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/bpf-cgroup.h> 56 #include <linux/ethtool.h> 57 #include <linux/mm.h> 58 #include <linux/socket.h> 59 #include <linux/file.h> 60 #include <linux/net.h> 61 #include <linux/interrupt.h> 62 #include <linux/thread_info.h> 63 #include <linux/rcupdate.h> 64 #include <linux/netdevice.h> 65 #include <linux/proc_fs.h> 66 #include <linux/seq_file.h> 67 #include <linux/mutex.h> 68 #include <linux/if_bridge.h> 69 #include <linux/if_vlan.h> 70 #include <linux/ptp_classify.h> 71 #include <linux/init.h> 72 #include <linux/poll.h> 73 #include <linux/cache.h> 74 #include <linux/module.h> 75 #include <linux/highmem.h> 76 #include <linux/mount.h> 77 #include <linux/pseudo_fs.h> 78 #include <linux/security.h> 79 #include <linux/syscalls.h> 80 #include <linux/compat.h> 81 #include <linux/kmod.h> 82 #include <linux/audit.h> 83 #include <linux/wireless.h> 84 #include <linux/nsproxy.h> 85 #include <linux/magic.h> 86 #include <linux/slab.h> 87 #include <linux/xattr.h> 88 #include <linux/nospec.h> 89 #include <linux/indirect_call_wrapper.h> 90 91 #include <linux/uaccess.h> 92 #include <asm/unistd.h> 93 94 #include <net/compat.h> 95 #include <net/wext.h> 96 #include <net/cls_cgroup.h> 97 98 #include <net/sock.h> 99 #include <linux/netfilter.h> 100 101 #include <linux/if_tun.h> 102 #include <linux/ipv6_route.h> 103 #include <linux/route.h> 104 #include <linux/termios.h> 105 #include <linux/sockios.h> 106 #include <net/busy_poll.h> 107 #include <linux/errqueue.h> 108 #include <linux/ptp_clock_kernel.h> 109 #include <trace/events/sock.h> 110 111 #ifdef CONFIG_NET_RX_BUSY_POLL 112 unsigned int sysctl_net_busy_read __read_mostly; 113 unsigned int sysctl_net_busy_poll __read_mostly; 114 #endif 115 116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to); 117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from); 118 static int sock_mmap(struct file *file, struct vm_area_struct *vma); 119 120 static int sock_close(struct inode *inode, struct file *file); 121 static __poll_t sock_poll(struct file *file, 122 struct poll_table_struct *wait); 123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 124 #ifdef CONFIG_COMPAT 125 static long compat_sock_ioctl(struct file *file, 126 unsigned int cmd, unsigned long arg); 127 #endif 128 static int sock_fasync(int fd, struct file *filp, int on); 129 static ssize_t sock_sendpage(struct file *file, struct page *page, 130 int offset, size_t size, loff_t *ppos, int more); 131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 132 struct pipe_inode_info *pipe, size_t len, 133 unsigned int flags); 134 135 #ifdef CONFIG_PROC_FS 136 static void sock_show_fdinfo(struct seq_file *m, struct file *f) 137 { 138 struct socket *sock = f->private_data; 139 140 if (sock->ops->show_fdinfo) 141 sock->ops->show_fdinfo(m, sock); 142 } 143 #else 144 #define sock_show_fdinfo NULL 145 #endif 146 147 /* 148 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear 149 * in the operation structures but are done directly via the socketcall() multiplexor. 150 */ 151 152 static const struct file_operations socket_file_ops = { 153 .owner = THIS_MODULE, 154 .llseek = no_llseek, 155 .read_iter = sock_read_iter, 156 .write_iter = sock_write_iter, 157 .poll = sock_poll, 158 .unlocked_ioctl = sock_ioctl, 159 #ifdef CONFIG_COMPAT 160 .compat_ioctl = compat_sock_ioctl, 161 #endif 162 .mmap = sock_mmap, 163 .release = sock_close, 164 .fasync = sock_fasync, 165 .sendpage = sock_sendpage, 166 .splice_write = generic_splice_sendpage, 167 .splice_read = sock_splice_read, 168 .show_fdinfo = sock_show_fdinfo, 169 }; 170 171 static const char * const pf_family_names[] = { 172 [PF_UNSPEC] = "PF_UNSPEC", 173 [PF_UNIX] = "PF_UNIX/PF_LOCAL", 174 [PF_INET] = "PF_INET", 175 [PF_AX25] = "PF_AX25", 176 [PF_IPX] = "PF_IPX", 177 [PF_APPLETALK] = "PF_APPLETALK", 178 [PF_NETROM] = "PF_NETROM", 179 [PF_BRIDGE] = "PF_BRIDGE", 180 [PF_ATMPVC] = "PF_ATMPVC", 181 [PF_X25] = "PF_X25", 182 [PF_INET6] = "PF_INET6", 183 [PF_ROSE] = "PF_ROSE", 184 [PF_DECnet] = "PF_DECnet", 185 [PF_NETBEUI] = "PF_NETBEUI", 186 [PF_SECURITY] = "PF_SECURITY", 187 [PF_KEY] = "PF_KEY", 188 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE", 189 [PF_PACKET] = "PF_PACKET", 190 [PF_ASH] = "PF_ASH", 191 [PF_ECONET] = "PF_ECONET", 192 [PF_ATMSVC] = "PF_ATMSVC", 193 [PF_RDS] = "PF_RDS", 194 [PF_SNA] = "PF_SNA", 195 [PF_IRDA] = "PF_IRDA", 196 [PF_PPPOX] = "PF_PPPOX", 197 [PF_WANPIPE] = "PF_WANPIPE", 198 [PF_LLC] = "PF_LLC", 199 [PF_IB] = "PF_IB", 200 [PF_MPLS] = "PF_MPLS", 201 [PF_CAN] = "PF_CAN", 202 [PF_TIPC] = "PF_TIPC", 203 [PF_BLUETOOTH] = "PF_BLUETOOTH", 204 [PF_IUCV] = "PF_IUCV", 205 [PF_RXRPC] = "PF_RXRPC", 206 [PF_ISDN] = "PF_ISDN", 207 [PF_PHONET] = "PF_PHONET", 208 [PF_IEEE802154] = "PF_IEEE802154", 209 [PF_CAIF] = "PF_CAIF", 210 [PF_ALG] = "PF_ALG", 211 [PF_NFC] = "PF_NFC", 212 [PF_VSOCK] = "PF_VSOCK", 213 [PF_KCM] = "PF_KCM", 214 [PF_QIPCRTR] = "PF_QIPCRTR", 215 [PF_SMC] = "PF_SMC", 216 [PF_XDP] = "PF_XDP", 217 [PF_MCTP] = "PF_MCTP", 218 }; 219 220 /* 221 * The protocol list. Each protocol is registered in here. 222 */ 223 224 static DEFINE_SPINLOCK(net_family_lock); 225 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; 226 227 /* 228 * Support routines. 229 * Move socket addresses back and forth across the kernel/user 230 * divide and look after the messy bits. 231 */ 232 233 /** 234 * move_addr_to_kernel - copy a socket address into kernel space 235 * @uaddr: Address in user space 236 * @kaddr: Address in kernel space 237 * @ulen: Length in user space 238 * 239 * The address is copied into kernel space. If the provided address is 240 * too long an error code of -EINVAL is returned. If the copy gives 241 * invalid addresses -EFAULT is returned. On a success 0 is returned. 242 */ 243 244 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr) 245 { 246 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) 247 return -EINVAL; 248 if (ulen == 0) 249 return 0; 250 if (copy_from_user(kaddr, uaddr, ulen)) 251 return -EFAULT; 252 return audit_sockaddr(ulen, kaddr); 253 } 254 255 /** 256 * move_addr_to_user - copy an address to user space 257 * @kaddr: kernel space address 258 * @klen: length of address in kernel 259 * @uaddr: user space address 260 * @ulen: pointer to user length field 261 * 262 * The value pointed to by ulen on entry is the buffer length available. 263 * This is overwritten with the buffer space used. -EINVAL is returned 264 * if an overlong buffer is specified or a negative buffer size. -EFAULT 265 * is returned if either the buffer or the length field are not 266 * accessible. 267 * After copying the data up to the limit the user specifies, the true 268 * length of the data is written over the length limit the user 269 * specified. Zero is returned for a success. 270 */ 271 272 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen, 273 void __user *uaddr, int __user *ulen) 274 { 275 int err; 276 int len; 277 278 BUG_ON(klen > sizeof(struct sockaddr_storage)); 279 err = get_user(len, ulen); 280 if (err) 281 return err; 282 if (len > klen) 283 len = klen; 284 if (len < 0) 285 return -EINVAL; 286 if (len) { 287 if (audit_sockaddr(klen, kaddr)) 288 return -ENOMEM; 289 if (copy_to_user(uaddr, kaddr, len)) 290 return -EFAULT; 291 } 292 /* 293 * "fromlen shall refer to the value before truncation.." 294 * 1003.1g 295 */ 296 return __put_user(klen, ulen); 297 } 298 299 static struct kmem_cache *sock_inode_cachep __ro_after_init; 300 301 static struct inode *sock_alloc_inode(struct super_block *sb) 302 { 303 struct socket_alloc *ei; 304 305 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL); 306 if (!ei) 307 return NULL; 308 init_waitqueue_head(&ei->socket.wq.wait); 309 ei->socket.wq.fasync_list = NULL; 310 ei->socket.wq.flags = 0; 311 312 ei->socket.state = SS_UNCONNECTED; 313 ei->socket.flags = 0; 314 ei->socket.ops = NULL; 315 ei->socket.sk = NULL; 316 ei->socket.file = NULL; 317 318 return &ei->vfs_inode; 319 } 320 321 static void sock_free_inode(struct inode *inode) 322 { 323 struct socket_alloc *ei; 324 325 ei = container_of(inode, struct socket_alloc, vfs_inode); 326 kmem_cache_free(sock_inode_cachep, ei); 327 } 328 329 static void init_once(void *foo) 330 { 331 struct socket_alloc *ei = (struct socket_alloc *)foo; 332 333 inode_init_once(&ei->vfs_inode); 334 } 335 336 static void init_inodecache(void) 337 { 338 sock_inode_cachep = kmem_cache_create("sock_inode_cache", 339 sizeof(struct socket_alloc), 340 0, 341 (SLAB_HWCACHE_ALIGN | 342 SLAB_RECLAIM_ACCOUNT | 343 SLAB_MEM_SPREAD | SLAB_ACCOUNT), 344 init_once); 345 BUG_ON(sock_inode_cachep == NULL); 346 } 347 348 static const struct super_operations sockfs_ops = { 349 .alloc_inode = sock_alloc_inode, 350 .free_inode = sock_free_inode, 351 .statfs = simple_statfs, 352 }; 353 354 /* 355 * sockfs_dname() is called from d_path(). 356 */ 357 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) 358 { 359 return dynamic_dname(buffer, buflen, "socket:[%lu]", 360 d_inode(dentry)->i_ino); 361 } 362 363 static const struct dentry_operations sockfs_dentry_operations = { 364 .d_dname = sockfs_dname, 365 }; 366 367 static int sockfs_xattr_get(const struct xattr_handler *handler, 368 struct dentry *dentry, struct inode *inode, 369 const char *suffix, void *value, size_t size) 370 { 371 if (value) { 372 if (dentry->d_name.len + 1 > size) 373 return -ERANGE; 374 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1); 375 } 376 return dentry->d_name.len + 1; 377 } 378 379 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname" 380 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX) 381 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1) 382 383 static const struct xattr_handler sockfs_xattr_handler = { 384 .name = XATTR_NAME_SOCKPROTONAME, 385 .get = sockfs_xattr_get, 386 }; 387 388 static int sockfs_security_xattr_set(const struct xattr_handler *handler, 389 struct mnt_idmap *idmap, 390 struct dentry *dentry, struct inode *inode, 391 const char *suffix, const void *value, 392 size_t size, int flags) 393 { 394 /* Handled by LSM. */ 395 return -EAGAIN; 396 } 397 398 static const struct xattr_handler sockfs_security_xattr_handler = { 399 .prefix = XATTR_SECURITY_PREFIX, 400 .set = sockfs_security_xattr_set, 401 }; 402 403 static const struct xattr_handler *sockfs_xattr_handlers[] = { 404 &sockfs_xattr_handler, 405 &sockfs_security_xattr_handler, 406 NULL 407 }; 408 409 static int sockfs_init_fs_context(struct fs_context *fc) 410 { 411 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC); 412 if (!ctx) 413 return -ENOMEM; 414 ctx->ops = &sockfs_ops; 415 ctx->dops = &sockfs_dentry_operations; 416 ctx->xattr = sockfs_xattr_handlers; 417 return 0; 418 } 419 420 static struct vfsmount *sock_mnt __read_mostly; 421 422 static struct file_system_type sock_fs_type = { 423 .name = "sockfs", 424 .init_fs_context = sockfs_init_fs_context, 425 .kill_sb = kill_anon_super, 426 }; 427 428 /* 429 * Obtains the first available file descriptor and sets it up for use. 430 * 431 * These functions create file structures and maps them to fd space 432 * of the current process. On success it returns file descriptor 433 * and file struct implicitly stored in sock->file. 434 * Note that another thread may close file descriptor before we return 435 * from this function. We use the fact that now we do not refer 436 * to socket after mapping. If one day we will need it, this 437 * function will increment ref. count on file by 1. 438 * 439 * In any case returned fd MAY BE not valid! 440 * This race condition is unavoidable 441 * with shared fd spaces, we cannot solve it inside kernel, 442 * but we take care of internal coherence yet. 443 */ 444 445 /** 446 * sock_alloc_file - Bind a &socket to a &file 447 * @sock: socket 448 * @flags: file status flags 449 * @dname: protocol name 450 * 451 * Returns the &file bound with @sock, implicitly storing it 452 * in sock->file. If dname is %NULL, sets to "". 453 * 454 * On failure @sock is released, and an ERR pointer is returned. 455 * 456 * This function uses GFP_KERNEL internally. 