1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * NET An implementation of the SOCKET network access protocol. 4 * 5 * Version: @(#)socket.c 1.1.93 18/02/95 6 * 7 * Authors: Orest Zborowski, <obz@Kodak.COM> 8 * Ross Biro 9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 10 * 11 * Fixes: 12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in 13 * shutdown() 14 * Alan Cox : verify_area() fixes 15 * Alan Cox : Removed DDI 16 * Jonathan Kamens : SOCK_DGRAM reconnect bug 17 * Alan Cox : Moved a load of checks to the very 18 * top level. 19 * Alan Cox : Move address structures to/from user 20 * mode above the protocol layers. 21 * Rob Janssen : Allow 0 length sends. 22 * Alan Cox : Asynchronous I/O support (cribbed from the 23 * tty drivers). 24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style) 25 * Jeff Uphoff : Made max number of sockets command-line 26 * configurable. 27 * Matti Aarnio : Made the number of sockets dynamic, 28 * to be allocated when needed, and mr. 29 * Uphoff's max is used as max to be 30 * allowed to allocate. 31 * Linus : Argh. removed all the socket allocation 32 * altogether: it's in the inode now. 33 * Alan Cox : Made sock_alloc()/sock_release() public 34 * for NetROM and future kernel nfsd type 35 * stuff. 36 * Alan Cox : sendmsg/recvmsg basics. 37 * Tom Dyas : Export net symbols. 38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n". 39 * Alan Cox : Added thread locking to sys_* calls 40 * for sockets. May have errors at the 41 * moment. 42 * Kevin Buhr : Fixed the dumb errors in the above. 43 * Andi Kleen : Some small cleanups, optimizations, 44 * and fixed a copy_from_user() bug. 45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0) 46 * Tigran Aivazian : Made listen(2) backlog sanity checks 47 * protocol-independent 48 * 49 * This module is effectively the top level interface to the BSD socket 50 * paradigm. 51 * 52 * Based upon Swansea University Computer Society NET3.039 53 */ 54 55 #include <linux/mm.h> 56 #include <linux/socket.h> 57 #include <linux/file.h> 58 #include <linux/net.h> 59 #include <linux/interrupt.h> 60 #include <linux/thread_info.h> 61 #include <linux/rcupdate.h> 62 #include <linux/netdevice.h> 63 #include <linux/proc_fs.h> 64 #include <linux/seq_file.h> 65 #include <linux/mutex.h> 66 #include <linux/if_bridge.h> 67 #include <linux/if_frad.h> 68 #include <linux/if_vlan.h> 69 #include <linux/ptp_classify.h> 70 #include <linux/init.h> 71 #include <linux/poll.h> 72 #include <linux/cache.h> 73 #include <linux/module.h> 74 #include <linux/highmem.h> 75 #include <linux/mount.h> 76 #include <linux/pseudo_fs.h> 77 #include <linux/security.h> 78 #include <linux/syscalls.h> 79 #include <linux/compat.h> 80 #include <linux/kmod.h> 81 #include <linux/audit.h> 82 #include <linux/wireless.h> 83 #include <linux/nsproxy.h> 84 #include <linux/magic.h> 85 #include <linux/slab.h> 86 #include <linux/xattr.h> 87 #include <linux/nospec.h> 88 #include <linux/indirect_call_wrapper.h> 89 90 #include <linux/uaccess.h> 91 #include <asm/unistd.h> 92 93 #include <net/compat.h> 94 #include <net/wext.h> 95 #include <net/cls_cgroup.h> 96 97 #include <net/sock.h> 98 #include <linux/netfilter.h> 99 100 #include <linux/if_tun.h> 101 #include <linux/ipv6_route.h> 102 #include <linux/route.h> 103 #include <linux/termios.h> 104 #include <linux/sockios.h> 105 #include <net/busy_poll.h> 106 #include <linux/errqueue.h> 107 108 #ifdef CONFIG_NET_RX_BUSY_POLL 109 unsigned int sysctl_net_busy_read __read_mostly; 110 unsigned int sysctl_net_busy_poll __read_mostly; 111 #endif 112 113 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to); 114 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from); 115 static int sock_mmap(struct file *file, struct vm_area_struct *vma); 116 117 static int sock_close(struct inode *inode, struct file *file); 118 static __poll_t sock_poll(struct file *file, 119 struct poll_table_struct *wait); 120 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 121 #ifdef CONFIG_COMPAT 122 static long compat_sock_ioctl(struct file *file, 123 unsigned int cmd, unsigned long arg); 124 #endif 125 static int sock_fasync(int fd, struct file *filp, int on); 126 static ssize_t sock_sendpage(struct file *file, struct page *page, 127 int offset, size_t size, loff_t *ppos, int more); 128 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 129 struct pipe_inode_info *pipe, size_t len, 130 unsigned int flags); 131 132 #ifdef CONFIG_PROC_FS 133 static void sock_show_fdinfo(struct seq_file *m, struct file *f) 134 { 135 struct socket *sock = f->private_data; 136 137 if (sock->ops->show_fdinfo) 138 sock->ops->show_fdinfo(m, sock); 139 } 140 #else 141 #define sock_show_fdinfo NULL 142 #endif 143 144 /* 145 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear 146 * in the operation structures but are done directly via the socketcall() multiplexor. 147 */ 148 149 static const struct file_operations socket_file_ops = { 150 .owner = THIS_MODULE, 151 .llseek = no_llseek, 152 .read_iter = sock_read_iter, 153 .write_iter = sock_write_iter, 154 .poll = sock_poll, 155 .unlocked_ioctl = sock_ioctl, 156 #ifdef CONFIG_COMPAT 157 .compat_ioctl = compat_sock_ioctl, 158 #endif 159 .mmap = sock_mmap, 160 .release = sock_close, 161 .fasync = sock_fasync, 162 .sendpage = sock_sendpage, 163 .splice_write = generic_splice_sendpage, 164 .splice_read = sock_splice_read, 165 .show_fdinfo = sock_show_fdinfo, 166 }; 167 168 /* 169 * The protocol list. Each protocol is registered in here. 170 */ 171 172 static DEFINE_SPINLOCK(net_family_lock); 173 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; 174 175 /* 176 * Support routines. 177 * Move socket addresses back and forth across the kernel/user 178 * divide and look after the messy bits. 179 */ 180 181 /** 182 * move_addr_to_kernel - copy a socket address into kernel space 183 * @uaddr: Address in user space 184 * @kaddr: Address in kernel space 185 * @ulen: Length in user space 186 * 187 * The address is copied into kernel space. If the provided address is 188 * too long an error code of -EINVAL is returned. If the copy gives 189 * invalid addresses -EFAULT is returned. On a success 0 is returned. 190 */ 191 192 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr) 193 { 194 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) 195 return -EINVAL; 196 if (ulen == 0) 197 return 0; 198 if (copy_from_user(kaddr, uaddr, ulen)) 199 return -EFAULT; 200 return audit_sockaddr(ulen, kaddr); 201 } 202 203 /** 204 * move_addr_to_user - copy an address to user space 205 * @kaddr: kernel space address 206 * @klen: length of address in kernel 207 * @uaddr: user space address 208 * @ulen: pointer to user length field 209 * 210 * The value pointed to by ulen on entry is the buffer length available. 211 * This is overwritten with the buffer space used. -EINVAL is returned 212 * if an overlong buffer is specified or a negative buffer size. -EFAULT 213 * is returned if either the buffer or the length field are not 214 * accessible. 215 * After copying the data up to the limit the user specifies, the true 216 * length of the data is written over the length limit the user 217 * specified. Zero is returned for a success. 218 */ 219 220 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen, 221 void __user *uaddr, int __user *ulen) 222 { 223 int err; 224 int len; 225 226 BUG_ON(klen > sizeof(struct sockaddr_storage)); 227 err = get_user(len, ulen); 228 if (err) 229 return err; 230 if (len > klen) 231 len = klen; 232 if (len < 0) 233 return -EINVAL; 234 if (len) { 235 if (audit_sockaddr(klen, kaddr)) 236 return -ENOMEM; 237 if (copy_to_user(uaddr, kaddr, len)) 238 return -EFAULT; 239 } 240 /* 241 * "fromlen shall refer to the value before truncation.." 242 * 1003.1g 243 */ 244 return __put_user(klen, ulen); 245 } 246 247 static struct kmem_cache *sock_inode_cachep __ro_after_init; 248 249 static struct inode *sock_alloc_inode(struct super_block *sb) 250 { 251 struct socket_alloc *ei; 252 253 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); 254 if (!ei) 255 return NULL; 256 init_waitqueue_head(&ei->socket.wq.wait); 257 ei->socket.wq.fasync_list = NULL; 258 ei->socket.wq.flags = 0; 259 260 ei->socket.state = SS_UNCONNECTED; 261 ei->socket.flags = 0; 262 ei->socket.ops = NULL; 263 ei->socket.sk = NULL; 264 ei->socket.file = NULL; 265 266 return &ei->vfs_inode; 267 } 268 269 static void sock_free_inode(struct inode *inode) 270 { 271 struct socket_alloc *ei; 272 273 ei = container_of(inode, struct socket_alloc, vfs_inode); 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 .free_inode = sock_free_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 int sockfs_init_fs_context(struct fs_context *fc) 357 { 358 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC); 359 if (!ctx) 360 return -ENOMEM; 361 ctx->ops = &sockfs_ops; 362 ctx->dops = &sockfs_dentry_operations; 363 ctx->xattr = sockfs_xattr_handlers; 364 return 0; 365 } 366 367 static struct vfsmount *sock_mnt __read_mostly; 368 369 static struct file_system_type sock_fs_type = { 370 .name = "sockfs", 371 .init_fs_context = sockfs_init_fs_context, 372 .kill_sb = kill_anon_super, 373 }; 374 375 /* 376 * Obtains the first available file descriptor and sets it up for use. 377 * 378 * These functions create file structures and maps them to fd space 379 * of the current process. On success it returns file descriptor 380 * and file struct implicitly stored in sock->file. 381 * Note that another thread may close file descriptor before we return 382 * from this function. We use the fact that now we do not refer 383 * to socket after mapping. If one day we will need it, this 384 * function will increment ref. count on file by 1. 385 * 386 * In any case returned fd MAY BE not valid! 387 * This race condition is unavoidable 388 * with shared fd spaces, we cannot solve it inside kernel, 389 * but we take care of internal coherence yet. 390 */ 391 392 /** 393 * sock_alloc_file - Bind a &socket to a &file 394 * @sock: socket 395 * @flags: file status flags 396 * @dname: protocol name 397 * 398 * Returns the &file bound with @sock, implicitly storing it 399 * in sock->file. If dname is %NULL, sets to "". 400 * On failure the return is a ERR pointer (see linux/err.h). 401 * This function uses GFP_KERNEL internally. 402 */ 403 404 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname) 405 { 406 struct file *file; 407 408 if (!dname) 409 dname = sock->sk ? sock->sk->sk_prot_creator->name : ""; 410 411 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname, 412 O_RDWR | (flags & O_NONBLOCK), 413 &socket_file_ops); 414 if (IS_ERR(file)) { 415 sock_release(sock); 416 return file; 417 } 418 419 sock->file = file; 420 file->private_data = sock; 421 stream_open(SOCK_INODE(sock), file); 422 return file; 423 } 424 EXPORT_SYMBOL(sock_alloc_file); 425 426 static int sock_map_fd(struct socket *sock, int flags) 427 { 428 struct file *newfile; 429 int fd = get_unused_fd_flags(flags); 430 if (unlikely(fd < 0)) { 431 sock_release(sock); 432 return fd; 433 } 434 435 newfile = sock_alloc_file(sock, flags, NULL); 436 if (!IS_ERR(newfile)) { 437 fd_install(fd, newfile); 438 return fd; 439 } 440 441 put_unused_fd(fd); 442 return PTR_ERR(newfile); 443 } 444 445 /** 446 * sock_from_file - Return the &socket bounded to @file. 447 * @file: file 448 * @err: pointer to an error code return 449 * 450 * On failure returns %NULL and assigns -ENOTSOCK to @err. 451 */ 452 453 struct socket *sock_from_file(struct file *file, int *err) 454 { 455 if (file->f_op == &socket_file_ops) 456 return file->private_data; /* set in sock_map_fd */ 457 458 *err = -ENOTSOCK; 459 return NULL; 460 } 461 EXPORT_SYMBOL(sock_from_file); 462 463 /** 464 * sockfd_lookup - Go from a file number to its socket slot 465 * @fd: file handle 466 * @err: pointer to an error code return 467 * 468 * The file handle passed in is locked and the socket it is bound 469 * to is returned. If an error occurs the err pointer is overwritten 470 * with a negative errno code and NULL is returned. The function checks 471 * for both invalid handles and passing a handle which is not a socket. 472 * 473 * On a success the socket object pointer is returned. 474 */ 475 476 struct socket *sockfd_lookup(int fd, int *err) 477 { 478 struct file *file; 479 struct socket *sock; 480 481 file = fget(fd); 482 if (!file) { 483 *err = -EBADF; 484 return NULL; 485 } 486 487 sock = sock_from_file(file, err); 488 if (!sock) 489 fput(file); 490 return sock; 491 } 492 EXPORT_SYMBOL(sockfd_lookup); 493 494 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) 495 { 496 struct fd f = fdget(fd); 497 struct socket *sock; 498 499 *err = -EBADF; 500 if (f.file) { 501 sock = sock_from_file(f.file, err); 502 if (likely(sock)) { 503 *fput_needed = f.flags; 504 return sock; 505 } 506 fdput(f); 507 } 508 return NULL; 509 } 510 511 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer, 512 size_t size) 513 { 514 ssize_t len; 515 ssize_t used = 0; 516 517 len = security_inode_listsecurity(d_inode(dentry), buffer, size); 518 if (len < 0) 519 return len; 520 used += len; 521 if (buffer) { 522 if (size < used) 523 return -ERANGE; 524 buffer += len; 525 } 526 527 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1); 528 used += len; 529 if (buffer) { 530 if (size < used) 531 return -ERANGE; 532 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len); 533 buffer += len; 534 } 535 536 return used; 537 } 538 539 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr) 540 { 541 int err = simple_setattr(dentry, iattr); 542 543 if (!err && (iattr->ia_valid & ATTR_UID)) { 544 struct socket *sock = SOCKET_I(d_inode(dentry)); 545 546 if (sock->sk) 547 sock->sk->sk_uid = iattr->ia_uid; 548 else 549 err = -ENOENT; 550 } 551 552 return err; 553 } 554 555 static const struct inode_operations sockfs_inode_ops = { 556 .listxattr = sockfs_listxattr, 557 .setattr = sockfs_setattr, 558 }; 559 560 /** 561 * sock_alloc - allocate a socket 562 * 563 * Allocate a new inode and socket object. The two are bound together 564 * and initialised. The socket is then returned. If we are out of inodes 565 * NULL is returned. This functions uses GFP_KERNEL internally. 