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_alloc_file */ 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 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 934 struct sk_buff *skb) 935 { 936 sock_recv_timestamp(msg, sk, skb); 937 sock_recv_drops(msg, sk, skb); 938 } 939 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops); 940 941 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *, 942 size_t, int)); 943 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *, 944 size_t, int)); 945 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 946 int flags) 947 { 948 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg, 949 inet_recvmsg, sock, msg, msg_data_left(msg), 950 flags); 951 } 952 953 /** 954 * sock_recvmsg - receive a message from @sock 955 * @sock: socket 956 * @msg: message to receive 957 * @flags: message flags 958 * 959 * Receives @msg from @sock, passing through LSM. Returns the total number 960 * of bytes received, or an error. 961 */ 962 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) 963 { 964 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); 965 966 return err ?: sock_recvmsg_nosec(sock, msg, flags); 967 } 968 EXPORT_SYMBOL(sock_recvmsg); 969 970 /** 971 * kernel_recvmsg - Receive a message from a socket (kernel space) 972 * @sock: The socket to receive the message from 973 * @msg: Received message 974 * @vec: Input s/g array for message data 975 * @num: Size of input s/g array 976 * @size: Number of bytes to read 977 * @flags: Message flags (MSG_DONTWAIT, etc...) 978 * 979 * On return the msg structure contains the scatter/gather array passed in the 980 * vec argument. The array is modified so that it consists of the unfilled 981 * portion of the original array. 982 * 983 * The returned value is the total number of bytes received, or an error. 984 */ 985 986 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 987 struct kvec *vec, size_t num, size_t size, int flags) 988 { 989 msg->msg_control_is_user = false; 990 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size); 991 return sock_recvmsg(sock, msg, flags); 992 } 993 EXPORT_SYMBOL(kernel_recvmsg); 994 995 static ssize_t sock_sendpage(struct file *file, struct page *page, 996 int offset, size_t size, loff_t *ppos, int more) 997 { 998 struct socket *sock; 999 int flags; 1000 1001 sock = file->private_data; 1002 1003 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 1004 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */ 1005 flags |= more; 1006 1007 return kernel_sendpage(sock, page, offset, size, flags); 1008 } 1009 1010 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 1011 struct pipe_inode_info *pipe, size_t len, 1012 unsigned int flags) 1013 { 1014 struct socket *sock = file->private_data; 1015 1016 if (unlikely(!sock->ops->splice_read)) 1017 return generic_file_splice_read(file, ppos, pipe, len, flags); 1018 1019 return sock->ops->splice_read(sock, ppos, pipe, len, flags); 1020 } 1021 1022 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) 1023 { 1024 struct file *file = iocb->ki_filp; 1025 struct socket *sock = file->private_data; 1026 struct msghdr msg = {.msg_iter = *to, 1027 .msg_iocb = iocb}; 1028 ssize_t res; 1029 1030 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1031 msg.msg_flags = MSG_DONTWAIT; 1032 1033 if (iocb->ki_pos != 0) 1034 return -ESPIPE; 1035 1036 if (!iov_iter_count(to)) /* Match SYS5 behaviour */ 1037 return 0; 1038 1039 res = sock_recvmsg(sock, &msg, msg.msg_flags); 1040 *to = msg.msg_iter; 1041 return res; 1042 } 1043 1044 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) 1045 { 1046 struct file *file = iocb->ki_filp; 1047 struct socket *sock = file->private_data; 1048 struct msghdr msg = {.msg_iter = *from, 1049 .msg_iocb = iocb}; 1050 ssize_t res; 1051 1052 if (iocb->ki_pos != 0) 1053 return -ESPIPE; 1054 1055 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1056 msg.msg_flags = MSG_DONTWAIT; 1057 1058 if (sock->type == SOCK_SEQPACKET) 1059 msg.msg_flags |= MSG_EOR; 1060 1061 res = sock_sendmsg(sock, &msg); 1062 *from = msg.msg_iter; 1063 return res; 1064 } 1065 1066 /* 1067 * Atomic setting of ioctl hooks to avoid race 1068 * with module unload. 1069 */ 1070 1071 static DEFINE_MUTEX(br_ioctl_mutex); 1072 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br, 1073 unsigned int cmd, struct ifreq *ifr, 1074 void __user *uarg); 1075 1076 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br, 1077 unsigned int cmd, struct ifreq *ifr, 1078 void __user *uarg)) 1079 { 1080 mutex_lock(&br_ioctl_mutex); 1081 br_ioctl_hook = hook; 1082 mutex_unlock(&br_ioctl_mutex); 1083 } 1084 EXPORT_SYMBOL(brioctl_set); 1085 1086 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd, 1087 struct ifreq *ifr, void __user *uarg) 1088 { 1089 int err = -ENOPKG; 1090 1091 if (!br_ioctl_hook) 1092 request_module("bridge"); 1093 1094 mutex_lock(&br_ioctl_mutex); 1095 if (br_ioctl_hook) 1096 err = br_ioctl_hook(net, br, cmd, ifr, uarg); 1097 mutex_unlock(&br_ioctl_mutex); 1098 1099 return err; 1100 } 1101 1102 static DEFINE_MUTEX(vlan_ioctl_mutex); 1103 static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 1104 1105 void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 1106 { 1107 mutex_lock(&vlan_ioctl_mutex); 1108 vlan_ioctl_hook = hook; 1109 mutex_unlock(&vlan_ioctl_mutex); 1110 } 1111 EXPORT_SYMBOL(vlan_ioctl_set); 1112 1113 static long sock_do_ioctl(struct net *net, struct socket *sock, 1114 unsigned int cmd, unsigned long arg) 1115 { 1116 struct ifreq ifr; 1117 bool need_copyout; 1118 int err; 1119 void __user *argp = (void __user *)arg; 1120 void __user *data; 1121 1122 err = sock->ops->ioctl(sock, cmd, arg); 1123 1124 /* 1125 * If this ioctl is unknown try to hand it down 1126 * to the NIC driver. 1127 */ 1128 if (err != -ENOIOCTLCMD) 1129 return err; 1130 1131 if (!is_socket_ioctl_cmd(cmd)) 1132 return -ENOTTY; 1133 1134 if (get_user_ifreq(&ifr, &data, argp)) 1135 return -EFAULT; 1136 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); 1137 if (!err && need_copyout) 1138 if (put_user_ifreq(&ifr, argp)) 1139 return -EFAULT; 1140 1141 return err; 1142 } 1143 1144 /* 1145 * With an ioctl, arg may well be a user mode pointer, but we don't know 1146 * what to do with it - that's up to the protocol still. 1147 */ 1148 1149 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 1150 { 1151 struct socket *sock; 1152 struct sock *sk; 1153 void __user *argp = (void __user *)arg; 1154 int pid, err; 1155 struct net *net; 1156 1157 sock = file->private_data; 1158 sk = sock->sk; 1159 net = sock_net(sk); 1160 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { 1161 struct ifreq ifr; 1162 void __user *data; 1163 bool need_copyout; 1164 if (get_user_ifreq(&ifr, &data, argp)) 1165 return -EFAULT; 1166 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); 1167 if (!err && need_copyout) 1168 if (put_user_ifreq(&ifr, argp)) 1169 return -EFAULT; 1170 } else 1171 #ifdef CONFIG_WEXT_CORE 1172 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 1173 err = wext_handle_ioctl(net, cmd, argp); 1174 } else 1175 #endif 1176 switch (cmd) { 1177 case FIOSETOWN: 1178 case SIOCSPGRP: 1179 err = -EFAULT; 1180 if (get_user(pid, (int __user *)argp)) 1181 break; 1182 err = f_setown(sock->file, pid, 1); 1183 break; 1184 case FIOGETOWN: 1185 case SIOCGPGRP: 1186 err = put_user(f_getown(sock->file), 1187 (int __user *)argp); 1188 break; 1189 case SIOCGIFBR: 1190 case SIOCSIFBR: 1191 case SIOCBRADDBR: 1192 case SIOCBRDELBR: 1193 err = br_ioctl_call(net, NULL, cmd, NULL, argp); 1194 break; 1195 case SIOCGIFVLAN: 1196 case SIOCSIFVLAN: 1197 err = -ENOPKG; 1198 if (!vlan_ioctl_hook) 1199 request_module("8021q"); 1200 1201 mutex_lock(&vlan_ioctl_mutex); 1202 if (vlan_ioctl_hook) 1203 err = vlan_ioctl_hook(net, argp); 1204 mutex_unlock(&vlan_ioctl_mutex); 1205 break; 1206 case SIOCGSKNS: 1207 err = -EPERM; 1208 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1209 break; 1210 1211 err = open_related_ns(&net->ns, get_net_ns); 1212 break; 1213 case SIOCGSTAMP_OLD: 1214 case SIOCGSTAMPNS_OLD: 1215 if (!sock->ops->gettstamp) { 1216 err = -ENOIOCTLCMD; 1217 break; 1218 } 1219 err = sock->ops->gettstamp(sock, argp, 1220 cmd == SIOCGSTAMP_OLD, 1221 !IS_ENABLED(CONFIG_64BIT)); 1222 break; 1223 case SIOCGSTAMP_NEW: 1224 case SIOCGSTAMPNS_NEW: 1225 if (!sock->ops->gettstamp) { 1226 err = -ENOIOCTLCMD; 1227 break; 1228 } 1229 err = sock->ops->gettstamp(sock, argp, 1230 cmd == SIOCGSTAMP_NEW, 1231 false); 1232 break; 1233 1234 case SIOCGIFCONF: 1235 err = dev_ifconf(net, argp); 1236 break; 1237 1238 default: 1239 err = sock_do_ioctl(net, sock, cmd, arg); 1240 break; 1241 } 1242 return err; 1243 } 1244 1245 /** 1246 * sock_create_lite - creates a socket 1247 * @family: protocol family (AF_INET, ...) 1248 * @type: communication type (SOCK_STREAM, ...) 1249 * @protocol: protocol (0, ...) 1250 * @res: new socket 1251 * 1252 * Creates a new socket and assigns it to @res, passing through LSM. 1253 * The new socket initialization is not complete, see kernel_accept(). 