1 /* 2 * NET An implementation of the SOCKET network access protocol. 3 * 4 * Version: @(#)socket.c 1.1.93 18/02/95 5 * 6 * Authors: Orest Zborowski, <obz@Kodak.COM> 7 * Ross Biro 8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 9 * 10 * Fixes: 11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in 12 * shutdown() 13 * Alan Cox : verify_area() fixes 14 * Alan Cox : Removed DDI 15 * Jonathan Kamens : SOCK_DGRAM reconnect bug 16 * Alan Cox : Moved a load of checks to the very 17 * top level. 18 * Alan Cox : Move address structures to/from user 19 * mode above the protocol layers. 20 * Rob Janssen : Allow 0 length sends. 21 * Alan Cox : Asynchronous I/O support (cribbed from the 22 * tty drivers). 23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style) 24 * Jeff Uphoff : Made max number of sockets command-line 25 * configurable. 26 * Matti Aarnio : Made the number of sockets dynamic, 27 * to be allocated when needed, and mr. 28 * Uphoff's max is used as max to be 29 * allowed to allocate. 30 * Linus : Argh. removed all the socket allocation 31 * altogether: it's in the inode now. 32 * Alan Cox : Made sock_alloc()/sock_release() public 33 * for NetROM and future kernel nfsd type 34 * stuff. 35 * Alan Cox : sendmsg/recvmsg basics. 36 * Tom Dyas : Export net symbols. 37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n". 38 * Alan Cox : Added thread locking to sys_* calls 39 * for sockets. May have errors at the 40 * moment. 41 * Kevin Buhr : Fixed the dumb errors in the above. 42 * Andi Kleen : Some small cleanups, optimizations, 43 * and fixed a copy_from_user() bug. 44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0) 45 * Tigran Aivazian : Made listen(2) backlog sanity checks 46 * protocol-independent 47 * 48 * 49 * This program is free software; you can redistribute it and/or 50 * modify it under the terms of the GNU General Public License 51 * as published by the Free Software Foundation; either version 52 * 2 of the License, or (at your option) any later version. 53 * 54 * 55 * This module is effectively the top level interface to the BSD socket 56 * paradigm. 57 * 58 * Based upon Swansea University Computer Society NET3.039 59 */ 60 61 #include <linux/mm.h> 62 #include <linux/socket.h> 63 #include <linux/file.h> 64 #include <linux/net.h> 65 #include <linux/interrupt.h> 66 #include <linux/thread_info.h> 67 #include <linux/rcupdate.h> 68 #include <linux/netdevice.h> 69 #include <linux/proc_fs.h> 70 #include <linux/seq_file.h> 71 #include <linux/mutex.h> 72 #include <linux/if_bridge.h> 73 #include <linux/if_frad.h> 74 #include <linux/if_vlan.h> 75 #include <linux/ptp_classify.h> 76 #include <linux/init.h> 77 #include <linux/poll.h> 78 #include <linux/cache.h> 79 #include <linux/module.h> 80 #include <linux/highmem.h> 81 #include <linux/mount.h> 82 #include <linux/security.h> 83 #include <linux/syscalls.h> 84 #include <linux/compat.h> 85 #include <linux/kmod.h> 86 #include <linux/audit.h> 87 #include <linux/wireless.h> 88 #include <linux/nsproxy.h> 89 #include <linux/magic.h> 90 #include <linux/slab.h> 91 #include <linux/xattr.h> 92 #include <linux/nospec.h> 93 94 #include <linux/uaccess.h> 95 #include <asm/unistd.h> 96 97 #include <net/compat.h> 98 #include <net/wext.h> 99 #include <net/cls_cgroup.h> 100 101 #include <net/sock.h> 102 #include <linux/netfilter.h> 103 104 #include <linux/if_tun.h> 105 #include <linux/ipv6_route.h> 106 #include <linux/route.h> 107 #include <linux/sockios.h> 108 #include <net/busy_poll.h> 109 #include <linux/errqueue.h> 110 111 #ifdef CONFIG_NET_RX_BUSY_POLL 112 unsigned int sysctl_net_busy_read __read_mostly; 113 unsigned int sysctl_net_busy_poll __read_mostly; 114 #endif 115 116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to); 117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from); 118 static int sock_mmap(struct file *file, struct vm_area_struct *vma); 119 120 static int sock_close(struct inode *inode, struct file *file); 121 static __poll_t sock_poll(struct file *file, 122 struct poll_table_struct *wait); 123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 124 #ifdef CONFIG_COMPAT 125 static long compat_sock_ioctl(struct file *file, 126 unsigned int cmd, unsigned long arg); 127 #endif 128 static int sock_fasync(int fd, struct file *filp, int on); 129 static ssize_t sock_sendpage(struct file *file, struct page *page, 130 int offset, size_t size, loff_t *ppos, int more); 131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 132 struct pipe_inode_info *pipe, size_t len, 133 unsigned int flags); 134 135 /* 136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear 137 * in the operation structures but are done directly via the socketcall() multiplexor. 138 */ 139 140 static const struct file_operations socket_file_ops = { 141 .owner = THIS_MODULE, 142 .llseek = no_llseek, 143 .read_iter = sock_read_iter, 144 .write_iter = sock_write_iter, 145 .poll = sock_poll, 146 .unlocked_ioctl = sock_ioctl, 147 #ifdef CONFIG_COMPAT 148 .compat_ioctl = compat_sock_ioctl, 149 #endif 150 .mmap = sock_mmap, 151 .release = sock_close, 152 .fasync = sock_fasync, 153 .sendpage = sock_sendpage, 154 .splice_write = generic_splice_sendpage, 155 .splice_read = sock_splice_read, 156 }; 157 158 /* 159 * The protocol list. Each protocol is registered in here. 160 */ 161 162 static DEFINE_SPINLOCK(net_family_lock); 163 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; 164 165 /* 166 * Support routines. 167 * Move socket addresses back and forth across the kernel/user 168 * divide and look after the messy bits. 169 */ 170 171 /** 172 * move_addr_to_kernel - copy a socket address into kernel space 173 * @uaddr: Address in user space 174 * @kaddr: Address in kernel space 175 * @ulen: Length in user space 176 * 177 * The address is copied into kernel space. If the provided address is 178 * too long an error code of -EINVAL is returned. If the copy gives 179 * invalid addresses -EFAULT is returned. On a success 0 is returned. 180 */ 181 182 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr) 183 { 184 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) 185 return -EINVAL; 186 if (ulen == 0) 187 return 0; 188 if (copy_from_user(kaddr, uaddr, ulen)) 189 return -EFAULT; 190 return audit_sockaddr(ulen, kaddr); 191 } 192 193 /** 194 * move_addr_to_user - copy an address to user space 195 * @kaddr: kernel space address 196 * @klen: length of address in kernel 197 * @uaddr: user space address 198 * @ulen: pointer to user length field 199 * 200 * The value pointed to by ulen on entry is the buffer length available. 201 * This is overwritten with the buffer space used. -EINVAL is returned 202 * if an overlong buffer is specified or a negative buffer size. -EFAULT 203 * is returned if either the buffer or the length field are not 204 * accessible. 205 * After copying the data up to the limit the user specifies, the true 206 * length of the data is written over the length limit the user 207 * specified. Zero is returned for a success. 208 */ 209 210 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen, 211 void __user *uaddr, int __user *ulen) 212 { 213 int err; 214 int len; 215 216 BUG_ON(klen > sizeof(struct sockaddr_storage)); 217 err = get_user(len, ulen); 218 if (err) 219 return err; 220 if (len > klen) 221 len = klen; 222 if (len < 0) 223 return -EINVAL; 224 if (len) { 225 if (audit_sockaddr(klen, kaddr)) 226 return -ENOMEM; 227 if (copy_to_user(uaddr, kaddr, len)) 228 return -EFAULT; 229 } 230 /* 231 * "fromlen shall refer to the value before truncation.." 232 * 1003.1g 233 */ 234 return __put_user(klen, ulen); 235 } 236 237 static struct kmem_cache *sock_inode_cachep __ro_after_init; 238 239 static struct inode *sock_alloc_inode(struct super_block *sb) 240 { 241 struct socket_alloc *ei; 242 struct socket_wq *wq; 243 244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); 245 if (!ei) 246 return NULL; 247 wq = kmalloc(sizeof(*wq), GFP_KERNEL); 248 if (!wq) { 249 kmem_cache_free(sock_inode_cachep, ei); 250 return NULL; 251 } 252 init_waitqueue_head(&wq->wait); 253 wq->fasync_list = NULL; 254 wq->flags = 0; 255 RCU_INIT_POINTER(ei->socket.wq, wq); 256 257 ei->socket.state = SS_UNCONNECTED; 258 ei->socket.flags = 0; 259 ei->socket.ops = NULL; 260 ei->socket.sk = NULL; 261 ei->socket.file = NULL; 262 263 return &ei->vfs_inode; 264 } 265 266 static void sock_destroy_inode(struct inode *inode) 267 { 268 struct socket_alloc *ei; 269 struct socket_wq *wq; 270 271 ei = container_of(inode, struct socket_alloc, vfs_inode); 272 wq = rcu_dereference_protected(ei->socket.wq, 1); 273 kfree_rcu(wq, rcu); 274 kmem_cache_free(sock_inode_cachep, ei); 275 } 276 277 static void init_once(void *foo) 278 { 279 struct socket_alloc *ei = (struct socket_alloc *)foo; 280 281 inode_init_once(&ei->vfs_inode); 282 } 283 284 static void init_inodecache(void) 285 { 286 sock_inode_cachep = kmem_cache_create("sock_inode_cache", 287 sizeof(struct socket_alloc), 288 0, 289 (SLAB_HWCACHE_ALIGN | 290 SLAB_RECLAIM_ACCOUNT | 291 SLAB_MEM_SPREAD | SLAB_ACCOUNT), 292 init_once); 293 BUG_ON(sock_inode_cachep == NULL); 294 } 295 296 static const struct super_operations sockfs_ops = { 297 .alloc_inode = sock_alloc_inode, 298 .destroy_inode = sock_destroy_inode, 299 .statfs = simple_statfs, 300 }; 301 302 /* 303 * sockfs_dname() is called from d_path(). 304 */ 305 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) 306 { 307 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", 308 d_inode(dentry)->i_ino); 309 } 310 311 static const struct dentry_operations sockfs_dentry_operations = { 312 .d_dname = sockfs_dname, 313 }; 314 315 static int sockfs_xattr_get(const struct xattr_handler *handler, 316 struct dentry *dentry, struct inode *inode, 317 const char *suffix, void *value, size_t size) 318 { 319 if (value) { 320 if (dentry->d_name.len + 1 > size) 321 return -ERANGE; 322 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1); 323 } 324 return dentry->d_name.len + 1; 325 } 326 327 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname" 328 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX) 329 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1) 330 331 static const struct xattr_handler sockfs_xattr_handler = { 332 .name = XATTR_NAME_SOCKPROTONAME, 333 .get = sockfs_xattr_get, 334 }; 335 336 static int sockfs_security_xattr_set(const struct xattr_handler *handler, 337 struct dentry *dentry, struct inode *inode, 338 const char *suffix, const void *value, 339 size_t size, int flags) 340 { 341 /* Handled by LSM. */ 342 return -EAGAIN; 343 } 344 345 static const struct xattr_handler sockfs_security_xattr_handler = { 346 .prefix = XATTR_SECURITY_PREFIX, 347 .set = sockfs_security_xattr_set, 348 }; 349 350 static const struct xattr_handler *sockfs_xattr_handlers[] = { 351 &sockfs_xattr_handler, 352 &sockfs_security_xattr_handler, 353 NULL 354 }; 355 356 static struct dentry *sockfs_mount(struct file_system_type *fs_type, 357 int flags, const char *dev_name, void *data) 358 { 359 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops, 360 sockfs_xattr_handlers, 361 &sockfs_dentry_operations, SOCKFS_MAGIC); 362 } 363 364 static struct vfsmount *sock_mnt __read_mostly; 365 366 static struct file_system_type sock_fs_type = { 367 .name = "sockfs", 368 .mount = sockfs_mount, 369 .kill_sb = kill_anon_super, 370 }; 371 372 /* 373 * Obtains the first available file descriptor and sets it up for use. 374 * 375 * These functions create file structures and maps them to fd space 376 * of the current process. On success it returns file descriptor 377 * and file struct implicitly stored in sock->file. 378 * Note that another thread may close file descriptor before we return 379 * from this function. We use the fact that now we do not refer 380 * to socket after mapping. If one day we will need it, this 381 * function will increment ref. count on file by 1. 