457 */ 458 459 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname) 460 { 461 struct file *file; 462 463 if (!dname) 464 dname = sock->sk ? sock->sk->sk_prot_creator->name : ""; 465 466 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname, 467 O_RDWR | (flags & O_NONBLOCK), 468 &socket_file_ops); 469 if (IS_ERR(file)) { 470 sock_release(sock); 471 return file; 472 } 473 474 sock->file = file; 475 file->private_data = sock; 476 stream_open(SOCK_INODE(sock), file); 477 return file; 478 } 479 EXPORT_SYMBOL(sock_alloc_file); 480 481 static int sock_map_fd(struct socket *sock, int flags) 482 { 483 struct file *newfile; 484 int fd = get_unused_fd_flags(flags); 485 if (unlikely(fd < 0)) { 486 sock_release(sock); 487 return fd; 488 } 489 490 newfile = sock_alloc_file(sock, flags, NULL); 491 if (!IS_ERR(newfile)) { 492 fd_install(fd, newfile); 493 return fd; 494 } 495 496 put_unused_fd(fd); 497 return PTR_ERR(newfile); 498 } 499 500 /** 501 * sock_from_file - Return the &socket bounded to @file. 502 * @file: file 503 * 504 * On failure returns %NULL. 505 */ 506 507 struct socket *sock_from_file(struct file *file) 508 { 509 if (file->f_op == &socket_file_ops) 510 return file->private_data; /* set in sock_alloc_file */ 511 512 return NULL; 513 } 514 EXPORT_SYMBOL(sock_from_file); 515 516 /** 517 * sockfd_lookup - Go from a file number to its socket slot 518 * @fd: file handle 519 * @err: pointer to an error code return 520 * 521 * The file handle passed in is locked and the socket it is bound 522 * to is returned. If an error occurs the err pointer is overwritten 523 * with a negative errno code and NULL is returned. The function checks 524 * for both invalid handles and passing a handle which is not a socket. 525 * 526 * On a success the socket object pointer is returned. 527 */ 528 529 struct socket *sockfd_lookup(int fd, int *err) 530 { 531 struct file *file; 532 struct socket *sock; 533 534 file = fget(fd); 535 if (!file) { 536 *err = -EBADF; 537 return NULL; 538 } 539 540 sock = sock_from_file(file); 541 if (!sock) { 542 *err = -ENOTSOCK; 543 fput(file); 544 } 545 return sock; 546 } 547 EXPORT_SYMBOL(sockfd_lookup); 548 549 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) 550 { 551 struct fd f = fdget(fd); 552 struct socket *sock; 553 554 *err = -EBADF; 555 if (f.file) { 556 sock = sock_from_file(f.file); 557 if (likely(sock)) { 558 *fput_needed = f.flags & FDPUT_FPUT; 559 return sock; 560 } 561 *err = -ENOTSOCK; 562 fdput(f); 563 } 564 return NULL; 565 } 566 567 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer, 568 size_t size) 569 { 570 ssize_t len; 571 ssize_t used = 0; 572 573 len = security_inode_listsecurity(d_inode(dentry), buffer, size); 574 if (len < 0) 575 return len; 576 used += len; 577 if (buffer) { 578 if (size < used) 579 return -ERANGE; 580 buffer += len; 581 } 582 583 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1); 584 used += len; 585 if (buffer) { 586 if (size < used) 587 return -ERANGE; 588 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len); 589 buffer += len; 590 } 591 592 return used; 593 } 594 595 static int sockfs_setattr(struct mnt_idmap *idmap, 596 struct dentry *dentry, struct iattr *iattr) 597 { 598 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr); 599 600 if (!err && (iattr->ia_valid & ATTR_UID)) { 601 struct socket *sock = SOCKET_I(d_inode(dentry)); 602 603 if (sock->sk) 604 sock->sk->sk_uid = iattr->ia_uid; 605 else 606 err = -ENOENT; 607 } 608 609 return err; 610 } 611 612 static const struct inode_operations sockfs_inode_ops = { 613 .listxattr = sockfs_listxattr, 614 .setattr = sockfs_setattr, 615 }; 616 617 /** 618 * sock_alloc - allocate a socket 619 * 620 * Allocate a new inode and socket object. The two are bound together 621 * and initialised. The socket is then returned. If we are out of inodes 622 * NULL is returned. This functions uses GFP_KERNEL internally. 623 */ 624 625 struct socket *sock_alloc(void) 626 { 627 struct inode *inode; 628 struct socket *sock; 629 630 inode = new_inode_pseudo(sock_mnt->mnt_sb); 631 if (!inode) 632 return NULL; 633 634 sock = SOCKET_I(inode); 635 636 inode->i_ino = get_next_ino(); 637 inode->i_mode = S_IFSOCK | S_IRWXUGO; 638 inode->i_uid = current_fsuid(); 639 inode->i_gid = current_fsgid(); 640 inode->i_op = &sockfs_inode_ops; 641 642 return sock; 643 } 644 EXPORT_SYMBOL(sock_alloc); 645 646 static void __sock_release(struct socket *sock, struct inode *inode) 647 { 648 if (sock->ops) { 649 struct module *owner = sock->ops->owner; 650 651 if (inode) 652 inode_lock(inode); 653 sock->ops->release(sock); 654 sock->sk = NULL; 655 if (inode) 656 inode_unlock(inode); 657 sock->ops = NULL; 658 module_put(owner); 659 } 660 661 if (sock->wq.fasync_list) 662 pr_err("%s: fasync list not empty!\n", __func__); 663 664 if (!sock->file) { 665 iput(SOCK_INODE(sock)); 666 return; 667 } 668 sock->file = NULL; 669 } 670 671 /** 672 * sock_release - close a socket 673 * @sock: socket to close 674 * 675 * The socket is released from the protocol stack if it has a release 676 * callback, and the inode is then released if the socket is bound to 677 * an inode not a file. 678 */ 679 void sock_release(struct socket *sock) 680 { 681 __sock_release(sock, NULL); 682 } 683 EXPORT_SYMBOL(sock_release); 684 685 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags) 686 { 687 u8 flags = *tx_flags; 688 689 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) { 690 flags |= SKBTX_HW_TSTAMP; 691 692 /* PTP hardware clocks can provide a free running cycle counter 693 * as a time base for virtual clocks. Tell driver to use the 694 * free running cycle counter for timestamp if socket is bound 695 * to virtual clock. 696 */ 697 if (tsflags & SOF_TIMESTAMPING_BIND_PHC) 698 flags |= SKBTX_HW_TSTAMP_USE_CYCLES; 699 } 700 701 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE) 702 flags |= SKBTX_SW_TSTAMP; 703 704 if (tsflags & SOF_TIMESTAMPING_TX_SCHED) 705 flags |= SKBTX_SCHED_TSTAMP; 706 707 *tx_flags = flags; 708 } 709 EXPORT_SYMBOL(__sock_tx_timestamp); 710 711 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *, 712 size_t)); 713 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *, 714 size_t)); 715 716 static noinline void call_trace_sock_send_length(struct sock *sk, int ret, 717 int flags) 718 { 719 trace_sock_send_length(sk, ret, 0); 720 } 721 722 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) 723 { 724 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg, 725 inet_sendmsg, sock, msg, 726 msg_data_left(msg)); 727 BUG_ON(ret == -EIOCBQUEUED); 728 729 if (trace_sock_send_length_enabled()) 730 call_trace_sock_send_length(sock->sk, ret, 0); 731 return ret; 732 } 733 734 /** 735 * sock_sendmsg - send a message through @sock 736 * @sock: socket 737 * @msg: message to send 738 * 739 * Sends @msg through @sock, passing through LSM. 740 * Returns the number of bytes sent, or an error code. 741 */ 742 int sock_sendmsg(struct socket *sock, struct msghdr *msg) 743 { 744 int err = security_socket_sendmsg(sock, msg, 745 msg_data_left(msg)); 746 747 return err ?: sock_sendmsg_nosec(sock, msg); 748 } 749 EXPORT_SYMBOL(sock_sendmsg); 750 751 /** 752 * kernel_sendmsg - send a message through @sock (kernel-space) 753 * @sock: socket 754 * @msg: message header 755 * @vec: kernel vec 756 * @num: vec array length 757 * @size: total message data size 758 * 759 * Builds the message data with @vec and sends it through @sock. 760 * Returns the number of bytes sent, or an error code. 761 */ 762 763 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, 764 struct kvec *vec, size_t num, size_t size) 765 { 766 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size); 767 return sock_sendmsg(sock, msg); 768 } 769 EXPORT_SYMBOL(kernel_sendmsg); 770 771 /** 772 * kernel_sendmsg_locked - send a message through @sock (kernel-space) 773 * @sk: sock 774 * @msg: message header 775 * @vec: output s/g array 776 * @num: output s/g array length 777 * @size: total message data size 778 * 779 * Builds the message data with @vec and sends it through @sock. 780 * Returns the number of bytes sent, or an error code. 781 * Caller must hold @sk. 782 */ 783 784 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg, 785 struct kvec *vec, size_t num, size_t size) 786 { 787 struct socket *sock = sk->sk_socket; 788 789 if (!sock->ops->sendmsg_locked) 790 return sock_no_sendmsg_locked(sk, msg, size); 791 792 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size); 793 794 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg)); 795 } 796 EXPORT_SYMBOL(kernel_sendmsg_locked); 797 798 static bool skb_is_err_queue(const struct sk_buff *skb) 799 { 800 /* pkt_type of skbs enqueued on the error queue are set to 801 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do 802 * in recvmsg, since skbs received on a local socket will never 803 * have a pkt_type of PACKET_OUTGOING. 804 */ 805 return skb->pkt_type == PACKET_OUTGOING; 806 } 807 808 /* On transmit, software and hardware timestamps are returned independently. 809 * As the two skb clones share the hardware timestamp, which may be updated 810 * before the software timestamp is received, a hardware TX timestamp may be 811 * returned only if there is no software TX timestamp. Ignore false software 812 * timestamps, which may be made in the __sock_recv_timestamp() call when the 813 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a 814 * hardware timestamp. 815 */ 816 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp) 817 { 818 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb); 819 } 820 821 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index) 822 { 823 bool cycles = sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC; 824 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); 825 struct net_device *orig_dev; 826 ktime_t hwtstamp; 827 828 rcu_read_lock(); 829 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 830 if (orig_dev) { 831 *if_index = orig_dev->ifindex; 832 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles); 833 } else { 834 hwtstamp = shhwtstamps->hwtstamp; 835 } 836 rcu_read_unlock(); 837 838 return hwtstamp; 839 } 840 841 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb, 842 int if_index) 843 { 844 struct scm_ts_pktinfo ts_pktinfo; 845 struct net_device *orig_dev; 846 847 if (!skb_mac_header_was_set(skb)) 848 return; 849 850 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo)); 851 852 if (!if_index) { 853 rcu_read_lock(); 854 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 855 if (orig_dev) 856 if_index = orig_dev->ifindex; 857 rcu_read_unlock(); 858 } 859 ts_pktinfo.if_index = if_index; 860 861 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb); 862 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO, 863 sizeof(ts_pktinfo), &ts_pktinfo); 864 } 865 866 /* 867 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) 868 */ 869 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 870 struct sk_buff *skb) 871 { 872 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); 873 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW); 874 struct scm_timestamping_internal tss; 875 876 int empty = 1, false_tstamp = 0; 877 struct skb_shared_hwtstamps *shhwtstamps = 878 skb_hwtstamps(skb); 879 int if_index; 880 ktime_t hwtstamp; 881 882 /* Race occurred between timestamp enabling and packet 883 receiving. Fill in the current time for now. */ 884 if (need_software_tstamp && skb->tstamp == 0) { 885 __net_timestamp(skb); 886 false_tstamp = 1; 887 } 888 889 if (need_software_tstamp) { 890 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { 891 if (new_tstamp) { 892 struct __kernel_sock_timeval tv; 893 894 skb_get_new_timestamp(skb, &tv); 895 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW, 896 sizeof(tv), &tv); 897 } else { 898 struct __kernel_old_timeval tv; 899 900 skb_get_timestamp(skb, &tv); 901 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD, 902 sizeof(tv), &tv); 903 } 904 } else { 905 if (new_tstamp) { 906 struct __kernel_timespec ts; 907 908 skb_get_new_timestampns(skb, &ts); 909 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW, 910 sizeof(ts), &ts); 911 } else { 912 struct __kernel_old_timespec ts; 913 914 skb_get_timestampns(skb, &ts); 915 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD, 916 sizeof(ts), &ts); 917 } 918 } 919 } 920 921 memset(&tss, 0, sizeof(tss)); 922 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) && 923 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0)) 924 empty = 0; 925 if (shhwtstamps && 926 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) && 927 !skb_is_swtx_tstamp(skb, false_tstamp)) { 928 if_index = 0; 929 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV) 