originally given (as the user buffer size),
3196                          * which has been range-checked.
3197                          */
3198                         if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3199                                 return -EFAULT;
3200                         if (actual_rule_cnt < rule_cnt)
3201                                 rule_cnt = actual_rule_cnt;
3202                         if (copy_in_user(&compat_rxnfc->rule_locs[0],
3203                                          &rxnfc->rule_locs[0],
3204                                          rule_cnt * sizeof(u32)))
3205                                 return -EFAULT;
3206                 }
3207         }
3208 
3209         return 0;
3210 }
3211 
3212 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3213 {
3214         compat_uptr_t uptr32;
3215         struct ifreq ifr;
3216         void __user *saved;
3217         int err;
3218 
3219         if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3220                 return -EFAULT;
3221 
3222         if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3223                 return -EFAULT;
3224 
3225         saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3226         ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3227 
3228         err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3229         if (!err) {
3230                 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3231                 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3232                         err = -EFAULT;
3233         }
3234         return err;
3235 }
3236 
3237 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3238 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3239                                  struct compat_ifreq __user *u_ifreq32)
3240 {
3241         struct ifreq ifreq;
3242         u32 data32;
3243 
3244         if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3245                 return -EFAULT;
3246         if (get_user(data32, &u_ifreq32->ifr_data))
3247                 return -EFAULT;
3248         ifreq.ifr_data = compat_ptr(data32);
3249 
3250         return dev_ioctl(net, cmd, &ifreq, NULL);
3251 }
3252 
3253 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3254                               unsigned int cmd,
3255                               struct compat_ifreq __user *uifr32)
3256 {
3257         struct ifreq __user *uifr;
3258         int err;
3259 
3260         /* Handle the fact that while struct ifreq has the same *layout* on
3261          * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3262          * which are handled elsewhere, it still has different *size* due to
3263          * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3264          * resulting in struct ifreq being 32 and 40 bytes respectively).
3265          * As a result, if the struct happens to be at the end of a page and
3266          * the next page isn't readable/writable, we get a fault. To prevent
3267          * that, copy back and forth to the full size.
3268          */
3269 
3270         uifr = compat_alloc_user_space(sizeof(*uifr));
3271         if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3272                 return -EFAULT;
3273 
3274         err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3275 
3276         if (!err) {
3277                 switch (cmd) {
3278                 case SIOCGIFFLAGS:
3279                 case SIOCGIFMETRIC:
3280                 case SIOCGIFMTU:
3281                 case SIOCGIFMEM:
3282                 case SIOCGIFHWADDR:
3283                 case SIOCGIFINDEX:
3284                 case SIOCGIFADDR:
3285                 case SIOCGIFBRDADDR:
3286                 case SIOCGIFDSTADDR:
3287                 case SIOCGIFNETMASK:
3288                 case SIOCGIFPFLAGS:
3289                 case SIOCGIFTXQLEN:
3290                 case SIOCGMIIPHY:
3291                 case SIOCGMIIREG:
3292                 case SIOCGIFNAME:
3293                         if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3294                                 err = -EFAULT;
3295                         break;
3296                 }
3297         }
3298         return err;
3299 }
3300 
3301 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3302                         struct compat_ifreq __user *uifr32)
3303 {
3304         struct ifreq ifr;
3305         struct compat_ifmap __user *uifmap32;
3306         int err;
3307 
3308         uifmap32 = &uifr32->ifr_ifru.ifru_map;
3309         err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3310         err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3311         err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3312         err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3313         err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3314         err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3315         err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3316         if (err)
3317                 return -EFAULT;
3318 
3319         err = dev_ioctl(net, cmd, &ifr, NULL);