566 */ 567 568 struct socket *sock_alloc(void) 569 { 570 struct inode *inode; 571 struct socket *sock; 572 573 inode = new_inode_pseudo(sock_mnt->mnt_sb); 574 if (!inode) 575 return NULL; 576 577 sock = SOCKET_I(inode); 578 579 inode->i_ino = get_next_ino(); 580 inode->i_mode = S_IFSOCK | S_IRWXUGO; 581 inode->i_uid = current_fsuid(); 582 inode->i_gid = current_fsgid(); 583 inode->i_op = &sockfs_inode_ops; 584 585 return sock; 586 } 587 EXPORT_SYMBOL(sock_alloc); 588 589 /** 590 * sock_release - close a socket 591 * @sock: socket to close 592 * 593 * The socket is released from the protocol stack if it has a release 594 * callback, and the inode is then released if the socket is bound to 595 * an inode not a file. 596 */ 597 598 static void __sock_release(struct socket *sock, struct inode *inode) 599 { 600 if (sock->ops) { 601 struct module *owner = sock->ops->owner; 602 603 if (inode) 604 inode_lock(inode); 605 sock->ops->release(sock); 606 sock->sk = NULL; 607 if (inode) 608 inode_unlock(inode); 609 sock->ops = NULL; 610 module_put(owner); 611 } 612 613 if (sock->wq.fasync_list) 614 pr_err("%s: fasync list not empty!\n", __func__); 615 616 if (!sock->file) { 617 iput(SOCK_INODE(sock)); 618 return; 619 } 620 sock->file = NULL; 621 } 622 623 void sock_release(struct socket *sock) 624 { 625 __sock_release(sock, NULL); 626 } 627 EXPORT_SYMBOL(sock_release); 628 629 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags) 630 { 631 u8 flags = *tx_flags; 632 633 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) 634 flags |= SKBTX_HW_TSTAMP; 635 636 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE) 637 flags |= SKBTX_SW_TSTAMP; 638 639 if (tsflags & SOF_TIMESTAMPING_TX_SCHED) 640 flags |= SKBTX_SCHED_TSTAMP; 641 642 *tx_flags = flags; 643 } 644 EXPORT_SYMBOL(__sock_tx_timestamp); 645 646 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *, 647 size_t)); 648 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *, 649 size_t)); 650 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) 651 { 652 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg, 653 inet_sendmsg, sock, msg, 654 msg_data_left(msg)); 655 BUG_ON(ret == -EIOCBQUEUED); 656 return ret; 657 } 658 659 /** 660 * sock_sendmsg - send a message through @sock 661 * @sock: socket 662 * @msg: message to send 663 * 664 * Sends @msg through @sock, passing through LSM. 665 * Returns the number of bytes sent, or an error code. 666 */ 667 int sock_sendmsg(struct socket *sock, struct msghdr *msg) 668 { 669 int err = security_socket_sendmsg(sock, msg, 670 msg_data_left(msg)); 671 672 return err ?: sock_sendmsg_nosec(sock, msg); 673 } 674 EXPORT_SYMBOL(sock_sendmsg); 675 676 /** 677 * kernel_sendmsg - send a message through @sock (kernel-space) 678 * @sock: socket 679 * @msg: message header 680 * @vec: kernel vec 681 * @num: vec array length 682 * @size: total message data size 683 * 684 * Builds the message data with @vec and sends it through @sock. 685 * Returns the number of bytes sent, or an error code. 686 */ 687 688 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, 689 struct kvec *vec, size_t num, size_t size) 690 { 691 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size); 692 return sock_sendmsg(sock, msg); 693 } 694 EXPORT_SYMBOL(kernel_sendmsg); 695 696 /** 697 * kernel_sendmsg_locked - send a message through @sock (kernel-space) 698 * @sk: sock 699 * @msg: message header 700 * @vec: output s/g array 701 * @num: output s/g array length 702 * @size: total message data size 703 * 704 * Builds the message data with @vec and sends it through @sock. 705 * Returns the number of bytes sent, or an error code. 706 * Caller must hold @sk. 707 */ 708 709 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg, 710 struct kvec *vec, size_t num, size_t size) 711 { 712 struct socket *sock = sk->sk_socket; 713 714 if (!sock->ops->sendmsg_locked) 715 return sock_no_sendmsg_locked(sk, msg, size); 716 717 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size); 718 719 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg)); 720 } 721 EXPORT_SYMBOL(kernel_sendmsg_locked); 722 723 static bool skb_is_err_queue(const struct sk_buff *skb) 724 { 725 /* pkt_type of skbs enqueued on the error queue are set to 726 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do 727 * in recvmsg, since skbs received on a local socket will never 728 * have a pkt_type of PACKET_OUTGOING. 729 */ 730 return skb->pkt_type == PACKET_OUTGOING; 731 } 732 733 /* On transmit, software and hardware timestamps are returned independently. 734 * As the two skb clones share the hardware timestamp, which may be updated 735 * before the software timestamp is received, a hardware TX timestamp may be 736 * returned only if there is no software TX timestamp. Ignore false software 737 * timestamps, which may be made in the __sock_recv_timestamp() call when the 738 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a 739 * hardware timestamp. 740 */ 741 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp) 742 { 743 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb); 744 } 745 746 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb) 747 { 748 struct scm_ts_pktinfo ts_pktinfo; 749 struct net_device *orig_dev; 750 751 if (!skb_mac_header_was_set(skb)) 752 return; 753 754 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo)); 755 756 rcu_read_lock(); 757 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 758 if (orig_dev) 759 ts_pktinfo.if_index = orig_dev->ifindex; 760 rcu_read_unlock(); 761 762 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb); 763 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO, 764 sizeof(ts_pktinfo), &ts_pktinfo); 765 } 766 767 /* 768 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) 769 */ 770 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 771 struct sk_buff *skb) 772 { 773 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); 774 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW); 775 struct scm_timestamping_internal tss; 776 777 int empty = 1, false_tstamp = 0; 778 struct skb_shared_hwtstamps *shhwtstamps = 779 skb_hwtstamps(skb); 780 781 /* Race occurred between timestamp enabling and packet 782 receiving. Fill in the current time for now. */ 783 if (need_software_tstamp && skb->tstamp == 0) { 784 __net_timestamp(skb); 785 false_tstamp = 1; 786 } 787 788 if (need_software_tstamp) { 789 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { 790 if (new_tstamp) { 791 struct __kernel_sock_timeval tv; 792 793 skb_get_new_timestamp(skb, &tv); 794 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW, 795 sizeof(tv), &tv); 796 } else { 797 struct __kernel_old_timeval tv; 798 799 skb_get_timestamp(skb, &tv); 800 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD, 801 sizeof(tv), &tv); 802 } 803 } else { 804 if (new_tstamp) { 805 struct __kernel_timespec ts; 806 807 skb_get_new_timestampns(skb, &ts); 808 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW, 809 sizeof(ts), &ts); 810 } else { 811 struct __kernel_old_timespec ts; 812 813 skb_get_timestampns(skb, &ts); 814 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD, 815 sizeof(ts), &ts); 816 } 817 } 818 } 819 820 memset(&tss, 0, sizeof(tss)); 821 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) && 822 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0)) 823 empty = 0; 824 if (shhwtstamps && 825 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) && 826 !skb_is_swtx_tstamp(skb, false_tstamp) && 827 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) { 828 empty = 0; 829 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) && 830 !skb_is_err_queue(skb)) 831 put_ts_pktinfo(msg, skb); 832 } 833 if (!empty) { 834 if (sock_flag(sk, SOCK_TSTAMP_NEW)) 835 put_cmsg_scm_timestamping64(msg, &tss); 836 else 837 put_cmsg_scm_timestamping(msg, &tss); 838 839 if (skb_is_err_queue(skb) && skb->len && 840 SKB_EXT_ERR(skb)->opt_stats) 841 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS, 842 skb->len, skb->data); 843 } 844 } 845 EXPORT_SYMBOL_GPL(__sock_recv_timestamp); 846 847 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, 848 struct sk_buff *skb) 849 { 850 int ack; 851 852 if (!sock_flag(sk, SOCK_WIFI_STATUS)) 853 return; 854 if (!skb->wifi_acked_valid) 855 return; 856 857 ack = skb->wifi_acked; 858 859 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); 860 } 861 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); 862 863 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, 864 struct sk_buff *skb) 865 { 866 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount) 867 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, 868 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount); 869 } 870 871 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 872 struct sk_buff *skb) 873 { 874 sock_recv_timestamp(msg, sk, skb); 875 sock_recv_drops(msg, sk, skb); 876 } 877 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops); 878 879 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *, 880 size_t, int)); 881 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *, 882 size_t, int)); 883 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 884 int flags) 885 { 886 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg, 887 inet_recvmsg, sock, msg, msg_data_left(msg), 888 flags); 889 } 890 891 /** 892 * sock_recvmsg - receive a message from @sock 893 * @sock: socket 894 * @msg: message to receive 895 * @flags: message flags 896 * 897 * Receives @msg from @sock, passing through LSM. Returns the total number 898 * of bytes received, or an error. 899 */ 900 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) 901 { 902 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); 903 904 return err ?: sock_recvmsg_nosec(sock, msg, flags); 905 } 906 EXPORT_SYMBOL(sock_recvmsg); 907 908 /** 909 * kernel_recvmsg - Receive a message from a socket (kernel space) 910 * @sock: The socket to receive the message from 911 * @msg: Received message 912 * @vec: Input s/g array for message data 913 * @num: Size of input s/g array 914 * @size: Number of bytes to read 915 * @flags: Message flags (MSG_DONTWAIT, etc...) 916 * 917 * On return the msg structure contains the scatter/gather array passed in the 918 * vec argument. The array is modified so that it consists of the unfilled 919 * portion of the original array. 920 * 921 * The returned value is the total number of bytes received, or an error. 922 */ 923 924 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 925 struct kvec *vec, size_t num, size_t size, int flags) 926 { 927 mm_segment_t oldfs = get_fs(); 928 int result; 929 930 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size); 931 set_fs(KERNEL_DS); 932 result = sock_recvmsg(sock, msg, flags); 933 set_fs(oldfs); 934 return result; 935 } 936 EXPORT_SYMBOL(kernel_recvmsg); 937 938 static ssize_t sock_sendpage(struct file *file, struct page *page, 939 int offset, size_t size, loff_t *ppos, int more) 940 { 941 struct socket *sock; 942 int flags; 943 944 sock = file->private_data; 945 946 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 947 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */ 948 flags |= more; 949 950 return kernel_sendpage(sock, page, offset, size, flags); 951 } 952 953 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 954 struct pipe_inode_info *pipe, size_t len, 955 unsigned int flags) 956 { 957 struct socket *sock = file->private_data; 958 959 if (unlikely(!sock->ops->splice_read)) 960 return generic_file_splice_read(file, ppos, pipe, len, flags); 961 962 return sock->ops->splice_read(sock, ppos, pipe, len, flags); 963 } 964 965 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) 966 { 967 struct file *file = iocb->ki_filp; 968 struct socket *sock = file->private_data; 969 struct msghdr msg = {.msg_iter = *to, 970 .msg_iocb = iocb}; 971 ssize_t res; 972 973 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 974 msg.msg_flags = MSG_DONTWAIT; 975 976 if (iocb->ki_pos != 0) 977 return -ESPIPE; 978 979 if (!iov_iter_count(to)) /* Match SYS5 behaviour */ 980 return 0; 981 982 res = sock_recvmsg(sock, &msg, msg.msg_flags); 983 *to = msg.msg_iter; 984 return res; 985 } 986 987 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) 988 { 989 struct file *file = iocb->ki_filp; 990 struct socket *sock = file->private_data; 991 struct msghdr msg = {.msg_iter = *from, 992 .msg_iocb = iocb}; 993 ssize_t res; 994 995 if (iocb->ki_pos != 0) 996 return -ESPIPE; 997 998 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 999 msg.msg_flags = MSG_DONTWAIT; 1000 1001 if (sock->type == SOCK_SEQPACKET) 1002 msg.msg_flags |= MSG_EOR; 1003 1004 res = sock_sendmsg(sock, &msg); 1005 *from = msg.msg_iter; 1006 return res; 1007 } 1008 1009 /* 1010 * Atomic setting of ioctl hooks to avoid race 1011 * with module unload. 