930 hwtstamp = get_timestamp(sk, skb, &if_index); 931 else 932 hwtstamp = shhwtstamps->hwtstamp; 933 934 if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC) 935 hwtstamp = ptp_convert_timestamp(&hwtstamp, 936 sk->sk_bind_phc); 937 938 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) { 939 empty = 0; 940 941 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) && 942 !skb_is_err_queue(skb)) 943 put_ts_pktinfo(msg, skb, if_index); 944 } 945 } 946 if (!empty) { 947 if (sock_flag(sk, SOCK_TSTAMP_NEW)) 948 put_cmsg_scm_timestamping64(msg, &tss); 949 else 950 put_cmsg_scm_timestamping(msg, &tss); 951 952 if (skb_is_err_queue(skb) && skb->len && 953 SKB_EXT_ERR(skb)->opt_stats) 954 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS, 955 skb->len, skb->data); 956 } 957 } 958 EXPORT_SYMBOL_GPL(__sock_recv_timestamp); 959 960 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, 961 struct sk_buff *skb) 962 { 963 int ack; 964 965 if (!sock_flag(sk, SOCK_WIFI_STATUS)) 966 return; 967 if (!skb->wifi_acked_valid) 968 return; 969 970 ack = skb->wifi_acked; 971 972 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); 973 } 974 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); 975 976 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, 977 struct sk_buff *skb) 978 { 979 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount) 980 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, 981 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount); 982 } 983 984 static void sock_recv_mark(struct msghdr *msg, struct sock *sk, 985 struct sk_buff *skb) 986 { 987 if (sock_flag(sk, SOCK_RCVMARK) && skb) { 988 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */ 989 __u32 mark = skb->mark; 990 991 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark); 992 } 993 } 994 995 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk, 996 struct sk_buff *skb) 997 { 998 sock_recv_timestamp(msg, sk, skb); 999 sock_recv_drops(msg, sk, skb); 1000 sock_recv_mark(msg, sk, skb); 1001 } 1002 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs); 1003 1004 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *, 1005 size_t, int)); 1006 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *, 1007 size_t, int)); 1008 1009 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags) 1010 { 1011 trace_sock_recv_length(sk, ret, flags); 1012 } 1013 1014 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 1015 int flags) 1016 { 1017 int ret = INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg, 1018 inet_recvmsg, sock, msg, 1019 msg_data_left(msg), flags); 1020 if (trace_sock_recv_length_enabled()) 1021 call_trace_sock_recv_length(sock->sk, ret, flags); 1022 return ret; 1023 } 1024 1025 /** 1026 * sock_recvmsg - receive a message from @sock 1027 * @sock: socket 1028 * @msg: message to receive 1029 * @flags: message flags 1030 * 1031 * Receives @msg from @sock, passing through LSM. Returns the total number 1032 * of bytes received, or an error. 1033 */ 1034 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) 1035 { 1036 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); 1037 1038 return err ?: sock_recvmsg_nosec(sock, msg, flags); 1039 } 1040 EXPORT_SYMBOL(sock_recvmsg); 1041 1042 /** 1043 * kernel_recvmsg - Receive a message from a socket (kernel space) 1044 * @sock: The socket to receive the message from 1045 * @msg: Received message 1046 * @vec: Input s/g array for message data 1047 * @num: Size of input s/g array 1048 * @size: Number of bytes to read 1049 * @flags: Message flags (MSG_DONTWAIT, etc...) 1050 * 1051 * On return the msg structure contains the scatter/gather array passed in the 1052 * vec argument. The array is modified so that it consists of the unfilled 1053 * portion of the original array. 1054 * 1055 * The returned value is the total number of bytes received, or an error. 1056 */ 1057 1058 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 1059 struct kvec *vec, size_t num, size_t size, int flags) 1060 { 1061 msg->msg_control_is_user = false; 1062 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size); 1063 return sock_recvmsg(sock, msg, flags); 1064 } 1065 EXPORT_SYMBOL(kernel_recvmsg); 1066 1067 static ssize_t sock_sendpage(struct file *file, struct page *page, 1068 int offset, size_t size, loff_t *ppos, int more) 1069 { 1070 struct socket *sock; 1071 int flags; 1072 int ret; 1073 1074 sock = file->private_data; 1075 1076 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 1077 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */ 1078 flags |= more; 1079 1080 ret = kernel_sendpage(sock, page, offset, size, flags); 1081 1082 if (trace_sock_send_length_enabled()) 1083 call_trace_sock_send_length(sock->sk, ret, 0); 1084 return ret; 1085 } 1086 1087 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 1088 struct pipe_inode_info *pipe, size_t len, 1089 unsigned int flags) 1090 { 1091 struct socket *sock = file->private_data; 1092 1093 if (unlikely(!sock->ops->splice_read)) 1094 return generic_file_splice_read(file, ppos, pipe, len, flags); 1095 1096 return sock->ops->splice_read(sock, ppos, pipe, len, flags); 1097 } 1098 1099 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) 1100 { 1101 struct file *file = iocb->ki_filp; 1102 struct socket *sock = file->private_data; 1103 struct msghdr msg = {.msg_iter = *to, 1104 .msg_iocb = iocb}; 1105 ssize_t res; 1106 1107 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1108 msg.msg_flags = MSG_DONTWAIT; 1109 1110 if (iocb->ki_pos != 0) 1111 return -ESPIPE; 1112 1113 if (!iov_iter_count(to)) /* Match SYS5 behaviour */ 1114 return 0; 1115 1116 res = sock_recvmsg(sock, &msg, msg.msg_flags); 1117 *to = msg.msg_iter; 1118 return res; 1119 } 1120 1121 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) 1122 { 1123 struct file *file = iocb->ki_filp; 1124 struct socket *sock = file->private_data; 1125 struct msghdr msg = {.msg_iter = *from, 1126 .msg_iocb = iocb}; 1127 ssize_t res; 1128 1129 if (iocb->ki_pos != 0) 1130 return -ESPIPE; 1131 1132 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1133 msg.msg_flags = MSG_DONTWAIT; 1134 1135 if (sock->type == SOCK_SEQPACKET) 1136 msg.msg_flags |= MSG_EOR; 1137 1138 res = sock_sendmsg(sock, &msg); 1139 *from = msg.msg_iter; 1140 return res; 1141 } 1142 1143 /* 1144 * Atomic setting of ioctl hooks to avoid race 1145 * with module unload. 1146 */ 1147 1148 static DEFINE_MUTEX(br_ioctl_mutex); 1149 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br, 1150 unsigned int cmd, struct ifreq *ifr, 1151 void __user *uarg); 1152 1153 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br, 1154 unsigned int cmd, struct ifreq *ifr, 1155 void __user *uarg)) 1156 { 1157 mutex_lock(&br_ioctl_mutex); 1158 br_ioctl_hook = hook; 1159 mutex_unlock(&br_ioctl_mutex); 1160 } 1161 EXPORT_SYMBOL(brioctl_set); 1162 1163 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd, 1164 struct ifreq *ifr, void __user *uarg) 1165 { 1166 int err = -ENOPKG; 1167 1168 if (!br_ioctl_hook) 1169 request_module("bridge"); 1170 1171 mutex_lock(&br_ioctl_mutex); 1172 if (br_ioctl_hook) 1173 err = br_ioctl_hook(net, br, cmd, ifr, uarg); 1174 mutex_unlock(&br_ioctl_mutex); 1175 1176 return err; 1177 } 1178 1179 static DEFINE_MUTEX(vlan_ioctl_mutex); 1180 static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 1181 1182 void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 1183 { 1184 mutex_lock(&vlan_ioctl_mutex); 1185 vlan_ioctl_hook = hook; 1186 mutex_unlock(&vlan_ioctl_mutex); 1187 } 1188 EXPORT_SYMBOL(vlan_ioctl_set); 1189 1190 static long sock_do_ioctl(struct net *net, struct socket *sock, 1191 unsigned int cmd, unsigned long arg) 1192 { 1193 struct ifreq ifr; 1194 bool need_copyout; 1195 int err; 1196 void __user *argp = (void __user *)arg; 1197 void __user *data; 1198 1199 err = sock->ops->ioctl(sock, cmd, arg); 1200 1201 /* 1202 * If this ioctl is unknown try to hand it down 1203 * to the NIC driver. 1204 */ 1205 if (err != -ENOIOCTLCMD) 1206 return err; 1207 1208 if (!is_socket_ioctl_cmd(cmd)) 1209 return -ENOTTY; 1210 1211 if (get_user_ifreq(&ifr, &data, argp)) 1212 return -EFAULT; 1213 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); 1214 if (!err && need_copyout) 1215 if (put_user_ifreq(&ifr, argp)) 1216 return -EFAULT; 1217 1218 return err; 1219 } 1220 1221 /* 1222 * With an ioctl, arg may well be a user mode pointer, but we don't know 1223 * what to do with it - that's up to the protocol still. 1224 */ 1225 1226 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 1227 { 1228 struct socket *sock; 1229 struct sock *sk; 1230 void __user *argp = (void __user *)arg; 1231 int pid, err; 1232 struct net *net; 1233 1234 sock = file->private_data; 1235 sk = sock->sk; 1236 net = sock_net(sk); 1237 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { 1238 struct ifreq ifr; 1239 void __user *data; 1240 bool need_copyout; 1241 if (get_user_ifreq(&ifr, &data, argp)) 1242 return -EFAULT; 1243 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); 1244 if (!err && need_copyout) 1245 if (put_user_ifreq(&ifr, argp)) 1246 return -EFAULT; 1247 } else 1248 #ifdef CONFIG_WEXT_CORE 1249 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 1250 err = wext_handle_ioctl(net, cmd, argp); 1251 } else 1252 #endif 1253 switch (cmd) { 1254 case FIOSETOWN: 1255 case SIOCSPGRP: 1256 err = -EFAULT; 1257 if (get_user(pid, (int __user *)argp)) 1258 break; 1259 err = f_setown(sock->file, pid, 1); 1260 break; 1261 case FIOGETOWN: 1262 case SIOCGPGRP: 1263 err = put_user(f_getown(sock->file), 1264 (int __user *)argp); 1265 break; 1266 case SIOCGIFBR: 1267 case SIOCSIFBR: 1268 case SIOCBRADDBR: 1269 case SIOCBRDELBR: 1270 err = br_ioctl_call(net, NULL, cmd, NULL, argp); 1271 break; 1272 case SIOCGIFVLAN: 1273 case SIOCSIFVLAN: 1274 err = -ENOPKG; 1275 if (!vlan_ioctl_hook) 1276 request_module("8021q"); 1277 1278 mutex_lock(&vlan_ioctl_mutex); 1279 if (vlan_ioctl_hook) 1280 err = vlan_ioctl_hook(net, argp); 1281 mutex_unlock(&vlan_ioctl_mutex); 1282 break; 1283 case SIOCGSKNS: 1284 err = -EPERM; 1285 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1286 break; 1287 1288 err = open_related_ns(&net->ns, get_net_ns); 1289 break; 1290 case SIOCGSTAMP_OLD: 1291 case SIOCGSTAMPNS_OLD: 1292 if (!sock->ops->gettstamp) { 1293 err = -ENOIOCTLCMD; 1294 break; 1295 } 1296 err = sock->ops->gettstamp(sock, argp, 1297 cmd == SIOCGSTAMP_OLD, 1298 !IS_ENABLED(CONFIG_64BIT)); 1299 break; 1300 case SIOCGSTAMP_NEW: 1301 case SIOCGSTAMPNS_NEW: 1302 if (!sock->ops->gettstamp) { 1303 err = -ENOIOCTLCMD; 1304 break; 1305 } 1306 err = sock->ops->gettstamp(sock, argp, 1307 cmd == SIOCGSTAMP_NEW, 1308 false); 1309 break; 1310 1311 case SIOCGIFCONF: 1312 err = dev_ifconf(net, argp); 1313 break; 1314 1315 default: 1316 err = sock_do_ioctl(net, sock, cmd, arg); 1317 break; 1318 } 1319 return err; 1320 } 1321 1322 /** 1323 * sock_create_lite - creates a socket 1324 * @family: protocol family (AF_INET, ...) 1325 * @type: communication type (SOCK_STREAM, ...) 1326 * @protocol: protocol (0, ...) 1327 * @res: new socket 1328 * 1329 * Creates a new socket and assigns it to @res, passing through LSM. 1330 * The new socket initialization is not complete, see kernel_accept(). 1331 * Returns 0 or an error. On failure @res is set to %NULL. 1332 * This function internally uses GFP_KERNEL. 1333 */ 1334 1335 int sock_create_lite(int family, int type, int protocol, struct socket **res) 1336 { 1337 int err; 1338 struct socket *sock = NULL; 1339 1340 err = security_socket_create(family, type, protocol, 1); 1341 if (err) 1342 goto out; 1343 1344 sock = sock_alloc(); 1345 if (!sock) { 1346 err = -ENOMEM; 1347 goto out; 1348 } 1349 1350 sock->type = type; 1351 err = security_socket_post_create(sock, family, type, protocol, 1); 1352 if (err) 1353 goto out_release; 1354 1355 out: 1356 *res = sock; 1357 return err; 1358 out_release: 1359 sock_release(sock); 1360 sock = NULL; 1361 goto out; 1362 } 1363 EXPORT_SYMBOL(sock_create_lite); 1364 1365 /* No kernel lock held - perfect */ 1366 static __poll_t sock_poll(struct file *file, poll_table *wait) 1367 { 1368 struct socket *sock = file->private_data; 1369 __poll_t events = poll_requested_events(wait), flag = 0; 1370 1371 if (!sock->ops->poll) 1372 return 0; 1373 1374 if (sk_can_busy_loop(sock->sk)) { 1375 /* poll once if requested by the syscall */ 1376 if (events & POLL_BUSY_LOOP) 1377 sk_busy_loop(sock->sk, 1); 1378 1379 /* if this socket can poll_ll, tell the system call */ 1380 flag = POLL_BUSY_LOOP; 1381 } 1382 1383 return sock->ops->poll(file, sock, wait) | flag; 1384 } 1385 1386 static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1387 { 1388 struct socket *sock = file->private_data; 1389 1390 return sock->ops->mmap(file, sock, vma); 1391 } 1392 1393 static int sock_close(struct inode *inode, struct file *filp) 1394 { 1395 __sock_release(SOCKET_I(inode), inode); 1396 return 0; 1397 } 1398 1399 /* 1400 * Update the socket async list 1401 * 1402 * Fasync_list locking strategy. 