382 * 383 * In any case returned fd MAY BE not valid! 384 * This race condition is unavoidable 385 * with shared fd spaces, we cannot solve it inside kernel, 386 * but we take care of internal coherence yet. 387 */ 388 389 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname) 390 { 391 struct qstr name = { .name = "" }; 392 struct path path; 393 struct file *file; 394 395 if (dname) { 396 name.name = dname; 397 name.len = strlen(name.name); 398 } else if (sock->sk) { 399 name.name = sock->sk->sk_prot_creator->name; 400 name.len = strlen(name.name); 401 } 402 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name); 403 if (unlikely(!path.dentry)) { 404 sock_release(sock); 405 return ERR_PTR(-ENOMEM); 406 } 407 path.mnt = mntget(sock_mnt); 408 409 d_instantiate(path.dentry, SOCK_INODE(sock)); 410 411 file = alloc_file(&path, FMODE_READ | FMODE_WRITE, 412 &socket_file_ops); 413 if (IS_ERR(file)) { 414 /* drop dentry, keep inode for a bit */ 415 ihold(d_inode(path.dentry)); 416 path_put(&path); 417 /* ... and now kill it properly */ 418 sock_release(sock); 419 return file; 420 } 421 422 sock->file = file; 423 file->f_flags = O_RDWR | (flags & O_NONBLOCK); 424 file->private_data = sock; 425 return file; 426 } 427 EXPORT_SYMBOL(sock_alloc_file); 428 429 static int sock_map_fd(struct socket *sock, int flags) 430 { 431 struct file *newfile; 432 int fd = get_unused_fd_flags(flags); 433 if (unlikely(fd < 0)) { 434 sock_release(sock); 435 return fd; 436 } 437 438 newfile = sock_alloc_file(sock, flags, NULL); 439 if (likely(!IS_ERR(newfile))) { 440 fd_install(fd, newfile); 441 return fd; 442 } 443 444 put_unused_fd(fd); 445 return PTR_ERR(newfile); 446 } 447 448 struct socket *sock_from_file(struct file *file, int *err) 449 { 450 if (file->f_op == &socket_file_ops) 451 return file->private_data; /* set in sock_map_fd */ 452 453 *err = -ENOTSOCK; 454 return NULL; 455 } 456 EXPORT_SYMBOL(sock_from_file); 457 458 /** 459 * sockfd_lookup - Go from a file number to its socket slot 460 * @fd: file handle 461 * @err: pointer to an error code return 462 * 463 * The file handle passed in is locked and the socket it is bound 464 * to is returned. If an error occurs the err pointer is overwritten 465 * with a negative errno code and NULL is returned. The function checks 466 * for both invalid handles and passing a handle which is not a socket. 467 * 468 * On a success the socket object pointer is returned. 469 */ 470 471 struct socket *sockfd_lookup(int fd, int *err) 472 { 473 struct file *file; 474 struct socket *sock; 475 476 file = fget(fd); 477 if (!file) { 478 *err = -EBADF; 479 return NULL; 480 } 481 482 sock = sock_from_file(file, err); 483 if (!sock) 484 fput(file); 485 return sock; 486 } 487 EXPORT_SYMBOL(sockfd_lookup); 488 489 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) 490 { 491 struct fd f = fdget(fd); 492 struct socket *sock; 493 494 *err = -EBADF; 495 if (f.file) { 496 sock = sock_from_file(f.file, err); 497 if (likely(sock)) { 498 *fput_needed = f.flags; 499 return sock; 500 } 501 fdput(f); 502 } 503 return NULL; 504 } 505 506 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer, 507 size_t size) 508 { 509 ssize_t len; 510 ssize_t used = 0; 511 512 len = security_inode_listsecurity(d_inode(dentry), buffer, size); 513 if (len < 0) 514 return len; 515 used += len; 516 if (buffer) { 517 if (size < used) 518 return -ERANGE; 519 buffer += len; 520 } 521 522 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1); 523 used += len; 524 if (buffer) { 525 if (size < used) 526 return -ERANGE; 527 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len); 528 buffer += len; 529 } 530 531 return used; 532 } 533 534 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr) 535 { 536 int err = simple_setattr(dentry, iattr); 537 538 if (!err && (iattr->ia_valid & ATTR_UID)) { 539 struct socket *sock = SOCKET_I(d_inode(dentry)); 540 541 if (sock->sk) 542 sock->sk->sk_uid = iattr->ia_uid; 543 else 544 err = -ENOENT; 545 } 546 547 return err; 548 } 549 550 static const struct inode_operations sockfs_inode_ops = { 551 .listxattr = sockfs_listxattr, 552 .setattr = sockfs_setattr, 553 }; 554 555 /** 556 * sock_alloc - allocate a socket 557 * 558 * Allocate a new inode and socket object. The two are bound together 559 * and initialised. The socket is then returned. If we are out of inodes 560 * NULL is returned. 561 */ 562 563 struct socket *sock_alloc(void) 564 { 565 struct inode *inode; 566 struct socket *sock; 567 568 inode = new_inode_pseudo(sock_mnt->mnt_sb); 569 if (!inode) 570 return NULL; 571 572 sock = SOCKET_I(inode); 573 574 inode->i_ino = get_next_ino(); 575 inode->i_mode = S_IFSOCK | S_IRWXUGO; 576 inode->i_uid = current_fsuid(); 577 inode->i_gid = current_fsgid(); 578 inode->i_op = &sockfs_inode_ops; 579 580 return sock; 581 } 582 EXPORT_SYMBOL(sock_alloc); 583 584 /** 585 * sock_release - close a socket 586 * @sock: socket to close 587 * 588 * The socket is released from the protocol stack if it has a release 589 * callback, and the inode is then released if the socket is bound to 590 * an inode not a file. 591 */ 592 593 static void __sock_release(struct socket *sock, struct inode *inode) 594 { 595 if (sock->ops) { 596 struct module *owner = sock->ops->owner; 597 598 if (inode) 599 inode_lock(inode); 600 sock->ops->release(sock); 601 if (inode) 602 inode_unlock(inode); 603 sock->ops = NULL; 604 module_put(owner); 605 } 606 607 if (rcu_dereference_protected(sock->wq, 1)->fasync_list) 608 pr_err("%s: fasync list not empty!\n", __func__); 609 610 if (!sock->file) { 611 iput(SOCK_INODE(sock)); 612 return; 613 } 614 sock->file = NULL; 615 } 616 617 void sock_release(struct socket *sock) 618 { 619 __sock_release(sock, NULL); 620 } 621 EXPORT_SYMBOL(sock_release); 622 623 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags) 624 { 625 u8 flags = *tx_flags; 626 627 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) 628 flags |= SKBTX_HW_TSTAMP; 629 630 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE) 631 flags |= SKBTX_SW_TSTAMP; 632 633 if (tsflags & SOF_TIMESTAMPING_TX_SCHED) 634 flags |= SKBTX_SCHED_TSTAMP; 635 636 *tx_flags = flags; 637 } 638 EXPORT_SYMBOL(__sock_tx_timestamp); 639 640 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) 641 { 642 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg)); 643 BUG_ON(ret == -EIOCBQUEUED); 644 return ret; 645 } 646 647 int sock_sendmsg(struct socket *sock, struct msghdr *msg) 648 { 649 int err = security_socket_sendmsg(sock, msg, 650 msg_data_left(msg)); 651 652 return err ?: sock_sendmsg_nosec(sock, msg); 653 } 654 EXPORT_SYMBOL(sock_sendmsg); 655 656 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, 657 struct kvec *vec, size_t num, size_t size) 658 { 659 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size); 660 return sock_sendmsg(sock, msg); 661 } 662 EXPORT_SYMBOL(kernel_sendmsg); 663 664 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg, 665 struct kvec *vec, size_t num, size_t size) 666 { 667 struct socket *sock = sk->sk_socket; 668 669 if (!sock->ops->sendmsg_locked) 670 return sock_no_sendmsg_locked(sk, msg, size); 671 672 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size); 673 674 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg)); 675 } 676 EXPORT_SYMBOL(kernel_sendmsg_locked); 677 678 static bool skb_is_err_queue(const struct sk_buff *skb) 679 { 680 /* pkt_type of skbs enqueued on the error queue are set to 681 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do 682 * in recvmsg, since skbs received on a local socket will never 683 * have a pkt_type of PACKET_OUTGOING. 684 */ 685 return skb->pkt_type == PACKET_OUTGOING; 686 } 687 688 /* On transmit, software and hardware timestamps are returned independently. 689 * As the two skb clones share the hardware timestamp, which may be updated 690 * before the software timestamp is received, a hardware TX timestamp may be 691 * returned only if there is no software TX timestamp. Ignore false software 692 * timestamps, which may be made in the __sock_recv_timestamp() call when the 693 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a 694 * hardware timestamp. 695 */ 696 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp) 697 { 698 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb); 699 } 700 701 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb) 702 { 703 struct scm_ts_pktinfo ts_pktinfo; 704 struct net_device *orig_dev; 705 706 if (!skb_mac_header_was_set(skb)) 707 return; 708 709 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo)); 710 711 rcu_read_lock(); 712 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 713 if (orig_dev) 714 ts_pktinfo.if_index = orig_dev->ifindex; 715 rcu_read_unlock(); 716 717 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb); 718 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO, 719 sizeof(ts_pktinfo), &ts_pktinfo); 720 } 721 722 /* 723 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) 724 */ 725 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 726 struct sk_buff *skb) 727 { 728 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); 729 struct scm_timestamping tss; 730 int empty = 1, false_tstamp = 0; 731 struct skb_shared_hwtstamps *shhwtstamps = 732 skb_hwtstamps(skb); 733 734 /* Race occurred between timestamp enabling and packet 735 receiving. Fill in the current time for now. */ 736 if (need_software_tstamp && skb->tstamp == 0) { 737 __net_timestamp(skb); 738 false_tstamp = 1; 739 } 740 741 if (need_software_tstamp) { 742 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { 743 struct timeval tv; 744 skb_get_timestamp(skb, &tv); 745 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, 746 sizeof(tv), &tv); 747 } else { 748 struct timespec ts; 749 skb_get_timestampns(skb, &ts); 750 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, 751 sizeof(ts), &ts); 752 } 753 } 754 755 memset(&tss, 0, sizeof(tss)); 756 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) && 757 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0)) 758 empty = 0; 759 if (shhwtstamps && 760 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) && 761 !skb_is_swtx_tstamp(skb, false_tstamp) && 762 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) { 763 empty = 0; 764 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) && 765 !skb_is_err_queue(skb)) 766 put_ts_pktinfo(msg, skb); 767 } 768 if (!empty) { 769 put_cmsg(msg, SOL_SOCKET, 770 SCM_TIMESTAMPING, sizeof(tss), &tss); 771 772 if (skb_is_err_queue(skb) && skb->len && 773 SKB_EXT_ERR(skb)->opt_stats) 774 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS, 775 skb->len, skb->data); 776 } 777 } 778 EXPORT_SYMBOL_GPL(__sock_recv_timestamp); 779 780 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, 781 struct sk_buff *skb) 782 { 783 int ack; 784 785 if (!sock_flag(sk, SOCK_WIFI_STATUS)) 786 return; 787 if (!skb->wifi_acked_valid) 788 return; 789 790 ack = skb->wifi_acked; 791 792 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); 793 } 794 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); 795 796 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, 797 struct sk_buff *skb) 798 { 799 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount) 800 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, 801 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount); 802 } 803 804 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 805 struct sk_buff *skb) 806 { 807 sock_recv_timestamp(msg, sk, skb); 808 sock_recv_drops(msg, sk, skb); 809 } 810 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops); 811 812 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 813 int flags) 814 { 815 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags); 816 } 817 818 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) 819 { 820 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); 821 822 return err ?: sock_recvmsg_nosec(sock, msg, flags); 823 } 824 EXPORT_SYMBOL(sock_recvmsg); 825 826 /** 827 * kernel_recvmsg - Receive a message from a socket (kernel space) 828 * @sock: The socket to receive the message from 829 * @msg: Received message 830 * @vec: Input s/g array for message data 831 * @num: Size of input s/g array 832 * @size: Number of bytes to read 833 * @flags: Message flags (MSG_DONTWAIT, etc...) 834 * 835 * On return the msg structure contains the scatter/gather array passed in the 836 * vec argument. The array is modified so that it consists of the unfilled 837 * portion of the original array. 