3320 
3321         if (cmd == SIOCGIFMAP && !err) {
3322                 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3323                 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3324                 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3325                 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3326                 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3327                 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3328                 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3329                 if (err)
3330                         err = -EFAULT;
3331         }
3332         return err;
3333 }
3334 
3335 struct rtentry32 {
3336         u32             rt_pad1;
3337         struct sockaddr rt_dst;         /* target address               */
3338         struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3339         struct sockaddr rt_genmask;     /* target network mask (IP)     */
3340         unsigned short  rt_flags;
3341         short           rt_pad2;
3342         u32             rt_pad3;
3343         unsigned char   rt_tos;
3344         unsigned char   rt_class;
3345         short           rt_pad4;
3346         short           rt_metric;      /* +1 for binary compatibility! */
3347         /* char * */ u32 rt_dev;        /* forcing the device at add    */
3348         u32             rt_mtu;         /* per route MTU/Window         */
3349         u32             rt_window;      /* Window clamping              */
3350         unsigned short  rt_irtt;        /* Initial RTT                  */
3351 };
3352 
3353 struct in6_rtmsg32 {
3354         struct in6_addr         rtmsg_dst;
3355         struct in6_addr         rtmsg_src;
3356         struct in6_addr         rtmsg_gateway;
3357         u32                     rtmsg_type;
3358         u16                     rtmsg_dst_len;
3359         u16                     rtmsg_src_len;
3360         u32                     rtmsg_metric;
3361         u32                     rtmsg_info;
3362         u32                     rtmsg_flags;
3363         s32                     rtmsg_ifindex;
3364 };
3365 
3366 static int routing_ioctl(struct net *net, struct socket *sock,
3367                          unsigned int cmd, void __user *argp)
3368 {
3369         int ret;
3370         void *r = NULL;
3371         struct in6_rtmsg r6;
3372         struct rtentry r4;
3373         char devname[16];
3374         u32 rtdev;
3375         mm_segment_t old_fs = get_fs();
3376 
3377         if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3378                 struct in6_rtmsg32 __user *ur6 = argp;
3379                 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3380                         3 * sizeof(struct in6_addr));
3381                 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3382                 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3383                 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3384                 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3385                 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3386                 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3387                 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3388 
3389                 r = (void *) &r6;
3390         } else { /* ipv4 */
3391                 struct rtentry32 __user *ur4 = argp;
3392                 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3393                                         3 * sizeof(struct sockaddr));
3394                 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3395                 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3396                 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3397                 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3398                 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3399                 ret |= get_user(rtdev, &(ur4->rt_dev));
3400                 if (rtdev) {
3401                         ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3402                         r4.rt_dev = (char __user __force *)devname;
3403                         devname[15] = 0;
3404                 } else
3405                         r4.rt_dev = NULL;
3406 
3407                 r = (void *) &r4;
3408         }
3409 
3410         if (ret) {
3411                 ret = -EFAULT;
3412                 goto out;
3413         }
3414 
3415         set_fs(KERNEL_DS);
3416         ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3417         set_fs(old_fs);
3418 
3419 out:
3420         return ret;
3421 }
3422 
3423 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3424  * for some operations; this forces use of the newer bridge-utils that
3425  * use compatible ioctls
3426  */
3427 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3428 {
3429         compat_ulong_t tmp;
3430 
3431         if (get_user(tmp, argp))
3432                 return -EFAULT;
3433         if (tmp == BRCTL_GET_VERSION)
3434                 return BRCTL_VERSION + 1;