1254 * Returns 0 or an error. On failure @res is set to %NULL. 1255 * This function internally uses GFP_KERNEL. 1256 */ 1257 1258 int sock_create_lite(int family, int type, int protocol, struct socket **res) 1259 { 1260 int err; 1261 struct socket *sock = NULL; 1262 1263 err = security_socket_create(family, type, protocol, 1); 1264 if (err) 1265 goto out; 1266 1267 sock = sock_alloc(); 1268 if (!sock) { 1269 err = -ENOMEM; 1270 goto out; 1271 } 1272 1273 sock->type = type; 1274 err = security_socket_post_create(sock, family, type, protocol, 1); 1275 if (err) 1276 goto out_release; 1277 1278 out: 1279 *res = sock; 1280 return err; 1281 out_release: 1282 sock_release(sock); 1283 sock = NULL; 1284 goto out; 1285 } 1286 EXPORT_SYMBOL(sock_create_lite); 1287 1288 /* No kernel lock held - perfect */ 1289 static __poll_t sock_poll(struct file *file, poll_table *wait) 1290 { 1291 struct socket *sock = file->private_data; 1292 __poll_t events = poll_requested_events(wait), flag = 0; 1293 1294 if (!sock->ops->poll) 1295 return 0; 1296 1297 if (sk_can_busy_loop(sock->sk)) { 1298 /* poll once if requested by the syscall */ 1299 if (events & POLL_BUSY_LOOP) 1300 sk_busy_loop(sock->sk, 1); 1301 1302 /* if this socket can poll_ll, tell the system call */ 1303 flag = POLL_BUSY_LOOP; 1304 } 1305 1306 return sock->ops->poll(file, sock, wait) | flag; 1307 } 1308 1309 static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1310 { 1311 struct socket *sock = file->private_data; 1312 1313 return sock->ops->mmap(file, sock, vma); 1314 } 1315 1316 static int sock_close(struct inode *inode, struct file *filp) 1317 { 1318 __sock_release(SOCKET_I(inode), inode); 1319 return 0; 1320 } 1321 1322 /* 1323 * Update the socket async list 1324 * 1325 * Fasync_list locking strategy. 1326 * 1327 * 1. fasync_list is modified only under process context socket lock 1328 * i.e. under semaphore. 1329 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1330 * or under socket lock 1331 */ 1332 1333 static int sock_fasync(int fd, struct file *filp, int on) 1334 { 1335 struct socket *sock = filp->private_data; 1336 struct sock *sk = sock->sk; 1337 struct socket_wq *wq = &sock->wq; 1338 1339 if (sk == NULL) 1340 return -EINVAL; 1341 1342 lock_sock(sk); 1343 fasync_helper(fd, filp, on, &wq->fasync_list); 1344 1345 if (!wq->fasync_list) 1346 sock_reset_flag(sk, SOCK_FASYNC); 1347 else 1348 sock_set_flag(sk, SOCK_FASYNC); 1349 1350 release_sock(sk); 1351 return 0; 1352 } 1353 1354 /* This function may be called only under rcu_lock */ 1355 1356 int sock_wake_async(struct socket_wq *wq, int how, int band) 1357 { 1358 if (!wq || !wq->fasync_list) 1359 return -1; 1360 1361 switch (how) { 1362 case SOCK_WAKE_WAITD: 1363 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) 1364 break; 1365 goto call_kill; 1366 case SOCK_WAKE_SPACE: 1367 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) 1368 break; 1369 fallthrough; 1370 case SOCK_WAKE_IO: 1371 call_kill: 1372 kill_fasync(&wq->fasync_list, SIGIO, band); 1373 break; 1374 case SOCK_WAKE_URG: 1375 kill_fasync(&wq->fasync_list, SIGURG, band); 1376 } 1377 1378 return 0; 1379 } 1380 EXPORT_SYMBOL(sock_wake_async); 1381 1382 /** 1383 * __sock_create - creates a socket 1384 * @net: net namespace 1385 * @family: protocol family (AF_INET, ...) 1386 * @type: communication type (SOCK_STREAM, ...) 1387 * @protocol: protocol (0, ...) 1388 * @res: new socket 1389 * @kern: boolean for kernel space sockets 1390 * 1391 * Creates a new socket and assigns it to @res, passing through LSM. 1392 * Returns 0 or an error. On failure @res is set to %NULL. @kern must 1393 * be set to true if the socket resides in kernel space. 1394 * This function internally uses GFP_KERNEL. 1395 */ 1396 1397 int __sock_create(struct net *net, int family, int type, int protocol, 1398 struct socket **res, int kern) 1399 { 1400 int err; 1401 struct socket *sock; 1402 const struct net_proto_family *pf; 1403 1404 /* 1405 * Check protocol is in range 1406 */ 1407 if (family < 0 || family >= NPROTO) 1408 return -EAFNOSUPPORT; 1409 if (type < 0 || type >= SOCK_MAX) 1410 return -EINVAL; 1411 1412 /* Compatibility. 1413 1414 This uglymoron is moved from INET layer to here to avoid 1415 deadlock in module load. 1416 */ 1417 if (family == PF_INET && type == SOCK_PACKET) { 1418 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1419 current->comm); 1420 family = PF_PACKET; 1421 } 1422 1423 err = security_socket_create(family, type, protocol, kern); 1424 if (err) 1425 return err; 1426 1427 /* 1428 * Allocate the socket and allow the family to set things up. if 1429 * the protocol is 0, the family is instructed to select an appropriate 1430 * default. 1431 */ 1432 sock = sock_alloc(); 1433 if (!sock) { 1434 net_warn_ratelimited("socket: no more sockets\n"); 1435 return -ENFILE; /* Not exactly a match, but its the 1436 closest posix thing */ 1437 } 1438 1439 sock->type = type; 1440 1441 #ifdef CONFIG_MODULES 1442 /* Attempt to load a protocol module if the find failed. 1443 * 1444 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1445 * requested real, full-featured networking support upon configuration. 1446 * Otherwise module support will break! 1447 */ 1448 if (rcu_access_pointer(net_families[family]) == NULL) 1449 request_module("net-pf-%d", family); 1450 #endif 1451 1452 rcu_read_lock(); 1453 pf = rcu_dereference(net_families[family]); 1454 err = -EAFNOSUPPORT; 1455 if (!pf) 1456 goto out_release; 1457 1458 /* 1459 * We will call the ->create function, that possibly is in a loadable 1460 * module, so we have to bump that loadable module refcnt first. 1461 */ 1462 if (!try_module_get(pf->owner)) 1463 goto out_release; 1464 1465 /* Now protected by module ref count */ 1466 rcu_read_unlock(); 1467 1468 err = pf->create(net, sock, protocol, kern); 1469 if (err < 0) 1470 goto out_module_put; 1471 1472 /* 1473 * Now to bump the refcnt of the [loadable] module that owns this 1474 * socket at sock_release time we decrement its refcnt. 1475 */ 1476 if (!try_module_get(sock->ops->owner)) 1477 goto out_module_busy; 1478 1479 /* 1480 * Now that we're done with the ->create function, the [loadable] 1481 * module can have its refcnt decremented 1482 */ 1483 module_put(pf->owner); 1484 err = security_socket_post_create(sock, family, type, protocol, kern); 1485 if (err) 1486 goto out_sock_release; 1487 *res = sock; 1488 1489 return 0; 1490 1491 out_module_busy: 1492 err = -EAFNOSUPPORT; 1493 out_module_put: 1494 sock->ops = NULL; 1495 module_put(pf->owner); 1496 out_sock_release: 1497 sock_release(sock); 1498 return err; 1499 1500 out_release: 1501 rcu_read_unlock(); 1502 goto out_sock_release; 1503 } 1504 EXPORT_SYMBOL(__sock_create); 1505 1506 /** 1507 * sock_create - creates a socket 1508 * @family: protocol family (AF_INET, ...) 1509 * @type: communication type (SOCK_STREAM, ...) 1510 * @protocol: protocol (0, ...) 1511 * @res: new socket 1512 * 1513 * A wrapper around __sock_create(). 1514 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1515 */ 1516 1517 int sock_create(int family, int type, int protocol, struct socket **res) 1518 { 1519 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1520 } 1521 EXPORT_SYMBOL(sock_create); 1522 1523 /** 1524 * sock_create_kern - creates a socket (kernel space) 1525 * @net: net namespace 1526 * @family: protocol family (AF_INET, ...) 1527 * @type: communication type (SOCK_STREAM, ...) 1528 * @protocol: protocol (0, ...) 1529 * @res: new socket 1530 * 1531 * A wrapper around __sock_create(). 1532 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1533 */ 1534 1535 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) 1536 { 1537 return __sock_create(net, family, type, protocol, res, 1); 1538 } 1539 EXPORT_SYMBOL(sock_create_kern); 1540 1541 static struct socket *__sys_socket_create(int family, int type, int protocol) 1542 { 1543 struct socket *sock; 1544 int retval; 1545 1546 /* Check the SOCK_* constants for consistency. */ 1547 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1548 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1549 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1550 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1551 1552 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1553 return ERR_PTR(-EINVAL); 1554 type &= SOCK_TYPE_MASK; 1555 1556 retval = sock_create(family, type, protocol, &sock); 1557 if (retval < 0) 1558 return ERR_PTR(retval); 1559 1560 return sock; 1561 } 1562 1563 struct file *__sys_socket_file(int family, int type, int protocol) 1564 { 1565 struct socket *sock; 1566 struct file *file; 1567 int flags; 1568 1569 sock = __sys_socket_create(family, type, protocol); 1570 if (IS_ERR(sock)) 1571 return ERR_CAST(sock); 1572 1573 flags = type & ~SOCK_TYPE_MASK; 1574 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1575 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1576 1577 file = sock_alloc_file(sock, flags, NULL); 1578 if (IS_ERR(file)) 1579 sock_release(sock); 1580 1581 return file; 1582 } 1583 1584 int __sys_socket(int family, int type, int protocol) 1585 { 1586 struct socket *sock; 1587 int flags; 1588 1589 sock = __sys_socket_create(family, type, protocol); 1590 if (IS_ERR(sock)) 1591 return PTR_ERR(sock); 1592 1593 flags = type & ~SOCK_TYPE_MASK; 1594 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1595 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1596 1597 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1598 } 1599 1600 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1601 { 1602 return __sys_socket(family, type, protocol); 1603 } 1604 1605 /* 1606 * Create a pair of connected sockets. 