1012 */ 1013 1014 static DEFINE_MUTEX(br_ioctl_mutex); 1015 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg); 1016 1017 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) 1018 { 1019 mutex_lock(&br_ioctl_mutex); 1020 br_ioctl_hook = hook; 1021 mutex_unlock(&br_ioctl_mutex); 1022 } 1023 EXPORT_SYMBOL(brioctl_set); 1024 1025 static DEFINE_MUTEX(vlan_ioctl_mutex); 1026 static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 1027 1028 void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 1029 { 1030 mutex_lock(&vlan_ioctl_mutex); 1031 vlan_ioctl_hook = hook; 1032 mutex_unlock(&vlan_ioctl_mutex); 1033 } 1034 EXPORT_SYMBOL(vlan_ioctl_set); 1035 1036 static DEFINE_MUTEX(dlci_ioctl_mutex); 1037 static int (*dlci_ioctl_hook) (unsigned int, void __user *); 1038 1039 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) 1040 { 1041 mutex_lock(&dlci_ioctl_mutex); 1042 dlci_ioctl_hook = hook; 1043 mutex_unlock(&dlci_ioctl_mutex); 1044 } 1045 EXPORT_SYMBOL(dlci_ioctl_set); 1046 1047 static long sock_do_ioctl(struct net *net, struct socket *sock, 1048 unsigned int cmd, unsigned long arg) 1049 { 1050 int err; 1051 void __user *argp = (void __user *)arg; 1052 1053 err = sock->ops->ioctl(sock, cmd, arg); 1054 1055 /* 1056 * If this ioctl is unknown try to hand it down 1057 * to the NIC driver. 1058 */ 1059 if (err != -ENOIOCTLCMD) 1060 return err; 1061 1062 if (cmd == SIOCGIFCONF) { 1063 struct ifconf ifc; 1064 if (copy_from_user(&ifc, argp, sizeof(struct ifconf))) 1065 return -EFAULT; 1066 rtnl_lock(); 1067 err = dev_ifconf(net, &ifc, sizeof(struct ifreq)); 1068 rtnl_unlock(); 1069 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf))) 1070 err = -EFAULT; 1071 } else { 1072 struct ifreq ifr; 1073 bool need_copyout; 1074 if (copy_from_user(&ifr, argp, sizeof(struct ifreq))) 1075 return -EFAULT; 1076 err = dev_ioctl(net, cmd, &ifr, &need_copyout); 1077 if (!err && need_copyout) 1078 if (copy_to_user(argp, &ifr, sizeof(struct ifreq))) 1079 return -EFAULT; 1080 } 1081 return err; 1082 } 1083 1084 /* 1085 * With an ioctl, arg may well be a user mode pointer, but we don't know 1086 * what to do with it - that's up to the protocol still. 1087 */ 1088 1089 /** 1090 * get_net_ns - increment the refcount of the network namespace 1091 * @ns: common namespace (net) 1092 * 1093 * Returns the net's common namespace. 1094 */ 1095 1096 struct ns_common *get_net_ns(struct ns_common *ns) 1097 { 1098 return &get_net(container_of(ns, struct net, ns))->ns; 1099 } 1100 EXPORT_SYMBOL_GPL(get_net_ns); 1101 1102 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 1103 { 1104 struct socket *sock; 1105 struct sock *sk; 1106 void __user *argp = (void __user *)arg; 1107 int pid, err; 1108 struct net *net; 1109 1110 sock = file->private_data; 1111 sk = sock->sk; 1112 net = sock_net(sk); 1113 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { 1114 struct ifreq ifr; 1115 bool need_copyout; 1116 if (copy_from_user(&ifr, argp, sizeof(struct ifreq))) 1117 return -EFAULT; 1118 err = dev_ioctl(net, cmd, &ifr, &need_copyout); 1119 if (!err && need_copyout) 1120 if (copy_to_user(argp, &ifr, sizeof(struct ifreq))) 1121 return -EFAULT; 1122 } else 1123 #ifdef CONFIG_WEXT_CORE 1124 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 1125 err = wext_handle_ioctl(net, cmd, argp); 1126 } else 1127 #endif 1128 switch (cmd) { 1129 case FIOSETOWN: 1130 case SIOCSPGRP: 1131 err = -EFAULT; 1132 if (get_user(pid, (int __user *)argp)) 1133 break; 1134 err = f_setown(sock->file, pid, 1); 1135 break; 1136 case FIOGETOWN: 1137 case SIOCGPGRP: 1138 err = put_user(f_getown(sock->file), 1139 (int __user *)argp); 1140 break; 1141 case SIOCGIFBR: 1142 case SIOCSIFBR: 1143 case SIOCBRADDBR: 1144 case SIOCBRDELBR: 1145 err = -ENOPKG; 1146 if (!br_ioctl_hook) 1147 request_module("bridge"); 1148 1149 mutex_lock(&br_ioctl_mutex); 1150 if (br_ioctl_hook) 1151 err = br_ioctl_hook(net, cmd, argp); 1152 mutex_unlock(&br_ioctl_mutex); 1153 break; 1154 case SIOCGIFVLAN: 1155 case SIOCSIFVLAN: 1156 err = -ENOPKG; 1157 if (!vlan_ioctl_hook) 1158 request_module("8021q"); 1159 1160 mutex_lock(&vlan_ioctl_mutex); 1161 if (vlan_ioctl_hook) 1162 err = vlan_ioctl_hook(net, argp); 1163 mutex_unlock(&vlan_ioctl_mutex); 1164 break; 1165 case SIOCADDDLCI: 1166 case SIOCDELDLCI: 1167 err = -ENOPKG; 1168 if (!dlci_ioctl_hook) 1169 request_module("dlci"); 1170 1171 mutex_lock(&dlci_ioctl_mutex); 1172 if (dlci_ioctl_hook) 1173 err = dlci_ioctl_hook(cmd, argp); 1174 mutex_unlock(&dlci_ioctl_mutex); 1175 break; 1176 case SIOCGSKNS: 1177 err = -EPERM; 1178 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1179 break; 1180 1181 err = open_related_ns(&net->ns, get_net_ns); 1182 break; 1183 case SIOCGSTAMP_OLD: 1184 case SIOCGSTAMPNS_OLD: 1185 if (!sock->ops->gettstamp) { 1186 err = -ENOIOCTLCMD; 1187 break; 1188 } 1189 err = sock->ops->gettstamp(sock, argp, 1190 cmd == SIOCGSTAMP_OLD, 1191 !IS_ENABLED(CONFIG_64BIT)); 1192 break; 1193 case SIOCGSTAMP_NEW: 1194 case SIOCGSTAMPNS_NEW: 1195 if (!sock->ops->gettstamp) { 1196 err = -ENOIOCTLCMD; 1197 break; 1198 } 1199 err = sock->ops->gettstamp(sock, argp, 1200 cmd == SIOCGSTAMP_NEW, 1201 false); 1202 break; 1203 default: 1204 err = sock_do_ioctl(net, sock, cmd, arg); 1205 break; 1206 } 1207 return err; 1208 } 1209 1210 /** 1211 * sock_create_lite - creates a socket 1212 * @family: protocol family (AF_INET, ...) 1213 * @type: communication type (SOCK_STREAM, ...) 1214 * @protocol: protocol (0, ...) 1215 * @res: new socket 1216 * 1217 * Creates a new socket and assigns it to @res, passing through LSM. 1218 * The new socket initialization is not complete, see kernel_accept(). 1219 * Returns 0 or an error. On failure @res is set to %NULL. 1220 * This function internally uses GFP_KERNEL. 1221 */ 1222 1223 int sock_create_lite(int family, int type, int protocol, struct socket **res) 1224 { 1225 int err; 1226 struct socket *sock = NULL; 1227 1228 err = security_socket_create(family, type, protocol, 1); 1229 if (err) 1230 goto out; 1231 1232 sock = sock_alloc(); 1233 if (!sock) { 1234 err = -ENOMEM; 1235 goto out; 1236 } 1237 1238 sock->type = type; 1239 err = security_socket_post_create(sock, family, type, protocol, 1); 1240 if (err) 1241 goto out_release; 1242 1243 out: 1244 *res = sock; 1245 return err; 1246 out_release: 1247 sock_release(sock); 1248 sock = NULL; 1249 goto out; 1250 } 1251 EXPORT_SYMBOL(sock_create_lite); 1252 1253 /* No kernel lock held - perfect */ 1254 static __poll_t sock_poll(struct file *file, poll_table *wait) 1255 { 1256 struct socket *sock = file->private_data; 1257 __poll_t events = poll_requested_events(wait), flag = 0; 1258 1259 if (!sock->ops->poll) 1260 return 0; 1261 1262 if (sk_can_busy_loop(sock->sk)) { 1263 /* poll once if requested by the syscall */ 1264 if (events & POLL_BUSY_LOOP) 1265 sk_busy_loop(sock->sk, 1); 1266 1267 /* if this socket can poll_ll, tell the system call */ 1268 flag = POLL_BUSY_LOOP; 1269 } 1270 1271 return sock->ops->poll(file, sock, wait) | flag; 1272 } 1273 1274 static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1275 { 1276 struct socket *sock = file->private_data; 1277 1278 return sock->ops->mmap(file, sock, vma); 1279 } 1280 1281 static int sock_close(struct inode *inode, struct file *filp) 1282 { 1283 __sock_release(SOCKET_I(inode), inode); 1284 return 0; 1285 } 1286 1287 /* 1288 * Update the socket async list 1289 * 1290 * Fasync_list locking strategy. 1291 * 1292 * 1. fasync_list is modified only under process context socket lock 1293 * i.e. under semaphore. 1294 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1295 * or under socket lock 1296 */ 1297 1298 static int sock_fasync(int fd, struct file *filp, int on) 1299 { 1300 struct socket *sock = filp->private_data; 1301 struct sock *sk = sock->sk; 1302 struct socket_wq *wq = &sock->wq; 1303 1304 if (sk == NULL) 1305 return -EINVAL; 1306 1307 lock_sock(sk); 1308 fasync_helper(fd, filp, on, &wq->fasync_list); 1309 1310 if (!wq->fasync_list) 1311 sock_reset_flag(sk, SOCK_FASYNC); 1312 else 1313 sock_set_flag(sk, SOCK_FASYNC); 1314 1315 release_sock(sk); 1316 return 0; 1317 } 1318 1319 /* This function may be called only under rcu_lock */ 1320 1321 int sock_wake_async(struct socket_wq *wq, int how, int band) 1322 { 1323 if (!wq || !wq->fasync_list) 1324 return -1; 1325 1326 switch (how) { 1327 case SOCK_WAKE_WAITD: 1328 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) 1329 break; 1330 goto call_kill; 1331 case SOCK_WAKE_SPACE: 1332 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) 1333 break; 1334 /* fall through */ 1335 case SOCK_WAKE_IO: 1336 call_kill: 1337 kill_fasync(&wq->fasync_list, SIGIO, band); 1338 break; 1339 case SOCK_WAKE_URG: 1340 kill_fasync(&wq->fasync_list, SIGURG, band); 1341 } 1342 1343 return 0; 1344 } 1345 EXPORT_SYMBOL(sock_wake_async); 1346 1347 /** 1348 * __sock_create - creates a socket 1349 * @net: net namespace 1350 * @family: protocol family (AF_INET, ...) 1351 * @type: communication type (SOCK_STREAM, ...) 1352 * @protocol: protocol (0, ...) 1353 * @res: new socket 1354 * @kern: boolean for kernel space sockets 1355 * 1356 * Creates a new socket and assigns it to @res, passing through LSM. 1357 * Returns 0 or an error. On failure @res is set to %NULL. @kern must 1358 * be set to true if the socket resides in kernel space. 1359 * This function internally uses GFP_KERNEL. 1360 */ 1361 1362 int __sock_create(struct net *net, int family, int type, int protocol, 1363 struct socket **res, int kern) 1364 { 1365 int err; 1366 struct socket *sock; 1367 const struct net_proto_family *pf; 1368 1369 /* 1370 * Check protocol is in range 1371 */ 1372 if (family < 0 || family >= NPROTO) 1373 return -EAFNOSUPPORT; 1374 if (type < 0 || type >= SOCK_MAX) 1375 return -EINVAL; 1376 1377 /* Compatibility. 1378 1379 This uglymoron is moved from INET layer to here to avoid 1380 deadlock in module load. 1381 */ 1382 if (family == PF_INET && type == SOCK_PACKET) { 1383 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1384 current->comm); 1385 family = PF_PACKET; 1386 } 1387 1388 err = security_socket_create(family, type, protocol, kern); 1389 if (err) 1390 return err; 1391 1392 /* 1393 * Allocate the socket and allow the family to set things up. if 1394 * the protocol is 0, the family is instructed to select an appropriate 1395 * default. 1396 */ 1397 sock = sock_alloc(); 1398 if (!sock) { 1399 net_warn_ratelimited("socket: no more sockets\n"); 1400 return -ENFILE; /* Not exactly a match, but its the 1401 closest posix thing */ 1402 } 1403 1404 sock->type = type; 1405 1406 #ifdef CONFIG_MODULES 1407 /* Attempt to load a protocol module if the find failed. 1408 * 1409 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1410 * requested real, full-featured networking support upon configuration. 1411 * Otherwise module support will break! 1412 */ 1413 if (rcu_access_pointer(net_families[family]) == NULL) 1414 request_module("net-pf-%d", family); 1415 #endif 1416 1417 rcu_read_lock(); 1418 pf = rcu_dereference(net_families[family]); 1419 err = -EAFNOSUPPORT; 1420 if (!pf) 1421 goto out_release; 1422 1423 /* 1424 * We will call the ->create function, that possibly is in a loadable 1425 * module, so we have to bump that loadable module refcnt first. 1426 */ 1427 if (!try_module_get(pf->owner)) 1428 goto out_release; 1429 1430 /* Now protected by module ref count */ 1431 rcu_read_unlock(); 1432 1433 err = pf->create(net, sock, protocol, kern); 1434 if (err < 0) 1435 goto out_module_put; 1436 1437 /* 1438 * Now to bump the refcnt of the [loadable] module that owns this 1439 * socket at sock_release time we decrement its refcnt. 1440 */ 1441 if (!try_module_get(sock->ops->owner)) 1442 goto out_module_busy; 1443 1444 /* 1445 * Now that we're done with the ->create function, the [loadable] 1446 * module can have its refcnt decremented 1447 */ 1448 module_put(pf->owner); 1449 err = security_socket_post_create(sock, family, type, protocol, kern); 1450 if (err) 1451 goto out_sock_release; 1452 *res = sock; 1453 1454 return 0; 1455 1456 out_module_busy: 1457 err = -EAFNOSUPPORT; 1458 out_module_put: 1459 sock->ops = NULL; 1460 module_put(pf->owner); 1461 out_sock_release: 1462 sock_release(sock); 1463 return err; 1464 1465 out_release: 1466 rcu_read_unlock(); 1467 goto out_sock_release; 1468 } 1469 EXPORT_SYMBOL(__sock_create); 1470 1471 /** 1472 * sock_create - creates a socket 1473 * @family: protocol family (AF_INET, ...) 1474 * @type: communication type (SOCK_STREAM, ...) 1475 * @protocol: protocol (0, ...) 1476 * @res: new socket 1477 * 1478 * A wrapper around __sock_create(). 