838 * 839 * The returned value is the total number of bytes received, or an error. 840 */ 841 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 842 struct kvec *vec, size_t num, size_t size, int flags) 843 { 844 mm_segment_t oldfs = get_fs(); 845 int result; 846 847 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size); 848 set_fs(KERNEL_DS); 849 result = sock_recvmsg(sock, msg, flags); 850 set_fs(oldfs); 851 return result; 852 } 853 EXPORT_SYMBOL(kernel_recvmsg); 854 855 static ssize_t sock_sendpage(struct file *file, struct page *page, 856 int offset, size_t size, loff_t *ppos, int more) 857 { 858 struct socket *sock; 859 int flags; 860 861 sock = file->private_data; 862 863 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 864 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */ 865 flags |= more; 866 867 return kernel_sendpage(sock, page, offset, size, flags); 868 } 869 870 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 871 struct pipe_inode_info *pipe, size_t len, 872 unsigned int flags) 873 { 874 struct socket *sock = file->private_data; 875 876 if (unlikely(!sock->ops->splice_read)) 877 return -EINVAL; 878 879 return sock->ops->splice_read(sock, ppos, pipe, len, flags); 880 } 881 882 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) 883 { 884 struct file *file = iocb->ki_filp; 885 struct socket *sock = file->private_data; 886 struct msghdr msg = {.msg_iter = *to, 887 .msg_iocb = iocb}; 888 ssize_t res; 889 890 if (file->f_flags & O_NONBLOCK) 891 msg.msg_flags = MSG_DONTWAIT; 892 893 if (iocb->ki_pos != 0) 894 return -ESPIPE; 895 896 if (!iov_iter_count(to)) /* Match SYS5 behaviour */ 897 return 0; 898 899 res = sock_recvmsg(sock, &msg, msg.msg_flags); 900 *to = msg.msg_iter; 901 return res; 902 } 903 904 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) 905 { 906 struct file *file = iocb->ki_filp; 907 struct socket *sock = file->private_data; 908 struct msghdr msg = {.msg_iter = *from, 909 .msg_iocb = iocb}; 910 ssize_t res; 911 912 if (iocb->ki_pos != 0) 913 return -ESPIPE; 914 915 if (file->f_flags & O_NONBLOCK) 916 msg.msg_flags = MSG_DONTWAIT; 917 918 if (sock->type == SOCK_SEQPACKET) 919 msg.msg_flags |= MSG_EOR; 920 921 res = sock_sendmsg(sock, &msg); 922 *from = msg.msg_iter; 923 return res; 924 } 925 926 /* 927 * Atomic setting of ioctl hooks to avoid race 928 * with module unload. 929 */ 930 931 static DEFINE_MUTEX(br_ioctl_mutex); 932 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg); 933 934 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) 935 { 936 mutex_lock(&br_ioctl_mutex); 937 br_ioctl_hook = hook; 938 mutex_unlock(&br_ioctl_mutex); 939 } 940 EXPORT_SYMBOL(brioctl_set); 941 942 static DEFINE_MUTEX(vlan_ioctl_mutex); 943 static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 944 945 void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 946 { 947 mutex_lock(&vlan_ioctl_mutex); 948 vlan_ioctl_hook = hook; 949 mutex_unlock(&vlan_ioctl_mutex); 950 } 951 EXPORT_SYMBOL(vlan_ioctl_set); 952 953 static DEFINE_MUTEX(dlci_ioctl_mutex); 954 static int (*dlci_ioctl_hook) (unsigned int, void __user *); 955 956 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) 957 { 958 mutex_lock(&dlci_ioctl_mutex); 959 dlci_ioctl_hook = hook; 960 mutex_unlock(&dlci_ioctl_mutex); 961 } 962 EXPORT_SYMBOL(dlci_ioctl_set); 963 964 static long sock_do_ioctl(struct net *net, struct socket *sock, 965 unsigned int cmd, unsigned long arg) 966 { 967 int err; 968 void __user *argp = (void __user *)arg; 969 970 err = sock->ops->ioctl(sock, cmd, arg); 971 972 /* 973 * If this ioctl is unknown try to hand it down 974 * to the NIC driver. 975 */ 976 if (err != -ENOIOCTLCMD) 977 return err; 978 979 if (cmd == SIOCGIFCONF) { 980 struct ifconf ifc; 981 if (copy_from_user(&ifc, argp, sizeof(struct ifconf))) 982 return -EFAULT; 983 rtnl_lock(); 984 err = dev_ifconf(net, &ifc, sizeof(struct ifreq)); 985 rtnl_unlock(); 986 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf))) 987 err = -EFAULT; 988 } else { 989 struct ifreq ifr; 990 bool need_copyout; 991 if (copy_from_user(&ifr, argp, sizeof(struct ifreq))) 992 return -EFAULT; 993 err = dev_ioctl(net, cmd, &ifr, &need_copyout); 994 if (!err && need_copyout) 995 if (copy_to_user(argp, &ifr, sizeof(struct ifreq))) 996 return -EFAULT; 997 } 998 return err; 999 } 1000 1001 /* 1002 * With an ioctl, arg may well be a user mode pointer, but we don't know 1003 * what to do with it - that's up to the protocol still. 1004 */ 1005 1006 struct ns_common *get_net_ns(struct ns_common *ns) 1007 { 1008 return &get_net(container_of(ns, struct net, ns))->ns; 1009 } 1010 EXPORT_SYMBOL_GPL(get_net_ns); 1011 1012 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 1013 { 1014 struct socket *sock; 1015 struct sock *sk; 1016 void __user *argp = (void __user *)arg; 1017 int pid, err; 1018 struct net *net; 1019 1020 sock = file->private_data; 1021 sk = sock->sk; 1022 net = sock_net(sk); 1023 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { 1024 struct ifreq ifr; 1025 bool need_copyout; 1026 if (copy_from_user(&ifr, argp, sizeof(struct ifreq))) 1027 return -EFAULT; 1028 err = dev_ioctl(net, cmd, &ifr, &need_copyout); 1029 if (!err && need_copyout) 1030 if (copy_to_user(argp, &ifr, sizeof(struct ifreq))) 1031 return -EFAULT; 1032 } else 1033 #ifdef CONFIG_WEXT_CORE 1034 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 1035 err = wext_handle_ioctl(net, cmd, argp); 1036 } else 1037 #endif 1038 switch (cmd) { 1039 case FIOSETOWN: 1040 case SIOCSPGRP: 1041 err = -EFAULT; 1042 if (get_user(pid, (int __user *)argp)) 1043 break; 1044 err = f_setown(sock->file, pid, 1); 1045 break; 1046 case FIOGETOWN: 1047 case SIOCGPGRP: 1048 err = put_user(f_getown(sock->file), 1049 (int __user *)argp); 1050 break; 1051 case SIOCGIFBR: 1052 case SIOCSIFBR: 1053 case SIOCBRADDBR: 1054 case SIOCBRDELBR: 1055 err = -ENOPKG; 1056 if (!br_ioctl_hook) 1057 request_module("bridge"); 1058 1059 mutex_lock(&br_ioctl_mutex); 1060 if (br_ioctl_hook) 1061 err = br_ioctl_hook(net, cmd, argp); 1062 mutex_unlock(&br_ioctl_mutex); 1063 break; 1064 case SIOCGIFVLAN: 1065 case SIOCSIFVLAN: 1066 err = -ENOPKG; 1067 if (!vlan_ioctl_hook) 1068 request_module("8021q"); 1069 1070 mutex_lock(&vlan_ioctl_mutex); 1071 if (vlan_ioctl_hook) 1072 err = vlan_ioctl_hook(net, argp); 1073 mutex_unlock(&vlan_ioctl_mutex); 1074 break; 1075 case SIOCADDDLCI: 1076 case SIOCDELDLCI: 1077 err = -ENOPKG; 1078 if (!dlci_ioctl_hook) 1079 request_module("dlci"); 1080 1081 mutex_lock(&dlci_ioctl_mutex); 1082 if (dlci_ioctl_hook) 1083 err = dlci_ioctl_hook(cmd, argp); 1084 mutex_unlock(&dlci_ioctl_mutex); 1085 break; 1086 case SIOCGSKNS: 1087 err = -EPERM; 1088 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1089 break; 1090 1091 err = open_related_ns(&net->ns, get_net_ns); 1092 break; 1093 default: 1094 err = sock_do_ioctl(net, sock, cmd, arg); 1095 break; 1096 } 1097 return err; 1098 } 1099 1100 int sock_create_lite(int family, int type, int protocol, struct socket **res) 1101 { 1102 int err; 1103 struct socket *sock = NULL; 1104 1105 err = security_socket_create(family, type, protocol, 1); 1106 if (err) 1107 goto out; 1108 1109 sock = sock_alloc(); 1110 if (!sock) { 1111 err = -ENOMEM; 1112 goto out; 1113 } 1114 1115 sock->type = type; 1116 err = security_socket_post_create(sock, family, type, protocol, 1); 1117 if (err) 1118 goto out_release; 1119 1120 out: 1121 *res = sock; 1122 return err; 1123 out_release: 1124 sock_release(sock); 1125 sock = NULL; 1126 goto out; 1127 } 1128 EXPORT_SYMBOL(sock_create_lite); 1129 1130 /* No kernel lock held - perfect */ 1131 static __poll_t sock_poll(struct file *file, poll_table *wait) 1132 { 1133 struct socket *sock = file->private_data; 1134 __poll_t events = poll_requested_events(wait); 1135 1136 sock_poll_busy_loop(sock, events); 1137 if (!sock->ops->poll) 1138 return 0; 1139 return sock->ops->poll(file, sock, wait) | sock_poll_busy_flag(sock); 1140 } 1141 1142 static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1143 { 1144 struct socket *sock = file->private_data; 1145 1146 return sock->ops->mmap(file, sock, vma); 1147 } 1148 1149 static int sock_close(struct inode *inode, struct file *filp) 1150 { 1151 __sock_release(SOCKET_I(inode), inode); 1152 return 0; 1153 } 1154 1155 /* 1156 * Update the socket async list 1157 * 1158 * Fasync_list locking strategy. 1159 * 1160 * 1. fasync_list is modified only under process context socket lock 1161 * i.e. under semaphore. 1162 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1163 * or under socket lock 1164 */ 1165 1166 static int sock_fasync(int fd, struct file *filp, int on) 1167 { 1168 struct socket *sock = filp->private_data; 1169 struct sock *sk = sock->sk; 1170 struct socket_wq *wq; 1171 1172 if (sk == NULL) 1173 return -EINVAL; 1174 1175 lock_sock(sk); 1176 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk)); 1177 fasync_helper(fd, filp, on, &wq->fasync_list); 1178 1179 if (!wq->fasync_list) 1180 sock_reset_flag(sk, SOCK_FASYNC); 1181 else 1182 sock_set_flag(sk, SOCK_FASYNC); 1183 1184 release_sock(sk); 1185 return 0; 1186 } 1187 1188 /* This function may be called only under rcu_lock */ 1189 1190 int sock_wake_async(struct socket_wq *wq, int how, int band) 1191 { 1192 if (!wq || !wq->fasync_list) 1193 return -1; 1194 1195 switch (how) { 1196 case SOCK_WAKE_WAITD: 1197 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) 1198 break; 1199 goto call_kill; 1200 case SOCK_WAKE_SPACE: 1201 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) 1202 break; 1203 /* fall through */ 1204 case SOCK_WAKE_IO: 1205 call_kill: 1206 kill_fasync(&wq->fasync_list, SIGIO, band); 1207 break; 1208 case SOCK_WAKE_URG: 1209 kill_fasync(&wq->fasync_list, SIGURG, band); 1210 } 1211 1212 return 0; 1213 } 1214 EXPORT_SYMBOL(sock_wake_async); 1215 1216 int __sock_create(struct net *net, int family, int type, int protocol, 1217 struct socket **res, int kern) 1218 { 1219 int err; 1220 struct socket *sock; 1221 const struct net_proto_family *pf; 1222 1223 /* 1224 * Check protocol is in range 1225 */ 1226 if (family < 0 || family >= NPROTO) 1227 return -EAFNOSUPPORT; 1228 if (type < 0 || type >= SOCK_MAX) 1229 return -EINVAL; 1230 1231 /* Compatibility. 1232 1233 This uglymoron is moved from INET layer to here to avoid 1234 deadlock in module load. 1235 */ 1236 if (family == PF_INET && type == SOCK_PACKET) { 1237 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1238 current->comm); 1239 family = PF_PACKET; 1240 } 1241 1242 err = security_socket_create(family, type, protocol, kern); 1243 if (err) 1244 return err; 1245 1246 /* 1247 * Allocate the socket and allow the family to set things up. if 1248 * the protocol is 0, the family is instructed to select an appropriate 1249 * default. 1250 */ 1251 sock = sock_alloc(); 1252 if (!sock) { 1253 net_warn_ratelimited("socket: no more sockets\n"); 1254 return -ENFILE; /* Not exactly a match, but its the 1255 closest posix thing */ 1256 } 1257 1258 sock->type = type; 1259 1260 #ifdef CONFIG_MODULES 1261 /* Attempt to load a protocol module if the find failed. 