1607 */ 1608 1609 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) 1610 { 1611 struct socket *sock1, *sock2; 1612 int fd1, fd2, err; 1613 struct file *newfile1, *newfile2; 1614 int flags; 1615 1616 flags = type & ~SOCK_TYPE_MASK; 1617 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1618 return -EINVAL; 1619 type &= SOCK_TYPE_MASK; 1620 1621 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1622 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1623 1624 /* 1625 * reserve descriptors and make sure we won't fail 1626 * to return them to userland. 1627 */ 1628 fd1 = get_unused_fd_flags(flags); 1629 if (unlikely(fd1 < 0)) 1630 return fd1; 1631 1632 fd2 = get_unused_fd_flags(flags); 1633 if (unlikely(fd2 < 0)) { 1634 put_unused_fd(fd1); 1635 return fd2; 1636 } 1637 1638 err = put_user(fd1, &usockvec[0]); 1639 if (err) 1640 goto out; 1641 1642 err = put_user(fd2, &usockvec[1]); 1643 if (err) 1644 goto out; 1645 1646 /* 1647 * Obtain the first socket and check if the underlying protocol 1648 * supports the socketpair call. 1649 */ 1650 1651 err = sock_create(family, type, protocol, &sock1); 1652 if (unlikely(err < 0)) 1653 goto out; 1654 1655 err = sock_create(family, type, protocol, &sock2); 1656 if (unlikely(err < 0)) { 1657 sock_release(sock1); 1658 goto out; 1659 } 1660 1661 err = security_socket_socketpair(sock1, sock2); 1662 if (unlikely(err)) { 1663 sock_release(sock2); 1664 sock_release(sock1); 1665 goto out; 1666 } 1667 1668 err = sock1->ops->socketpair(sock1, sock2); 1669 if (unlikely(err < 0)) { 1670 sock_release(sock2); 1671 sock_release(sock1); 1672 goto out; 1673 } 1674 1675 newfile1 = sock_alloc_file(sock1, flags, NULL); 1676 if (IS_ERR(newfile1)) { 1677 err = PTR_ERR(newfile1); 1678 sock_release(sock2); 1679 goto out; 1680 } 1681 1682 newfile2 = sock_alloc_file(sock2, flags, NULL); 1683 if (IS_ERR(newfile2)) { 1684 err = PTR_ERR(newfile2); 1685 fput(newfile1); 1686 goto out; 1687 } 1688 1689 audit_fd_pair(fd1, fd2); 1690 1691 fd_install(fd1, newfile1); 1692 fd_install(fd2, newfile2); 1693 return 0; 1694 1695 out: 1696 put_unused_fd(fd2); 1697 put_unused_fd(fd1); 1698 return err; 1699 } 1700 1701 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1702 int __user *, usockvec) 1703 { 1704 return __sys_socketpair(family, type, protocol, usockvec); 1705 } 1706 1707 /* 1708 * Bind a name to a socket. Nothing much to do here since it's 1709 * the protocol's responsibility to handle the local address. 1710 * 1711 * We move the socket address to kernel space before we call 1712 * the protocol layer (having also checked the address is ok). 1713 */ 1714 1715 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) 1716 { 1717 struct socket *sock; 1718 struct sockaddr_storage address; 1719 int err, fput_needed; 1720 1721 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1722 if (sock) { 1723 err = move_addr_to_kernel(umyaddr, addrlen, &address); 1724 if (!err) { 1725 err = security_socket_bind(sock, 1726 (struct sockaddr *)&address, 1727 addrlen); 1728 if (!err) 1729 err = sock->ops->bind(sock, 1730 (struct sockaddr *) 1731 &address, addrlen); 1732 } 1733 fput_light(sock->file, fput_needed); 1734 } 1735 return err; 1736 } 1737 1738 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1739 { 1740 return __sys_bind(fd, umyaddr, addrlen); 1741 } 1742 1743 /* 1744 * Perform a listen. Basically, we allow the protocol to do anything 1745 * necessary for a listen, and if that works, we mark the socket as 1746 * ready for listening. 1747 */ 1748 1749 int __sys_listen(int fd, int backlog) 1750 { 1751 struct socket *sock; 1752 int err, fput_needed; 1753 int somaxconn; 1754 1755 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1756 if (sock) { 1757 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; 1758 if ((unsigned int)backlog > somaxconn) 1759 backlog = somaxconn; 1760 1761 err = security_socket_listen(sock, backlog); 1762 if (!err) 1763 err = sock->ops->listen(sock, backlog); 1764 1765 fput_light(sock->file, fput_needed); 1766 } 1767 return err; 1768 } 1769 1770 SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1771 { 1772 return __sys_listen(fd, backlog); 1773 } 1774 1775 struct file *do_accept(struct file *file, unsigned file_flags, 1776 struct sockaddr __user *upeer_sockaddr, 1777 int __user *upeer_addrlen, int flags) 1778 { 1779 struct socket *sock, *newsock; 1780 struct file *newfile; 1781 int err, len; 1782 struct sockaddr_storage address; 1783 1784 sock = sock_from_file(file); 1785 if (!sock) 1786 return ERR_PTR(-ENOTSOCK); 1787 1788 newsock = sock_alloc(); 1789 if (!newsock) 1790 return ERR_PTR(-ENFILE); 1791 1792 newsock->type = sock->type; 1793 newsock->ops = sock->ops; 1794 1795 /* 1796 * We don't need try_module_get here, as the listening socket (sock) 1797 * has the protocol module (sock->ops->owner) held. 1798 */ 1799 __module_get(newsock->ops->owner); 1800 1801 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); 1802 if (IS_ERR(newfile)) 1803 return newfile; 1804 1805 err = security_socket_accept(sock, newsock); 1806 if (err) 1807 goto out_fd; 1808 1809 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags, 1810 false); 1811 if (err < 0) 1812 goto out_fd; 1813 1814 if (upeer_sockaddr) { 1815 len = newsock->ops->getname(newsock, 1816 (struct sockaddr *)&address, 2); 1817 if (len < 0) { 1818 err = -ECONNABORTED; 1819 goto out_fd; 1820 } 1821 err = move_addr_to_user(&address, 1822 len, upeer_sockaddr, upeer_addrlen); 1823 if (err < 0) 1824 goto out_fd; 1825 } 1826 1827 /* File flags are not inherited via accept() unlike another OSes. */ 1828 return newfile; 1829 out_fd: 1830 fput(newfile); 1831 return ERR_PTR(err); 1832 } 1833 1834 int __sys_accept4_file(struct file *file, unsigned file_flags, 1835 struct sockaddr __user *upeer_sockaddr, 1836 int __user *upeer_addrlen, int flags, 1837 unsigned long nofile) 1838 { 1839 struct file *newfile; 1840 int newfd; 1841 1842 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1843 return -EINVAL; 1844 1845 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1846 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1847 1848 newfd = __get_unused_fd_flags(flags, nofile); 1849 if (unlikely(newfd < 0)) 1850 return newfd; 1851 1852 newfile = do_accept(file, file_flags, upeer_sockaddr, upeer_addrlen, 1853 flags); 1854 if (IS_ERR(newfile)) { 1855 put_unused_fd(newfd); 1856 return PTR_ERR(newfile); 1857 } 1858 fd_install(newfd, newfile); 1859 return newfd; 1860 } 1861 1862 /* 1863 * For accept, we attempt to create a new socket, set up the link 1864 * with the client, wake up the client, then return the new 1865 * connected fd. We collect the address of the connector in kernel 1866 * space and move it to user at the very end. This is unclean because 1867 * we open the socket then return an error. 1868 * 1869 * 1003.1g adds the ability to recvmsg() to query connection pending 1870 * status to recvmsg. We need to add that support in a way thats 1871 * clean when we restructure accept also. 1872 */ 1873 1874 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, 1875 int __user *upeer_addrlen, int flags) 1876 { 1877 int ret = -EBADF; 1878 struct fd f; 1879 1880 f = fdget(fd); 1881 if (f.file) { 1882 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr, 1883 upeer_addrlen, flags, 1884 rlimit(RLIMIT_NOFILE)); 1885 fdput(f); 1886 } 1887 1888 return ret; 1889 } 1890 1891 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 1892 int __user *, upeer_addrlen, int, flags) 1893 { 1894 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); 1895 } 1896 1897 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 1898 int __user *, upeer_addrlen) 1899 { 1900 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 1901 } 1902 1903 /* 1904 * Attempt to connect to a socket with the server address. The address 1905 * is in user space so we verify it is OK and move it to kernel space. 1906 * 1907 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 1908 * break bindings 1909 * 1910 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 1911 * other SEQPACKET protocols that take time to connect() as it doesn't 1912 * include the -EINPROGRESS status for such sockets. 