1479 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1480 */ 1481 1482 int sock_create(int family, int type, int protocol, struct socket **res) 1483 { 1484 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1485 } 1486 EXPORT_SYMBOL(sock_create); 1487 1488 /** 1489 * sock_create_kern - creates a socket (kernel space) 1490 * @net: net namespace 1491 * @family: protocol family (AF_INET, ...) 1492 * @type: communication type (SOCK_STREAM, ...) 1493 * @protocol: protocol (0, ...) 1494 * @res: new socket 1495 * 1496 * A wrapper around __sock_create(). 1497 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1498 */ 1499 1500 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) 1501 { 1502 return __sock_create(net, family, type, protocol, res, 1); 1503 } 1504 EXPORT_SYMBOL(sock_create_kern); 1505 1506 int __sys_socket(int family, int type, int protocol) 1507 { 1508 int retval; 1509 struct socket *sock; 1510 int flags; 1511 1512 /* Check the SOCK_* constants for consistency. */ 1513 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1514 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1515 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1516 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1517 1518 flags = type & ~SOCK_TYPE_MASK; 1519 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1520 return -EINVAL; 1521 type &= SOCK_TYPE_MASK; 1522 1523 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1524 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1525 1526 retval = sock_create(family, type, protocol, &sock); 1527 if (retval < 0) 1528 return retval; 1529 1530 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1531 } 1532 1533 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1534 { 1535 return __sys_socket(family, type, protocol); 1536 } 1537 1538 /* 1539 * Create a pair of connected sockets. 1540 */ 1541 1542 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) 1543 { 1544 struct socket *sock1, *sock2; 1545 int fd1, fd2, err; 1546 struct file *newfile1, *newfile2; 1547 int flags; 1548 1549 flags = type & ~SOCK_TYPE_MASK; 1550 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1551 return -EINVAL; 1552 type &= SOCK_TYPE_MASK; 1553 1554 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1555 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1556 1557 /* 1558 * reserve descriptors and make sure we won't fail 1559 * to return them to userland. 1560 */ 1561 fd1 = get_unused_fd_flags(flags); 1562 if (unlikely(fd1 < 0)) 1563 return fd1; 1564 1565 fd2 = get_unused_fd_flags(flags); 1566 if (unlikely(fd2 < 0)) { 1567 put_unused_fd(fd1); 1568 return fd2; 1569 } 1570 1571 err = put_user(fd1, &usockvec[0]); 1572 if (err) 1573 goto out; 1574 1575 err = put_user(fd2, &usockvec[1]); 1576 if (err) 1577 goto out; 1578 1579 /* 1580 * Obtain the first socket and check if the underlying protocol 1581 * supports the socketpair call. 1582 */ 1583 1584 err = sock_create(family, type, protocol, &sock1); 1585 if (unlikely(err < 0)) 1586 goto out; 1587 1588 err = sock_create(family, type, protocol, &sock2); 1589 if (unlikely(err < 0)) { 1590 sock_release(sock1); 1591 goto out; 1592 } 1593 1594 err = security_socket_socketpair(sock1, sock2); 1595 if (unlikely(err)) { 1596 sock_release(sock2); 1597 sock_release(sock1); 1598 goto out; 1599 } 1600 1601 err = sock1->ops->socketpair(sock1, sock2); 1602 if (unlikely(err < 0)) { 1603 sock_release(sock2); 1604 sock_release(sock1); 1605 goto out; 1606 } 1607 1608 newfile1 = sock_alloc_file(sock1, flags, NULL); 1609 if (IS_ERR(newfile1)) { 1610 err = PTR_ERR(newfile1); 1611 sock_release(sock2); 1612 goto out; 1613 } 1614 1615 newfile2 = sock_alloc_file(sock2, flags, NULL); 1616 if (IS_ERR(newfile2)) { 1617 err = PTR_ERR(newfile2); 1618 fput(newfile1); 1619 goto out; 1620 } 1621 1622 audit_fd_pair(fd1, fd2); 1623 1624 fd_install(fd1, newfile1); 1625 fd_install(fd2, newfile2); 1626 return 0; 1627 1628 out: 1629 put_unused_fd(fd2); 1630 put_unused_fd(fd1); 1631 return err; 1632 } 1633 1634 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1635 int __user *, usockvec) 1636 { 1637 return __sys_socketpair(family, type, protocol, usockvec); 1638 } 1639 1640 /* 1641 * Bind a name to a socket. Nothing much to do here since it's 1642 * the protocol's responsibility to handle the local address. 1643 * 1644 * We move the socket address to kernel space before we call 1645 * the protocol layer (having also checked the address is ok). 1646 */ 1647 1648 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) 1649 { 1650 struct socket *sock; 1651 struct sockaddr_storage address; 1652 int err, fput_needed; 1653 1654 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1655 if (sock) { 1656 err = move_addr_to_kernel(umyaddr, addrlen, &address); 1657 if (!err) { 1658 err = security_socket_bind(sock, 1659 (struct sockaddr *)&address, 1660 addrlen); 1661 if (!err) 1662 err = sock->ops->bind(sock, 1663 (struct sockaddr *) 1664 &address, addrlen); 1665 } 1666 fput_light(sock->file, fput_needed); 1667 } 1668 return err; 1669 } 1670 1671 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1672 { 1673 return __sys_bind(fd, umyaddr, addrlen); 1674 } 1675 1676 /* 1677 * Perform a listen. Basically, we allow the protocol to do anything 1678 * necessary for a listen, and if that works, we mark the socket as 1679 * ready for listening. 1680 */ 1681 1682 int __sys_listen(int fd, int backlog) 1683 { 1684 struct socket *sock; 1685 int err, fput_needed; 1686 int somaxconn; 1687 1688 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1689 if (sock) { 1690 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; 1691 if ((unsigned int)backlog > somaxconn) 1692 backlog = somaxconn; 1693 1694 err = security_socket_listen(sock, backlog); 1695 if (!err) 1696 err = sock->ops->listen(sock, backlog); 1697 1698 fput_light(sock->file, fput_needed); 1699 } 1700 return err; 1701 } 1702 1703 SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1704 { 1705 return __sys_listen(fd, backlog); 1706 } 1707 1708 int __sys_accept4_file(struct file *file, unsigned file_flags, 1709 struct sockaddr __user *upeer_sockaddr, 1710 int __user *upeer_addrlen, int flags, 1711 unsigned long nofile) 1712 { 1713 struct socket *sock, *newsock; 1714 struct file *newfile; 1715 int err, len, newfd; 1716 struct sockaddr_storage address; 1717 1718 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1719 return -EINVAL; 1720 1721 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1722 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1723 1724 sock = sock_from_file(file, &err); 1725 if (!sock) 1726 goto out; 1727 1728 err = -ENFILE; 1729 newsock = sock_alloc(); 1730 if (!newsock) 1731 goto out; 1732 1733 newsock->type = sock->type; 1734 newsock->ops = sock->ops; 1735 1736 /* 1737 * We don't need try_module_get here, as the listening socket (sock) 1738 * has the protocol module (sock->ops->owner) held. 1739 */ 1740 __module_get(newsock->ops->owner); 1741 1742 newfd = __get_unused_fd_flags(flags, nofile); 1743 if (unlikely(newfd < 0)) { 1744 err = newfd; 1745 sock_release(newsock); 1746 goto out; 1747 } 1748 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); 1749 if (IS_ERR(newfile)) { 1750 err = PTR_ERR(newfile); 1751 put_unused_fd(newfd); 1752 goto out; 1753 } 1754 1755 err = security_socket_accept(sock, newsock); 1756 if (err) 1757 goto out_fd; 1758 1759 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags, 1760 false); 1761 if (err < 0) 1762 goto out_fd; 1763 1764 if (upeer_sockaddr) { 1765 len = newsock->ops->getname(newsock, 1766 (struct sockaddr *)&address, 2); 1767 if (len < 0) { 1768 err = -ECONNABORTED; 1769 goto out_fd; 1770 } 1771 err = move_addr_to_user(&address, 1772 len, upeer_sockaddr, upeer_addrlen); 1773 if (err < 0) 1774 goto out_fd; 1775 } 1776 1777 /* File flags are not inherited via accept() unlike another OSes. */ 1778 1779 fd_install(newfd, newfile); 1780 err = newfd; 1781 out: 1782 return err; 1783 out_fd: 1784 fput(newfile); 1785 put_unused_fd(newfd); 1786 goto out; 1787 1788 } 1789 1790 /* 1791 * For accept, we attempt to create a new socket, set up the link 1792 * with the client, wake up the client, then return the new 1793 * connected fd. We collect the address of the connector in kernel 1794 * space and move it to user at the very end. This is unclean because 1795 * we open the socket then return an error. 1796 * 1797 * 1003.1g adds the ability to recvmsg() to query connection pending 1798 * status to recvmsg. We need to add that support in a way thats 1799 * clean when we restructure accept also. 1800 */ 1801 1802 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, 1803 int __user *upeer_addrlen, int flags) 1804 { 1805 int ret = -EBADF; 1806 struct fd f; 1807 1808 f = fdget(fd); 1809 if (f.file) { 1810 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr, 1811 upeer_addrlen, flags, 1812 rlimit(RLIMIT_NOFILE)); 1813 if (f.flags) 1814 fput(f.file); 1815 } 1816 1817 return ret; 1818 } 1819 1820 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 1821 int __user *, upeer_addrlen, int, flags) 1822 { 1823 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); 1824 } 1825 1826 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 1827 int __user *, upeer_addrlen) 1828 { 1829 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 1830 } 1831 1832 /* 1833 * Attempt to connect to a socket with the server address. The address 1834 * is in user space so we verify it is OK and move it to kernel space. 1835 * 1836 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 1837 * break bindings 1838 * 1839 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 1840 * other SEQPACKET protocols that take time to connect() as it doesn't 1841 * include the -EINPROGRESS status for such sockets. 1842 */ 1843 1844 int __sys_connect_file(struct file *file, struct sockaddr_storage *address, 1845 int addrlen, int file_flags) 1846 { 1847 struct socket *sock; 1848 int err; 1849 1850 sock = sock_from_file(file, &err); 1851 if (!sock) 1852 goto out; 1853 1854 err = 1855 security_socket_connect(sock, (struct sockaddr *)address, addrlen); 1856 if (err) 1857 goto out; 1858 1859 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen, 1860 sock->file->f_flags | file_flags); 1861 out: 1862 return err; 1863 } 1864 1865 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) 1866 { 1867 int ret = -EBADF; 1868 struct fd f; 1869 1870 f = fdget(fd); 1871 if (f.file) { 1872 struct sockaddr_storage address; 1873 1874 ret = move_addr_to_kernel(uservaddr, addrlen, &address); 1875 if (!ret) 1876 ret = __sys_connect_file(f.file, &address, addrlen, 0); 1877 if (f.flags) 1878 fput(f.file); 1879 } 1880 1881 return ret; 1882 } 1883 1884 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 1885 int, addrlen) 1886 { 1887 return __sys_connect(fd, uservaddr, addrlen); 1888 } 1889 1890 /* 1891 * Get the local address ('name') of a socket object. Move the obtained 1892 * name to user space. 1893 */ 1894 1895 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, 1896 int __user *usockaddr_len) 1897 { 1898 struct socket *sock; 1899 struct sockaddr_storage address; 1900 int err, fput_needed; 1901 1902 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1903 if (!sock) 1904 goto out; 1905 1906 err = security_socket_getsockname(sock); 1907 if (err) 1908 goto out_put; 1909 1910 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0); 1911 if (err < 0) 1912 goto out_put; 1913 /* "err" is actually length in this case */ 1914 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); 1915 1916 out_put: 1917 fput_light(sock->file, fput_needed); 1918 out: 1919 return err; 1920 } 1921 1922 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 1923 int __user *, usockaddr_len) 1924 { 1925 return __sys_getsockname(fd, usockaddr, usockaddr_len); 1926 } 1927 1928 /* 1929 * Get the remote address ('name') of a socket object. Move the obtained 1930 * name to user space. 1931 */ 1932 1933 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, 1934 int __user *usockaddr_len) 1935 { 1936 struct socket *sock; 1937 struct sockaddr_storage address; 1938 int err, fput_needed; 1939 1940 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1941 if (sock != NULL) { 1942 err = security_socket_getpeername(sock); 1943 if (err) { 1944 fput_light(sock->file, fput_needed); 1945 return err; 1946 } 1947 1948 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1); 1949 if (err >= 0) 1950 /* "err" is actually length in this case */ 1951 err = move_addr_to_user(&address, err, usockaddr, 1952 usockaddr_len); 1953 fput_light(sock->file, fput_needed); 1954 } 1955 return err; 1956 } 1957 1958 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 1959 int __user *, usockaddr_len) 1960 { 1961 return __sys_getpeername(fd, usockaddr, usockaddr_len); 1962 } 1963 1964 /* 1965 * Send a datagram to a given address. We move the address into kernel 1966 * space and check the user space data area is readable before invoking 1967 * the protocol. 