1262 * 1263 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1264 * requested real, full-featured networking support upon configuration. 1265 * Otherwise module support will break! 1266 */ 1267 if (rcu_access_pointer(net_families[family]) == NULL) 1268 request_module("net-pf-%d", family); 1269 #endif 1270 1271 rcu_read_lock(); 1272 pf = rcu_dereference(net_families[family]); 1273 err = -EAFNOSUPPORT; 1274 if (!pf) 1275 goto out_release; 1276 1277 /* 1278 * We will call the ->create function, that possibly is in a loadable 1279 * module, so we have to bump that loadable module refcnt first. 1280 */ 1281 if (!try_module_get(pf->owner)) 1282 goto out_release; 1283 1284 /* Now protected by module ref count */ 1285 rcu_read_unlock(); 1286 1287 err = pf->create(net, sock, protocol, kern); 1288 if (err < 0) 1289 goto out_module_put; 1290 1291 /* 1292 * Now to bump the refcnt of the [loadable] module that owns this 1293 * socket at sock_release time we decrement its refcnt. 1294 */ 1295 if (!try_module_get(sock->ops->owner)) 1296 goto out_module_busy; 1297 1298 /* 1299 * Now that we're done with the ->create function, the [loadable] 1300 * module can have its refcnt decremented 1301 */ 1302 module_put(pf->owner); 1303 err = security_socket_post_create(sock, family, type, protocol, kern); 1304 if (err) 1305 goto out_sock_release; 1306 *res = sock; 1307 1308 return 0; 1309 1310 out_module_busy: 1311 err = -EAFNOSUPPORT; 1312 out_module_put: 1313 sock->ops = NULL; 1314 module_put(pf->owner); 1315 out_sock_release: 1316 sock_release(sock); 1317 return err; 1318 1319 out_release: 1320 rcu_read_unlock(); 1321 goto out_sock_release; 1322 } 1323 EXPORT_SYMBOL(__sock_create); 1324 1325 int sock_create(int family, int type, int protocol, struct socket **res) 1326 { 1327 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1328 } 1329 EXPORT_SYMBOL(sock_create); 1330 1331 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) 1332 { 1333 return __sock_create(net, family, type, protocol, res, 1); 1334 } 1335 EXPORT_SYMBOL(sock_create_kern); 1336 1337 int __sys_socket(int family, int type, int protocol) 1338 { 1339 int retval; 1340 struct socket *sock; 1341 int flags; 1342 1343 /* Check the SOCK_* constants for consistency. */ 1344 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1345 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1346 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1347 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1348 1349 flags = type & ~SOCK_TYPE_MASK; 1350 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1351 return -EINVAL; 1352 type &= SOCK_TYPE_MASK; 1353 1354 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1355 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1356 1357 retval = sock_create(family, type, protocol, &sock); 1358 if (retval < 0) 1359 return retval; 1360 1361 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1362 } 1363 1364 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1365 { 1366 return __sys_socket(family, type, protocol); 1367 } 1368 1369 /* 1370 * Create a pair of connected sockets. 1371 */ 1372 1373 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) 1374 { 1375 struct socket *sock1, *sock2; 1376 int fd1, fd2, err; 1377 struct file *newfile1, *newfile2; 1378 int flags; 1379 1380 flags = type & ~SOCK_TYPE_MASK; 1381 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1382 return -EINVAL; 1383 type &= SOCK_TYPE_MASK; 1384 1385 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1386 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1387 1388 /* 1389 * reserve descriptors and make sure we won't fail 1390 * to return them to userland. 1391 */ 1392 fd1 = get_unused_fd_flags(flags); 1393 if (unlikely(fd1 < 0)) 1394 return fd1; 1395 1396 fd2 = get_unused_fd_flags(flags); 1397 if (unlikely(fd2 < 0)) { 1398 put_unused_fd(fd1); 1399 return fd2; 1400 } 1401 1402 err = put_user(fd1, &usockvec[0]); 1403 if (err) 1404 goto out; 1405 1406 err = put_user(fd2, &usockvec[1]); 1407 if (err) 1408 goto out; 1409 1410 /* 1411 * Obtain the first socket and check if the underlying protocol 1412 * supports the socketpair call. 1413 */ 1414 1415 err = sock_create(family, type, protocol, &sock1); 1416 if (unlikely(err < 0)) 1417 goto out; 1418 1419 err = sock_create(family, type, protocol, &sock2); 1420 if (unlikely(err < 0)) { 1421 sock_release(sock1); 1422 goto out; 1423 } 1424 1425 err = security_socket_socketpair(sock1, sock2); 1426 if (unlikely(err)) { 1427 sock_release(sock2); 1428 sock_release(sock1); 1429 goto out; 1430 } 1431 1432 err = sock1->ops->socketpair(sock1, sock2); 1433 if (unlikely(err < 0)) { 1434 sock_release(sock2); 1435 sock_release(sock1); 1436 goto out; 1437 } 1438 1439 newfile1 = sock_alloc_file(sock1, flags, NULL); 1440 if (IS_ERR(newfile1)) { 1441 err = PTR_ERR(newfile1); 1442 sock_release(sock2); 1443 goto out; 1444 } 1445 1446 newfile2 = sock_alloc_file(sock2, flags, NULL); 1447 if (IS_ERR(newfile2)) { 1448 err = PTR_ERR(newfile2); 1449 fput(newfile1); 1450 goto out; 1451 } 1452 1453 audit_fd_pair(fd1, fd2); 1454 1455 fd_install(fd1, newfile1); 1456 fd_install(fd2, newfile2); 1457 return 0; 1458 1459 out: 1460 put_unused_fd(fd2); 1461 put_unused_fd(fd1); 1462 return err; 1463 } 1464 1465 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1466 int __user *, usockvec) 1467 { 1468 return __sys_socketpair(family, type, protocol, usockvec); 1469 } 1470 1471 /* 1472 * Bind a name to a socket. Nothing much to do here since it's 1473 * the protocol's responsibility to handle the local address. 1474 * 1475 * We move the socket address to kernel space before we call 1476 * the protocol layer (having also checked the address is ok). 1477 */ 1478 1479 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) 1480 { 1481 struct socket *sock; 1482 struct sockaddr_storage address; 1483 int err, fput_needed; 1484 1485 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1486 if (sock) { 1487 err = move_addr_to_kernel(umyaddr, addrlen, &address); 1488 if (err >= 0) { 1489 err = security_socket_bind(sock, 1490 (struct sockaddr *)&address, 1491 addrlen); 1492 if (!err) 1493 err = sock->ops->bind(sock, 1494 (struct sockaddr *) 1495 &address, addrlen); 1496 } 1497 fput_light(sock->file, fput_needed); 1498 } 1499 return err; 1500 } 1501 1502 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1503 { 1504 return __sys_bind(fd, umyaddr, addrlen); 1505 } 1506 1507 /* 1508 * Perform a listen. Basically, we allow the protocol to do anything 1509 * necessary for a listen, and if that works, we mark the socket as 1510 * ready for listening. 1511 */ 1512 1513 int __sys_listen(int fd, int backlog) 1514 { 1515 struct socket *sock; 1516 int err, fput_needed; 1517 int somaxconn; 1518 1519 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1520 if (sock) { 1521 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; 1522 if ((unsigned int)backlog > somaxconn) 1523 backlog = somaxconn; 1524 1525 err = security_socket_listen(sock, backlog); 1526 if (!err) 1527 err = sock->ops->listen(sock, backlog); 1528 1529 fput_light(sock->file, fput_needed); 1530 } 1531 return err; 1532 } 1533 1534 SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1535 { 1536 return __sys_listen(fd, backlog); 1537 } 1538 1539 /* 1540 * For accept, we attempt to create a new socket, set up the link 1541 * with the client, wake up the client, then return the new 1542 * connected fd. We collect the address of the connector in kernel 1543 * space and move it to user at the very end. This is unclean because 1544 * we open the socket then return an error. 1545 * 1546 * 1003.1g adds the ability to recvmsg() to query connection pending 1547 * status to recvmsg. We need to add that support in a way thats 1548 * clean when we restructure accept also. 1549 */ 1550 1551 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, 1552 int __user *upeer_addrlen, int flags) 1553 { 1554 struct socket *sock, *newsock; 1555 struct file *newfile; 1556 int err, len, newfd, fput_needed; 1557 struct sockaddr_storage address; 1558 1559 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1560 return -EINVAL; 1561 1562 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1563 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1564 1565 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1566 if (!sock) 1567 goto out; 1568 1569 err = -ENFILE; 1570 newsock = sock_alloc(); 1571 if (!newsock) 1572 goto out_put; 1573 1574 newsock->type = sock->type; 1575 newsock->ops = sock->ops; 1576 1577 /* 1578 * We don't need try_module_get here, as the listening socket (sock) 1579 * has the protocol module (sock->ops->owner) held. 1580 */ 1581 __module_get(newsock->ops->owner); 1582 1583 newfd = get_unused_fd_flags(flags); 1584 if (unlikely(newfd < 0)) { 1585 err = newfd; 1586 sock_release(newsock); 1587 goto out_put; 1588 } 1589 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); 1590 if (IS_ERR(newfile)) { 1591 err = PTR_ERR(newfile); 1592 put_unused_fd(newfd); 1593 goto out_put; 1594 } 1595 1596 err = security_socket_accept(sock, newsock); 1597 if (err) 1598 goto out_fd; 1599 1600 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false); 1601 if (err < 0) 1602 goto out_fd; 1603 1604 if (upeer_sockaddr) { 1605 len = newsock->ops->getname(newsock, 1606 (struct sockaddr *)&address, 2); 1607 if (len < 0) { 1608 err = -ECONNABORTED; 1609 goto out_fd; 1610 } 1611 err = move_addr_to_user(&address, 1612 len, upeer_sockaddr, upeer_addrlen); 1613 if (err < 0) 1614 goto out_fd; 1615 } 1616 1617 /* File flags are not inherited via accept() unlike another OSes. */ 1618 1619 fd_install(newfd, newfile); 1620 err = newfd; 1621 1622 out_put: 1623 fput_light(sock->file, fput_needed); 1624 out: 1625 return err; 1626 out_fd: 1627 fput(newfile); 1628 put_unused_fd(newfd); 1629 goto out_put; 1630 } 1631 1632 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 1633 int __user *, upeer_addrlen, int, flags) 1634 { 1635 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); 1636 } 1637 1638 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 1639 int __user *, upeer_addrlen) 1640 { 1641 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 1642 } 1643 1644 /* 1645 * Attempt to connect to a socket with the server address. The address 1646 * is in user space so we verify it is OK and move it to kernel space. 1647 * 1648 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 1649 * break bindings 1650 * 1651 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 1652 * other SEQPACKET protocols that take time to connect() as it doesn't 1653 * include the -EINPROGRESS status for such sockets. 1654 */ 1655 1656 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) 1657 { 1658 struct socket *sock; 1659 struct sockaddr_storage address; 1660 int err, fput_needed; 1661 1662 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1663 if (!sock) 1664 goto out; 1665 err = move_addr_to_kernel(uservaddr, addrlen, &address); 1666 if (err < 0) 1667 goto out_put; 1668 1669 err = 1670 security_socket_connect(sock, (struct sockaddr *)&address, addrlen); 1671 if (err) 1672 goto out_put; 1673 1674 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, 1675 sock->file->f_flags); 1676 out_put: 1677 fput_light(sock->file, fput_needed); 1678 out: 1679 return err; 1680 } 1681 1682 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 1683 int, addrlen) 1684 { 1685 return __sys_connect(fd, uservaddr, addrlen); 1686 } 1687 1688 /* 1689 * Get the local address ('name') of a socket object. Move the obtained 1690 * name to user space. 