1403 * 1404 * 1. fasync_list is modified only under process context socket lock 1405 * i.e. under semaphore. 1406 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1407 * or under socket lock 1408 */ 1409 1410 static int sock_fasync(int fd, struct file *filp, int on) 1411 { 1412 struct socket *sock = filp->private_data; 1413 struct sock *sk = sock->sk; 1414 struct socket_wq *wq = &sock->wq; 1415 1416 if (sk == NULL) 1417 return -EINVAL; 1418 1419 lock_sock(sk); 1420 fasync_helper(fd, filp, on, &wq->fasync_list); 1421 1422 if (!wq->fasync_list) 1423 sock_reset_flag(sk, SOCK_FASYNC); 1424 else 1425 sock_set_flag(sk, SOCK_FASYNC); 1426 1427 release_sock(sk); 1428 return 0; 1429 } 1430 1431 /* This function may be called only under rcu_lock */ 1432 1433 int sock_wake_async(struct socket_wq *wq, int how, int band) 1434 { 1435 if (!wq || !wq->fasync_list) 1436 return -1; 1437 1438 switch (how) { 1439 case SOCK_WAKE_WAITD: 1440 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) 1441 break; 1442 goto call_kill; 1443 case SOCK_WAKE_SPACE: 1444 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) 1445 break; 1446 fallthrough; 1447 case SOCK_WAKE_IO: 1448 call_kill: 1449 kill_fasync(&wq->fasync_list, SIGIO, band); 1450 break; 1451 case SOCK_WAKE_URG: 1452 kill_fasync(&wq->fasync_list, SIGURG, band); 1453 } 1454 1455 return 0; 1456 } 1457 EXPORT_SYMBOL(sock_wake_async); 1458 1459 /** 1460 * __sock_create - creates a socket 1461 * @net: net namespace 1462 * @family: protocol family (AF_INET, ...) 1463 * @type: communication type (SOCK_STREAM, ...) 1464 * @protocol: protocol (0, ...) 1465 * @res: new socket 1466 * @kern: boolean for kernel space sockets 1467 * 1468 * Creates a new socket and assigns it to @res, passing through LSM. 1469 * Returns 0 or an error. On failure @res is set to %NULL. @kern must 1470 * be set to true if the socket resides in kernel space. 1471 * This function internally uses GFP_KERNEL. 1472 */ 1473 1474 int __sock_create(struct net *net, int family, int type, int protocol, 1475 struct socket **res, int kern) 1476 { 1477 int err; 1478 struct socket *sock; 1479 const struct net_proto_family *pf; 1480 1481 /* 1482 * Check protocol is in range 1483 */ 1484 if (family < 0 || family >= NPROTO) 1485 return -EAFNOSUPPORT; 1486 if (type < 0 || type >= SOCK_MAX) 1487 return -EINVAL; 1488 1489 /* Compatibility. 1490 1491 This uglymoron is moved from INET layer to here to avoid 1492 deadlock in module load. 1493 */ 1494 if (family == PF_INET && type == SOCK_PACKET) { 1495 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1496 current->comm); 1497 family = PF_PACKET; 1498 } 1499 1500 err = security_socket_create(family, type, protocol, kern); 1501 if (err) 1502 return err; 1503 1504 /* 1505 * Allocate the socket and allow the family to set things up. if 1506 * the protocol is 0, the family is instructed to select an appropriate 1507 * default. 1508 */ 1509 sock = sock_alloc(); 1510 if (!sock) { 1511 net_warn_ratelimited("socket: no more sockets\n"); 1512 return -ENFILE; /* Not exactly a match, but its the 1513 closest posix thing */ 1514 } 1515 1516 sock->type = type; 1517 1518 #ifdef CONFIG_MODULES 1519 /* Attempt to load a protocol module if the find failed. 1520 * 1521 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1522 * requested real, full-featured networking support upon configuration. 1523 * Otherwise module support will break! 1524 */ 1525 if (rcu_access_pointer(net_families[family]) == NULL) 1526 request_module("net-pf-%d", family); 1527 #endif 1528 1529 rcu_read_lock(); 1530 pf = rcu_dereference(net_families[family]); 1531 err = -EAFNOSUPPORT; 1532 if (!pf) 1533 goto out_release; 1534 1535 /* 1536 * We will call the ->create function, that possibly is in a loadable 1537 * module, so we have to bump that loadable module refcnt first. 1538 */ 1539 if (!try_module_get(pf->owner)) 1540 goto out_release; 1541 1542 /* Now protected by module ref count */ 1543 rcu_read_unlock(); 1544 1545 err = pf->create(net, sock, protocol, kern); 1546 if (err < 0) 1547 goto out_module_put; 1548 1549 /* 1550 * Now to bump the refcnt of the [loadable] module that owns this 1551 * socket at sock_release time we decrement its refcnt. 1552 */ 1553 if (!try_module_get(sock->ops->owner)) 1554 goto out_module_busy; 1555 1556 /* 1557 * Now that we're done with the ->create function, the [loadable] 1558 * module can have its refcnt decremented 1559 */ 1560 module_put(pf->owner); 1561 err = security_socket_post_create(sock, family, type, protocol, kern); 1562 if (err) 1563 goto out_sock_release; 1564 *res = sock; 1565 1566 return 0; 1567 1568 out_module_busy: 1569 err = -EAFNOSUPPORT; 1570 out_module_put: 1571 sock->ops = NULL; 1572 module_put(pf->owner); 1573 out_sock_release: 1574 sock_release(sock); 1575 return err; 1576 1577 out_release: 1578 rcu_read_unlock(); 1579 goto out_sock_release; 1580 } 1581 EXPORT_SYMBOL(__sock_create); 1582 1583 /** 1584 * sock_create - creates a socket 1585 * @family: protocol family (AF_INET, ...) 1586 * @type: communication type (SOCK_STREAM, ...) 1587 * @protocol: protocol (0, ...) 1588 * @res: new socket 1589 * 1590 * A wrapper around __sock_create(). 1591 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1592 */ 1593 1594 int sock_create(int family, int type, int protocol, struct socket **res) 1595 { 1596 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1597 } 1598 EXPORT_SYMBOL(sock_create); 1599 1600 /** 1601 * sock_create_kern - creates a socket (kernel space) 1602 * @net: net namespace 1603 * @family: protocol family (AF_INET, ...) 1604 * @type: communication type (SOCK_STREAM, ...) 1605 * @protocol: protocol (0, ...) 1606 * @res: new socket 1607 * 1608 * A wrapper around __sock_create(). 1609 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1610 */ 1611 1612 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) 1613 { 1614 return __sock_create(net, family, type, protocol, res, 1); 1615 } 1616 EXPORT_SYMBOL(sock_create_kern); 1617 1618 static struct socket *__sys_socket_create(int family, int type, int protocol) 1619 { 1620 struct socket *sock; 1621 int retval; 1622 1623 /* Check the SOCK_* constants for consistency. */ 1624 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1625 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1626 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1627 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1628 1629 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1630 return ERR_PTR(-EINVAL); 1631 type &= SOCK_TYPE_MASK; 1632 1633 retval = sock_create(family, type, protocol, &sock); 1634 if (retval < 0) 1635 return ERR_PTR(retval); 1636 1637 return sock; 1638 } 1639 1640 struct file *__sys_socket_file(int family, int type, int protocol) 1641 { 1642 struct socket *sock; 1643 int flags; 1644 1645 sock = __sys_socket_create(family, type, protocol); 1646 if (IS_ERR(sock)) 1647 return ERR_CAST(sock); 1648 1649 flags = type & ~SOCK_TYPE_MASK; 1650 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1651 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1652 1653 return sock_alloc_file(sock, flags, NULL); 1654 } 1655 1656 int __sys_socket(int family, int type, int protocol) 1657 { 1658 struct socket *sock; 1659 int flags; 1660 1661 sock = __sys_socket_create(family, type, protocol); 1662 if (IS_ERR(sock)) 1663 return PTR_ERR(sock); 1664 1665 flags = type & ~SOCK_TYPE_MASK; 1666 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1667 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1668 1669 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1670 } 1671 1672 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1673 { 1674 return __sys_socket(family, type, protocol); 1675 } 1676 1677 /* 1678 * Create a pair of connected sockets. 1679 */ 1680 1681 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) 1682 { 1683 struct socket *sock1, *sock2; 1684 int fd1, fd2, err; 1685 struct file *newfile1, *newfile2; 1686 int flags; 1687 1688 flags = type & ~SOCK_TYPE_MASK; 1689 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1690 return -EINVAL; 1691 type &= SOCK_TYPE_MASK; 1692 1693 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1694 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1695 1696 /* 1697 * reserve descriptors and make sure we won't fail 1698 * to return them to userland. 1699 */ 1700 fd1 = get_unused_fd_flags(flags); 1701 if (unlikely(fd1 < 0)) 1702 return fd1; 1703 1704 fd2 = get_unused_fd_flags(flags); 1705 if (unlikely(fd2 < 0)) { 1706 put_unused_fd(fd1); 1707 return fd2; 1708 } 1709 1710 err = put_user(fd1, &usockvec[0]); 1711 if (err) 1712 goto out; 1713 1714 err = put_user(fd2, &usockvec[1]); 1715 if (err) 1716 goto out; 1717 1718 /* 1719 * Obtain the first socket and check if the underlying protocol 1720 * supports the socketpair call. 1721 */ 1722 1723 err = sock_create(family, type, protocol, &sock1); 1724 if (unlikely(err < 0)) 1725 goto out; 1726 1727 err = sock_create(family, type, protocol, &sock2); 1728 if (unlikely(err < 0)) { 1729 sock_release(sock1); 1730 goto out; 1731 } 1732 1733 err = security_socket_socketpair(sock1, sock2); 1734 if (unlikely(err)) { 1735 sock_release(sock2); 1736 sock_release(sock1); 1737 goto out; 1738 } 1739 1740 err = sock1->ops->socketpair(sock1, sock2); 1741 if (unlikely(err < 0)) { 1742 sock_release(sock2); 1743 sock_release(sock1); 1744 goto out; 1745 } 1746 1747 newfile1 = sock_alloc_file(sock1, flags, NULL); 1748 if (IS_ERR(newfile1)) { 1749 err = PTR_ERR(newfile1); 1750 sock_release(sock2); 1751 goto out; 1752 } 1753 1754 newfile2 = sock_alloc_file(sock2, flags, NULL); 1755 if (IS_ERR(newfile2)) { 1756 err = PTR_ERR(newfile2); 1757 fput(newfile1); 1758 goto out; 1759 } 1760 1761 audit_fd_pair(fd1, fd2); 1762 1763 fd_install(fd1, newfile1); 1764 fd_install(fd2, newfile2); 1765 return 0; 1766 1767 out: 1768 put_unused_fd(fd2); 1769 put_unused_fd(fd1); 1770 return err; 1771 } 1772 1773 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1774 int __user *, usockvec) 1775 { 1776 return __sys_socketpair(family, type, protocol, usockvec); 1777 } 1778 1779 /* 1780 * Bind a name to a socket. Nothing much to do here since it's 1781 * the protocol's responsibility to handle the local address. 1782 * 1783 * We move the socket address to kernel space before we call 1784 * the protocol layer (having also checked the address is ok). 1785 */ 1786 1787 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) 1788 { 1789 struct socket *sock; 1790 struct sockaddr_storage address; 1791 int err, fput_needed; 1792 1793 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1794 if (sock) { 1795 err = move_addr_to_kernel(umyaddr, addrlen, &address); 1796 if (!err) { 1797 err = security_socket_bind(sock, 1798 (struct sockaddr *)&address, 1799 addrlen); 1800 if (!err) 1801 err = sock->ops->bind(sock, 1802 (struct sockaddr *) 1803 &address, addrlen); 1804 } 1805 fput_light(sock->file, fput_needed); 1806 } 1807 return err; 1808 } 1809 1810 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1811 { 1812 return __sys_bind(fd, umyaddr, addrlen); 1813 } 1814 1815 /* 1816 * Perform a listen. Basically, we allow the protocol to do anything 1817 * necessary for a listen, and if that works, we mark the socket as 1818 * ready for listening. 