1968 */ 1969 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, 1970 struct sockaddr __user *addr, int addr_len) 1971 { 1972 struct socket *sock; 1973 struct sockaddr_storage address; 1974 int err; 1975 struct msghdr msg; 1976 struct iovec iov; 1977 int fput_needed; 1978 1979 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter); 1980 if (unlikely(err)) 1981 return err; 1982 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1983 if (!sock) 1984 goto out; 1985 1986 msg.msg_name = NULL; 1987 msg.msg_control = NULL; 1988 msg.msg_controllen = 0; 1989 msg.msg_namelen = 0; 1990 if (addr) { 1991 err = move_addr_to_kernel(addr, addr_len, &address); 1992 if (err < 0) 1993 goto out_put; 1994 msg.msg_name = (struct sockaddr *)&address; 1995 msg.msg_namelen = addr_len; 1996 } 1997 if (sock->file->f_flags & O_NONBLOCK) 1998 flags |= MSG_DONTWAIT; 1999 msg.msg_flags = flags; 2000 err = sock_sendmsg(sock, &msg); 2001 2002 out_put: 2003 fput_light(sock->file, fput_needed); 2004 out: 2005 return err; 2006 } 2007 2008 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 2009 unsigned int, flags, struct sockaddr __user *, addr, 2010 int, addr_len) 2011 { 2012 return __sys_sendto(fd, buff, len, flags, addr, addr_len); 2013 } 2014 2015 /* 2016 * Send a datagram down a socket. 2017 */ 2018 2019 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 2020 unsigned int, flags) 2021 { 2022 return __sys_sendto(fd, buff, len, flags, NULL, 0); 2023 } 2024 2025 /* 2026 * Receive a frame from the socket and optionally record the address of the 2027 * sender. We verify the buffers are writable and if needed move the 2028 * sender address from kernel to user space. 2029 */ 2030 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, 2031 struct sockaddr __user *addr, int __user *addr_len) 2032 { 2033 struct socket *sock; 2034 struct iovec iov; 2035 struct msghdr msg; 2036 struct sockaddr_storage address; 2037 int err, err2; 2038 int fput_needed; 2039 2040 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter); 2041 if (unlikely(err)) 2042 return err; 2043 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2044 if (!sock) 2045 goto out; 2046 2047 msg.msg_control = NULL; 2048 msg.msg_controllen = 0; 2049 /* Save some cycles and don't copy the address if not needed */ 2050 msg.msg_name = addr ? (struct sockaddr *)&address : NULL; 2051 /* We assume all kernel code knows the size of sockaddr_storage */ 2052 msg.msg_namelen = 0; 2053 msg.msg_iocb = NULL; 2054 msg.msg_flags = 0; 2055 if (sock->file->f_flags & O_NONBLOCK) 2056 flags |= MSG_DONTWAIT; 2057 err = sock_recvmsg(sock, &msg, flags); 2058 2059 if (err >= 0 && addr != NULL) { 2060 err2 = move_addr_to_user(&address, 2061 msg.msg_namelen, addr, addr_len); 2062 if (err2 < 0) 2063 err = err2; 2064 } 2065 2066 fput_light(sock->file, fput_needed); 2067 out: 2068 return err; 2069 } 2070 2071 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 2072 unsigned int, flags, struct sockaddr __user *, addr, 2073 int __user *, addr_len) 2074 { 2075 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); 2076 } 2077 2078 /* 2079 * Receive a datagram from a socket. 2080 */ 2081 2082 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, 2083 unsigned int, flags) 2084 { 2085 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 2086 } 2087 2088 /* 2089 * Set a socket option. Because we don't know the option lengths we have 2090 * to pass the user mode parameter for the protocols to sort out. 2091 */ 2092 2093 static int __sys_setsockopt(int fd, int level, int optname, 2094 char __user *optval, int optlen) 2095 { 2096 mm_segment_t oldfs = get_fs(); 2097 char *kernel_optval = NULL; 2098 int err, fput_needed; 2099 struct socket *sock; 2100 2101 if (optlen < 0) 2102 return -EINVAL; 2103 2104 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2105 if (sock != NULL) { 2106 err = security_socket_setsockopt(sock, level, optname); 2107 if (err) 2108 goto out_put; 2109 2110 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, 2111 &optname, optval, &optlen, 2112 &kernel_optval); 2113 2114 if (err < 0) { 2115 goto out_put; 2116 } else if (err > 0) { 2117 err = 0; 2118 goto out_put; 2119 } 2120 2121 if (kernel_optval) { 2122 set_fs(KERNEL_DS); 2123 optval = (char __user __force *)kernel_optval; 2124 } 2125 2126 if (level == SOL_SOCKET) 2127 err = 2128 sock_setsockopt(sock, level, optname, optval, 2129 optlen); 2130 else 2131 err = 2132 sock->ops->setsockopt(sock, level, optname, optval, 2133 optlen); 2134 2135 if (kernel_optval) { 2136 set_fs(oldfs); 2137 kfree(kernel_optval); 2138 } 2139 out_put: 2140 fput_light(sock->file, fput_needed); 2141 } 2142 return err; 2143 } 2144 2145 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 2146 char __user *, optval, int, optlen) 2147 { 2148 return __sys_setsockopt(fd, level, optname, optval, optlen); 2149 } 2150 2151 /* 2152 * Get a socket option. Because we don't know the option lengths we have 2153 * to pass a user mode parameter for the protocols to sort out. 2154 */ 2155 2156 static int __sys_getsockopt(int fd, int level, int optname, 2157 char __user *optval, int __user *optlen) 2158 { 2159 int err, fput_needed; 2160 struct socket *sock; 2161 int max_optlen; 2162 2163 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2164 if (sock != NULL) { 2165 err = security_socket_getsockopt(sock, level, optname); 2166 if (err) 2167 goto out_put; 2168 2169 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen); 2170 2171 if (level == SOL_SOCKET) 2172 err = 2173 sock_getsockopt(sock, level, optname, optval, 2174 optlen); 2175 else 2176 err = 2177 sock->ops->getsockopt(sock, level, optname, optval, 2178 optlen); 2179 2180 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname, 2181 optval, optlen, 2182 max_optlen, err); 2183 out_put: 2184 fput_light(sock->file, fput_needed); 2185 } 2186 return err; 2187 } 2188 2189 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 2190 char __user *, optval, int __user *, optlen) 2191 { 2192 return __sys_getsockopt(fd, level, optname, optval, optlen); 2193 } 2194 2195 /* 2196 * Shutdown a socket. 2197 */ 2198 2199 int __sys_shutdown(int fd, int how) 2200 { 2201 int err, fput_needed; 2202 struct socket *sock; 2203 2204 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2205 if (sock != NULL) { 2206 err = security_socket_shutdown(sock, how); 2207 if (!err) 2208 err = sock->ops->shutdown(sock, how); 2209 fput_light(sock->file, fput_needed); 2210 } 2211 return err; 2212 } 2213 2214 SYSCALL_DEFINE2(shutdown, int, fd, int, how) 2215 { 2216 return __sys_shutdown(fd, how); 2217 } 2218 2219 /* A couple of helpful macros for getting the address of the 32/64 bit 2220 * fields which are the same type (int / unsigned) on our platforms. 2221 */ 2222 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 2223 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 2224 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 2225 2226 struct used_address { 2227 struct sockaddr_storage name; 2228 unsigned int name_len; 2229 }; 2230 2231 static int copy_msghdr_from_user(struct msghdr *kmsg, 2232 struct user_msghdr __user *umsg, 2233 struct sockaddr __user **save_addr, 2234 struct iovec **iov) 2235 { 2236 struct user_msghdr msg; 2237 ssize_t err; 2238 2239 if (copy_from_user(&msg, umsg, sizeof(*umsg))) 2240 return -EFAULT; 2241 2242 kmsg->msg_control = (void __force *)msg.msg_control; 2243 kmsg->msg_controllen = msg.msg_controllen; 2244 kmsg->msg_flags = msg.msg_flags; 2245 2246 kmsg->msg_namelen = msg.msg_namelen; 2247 if (!msg.msg_name) 2248 kmsg->msg_namelen = 0; 2249 2250 if (kmsg->msg_namelen < 0) 2251 return -EINVAL; 2252 2253 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) 2254 kmsg->msg_namelen = sizeof(struct sockaddr_storage); 2255 2256 if (save_addr) 2257 *save_addr = msg.msg_name; 2258 2259 if (msg.msg_name && kmsg->msg_namelen) { 2260 if (!save_addr) { 2261 err = move_addr_to_kernel(msg.msg_name, 2262 kmsg->msg_namelen, 2263 kmsg->msg_name); 2264 if (err < 0) 2265 return err; 2266 } 2267 } else { 2268 kmsg->msg_name = NULL; 2269 kmsg->msg_namelen = 0; 2270 } 2271 2272 if (msg.msg_iovlen > UIO_MAXIOV) 2273 return -EMSGSIZE; 2274 2275 kmsg->msg_iocb = NULL; 2276 2277 err = import_iovec(save_addr ? READ : WRITE, 2278 msg.msg_iov, msg.msg_iovlen, 2279 UIO_FASTIOV, iov, &kmsg->msg_iter); 2280 return err < 0 ? err : 0; 2281 } 2282 2283 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, 2284 unsigned int flags, struct used_address *used_address, 2285 unsigned int allowed_msghdr_flags) 2286 { 2287 unsigned char ctl[sizeof(struct cmsghdr) + 20] 2288 __aligned(sizeof(__kernel_size_t)); 2289 /* 20 is size of ipv6_pktinfo */ 2290 unsigned char *ctl_buf = ctl; 2291 int ctl_len; 2292 ssize_t err; 2293 2294 err = -ENOBUFS; 2295 2296 if (msg_sys->msg_controllen > INT_MAX) 2297 goto out; 2298 flags |= (msg_sys->msg_flags & allowed_msghdr_flags); 2299 ctl_len = msg_sys->msg_controllen; 2300 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 2301 err = 2302 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, 2303 sizeof(ctl)); 2304 if (err) 2305 goto out; 2306 ctl_buf = msg_sys->msg_control; 2307 ctl_len = msg_sys->msg_controllen; 2308 } else if (ctl_len) { 2309 BUILD_BUG_ON(sizeof(struct cmsghdr) != 2310 CMSG_ALIGN(sizeof(struct cmsghdr))); 2311 if (ctl_len > sizeof(ctl)) { 2312 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 2313 if (ctl_buf == NULL) 2314 goto out; 2315 } 2316 err = -EFAULT; 2317 /* 2318 * Careful! Before this, msg_sys->msg_control contains a user pointer. 2319 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted 2320 * checking falls down on this. 2321 */ 2322 if (copy_from_user(ctl_buf, 2323 (void __user __force *)msg_sys->msg_control, 2324 ctl_len)) 2325 goto out_freectl; 2326 msg_sys->msg_control = ctl_buf; 2327 } 2328 msg_sys->msg_flags = flags; 2329 2330 if (sock->file->f_flags & O_NONBLOCK) 2331 msg_sys->msg_flags |= MSG_DONTWAIT; 2332 /* 2333 * If this is sendmmsg() and current destination address is same as 2334 * previously succeeded address, omit asking LSM's decision. 2335 * used_address->name_len is initialized to UINT_MAX so that the first 2336 * destination address never matches. 2337 */ 2338 if (used_address && msg_sys->msg_name && 2339 used_address->name_len == msg_sys->msg_namelen && 2340 !memcmp(&used_address->name, msg_sys->msg_name, 2341 used_address->name_len)) { 2342 err = sock_sendmsg_nosec(sock, msg_sys); 2343 goto out_freectl; 2344 } 2345 err = sock_sendmsg(sock, msg_sys); 2346 /* 2347 * If this is sendmmsg() and sending to current destination address was 2348 * successful, remember it. 2349 */ 2350 if (used_address && err >= 0) { 2351 used_address->name_len = msg_sys->msg_namelen; 2352 if (msg_sys->msg_name) 2353 memcpy(&used_address->name, msg_sys->msg_name, 2354 used_address->name_len); 2355 } 2356 2357 out_freectl: 2358 if (ctl_buf != ctl) 2359 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 2360 out: 2361 return err; 2362 } 2363 2364 int sendmsg_copy_msghdr(struct msghdr *msg, 2365 struct user_msghdr __user *umsg, unsigned flags, 2366 struct iovec **iov) 2367 { 2368 int err; 2369 2370 if (flags & MSG_CMSG_COMPAT) { 2371 struct compat_msghdr __user *msg_compat; 2372 2373 msg_compat = (struct compat_msghdr __user *) umsg; 2374 err = get_compat_msghdr(msg, msg_compat, NULL, iov); 2375 } else { 2376 err = copy_msghdr_from_user(msg, umsg, NULL, iov); 2377 } 2378 if (err < 0) 2379 return err; 2380 2381 return 0; 2382 } 2383 2384 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, 2385 struct msghdr *msg_sys, unsigned int flags, 2386 struct used_address *used_address, 2387 unsigned int allowed_msghdr_flags) 2388 { 2389 struct sockaddr_storage address; 2390 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2391 ssize_t err; 2392 2393 msg_sys->msg_name = &address; 2394 2395 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov); 2396 if (err < 0) 2397 return err; 2398 2399 err = ____sys_sendmsg(sock, msg_sys, flags, used_address, 2400 allowed_msghdr_flags); 2401 kfree(iov); 2402 return err; 2403 } 2404 2405 /* 2406 * BSD sendmsg interface 2407 */ 2408 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg, 2409 unsigned int flags) 2410 { 2411 /* disallow ancillary data requests from this path */ 2412 if (msg->msg_control || msg->msg_controllen) 2413 return -EINVAL; 2414 2415 return ____sys_sendmsg(sock, msg, flags, NULL, 0); 2416 } 2417 2418 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2419 bool forbid_cmsg_compat) 2420 { 2421 int fput_needed, err; 2422 struct msghdr msg_sys; 2423 struct socket *sock; 2424 2425 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2426 return -EINVAL; 2427 2428 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2429 if (!sock) 2430 goto out; 2431 2432 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); 2433 2434 fput_light(sock->file, fput_needed); 2435 out: 2436 return err; 2437 } 2438 2439 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) 2440 { 2441 return __sys_sendmsg(fd, msg, flags, true); 2442 } 2443 2444 /* 2445 * Linux sendmmsg interface 2446 */ 2447 2448 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2449 unsigned int flags, bool forbid_cmsg_compat) 2450 { 2451 int fput_needed, err, datagrams; 2452 struct socket *sock; 2453 struct mmsghdr __user *entry; 2454 struct compat_mmsghdr __user *compat_entry; 2455 struct msghdr msg_sys; 2456 struct used_address used_address; 2457 unsigned int oflags = flags; 2458 2459 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2460 