1913 */ 1914 1915 int __sys_connect_file(struct file *file, struct sockaddr_storage *address, 1916 int addrlen, int file_flags) 1917 { 1918 struct socket *sock; 1919 int err; 1920 1921 sock = sock_from_file(file); 1922 if (!sock) { 1923 err = -ENOTSOCK; 1924 goto out; 1925 } 1926 1927 err = 1928 security_socket_connect(sock, (struct sockaddr *)address, addrlen); 1929 if (err) 1930 goto out; 1931 1932 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen, 1933 sock->file->f_flags | file_flags); 1934 out: 1935 return err; 1936 } 1937 1938 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) 1939 { 1940 int ret = -EBADF; 1941 struct fd f; 1942 1943 f = fdget(fd); 1944 if (f.file) { 1945 struct sockaddr_storage address; 1946 1947 ret = move_addr_to_kernel(uservaddr, addrlen, &address); 1948 if (!ret) 1949 ret = __sys_connect_file(f.file, &address, addrlen, 0); 1950 fdput(f); 1951 } 1952 1953 return ret; 1954 } 1955 1956 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 1957 int, addrlen) 1958 { 1959 return __sys_connect(fd, uservaddr, addrlen); 1960 } 1961 1962 /* 1963 * Get the local address ('name') of a socket object. Move the obtained 1964 * name to user space. 1965 */ 1966 1967 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, 1968 int __user *usockaddr_len) 1969 { 1970 struct socket *sock; 1971 struct sockaddr_storage address; 1972 int err, fput_needed; 1973 1974 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1975 if (!sock) 1976 goto out; 1977 1978 err = security_socket_getsockname(sock); 1979 if (err) 1980 goto out_put; 1981 1982 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0); 1983 if (err < 0) 1984 goto out_put; 1985 /* "err" is actually length in this case */ 1986 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); 1987 1988 out_put: 1989 fput_light(sock->file, fput_needed); 1990 out: 1991 return err; 1992 } 1993 1994 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 1995 int __user *, usockaddr_len) 1996 { 1997 return __sys_getsockname(fd, usockaddr, usockaddr_len); 1998 } 1999 2000 /* 2001 * Get the remote address ('name') of a socket object. Move the obtained 2002 * name to user space. 2003 */ 2004 2005 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, 2006 int __user *usockaddr_len) 2007 { 2008 struct socket *sock; 2009 struct sockaddr_storage address; 2010 int err, fput_needed; 2011 2012 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2013 if (sock != NULL) { 2014 err = security_socket_getpeername(sock); 2015 if (err) { 2016 fput_light(sock->file, fput_needed); 2017 return err; 2018 } 2019 2020 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1); 2021 if (err >= 0) 2022 /* "err" is actually length in this case */ 2023 err = move_addr_to_user(&address, err, usockaddr, 2024 usockaddr_len); 2025 fput_light(sock->file, fput_needed); 2026 } 2027 return err; 2028 } 2029 2030 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 2031 int __user *, usockaddr_len) 2032 { 2033 return __sys_getpeername(fd, usockaddr, usockaddr_len); 2034 } 2035 2036 /* 2037 * Send a datagram to a given address. We move the address into kernel 2038 * space and check the user space data area is readable before invoking 2039 * the protocol. 2040 */ 2041 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, 2042 struct sockaddr __user *addr, int addr_len) 2043 { 2044 struct socket *sock; 2045 struct sockaddr_storage address; 2046 int err; 2047 struct msghdr msg; 2048 struct iovec iov; 2049 int fput_needed; 2050 2051 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter); 2052 if (unlikely(err)) 2053 return err; 2054 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2055 if (!sock) 2056 goto out; 2057 2058 msg.msg_name = NULL; 2059 msg.msg_control = NULL; 2060 msg.msg_controllen = 0; 2061 msg.msg_namelen = 0; 2062 if (addr) { 2063 err = move_addr_to_kernel(addr, addr_len, &address); 2064 if (err < 0) 2065 goto out_put; 2066 msg.msg_name = (struct sockaddr *)&address; 2067 msg.msg_namelen = addr_len; 2068 } 2069 if (sock->file->f_flags & O_NONBLOCK) 2070 flags |= MSG_DONTWAIT; 2071 msg.msg_flags = flags; 2072 err = sock_sendmsg(sock, &msg); 2073 2074 out_put: 2075 fput_light(sock->file, fput_needed); 2076 out: 2077 return err; 2078 } 2079 2080 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 2081 unsigned int, flags, struct sockaddr __user *, addr, 2082 int, addr_len) 2083 { 2084 return __sys_sendto(fd, buff, len, flags, addr, addr_len); 2085 } 2086 2087 /* 2088 * Send a datagram down a socket. 2089 */ 2090 2091 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 2092 unsigned int, flags) 2093 { 2094 return __sys_sendto(fd, buff, len, flags, NULL, 0); 2095 } 2096 2097 /* 2098 * Receive a frame from the socket and optionally record the address of the 2099 * sender. We verify the buffers are writable and if needed move the 2100 * sender address from kernel to user space. 2101 */ 2102 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, 2103 struct sockaddr __user *addr, int __user *addr_len) 2104 { 2105 struct socket *sock; 2106 struct iovec iov; 2107 struct msghdr msg; 2108 struct sockaddr_storage address; 2109 int err, err2; 2110 int fput_needed; 2111 2112 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter); 2113 if (unlikely(err)) 2114 return err; 2115 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2116 if (!sock) 2117 goto out; 2118 2119 msg.msg_control = NULL; 2120 msg.msg_controllen = 0; 2121 /* Save some cycles and don't copy the address if not needed */ 2122 msg.msg_name = addr ? (struct sockaddr *)&address : NULL; 2123 /* We assume all kernel code knows the size of sockaddr_storage */ 2124 msg.msg_namelen = 0; 2125 msg.msg_iocb = NULL; 2126 msg.msg_flags = 0; 2127 if (sock->file->f_flags & O_NONBLOCK) 2128 flags |= MSG_DONTWAIT; 2129 err = sock_recvmsg(sock, &msg, flags); 2130 2131 if (err >= 0 && addr != NULL) { 2132 err2 = move_addr_to_user(&address, 2133 msg.msg_namelen, addr, addr_len); 2134 if (err2 < 0) 2135 err = err2; 2136 } 2137 2138 fput_light(sock->file, fput_needed); 2139 out: 2140 return err; 2141 } 2142 2143 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 2144 unsigned int, flags, struct sockaddr __user *, addr, 2145 int __user *, addr_len) 2146 { 2147 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); 2148 } 2149 2150 /* 2151 * Receive a datagram from a socket. 2152 */ 2153 2154 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, 2155 unsigned int, flags) 2156 { 2157 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 2158 } 2159 2160 static bool sock_use_custom_sol_socket(const struct socket *sock) 2161 { 2162 const struct sock *sk = sock->sk; 2163 2164 /* Use sock->ops->setsockopt() for MPTCP */ 2165 return IS_ENABLED(CONFIG_MPTCP) && 2166 sk->sk_protocol == IPPROTO_MPTCP && 2167 sk->sk_type == SOCK_STREAM && 2168 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6); 2169 } 2170 2171 /* 2172 * Set a socket option. Because we don't know the option lengths we have 2173 * to pass the user mode parameter for the protocols to sort out. 2174 */ 2175 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval, 2176 int optlen) 2177 { 2178 sockptr_t optval = USER_SOCKPTR(user_optval); 2179 char *kernel_optval = NULL; 2180 int err, fput_needed; 2181 struct socket *sock; 2182 2183 if (optlen < 0) 2184 return -EINVAL; 2185 2186 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2187 if (!sock) 2188 return err; 2189 2190 err = security_socket_setsockopt(sock, level, optname); 2191 if (err) 2192 goto out_put; 2193 2194 if (!in_compat_syscall()) 2195 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname, 2196 user_optval, &optlen, 2197 &kernel_optval); 2198 if (err < 0) 2199 goto out_put; 2200 if (err > 0) { 2201 err = 0; 2202 goto out_put; 2203 } 2204 2205 if (kernel_optval) 2206 optval = KERNEL_SOCKPTR(kernel_optval); 2207 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock)) 2208 err = sock_setsockopt(sock, level, optname, optval, optlen); 2209 else if (unlikely(!sock->ops->setsockopt)) 2210 err = -EOPNOTSUPP; 2211 else 2212 err = sock->ops->setsockopt(sock, level, optname, optval, 2213 optlen); 2214 kfree(kernel_optval); 2215 out_put: 2216 fput_light(sock->file, fput_needed); 2217 return err; 2218 } 2219 2220 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 2221 char __user *, optval, int, optlen) 2222 { 2223 return __sys_setsockopt(fd, level, optname, optval, optlen); 2224 } 2225 2226 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level, 2227 int optname)); 2228 2229 /* 2230 * Get a socket option. Because we don't know the option lengths we have 2231 * to pass a user mode parameter for the protocols to sort out. 2232 */ 2233 int __sys_getsockopt(int fd, int level, int optname, char __user *optval, 2234 int __user *optlen) 2235 { 2236 int err, fput_needed; 2237 struct socket *sock; 2238 int max_optlen; 2239 2240 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2241 if (!sock) 2242 return err; 2243 2244 err = security_socket_getsockopt(sock, level, optname); 2245 if (err) 2246 goto out_put; 2247 2248 if (!in_compat_syscall()) 2249 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen); 2250 2251 if (level == SOL_SOCKET) 2252 err = sock_getsockopt(sock, level, optname, optval, optlen); 2253 else if (unlikely(!sock->ops->getsockopt)) 2254 err = -EOPNOTSUPP; 2255 else 2256 err = sock->ops->getsockopt(sock, level, optname, optval, 2257 optlen); 2258 2259 if (!in_compat_syscall()) 2260 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname, 2261 optval, optlen, max_optlen, 2262 err); 2263 out_put: 2264 fput_light(sock->file, fput_needed); 2265 return err; 2266 } 2267 2268 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 2269 char __user *, optval, int __user *, optlen) 2270 { 2271 return __sys_getsockopt(fd, level, optname, optval, optlen); 2272 } 2273 2274 /* 2275 * Shutdown a socket. 