1819 */ 1820 1821 int __sys_listen(int fd, int backlog) 1822 { 1823 struct socket *sock; 1824 int err, fput_needed; 1825 int somaxconn; 1826 1827 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1828 if (sock) { 1829 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn); 1830 if ((unsigned int)backlog > somaxconn) 1831 backlog = somaxconn; 1832 1833 err = security_socket_listen(sock, backlog); 1834 if (!err) 1835 err = sock->ops->listen(sock, backlog); 1836 1837 fput_light(sock->file, fput_needed); 1838 } 1839 return err; 1840 } 1841 1842 SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1843 { 1844 return __sys_listen(fd, backlog); 1845 } 1846 1847 struct file *do_accept(struct file *file, unsigned file_flags, 1848 struct sockaddr __user *upeer_sockaddr, 1849 int __user *upeer_addrlen, int flags) 1850 { 1851 struct socket *sock, *newsock; 1852 struct file *newfile; 1853 int err, len; 1854 struct sockaddr_storage address; 1855 1856 sock = sock_from_file(file); 1857 if (!sock) 1858 return ERR_PTR(-ENOTSOCK); 1859 1860 newsock = sock_alloc(); 1861 if (!newsock) 1862 return ERR_PTR(-ENFILE); 1863 1864 newsock->type = sock->type; 1865 newsock->ops = sock->ops; 1866 1867 /* 1868 * We don't need try_module_get here, as the listening socket (sock) 1869 * has the protocol module (sock->ops->owner) held. 1870 */ 1871 __module_get(newsock->ops->owner); 1872 1873 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); 1874 if (IS_ERR(newfile)) 1875 return newfile; 1876 1877 err = security_socket_accept(sock, newsock); 1878 if (err) 1879 goto out_fd; 1880 1881 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags, 1882 false); 1883 if (err < 0) 1884 goto out_fd; 1885 1886 if (upeer_sockaddr) { 1887 len = newsock->ops->getname(newsock, 1888 (struct sockaddr *)&address, 2); 1889 if (len < 0) { 1890 err = -ECONNABORTED; 1891 goto out_fd; 1892 } 1893 err = move_addr_to_user(&address, 1894 len, upeer_sockaddr, upeer_addrlen); 1895 if (err < 0) 1896 goto out_fd; 1897 } 1898 1899 /* File flags are not inherited via accept() unlike another OSes. */ 1900 return newfile; 1901 out_fd: 1902 fput(newfile); 1903 return ERR_PTR(err); 1904 } 1905 1906 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr, 1907 int __user *upeer_addrlen, int flags) 1908 { 1909 struct file *newfile; 1910 int newfd; 1911 1912 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1913 return -EINVAL; 1914 1915 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1916 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1917 1918 newfd = get_unused_fd_flags(flags); 1919 if (unlikely(newfd < 0)) 1920 return newfd; 1921 1922 newfile = do_accept(file, 0, upeer_sockaddr, upeer_addrlen, 1923 flags); 1924 if (IS_ERR(newfile)) { 1925 put_unused_fd(newfd); 1926 return PTR_ERR(newfile); 1927 } 1928 fd_install(newfd, newfile); 1929 return newfd; 1930 } 1931 1932 /* 1933 * For accept, we attempt to create a new socket, set up the link 1934 * with the client, wake up the client, then return the new 1935 * connected fd. We collect the address of the connector in kernel 1936 * space and move it to user at the very end. This is unclean because 1937 * we open the socket then return an error. 1938 * 1939 * 1003.1g adds the ability to recvmsg() to query connection pending 1940 * status to recvmsg. We need to add that support in a way thats 1941 * clean when we restructure accept also. 1942 */ 1943 1944 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, 1945 int __user *upeer_addrlen, int flags) 1946 { 1947 int ret = -EBADF; 1948 struct fd f; 1949 1950 f = fdget(fd); 1951 if (f.file) { 1952 ret = __sys_accept4_file(f.file, upeer_sockaddr, 1953 upeer_addrlen, flags); 1954 fdput(f); 1955 } 1956 1957 return ret; 1958 } 1959 1960 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 1961 int __user *, upeer_addrlen, int, flags) 1962 { 1963 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); 1964 } 1965 1966 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 1967 int __user *, upeer_addrlen) 1968 { 1969 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 1970 } 1971 1972 /* 1973 * Attempt to connect to a socket with the server address. The address 1974 * is in user space so we verify it is OK and move it to kernel space. 1975 * 1976 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 1977 * break bindings 1978 * 1979 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 1980 * other SEQPACKET protocols that take time to connect() as it doesn't 1981 * include the -EINPROGRESS status for such sockets. 1982 */ 1983 1984 int __sys_connect_file(struct file *file, struct sockaddr_storage *address, 1985 int addrlen, int file_flags) 1986 { 1987 struct socket *sock; 1988 int err; 1989 1990 sock = sock_from_file(file); 1991 if (!sock) { 1992 err = -ENOTSOCK; 1993 goto out; 1994 } 1995 1996 err = 1997 security_socket_connect(sock, (struct sockaddr *)address, addrlen); 1998 if (err) 1999 goto out; 2000 2001 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen, 2002 sock->file->f_flags | file_flags); 2003 out: 2004 return err; 2005 } 2006 2007 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) 2008 { 2009 int ret = -EBADF; 2010 struct fd f; 2011 2012 f = fdget(fd); 2013 if (f.file) { 2014 struct sockaddr_storage address; 2015 2016 ret = move_addr_to_kernel(uservaddr, addrlen, &address); 2017 if (!ret) 2018 ret = __sys_connect_file(f.file, &address, addrlen, 0); 2019 fdput(f); 2020 } 2021 2022 return ret; 2023 } 2024 2025 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 2026 int, addrlen) 2027 { 2028 return __sys_connect(fd, uservaddr, addrlen); 2029 } 2030 2031 /* 2032 * Get the local address ('name') of a socket object. Move the obtained 2033 * name to user space. 2034 */ 2035 2036 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, 2037 int __user *usockaddr_len) 2038 { 2039 struct socket *sock; 2040 struct sockaddr_storage address; 2041 int err, fput_needed; 2042 2043 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2044 if (!sock) 2045 goto out; 2046 2047 err = security_socket_getsockname(sock); 2048 if (err) 2049 goto out_put; 2050 2051 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0); 2052 if (err < 0) 2053 goto out_put; 2054 /* "err" is actually length in this case */ 2055 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); 2056 2057 out_put: 2058 fput_light(sock->file, fput_needed); 2059 out: 2060 return err; 2061 } 2062 2063 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 2064 int __user *, usockaddr_len) 2065 { 2066 return __sys_getsockname(fd, usockaddr, usockaddr_len); 2067 } 2068 2069 /* 2070 * Get the remote address ('name') of a socket object. Move the obtained 2071 * name to user space. 2072 */ 2073 2074 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, 2075 int __user *usockaddr_len) 2076 { 2077 struct socket *sock; 2078 struct sockaddr_storage address; 2079 int err, fput_needed; 2080 2081 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2082 if (sock != NULL) { 2083 err = security_socket_getpeername(sock); 2084 if (err) { 2085 fput_light(sock->file, fput_needed); 2086 return err; 2087 } 2088 2089 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1); 2090 if (err >= 0) 2091 /* "err" is actually length in this case */ 2092 err = move_addr_to_user(&address, err, usockaddr, 2093 usockaddr_len); 2094 fput_light(sock->file, fput_needed); 2095 } 2096 return err; 2097 } 2098 2099 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 2100 int __user *, usockaddr_len) 2101 { 2102 return __sys_getpeername(fd, usockaddr, usockaddr_len); 2103 } 2104 2105 /* 2106 * Send a datagram to a given address. We move the address into kernel 2107 * space and check the user space data area is readable before invoking 2108 * the protocol. 2109 */ 2110 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, 2111 struct sockaddr __user *addr, int addr_len) 2112 { 2113 struct socket *sock; 2114 struct sockaddr_storage address; 2115 int err; 2116 struct msghdr msg; 2117 struct iovec iov; 2118 int fput_needed; 2119 2120 err = import_single_range(ITER_SOURCE, buff, len, &iov, &msg.msg_iter); 2121 if (unlikely(err)) 2122 return err; 2123 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2124 if (!sock) 2125 goto out; 2126 2127 msg.msg_name = NULL; 2128 msg.msg_control = NULL; 2129 msg.msg_controllen = 0; 2130 msg.msg_namelen = 0; 2131 msg.msg_ubuf = NULL; 2132 if (addr) { 2133 err = move_addr_to_kernel(addr, addr_len, &address); 2134 if (err < 0) 2135 goto out_put; 2136 msg.msg_name = (struct sockaddr *)&address; 2137 msg.msg_namelen = addr_len; 2138 } 2139 if (sock->file->f_flags & O_NONBLOCK) 2140 flags |= MSG_DONTWAIT; 2141 msg.msg_flags = flags; 2142 err = sock_sendmsg(sock, &msg); 2143 2144 out_put: 2145 fput_light(sock->file, fput_needed); 2146 out: 2147 return err; 2148 } 2149 2150 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 2151 unsigned int, flags, struct sockaddr __user *, addr, 2152 int, addr_len) 2153 { 2154 return __sys_sendto(fd, buff, len, flags, addr, addr_len); 2155 } 2156 2157 /* 2158 * Send a datagram down a socket. 2159 */ 2160 2161 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 2162 unsigned int, flags) 2163 { 2164 return __sys_sendto(fd, buff, len, flags, NULL, 0); 2165 } 2166 2167 /* 2168 * Receive a frame from the socket and optionally record the address of the 2169 * sender. We verify the buffers are writable and if needed move the 2170 * sender address from kernel to user space. 2171 */ 2172 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, 2173 struct sockaddr __user *addr, int __user *addr_len) 2174 { 2175 struct sockaddr_storage address; 2176 struct msghdr msg = { 2177 /* Save some cycles and don't copy the address if not needed */ 2178 .msg_name = addr ? (struct sockaddr *)&address : NULL, 2179 }; 2180 struct socket *sock; 2181 struct iovec iov; 2182 int err, err2; 2183 int fput_needed; 2184 2185 err = import_single_range(ITER_DEST, ubuf, size, &iov, &msg.msg_iter); 2186 if (unlikely(err)) 2187 return err; 2188 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2189 if (!sock) 2190 goto out; 2191 2192 if (sock->file->f_flags & O_NONBLOCK) 2193 flags |= MSG_DONTWAIT; 2194 err = sock_recvmsg(sock, &msg, flags); 2195 2196 if (err >= 0 && addr != NULL) { 2197 err2 = move_addr_to_user(&address, 2198 msg.msg_namelen, addr, addr_len); 2199 if (err2 < 0) 2200 err = err2; 2201 } 2202 2203 fput_light(sock->file, fput_needed); 2204 out: 2205 return err; 2206 } 2207 2208 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 2209 unsigned int, flags, struct sockaddr __user *, addr, 2210 int __user *, addr_len) 2211 { 2212 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); 2213 } 2214 2215 /* 2216 * Receive a datagram from a socket. 2217 */ 2218 2219 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, 2220 unsigned int, flags) 2221 { 2222 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 2223 } 2224 2225 static bool sock_use_custom_sol_socket(const struct socket *sock) 2226 { 2227 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags); 2228 } 2229 2230 /* 2231 * Set a socket option. Because we don't know the option lengths we have 2232 * to pass the user mode parameter for the protocols to sort out. 2233 */ 2234 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval, 2235 int optlen) 2236 { 2237 sockptr_t optval = USER_SOCKPTR(user_optval); 2238 char *kernel_optval = NULL; 2239 int err, fput_needed; 2240 struct socket *sock; 2241 2242 if (optlen < 0) 2243 return -EINVAL; 2244 2245 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2246 if (!sock) 2247 return err; 2248 2249 err = security_socket_setsockopt(sock, level, optname); 2250 if (err) 2251 goto out_put; 2252 2253 if (!in_compat_syscall()) 2254 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname, 2255 user_optval, &optlen, 2256 &kernel_optval); 2257 if (err < 0) 2258 goto out_put; 2259 if (err > 0) { 2260 err = 0; 2261 goto out_put; 2262 } 2263 2264 if (kernel_optval) 2265 optval = KERNEL_SOCKPTR(kernel_optval); 2266 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock)) 2267 err = sock_setsockopt(sock, level, optname, optval, optlen); 2268 else if (unlikely(!sock->ops->setsockopt)) 2269 err = -EOPNOTSUPP; 2270 else 2271 err = sock->ops->setsockopt(sock, level, optname, optval, 2272 optlen); 2273 kfree(kernel_optval); 2274 out_put: 2275 fput_light(sock->file, fput_needed); 2276 return err; 2277 } 2278 2279 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 2280 char __user *, optval, int, optlen) 2281 { 2282 return __sys_setsockopt(fd, level, optname, optval, optlen); 2283 } 2284 2285 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level, 2286 int optname)); 2287 2288 /* 2289 * Get a socket option. Because we don't know the option lengths we have 2290 * to pass a user mode parameter for the protocols to sort out. 2291 */ 2292 int __sys_getsockopt(int fd, int level, int optname, char __user *optval, 2293 int __user *optlen) 2294 { 2295 int err, fput_needed; 2296 struct socket *sock; 2297 int max_optlen; 2298 2299 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2300 if (!sock) 2301 return err; 2302 2303 err = security_socket_getsockopt(sock, level, optname); 2304 if (err) 2305 goto out_put; 2306 2307 if (!in_compat_syscall()) 2308 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen); 2309 2310 if (level == SOL_SOCKET) 2311 err = sock_getsockopt(sock, level, optname, optval, optlen); 2312 else if (unlikely(!sock->ops->getsockopt)) 2313 err = -EOPNOTSUPP; 2314 else 2315 err = sock->ops->getsockopt(sock, level, optname, optval, 2316 optlen); 2317 2318 if (!in_compat_syscall()) 2319 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname, 2320 optval, optlen, max_optlen, 2321 err); 2322 out_put: 2323 fput_light(sock->file, fput_needed); 2324 return err; 2325 } 2326 2327 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 2328 char __user *, optval, int __user *, optlen) 2329 { 2330 return __sys_getsockopt(fd, level, optname, optval, optlen); 2331 } 2332 2333 /* 2334 * Shutdown a socket. 