return -EINVAL; 2461 2462 if (vlen > UIO_MAXIOV) 2463 vlen = UIO_MAXIOV; 2464 2465 datagrams = 0; 2466 2467 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2468 if (!sock) 2469 return err; 2470 2471 used_address.name_len = UINT_MAX; 2472 entry = mmsg; 2473 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2474 err = 0; 2475 flags |= MSG_BATCH; 2476 2477 while (datagrams < vlen) { 2478 if (datagrams == vlen - 1) 2479 flags = oflags; 2480 2481 if (MSG_CMSG_COMPAT & flags) { 2482 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, 2483 &msg_sys, flags, &used_address, MSG_EOR); 2484 if (err < 0) 2485 break; 2486 err = __put_user(err, &compat_entry->msg_len); 2487 ++compat_entry; 2488 } else { 2489 err = ___sys_sendmsg(sock, 2490 (struct user_msghdr __user *)entry, 2491 &msg_sys, flags, &used_address, MSG_EOR); 2492 if (err < 0) 2493 break; 2494 err = put_user(err, &entry->msg_len); 2495 ++entry; 2496 } 2497 2498 if (err) 2499 break; 2500 ++datagrams; 2501 if (msg_data_left(&msg_sys)) 2502 break; 2503 cond_resched(); 2504 } 2505 2506 fput_light(sock->file, fput_needed); 2507 2508 /* We only return an error if no datagrams were able to be sent */ 2509 if (datagrams != 0) 2510 return datagrams; 2511 2512 return err; 2513 } 2514 2515 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, 2516 unsigned int, vlen, unsigned int, flags) 2517 { 2518 return __sys_sendmmsg(fd, mmsg, vlen, flags, true); 2519 } 2520 2521 int recvmsg_copy_msghdr(struct msghdr *msg, 2522 struct user_msghdr __user *umsg, unsigned flags, 2523 struct sockaddr __user **uaddr, 2524 struct iovec **iov) 2525 { 2526 ssize_t err; 2527 2528 if (MSG_CMSG_COMPAT & flags) { 2529 struct compat_msghdr __user *msg_compat; 2530 2531 msg_compat = (struct compat_msghdr __user *) umsg; 2532 err = get_compat_msghdr(msg, msg_compat, uaddr, iov); 2533 } else { 2534 err = copy_msghdr_from_user(msg, umsg, uaddr, iov); 2535 } 2536 if (err < 0) 2537 return err; 2538 2539 return 0; 2540 } 2541 2542 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys, 2543 struct user_msghdr __user *msg, 2544 struct sockaddr __user *uaddr, 2545 unsigned int flags, int nosec) 2546 { 2547 struct compat_msghdr __user *msg_compat = 2548 (struct compat_msghdr __user *) msg; 2549 int __user *uaddr_len = COMPAT_NAMELEN(msg); 2550 struct sockaddr_storage addr; 2551 unsigned long cmsg_ptr; 2552 int len; 2553 ssize_t err; 2554 2555 msg_sys->msg_name = &addr; 2556 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2557 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2558 2559 /* We assume all kernel code knows the size of sockaddr_storage */ 2560 msg_sys->msg_namelen = 0; 2561 2562 if (sock->file->f_flags & O_NONBLOCK) 2563 flags |= MSG_DONTWAIT; 2564 2565 if (unlikely(nosec)) 2566 err = sock_recvmsg_nosec(sock, msg_sys, flags); 2567 else 2568 err = sock_recvmsg(sock, msg_sys, flags); 2569 2570 if (err < 0) 2571 goto out; 2572 len = err; 2573 2574 if (uaddr != NULL) { 2575 err = move_addr_to_user(&addr, 2576 msg_sys->msg_namelen, uaddr, 2577 uaddr_len); 2578 if (err < 0) 2579 goto out; 2580 } 2581 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2582 COMPAT_FLAGS(msg)); 2583 if (err) 2584 goto out; 2585 if (MSG_CMSG_COMPAT & flags) 2586 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2587 &msg_compat->msg_controllen); 2588 else 2589 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2590 &msg->msg_controllen); 2591 if (err) 2592 goto out; 2593 err = len; 2594 out: 2595 return err; 2596 } 2597 2598 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, 2599 struct msghdr *msg_sys, unsigned int flags, int nosec) 2600 { 2601 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2602 /* user mode address pointers */ 2603 struct sockaddr __user *uaddr; 2604 ssize_t err; 2605 2606 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov); 2607 if (err < 0) 2608 return err; 2609 2610 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec); 2611 kfree(iov); 2612 return err; 2613 } 2614 2615 /* 2616 * BSD recvmsg interface 2617 */ 2618 2619 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg, 2620 struct user_msghdr __user *umsg, 2621 struct sockaddr __user *uaddr, unsigned int flags) 2622 { 2623 /* disallow ancillary data requests from this path */ 2624 if (msg->msg_control || msg->msg_controllen) 2625 return -EINVAL; 2626 2627 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0); 2628 } 2629 2630 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2631 bool forbid_cmsg_compat) 2632 { 2633 int fput_needed, err; 2634 struct msghdr msg_sys; 2635 struct socket *sock; 2636 2637 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2638 return -EINVAL; 2639 2640 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2641 if (!sock) 2642 goto out; 2643 2644 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2645 2646 fput_light(sock->file, fput_needed); 2647 out: 2648 return err; 2649 } 2650 2651 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, 2652 unsigned int, flags) 2653 { 2654 return __sys_recvmsg(fd, msg, flags, true); 2655 } 2656 2657 /* 2658 * Linux recvmmsg interface 2659 */ 2660 2661 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2662 unsigned int vlen, unsigned int flags, 2663 struct timespec64 *timeout) 2664 { 2665 int fput_needed, err, datagrams; 2666 struct socket *sock; 2667 struct mmsghdr __user *entry; 2668 struct compat_mmsghdr __user *compat_entry; 2669 struct msghdr msg_sys; 2670 struct timespec64 end_time; 2671 struct timespec64 timeout64; 2672 2673 if (timeout && 2674 poll_select_set_timeout(&end_time, timeout->tv_sec, 2675 timeout->tv_nsec)) 2676 return -EINVAL; 2677 2678 datagrams = 0; 2679 2680 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2681 if (!sock) 2682 return err; 2683 2684 if (likely(!(flags & MSG_ERRQUEUE))) { 2685 err = sock_error(sock->sk); 2686 if (err) { 2687 datagrams = err; 2688 goto out_put; 2689 } 2690 } 2691 2692 entry = mmsg; 2693 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2694 2695 while (datagrams < vlen) { 2696 /* 2697 * No need to ask LSM for more than the first datagram. 2698 */ 2699 if (MSG_CMSG_COMPAT & flags) { 2700 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, 2701 &msg_sys, flags & ~MSG_WAITFORONE, 2702 datagrams); 2703 if (err < 0) 2704 break; 2705 err = __put_user(err, &compat_entry->msg_len); 2706 ++compat_entry; 2707 } else { 2708 err = ___sys_recvmsg(sock, 2709 (struct user_msghdr __user *)entry, 2710 &msg_sys, flags & ~MSG_WAITFORONE, 2711 datagrams); 2712 if (err < 0) 2713 break; 2714 err = put_user(err, &entry->msg_len); 2715 ++entry; 2716 } 2717 2718 if (err) 2719 break; 2720 ++datagrams; 2721 2722 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2723 if (flags & MSG_WAITFORONE) 2724 flags |= MSG_DONTWAIT; 2725 2726 if (timeout) { 2727 ktime_get_ts64(&timeout64); 2728 *timeout = timespec64_sub(end_time, timeout64); 2729 if (timeout->tv_sec < 0) { 2730 timeout->tv_sec = timeout->tv_nsec = 0; 2731 break; 2732 } 2733 2734 /* Timeout, return less than vlen datagrams */ 2735 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2736 break; 2737 } 2738 2739 /* Out of band data, return right away */ 2740 if (msg_sys.msg_flags & MSG_OOB) 2741 break; 2742 cond_resched(); 2743 } 2744 2745 if (err == 0) 2746 goto out_put; 2747 2748 if (datagrams == 0) { 2749 datagrams = err; 2750 goto out_put; 2751 } 2752 2753 /* 2754 * We may return less entries than requested (vlen) if the 2755 * sock is non block and there aren't enough datagrams... 2756 */ 2757 if (err != -EAGAIN) { 2758 /* 2759 * ... or if recvmsg returns an error after we 2760 * received some datagrams, where we record the 2761 * error to return on the next call or if the 2762 * app asks about it using getsockopt(SO_ERROR). 2763 */ 2764 sock->sk->sk_err = -err; 2765 } 2766 out_put: 2767 fput_light(sock->file, fput_needed); 2768 2769 return datagrams; 2770 } 2771 2772 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2773 unsigned int vlen, unsigned int flags, 2774 struct __kernel_timespec __user *timeout, 2775 struct old_timespec32 __user *timeout32) 2776 { 2777 int datagrams; 2778 struct timespec64 timeout_sys; 2779 2780 if (timeout && get_timespec64(&timeout_sys, timeout)) 2781 return -EFAULT; 2782 2783 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32)) 2784 return -EFAULT; 2785 2786 if (!timeout && !timeout32) 2787 return do_recvmmsg(fd, mmsg, vlen, flags, NULL); 2788 2789 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 2790 2791 if (datagrams <= 0) 2792 return datagrams; 2793 2794 if (timeout && put_timespec64(&timeout_sys, timeout)) 2795 datagrams = -EFAULT; 2796 2797 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32)) 2798 datagrams = -EFAULT; 2799 2800 return datagrams; 2801 } 2802 2803 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 2804 unsigned int, vlen, unsigned int, flags, 2805 struct __kernel_timespec __user *, timeout) 2806 { 2807 if (flags & MSG_CMSG_COMPAT) 2808 return -EINVAL; 2809 2810 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL); 2811 } 2812 2813 #ifdef CONFIG_COMPAT_32BIT_TIME 2814 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg, 2815 unsigned int, vlen, unsigned int, flags, 2816 struct old_timespec32 __user *, timeout) 2817 { 2818 if (flags & MSG_CMSG_COMPAT) 2819 return -EINVAL; 2820 2821 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout); 2822 } 2823 #endif 2824 2825 #ifdef __ARCH_WANT_SYS_SOCKETCALL 2826 /* Argument list sizes for sys_socketcall */ 2827 #define AL(x) ((x) * sizeof(unsigned long)) 2828 static const unsigned char nargs[21] = { 2829 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 2830 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 2831 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 2832 AL(4), AL(5), AL(4) 2833 }; 2834 2835 #undef AL 2836 2837 /* 2838 * System call vectors. 2839 * 2840 * Argument checking cleaned up. Saved 20% in size. 2841 * This function doesn't need to set the kernel lock because 2842 * it is set by the callees. 2843 */ 2844 2845 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 2846 { 2847 unsigned long a[AUDITSC_ARGS]; 2848 unsigned long a0, a1; 2849 int err; 2850 unsigned int len; 2851 2852 if (call < 1 || call > SYS_SENDMMSG) 2853 return -EINVAL; 2854 call = array_index_nospec(call, SYS_SENDMMSG + 1); 2855 2856 len = nargs[call]; 2857 if (len > sizeof(a)) 2858 return -EINVAL; 2859 2860 /* copy_from_user should be SMP safe. */ 2861 if (copy_from_user(a, args, len)) 2862 return -EFAULT; 2863 2864 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); 2865 if (err) 2866 return err; 2867 2868 a0 = a[0]; 2869 a1 = a[1]; 2870 2871 switch (call) { 2872 case SYS_SOCKET: 2873 err = __sys_socket(a0, a1, a[2]); 2874 break; 2875 case SYS_BIND: 2876 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 2877 break; 2878 case SYS_CONNECT: 2879 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 2880 break; 2881 case SYS_LISTEN: 2882 err = __sys_listen(a0, a1); 2883 break; 2884 case SYS_ACCEPT: 2885 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 2886 (int __user *)a[2], 0); 2887 break; 2888 case SYS_GETSOCKNAME: 2889 err = 2890 __sys_getsockname(a0, (struct sockaddr __user *)a1, 2891 (int __user *)a[2]); 2892 break; 2893 case SYS_GETPEERNAME: 2894 err = 2895 __sys_getpeername(a0, (struct sockaddr __user *)a1, 2896 (int __user *)a[2]); 2897 break; 2898 case SYS_SOCKETPAIR: 2899 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 2900 break; 2901 case SYS_SEND: 2902 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 2903 NULL, 0); 2904 break; 2905 case SYS_SENDTO: 2906 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 2907 (struct sockaddr __user *)a[4], a[5]); 2908 break; 2909 case SYS_RECV: 2910 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 2911 NULL, NULL); 2912 break; 2913 case SYS_RECVFROM: 2914 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 2915 (struct sockaddr __user *)a[4], 2916 (int __user *)a[5]); 2917 break; 2918 case SYS_SHUTDOWN: 2919 err = __sys_shutdown(a0, a1); 2920 break; 2921 case SYS_SETSOCKOPT: 2922 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], 2923 a[4]); 2924 break; 2925 case SYS_GETSOCKOPT: 2926 err = 2927 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 2928 (int __user *)a[4]); 2929 break; 2930 case SYS_SENDMSG: 2931 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, 2932 a[2], true); 2933 break; 2934 case SYS_SENDMMSG: 2935 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], 2936 a[3], true); 2937 break; 2938 case SYS_RECVMSG: 2939 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, 2940 a[2], true); 2941 break; 2942 case SYS_RECVMMSG: 2943 if (IS_ENABLED(CONFIG_64BIT)) 2944 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 2945 a[2], a[3], 2946 (struct __kernel_timespec __user *)a[4], 2947 NULL); 2948 else 2949 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 2950 a[2], a[3], NULL, 2951 (struct old_timespec32 __user *)a[4]); 2952 break; 2953 case SYS_ACCEPT4: 2954 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 2955 (int __user *)a[2], a[3]); 2956 break; 2957 default: 2958 err = -EINVAL; 2959 break; 2960 } 2961 return err; 2962 } 2963 2964 #endif /* __ARCH_WANT_SYS_SOCKETCALL */ 2965 2966 /** 2967 * sock_register - add a socket protocol handler 2968 * @ops: description of protocol 2969 * 2970 * This function is called by a protocol handler that wants to 2971 * advertise its address family, and have it linked into the 2972 * socket interface. The value ops->family corresponds to the 2973 * socket system call protocol family. 