3435         return -EINVAL;
3436 }
3437 
3438 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3439                          unsigned int cmd, unsigned long arg)
3440 {
3441         void __user *argp = compat_ptr(arg);
3442         struct sock *sk = sock->sk;
3443         struct net *net = sock_net(sk);
3444 
3445         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3446                 return compat_ifr_data_ioctl(net, cmd, argp);
3447 
3448         switch (cmd) {
3449         case SIOCSIFBR:
3450         case SIOCGIFBR:
3451                 return old_bridge_ioctl(argp);
3452         case SIOCGIFCONF:
3453                 return compat_dev_ifconf(net, argp);
3454         case SIOCETHTOOL:
3455                 return ethtool_ioctl(net, argp);
3456         case SIOCWANDEV:
3457                 return compat_siocwandev(net, argp);
3458         case SIOCGIFMAP:
3459         case SIOCSIFMAP:
3460                 return compat_sioc_ifmap(net, cmd, argp);
3461         case SIOCADDRT:
3462         case SIOCDELRT:
3463                 return routing_ioctl(net, sock, cmd, argp);
3464         case SIOCGSTAMP_OLD:
3465         case SIOCGSTAMPNS_OLD:
3466                 if (!sock->ops->gettstamp)
3467                         return -ENOIOCTLCMD;
3468                 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3469                                             !COMPAT_USE_64BIT_TIME);
3470 
3471         case SIOCBONDSLAVEINFOQUERY:
3472         case SIOCBONDINFOQUERY:
3473         case SIOCSHWTSTAMP:
3474         case SIOCGHWTSTAMP:
3475                 return compat_ifr_data_ioctl(net, cmd, argp);
3476 
3477         case FIOSETOWN:
3478         case SIOCSPGRP:
3479         case FIOGETOWN:
3480         case SIOCGPGRP:
3481         case SIOCBRADDBR:
3482         case SIOCBRDELBR:
3483         case SIOCGIFVLAN:
3484         case SIOCSIFVLAN:
3485         case SIOCADDDLCI:
3486         case SIOCDELDLCI:
3487         case SIOCGSKNS:
3488         case SIOCGSTAMP_NEW:
3489         case SIOCGSTAMPNS_NEW:
3490                 return sock_ioctl(file, cmd, arg);
3491 
3492         case SIOCGIFFLAGS:
3493         case SIOCSIFFLAGS:
3494         case SIOCGIFMETRIC:
3495         case SIOCSIFMETRIC:
3496         case SIOCGIFMTU:
3497         case SIOCSIFMTU:
3498         case SIOCGIFMEM:
3499         case SIOCSIFMEM:
3500         case SIOCGIFHWADDR:
3501         case SIOCSIFHWADDR:
3502         case SIOCADDMULTI:
3503         case SIOCDELMULTI:
3504         case SIOCGIFINDEX:
3505         case SIOCGIFADDR:
3506         case SIOCSIFADDR:
3507         case SIOCSIFHWBROADCAST:
3508         case SIOCDIFADDR:
3509         case SIOCGIFBRDADDR:
3510         case SIOCSIFBRDADDR:
3511         case SIOCGIFDSTADDR:
3512         case SIOCSIFDSTADDR:
3513         case SIOCGIFNETMASK:
3514         case SIOCSIFNETMASK:
3515         case SIOCSIFPFLAGS:
3516         case SIOCGIFPFLAGS:
3517         case SIOCGIFTXQLEN:
3518         case SIOCSIFTXQLEN:
3519         case SIOCBRADDIF:
3520         case SIOCBRDELIF:
3521         case SIOCGIFNAME:
3522         case SIOCSIFNAME:
3523         case SIOCGMIIPHY:
3524         case SIOCGMIIREG:
3525         case SIOCSMIIREG:
3526         case SIOCBONDENSLAVE:
3527         case SIOCBONDRELEASE:
3528         case SIOCBONDSETHWADDR:
3529         case SIOCBONDCHANGEACTIVE:
3530                 return compat_ifreq_ioctl(net, sock, cmd, argp);
3531 
3532         case SIOCSARP:
3533         case SIOCGARP:
3534         case SIOCDARP:
3535         case SIOCOUTQNSD:
3536         case SIOCATMARK:
3537                 return sock_do_ioctl(net, sock, cmd, arg);
3538         }
3539 
3540         return -ENOIOCTLCMD;
3541 }
3542 
3543 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3544                               unsigned long arg)
3545 {
3546         struct socket *sock = file->private_data;
3547         int ret = -ENOIOCTLCMD;
3548         struct sock *sk;
3549         struct net *net;
3550 
3551         sk = sock->sk;
3552         net = sock_net(sk);
3553 
3554         if (sock->ops->compat_ioctl)
3555                 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3556 
3557         if (ret == -ENOIOCTLCMD &&
3558             (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3559                 ret = compat_wext_handle_ioctl(net, cmd, arg);
3560 
3561         if (ret == -ENOIOCTLCMD)
3562                 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3563 
3564         return ret;
3565 }
3566 #endif
3567 
3568 /**
3569  *      kernel_bind - bind an address to a socket (kernel space)
3570  *      @sock: socket
3571  *      @addr: address
3572  *      @addrlen: length of address
3573  *
3574  *      Returns 0 or an error.
3575  */
3576 
3577 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3578 {
3579         return sock->ops->bind(sock, addr, addrlen);
3580 }
3581 EXPORT_SYMBOL(kernel_bind);
3582 
3583 /**
3584  *      kernel_listen - move socket to listening state (kernel space)
3585  *      @sock: socket
3586  *      @backlog: pending connections queue size
3587  *
3588  *      Returns 0 or an error.