2335 */ 2336 2337 int __sys_shutdown_sock(struct socket *sock, int how) 2338 { 2339 int err; 2340 2341 err = security_socket_shutdown(sock, how); 2342 if (!err) 2343 err = sock->ops->shutdown(sock, how); 2344 2345 return err; 2346 } 2347 2348 int __sys_shutdown(int fd, int how) 2349 { 2350 int err, fput_needed; 2351 struct socket *sock; 2352 2353 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2354 if (sock != NULL) { 2355 err = __sys_shutdown_sock(sock, how); 2356 fput_light(sock->file, fput_needed); 2357 } 2358 return err; 2359 } 2360 2361 SYSCALL_DEFINE2(shutdown, int, fd, int, how) 2362 { 2363 return __sys_shutdown(fd, how); 2364 } 2365 2366 /* A couple of helpful macros for getting the address of the 32/64 bit 2367 * fields which are the same type (int / unsigned) on our platforms. 2368 */ 2369 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 2370 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 2371 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 2372 2373 struct used_address { 2374 struct sockaddr_storage name; 2375 unsigned int name_len; 2376 }; 2377 2378 int __copy_msghdr(struct msghdr *kmsg, 2379 struct user_msghdr *msg, 2380 struct sockaddr __user **save_addr) 2381 { 2382 ssize_t err; 2383 2384 kmsg->msg_control_is_user = true; 2385 kmsg->msg_get_inq = 0; 2386 kmsg->msg_control_user = msg->msg_control; 2387 kmsg->msg_controllen = msg->msg_controllen; 2388 kmsg->msg_flags = msg->msg_flags; 2389 2390 kmsg->msg_namelen = msg->msg_namelen; 2391 if (!msg->msg_name) 2392 kmsg->msg_namelen = 0; 2393 2394 if (kmsg->msg_namelen < 0) 2395 return -EINVAL; 2396 2397 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) 2398 kmsg->msg_namelen = sizeof(struct sockaddr_storage); 2399 2400 if (save_addr) 2401 *save_addr = msg->msg_name; 2402 2403 if (msg->msg_name && kmsg->msg_namelen) { 2404 if (!save_addr) { 2405 err = move_addr_to_kernel(msg->msg_name, 2406 kmsg->msg_namelen, 2407 kmsg->msg_name); 2408 if (err < 0) 2409 return err; 2410 } 2411 } else { 2412 kmsg->msg_name = NULL; 2413 kmsg->msg_namelen = 0; 2414 } 2415 2416 if (msg->msg_iovlen > UIO_MAXIOV) 2417 return -EMSGSIZE; 2418 2419 kmsg->msg_iocb = NULL; 2420 kmsg->msg_ubuf = NULL; 2421 return 0; 2422 } 2423 2424 static int copy_msghdr_from_user(struct msghdr *kmsg, 2425 struct user_msghdr __user *umsg, 2426 struct sockaddr __user **save_addr, 2427 struct iovec **iov) 2428 { 2429 struct user_msghdr msg; 2430 ssize_t err; 2431 2432 if (copy_from_user(&msg, umsg, sizeof(*umsg))) 2433 return -EFAULT; 2434 2435 err = __copy_msghdr(kmsg, &msg, save_addr); 2436 if (err) 2437 return err; 2438 2439 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE, 2440 msg.msg_iov, msg.msg_iovlen, 2441 UIO_FASTIOV, iov, &kmsg->msg_iter); 2442 return err < 0 ? err : 0; 2443 } 2444 2445 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, 2446 unsigned int flags, struct used_address *used_address, 2447 unsigned int allowed_msghdr_flags) 2448 { 2449 unsigned char ctl[sizeof(struct cmsghdr) + 20] 2450 __aligned(sizeof(__kernel_size_t)); 2451 /* 20 is size of ipv6_pktinfo */ 2452 unsigned char *ctl_buf = ctl; 2453 int ctl_len; 2454 ssize_t err; 2455 2456 err = -ENOBUFS; 2457 2458 if (msg_sys->msg_controllen > INT_MAX) 2459 goto out; 2460 flags |= (msg_sys->msg_flags & allowed_msghdr_flags); 2461 ctl_len = msg_sys->msg_controllen; 2462 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 2463 err = 2464 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, 2465 sizeof(ctl)); 2466 if (err) 2467 goto out; 2468 ctl_buf = msg_sys->msg_control; 2469 ctl_len = msg_sys->msg_controllen; 2470 } else if (ctl_len) { 2471 BUILD_BUG_ON(sizeof(struct cmsghdr) != 2472 CMSG_ALIGN(sizeof(struct cmsghdr))); 2473 if (ctl_len > sizeof(ctl)) { 2474 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 2475 if (ctl_buf == NULL) 2476 goto out; 2477 } 2478 err = -EFAULT; 2479 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len)) 2480 goto out_freectl; 2481 msg_sys->msg_control = ctl_buf; 2482 msg_sys->msg_control_is_user = false; 2483 } 2484 msg_sys->msg_flags = flags; 2485 2486 if (sock->file->f_flags & O_NONBLOCK) 2487 msg_sys->msg_flags |= MSG_DONTWAIT; 2488 /* 2489 * If this is sendmmsg() and current destination address is same as 2490 * previously succeeded address, omit asking LSM's decision. 2491 * used_address->name_len is initialized to UINT_MAX so that the first 2492 * destination address never matches. 2493 */ 2494 if (used_address && msg_sys->msg_name && 2495 used_address->name_len == msg_sys->msg_namelen && 2496 !memcmp(&used_address->name, msg_sys->msg_name, 2497 used_address->name_len)) { 2498 err = sock_sendmsg_nosec(sock, msg_sys); 2499 goto out_freectl; 2500 } 2501 err = sock_sendmsg(sock, msg_sys); 2502 /* 2503 * If this is sendmmsg() and sending to current destination address was 2504 * successful, remember it. 2505 */ 2506 if (used_address && err >= 0) { 2507 used_address->name_len = msg_sys->msg_namelen; 2508 if (msg_sys->msg_name) 2509 memcpy(&used_address->name, msg_sys->msg_name, 2510 used_address->name_len); 2511 } 2512 2513 out_freectl: 2514 if (ctl_buf != ctl) 2515 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 2516 out: 2517 return err; 2518 } 2519 2520 int sendmsg_copy_msghdr(struct msghdr *msg, 2521 struct user_msghdr __user *umsg, unsigned flags, 2522 struct iovec **iov) 2523 { 2524 int err; 2525 2526 if (flags & MSG_CMSG_COMPAT) { 2527 struct compat_msghdr __user *msg_compat; 2528 2529 msg_compat = (struct compat_msghdr __user *) umsg; 2530 err = get_compat_msghdr(msg, msg_compat, NULL, iov); 2531 } else { 2532 err = copy_msghdr_from_user(msg, umsg, NULL, iov); 2533 } 2534 if (err < 0) 2535 return err; 2536 2537 return 0; 2538 } 2539 2540 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, 2541 struct msghdr *msg_sys, unsigned int flags, 2542 struct used_address *used_address, 2543 unsigned int allowed_msghdr_flags) 2544 { 2545 struct sockaddr_storage address; 2546 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2547 ssize_t err; 2548 2549 msg_sys->msg_name = &address; 2550 2551 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov); 2552 if (err < 0) 2553 return err; 2554 2555 err = ____sys_sendmsg(sock, msg_sys, flags, used_address, 2556 allowed_msghdr_flags); 2557 kfree(iov); 2558 return err; 2559 } 2560 2561 /* 2562 * BSD sendmsg interface 2563 */ 2564 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg, 2565 unsigned int flags) 2566 { 2567 return ____sys_sendmsg(sock, msg, flags, NULL, 0); 2568 } 2569 2570 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2571 bool forbid_cmsg_compat) 2572 { 2573 int fput_needed, err; 2574 struct msghdr msg_sys; 2575 struct socket *sock; 2576 2577 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2578 return -EINVAL; 2579 2580 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2581 if (!sock) 2582 goto out; 2583 2584 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); 2585 2586 fput_light(sock->file, fput_needed); 2587 out: 2588 return err; 2589 } 2590 2591 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) 2592 { 2593 return __sys_sendmsg(fd, msg, flags, true); 2594 } 2595 2596 /* 2597 * Linux sendmmsg interface 2598 */ 2599 2600 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2601 unsigned int flags, bool forbid_cmsg_compat) 2602 { 2603 int fput_needed, err, datagrams; 2604 struct socket *sock; 2605 struct mmsghdr __user *entry; 2606 struct compat_mmsghdr __user *compat_entry; 2607 struct msghdr msg_sys; 2608 struct used_address used_address; 2609 unsigned int oflags = flags; 2610 2611 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2612 return -EINVAL; 2613 2614 if (vlen > UIO_MAXIOV) 2615 vlen = UIO_MAXIOV; 2616 2617 datagrams = 0; 2618 2619 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2620 if (!sock) 2621 return err; 2622 2623 used_address.name_len = UINT_MAX; 2624 entry = mmsg; 2625 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2626 err = 0; 2627 flags |= MSG_BATCH; 2628 2629 while (datagrams < vlen) { 2630 if (datagrams == vlen - 1) 2631 flags = oflags; 2632 2633 if (MSG_CMSG_COMPAT & flags) { 2634 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, 2635 &msg_sys, flags, &used_address, MSG_EOR); 2636 if (err < 0) 2637 break; 2638 err = __put_user(err, &compat_entry->msg_len); 2639 ++compat_entry; 2640 } else { 2641 err = ___sys_sendmsg(sock, 2642 (struct user_msghdr __user *)entry, 2643 &msg_sys, flags, &used_address, MSG_EOR); 2644 if (err < 0) 2645 break; 2646 err = put_user(err, &entry->msg_len); 2647 ++entry; 2648 } 2649 2650 if (err) 2651 break; 2652 ++datagrams; 2653 if (msg_data_left(&msg_sys)) 2654 break; 2655 cond_resched(); 2656 } 2657 2658 fput_light(sock->file, fput_needed); 2659 2660 /* We only return an error if no datagrams were able to be sent */ 2661 if (datagrams != 0) 2662 return datagrams; 2663 2664 return err; 2665 } 2666 2667 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, 2668 unsigned int, vlen, unsigned int, flags) 2669 { 2670 return __sys_sendmmsg(fd, mmsg, vlen, flags, true); 2671 } 2672 2673 int recvmsg_copy_msghdr(struct msghdr *msg, 2674 struct user_msghdr __user *umsg, unsigned flags, 2675 struct sockaddr __user **uaddr, 2676 struct iovec **iov) 2677 { 2678 ssize_t err; 2679 2680 if (MSG_CMSG_COMPAT & flags) { 2681 struct compat_msghdr __user *msg_compat; 2682 2683 msg_compat = (struct compat_msghdr __user *) umsg; 2684 err = get_compat_msghdr(msg, msg_compat, uaddr, iov); 2685 } else { 2686 err = copy_msghdr_from_user(msg, umsg, uaddr, iov); 2687 } 2688 if (err < 0) 2689 return err; 2690 2691 return 0; 2692 } 2693 2694 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys, 2695 struct user_msghdr __user *msg, 2696 struct sockaddr __user *uaddr, 2697 unsigned int flags, int nosec) 2698 { 2699 struct compat_msghdr __user *msg_compat = 2700 (struct compat_msghdr __user *) msg; 2701 int __user *uaddr_len = COMPAT_NAMELEN(msg); 2702 struct sockaddr_storage addr; 2703 unsigned long cmsg_ptr; 2704 int len; 2705 ssize_t err; 2706 2707 msg_sys->msg_name = &addr; 2708 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2709 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2710 2711 /* We assume all kernel code knows the size of sockaddr_storage */ 2712 msg_sys->msg_namelen = 0; 2713 2714 if (sock->file->f_flags & O_NONBLOCK) 2715 flags |= MSG_DONTWAIT; 2716 2717 if (unlikely(nosec)) 2718 err = sock_recvmsg_nosec(sock, msg_sys, flags); 2719 else 2720 err = sock_recvmsg(sock, msg_sys, flags); 2721 2722 if (err < 0) 2723 goto out; 2724 len = err; 2725 2726 if (uaddr != NULL) { 2727 err = move_addr_to_user(&addr, 2728 msg_sys->msg_namelen, uaddr, 2729 uaddr_len); 2730 if (err < 0) 2731 goto out; 2732 } 2733 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2734 COMPAT_FLAGS(msg)); 2735 if (err) 2736 goto out; 2737 if (MSG_CMSG_COMPAT & flags) 2738 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2739 &msg_compat->msg_controllen); 2740 else 2741 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2742 &msg->msg_controllen); 2743 if (err) 2744 goto out; 2745 err = len; 2746 out: 2747 return err; 2748 } 2749 2750 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, 2751 struct msghdr *msg_sys, unsigned int flags, int nosec) 