2276 */ 2277 2278 int __sys_shutdown_sock(struct socket *sock, int how) 2279 { 2280 int err; 2281 2282 err = security_socket_shutdown(sock, how); 2283 if (!err) 2284 err = sock->ops->shutdown(sock, how); 2285 2286 return err; 2287 } 2288 2289 int __sys_shutdown(int fd, int how) 2290 { 2291 int err, fput_needed; 2292 struct socket *sock; 2293 2294 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2295 if (sock != NULL) { 2296 err = __sys_shutdown_sock(sock, how); 2297 fput_light(sock->file, fput_needed); 2298 } 2299 return err; 2300 } 2301 2302 SYSCALL_DEFINE2(shutdown, int, fd, int, how) 2303 { 2304 return __sys_shutdown(fd, how); 2305 } 2306 2307 /* A couple of helpful macros for getting the address of the 32/64 bit 2308 * fields which are the same type (int / unsigned) on our platforms. 2309 */ 2310 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 2311 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 2312 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 2313 2314 struct used_address { 2315 struct sockaddr_storage name; 2316 unsigned int name_len; 2317 }; 2318 2319 int __copy_msghdr_from_user(struct msghdr *kmsg, 2320 struct user_msghdr __user *umsg, 2321 struct sockaddr __user **save_addr, 2322 struct iovec __user **uiov, size_t *nsegs) 2323 { 2324 struct user_msghdr msg; 2325 ssize_t err; 2326 2327 if (copy_from_user(&msg, umsg, sizeof(*umsg))) 2328 return -EFAULT; 2329 2330 kmsg->msg_control_is_user = true; 2331 kmsg->msg_control_user = msg.msg_control; 2332 kmsg->msg_controllen = msg.msg_controllen; 2333 kmsg->msg_flags = msg.msg_flags; 2334 2335 kmsg->msg_namelen = msg.msg_namelen; 2336 if (!msg.msg_name) 2337 kmsg->msg_namelen = 0; 2338 2339 if (kmsg->msg_namelen < 0) 2340 return -EINVAL; 2341 2342 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) 2343 kmsg->msg_namelen = sizeof(struct sockaddr_storage); 2344 2345 if (save_addr) 2346 *save_addr = msg.msg_name; 2347 2348 if (msg.msg_name && kmsg->msg_namelen) { 2349 if (!save_addr) { 2350 err = move_addr_to_kernel(msg.msg_name, 2351 kmsg->msg_namelen, 2352 kmsg->msg_name); 2353 if (err < 0) 2354 return err; 2355 } 2356 } else { 2357 kmsg->msg_name = NULL; 2358 kmsg->msg_namelen = 0; 2359 } 2360 2361 if (msg.msg_iovlen > UIO_MAXIOV) 2362 return -EMSGSIZE; 2363 2364 kmsg->msg_iocb = NULL; 2365 *uiov = msg.msg_iov; 2366 *nsegs = msg.msg_iovlen; 2367 return 0; 2368 } 2369 2370 static int copy_msghdr_from_user(struct msghdr *kmsg, 2371 struct user_msghdr __user *umsg, 2372 struct sockaddr __user **save_addr, 2373 struct iovec **iov) 2374 { 2375 struct user_msghdr msg; 2376 ssize_t err; 2377 2378 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov, 2379 &msg.msg_iovlen); 2380 if (err) 2381 return err; 2382 2383 err = import_iovec(save_addr ? READ : WRITE, 2384 msg.msg_iov, msg.msg_iovlen, 2385 UIO_FASTIOV, iov, &kmsg->msg_iter); 2386 return err < 0 ? err : 0; 2387 } 2388 2389 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, 2390 unsigned int flags, struct used_address *used_address, 2391 unsigned int allowed_msghdr_flags) 2392 { 2393 unsigned char ctl[sizeof(struct cmsghdr) + 20] 2394 __aligned(sizeof(__kernel_size_t)); 2395 /* 20 is size of ipv6_pktinfo */ 2396 unsigned char *ctl_buf = ctl; 2397 int ctl_len; 2398 ssize_t err; 2399 2400 err = -ENOBUFS; 2401 2402 if (msg_sys->msg_controllen > INT_MAX) 2403 goto out; 2404 flags |= (msg_sys->msg_flags & allowed_msghdr_flags); 2405 ctl_len = msg_sys->msg_controllen; 2406 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 2407 err = 2408 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, 2409 sizeof(ctl)); 2410 if (err) 2411 goto out; 2412 ctl_buf = msg_sys->msg_control; 2413 ctl_len = msg_sys->msg_controllen; 2414 } else if (ctl_len) { 2415 BUILD_BUG_ON(sizeof(struct cmsghdr) != 2416 CMSG_ALIGN(sizeof(struct cmsghdr))); 2417 if (ctl_len > sizeof(ctl)) { 2418 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 2419 if (ctl_buf == NULL) 2420 goto out; 2421 } 2422 err = -EFAULT; 2423 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len)) 2424 goto out_freectl; 2425 msg_sys->msg_control = ctl_buf; 2426 msg_sys->msg_control_is_user = false; 2427 } 2428 msg_sys->msg_flags = flags; 2429 2430 if (sock->file->f_flags & O_NONBLOCK) 2431 msg_sys->msg_flags |= MSG_DONTWAIT; 2432 /* 2433 * If this is sendmmsg() and current destination address is same as 2434 * previously succeeded address, omit asking LSM's decision. 2435 * used_address->name_len is initialized to UINT_MAX so that the first 2436 * destination address never matches. 2437 */ 2438 if (used_address && msg_sys->msg_name && 2439 used_address->name_len == msg_sys->msg_namelen && 2440 !memcmp(&used_address->name, msg_sys->msg_name, 2441 used_address->name_len)) { 2442 err = sock_sendmsg_nosec(sock, msg_sys); 2443 goto out_freectl; 2444 } 2445 err = sock_sendmsg(sock, msg_sys); 2446 /* 2447 * If this is sendmmsg() and sending to current destination address was 2448 * successful, remember it. 2449 */ 2450 if (used_address && err >= 0) { 2451 used_address->name_len = msg_sys->msg_namelen; 2452 if (msg_sys->msg_name) 2453 memcpy(&used_address->name, msg_sys->msg_name, 2454 used_address->name_len); 2455 } 2456 2457 out_freectl: 2458 if (ctl_buf != ctl) 2459 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 2460 out: 2461 return err; 2462 } 2463 2464 int sendmsg_copy_msghdr(struct msghdr *msg, 2465 struct user_msghdr __user *umsg, unsigned flags, 2466 struct iovec **iov) 2467 { 2468 int err; 2469 2470 if (flags & MSG_CMSG_COMPAT) { 2471 struct compat_msghdr __user *msg_compat; 2472 2473 msg_compat = (struct compat_msghdr __user *) umsg; 2474 err = get_compat_msghdr(msg, msg_compat, NULL, iov); 2475 } else { 2476 err = copy_msghdr_from_user(msg, umsg, NULL, iov); 2477 } 2478 if (err < 0) 2479 return err; 2480 2481 return 0; 2482 } 2483 2484 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, 2485 struct msghdr *msg_sys, unsigned int flags, 2486 struct used_address *used_address, 2487 unsigned int allowed_msghdr_flags) 2488 { 2489 struct sockaddr_storage address; 2490 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2491 ssize_t err; 2492 2493 msg_sys->msg_name = &address; 2494 2495 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov); 2496 if (err < 0) 2497 return err; 2498 2499 err = ____sys_sendmsg(sock, msg_sys, flags, used_address, 2500 allowed_msghdr_flags); 2501 kfree(iov); 2502 return err; 2503 } 2504 2505 /* 2506 * BSD sendmsg interface 2507 */ 2508 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg, 2509 unsigned int flags) 2510 { 2511 return ____sys_sendmsg(sock, msg, flags, NULL, 0); 2512 } 2513 2514 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2515 bool forbid_cmsg_compat) 2516 { 2517 int fput_needed, err; 2518 struct msghdr msg_sys; 2519 struct socket *sock; 2520 2521 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2522 return -EINVAL; 2523 2524 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2525 if (!sock) 2526 goto out; 2527 2528 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); 2529 2530 fput_light(sock->file, fput_needed); 2531 out: 2532 return err; 2533 } 2534 2535 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) 2536 { 2537 return __sys_sendmsg(fd, msg, flags, true); 2538 } 2539 2540 /* 2541 * Linux sendmmsg interface 2542 */ 2543 2544 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2545 unsigned int flags, bool forbid_cmsg_compat) 2546 { 2547 int fput_needed, err, datagrams; 2548 struct socket *sock; 2549 struct mmsghdr __user *entry; 2550 struct compat_mmsghdr __user *compat_entry; 2551 struct msghdr msg_sys; 2552 struct used_address used_address; 2553 unsigned int oflags = flags; 2554 2555 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2556 return -EINVAL; 2557 2558 if (vlen > UIO_MAXIOV) 2559 vlen = UIO_MAXIOV; 2560 2561 datagrams = 0; 2562 2563 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2564 if (!sock) 2565 return err; 2566 2567 used_address.name_len = UINT_MAX; 2568 entry = mmsg; 2569 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2570 err = 0; 2571 flags |= MSG_BATCH; 2572 2573 while (datagrams < vlen) { 2574 if (datagrams == vlen - 1) 2575 flags = oflags; 2576 2577 if (MSG_CMSG_COMPAT & flags) { 2578 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, 2579 &msg_sys, flags, &used_address, MSG_EOR); 2580 if (err < 0) 2581 break; 2582 err = __put_user(err, &compat_entry->msg_len); 2583 ++compat_entry; 2584 } else { 2585 err = ___sys_sendmsg(sock, 2586 (struct user_msghdr __user *)entry, 2587 &msg_sys, flags, &used_address, MSG_EOR); 2588 if (err < 0) 2589 break; 2590 err = put_user(err, &entry->msg_len); 2591 ++entry; 2592 } 2593 2594 if (err) 2595 break; 2596 ++datagrams; 2597 if (msg_data_left(&msg_sys)) 