3589  */
3590 
3591 int kernel_listen(struct socket *sock, int backlog)
3592 {
3593         return sock->ops->listen(sock, backlog);
3594 }
3595 EXPORT_SYMBOL(kernel_listen);
3596 
3597 /**
3598  *      kernel_accept - accept a connection (kernel space)
3599  *      @sock: listening socket
3600  *      @newsock: new connected socket
3601  *      @flags: flags
3602  *
3603  *      @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3604  *      If it fails, @newsock is guaranteed to be %NULL.
3605  *      Returns 0 or an error.
3606  */
3607 
3608 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3609 {
3610         struct sock *sk = sock->sk;
3611         int err;
3612 
3613         err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3614                                newsock);
3615         if (err < 0)
3616                 goto done;
3617 
3618         err = sock->ops->accept(sock, *newsock, flags, true);
3619         if (err < 0) {
3620                 sock_release(*newsock);
3621                 *newsock = NULL;
3622                 goto done;
3623         }
3624 
3625         (*newsock)->ops = sock->ops;
3626         __module_get((*newsock)->ops->owner);
3627 
3628 done:
3629         return err;
3630 }
3631 EXPORT_SYMBOL(kernel_accept);
3632 
3633 /**
3634  *      kernel_connect - connect a socket (kernel space)
3635  *      @sock: socket
3636  *      @addr: address
3637  *      @addrlen: address length
3638  *      @flags: flags (O_NONBLOCK, ...)
3639  *
3640  *      For datagram sockets, @addr is the addres to which datagrams are sent
3641  *      by default, and the only address from which datagrams are received.
3642  *      For stream sockets, attempts to connect to @addr.
3643  *      Returns 0 or an error code.
3644  */
3645 
3646 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3647                    int flags)
3648 {
3649         return sock->ops->connect(sock, addr, addrlen, flags);
3650 }
3651 EXPORT_SYMBOL(kernel_connect);
3652 
3653 /**
3654  *      kernel_getsockname - get the address which the socket is bound (kernel space)
3655  *      @sock: socket
3656  *      @addr: address holder
3657  *
3658  *      Fills the @addr pointer with the address which the socket is bound.
3659  *      Returns 0 or an error code.
3660  */
3661 
3662 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3663 {
3664         return sock->ops->getname(sock, addr, 0);
3665 }
3666 EXPORT_SYMBOL(kernel_getsockname);
3667 
3668 /**
3669  *      kernel_peername - get the address which the socket is connected (kernel space)
3670  *      @sock: socket
3671  *      @addr: address holder
3672  *
3673  *      Fills the @addr pointer with the address which the socket is connected.
3674  *      Returns 0 or an error code.
3675  */
3676 
3677 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3678 {
3679         return sock->ops->getname(sock, addr, 1);
3680 }
3681 EXPORT_SYMBOL(kernel_getpeername);
3682 
3683 /**
3684  *      kernel_getsockopt - get a socket option (kernel space)
3685  *      @sock: socket
3686  *      @level: API level (SOL_SOCKET, ...)
3687  *      @optname: option tag
3688  *      @optval: option value
3689  *      @optlen: option length
3690  *
3691  *      Assigns the option length to @optlen.
3692  *      Returns 0 or an error.
3693  */
3694 
3695 int kernel_getsockopt(struct socket *sock, int level, int optname,
3696                         char *optval, int *optlen)
3697 {
3698         mm_segment_t oldfs = get_fs();
3699         char __user *uoptval;
3700         int __user *uoptlen;
3701         int err;
3702 
3703         uoptval = (char __user __force *) optval;
3704         uoptlen = (int __user __force *) optlen;
3705 
3706         set_fs(KERNEL_DS);
3707         if (level == SOL_SOCKET)
3708                 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3709         else
3710                 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3711                                             uoptlen);
3712         set_fs(oldfs);
3713         return err;
3714 }
3715 EXPORT_SYMBOL(kernel_getsockopt);
3716 
3717 /**
3718  *      kernel_setsockopt - set a socket option (kernel space)
3719  *      @sock: socket
3720  *      @level: API level (SOL_SOCKET, ...)
3721  *      @optname: option tag
3722  *      @optval: option value
3723  *      @optlen: option length
3724  *
3725  *      Returns 0 or an error.