2752 { 2753 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2754 /* user mode address pointers */ 2755 struct sockaddr __user *uaddr; 2756 ssize_t err; 2757 2758 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov); 2759 if (err < 0) 2760 return err; 2761 2762 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec); 2763 kfree(iov); 2764 return err; 2765 } 2766 2767 /* 2768 * BSD recvmsg interface 2769 */ 2770 2771 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg, 2772 struct user_msghdr __user *umsg, 2773 struct sockaddr __user *uaddr, unsigned int flags) 2774 { 2775 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0); 2776 } 2777 2778 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2779 bool forbid_cmsg_compat) 2780 { 2781 int fput_needed, err; 2782 struct msghdr msg_sys; 2783 struct socket *sock; 2784 2785 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2786 return -EINVAL; 2787 2788 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2789 if (!sock) 2790 goto out; 2791 2792 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2793 2794 fput_light(sock->file, fput_needed); 2795 out: 2796 return err; 2797 } 2798 2799 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, 2800 unsigned int, flags) 2801 { 2802 return __sys_recvmsg(fd, msg, flags, true); 2803 } 2804 2805 /* 2806 * Linux recvmmsg interface 2807 */ 2808 2809 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2810 unsigned int vlen, unsigned int flags, 2811 struct timespec64 *timeout) 2812 { 2813 int fput_needed, err, datagrams; 2814 struct socket *sock; 2815 struct mmsghdr __user *entry; 2816 struct compat_mmsghdr __user *compat_entry; 2817 struct msghdr msg_sys; 2818 struct timespec64 end_time; 2819 struct timespec64 timeout64; 2820 2821 if (timeout && 2822 poll_select_set_timeout(&end_time, timeout->tv_sec, 2823 timeout->tv_nsec)) 2824 return -EINVAL; 2825 2826 datagrams = 0; 2827 2828 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2829 if (!sock) 2830 return err; 2831 2832 if (likely(!(flags & MSG_ERRQUEUE))) { 2833 err = sock_error(sock->sk); 2834 if (err) { 2835 datagrams = err; 2836 goto out_put; 2837 } 2838 } 2839 2840 entry = mmsg; 2841 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2842 2843 while (datagrams < vlen) { 2844 /* 2845 * No need to ask LSM for more than the first datagram. 2846 */ 2847 if (MSG_CMSG_COMPAT & flags) { 2848 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, 2849 &msg_sys, flags & ~MSG_WAITFORONE, 2850 datagrams); 2851 if (err < 0) 2852 break; 2853 err = __put_user(err, &compat_entry->msg_len); 2854 ++compat_entry; 2855 } else { 2856 err = ___sys_recvmsg(sock, 2857 (struct user_msghdr __user *)entry, 2858 &msg_sys, flags & ~MSG_WAITFORONE, 2859 datagrams); 2860 if (err < 0) 2861 break; 2862 err = put_user(err, &entry->msg_len); 2863 ++entry; 2864 } 2865 2866 if (err) 2867 break; 2868 ++datagrams; 2869 2870 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2871 if (flags & MSG_WAITFORONE) 2872 flags |= MSG_DONTWAIT; 2873 2874 if (timeout) { 2875 ktime_get_ts64(&timeout64); 2876 *timeout = timespec64_sub(end_time, timeout64); 2877 if (timeout->tv_sec < 0) { 2878 timeout->tv_sec = timeout->tv_nsec = 0; 2879 break; 2880 } 2881 2882 /* Timeout, return less than vlen datagrams */ 2883 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2884 break; 2885 } 2886 2887 /* Out of band data, return right away */ 2888 if (msg_sys.msg_flags & MSG_OOB) 2889 break; 2890 cond_resched(); 2891 } 2892 2893 if (err == 0) 2894 goto out_put; 2895 2896 if (datagrams == 0) { 2897 datagrams = err; 2898 goto out_put; 2899 } 2900 2901 /* 2902 * We may return less entries than requested (vlen) if the 2903 * sock is non block and there aren't enough datagrams... 2904 */ 2905 if (err != -EAGAIN) { 2906 /* 2907 * ... or if recvmsg returns an error after we 2908 * received some datagrams, where we record the 2909 * error to return on the next call or if the 2910 * app asks about it using getsockopt(SO_ERROR). 2911 */ 2912 sock->sk->sk_err = -err; 2913 } 2914 out_put: 2915 fput_light(sock->file, fput_needed); 2916 2917 return datagrams; 2918 } 2919 2920 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2921 unsigned int vlen, unsigned int flags, 2922 struct __kernel_timespec __user *timeout, 2923 struct old_timespec32 __user *timeout32) 2924 { 2925 int datagrams; 2926 struct timespec64 timeout_sys; 2927 2928 if (timeout && get_timespec64(&timeout_sys, timeout)) 2929 return -EFAULT; 2930 2931 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32)) 2932 return -EFAULT; 2933 2934 if (!timeout && !timeout32) 2935 return do_recvmmsg(fd, mmsg, vlen, flags, NULL); 2936 2937 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 2938 2939 if (datagrams <= 0) 2940 return datagrams; 2941 2942 if (timeout && put_timespec64(&timeout_sys, timeout)) 2943 datagrams = -EFAULT; 2944 2945 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32)) 2946 datagrams = -EFAULT; 2947 2948 return datagrams; 2949 } 2950 2951 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 2952 unsigned int, vlen, unsigned int, flags, 2953 struct __kernel_timespec __user *, timeout) 2954 { 2955 if (flags & MSG_CMSG_COMPAT) 2956 return -EINVAL; 2957 2958 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL); 2959 } 2960 2961 #ifdef CONFIG_COMPAT_32BIT_TIME 2962 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg, 2963 unsigned int, vlen, unsigned int, flags, 2964 struct old_timespec32 __user *, timeout) 2965 { 2966 if (flags & MSG_CMSG_COMPAT) 2967 return -EINVAL; 2968 2969 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout); 2970 } 2971 #endif 2972 2973 #ifdef __ARCH_WANT_SYS_SOCKETCALL 2974 /* Argument list sizes for sys_socketcall */ 2975 #define AL(x) ((x) * sizeof(unsigned long)) 2976 static const unsigned char nargs[21] = { 2977 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 2978 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 2979 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 2980 AL(4), AL(5), AL(4) 2981 }; 2982 2983 #undef AL 2984 2985 /* 2986 * System call vectors. 2987 * 2988 * Argument checking cleaned up. Saved 20% in size. 2989 * This function doesn't need to set the kernel lock because 2990 * it is set by the callees. 2991 */ 2992 2993 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 2994 { 2995 unsigned long a[AUDITSC_ARGS]; 2996 unsigned long a0, a1; 2997 int err; 2998 unsigned int len; 2999 3000 if (call < 1 || call > SYS_SENDMMSG) 3001 return -EINVAL; 3002 call = array_index_nospec(call, SYS_SENDMMSG + 1); 3003 3004 len = nargs[call]; 3005 if (len > sizeof(a)) 3006 return -EINVAL; 3007 3008 /* copy_from_user should be SMP safe. */ 3009 if (copy_from_user(a, args, len)) 3010 return -EFAULT; 3011 3012 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); 3013 if (err) 3014 return err; 3015 3016 a0 = a[0]; 3017 a1 = a[1]; 3018 3019 switch (call) { 3020 case SYS_SOCKET: 3021 err = __sys_socket(a0, a1, a[2]); 3022 break; 3023 case SYS_BIND: 3024 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 3025 break; 3026 case SYS_CONNECT: 3027 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 3028 break; 3029 case SYS_LISTEN: 3030 err = __sys_listen(a0, a1); 3031 break; 3032 case SYS_ACCEPT: 3033 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 3034 (int __user *)a[2], 0); 3035 break; 3036 case SYS_GETSOCKNAME: 3037 err = 3038 __sys_getsockname(a0, (struct sockaddr __user *)a1, 3039 (int __user *)a[2]); 3040 break; 3041 case SYS_GETPEERNAME: 3042 err = 3043 __sys_getpeername(a0, (struct sockaddr __user *)a1, 3044 (int __user *)a[2]); 3045 break; 3046 case SYS_SOCKETPAIR: 3047 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 3048 break; 3049 case SYS_SEND: 3050 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 3051 NULL, 0); 3052 break; 3053 case SYS_SENDTO: 3054 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 3055 (struct sockaddr __user *)a[4], a[5]); 3056 break; 3057 case SYS_RECV: 3058 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 3059 NULL, NULL); 3060 break; 3061 case SYS_RECVFROM: 3062 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 3063 (struct sockaddr __user *)a[4], 3064 (int __user *)a[5]); 3065 break; 3066 case SYS_SHUTDOWN: 3067 err = __sys_shutdown(a0, a1); 3068 break; 3069 case SYS_SETSOCKOPT: 3070 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], 3071 a[4]); 3072 break; 3073 case SYS_GETSOCKOPT: 3074 err = 3075 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 3076 (int __user *)a[4]); 3077 break; 3078 case SYS_SENDMSG: 3079 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, 3080 a[2], true); 3081 break; 3082 case SYS_SENDMMSG: 3083 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], 3084 a[3], true); 3085 break; 3086 case SYS_RECVMSG: 3087 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, 3088 a[2], true); 3089 break; 3090 case SYS_RECVMMSG: 3091 if (IS_ENABLED(CONFIG_64BIT)) 3092 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 3093 a[2], a[3], 3094 (struct __kernel_timespec __user *)a[4], 3095 NULL); 3096 else 3097 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 3098 a[2], a[3], NULL, 3099 (struct old_timespec32 __user *)a[4]); 3100 break; 3101 case SYS_ACCEPT4: 3102 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 3103 (int __user *)a[2], a[3]); 3104 break; 3105 default: 3106 err = -EINVAL; 3107 break; 3108 } 3109 return err; 3110 } 3111 3112 #endif /* __ARCH_WANT_SYS_SOCKETCALL */ 3113 3114 /** 3115 * sock_register - add a socket protocol handler 3116 * @ops: description of protocol 3117 * 3118 * This function is called by a protocol handler that wants to 3119 * advertise its address family, and have it linked into the 3120 * socket interface. The value ops->family corresponds to the 3121 * socket system call protocol family. 3122 */ 3123 int sock_register(const struct net_proto_family *ops) 3124 { 3125 int err; 3126 3127 if (ops->family >= NPROTO) { 3128 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); 3129 return -ENOBUFS; 3130 } 3131 3132 spin_lock(&net_family_lock); 3133 if (rcu_dereference_protected(net_families[ops->family], 3134 lockdep_is_held(&net_family_lock))) 3135 err = -EEXIST; 3136 else { 3137 rcu_assign_pointer(net_families[ops->family], ops); 3138 err = 0; 3139 } 3140 spin_unlock(&net_family_lock); 3141 3142 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]); 3143 return err; 3144 } 3145 EXPORT_SYMBOL(sock_register); 3146 3147 /** 3148 * sock_unregister - remove a protocol handler 3149 * @family: protocol family to remove 3150 * 3151 * This function is called by a protocol handler that wants to 3152 * remove its address family, and have it unlinked from the 3153 * new socket creation. 3154 * 3155 * If protocol handler is a module, then it can use module reference 3156 * counts to protect against new references. If protocol handler is not 3157 * a module then it needs to provide its own protection in 3158 * the ops->create routine. 3159 */ 3160 void sock_unregister(int family) 3161 { 3162 BUG_ON(family < 0 || family >= NPROTO); 3163 3164 spin_lock(&net_family_lock); 3165 RCU_INIT_POINTER(net_families[family], NULL); 3166 spin_unlock(&net_family_lock); 3167 3168 synchronize_rcu(); 3169 3170 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]); 3171 } 3172 EXPORT_SYMBOL(sock_unregister); 3173 3174 bool sock_is_registered(int family) 3175 { 3176 return family < NPROTO && rcu_access_pointer(net_families[family]); 3177 } 3178 3179 static int __init sock_init(void) 3180 { 3181 int err; 3182 /* 3183 * Initialize the network sysctl infrastructure. 3184 */ 3185 err = net_sysctl_init(); 3186 if (err) 3187 goto out; 3188 3189 /* 3190 * Initialize skbuff SLAB cache 3191 */ 3192 skb_init(); 3193 3194 /* 3195 * Initialize the protocols module. 3196 */ 3197 3198 init_inodecache(); 3199 3200 err = register_filesystem(&sock_fs_type); 3201 if (err) 3202 goto out; 3203 sock_mnt = kern_mount(&sock_fs_type); 3204 if (IS_ERR(sock_mnt)) { 3205 err = PTR_ERR(sock_mnt); 3206 goto out_mount; 3207 } 3208 3209 /* The real protocol initialization is performed in later initcalls. 