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