1691 */ 1692 1693 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, 1694 int __user *usockaddr_len) 1695 { 1696 struct socket *sock; 1697 struct sockaddr_storage address; 1698 int err, fput_needed; 1699 1700 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1701 if (!sock) 1702 goto out; 1703 1704 err = security_socket_getsockname(sock); 1705 if (err) 1706 goto out_put; 1707 1708 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0); 1709 if (err < 0) 1710 goto out_put; 1711 /* "err" is actually length in this case */ 1712 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); 1713 1714 out_put: 1715 fput_light(sock->file, fput_needed); 1716 out: 1717 return err; 1718 } 1719 1720 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 1721 int __user *, usockaddr_len) 1722 { 1723 return __sys_getsockname(fd, usockaddr, usockaddr_len); 1724 } 1725 1726 /* 1727 * Get the remote address ('name') of a socket object. Move the obtained 1728 * name to user space. 1729 */ 1730 1731 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, 1732 int __user *usockaddr_len) 1733 { 1734 struct socket *sock; 1735 struct sockaddr_storage address; 1736 int err, fput_needed; 1737 1738 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1739 if (sock != NULL) { 1740 err = security_socket_getpeername(sock); 1741 if (err) { 1742 fput_light(sock->file, fput_needed); 1743 return err; 1744 } 1745 1746 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1); 1747 if (err >= 0) 1748 /* "err" is actually length in this case */ 1749 err = move_addr_to_user(&address, err, usockaddr, 1750 usockaddr_len); 1751 fput_light(sock->file, fput_needed); 1752 } 1753 return err; 1754 } 1755 1756 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 1757 int __user *, usockaddr_len) 1758 { 1759 return __sys_getpeername(fd, usockaddr, usockaddr_len); 1760 } 1761 1762 /* 1763 * Send a datagram to a given address. We move the address into kernel 1764 * space and check the user space data area is readable before invoking 1765 * the protocol. 1766 */ 1767 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, 1768 struct sockaddr __user *addr, int addr_len) 1769 { 1770 struct socket *sock; 1771 struct sockaddr_storage address; 1772 int err; 1773 struct msghdr msg; 1774 struct iovec iov; 1775 int fput_needed; 1776 1777 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter); 1778 if (unlikely(err)) 1779 return err; 1780 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1781 if (!sock) 1782 goto out; 1783 1784 msg.msg_name = NULL; 1785 msg.msg_control = NULL; 1786 msg.msg_controllen = 0; 1787 msg.msg_namelen = 0; 1788 if (addr) { 1789 err = move_addr_to_kernel(addr, addr_len, &address); 1790 if (err < 0) 1791 goto out_put; 1792 msg.msg_name = (struct sockaddr *)&address; 1793 msg.msg_namelen = addr_len; 1794 } 1795 if (sock->file->f_flags & O_NONBLOCK) 1796 flags |= MSG_DONTWAIT; 1797 msg.msg_flags = flags; 1798 err = sock_sendmsg(sock, &msg); 1799 1800 out_put: 1801 fput_light(sock->file, fput_needed); 1802 out: 1803 return err; 1804 } 1805 1806 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 1807 unsigned int, flags, struct sockaddr __user *, addr, 1808 int, addr_len) 1809 { 1810 return __sys_sendto(fd, buff, len, flags, addr, addr_len); 1811 } 1812 1813 /* 1814 * Send a datagram down a socket. 1815 */ 1816 1817 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 1818 unsigned int, flags) 1819 { 1820 return __sys_sendto(fd, buff, len, flags, NULL, 0); 1821 } 1822 1823 /* 1824 * Receive a frame from the socket and optionally record the address of the 1825 * sender. We verify the buffers are writable and if needed move the 1826 * sender address from kernel to user space. 1827 */ 1828 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, 1829 struct sockaddr __user *addr, int __user *addr_len) 1830 { 1831 struct socket *sock; 1832 struct iovec iov; 1833 struct msghdr msg; 1834 struct sockaddr_storage address; 1835 int err, err2; 1836 int fput_needed; 1837 1838 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter); 1839 if (unlikely(err)) 1840 return err; 1841 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1842 if (!sock) 1843 goto out; 1844 1845 msg.msg_control = NULL; 1846 msg.msg_controllen = 0; 1847 /* Save some cycles and don't copy the address if not needed */ 1848 msg.msg_name = addr ? (struct sockaddr *)&address : NULL; 1849 /* We assume all kernel code knows the size of sockaddr_storage */ 1850 msg.msg_namelen = 0; 1851 msg.msg_iocb = NULL; 1852 msg.msg_flags = 0; 1853 if (sock->file->f_flags & O_NONBLOCK) 1854 flags |= MSG_DONTWAIT; 1855 err = sock_recvmsg(sock, &msg, flags); 1856 1857 if (err >= 0 && addr != NULL) { 1858 err2 = move_addr_to_user(&address, 1859 msg.msg_namelen, addr, addr_len); 1860 if (err2 < 0) 1861 err = err2; 1862 } 1863 1864 fput_light(sock->file, fput_needed); 1865 out: 1866 return err; 1867 } 1868 1869 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 1870 unsigned int, flags, struct sockaddr __user *, addr, 1871 int __user *, addr_len) 1872 { 1873 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); 1874 } 1875 1876 /* 1877 * Receive a datagram from a socket. 1878 */ 1879 1880 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, 1881 unsigned int, flags) 1882 { 1883 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 1884 } 1885 1886 /* 1887 * Set a socket option. Because we don't know the option lengths we have 1888 * to pass the user mode parameter for the protocols to sort out. 1889 */ 1890 1891 static int __sys_setsockopt(int fd, int level, int optname, 1892 char __user *optval, int optlen) 1893 { 1894 int err, fput_needed; 1895 struct socket *sock; 1896 1897 if (optlen < 0) 1898 return -EINVAL; 1899 1900 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1901 if (sock != NULL) { 1902 err = security_socket_setsockopt(sock, level, optname); 1903 if (err) 1904 goto out_put; 1905 1906 if (level == SOL_SOCKET) 1907 err = 1908 sock_setsockopt(sock, level, optname, optval, 1909 optlen); 1910 else 1911 err = 1912 sock->ops->setsockopt(sock, level, optname, optval, 1913 optlen); 1914 out_put: 1915 fput_light(sock->file, fput_needed); 1916 } 1917 return err; 1918 } 1919 1920 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 1921 char __user *, optval, int, optlen) 1922 { 1923 return __sys_setsockopt(fd, level, optname, optval, optlen); 1924 } 1925 1926 /* 1927 * Get a socket option. Because we don't know the option lengths we have 1928 * to pass a user mode parameter for the protocols to sort out. 1929 */ 1930 1931 static int __sys_getsockopt(int fd, int level, int optname, 1932 char __user *optval, int __user *optlen) 1933 { 1934 int err, fput_needed; 1935 struct socket *sock; 1936 1937 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1938 if (sock != NULL) { 1939 err = security_socket_getsockopt(sock, level, optname); 1940 if (err) 1941 goto out_put; 1942 1943 if (level == SOL_SOCKET) 1944 err = 1945 sock_getsockopt(sock, level, optname, optval, 1946 optlen); 1947 else 1948 err = 1949 sock->ops->getsockopt(sock, level, optname, optval, 1950 optlen); 1951 out_put: 1952 fput_light(sock->file, fput_needed); 1953 } 1954 return err; 1955 } 1956 1957 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 1958 char __user *, optval, int __user *, optlen) 1959 { 1960 return __sys_getsockopt(fd, level, optname, optval, optlen); 1961 } 1962 1963 /* 1964 * Shutdown a socket. 1965 */ 1966 1967 int __sys_shutdown(int fd, int how) 1968 { 1969 int err, fput_needed; 1970 struct socket *sock; 1971 1972 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1973 if (sock != NULL) { 1974 err = security_socket_shutdown(sock, how); 1975 if (!err) 1976 err = sock->ops->shutdown(sock, how); 1977 fput_light(sock->file, fput_needed); 1978 } 1979 return err; 1980 } 1981 1982 SYSCALL_DEFINE2(shutdown, int, fd, int, how) 1983 { 1984 return __sys_shutdown(fd, how); 1985 } 1986 1987 /* A couple of helpful macros for getting the address of the 32/64 bit 1988 * fields which are the same type (int / unsigned) on our platforms. 1989 */ 1990 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 1991 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 1992 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 1993 1994 struct used_address { 1995 struct sockaddr_storage name; 1996 unsigned int name_len; 1997 }; 1998 1999 static int copy_msghdr_from_user(struct msghdr *kmsg, 2000 struct user_msghdr __user *umsg, 2001 struct sockaddr __user **save_addr, 2002 struct iovec **iov) 2003 { 2004 struct user_msghdr msg; 2005 ssize_t err; 2006 2007 if (copy_from_user(&msg, umsg, sizeof(*umsg))) 2008 return -EFAULT; 2009 2010 kmsg->msg_control = (void __force *)msg.msg_control; 2011 kmsg->msg_controllen = msg.msg_controllen; 2012 kmsg->msg_flags = msg.msg_flags; 2013 2014 kmsg->msg_namelen = msg.msg_namelen; 2015 if (!msg.msg_name) 2016 kmsg->msg_namelen = 0; 2017 2018 if (kmsg->msg_namelen < 0) 2019 return -EINVAL; 2020 2021 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) 2022 kmsg->msg_namelen = sizeof(struct sockaddr_storage); 2023 2024 if (save_addr) 2025 *save_addr = msg.msg_name; 2026 2027 if (msg.msg_name && kmsg->msg_namelen) { 2028 if (!save_addr) { 2029 err = move_addr_to_kernel(msg.msg_name, 2030 kmsg->msg_namelen, 2031 kmsg->msg_name); 2032 if (err < 0) 2033 return err; 2034 } 2035 } else { 2036 kmsg->msg_name = NULL; 2037 kmsg->msg_namelen = 0; 2038 } 2039 2040 if (msg.msg_iovlen > UIO_MAXIOV) 2041 return -EMSGSIZE; 2042 2043 kmsg->msg_iocb = NULL; 2044 2045 return import_iovec(save_addr ? READ : WRITE, 2046 msg.msg_iov, msg.msg_iovlen, 2047 UIO_FASTIOV, iov, &kmsg->msg_iter); 2048 } 2049 2050 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, 2051 struct msghdr *msg_sys, unsigned int flags, 2052 struct used_address *used_address, 2053 unsigned int allowed_msghdr_flags) 2054 { 2055 struct compat_msghdr __user *msg_compat = 2056 (struct compat_msghdr __user *)msg; 2057 struct sockaddr_storage address; 2058 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2059 unsigned char ctl[sizeof(struct cmsghdr) + 20] 2060 __aligned(sizeof(__kernel_size_t)); 2061 /* 20 is size of ipv6_pktinfo */ 2062 unsigned char *ctl_buf = ctl; 2063 int ctl_len; 2064 ssize_t err; 2065 2066 msg_sys->msg_name = &address; 2067 2068 if (MSG_CMSG_COMPAT & flags) 2069 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov); 2070 else 2071 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov); 2072 if (err < 0) 2073 return err; 2074 2075 err = -ENOBUFS; 2076 2077 if (msg_sys->msg_controllen > INT_MAX) 2078 goto out_freeiov; 2079 flags |= (msg_sys->msg_flags & allowed_msghdr_flags); 2080 ctl_len = msg_sys->msg_controllen; 2081 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 2082 err = 2083 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, 2084 sizeof(ctl)); 2085 if (err) 2086 goto out_freeiov; 2087 ctl_buf = msg_sys->msg_control; 2088 ctl_len = msg_sys->msg_controllen; 2089 } else if (ctl_len) { 2090 BUILD_BUG_ON(sizeof(struct cmsghdr) != 2091 CMSG_ALIGN(sizeof(struct cmsghdr))); 2092 if (ctl_len > sizeof(ctl)) { 2093 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 2094 if (ctl_buf == NULL) 2095 goto out_freeiov; 2096 } 2097 err = -EFAULT; 2098 /* 2099 * Careful! Before this, msg_sys->msg_control contains a user pointer. 2100 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted 2101 * checking falls down on this. 2102 */ 2103 if (copy_from_user(ctl_buf, 2104 (void __user __force *)msg_sys->msg_control, 2105 ctl_len)) 2106 goto out_freectl; 2107 msg_sys->msg_control = ctl_buf; 2108 } 2109 msg_sys->msg_flags = flags; 2110 2111 if (sock->file->f_flags & O_NONBLOCK) 2112 msg_sys->msg_flags |= MSG_DONTWAIT; 2113 /* 2114 * If this is sendmmsg() and current destination address is same as 2115 * previously succeeded address, omit asking LSM's decision. 2116 * used_address->name_len is initialized to UINT_MAX so that the first 2117 * destination address never matches. 