3726  */
3727 
3728 int kernel_setsockopt(struct socket *sock, int level, int optname,
3729                         char *optval, unsigned int optlen)
3730 {
3731         mm_segment_t oldfs = get_fs();
3732         char __user *uoptval;
3733         int err;
3734 
3735         uoptval = (char __user __force *) optval;
3736 
3737         set_fs(KERNEL_DS);
3738         if (level == SOL_SOCKET)
3739                 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3740         else
3741                 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3742                                             optlen);
3743         set_fs(oldfs);
3744         return err;
3745 }
3746 EXPORT_SYMBOL(kernel_setsockopt);
3747 
3748 /**
3749  *      kernel_sendpage - send a &page through a socket (kernel space)
3750  *      @sock: socket
3751  *      @page: page
3752  *      @offset: page offset
3753  *      @size: total size in bytes
3754  *      @flags: flags (MSG_DONTWAIT, ...)
3755  *
3756  *      Returns the total amount sent in bytes or an error.
3757  */
3758 
3759 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3760                     size_t size, int flags)
3761 {
3762         if (sock->ops->sendpage)
3763                 return sock->ops->sendpage(sock, page, offset, size, flags);
3764 
3765         return sock_no_sendpage(sock, page, offset, size, flags);
3766 }
3767 EXPORT_SYMBOL(kernel_sendpage);
3768 
3769 /**
3770  *      kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3771  *      @sk: sock
3772  *      @page: page
3773  *      @offset: page offset
3774  *      @size: total size in bytes
3775  *      @flags: flags (MSG_DONTWAIT, ...)
3776  *
3777  *      Returns the total amount sent in bytes or an error.
3778  *      Caller must hold @sk.
3779  */
3780 
3781 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3782                            size_t size, int flags)
3783 {
3784         struct socket *sock = sk->sk_socket;
3785 
3786         if (sock->ops->sendpage_locked)
3787                 return sock->ops->sendpage_locked(sk, page, offset, size,
3788                                                   flags);
3789 
3790         return sock_no_sendpage_locked(sk, page, offset, size, flags);
3791 }
3792 EXPORT_SYMBOL(kernel_sendpage_locked);
3793 
3794 /**
3795  *      kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3796  *      @sock: socket
3797  *      @how: connection part
3798  *
3799  *      Returns 0 or an error.
3800  */
3801 
3802 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3803 {
3804         return sock->ops->shutdown(sock, how);
3805 }
3806 EXPORT_SYMBOL(kernel_sock_shutdown);
3807 
3808 /**
3809  *      kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3810  *      @sk: socket
3811  *
3812  *      This routine returns the IP overhead imposed by a socket i.e.
3813  *      the length of the underlying IP header, depending on whether
3814  *      this is an IPv4 or IPv6 socket and the length from IP options turned
3815  *      on at the socket. Assumes that the caller has a lock on the socket.
3816  */
3817 
3818 u32 kernel_sock_ip_overhead(struct sock *sk)
3819 {
3820         struct inet_sock *inet;
3821         struct ip_options_rcu *opt;
3822         u32 overhead = 0;
3823 #if IS_ENABLED(CONFIG_IPV6)
3824         struct ipv6_pinfo *np;
3825         struct ipv6_txoptions *optv6 = NULL;
3826 #endif /* IS_ENABLED(CONFIG_IPV6) */
3827 
3828         if (!sk)
3829                 return overhead;
3830 
3831         switch (sk->sk_family) {
3832         case AF_INET:
3833                 inet = inet_sk(sk);
3834                 overhead += sizeof(struct iphdr);
3835                 opt = rcu_dereference_protected(inet->inet_opt,
3836                                                 sock_owned_by_user(sk));
3837                 if (opt)
3838                         overhead += opt->opt.optlen;
3839                 return overhead;
3840 #if IS_ENABLED(CONFIG_IPV6)
3841         case AF_INET6:
3842                 np = inet6_sk(sk);
3843                 overhead += sizeof(struct ipv6hdr);
3844                 if (np)
3845                         optv6 = rcu_dereference_protected(np->opt,
3846                                                           sock_owned_by_user(sk));
3847                 if (optv6)
3848                         overhead += (optv6->opt_flen + optv6->opt_nflen);
3849                 return overhead;
3850 #endif /* IS_ENABLED(CONFIG_IPV6) */
3851         default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3852                 return overhead;
3853         }
3854 }
3855 EXPORT_SYMBOL(kernel_sock_ip_overhead);