2974 */ 2975 int sock_register(const struct net_proto_family *ops) 2976 { 2977 int err; 2978 2979 if (ops->family >= NPROTO) { 2980 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); 2981 return -ENOBUFS; 2982 } 2983 2984 spin_lock(&net_family_lock); 2985 if (rcu_dereference_protected(net_families[ops->family], 2986 lockdep_is_held(&net_family_lock))) 2987 err = -EEXIST; 2988 else { 2989 rcu_assign_pointer(net_families[ops->family], ops); 2990 err = 0; 2991 } 2992 spin_unlock(&net_family_lock); 2993 2994 pr_info("NET: Registered protocol family %d\n", ops->family); 2995 return err; 2996 } 2997 EXPORT_SYMBOL(sock_register); 2998 2999 /** 3000 * sock_unregister - remove a protocol handler 3001 * @family: protocol family to remove 3002 * 3003 * This function is called by a protocol handler that wants to 3004 * remove its address family, and have it unlinked from the 3005 * new socket creation. 3006 * 3007 * If protocol handler is a module, then it can use module reference 3008 * counts to protect against new references. If protocol handler is not 3009 * a module then it needs to provide its own protection in 3010 * the ops->create routine. 3011 */ 3012 void sock_unregister(int family) 3013 { 3014 BUG_ON(family < 0 || family >= NPROTO); 3015 3016 spin_lock(&net_family_lock); 3017 RCU_INIT_POINTER(net_families[family], NULL); 3018 spin_unlock(&net_family_lock); 3019 3020 synchronize_rcu(); 3021 3022 pr_info("NET: Unregistered protocol family %d\n", family); 3023 } 3024 EXPORT_SYMBOL(sock_unregister); 3025 3026 bool sock_is_registered(int family) 3027 { 3028 return family < NPROTO && rcu_access_pointer(net_families[family]); 3029 } 3030 3031 static int __init sock_init(void) 3032 { 3033 int err; 3034 /* 3035 * Initialize the network sysctl infrastructure. 3036 */ 3037 err = net_sysctl_init(); 3038 if (err) 3039 goto out; 3040 3041 /* 3042 * Initialize skbuff SLAB cache 3043 */ 3044 skb_init(); 3045 3046 /* 3047 * Initialize the protocols module. 3048 */ 3049 3050 init_inodecache(); 3051 3052 err = register_filesystem(&sock_fs_type); 3053 if (err) 3054 goto out_fs; 3055 sock_mnt = kern_mount(&sock_fs_type); 3056 if (IS_ERR(sock_mnt)) { 3057 err = PTR_ERR(sock_mnt); 3058 goto out_mount; 3059 } 3060 3061 /* The real protocol initialization is performed in later initcalls. 3062 */ 3063 3064 #ifdef CONFIG_NETFILTER 3065 err = netfilter_init(); 3066 if (err) 3067 goto out; 3068 #endif 3069 3070 ptp_classifier_init(); 3071 3072 out: 3073 return err; 3074 3075 out_mount: 3076 unregister_filesystem(&sock_fs_type); 3077 out_fs: 3078 goto out; 3079 } 3080 3081 core_initcall(sock_init); /* early initcall */ 3082 3083 #ifdef CONFIG_PROC_FS 3084 void socket_seq_show(struct seq_file *seq) 3085 { 3086 seq_printf(seq, "sockets: used %d\n", 3087 sock_inuse_get(seq->private)); 3088 } 3089 #endif /* CONFIG_PROC_FS */ 3090 3091 #ifdef CONFIG_COMPAT 3092 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32) 3093 { 3094 struct compat_ifconf ifc32; 3095 struct ifconf ifc; 3096 int err; 3097 3098 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf))) 3099 return -EFAULT; 3100 3101 ifc.ifc_len = ifc32.ifc_len; 3102 ifc.ifc_req = compat_ptr(ifc32.ifcbuf); 3103 3104 rtnl_lock(); 3105 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq)); 3106 rtnl_unlock(); 3107 if (err) 3108 return err; 3109 3110 ifc32.ifc_len = ifc.ifc_len; 3111 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf))) 3112 return -EFAULT; 3113 3114 return 0; 3115 } 3116 3117 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32) 3118 { 3119 struct compat_ethtool_rxnfc __user *compat_rxnfc; 3120 bool convert_in = false, convert_out = false; 3121 size_t buf_size = 0; 3122 struct ethtool_rxnfc __user *rxnfc = NULL; 3123 struct ifreq ifr; 3124 u32 rule_cnt = 0, actual_rule_cnt; 3125 u32 ethcmd; 3126 u32 data; 3127 int ret; 3128 3129 if (get_user(data, &ifr32->ifr_ifru.ifru_data)) 3130 return -EFAULT; 3131 3132 compat_rxnfc = compat_ptr(data); 3133 3134 if (get_user(ethcmd, &compat_rxnfc->cmd)) 3135 return -EFAULT; 3136 3137 /* Most ethtool structures are defined without padding. 3138 * Unfortunately struct ethtool_rxnfc is an exception. 3139 */ 3140 switch (ethcmd) { 3141 default: 3142 break; 3143 case ETHTOOL_GRXCLSRLALL: 3144 /* Buffer size is variable */ 3145 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt)) 3146 return -EFAULT; 3147 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32)) 3148 return -ENOMEM; 3149 buf_size += rule_cnt * sizeof(u32); 3150 /* fall through */ 3151 case ETHTOOL_GRXRINGS: 3152 case ETHTOOL_GRXCLSRLCNT: 3153 case ETHTOOL_GRXCLSRULE: 3154 case ETHTOOL_SRXCLSRLINS: 3155 convert_out = true; 3156 /* fall through */ 3157 case ETHTOOL_SRXCLSRLDEL: 3158 buf_size += sizeof(struct ethtool_rxnfc); 3159 convert_in = true; 3160 rxnfc = compat_alloc_user_space(buf_size); 3161 break; 3162 } 3163 3164 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ)) 3165 return -EFAULT; 3166 3167 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc; 3168 3169 if (convert_in) { 3170 /* We expect there to be holes between fs.m_ext and 3171 * fs.ring_cookie and at the end of fs, but nowhere else. 3172 */ 3173 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) + 3174 sizeof(compat_rxnfc->fs.m_ext) != 3175 offsetof(struct ethtool_rxnfc, fs.m_ext) + 3176 sizeof(rxnfc->fs.m_ext)); 3177 BUILD_BUG_ON( 3178 offsetof(struct compat_ethtool_rxnfc, fs.location) - 3179 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) != 3180 offsetof(struct ethtool_rxnfc, fs.location) - 3181 offsetof(struct ethtool_rxnfc, fs.ring_cookie)); 3182 3183 if (copy_in_user(rxnfc, compat_rxnfc, 3184 (void __user *)(&rxnfc->fs.m_ext + 1) - 3185 (void __user *)rxnfc) || 3186 copy_in_user(&rxnfc->fs.ring_cookie, 3187 &compat_rxnfc->fs.ring_cookie, 3188 (void __user *)(&rxnfc->fs.location + 1) - 3189 (void __user *)&rxnfc->fs.ring_cookie)) 3190 return -EFAULT; 3191 if (ethcmd == ETHTOOL_GRXCLSRLALL) { 3192 if (put_user(rule_cnt, &rxnfc->rule_cnt)) 3193 return -EFAULT; 3194 } else if (copy_in_user(&rxnfc->rule_cnt, 3195 &compat_rxnfc->rule_cnt, 3196 sizeof(rxnfc->rule_cnt))) 3197 return -EFAULT; 3198 } 3199 3200 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL); 3201 if (ret) 3202 return ret; 3203 3204 if (convert_out) { 3205 if (copy_in_user(compat_rxnfc, rxnfc, 3206 (const void __user *)(&rxnfc->fs.m_ext + 1) - 3207 (const void __user *)rxnfc) || 3208 copy_in_user(&compat_rxnfc->fs.ring_cookie, 3209 &rxnfc->fs.ring_cookie, 3210 (const void __user *)(&rxnfc->fs.location + 1) - 3211 (const void __user *)&rxnfc->fs.ring_cookie) || 3212 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt, 3213 sizeof(rxnfc->rule_cnt))) 3214 return -EFAULT; 3215 3216 if (ethcmd == ETHTOOL_GRXCLSRLALL) { 3217 /* As an optimisation, we only copy the actual 3218 * number of rules that the underlying 3219 * function returned. Since Mallory might 3220 * change the rule count in user memory, we 3221 * check that it is less than the rule count 3222 * originally given (as the user buffer size), 3223 * which has been range-checked. 3224 */ 3225 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt)) 3226 return -EFAULT; 3227 if (actual_rule_cnt < rule_cnt) 3228 rule_cnt = actual_rule_cnt; 3229 if (copy_in_user(&compat_rxnfc->rule_locs[0], 3230 &rxnfc->rule_locs[0], 3231 rule_cnt * sizeof(u32))) 3232 return -EFAULT; 3233 } 3234 } 3235 3236 return 0; 3237 } 3238 3239 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 3240 { 3241 compat_uptr_t uptr32; 3242 struct ifreq ifr; 3243 void __user *saved; 3244 int err; 3245 3246 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq))) 3247 return -EFAULT; 3248 3249 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 3250 return -EFAULT; 3251 3252 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; 3253 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); 3254 3255 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL); 3256 if (!err) { 3257 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; 3258 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq))) 3259 err = -EFAULT; 3260 } 3261 return err; 3262 } 3263 3264 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ 3265 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, 3266 struct compat_ifreq __user *u_ifreq32) 3267 { 3268 struct ifreq ifreq; 3269 u32 data32; 3270 3271 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ)) 3272 return -EFAULT; 3273 if (get_user(data32, &u_ifreq32->ifr_data)) 3274 return -EFAULT; 3275 ifreq.ifr_data = compat_ptr(data32); 3276 3277 return dev_ioctl(net, cmd, &ifreq, NULL); 3278 } 3279 3280 static int compat_ifreq_ioctl(struct net *net, struct socket *sock, 3281 unsigned int cmd, 3282 struct compat_ifreq __user *uifr32) 3283 { 3284 struct ifreq __user *uifr; 3285 int err; 3286 3287 /* Handle the fact that while struct ifreq has the same *layout* on 3288 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data, 3289 * which are handled elsewhere, it still has different *size* due to 3290 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit, 3291 * resulting in struct ifreq being 32 and 40 bytes respectively). 3292 * As a result, if the struct happens to be at the end of a page and 3293 * the next page isn't readable/writable, we get a fault. To prevent 3294 * that, copy back and forth to the full size. 3295 */ 3296 3297 uifr = compat_alloc_user_space(sizeof(*uifr)); 3298 if (copy_in_user(uifr, uifr32, sizeof(*uifr32))) 3299 return -EFAULT; 3300 3301 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr); 3302 3303 if (!err) { 3304 switch (cmd) { 3305 case SIOCGIFFLAGS: 3306 case SIOCGIFMETRIC: 3307 case SIOCGIFMTU: 3308 case SIOCGIFMEM: 3309 case SIOCGIFHWADDR: 3310 case SIOCGIFINDEX: 3311 case SIOCGIFADDR: 3312 case SIOCGIFBRDADDR: 3313 case SIOCGIFDSTADDR: 3314 case SIOCGIFNETMASK: 3315 case SIOCGIFPFLAGS: 3316 case SIOCGIFTXQLEN: 3317 case SIOCGMIIPHY: 3318 case SIOCGMIIREG: 3319 case SIOCGIFNAME: 3320 if (copy_in_user(uifr32, uifr, sizeof(*uifr32))) 3321 err = -EFAULT; 3322 break; 3323 } 3324 } 3325 return err; 3326 } 3327 3328 static int compat_sioc_ifmap(struct net *net, unsigned int cmd, 3329 struct compat_ifreq __user *uifr32) 3330 { 3331 struct ifreq ifr; 3332 struct compat_ifmap __user *uifmap32; 3333 int err; 3334 3335 uifmap32 = &uifr32->ifr_ifru.ifru_map; 3336 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name)); 3337 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 3338 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 3339 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 3340 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq); 3341 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma); 3342 err |= get_user(ifr.ifr_map.port, &uifmap32->port); 3343 if (err) 3344 return -EFAULT; 3345 3346 err = dev_ioctl(net, cmd, &ifr, NULL); 3347 3348 if (cmd == SIOCGIFMAP && !err) { 3349 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name)); 3350 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 3351 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 3352 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 3353 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq); 3354 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma); 3355 err |= put_user(ifr.ifr_map.port, &uifmap32->port); 3356 if (err) 3357 err = -EFAULT; 3358 } 3359 return err; 3360 } 3361 3362 struct rtentry32 { 3363 u32 rt_pad1; 3364 struct sockaddr rt_dst; /* target address */ 3365 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */ 3366 struct sockaddr rt_genmask; /* target network mask (IP) */ 3367 unsigned short rt_flags; 3368 short rt_pad2; 3369 u32 rt_pad3; 3370 unsigned char rt_tos; 3371 unsigned char rt_class; 3372 short rt_pad4; 3373 short rt_metric; /* +1 for binary compatibility! */ 3374 /* char * */ u32 rt_dev; /* forcing the device at add */ 3375 u32 rt_mtu; /* per route MTU/Window */ 3376 u32 rt_window; /* Window clamping */ 3377 unsigned short rt_irtt; /* Initial RTT */ 3378 }; 3379 3380 struct in6_rtmsg32 { 3381 struct in6_addr rtmsg_dst; 3382 struct in6_addr rtmsg_src; 3383 struct in6_addr rtmsg_gateway; 3384 u32 rtmsg_type; 3385 u16 rtmsg_dst_len; 3386 u16 rtmsg_src_len; 3387 u32 rtmsg_metric; 3388 u32 rtmsg_info; 3389 u32 rtmsg_flags; 3390 s32 rtmsg_ifindex; 3391 }; 3392 3393 static int routing_ioctl(struct net *net, struct socket *sock, 3394 unsigned int cmd, void __user *argp) 3395 { 3396 int ret; 3397 void *r = NULL; 3398 struct in6_rtmsg r6; 3399 struct rtentry r4; 3400 char devname[16]; 3401 u32 rtdev; 3402 mm_segment_t old_fs = get_fs(); 3403 3404 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */ 3405 struct in6_rtmsg32 __user *ur6 = argp; 3406 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst), 3407 3 * sizeof(struct in6_addr)); 3408 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type)); 3409 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len)); 