2598 break; 2599 cond_resched(); 2600 } 2601 2602 fput_light(sock->file, fput_needed); 2603 2604 /* We only return an error if no datagrams were able to be sent */ 2605 if (datagrams != 0) 2606 return datagrams; 2607 2608 return err; 2609 } 2610 2611 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, 2612 unsigned int, vlen, unsigned int, flags) 2613 { 2614 return __sys_sendmmsg(fd, mmsg, vlen, flags, true); 2615 } 2616 2617 int recvmsg_copy_msghdr(struct msghdr *msg, 2618 struct user_msghdr __user *umsg, unsigned flags, 2619 struct sockaddr __user **uaddr, 2620 struct iovec **iov) 2621 { 2622 ssize_t err; 2623 2624 if (MSG_CMSG_COMPAT & flags) { 2625 struct compat_msghdr __user *msg_compat; 2626 2627 msg_compat = (struct compat_msghdr __user *) umsg; 2628 err = get_compat_msghdr(msg, msg_compat, uaddr, iov); 2629 } else { 2630 err = copy_msghdr_from_user(msg, umsg, uaddr, iov); 2631 } 2632 if (err < 0) 2633 return err; 2634 2635 return 0; 2636 } 2637 2638 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys, 2639 struct user_msghdr __user *msg, 2640 struct sockaddr __user *uaddr, 2641 unsigned int flags, int nosec) 2642 { 2643 struct compat_msghdr __user *msg_compat = 2644 (struct compat_msghdr __user *) msg; 2645 int __user *uaddr_len = COMPAT_NAMELEN(msg); 2646 struct sockaddr_storage addr; 2647 unsigned long cmsg_ptr; 2648 int len; 2649 ssize_t err; 2650 2651 msg_sys->msg_name = &addr; 2652 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2653 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2654 2655 /* We assume all kernel code knows the size of sockaddr_storage */ 2656 msg_sys->msg_namelen = 0; 2657 2658 if (sock->file->f_flags & O_NONBLOCK) 2659 flags |= MSG_DONTWAIT; 2660 2661 if (unlikely(nosec)) 2662 err = sock_recvmsg_nosec(sock, msg_sys, flags); 2663 else 2664 err = sock_recvmsg(sock, msg_sys, flags); 2665 2666 if (err < 0) 2667 goto out; 2668 len = err; 2669 2670 if (uaddr != NULL) { 2671 err = move_addr_to_user(&addr, 2672 msg_sys->msg_namelen, uaddr, 2673 uaddr_len); 2674 if (err < 0) 2675 goto out; 2676 } 2677 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2678 COMPAT_FLAGS(msg)); 2679 if (err) 2680 goto out; 2681 if (MSG_CMSG_COMPAT & flags) 2682 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2683 &msg_compat->msg_controllen); 2684 else 2685 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2686 &msg->msg_controllen); 2687 if (err) 2688 goto out; 2689 err = len; 2690 out: 2691 return err; 2692 } 2693 2694 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, 2695 struct msghdr *msg_sys, unsigned int flags, int nosec) 2696 { 2697 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2698 /* user mode address pointers */ 2699 struct sockaddr __user *uaddr; 2700 ssize_t err; 2701 2702 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov); 2703 if (err < 0) 2704 return err; 2705 2706 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec); 2707 kfree(iov); 2708 return err; 2709 } 2710 2711 /* 2712 * BSD recvmsg interface 2713 */ 2714 2715 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg, 2716 struct user_msghdr __user *umsg, 2717 struct sockaddr __user *uaddr, unsigned int flags) 2718 { 2719 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0); 2720 } 2721 2722 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2723 bool forbid_cmsg_compat) 2724 { 2725 int fput_needed, err; 2726 struct msghdr msg_sys; 2727 struct socket *sock; 2728 2729 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2730 return -EINVAL; 2731 2732 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2733 if (!sock) 2734 goto out; 2735 2736 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2737 2738 fput_light(sock->file, fput_needed); 2739 out: 2740 return err; 2741 } 2742 2743 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, 2744 unsigned int, flags) 2745 { 2746 return __sys_recvmsg(fd, msg, flags, true); 2747 } 2748 2749 /* 2750 * Linux recvmmsg interface 2751 */ 2752 2753 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2754 unsigned int vlen, unsigned int flags, 2755 struct timespec64 *timeout) 2756 { 2757 int fput_needed, err, datagrams; 2758 struct socket *sock; 2759 struct mmsghdr __user *entry; 2760 struct compat_mmsghdr __user *compat_entry; 2761 struct msghdr msg_sys; 2762 struct timespec64 end_time; 2763 struct timespec64 timeout64; 2764 2765 if (timeout && 2766 poll_select_set_timeout(&end_time, timeout->tv_sec, 2767 timeout->tv_nsec)) 2768 return -EINVAL; 2769 2770 datagrams = 0; 2771 2772 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2773 if (!sock) 2774 return err; 2775 2776 if (likely(!(flags & MSG_ERRQUEUE))) { 2777 err = sock_error(sock->sk); 2778 if (err) { 2779 datagrams = err; 2780 goto out_put; 2781 } 2782 } 2783 2784 entry = mmsg; 2785 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2786 2787 while (datagrams < vlen) { 2788 /* 2789 * No need to ask LSM for more than the first datagram. 2790 */ 2791 if (MSG_CMSG_COMPAT & flags) { 2792 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, 2793 &msg_sys, flags & ~MSG_WAITFORONE, 2794 datagrams); 2795 if (err < 0) 2796 break; 2797 err = __put_user(err, &compat_entry->msg_len); 2798 ++compat_entry; 2799 } else { 2800 err = ___sys_recvmsg(sock, 2801 (struct user_msghdr __user *)entry, 2802 &msg_sys, flags & ~MSG_WAITFORONE, 2803 datagrams); 2804 if (err < 0) 2805 break; 2806 err = put_user(err, &entry->msg_len); 2807 ++entry; 2808 } 2809 2810 if (err) 2811 break; 2812 ++datagrams; 2813 2814 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2815 if (flags & MSG_WAITFORONE) 2816 flags |= MSG_DONTWAIT; 2817 2818 if (timeout) { 2819 ktime_get_ts64(&timeout64); 2820 *timeout = timespec64_sub(end_time, timeout64); 2821 if (timeout->tv_sec < 0) { 2822 timeout->tv_sec = timeout->tv_nsec = 0; 2823 break; 2824 } 2825 2826 /* Timeout, return less than vlen datagrams */ 2827 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2828 break; 2829 } 2830 2831 /* Out of band data, return right away */ 2832 if (msg_sys.msg_flags & MSG_OOB) 2833 break; 2834 cond_resched(); 2835 } 2836 2837 if (err == 0) 2838 goto out_put; 2839 2840 if (datagrams == 0) { 2841 datagrams = err; 2842 goto out_put; 2843 } 2844 2845 /* 2846 * We may return less entries than requested (vlen) if the 2847 * sock is non block and there aren't enough datagrams... 2848 */ 2849 if (err != -EAGAIN) { 2850 /* 2851 * ... or if recvmsg returns an error after we 2852 * received some datagrams, where we record the 2853 * error to return on the next call or if the 2854 * app asks about it using getsockopt(SO_ERROR). 2855 */ 2856 sock->sk->sk_err = -err; 2857 } 2858 out_put: 2859 fput_light(sock->file, fput_needed); 2860 2861 return datagrams; 2862 } 2863 2864 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2865 unsigned int vlen, unsigned int flags, 2866 struct __kernel_timespec __user *timeout, 2867 struct old_timespec32 __user *timeout32) 2868 { 2869 int datagrams; 2870 struct timespec64 timeout_sys; 2871 2872 if (timeout && get_timespec64(&timeout_sys, timeout)) 2873 return -EFAULT; 2874 2875 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32)) 2876 return -EFAULT; 2877 2878 if (!timeout && !timeout32) 2879 return do_recvmmsg(fd, mmsg, vlen, flags, NULL); 2880 2881 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 2882 2883 if (datagrams <= 0) 2884 return datagrams; 2885 2886 if (timeout && put_timespec64(&timeout_sys, timeout)) 2887 datagrams = -EFAULT; 2888 2889 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32)) 2890 datagrams = -EFAULT; 2891 2892 return datagrams; 2893 } 2894 2895 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 2896 unsigned int, vlen, unsigned int, flags, 2897 struct __kernel_timespec __user *, timeout) 2898 { 2899 if (flags & MSG_CMSG_COMPAT) 2900 return -EINVAL; 2901 2902 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL); 2903 } 2904 2905 #ifdef CONFIG_COMPAT_32BIT_TIME 2906 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg, 2907 unsigned int, vlen, unsigned int, flags, 2908 struct old_timespec32 __user *, timeout) 2909 { 2910 if (flags & MSG_CMSG_COMPAT) 2911 return -EINVAL; 2912 2913 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout); 2914 } 2915 #endif 2916 2917 #ifdef __ARCH_WANT_SYS_SOCKETCALL 2918 /* Argument list sizes for sys_socketcall */ 2919 #define AL(x) ((x) * sizeof(unsigned long)) 2920 static const unsigned char nargs[21] = { 2921 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 2922 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 2923 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 2924 AL(4), AL(5), AL(4) 2925 }; 2926 2927 #undef AL 2928 2929 /* 2930 * System call vectors. 2931 * 2932 * Argument checking cleaned up. Saved 20% in size. 2933 * This function doesn't need to set the kernel lock because 2934 * it is set by the callees. 