3210 */ 3211 3212 #ifdef CONFIG_NETFILTER 3213 err = netfilter_init(); 3214 if (err) 3215 goto out; 3216 #endif 3217 3218 ptp_classifier_init(); 3219 3220 out: 3221 return err; 3222 3223 out_mount: 3224 unregister_filesystem(&sock_fs_type); 3225 goto out; 3226 } 3227 3228 core_initcall(sock_init); /* early initcall */ 3229 3230 #ifdef CONFIG_PROC_FS 3231 void socket_seq_show(struct seq_file *seq) 3232 { 3233 seq_printf(seq, "sockets: used %d\n", 3234 sock_inuse_get(seq->private)); 3235 } 3236 #endif /* CONFIG_PROC_FS */ 3237 3238 /* Handle the fact that while struct ifreq has the same *layout* on 3239 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data, 3240 * which are handled elsewhere, it still has different *size* due to 3241 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit, 3242 * resulting in struct ifreq being 32 and 40 bytes respectively). 3243 * As a result, if the struct happens to be at the end of a page and 3244 * the next page isn't readable/writable, we get a fault. To prevent 3245 * that, copy back and forth to the full size. 3246 */ 3247 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg) 3248 { 3249 if (in_compat_syscall()) { 3250 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr; 3251 3252 memset(ifr, 0, sizeof(*ifr)); 3253 if (copy_from_user(ifr32, arg, sizeof(*ifr32))) 3254 return -EFAULT; 3255 3256 if (ifrdata) 3257 *ifrdata = compat_ptr(ifr32->ifr_data); 3258 3259 return 0; 3260 } 3261 3262 if (copy_from_user(ifr, arg, sizeof(*ifr))) 3263 return -EFAULT; 3264 3265 if (ifrdata) 3266 *ifrdata = ifr->ifr_data; 3267 3268 return 0; 3269 } 3270 EXPORT_SYMBOL(get_user_ifreq); 3271 3272 int put_user_ifreq(struct ifreq *ifr, void __user *arg) 3273 { 3274 size_t size = sizeof(*ifr); 3275 3276 if (in_compat_syscall()) 3277 size = sizeof(struct compat_ifreq); 3278 3279 if (copy_to_user(arg, ifr, size)) 3280 return -EFAULT; 3281 3282 return 0; 3283 } 3284 EXPORT_SYMBOL(put_user_ifreq); 3285 3286 #ifdef CONFIG_COMPAT 3287 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 3288 { 3289 compat_uptr_t uptr32; 3290 struct ifreq ifr; 3291 void __user *saved; 3292 int err; 3293 3294 if (get_user_ifreq(&ifr, NULL, uifr32)) 3295 return -EFAULT; 3296 3297 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 3298 return -EFAULT; 3299 3300 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; 3301 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); 3302 3303 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL); 3304 if (!err) { 3305 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; 3306 if (put_user_ifreq(&ifr, uifr32)) 3307 err = -EFAULT; 3308 } 3309 return err; 3310 } 3311 3312 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ 3313 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, 3314 struct compat_ifreq __user *u_ifreq32) 3315 { 3316 struct ifreq ifreq; 3317 void __user *data; 3318 3319 if (!is_socket_ioctl_cmd(cmd)) 3320 return -ENOTTY; 3321 if (get_user_ifreq(&ifreq, &data, u_ifreq32)) 3322 return -EFAULT; 3323 ifreq.ifr_data = data; 3324 3325 return dev_ioctl(net, cmd, &ifreq, data, NULL); 3326 } 3327 3328 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 3329 unsigned int cmd, unsigned long arg) 3330 { 3331 void __user *argp = compat_ptr(arg); 3332 struct sock *sk = sock->sk; 3333 struct net *net = sock_net(sk); 3334 3335 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 3336 return sock_ioctl(file, cmd, (unsigned long)argp); 3337 3338 switch (cmd) { 3339 case SIOCWANDEV: 3340 return compat_siocwandev(net, argp); 3341 case SIOCGSTAMP_OLD: 3342 case SIOCGSTAMPNS_OLD: 3343 if (!sock->ops->gettstamp) 3344 return -ENOIOCTLCMD; 3345 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD, 3346 !COMPAT_USE_64BIT_TIME); 3347 3348 case SIOCETHTOOL: 3349 case SIOCBONDSLAVEINFOQUERY: 3350 case SIOCBONDINFOQUERY: 3351 case SIOCSHWTSTAMP: 3352 case SIOCGHWTSTAMP: 3353 return compat_ifr_data_ioctl(net, cmd, argp); 3354 3355 case FIOSETOWN: 3356 case SIOCSPGRP: 3357 case FIOGETOWN: 3358 case SIOCGPGRP: 3359 case SIOCBRADDBR: 3360 case SIOCBRDELBR: 3361 case SIOCGIFVLAN: 3362 case SIOCSIFVLAN: 3363 case SIOCGSKNS: 3364 case SIOCGSTAMP_NEW: 3365 case SIOCGSTAMPNS_NEW: 3366 case SIOCGIFCONF: 3367 case SIOCSIFBR: 3368 case SIOCGIFBR: 3369 return sock_ioctl(file, cmd, arg); 3370 3371 case SIOCGIFFLAGS: 3372 case SIOCSIFFLAGS: 3373 case SIOCGIFMAP: 3374 case SIOCSIFMAP: 3375 case SIOCGIFMETRIC: 3376 case SIOCSIFMETRIC: 3377 case SIOCGIFMTU: 3378 case SIOCSIFMTU: 3379 case SIOCGIFMEM: 3380 case SIOCSIFMEM: 3381 case SIOCGIFHWADDR: 3382 case SIOCSIFHWADDR: 3383 case SIOCADDMULTI: 3384 case SIOCDELMULTI: 3385 case SIOCGIFINDEX: 3386 case SIOCGIFADDR: 3387 case SIOCSIFADDR: 3388 case SIOCSIFHWBROADCAST: 3389 case SIOCDIFADDR: 3390 case SIOCGIFBRDADDR: 3391 case SIOCSIFBRDADDR: 3392 case SIOCGIFDSTADDR: 3393 case SIOCSIFDSTADDR: 3394 case SIOCGIFNETMASK: 3395 case SIOCSIFNETMASK: 3396 case SIOCSIFPFLAGS: 3397 case SIOCGIFPFLAGS: 3398 case SIOCGIFTXQLEN: 3399 case SIOCSIFTXQLEN: 3400 case SIOCBRADDIF: 3401 case SIOCBRDELIF: 3402 case SIOCGIFNAME: 3403 case SIOCSIFNAME: 3404 case SIOCGMIIPHY: 3405 case SIOCGMIIREG: 3406 case SIOCSMIIREG: 3407 case SIOCBONDENSLAVE: 3408 case SIOCBONDRELEASE: 3409 case SIOCBONDSETHWADDR: 3410 case SIOCBONDCHANGEACTIVE: 3411 case SIOCSARP: 3412 case SIOCGARP: 3413 case SIOCDARP: 3414 case SIOCOUTQ: 3415 case SIOCOUTQNSD: 3416 case SIOCATMARK: 3417 return sock_do_ioctl(net, sock, cmd, arg); 3418 } 3419 3420 return -ENOIOCTLCMD; 3421 } 3422 3423 static long compat_sock_ioctl(struct file *file, unsigned int cmd, 3424 unsigned long arg) 3425 { 3426 struct socket *sock = file->private_data; 3427 int ret = -ENOIOCTLCMD; 3428 struct sock *sk; 3429 struct net *net; 3430 3431 sk = sock->sk; 3432 net = sock_net(sk); 3433 3434 if (sock->ops->compat_ioctl) 3435 ret = sock->ops->compat_ioctl(sock, cmd, arg); 3436 3437 if (ret == -ENOIOCTLCMD && 3438 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 3439 ret = compat_wext_handle_ioctl(net, cmd, arg); 3440 3441 if (ret == -ENOIOCTLCMD) 3442 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 3443 3444 return ret; 3445 } 3446 #endif 3447 3448 /** 3449 * kernel_bind - bind an address to a socket (kernel space) 3450 * @sock: socket 3451 * @addr: address 3452 * @addrlen: length of address 3453 * 3454 * Returns 0 or an error. 3455 */ 3456 3457 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3458 { 3459 return sock->ops->bind(sock, addr, addrlen); 3460 } 3461 EXPORT_SYMBOL(kernel_bind); 3462 3463 /** 3464 * kernel_listen - move socket to listening state (kernel space) 3465 * @sock: socket 3466 * @backlog: pending connections queue size 3467 * 3468 * Returns 0 or an error. 3469 */ 3470 3471 int kernel_listen(struct socket *sock, int backlog) 3472 { 3473 return sock->ops->listen(sock, backlog); 3474 } 3475 EXPORT_SYMBOL(kernel_listen); 3476 3477 /** 3478 * kernel_accept - accept a connection (kernel space) 3479 * @sock: listening socket 3480 * @newsock: new connected socket 3481 * @flags: flags 3482 * 3483 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0. 3484 * If it fails, @newsock is guaranteed to be %NULL. 3485 * Returns 0 or an error. 3486 */ 3487 3488 int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3489 { 3490 struct sock *sk = sock->sk; 3491 int err; 3492 3493 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3494 newsock); 3495 if (err < 0) 3496 goto done; 3497 3498 err = sock->ops->accept(sock, *newsock, flags, true); 3499 if (err < 0) { 3500 sock_release(*newsock); 3501 *newsock = NULL; 3502 goto done; 3503 } 3504 3505 (*newsock)->ops = sock->ops; 3506 __module_get((*newsock)->ops->owner); 3507 3508 done: 3509 return err; 3510 } 3511 EXPORT_SYMBOL(kernel_accept); 3512 3513 /** 3514 * kernel_connect - connect a socket (kernel space) 3515 * @sock: socket 3516 * @addr: address 3517 * @addrlen: address length 3518 * @flags: flags (O_NONBLOCK, ...) 3519 * 3520 * For datagram sockets, @addr is the address to which datagrams are sent 3521 * by default, and the only address from which datagrams are received. 3522 * For stream sockets, attempts to connect to @addr. 3523 * Returns 0 or an error code. 3524 */ 3525 3526 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3527 int flags) 3528 { 3529 return sock->ops->connect(sock, addr, addrlen, flags); 3530 } 3531 EXPORT_SYMBOL(kernel_connect); 3532 3533 /** 3534 * kernel_getsockname - get the address which the socket is bound (kernel space) 3535 * @sock: socket 3536 * @addr: address holder 3537 * 3538 * Fills the @addr pointer with the address which the socket is bound. 3539 * Returns the length of the address in bytes or an error code. 3540 */ 3541 3542 int kernel_getsockname(struct socket *sock, struct sockaddr *addr) 3543 { 3544 return sock->ops->getname(sock, addr, 0); 3545 } 3546 EXPORT_SYMBOL(kernel_getsockname); 3547 3548 /** 3549 * kernel_getpeername - get the address which the socket is connected (kernel space) 3550 * @sock: socket 3551 * @addr: address holder 3552 * 3553 * Fills the @addr pointer with the address which the socket is connected. 3554 * Returns the length of the address in bytes or an error code. 3555 */ 3556 3557 int kernel_getpeername(struct socket *sock, struct sockaddr *addr) 3558 { 3559 return sock->ops->getname(sock, addr, 1); 3560 } 3561 EXPORT_SYMBOL(kernel_getpeername); 3562 3563 /** 3564 * kernel_sendpage - send a &page through a socket (kernel space) 3565 * @sock: socket 3566 * @page: page 3567 * @offset: page offset 3568 * @size: total size in bytes 3569 * @flags: flags (MSG_DONTWAIT, ...) 3570 * 3571 * Returns the total amount sent in bytes or an error. 3572 */ 3573 3574 int kernel_sendpage(struct socket *sock, struct page *page, int offset, 3575 size_t size, int flags) 3576 { 3577 if (sock->ops->sendpage) { 3578 /* Warn in case the improper page to zero-copy send */ 3579 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send"); 3580 return sock->ops->sendpage(sock, page, offset, size, flags); 3581 } 3582 return sock_no_sendpage(sock, page, offset, size, flags); 3583 } 3584 EXPORT_SYMBOL(kernel_sendpage); 3585 3586 /** 3587 * kernel_sendpage_locked - send a &page through the locked sock (kernel space) 3588 * @sk: sock 3589 * @page: page 3590 * @offset: page offset 3591 * @size: total size in bytes 3592 * @flags: flags (MSG_DONTWAIT, ...) 3593 * 3594 * Returns the total amount sent in bytes or an error. 3595 * Caller must hold @sk. 3596 */ 3597 3598 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset, 3599 size_t size, int flags) 3600 { 3601 struct socket *sock = sk->sk_socket; 3602 3603 if (sock->ops->sendpage_locked) 3604 return sock->ops->sendpage_locked(sk, page, offset, size, 3605 flags); 3606 3607 return sock_no_sendpage_locked(sk, page, offset, size, flags); 3608 } 3609 EXPORT_SYMBOL(kernel_sendpage_locked); 3610 3611 /** 3612 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space) 3613 * @sock: socket 3614 * @how: connection part 3615 * 3616 * Returns 0 or an error. 3617 */ 3618 3619 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3620 { 3621 return sock->ops->shutdown(sock, how); 3622 } 3623 EXPORT_SYMBOL(kernel_sock_shutdown); 3624 3625 /** 3626 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket 3627 * @sk: socket 3628 * 3629 * This routine returns the IP overhead imposed by a socket i.e. 3630 * the length of the underlying IP header, depending on whether 3631 * this is an IPv4 or IPv6 socket and the length from IP options turned 3632 * on at the socket. Assumes that the caller has a lock on the socket. 3633 */ 3634 3635 u32 kernel_sock_ip_overhead(struct sock *sk) 3636 { 3637 struct inet_sock *inet; 3638 struct ip_options_rcu *opt; 3639 u32 overhead = 0; 3640 #if IS_ENABLED(CONFIG_IPV6) 3641 struct ipv6_pinfo *np; 3642 struct ipv6_txoptions *optv6 = NULL; 3643 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3644 3645 if (!sk) 3646 return overhead; 3647 3648 switch (sk->sk_family) { 3649 case AF_INET: 3650 inet = inet_sk(sk); 3651 overhead += sizeof(struct iphdr); 3652 opt = rcu_dereference_protected(inet->inet_opt, 3653 sock_owned_by_user(sk)); 3654 if (opt) 3655 overhead += opt->opt.optlen; 3656 return overhead; 3657 #if IS_ENABLED(CONFIG_IPV6) 3658 case AF_INET6: 3659 np = inet6_sk(sk); 3660 overhead += sizeof(struct ipv6hdr); 3661 if (np) 3662 optv6 = rcu_dereference_protected(np->opt, 3663 sock_owned_by_user(sk)); 3664 if (optv6) 3665 overhead += (optv6->opt_flen + optv6->opt_nflen); 3666 return overhead; 3667 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3668 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ 3669 return overhead; 3670 } 3671 } 3672 EXPORT_SYMBOL(kernel_sock_ip_overhead); 3673