2118 */ 2119 if (used_address && msg_sys->msg_name && 2120 used_address->name_len == msg_sys->msg_namelen && 2121 !memcmp(&used_address->name, msg_sys->msg_name, 2122 used_address->name_len)) { 2123 err = sock_sendmsg_nosec(sock, msg_sys); 2124 goto out_freectl; 2125 } 2126 err = sock_sendmsg(sock, msg_sys); 2127 /* 2128 * If this is sendmmsg() and sending to current destination address was 2129 * successful, remember it. 2130 */ 2131 if (used_address && err >= 0) { 2132 used_address->name_len = msg_sys->msg_namelen; 2133 if (msg_sys->msg_name) 2134 memcpy(&used_address->name, msg_sys->msg_name, 2135 used_address->name_len); 2136 } 2137 2138 out_freectl: 2139 if (ctl_buf != ctl) 2140 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 2141 out_freeiov: 2142 kfree(iov); 2143 return err; 2144 } 2145 2146 /* 2147 * BSD sendmsg interface 2148 */ 2149 2150 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2151 bool forbid_cmsg_compat) 2152 { 2153 int fput_needed, err; 2154 struct msghdr msg_sys; 2155 struct socket *sock; 2156 2157 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2158 return -EINVAL; 2159 2160 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2161 if (!sock) 2162 goto out; 2163 2164 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); 2165 2166 fput_light(sock->file, fput_needed); 2167 out: 2168 return err; 2169 } 2170 2171 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) 2172 { 2173 return __sys_sendmsg(fd, msg, flags, true); 2174 } 2175 2176 /* 2177 * Linux sendmmsg interface 2178 */ 2179 2180 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2181 unsigned int flags, bool forbid_cmsg_compat) 2182 { 2183 int fput_needed, err, datagrams; 2184 struct socket *sock; 2185 struct mmsghdr __user *entry; 2186 struct compat_mmsghdr __user *compat_entry; 2187 struct msghdr msg_sys; 2188 struct used_address used_address; 2189 unsigned int oflags = flags; 2190 2191 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2192 return -EINVAL; 2193 2194 if (vlen > UIO_MAXIOV) 2195 vlen = UIO_MAXIOV; 2196 2197 datagrams = 0; 2198 2199 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2200 if (!sock) 2201 return err; 2202 2203 used_address.name_len = UINT_MAX; 2204 entry = mmsg; 2205 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2206 err = 0; 2207 flags |= MSG_BATCH; 2208 2209 while (datagrams < vlen) { 2210 if (datagrams == vlen - 1) 2211 flags = oflags; 2212 2213 if (MSG_CMSG_COMPAT & flags) { 2214 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, 2215 &msg_sys, flags, &used_address, MSG_EOR); 2216 if (err < 0) 2217 break; 2218 err = __put_user(err, &compat_entry->msg_len); 2219 ++compat_entry; 2220 } else { 2221 err = ___sys_sendmsg(sock, 2222 (struct user_msghdr __user *)entry, 2223 &msg_sys, flags, &used_address, MSG_EOR); 2224 if (err < 0) 2225 break; 2226 err = put_user(err, &entry->msg_len); 2227 ++entry; 2228 } 2229 2230 if (err) 2231 break; 2232 ++datagrams; 2233 if (msg_data_left(&msg_sys)) 2234 break; 2235 cond_resched(); 2236 } 2237 2238 fput_light(sock->file, fput_needed); 2239 2240 /* We only return an error if no datagrams were able to be sent */ 2241 if (datagrams != 0) 2242 return datagrams; 2243 2244 return err; 2245 } 2246 2247 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, 2248 unsigned int, vlen, unsigned int, flags) 2249 { 2250 return __sys_sendmmsg(fd, mmsg, vlen, flags, true); 2251 } 2252 2253 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, 2254 struct msghdr *msg_sys, unsigned int flags, int nosec) 2255 { 2256 struct compat_msghdr __user *msg_compat = 2257 (struct compat_msghdr __user *)msg; 2258 struct iovec iovstack[UIO_FASTIOV]; 2259 struct iovec *iov = iovstack; 2260 unsigned long cmsg_ptr; 2261 int len; 2262 ssize_t err; 2263 2264 /* kernel mode address */ 2265 struct sockaddr_storage addr; 2266 2267 /* user mode address pointers */ 2268 struct sockaddr __user *uaddr; 2269 int __user *uaddr_len = COMPAT_NAMELEN(msg); 2270 2271 msg_sys->msg_name = &addr; 2272 2273 if (MSG_CMSG_COMPAT & flags) 2274 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov); 2275 else 2276 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov); 2277 if (err < 0) 2278 return err; 2279 2280 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2281 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2282 2283 /* We assume all kernel code knows the size of sockaddr_storage */ 2284 msg_sys->msg_namelen = 0; 2285 2286 if (sock->file->f_flags & O_NONBLOCK) 2287 flags |= MSG_DONTWAIT; 2288 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags); 2289 if (err < 0) 2290 goto out_freeiov; 2291 len = err; 2292 2293 if (uaddr != NULL) { 2294 err = move_addr_to_user(&addr, 2295 msg_sys->msg_namelen, uaddr, 2296 uaddr_len); 2297 if (err < 0) 2298 goto out_freeiov; 2299 } 2300 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2301 COMPAT_FLAGS(msg)); 2302 if (err) 2303 goto out_freeiov; 2304 if (MSG_CMSG_COMPAT & flags) 2305 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2306 &msg_compat->msg_controllen); 2307 else 2308 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2309 &msg->msg_controllen); 2310 if (err) 2311 goto out_freeiov; 2312 err = len; 2313 2314 out_freeiov: 2315 kfree(iov); 2316 return err; 2317 } 2318 2319 /* 2320 * BSD recvmsg interface 2321 */ 2322 2323 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2324 bool forbid_cmsg_compat) 2325 { 2326 int fput_needed, err; 2327 struct msghdr msg_sys; 2328 struct socket *sock; 2329 2330 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2331 return -EINVAL; 2332 2333 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2334 if (!sock) 2335 goto out; 2336 2337 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2338 2339 fput_light(sock->file, fput_needed); 2340 out: 2341 return err; 2342 } 2343 2344 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, 2345 unsigned int, flags) 2346 { 2347 return __sys_recvmsg(fd, msg, flags, true); 2348 } 2349 2350 /* 2351 * Linux recvmmsg interface 2352 */ 2353 2354 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2355 unsigned int flags, struct timespec *timeout) 2356 { 2357 int fput_needed, err, datagrams; 2358 struct socket *sock; 2359 struct mmsghdr __user *entry; 2360 struct compat_mmsghdr __user *compat_entry; 2361 struct msghdr msg_sys; 2362 struct timespec64 end_time; 2363 struct timespec64 timeout64; 2364 2365 if (timeout && 2366 poll_select_set_timeout(&end_time, timeout->tv_sec, 2367 timeout->tv_nsec)) 2368 return -EINVAL; 2369 2370 datagrams = 0; 2371 2372 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2373 if (!sock) 2374 return err; 2375 2376 if (likely(!(flags & MSG_ERRQUEUE))) { 2377 err = sock_error(sock->sk); 2378 if (err) { 2379 datagrams = err; 2380 goto out_put; 2381 } 2382 } 2383 2384 entry = mmsg; 2385 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2386 2387 while (datagrams < vlen) { 2388 /* 2389 * No need to ask LSM for more than the first datagram. 2390 */ 2391 if (MSG_CMSG_COMPAT & flags) { 2392 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, 2393 &msg_sys, flags & ~MSG_WAITFORONE, 2394 datagrams); 2395 if (err < 0) 2396 break; 2397 err = __put_user(err, &compat_entry->msg_len); 2398 ++compat_entry; 2399 } else { 2400 err = ___sys_recvmsg(sock, 2401 (struct user_msghdr __user *)entry, 2402 &msg_sys, flags & ~MSG_WAITFORONE, 2403 datagrams); 2404 if (err < 0) 2405 break; 2406 err = put_user(err, &entry->msg_len); 2407 ++entry; 2408 } 2409 2410 if (err) 2411 break; 2412 ++datagrams; 2413 2414 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2415 if (flags & MSG_WAITFORONE) 2416 flags |= MSG_DONTWAIT; 2417 2418 if (timeout) { 2419 ktime_get_ts64(&timeout64); 2420 *timeout = timespec64_to_timespec( 2421 timespec64_sub(end_time, timeout64)); 2422 if (timeout->tv_sec < 0) { 2423 timeout->tv_sec = timeout->tv_nsec = 0; 2424 break; 2425 } 2426 2427 /* Timeout, return less than vlen datagrams */ 2428 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2429 break; 2430 } 2431 2432 /* Out of band data, return right away */ 2433 if (msg_sys.msg_flags & MSG_OOB) 2434 break; 2435 cond_resched(); 2436 } 2437 2438 if (err == 0) 2439 goto out_put; 2440 2441 if (datagrams == 0) { 2442 datagrams = err; 2443 goto out_put; 2444 } 2445 2446 /* 2447 * We may return less entries than requested (vlen) if the 2448 * sock is non block and there aren't enough datagrams... 2449 */ 2450 if (err != -EAGAIN) { 2451 /* 2452 * ... or if recvmsg returns an error after we 2453 * received some datagrams, where we record the 2454 * error to return on the next call or if the 2455 * app asks about it using getsockopt(SO_ERROR). 2456 */ 2457 sock->sk->sk_err = -err; 2458 } 2459 out_put: 2460 fput_light(sock->file, fput_needed); 2461 2462 return datagrams; 2463 } 2464 2465 static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2466 unsigned int vlen, unsigned int flags, 2467 struct timespec __user *timeout) 2468 { 2469 int datagrams; 2470 struct timespec timeout_sys; 2471 2472 if (flags & MSG_CMSG_COMPAT) 2473 return -EINVAL; 2474 2475 if (!timeout) 2476 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL); 2477 2478 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys))) 2479 return -EFAULT; 2480 2481 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 2482 2483 if (datagrams > 0 && 2484 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys))) 2485 datagrams = -EFAULT; 2486 2487 return datagrams; 2488 } 2489 2490 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 2491 unsigned int, vlen, unsigned int, flags, 2492 struct timespec __user *, timeout) 2493 { 2494 return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout); 2495 } 2496 2497 #ifdef __ARCH_WANT_SYS_SOCKETCALL 2498 /* Argument list sizes for sys_socketcall */ 2499 #define AL(x) ((x) * sizeof(unsigned long)) 2500 static const unsigned char nargs[21] = { 2501 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 2502 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 2503 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 2504 AL(4), AL(5), AL(4) 2505 }; 2506 2507 #undef AL 2508 2509 /* 2510 * System call vectors. 2511 * 2512 * Argument checking cleaned up. Saved 20% in size. 2513 * This function doesn't need to set the kernel lock because 2514 * it is set by the callees. 