2935 */ 2936 2937 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 2938 { 2939 unsigned long a[AUDITSC_ARGS]; 2940 unsigned long a0, a1; 2941 int err; 2942 unsigned int len; 2943 2944 if (call < 1 || call > SYS_SENDMMSG) 2945 return -EINVAL; 2946 call = array_index_nospec(call, SYS_SENDMMSG + 1); 2947 2948 len = nargs[call]; 2949 if (len > sizeof(a)) 2950 return -EINVAL; 2951 2952 /* copy_from_user should be SMP safe. */ 2953 if (copy_from_user(a, args, len)) 2954 return -EFAULT; 2955 2956 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); 2957 if (err) 2958 return err; 2959 2960 a0 = a[0]; 2961 a1 = a[1]; 2962 2963 switch (call) { 2964 case SYS_SOCKET: 2965 err = __sys_socket(a0, a1, a[2]); 2966 break; 2967 case SYS_BIND: 2968 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 2969 break; 2970 case SYS_CONNECT: 2971 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 2972 break; 2973 case SYS_LISTEN: 2974 err = __sys_listen(a0, a1); 2975 break; 2976 case SYS_ACCEPT: 2977 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 2978 (int __user *)a[2], 0); 2979 break; 2980 case SYS_GETSOCKNAME: 2981 err = 2982 __sys_getsockname(a0, (struct sockaddr __user *)a1, 2983 (int __user *)a[2]); 2984 break; 2985 case SYS_GETPEERNAME: 2986 err = 2987 __sys_getpeername(a0, (struct sockaddr __user *)a1, 2988 (int __user *)a[2]); 2989 break; 2990 case SYS_SOCKETPAIR: 2991 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 2992 break; 2993 case SYS_SEND: 2994 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 2995 NULL, 0); 2996 break; 2997 case SYS_SENDTO: 2998 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 2999 (struct sockaddr __user *)a[4], a[5]); 3000 break; 3001 case SYS_RECV: 3002 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 3003 NULL, NULL); 3004 break; 3005 case SYS_RECVFROM: 3006 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 3007 (struct sockaddr __user *)a[4], 3008 (int __user *)a[5]); 3009 break; 3010 case SYS_SHUTDOWN: 3011 err = __sys_shutdown(a0, a1); 3012 break; 3013 case SYS_SETSOCKOPT: 3014 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], 3015 a[4]); 3016 break; 3017 case SYS_GETSOCKOPT: 3018 err = 3019 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 3020 (int __user *)a[4]); 3021 break; 3022 case SYS_SENDMSG: 3023 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, 3024 a[2], true); 3025 break; 3026 case SYS_SENDMMSG: 3027 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], 3028 a[3], true); 3029 break; 3030 case SYS_RECVMSG: 3031 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, 3032 a[2], true); 3033 break; 3034 case SYS_RECVMMSG: 3035 if (IS_ENABLED(CONFIG_64BIT)) 3036 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 3037 a[2], a[3], 3038 (struct __kernel_timespec __user *)a[4], 3039 NULL); 3040 else 3041 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 3042 a[2], a[3], NULL, 3043 (struct old_timespec32 __user *)a[4]); 3044 break; 3045 case SYS_ACCEPT4: 3046 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 3047 (int __user *)a[2], a[3]); 3048 break; 3049 default: 3050 err = -EINVAL; 3051 break; 3052 } 3053 return err; 3054 } 3055 3056 #endif /* __ARCH_WANT_SYS_SOCKETCALL */ 3057 3058 /** 3059 * sock_register - add a socket protocol handler 3060 * @ops: description of protocol 3061 * 3062 * This function is called by a protocol handler that wants to 3063 * advertise its address family, and have it linked into the 3064 * socket interface. The value ops->family corresponds to the 3065 * socket system call protocol family. 3066 */ 3067 int sock_register(const struct net_proto_family *ops) 3068 { 3069 int err; 3070 3071 if (ops->family >= NPROTO) { 3072 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); 3073 return -ENOBUFS; 3074 } 3075 3076 spin_lock(&net_family_lock); 3077 if (rcu_dereference_protected(net_families[ops->family], 3078 lockdep_is_held(&net_family_lock))) 3079 err = -EEXIST; 3080 else { 3081 rcu_assign_pointer(net_families[ops->family], ops); 3082 err = 0; 3083 } 3084 spin_unlock(&net_family_lock); 3085 3086 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]); 3087 return err; 3088 } 3089 EXPORT_SYMBOL(sock_register); 3090 3091 /** 3092 * sock_unregister - remove a protocol handler 3093 * @family: protocol family to remove 3094 * 3095 * This function is called by a protocol handler that wants to 3096 * remove its address family, and have it unlinked from the 3097 * new socket creation. 3098 * 3099 * If protocol handler is a module, then it can use module reference 3100 * counts to protect against new references. If protocol handler is not 3101 * a module then it needs to provide its own protection in 3102 * the ops->create routine. 3103 */ 3104 void sock_unregister(int family) 3105 { 3106 BUG_ON(family < 0 || family >= NPROTO); 3107 3108 spin_lock(&net_family_lock); 3109 RCU_INIT_POINTER(net_families[family], NULL); 3110 spin_unlock(&net_family_lock); 3111 3112 synchronize_rcu(); 3113 3114 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]); 3115 } 3116 EXPORT_SYMBOL(sock_unregister); 3117 3118 bool sock_is_registered(int family) 3119 { 3120 return family < NPROTO && rcu_access_pointer(net_families[family]); 3121 } 3122 3123 static int __init sock_init(void) 3124 { 3125 int err; 3126 /* 3127 * Initialize the network sysctl infrastructure. 3128 */ 3129 err = net_sysctl_init(); 3130 if (err) 3131 goto out; 3132 3133 /* 3134 * Initialize skbuff SLAB cache 3135 */ 3136 skb_init(); 3137 3138 /* 3139 * Initialize the protocols module. 3140 */ 3141 3142 init_inodecache(); 3143 3144 err = register_filesystem(&sock_fs_type); 3145 if (err) 3146 goto out; 3147 sock_mnt = kern_mount(&sock_fs_type); 3148 if (IS_ERR(sock_mnt)) { 3149 err = PTR_ERR(sock_mnt); 3150 goto out_mount; 3151 } 3152 3153 /* The real protocol initialization is performed in later initcalls. 3154 */ 3155 3156 #ifdef CONFIG_NETFILTER 3157 err = netfilter_init(); 3158 if (err) 3159 goto out; 3160 #endif 3161 3162 ptp_classifier_init(); 3163 3164 out: 3165 return err; 3166 3167 out_mount: 3168 unregister_filesystem(&sock_fs_type); 3169 goto out; 3170 } 3171 3172 core_initcall(sock_init); /* early initcall */ 3173 3174 #ifdef CONFIG_PROC_FS 3175 void socket_seq_show(struct seq_file *seq) 3176 { 3177 seq_printf(seq, "sockets: used %d\n", 3178 sock_inuse_get(seq->private)); 3179 } 3180 #endif /* CONFIG_PROC_FS */ 3181 3182 /* Handle the fact that while struct ifreq has the same *layout* on 3183 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data, 3184 * which are handled elsewhere, it still has different *size* due to 3185 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit, 3186 * resulting in struct ifreq being 32 and 40 bytes respectively). 3187 * As a result, if the struct happens to be at the end of a page and 3188 * the next page isn't readable/writable, we get a fault. To prevent 3189 * that, copy back and forth to the full size. 3190 */ 3191 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg) 3192 { 3193 if (in_compat_syscall()) { 3194 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr; 3195 3196 memset(ifr, 0, sizeof(*ifr)); 3197 if (copy_from_user(ifr32, arg, sizeof(*ifr32))) 3198 return -EFAULT; 3199 3200 if (ifrdata) 3201 *ifrdata = compat_ptr(ifr32->ifr_data); 3202 3203 return 0; 3204 } 3205 3206 if (copy_from_user(ifr, arg, sizeof(*ifr))) 3207 return -EFAULT; 3208 3209 if (ifrdata) 3210 *ifrdata = ifr->ifr_data; 3211 3212 return 0; 3213 } 3214 EXPORT_SYMBOL(get_user_ifreq); 3215 3216 int put_user_ifreq(struct ifreq *ifr, void __user *arg) 3217 { 3218 size_t size = sizeof(*ifr); 3219 3220 if (in_compat_syscall()) 3221 size = sizeof(struct compat_ifreq); 3222 3223 if (copy_to_user(arg, ifr, size)) 3224 return -EFAULT; 3225 3226 return 0; 3227 } 3228 EXPORT_SYMBOL(put_user_ifreq); 3229 3230 #ifdef CONFIG_COMPAT 3231 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 3232 { 3233 compat_uptr_t uptr32; 3234 struct ifreq ifr; 3235 void __user *saved; 3236 int err; 3237 3238 if (get_user_ifreq(&ifr, NULL, uifr32)) 3239 return -EFAULT; 3240 3241 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 3242 return -EFAULT; 3243 3244 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; 3245 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); 3246 3247 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL); 3248 if (!err) { 3249 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; 3250 if (put_user_ifreq(&ifr, uifr32)) 3251 err = -EFAULT; 3252 } 3253 return err; 3254 } 3255 3256 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ 3257 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, 3258 struct compat_ifreq __user *u_ifreq32) 3259 { 3260 struct ifreq ifreq; 3261 void __user *data; 3262 3263 if (!is_socket_ioctl_cmd(cmd)) 3264 return -ENOTTY; 3265 if (get_user_ifreq(&ifreq, &data, u_ifreq32)) 3266 return -EFAULT; 3267 ifreq.ifr_data = data; 3268 3269 return dev_ioctl(net, cmd, &ifreq, data, NULL); 3270 } 3271 3272 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 3273 unsigned int cmd, unsigned long arg) 3274 { 3275 void __user *argp = compat_ptr(arg); 3276 struct sock *sk = sock->sk; 3277 struct net *net = sock_net(sk); 3278 3279 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 3280 return sock_ioctl(file, cmd, (unsigned long)argp); 