3410 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len)); 3411 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric)); 3412 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info)); 3413 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags)); 3414 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex)); 3415 3416 r = (void *) &r6; 3417 } else { /* ipv4 */ 3418 struct rtentry32 __user *ur4 = argp; 3419 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst), 3420 3 * sizeof(struct sockaddr)); 3421 ret |= get_user(r4.rt_flags, &(ur4->rt_flags)); 3422 ret |= get_user(r4.rt_metric, &(ur4->rt_metric)); 3423 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu)); 3424 ret |= get_user(r4.rt_window, &(ur4->rt_window)); 3425 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt)); 3426 ret |= get_user(rtdev, &(ur4->rt_dev)); 3427 if (rtdev) { 3428 ret |= copy_from_user(devname, compat_ptr(rtdev), 15); 3429 r4.rt_dev = (char __user __force *)devname; 3430 devname[15] = 0; 3431 } else 3432 r4.rt_dev = NULL; 3433 3434 r = (void *) &r4; 3435 } 3436 3437 if (ret) { 3438 ret = -EFAULT; 3439 goto out; 3440 } 3441 3442 set_fs(KERNEL_DS); 3443 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r); 3444 set_fs(old_fs); 3445 3446 out: 3447 return ret; 3448 } 3449 3450 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE 3451 * for some operations; this forces use of the newer bridge-utils that 3452 * use compatible ioctls 3453 */ 3454 static int old_bridge_ioctl(compat_ulong_t __user *argp) 3455 { 3456 compat_ulong_t tmp; 3457 3458 if (get_user(tmp, argp)) 3459 return -EFAULT; 3460 if (tmp == BRCTL_GET_VERSION) 3461 return BRCTL_VERSION + 1; 3462 return -EINVAL; 3463 } 3464 3465 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 3466 unsigned int cmd, unsigned long arg) 3467 { 3468 void __user *argp = compat_ptr(arg); 3469 struct sock *sk = sock->sk; 3470 struct net *net = sock_net(sk); 3471 3472 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 3473 return compat_ifr_data_ioctl(net, cmd, argp); 3474 3475 switch (cmd) { 3476 case SIOCSIFBR: 3477 case SIOCGIFBR: 3478 return old_bridge_ioctl(argp); 3479 case SIOCGIFCONF: 3480 return compat_dev_ifconf(net, argp); 3481 case SIOCETHTOOL: 3482 return ethtool_ioctl(net, argp); 3483 case SIOCWANDEV: 3484 return compat_siocwandev(net, argp); 3485 case SIOCGIFMAP: 3486 case SIOCSIFMAP: 3487 return compat_sioc_ifmap(net, cmd, argp); 3488 case SIOCADDRT: 3489 case SIOCDELRT: 3490 return routing_ioctl(net, sock, cmd, argp); 3491 case SIOCGSTAMP_OLD: 3492 case SIOCGSTAMPNS_OLD: 3493 if (!sock->ops->gettstamp) 3494 return -ENOIOCTLCMD; 3495 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD, 3496 !COMPAT_USE_64BIT_TIME); 3497 3498 case SIOCBONDSLAVEINFOQUERY: 3499 case SIOCBONDINFOQUERY: 3500 case SIOCSHWTSTAMP: 3501 case SIOCGHWTSTAMP: 3502 return compat_ifr_data_ioctl(net, cmd, argp); 3503 3504 case FIOSETOWN: 3505 case SIOCSPGRP: 3506 case FIOGETOWN: 3507 case SIOCGPGRP: 3508 case SIOCBRADDBR: 3509 case SIOCBRDELBR: 3510 case SIOCGIFVLAN: 3511 case SIOCSIFVLAN: 3512 case SIOCADDDLCI: 3513 case SIOCDELDLCI: 3514 case SIOCGSKNS: 3515 case SIOCGSTAMP_NEW: 3516 case SIOCGSTAMPNS_NEW: 3517 return sock_ioctl(file, cmd, arg); 3518 3519 case SIOCGIFFLAGS: 3520 case SIOCSIFFLAGS: 3521 case SIOCGIFMETRIC: 3522 case SIOCSIFMETRIC: 3523 case SIOCGIFMTU: 3524 case SIOCSIFMTU: 3525 case SIOCGIFMEM: 3526 case SIOCSIFMEM: 3527 case SIOCGIFHWADDR: 3528 case SIOCSIFHWADDR: 3529 case SIOCADDMULTI: 3530 case SIOCDELMULTI: 3531 case SIOCGIFINDEX: 3532 case SIOCGIFADDR: 3533 case SIOCSIFADDR: 3534 case SIOCSIFHWBROADCAST: 3535 case SIOCDIFADDR: 3536 case SIOCGIFBRDADDR: 3537 case SIOCSIFBRDADDR: 3538 case SIOCGIFDSTADDR: 3539 case SIOCSIFDSTADDR: 3540 case SIOCGIFNETMASK: 3541 case SIOCSIFNETMASK: 3542 case SIOCSIFPFLAGS: 3543 case SIOCGIFPFLAGS: 3544 case SIOCGIFTXQLEN: 3545 case SIOCSIFTXQLEN: 3546 case SIOCBRADDIF: 3547 case SIOCBRDELIF: 3548 case SIOCGIFNAME: 3549 case SIOCSIFNAME: 3550 case SIOCGMIIPHY: 3551 case SIOCGMIIREG: 3552 case SIOCSMIIREG: 3553 case SIOCBONDENSLAVE: 3554 case SIOCBONDRELEASE: 3555 case SIOCBONDSETHWADDR: 3556 case SIOCBONDCHANGEACTIVE: 3557 return compat_ifreq_ioctl(net, sock, cmd, argp); 3558 3559 case SIOCSARP: 3560 case SIOCGARP: 3561 case SIOCDARP: 3562 case SIOCOUTQ: 3563 case SIOCOUTQNSD: 3564 case SIOCATMARK: 3565 return sock_do_ioctl(net, sock, cmd, arg); 3566 } 3567 3568 return -ENOIOCTLCMD; 3569 } 3570 3571 static long compat_sock_ioctl(struct file *file, unsigned int cmd, 3572 unsigned long arg) 3573 { 3574 struct socket *sock = file->private_data; 3575 int ret = -ENOIOCTLCMD; 3576 struct sock *sk; 3577 struct net *net; 3578 3579 sk = sock->sk; 3580 net = sock_net(sk); 3581 3582 if (sock->ops->compat_ioctl) 3583 ret = sock->ops->compat_ioctl(sock, cmd, arg); 3584 3585 if (ret == -ENOIOCTLCMD && 3586 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 3587 ret = compat_wext_handle_ioctl(net, cmd, arg); 3588 3589 if (ret == -ENOIOCTLCMD) 3590 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 3591 3592 return ret; 3593 } 3594 #endif 3595 3596 /** 3597 * kernel_bind - bind an address to a socket (kernel space) 3598 * @sock: socket 3599 * @addr: address 3600 * @addrlen: length of address 3601 * 3602 * Returns 0 or an error. 3603 */ 3604 3605 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3606 { 3607 return sock->ops->bind(sock, addr, addrlen); 3608 } 3609 EXPORT_SYMBOL(kernel_bind); 3610 3611 /** 3612 * kernel_listen - move socket to listening state (kernel space) 3613 * @sock: socket 3614 * @backlog: pending connections queue size 3615 * 3616 * Returns 0 or an error. 3617 */ 3618 3619 int kernel_listen(struct socket *sock, int backlog) 3620 { 3621 return sock->ops->listen(sock, backlog); 3622 } 3623 EXPORT_SYMBOL(kernel_listen); 3624 3625 /** 3626 * kernel_accept - accept a connection (kernel space) 3627 * @sock: listening socket 3628 * @newsock: new connected socket 3629 * @flags: flags 3630 * 3631 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0. 3632 * If it fails, @newsock is guaranteed to be %NULL. 3633 * Returns 0 or an error. 3634 */ 3635 3636 int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3637 { 3638 struct sock *sk = sock->sk; 3639 int err; 3640 3641 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3642 newsock); 3643 if (err < 0) 3644 goto done; 3645 3646 err = sock->ops->accept(sock, *newsock, flags, true); 3647 if (err < 0) { 3648 sock_release(*newsock); 3649 *newsock = NULL; 3650 goto done; 3651 } 3652 3653 (*newsock)->ops = sock->ops; 3654 __module_get((*newsock)->ops->owner); 3655 3656 done: 3657 return err; 3658 } 3659 EXPORT_SYMBOL(kernel_accept); 3660 3661 /** 3662 * kernel_connect - connect a socket (kernel space) 3663 * @sock: socket 3664 * @addr: address 3665 * @addrlen: address length 3666 * @flags: flags (O_NONBLOCK, ...) 3667 * 3668 * For datagram sockets, @addr is the addres to which datagrams are sent 3669 * by default, and the only address from which datagrams are received. 3670 * For stream sockets, attempts to connect to @addr. 3671 * Returns 0 or an error code. 3672 */ 3673 3674 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3675 int flags) 3676 { 3677 return sock->ops->connect(sock, addr, addrlen, flags); 3678 } 3679 EXPORT_SYMBOL(kernel_connect); 3680 3681 /** 3682 * kernel_getsockname - get the address which the socket is bound (kernel space) 3683 * @sock: socket 3684 * @addr: address holder 3685 * 3686 * Fills the @addr pointer with the address which the socket is bound. 3687 * Returns 0 or an error code. 3688 */ 3689 3690 int kernel_getsockname(struct socket *sock, struct sockaddr *addr) 3691 { 3692 return sock->ops->getname(sock, addr, 0); 3693 } 3694 EXPORT_SYMBOL(kernel_getsockname); 3695 3696 /** 3697 * kernel_peername - get the address which the socket is connected (kernel space) 3698 * @sock: socket 3699 * @addr: address holder 3700 * 3701 * Fills the @addr pointer with the address which the socket is connected. 3702 * Returns 0 or an error code. 3703 */ 3704 3705 int kernel_getpeername(struct socket *sock, struct sockaddr *addr) 3706 { 3707 return sock->ops->getname(sock, addr, 1); 3708 } 3709 EXPORT_SYMBOL(kernel_getpeername); 3710 3711 /** 3712 * kernel_getsockopt - get a socket option (kernel space) 3713 * @sock: socket 3714 * @level: API level (SOL_SOCKET, ...) 3715 * @optname: option tag 3716 * @optval: option value 3717 * @optlen: option length 3718 * 3719 * Assigns the option length to @optlen. 3720 * Returns 0 or an error. 3721 */ 3722 3723 int kernel_getsockopt(struct socket *sock, int level, int optname, 3724 char *optval, int *optlen) 3725 { 3726 mm_segment_t oldfs = get_fs(); 3727 char __user *uoptval; 3728 int __user *uoptlen; 3729 int err; 3730 3731 uoptval = (char __user __force *) optval; 3732 uoptlen = (int __user __force *) optlen; 3733 3734 set_fs(KERNEL_DS); 3735 if (level == SOL_SOCKET) 3736 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen); 3737 else 3738 err = sock->ops->getsockopt(sock, level, optname, uoptval, 3739 uoptlen); 3740 set_fs(oldfs); 3741 return err; 3742 } 3743 EXPORT_SYMBOL(kernel_getsockopt); 3744 3745 /** 3746 * kernel_setsockopt - set a socket option (kernel space) 3747 * @sock: socket 3748 * @level: API level (SOL_SOCKET, ...) 3749 * @optname: option tag 3750 * @optval: option value 3751 * @optlen: option length 3752 * 3753 * Returns 0 or an error. 3754 */ 3755 3756 int kernel_setsockopt(struct socket *sock, int level, int optname, 3757 char *optval, unsigned int optlen) 3758 { 3759 mm_segment_t oldfs = get_fs(); 3760 char __user *uoptval; 3761 int err; 3762 3763 uoptval = (char __user __force *) optval; 3764 3765 set_fs(KERNEL_DS); 3766 if (level == SOL_SOCKET) 3767 err = sock_setsockopt(sock, level, optname, uoptval, optlen); 3768 else 3769 err = sock->ops->setsockopt(sock, level, optname, uoptval, 3770 optlen); 3771 set_fs(oldfs); 3772 return err; 3773 } 3774 EXPORT_SYMBOL(kernel_setsockopt); 3775 3776 /** 3777 * kernel_sendpage - send a &page through a socket (kernel space) 3778 * @sock: socket 3779 * @page: page 3780 * @offset: page offset 3781 * @size: total size in bytes 3782 * @flags: flags (MSG_DONTWAIT, ...) 3783 * 3784 * Returns the total amount sent in bytes or an error. 3785 */ 3786 3787 int kernel_sendpage(struct socket *sock, struct page *page, int offset, 3788 size_t size, int flags) 3789 { 3790 if (sock->ops->sendpage) 3791 return sock->ops->sendpage(sock, page, offset, size, flags); 3792 3793 return sock_no_sendpage(sock, page, offset, size, flags); 3794 } 3795 EXPORT_SYMBOL(kernel_sendpage); 3796 3797 /** 3798 * kernel_sendpage_locked - send a &page through the locked sock (kernel space) 3799 * @sk: sock 3800 * @page: page 3801 * @offset: page offset 3802 * @size: total size in bytes 3803 * @flags: flags (MSG_DONTWAIT, ...) 3804 * 3805 * Returns the total amount sent in bytes or an error. 3806 * Caller must hold @sk. 3807 */ 3808 3809 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset, 3810 size_t size, int flags) 3811 { 3812 struct socket *sock = sk->sk_socket; 3813 3814 if (sock->ops->sendpage_locked) 3815 return sock->ops->sendpage_locked(sk, page, offset, size, 3816 flags); 3817 3818 return sock_no_sendpage_locked(sk, page, offset, size, flags); 3819 } 3820 EXPORT_SYMBOL(kernel_sendpage_locked); 3821 3822 /** 3823 * kernel_shutdown - shut down part of a full-duplex connection (kernel space) 3824 * @sock: socket 3825 * @how: connection part 3826 * 3827 * Returns 0 or an error. 3828 */ 3829 3830 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3831 { 3832 return sock->ops->shutdown(sock, how); 3833 } 3834 EXPORT_SYMBOL(kernel_sock_shutdown); 3835 3836 /** 3837 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket 3838 * @sk: socket 3839 * 3840 * This routine returns the IP overhead imposed by a socket i.e. 3841 * the length of the underlying IP header, depending on whether 3842 * this is an IPv4 or IPv6 socket and the length from IP options turned 3843 * on at the socket. Assumes that the caller has a lock on the socket. 3844 */ 3845 3846 u32 kernel_sock_ip_overhead(struct sock *sk) 3847 { 3848 struct inet_sock *inet; 3849 struct ip_options_rcu *opt; 3850 u32 overhead = 0; 3851 #if IS_ENABLED(CONFIG_IPV6) 3852 struct ipv6_pinfo *np; 3853 struct ipv6_txoptions *optv6 = NULL; 3854 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3855 3856 if (!sk) 3857 return overhead; 3858 3859 switch (sk->sk_family) { 3860 case AF_INET: 3861 inet = inet_sk(sk); 3862 overhead += sizeof(struct iphdr); 3863 opt = rcu_dereference_protected(inet->inet_opt, 3864 sock_owned_by_user(sk)); 3865 if (opt) 3866 overhead += opt->opt.optlen; 3867 return overhead; 3868 #if IS_ENABLED(CONFIG_IPV6) 3869 case AF_INET6: 3870 np = inet6_sk(sk); 3871 overhead += sizeof(struct ipv6hdr); 3872 if (np) 3873 optv6 = rcu_dereference_protected(np->opt, 3874 sock_owned_by_user(sk)); 3875 if (optv6) 3876 overhead += (optv6->opt_flen + optv6->opt_nflen); 3877 return overhead; 3878 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3879 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ 3880 return overhead; 3881 } 3882 } 3883 EXPORT_SYMBOL(kernel_sock_ip_overhead); 3884