2515 */ 2516 2517 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 2518 { 2519 unsigned long a[AUDITSC_ARGS]; 2520 unsigned long a0, a1; 2521 int err; 2522 unsigned int len; 2523 2524 if (call < 1 || call > SYS_SENDMMSG) 2525 return -EINVAL; 2526 call = array_index_nospec(call, SYS_SENDMMSG + 1); 2527 2528 len = nargs[call]; 2529 if (len > sizeof(a)) 2530 return -EINVAL; 2531 2532 /* copy_from_user should be SMP safe. */ 2533 if (copy_from_user(a, args, len)) 2534 return -EFAULT; 2535 2536 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); 2537 if (err) 2538 return err; 2539 2540 a0 = a[0]; 2541 a1 = a[1]; 2542 2543 switch (call) { 2544 case SYS_SOCKET: 2545 err = __sys_socket(a0, a1, a[2]); 2546 break; 2547 case SYS_BIND: 2548 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 2549 break; 2550 case SYS_CONNECT: 2551 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 2552 break; 2553 case SYS_LISTEN: 2554 err = __sys_listen(a0, a1); 2555 break; 2556 case SYS_ACCEPT: 2557 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 2558 (int __user *)a[2], 0); 2559 break; 2560 case SYS_GETSOCKNAME: 2561 err = 2562 __sys_getsockname(a0, (struct sockaddr __user *)a1, 2563 (int __user *)a[2]); 2564 break; 2565 case SYS_GETPEERNAME: 2566 err = 2567 __sys_getpeername(a0, (struct sockaddr __user *)a1, 2568 (int __user *)a[2]); 2569 break; 2570 case SYS_SOCKETPAIR: 2571 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 2572 break; 2573 case SYS_SEND: 2574 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 2575 NULL, 0); 2576 break; 2577 case SYS_SENDTO: 2578 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 2579 (struct sockaddr __user *)a[4], a[5]); 2580 break; 2581 case SYS_RECV: 2582 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 2583 NULL, NULL); 2584 break; 2585 case SYS_RECVFROM: 2586 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 2587 (struct sockaddr __user *)a[4], 2588 (int __user *)a[5]); 2589 break; 2590 case SYS_SHUTDOWN: 2591 err = __sys_shutdown(a0, a1); 2592 break; 2593 case SYS_SETSOCKOPT: 2594 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], 2595 a[4]); 2596 break; 2597 case SYS_GETSOCKOPT: 2598 err = 2599 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 2600 (int __user *)a[4]); 2601 break; 2602 case SYS_SENDMSG: 2603 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, 2604 a[2], true); 2605 break; 2606 case SYS_SENDMMSG: 2607 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], 2608 a[3], true); 2609 break; 2610 case SYS_RECVMSG: 2611 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, 2612 a[2], true); 2613 break; 2614 case SYS_RECVMMSG: 2615 err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], 2616 a[3], (struct timespec __user *)a[4]); 2617 break; 2618 case SYS_ACCEPT4: 2619 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 2620 (int __user *)a[2], a[3]); 2621 break; 2622 default: 2623 err = -EINVAL; 2624 break; 2625 } 2626 return err; 2627 } 2628 2629 #endif /* __ARCH_WANT_SYS_SOCKETCALL */ 2630 2631 /** 2632 * sock_register - add a socket protocol handler 2633 * @ops: description of protocol 2634 * 2635 * This function is called by a protocol handler that wants to 2636 * advertise its address family, and have it linked into the 2637 * socket interface. The value ops->family corresponds to the 2638 * socket system call protocol family. 2639 */ 2640 int sock_register(const struct net_proto_family *ops) 2641 { 2642 int err; 2643 2644 if (ops->family >= NPROTO) { 2645 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); 2646 return -ENOBUFS; 2647 } 2648 2649 spin_lock(&net_family_lock); 2650 if (rcu_dereference_protected(net_families[ops->family], 2651 lockdep_is_held(&net_family_lock))) 2652 err = -EEXIST; 2653 else { 2654 rcu_assign_pointer(net_families[ops->family], ops); 2655 err = 0; 2656 } 2657 spin_unlock(&net_family_lock); 2658 2659 pr_info("NET: Registered protocol family %d\n", ops->family); 2660 return err; 2661 } 2662 EXPORT_SYMBOL(sock_register); 2663 2664 /** 2665 * sock_unregister - remove a protocol handler 2666 * @family: protocol family to remove 2667 * 2668 * This function is called by a protocol handler that wants to 2669 * remove its address family, and have it unlinked from the 2670 * new socket creation. 2671 * 2672 * If protocol handler is a module, then it can use module reference 2673 * counts to protect against new references. If protocol handler is not 2674 * a module then it needs to provide its own protection in 2675 * the ops->create routine. 2676 */ 2677 void sock_unregister(int family) 2678 { 2679 BUG_ON(family < 0 || family >= NPROTO); 2680 2681 spin_lock(&net_family_lock); 2682 RCU_INIT_POINTER(net_families[family], NULL); 2683 spin_unlock(&net_family_lock); 2684 2685 synchronize_rcu(); 2686 2687 pr_info("NET: Unregistered protocol family %d\n", family); 2688 } 2689 EXPORT_SYMBOL(sock_unregister); 2690 2691 bool sock_is_registered(int family) 2692 { 2693 return family < NPROTO && rcu_access_pointer(net_families[family]); 2694 } 2695 2696 static int __init sock_init(void) 2697 { 2698 int err; 2699 /* 2700 * Initialize the network sysctl infrastructure. 2701 */ 2702 err = net_sysctl_init(); 2703 if (err) 2704 goto out; 2705 2706 /* 2707 * Initialize skbuff SLAB cache 2708 */ 2709 skb_init(); 2710 2711 /* 2712 * Initialize the protocols module. 2713 */ 2714 2715 init_inodecache(); 2716 2717 err = register_filesystem(&sock_fs_type); 2718 if (err) 2719 goto out_fs; 2720 sock_mnt = kern_mount(&sock_fs_type); 2721 if (IS_ERR(sock_mnt)) { 2722 err = PTR_ERR(sock_mnt); 2723 goto out_mount; 2724 } 2725 2726 /* The real protocol initialization is performed in later initcalls. 2727 */ 2728 2729 #ifdef CONFIG_NETFILTER 2730 err = netfilter_init(); 2731 if (err) 2732 goto out; 2733 #endif 2734 2735 ptp_classifier_init(); 2736 2737 out: 2738 return err; 2739 2740 out_mount: 2741 unregister_filesystem(&sock_fs_type); 2742 out_fs: 2743 goto out; 2744 } 2745 2746 core_initcall(sock_init); /* early initcall */ 2747 2748 #ifdef CONFIG_PROC_FS 2749 void socket_seq_show(struct seq_file *seq) 2750 { 2751 seq_printf(seq, "sockets: used %d\n", 2752 sock_inuse_get(seq->private)); 2753 } 2754 #endif /* CONFIG_PROC_FS */ 2755 2756 #ifdef CONFIG_COMPAT 2757 static int do_siocgstamp(struct net *net, struct socket *sock, 2758 unsigned int cmd, void __user *up) 2759 { 2760 mm_segment_t old_fs = get_fs(); 2761 struct timeval ktv; 2762 int err; 2763 2764 set_fs(KERNEL_DS); 2765 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv); 2766 set_fs(old_fs); 2767 if (!err) 2768 err = compat_put_timeval(&ktv, up); 2769 2770 return err; 2771 } 2772 2773 static int do_siocgstampns(struct net *net, struct socket *sock, 2774 unsigned int cmd, void __user *up) 2775 { 2776 mm_segment_t old_fs = get_fs(); 2777 struct timespec kts; 2778 int err; 2779 2780 set_fs(KERNEL_DS); 2781 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts); 2782 set_fs(old_fs); 2783 if (!err) 2784 err = compat_put_timespec(&kts, up); 2785 2786 return err; 2787 } 2788 2789 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32) 2790 { 2791 struct compat_ifconf ifc32; 2792 struct ifconf ifc; 2793 int err; 2794 2795 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf))) 2796 return -EFAULT; 2797 2798 ifc.ifc_len = ifc32.ifc_len; 2799 ifc.ifc_req = compat_ptr(ifc32.ifcbuf); 2800 2801 rtnl_lock(); 2802 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq)); 2803 rtnl_unlock(); 2804 if (err) 2805 return err; 2806 2807 ifc32.ifc_len = ifc.ifc_len; 2808 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf))) 2809 return -EFAULT; 2810 2811 return 0; 2812 } 2813 2814 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32) 2815 { 2816 struct compat_ethtool_rxnfc __user *compat_rxnfc; 2817 bool convert_in = false, convert_out = false; 2818 size_t buf_size = 0; 2819 struct ethtool_rxnfc __user *rxnfc = NULL; 2820 struct ifreq ifr; 2821 u32 rule_cnt = 0, actual_rule_cnt; 2822 u32 ethcmd; 2823 u32 data; 2824 int ret; 2825 2826 if (get_user(data, &ifr32->ifr_ifru.ifru_data)) 2827 return -EFAULT; 2828 2829 compat_rxnfc = compat_ptr(data); 2830 2831 if (get_user(ethcmd, &compat_rxnfc->cmd)) 2832 return -EFAULT; 2833 2834 /* Most ethtool structures are defined without padding. 2835 * Unfortunately struct ethtool_rxnfc is an exception. 2836 */ 2837 switch (ethcmd) { 2838 default: 2839 break; 2840 case ETHTOOL_GRXCLSRLALL: 2841 /* Buffer size is variable */ 2842 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt)) 2843 return -EFAULT; 2844 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32)) 2845 return -ENOMEM; 2846 buf_size += rule_cnt * sizeof(u32); 2847 /* fall through */ 2848 case ETHTOOL_GRXRINGS: 2849 case ETHTOOL_GRXCLSRLCNT: 2850 case ETHTOOL_GRXCLSRULE: 2851 case ETHTOOL_SRXCLSRLINS: 2852 convert_out = true; 2853 /* fall through */ 2854 case ETHTOOL_SRXCLSRLDEL: 2855 buf_size += sizeof(struct ethtool_rxnfc); 2856 convert_in = true; 2857 rxnfc = compat_alloc_user_space(buf_size); 2858 break; 2859 } 2860 2861 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ)) 2862 return -EFAULT; 2863 2864 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc; 2865 2866 if (convert_in) { 2867 /* We expect there to be holes between fs.m_ext and 2868 * fs.ring_cookie and at the end of fs, but nowhere else. 2869 */ 2870 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) + 2871 sizeof(compat_rxnfc->fs.m_ext) != 2872 offsetof(struct ethtool_rxnfc, fs.m_ext) + 2873 sizeof(rxnfc->fs.m_ext)); 2874 BUILD_BUG_ON( 2875 offsetof(struct compat_ethtool_rxnfc, fs.location) - 2876 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) != 2877 offsetof(struct ethtool_rxnfc, fs.location) - 2878 offsetof(struct ethtool_rxnfc, fs.ring_cookie)); 2879 2880 if (copy_in_user(rxnfc, compat_rxnfc, 2881 (void __user *)(&rxnfc->fs.m_ext + 1) - 2882 (void __user *)rxnfc) || 2883 copy_in_user(&rxnfc->fs.ring_cookie, 2884 &compat_rxnfc->fs.ring_cookie, 2885 (void __user *)(&rxnfc->fs.location + 1) - 2886 (void __user *)&rxnfc->fs.ring_cookie) || 2887 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt, 2888 sizeof(rxnfc->rule_cnt))) 2889 return -EFAULT; 2890 } 2891 2892 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL); 2893 if (ret) 2894 return ret; 2895 2896 if (convert_out) { 2897 if (copy_in_user(compat_rxnfc, rxnfc, 2898 (const void __user *)(&rxnfc->fs.m_ext + 1) - 2899 (const void __user *)rxnfc) || 2900 copy_in_user(&compat_rxnfc->fs.ring_cookie, 2901 &rxnfc->fs.ring_cookie, 2902 (const void __user *)(&rxnfc->fs.location + 1) - 2903 (const void __user *)&rxnfc->fs.ring_cookie) || 2904 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt, 2905 sizeof(rxnfc->rule_cnt))) 2906 return -EFAULT; 2907 2908 if (ethcmd == ETHTOOL_GRXCLSRLALL) { 2909 /* As an optimisation, we only copy the actual 2910 * number of rules that the underlying 2911 * function returned. Since Mallory might 2912 * change the rule count in user memory, we 2913 * check that it is less than the rule count 2914 * originally given (as the user buffer size), 2915 * which has been range-checked. 2916 */ 2917 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt)) 2918 return -EFAULT; 2919 if (actual_rule_cnt < rule_cnt) 2920 rule_cnt = actual_rule_cnt; 2921 if (copy_in_user(&compat_rxnfc->rule_locs[0], 2922 &rxnfc->rule_locs[0], 2923 rule_cnt * sizeof(u32))) 2924 return -EFAULT; 2925 } 2926 } 2927 2928 return 0; 2929 } 2930 2931 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 2932 { 2933 compat_uptr_t uptr32; 2934 struct ifreq ifr; 2935 void __user *saved; 2936 int err; 2937 2938 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq))) 2939 return -EFAULT; 2940 2941 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 2942 return -EFAULT; 2943 2944 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; 2945 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); 2946 2947 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL); 2948 if (!err) { 2949 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; 2950 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq))) 2951 err = -EFAULT; 2952 } 2953 return err; 2954 } 2955 2956 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ 2957 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, 2958 struct compat_ifreq __user *u_ifreq32) 2959 { 2960 struct ifreq ifreq; 2961 u32 data32; 2962 2963 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ)) 2964 return -EFAULT; 2965 if (get_user(data32, &u_ifreq32->ifr_data)) 2966 return -EFAULT; 2967 ifreq.ifr_data = compat_ptr(data32); 2968 2969 return dev_ioctl(net, cmd, &ifreq, NULL); 2970 } 2971 2972 static int compat_sioc_ifmap(struct net *net, unsigned int cmd, 2973 struct compat_ifreq __user *uifr32) 2974 { 2975 struct ifreq ifr; 2976 struct compat_ifmap __user *uifmap32; 2977 int err; 2978 2979 uifmap32 = &uifr32->ifr_ifru.ifru_map; 2980 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name)); 2981 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 2982 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 2983 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 2984 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq); 2985 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma); 2986 err |= get_user(ifr.ifr_map.port, &uifmap32->port); 2987 if (err) 2988 return -EFAULT; 2989 2990 err = dev_ioctl(net, cmd, &ifr, NULL); 2991 