3281 3282 switch (cmd) { 3283 case SIOCWANDEV: 3284 return compat_siocwandev(net, argp); 3285 case SIOCGSTAMP_OLD: 3286 case SIOCGSTAMPNS_OLD: 3287 if (!sock->ops->gettstamp) 3288 return -ENOIOCTLCMD; 3289 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD, 3290 !COMPAT_USE_64BIT_TIME); 3291 3292 case SIOCETHTOOL: 3293 case SIOCBONDSLAVEINFOQUERY: 3294 case SIOCBONDINFOQUERY: 3295 case SIOCSHWTSTAMP: 3296 case SIOCGHWTSTAMP: 3297 return compat_ifr_data_ioctl(net, cmd, argp); 3298 3299 case FIOSETOWN: 3300 case SIOCSPGRP: 3301 case FIOGETOWN: 3302 case SIOCGPGRP: 3303 case SIOCBRADDBR: 3304 case SIOCBRDELBR: 3305 case SIOCGIFVLAN: 3306 case SIOCSIFVLAN: 3307 case SIOCGSKNS: 3308 case SIOCGSTAMP_NEW: 3309 case SIOCGSTAMPNS_NEW: 3310 case SIOCGIFCONF: 3311 case SIOCSIFBR: 3312 case SIOCGIFBR: 3313 return sock_ioctl(file, cmd, arg); 3314 3315 case SIOCGIFFLAGS: 3316 case SIOCSIFFLAGS: 3317 case SIOCGIFMAP: 3318 case SIOCSIFMAP: 3319 case SIOCGIFMETRIC: 3320 case SIOCSIFMETRIC: 3321 case SIOCGIFMTU: 3322 case SIOCSIFMTU: 3323 case SIOCGIFMEM: 3324 case SIOCSIFMEM: 3325 case SIOCGIFHWADDR: 3326 case SIOCSIFHWADDR: 3327 case SIOCADDMULTI: 3328 case SIOCDELMULTI: 3329 case SIOCGIFINDEX: 3330 case SIOCGIFADDR: 3331 case SIOCSIFADDR: 3332 case SIOCSIFHWBROADCAST: 3333 case SIOCDIFADDR: 3334 case SIOCGIFBRDADDR: 3335 case SIOCSIFBRDADDR: 3336 case SIOCGIFDSTADDR: 3337 case SIOCSIFDSTADDR: 3338 case SIOCGIFNETMASK: 3339 case SIOCSIFNETMASK: 3340 case SIOCSIFPFLAGS: 3341 case SIOCGIFPFLAGS: 3342 case SIOCGIFTXQLEN: 3343 case SIOCSIFTXQLEN: 3344 case SIOCBRADDIF: 3345 case SIOCBRDELIF: 3346 case SIOCGIFNAME: 3347 case SIOCSIFNAME: 3348 case SIOCGMIIPHY: 3349 case SIOCGMIIREG: 3350 case SIOCSMIIREG: 3351 case SIOCBONDENSLAVE: 3352 case SIOCBONDRELEASE: 3353 case SIOCBONDSETHWADDR: 3354 case SIOCBONDCHANGEACTIVE: 3355 case SIOCSARP: 3356 case SIOCGARP: 3357 case SIOCDARP: 3358 case SIOCOUTQ: 3359 case SIOCOUTQNSD: 3360 case SIOCATMARK: 3361 return sock_do_ioctl(net, sock, cmd, arg); 3362 } 3363 3364 return -ENOIOCTLCMD; 3365 } 3366 3367 static long compat_sock_ioctl(struct file *file, unsigned int cmd, 3368 unsigned long arg) 3369 { 3370 struct socket *sock = file->private_data; 3371 int ret = -ENOIOCTLCMD; 3372 struct sock *sk; 3373 struct net *net; 3374 3375 sk = sock->sk; 3376 net = sock_net(sk); 3377 3378 if (sock->ops->compat_ioctl) 3379 ret = sock->ops->compat_ioctl(sock, cmd, arg); 3380 3381 if (ret == -ENOIOCTLCMD && 3382 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 3383 ret = compat_wext_handle_ioctl(net, cmd, arg); 3384 3385 if (ret == -ENOIOCTLCMD) 3386 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 3387 3388 return ret; 3389 } 3390 #endif 3391 3392 /** 3393 * kernel_bind - bind an address to a socket (kernel space) 3394 * @sock: socket 3395 * @addr: address 3396 * @addrlen: length of address 3397 * 3398 * Returns 0 or an error. 3399 */ 3400 3401 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3402 { 3403 return sock->ops->bind(sock, addr, addrlen); 3404 } 3405 EXPORT_SYMBOL(kernel_bind); 3406 3407 /** 3408 * kernel_listen - move socket to listening state (kernel space) 3409 * @sock: socket 3410 * @backlog: pending connections queue size 3411 * 3412 * Returns 0 or an error. 3413 */ 3414 3415 int kernel_listen(struct socket *sock, int backlog) 3416 { 3417 return sock->ops->listen(sock, backlog); 3418 } 3419 EXPORT_SYMBOL(kernel_listen); 3420 3421 /** 3422 * kernel_accept - accept a connection (kernel space) 3423 * @sock: listening socket 3424 * @newsock: new connected socket 3425 * @flags: flags 3426 * 3427 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0. 3428 * If it fails, @newsock is guaranteed to be %NULL. 3429 * Returns 0 or an error. 3430 */ 3431 3432 int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3433 { 3434 struct sock *sk = sock->sk; 3435 int err; 3436 3437 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3438 newsock); 3439 if (err < 0) 3440 goto done; 3441 3442 err = sock->ops->accept(sock, *newsock, flags, true); 3443 if (err < 0) { 3444 sock_release(*newsock); 3445 *newsock = NULL; 3446 goto done; 3447 } 3448 3449 (*newsock)->ops = sock->ops; 3450 __module_get((*newsock)->ops->owner); 3451 3452 done: 3453 return err; 3454 } 3455 EXPORT_SYMBOL(kernel_accept); 3456 3457 /** 3458 * kernel_connect - connect a socket (kernel space) 3459 * @sock: socket 3460 * @addr: address 3461 * @addrlen: address length 3462 * @flags: flags (O_NONBLOCK, ...) 3463 * 3464 * For datagram sockets, @addr is the address to which datagrams are sent 3465 * by default, and the only address from which datagrams are received. 3466 * For stream sockets, attempts to connect to @addr. 3467 * Returns 0 or an error code. 3468 */ 3469 3470 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3471 int flags) 3472 { 3473 return sock->ops->connect(sock, addr, addrlen, flags); 3474 } 3475 EXPORT_SYMBOL(kernel_connect); 3476 3477 /** 3478 * kernel_getsockname - get the address which the socket is bound (kernel space) 3479 * @sock: socket 3480 * @addr: address holder 3481 * 3482 * Fills the @addr pointer with the address which the socket is bound. 3483 * Returns the length of the address in bytes or an error code. 3484 */ 3485 3486 int kernel_getsockname(struct socket *sock, struct sockaddr *addr) 3487 { 3488 return sock->ops->getname(sock, addr, 0); 3489 } 3490 EXPORT_SYMBOL(kernel_getsockname); 3491 3492 /** 3493 * kernel_getpeername - get the address which the socket is connected (kernel space) 3494 * @sock: socket 3495 * @addr: address holder 3496 * 3497 * Fills the @addr pointer with the address which the socket is connected. 3498 * Returns the length of the address in bytes or an error code. 3499 */ 3500 3501 int kernel_getpeername(struct socket *sock, struct sockaddr *addr) 3502 { 3503 return sock->ops->getname(sock, addr, 1); 3504 } 3505 EXPORT_SYMBOL(kernel_getpeername); 3506 3507 /** 3508 * kernel_sendpage - send a &page through a socket (kernel space) 3509 * @sock: socket 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 */ 3517 3518 int kernel_sendpage(struct socket *sock, struct page *page, int offset, 3519 size_t size, int flags) 3520 { 3521 if (sock->ops->sendpage) { 3522 /* Warn in case the improper page to zero-copy send */ 3523 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send"); 3524 return sock->ops->sendpage(sock, page, offset, size, flags); 3525 } 3526 return sock_no_sendpage(sock, page, offset, size, flags); 3527 } 3528 EXPORT_SYMBOL(kernel_sendpage); 3529 3530 /** 3531 * kernel_sendpage_locked - send a &page through the locked sock (kernel space) 3532 * @sk: sock 3533 * @page: page 3534 * @offset: page offset 3535 * @size: total size in bytes 3536 * @flags: flags (MSG_DONTWAIT, ...) 3537 * 3538 * Returns the total amount sent in bytes or an error. 3539 * Caller must hold @sk. 3540 */ 3541 3542 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset, 3543 size_t size, int flags) 3544 { 3545 struct socket *sock = sk->sk_socket; 3546 3547 if (sock->ops->sendpage_locked) 3548 return sock->ops->sendpage_locked(sk, page, offset, size, 3549 flags); 3550 3551 return sock_no_sendpage_locked(sk, page, offset, size, flags); 3552 } 3553 EXPORT_SYMBOL(kernel_sendpage_locked); 3554 3555 /** 3556 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space) 3557 * @sock: socket 3558 * @how: connection part 3559 * 3560 * Returns 0 or an error. 3561 */ 3562 3563 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3564 { 3565 return sock->ops->shutdown(sock, how); 3566 } 3567 EXPORT_SYMBOL(kernel_sock_shutdown); 3568 3569 /** 3570 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket 3571 * @sk: socket 3572 * 3573 * This routine returns the IP overhead imposed by a socket i.e. 3574 * the length of the underlying IP header, depending on whether 3575 * this is an IPv4 or IPv6 socket and the length from IP options turned 3576 * on at the socket. Assumes that the caller has a lock on the socket. 3577 */ 3578 3579 u32 kernel_sock_ip_overhead(struct sock *sk) 3580 { 3581 struct inet_sock *inet; 3582 struct ip_options_rcu *opt; 3583 u32 overhead = 0; 3584 #if IS_ENABLED(CONFIG_IPV6) 3585 struct ipv6_pinfo *np; 3586 struct ipv6_txoptions *optv6 = NULL; 3587 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3588 3589 if (!sk) 3590 return overhead; 3591 3592 switch (sk->sk_family) { 3593 case AF_INET: 3594 inet = inet_sk(sk); 3595 overhead += sizeof(struct iphdr); 3596 opt = rcu_dereference_protected(inet->inet_opt, 3597 sock_owned_by_user(sk)); 3598 if (opt) 3599 overhead += opt->opt.optlen; 3600 return overhead; 3601 #if IS_ENABLED(CONFIG_IPV6) 3602 case AF_INET6: 3603 np = inet6_sk(sk); 3604 overhead += sizeof(struct ipv6hdr); 3605 if (np) 3606 optv6 = rcu_dereference_protected(np->opt, 3607 sock_owned_by_user(sk)); 3608 if (optv6) 3609 overhead += (optv6->opt_flen + optv6->opt_nflen); 3610 return overhead; 3611 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3612 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ 3613 return overhead; 3614 } 3615 } 3616 EXPORT_SYMBOL(kernel_sock_ip_overhead); 3617