2992 if (cmd == SIOCGIFMAP && !err) { 2993 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name)); 2994 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 2995 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 2996 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 2997 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq); 2998 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma); 2999 err |= put_user(ifr.ifr_map.port, &uifmap32->port); 3000 if (err) 3001 err = -EFAULT; 3002 } 3003 return err; 3004 } 3005 3006 struct rtentry32 { 3007 u32 rt_pad1; 3008 struct sockaddr rt_dst; /* target address */ 3009 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */ 3010 struct sockaddr rt_genmask; /* target network mask (IP) */ 3011 unsigned short rt_flags; 3012 short rt_pad2; 3013 u32 rt_pad3; 3014 unsigned char rt_tos; 3015 unsigned char rt_class; 3016 short rt_pad4; 3017 short rt_metric; /* +1 for binary compatibility! */ 3018 /* char * */ u32 rt_dev; /* forcing the device at add */ 3019 u32 rt_mtu; /* per route MTU/Window */ 3020 u32 rt_window; /* Window clamping */ 3021 unsigned short rt_irtt; /* Initial RTT */ 3022 }; 3023 3024 struct in6_rtmsg32 { 3025 struct in6_addr rtmsg_dst; 3026 struct in6_addr rtmsg_src; 3027 struct in6_addr rtmsg_gateway; 3028 u32 rtmsg_type; 3029 u16 rtmsg_dst_len; 3030 u16 rtmsg_src_len; 3031 u32 rtmsg_metric; 3032 u32 rtmsg_info; 3033 u32 rtmsg_flags; 3034 s32 rtmsg_ifindex; 3035 }; 3036 3037 static int routing_ioctl(struct net *net, struct socket *sock, 3038 unsigned int cmd, void __user *argp) 3039 { 3040 int ret; 3041 void *r = NULL; 3042 struct in6_rtmsg r6; 3043 struct rtentry r4; 3044 char devname[16]; 3045 u32 rtdev; 3046 mm_segment_t old_fs = get_fs(); 3047 3048 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */ 3049 struct in6_rtmsg32 __user *ur6 = argp; 3050 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst), 3051 3 * sizeof(struct in6_addr)); 3052 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type)); 3053 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len)); 3054 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len)); 3055 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric)); 3056 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info)); 3057 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags)); 3058 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex)); 3059 3060 r = (void *) &r6; 3061 } else { /* ipv4 */ 3062 struct rtentry32 __user *ur4 = argp; 3063 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst), 3064 3 * sizeof(struct sockaddr)); 3065 ret |= get_user(r4.rt_flags, &(ur4->rt_flags)); 3066 ret |= get_user(r4.rt_metric, &(ur4->rt_metric)); 3067 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu)); 3068 ret |= get_user(r4.rt_window, &(ur4->rt_window)); 3069 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt)); 3070 ret |= get_user(rtdev, &(ur4->rt_dev)); 3071 if (rtdev) { 3072 ret |= copy_from_user(devname, compat_ptr(rtdev), 15); 3073 r4.rt_dev = (char __user __force *)devname; 3074 devname[15] = 0; 3075 } else 3076 r4.rt_dev = NULL; 3077 3078 r = (void *) &r4; 3079 } 3080 3081 if (ret) { 3082 ret = -EFAULT; 3083 goto out; 3084 } 3085 3086 set_fs(KERNEL_DS); 3087 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r); 3088 set_fs(old_fs); 3089 3090 out: 3091 return ret; 3092 } 3093 3094 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE 3095 * for some operations; this forces use of the newer bridge-utils that 3096 * use compatible ioctls 3097 */ 3098 static int old_bridge_ioctl(compat_ulong_t __user *argp) 3099 { 3100 compat_ulong_t tmp; 3101 3102 if (get_user(tmp, argp)) 3103 return -EFAULT; 3104 if (tmp == BRCTL_GET_VERSION) 3105 return BRCTL_VERSION + 1; 3106 return -EINVAL; 3107 } 3108 3109 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 3110 unsigned int cmd, unsigned long arg) 3111 { 3112 void __user *argp = compat_ptr(arg); 3113 struct sock *sk = sock->sk; 3114 struct net *net = sock_net(sk); 3115 3116 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 3117 return compat_ifr_data_ioctl(net, cmd, argp); 3118 3119 switch (cmd) { 3120 case SIOCSIFBR: 3121 case SIOCGIFBR: 3122 return old_bridge_ioctl(argp); 3123 case SIOCGIFCONF: 3124 return compat_dev_ifconf(net, argp); 3125 case SIOCETHTOOL: 3126 return ethtool_ioctl(net, argp); 3127 case SIOCWANDEV: 3128 return compat_siocwandev(net, argp); 3129 case SIOCGIFMAP: 3130 case SIOCSIFMAP: 3131 return compat_sioc_ifmap(net, cmd, argp); 3132 case SIOCADDRT: 3133 case SIOCDELRT: 3134 return routing_ioctl(net, sock, cmd, argp); 3135 case SIOCGSTAMP: 3136 return do_siocgstamp(net, sock, cmd, argp); 3137 case SIOCGSTAMPNS: 3138 return do_siocgstampns(net, sock, cmd, argp); 3139 case SIOCBONDSLAVEINFOQUERY: 3140 case SIOCBONDINFOQUERY: 3141 case SIOCSHWTSTAMP: 3142 case SIOCGHWTSTAMP: 3143 return compat_ifr_data_ioctl(net, cmd, argp); 3144 3145 case FIOSETOWN: 3146 case SIOCSPGRP: 3147 case FIOGETOWN: 3148 case SIOCGPGRP: 3149 case SIOCBRADDBR: 3150 case SIOCBRDELBR: 3151 case SIOCGIFVLAN: 3152 case SIOCSIFVLAN: 3153 case SIOCADDDLCI: 3154 case SIOCDELDLCI: 3155 case SIOCGSKNS: 3156 return sock_ioctl(file, cmd, arg); 3157 3158 case SIOCGIFFLAGS: 3159 case SIOCSIFFLAGS: 3160 case SIOCGIFMETRIC: 3161 case SIOCSIFMETRIC: 3162 case SIOCGIFMTU: 3163 case SIOCSIFMTU: 3164 case SIOCGIFMEM: 3165 case SIOCSIFMEM: 3166 case SIOCGIFHWADDR: 3167 case SIOCSIFHWADDR: 3168 case SIOCADDMULTI: 3169 case SIOCDELMULTI: 3170 case SIOCGIFINDEX: 3171 case SIOCGIFADDR: 3172 case SIOCSIFADDR: 3173 case SIOCSIFHWBROADCAST: 3174 case SIOCDIFADDR: 3175 case SIOCGIFBRDADDR: 3176 case SIOCSIFBRDADDR: 3177 case SIOCGIFDSTADDR: 3178 case SIOCSIFDSTADDR: 3179 case SIOCGIFNETMASK: 3180 case SIOCSIFNETMASK: 3181 case SIOCSIFPFLAGS: 3182 case SIOCGIFPFLAGS: 3183 case SIOCGIFTXQLEN: 3184 case SIOCSIFTXQLEN: 3185 case SIOCBRADDIF: 3186 case SIOCBRDELIF: 3187 case SIOCSIFNAME: 3188 case SIOCGMIIPHY: 3189 case SIOCGMIIREG: 3190 case SIOCSMIIREG: 3191 case SIOCSARP: 3192 case SIOCGARP: 3193 case SIOCDARP: 3194 case SIOCATMARK: 3195 case SIOCBONDENSLAVE: 3196 case SIOCBONDRELEASE: 3197 case SIOCBONDSETHWADDR: 3198 case SIOCBONDCHANGEACTIVE: 3199 case SIOCGIFNAME: 3200 return sock_do_ioctl(net, sock, cmd, arg); 3201 } 3202 3203 return -ENOIOCTLCMD; 3204 } 3205 3206 static long compat_sock_ioctl(struct file *file, unsigned int cmd, 3207 unsigned long arg) 3208 { 3209 struct socket *sock = file->private_data; 3210 int ret = -ENOIOCTLCMD; 3211 struct sock *sk; 3212 struct net *net; 3213 3214 sk = sock->sk; 3215 net = sock_net(sk); 3216 3217 if (sock->ops->compat_ioctl) 3218 ret = sock->ops->compat_ioctl(sock, cmd, arg); 3219 3220 if (ret == -ENOIOCTLCMD && 3221 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 3222 ret = compat_wext_handle_ioctl(net, cmd, arg); 3223 3224 if (ret == -ENOIOCTLCMD) 3225 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 3226 3227 return ret; 3228 } 3229 #endif 3230 3231 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3232 { 3233 return sock->ops->bind(sock, addr, addrlen); 3234 } 3235 EXPORT_SYMBOL(kernel_bind); 3236 3237 int kernel_listen(struct socket *sock, int backlog) 3238 { 3239 return sock->ops->listen(sock, backlog); 3240 } 3241 EXPORT_SYMBOL(kernel_listen); 3242 3243 int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3244 { 3245 struct sock *sk = sock->sk; 3246 int err; 3247 3248 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3249 newsock); 3250 if (err < 0) 3251 goto done; 3252 3253 err = sock->ops->accept(sock, *newsock, flags, true); 3254 if (err < 0) { 3255 sock_release(*newsock); 3256 *newsock = NULL; 3257 goto done; 3258 } 3259 3260 (*newsock)->ops = sock->ops; 3261 __module_get((*newsock)->ops->owner); 3262 3263 done: 3264 return err; 3265 } 3266 EXPORT_SYMBOL(kernel_accept); 3267 3268 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3269 int flags) 3270 { 3271 return sock->ops->connect(sock, addr, addrlen, flags); 3272 } 3273 EXPORT_SYMBOL(kernel_connect); 3274 3275 int kernel_getsockname(struct socket *sock, struct sockaddr *addr) 3276 { 3277 return sock->ops->getname(sock, addr, 0); 3278 } 3279 EXPORT_SYMBOL(kernel_getsockname); 3280 3281 int kernel_getpeername(struct socket *sock, struct sockaddr *addr) 3282 { 3283 return sock->ops->getname(sock, addr, 1); 3284 } 3285 EXPORT_SYMBOL(kernel_getpeername); 3286 3287 int kernel_getsockopt(struct socket *sock, int level, int optname, 3288 char *optval, int *optlen) 3289 { 3290 mm_segment_t oldfs = get_fs(); 3291 char __user *uoptval; 3292 int __user *uoptlen; 3293 int err; 3294 3295 uoptval = (char __user __force *) optval; 3296 uoptlen = (int __user __force *) optlen; 3297 3298 set_fs(KERNEL_DS); 3299 if (level == SOL_SOCKET) 3300 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen); 3301 else 3302 err = sock->ops->getsockopt(sock, level, optname, uoptval, 3303 uoptlen); 3304 set_fs(oldfs); 3305 return err; 3306 } 3307 EXPORT_SYMBOL(kernel_getsockopt); 3308 3309 int kernel_setsockopt(struct socket *sock, int level, int optname, 3310 char *optval, unsigned int optlen) 3311 { 3312 mm_segment_t oldfs = get_fs(); 3313 char __user *uoptval; 3314 int err; 3315 3316 uoptval = (char __user __force *) optval; 3317 3318 set_fs(KERNEL_DS); 3319 if (level == SOL_SOCKET) 3320 err = sock_setsockopt(sock, level, optname, uoptval, optlen); 3321 else 3322 err = sock->ops->setsockopt(sock, level, optname, uoptval, 3323 optlen); 3324 set_fs(oldfs); 3325 return err; 3326 } 3327 EXPORT_SYMBOL(kernel_setsockopt); 3328 3329 int kernel_sendpage(struct socket *sock, struct page *page, int offset, 3330 size_t size, int flags) 3331 { 3332 if (sock->ops->sendpage) 3333 return sock->ops->sendpage(sock, page, offset, size, flags); 3334 3335 return sock_no_sendpage(sock, page, offset, size, flags); 3336 } 3337 EXPORT_SYMBOL(kernel_sendpage); 3338 3339 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset, 3340 size_t size, int flags) 3341 { 3342 struct socket *sock = sk->sk_socket; 3343 3344 if (sock->ops->sendpage_locked) 3345 return sock->ops->sendpage_locked(sk, page, offset, size, 3346 flags); 3347 3348 return sock_no_sendpage_locked(sk, page, offset, size, flags); 3349 } 3350 EXPORT_SYMBOL(kernel_sendpage_locked); 3351 3352 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3353 { 3354 return sock->ops->shutdown(sock, how); 3355 } 3356 EXPORT_SYMBOL(kernel_sock_shutdown); 3357 3358 /* This routine returns the IP overhead imposed by a socket i.e. 3359 * the length of the underlying IP header, depending on whether 3360 * this is an IPv4 or IPv6 socket and the length from IP options turned 3361 * on at the socket. Assumes that the caller has a lock on the socket. 3362 */ 3363 u32 kernel_sock_ip_overhead(struct sock *sk) 3364 { 3365 struct inet_sock *inet; 3366 struct ip_options_rcu *opt; 3367 u32 overhead = 0; 3368 #if IS_ENABLED(CONFIG_IPV6) 3369 struct ipv6_pinfo *np; 3370 struct ipv6_txoptions *optv6 = NULL; 3371 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3372 3373 if (!sk) 3374 return overhead; 3375 3376 switch (sk->sk_family) { 3377 case AF_INET: 3378 inet = inet_sk(sk); 3379 overhead += sizeof(struct iphdr); 3380 opt = rcu_dereference_protected(inet->inet_opt, 3381 sock_owned_by_user(sk)); 3382 if (opt) 3383 overhead += opt->opt.optlen; 3384 return overhead; 3385 #if IS_ENABLED(CONFIG_IPV6) 3386 case AF_INET6: 3387 np = inet6_sk(sk); 3388 overhead += sizeof(struct ipv6hdr); 3389 if (np) 3390 optv6 = rcu_dereference_protected(np->opt, 3391 sock_owned_by_user(sk)); 3392 if (optv6) 3393 overhead += (optv6->opt_flen + optv6->opt_nflen); 3394 return overhead; 3395 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3396 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ 3397 return overhead; 3398 } 3399 } 3400 EXPORT_SYMBOL(kernel_sock_ip_overhead); 3401