1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * NET An implementation of the SOCKET network access protocol. 4 * 5 * Version: @(#)socket.c 1.1.93 18/02/95 6 * 7 * Authors: Orest Zborowski, <obz@Kodak.COM> 8 * Ross Biro 9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 10 * 11 * Fixes: 12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in 13 * shutdown() 14 * Alan Cox : verify_area() fixes 15 * Alan Cox : Removed DDI 16 * Jonathan Kamens : SOCK_DGRAM reconnect bug 17 * Alan Cox : Moved a load of checks to the very 18 * top level. 19 * Alan Cox : Move address structures to/from user 20 * mode above the protocol layers. 21 * Rob Janssen : Allow 0 length sends. 22 * Alan Cox : Asynchronous I/O support (cribbed from the 23 * tty drivers). 24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style) 25 * Jeff Uphoff : Made max number of sockets command-line 26 * configurable. 27 * Matti Aarnio : Made the number of sockets dynamic, 28 * to be allocated when needed, and mr. 29 * Uphoff's max is used as max to be 30 * allowed to allocate. 31 * Linus : Argh. removed all the socket allocation 32 * altogether: it's in the inode now. 33 * Alan Cox : Made sock_alloc()/sock_release() public 34 * for NetROM and future kernel nfsd type 35 * stuff. 36 * Alan Cox : sendmsg/recvmsg basics. 37 * Tom Dyas : Export net symbols. 38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n". 39 * Alan Cox : Added thread locking to sys_* calls 40 * for sockets. May have errors at the 41 * moment. 42 * Kevin Buhr : Fixed the dumb errors in the above. 43 * Andi Kleen : Some small cleanups, optimizations, 44 * and fixed a copy_from_user() bug. 45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0) 46 * Tigran Aivazian : Made listen(2) backlog sanity checks 47 * protocol-independent 48 * 49 * This module is effectively the top level interface to the BSD socket 50 * paradigm. 51 * 52 * Based upon Swansea University Computer Society NET3.039 53 */ 54 55 #include <linux/bpf-cgroup.h> 56 #include <linux/ethtool.h> 57 #include <linux/mm.h> 58 #include <linux/socket.h> 59 #include <linux/file.h> 60 #include <linux/net.h> 61 #include <linux/interrupt.h> 62 #include <linux/thread_info.h> 63 #include <linux/rcupdate.h> 64 #include <linux/netdevice.h> 65 #include <linux/proc_fs.h> 66 #include <linux/seq_file.h> 67 #include <linux/mutex.h> 68 #include <linux/if_bridge.h> 69 #include <linux/if_vlan.h> 70 #include <linux/ptp_classify.h> 71 #include <linux/init.h> 72 #include <linux/poll.h> 73 #include <linux/cache.h> 74 #include <linux/module.h> 75 #include <linux/highmem.h> 76 #include <linux/mount.h> 77 #include <linux/pseudo_fs.h> 78 #include <linux/security.h> 79 #include <linux/syscalls.h> 80 #include <linux/compat.h> 81 #include <linux/kmod.h> 82 #include <linux/audit.h> 83 #include <linux/wireless.h> 84 #include <linux/nsproxy.h> 85 #include <linux/magic.h> 86 #include <linux/slab.h> 87 #include <linux/xattr.h> 88 #include <linux/nospec.h> 89 #include <linux/indirect_call_wrapper.h> 90 91 #include <linux/uaccess.h> 92 #include <asm/unistd.h> 93 94 #include <net/compat.h> 95 #include <net/wext.h> 96 #include <net/cls_cgroup.h> 97 98 #include <net/sock.h> 99 #include <linux/netfilter.h> 100 101 #include <linux/if_tun.h> 102 #include <linux/ipv6_route.h> 103 #include <linux/route.h> 104 #include <linux/termios.h> 105 #include <linux/sockios.h> 106 #include <net/busy_poll.h> 107 #include <linux/errqueue.h> 108 #include <linux/ptp_clock_kernel.h> 109 110 #ifdef CONFIG_NET_RX_BUSY_POLL 111 unsigned int sysctl_net_busy_read __read_mostly; 112 unsigned int sysctl_net_busy_poll __read_mostly; 113 #endif 114 115 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to); 116 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from); 117 static int sock_mmap(struct file *file, struct vm_area_struct *vma); 118 119 static int sock_close(struct inode *inode, struct file *file); 120 static __poll_t sock_poll(struct file *file, 121 struct poll_table_struct *wait); 122 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 123 #ifdef CONFIG_COMPAT 124 static long compat_sock_ioctl(struct file *file, 125 unsigned int cmd, unsigned long arg); 126 #endif 127 static int sock_fasync(int fd, struct file *filp, int on); 128 static ssize_t sock_sendpage(struct file *file, struct page *page, 129 int offset, size_t size, loff_t *ppos, int more); 130 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 131 struct pipe_inode_info *pipe, size_t len, 132 unsigned int flags); 133 134 #ifdef CONFIG_PROC_FS 135 static void sock_show_fdinfo(struct seq_file *m, struct file *f) 136 { 137 struct socket *sock = f->private_data; 138 139 if (sock->ops->show_fdinfo) 140 sock->ops->show_fdinfo(m, sock); 141 } 142 #else 143 #define sock_show_fdinfo NULL 144 #endif 145 146 /* 147 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear 148 * in the operation structures but are done directly via the socketcall() multiplexor. 149 */ 150 151 static const struct file_operations socket_file_ops = { 152 .owner = THIS_MODULE, 153 .llseek = no_llseek, 154 .read_iter = sock_read_iter, 155 .write_iter = sock_write_iter, 156 .poll = sock_poll, 157 .unlocked_ioctl = sock_ioctl, 158 #ifdef CONFIG_COMPAT 159 .compat_ioctl = compat_sock_ioctl, 160 #endif 161 .mmap = sock_mmap, 162 .release = sock_close, 163 .fasync = sock_fasync, 164 .sendpage = sock_sendpage, 165 .splice_write = generic_splice_sendpage, 166 .splice_read = sock_splice_read, 167 .show_fdinfo = sock_show_fdinfo, 168 }; 169 170 static const char * const pf_family_names[] = { 171 [PF_UNSPEC] = "PF_UNSPEC", 172 [PF_UNIX] = "PF_UNIX/PF_LOCAL", 173 [PF_INET] = "PF_INET", 174 [PF_AX25] = "PF_AX25", 175 [PF_IPX] = "PF_IPX", 176 [PF_APPLETALK] = "PF_APPLETALK", 177 [PF_NETROM] = "PF_NETROM", 178 [PF_BRIDGE] = "PF_BRIDGE", 179 [PF_ATMPVC] = "PF_ATMPVC", 180 [PF_X25] = "PF_X25", 181 [PF_INET6] = "PF_INET6", 182 [PF_ROSE] = "PF_ROSE", 183 [PF_DECnet] = "PF_DECnet", 184 [PF_NETBEUI] = "PF_NETBEUI", 185 [PF_SECURITY] = "PF_SECURITY", 186 [PF_KEY] = "PF_KEY", 187 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE", 188 [PF_PACKET] = "PF_PACKET", 189 [PF_ASH] = "PF_ASH", 190 [PF_ECONET] = "PF_ECONET", 191 [PF_ATMSVC] = "PF_ATMSVC", 192 [PF_RDS] = "PF_RDS", 193 [PF_SNA] = "PF_SNA", 194 [PF_IRDA] = "PF_IRDA", 195 [PF_PPPOX] = "PF_PPPOX", 196 [PF_WANPIPE] = "PF_WANPIPE", 197 [PF_LLC] = "PF_LLC", 198 [PF_IB] = "PF_IB", 199 [PF_MPLS] = "PF_MPLS", 200 [PF_CAN] = "PF_CAN", 201 [PF_TIPC] = "PF_TIPC", 202 [PF_BLUETOOTH] = "PF_BLUETOOTH", 203 [PF_IUCV] = "PF_IUCV", 204 [PF_RXRPC] = "PF_RXRPC", 205 [PF_ISDN] = "PF_ISDN", 206 [PF_PHONET] = "PF_PHONET", 207 [PF_IEEE802154] = "PF_IEEE802154", 208 [PF_CAIF] = "PF_CAIF", 209 [PF_ALG] = "PF_ALG", 210 [PF_NFC] = "PF_NFC", 211 [PF_VSOCK] = "PF_VSOCK", 212 [PF_KCM] = "PF_KCM", 213 [PF_QIPCRTR] = "PF_QIPCRTR", 214 [PF_SMC] = "PF_SMC", 215 [PF_XDP] = "PF_XDP", 216 [PF_MCTP] = "PF_MCTP", 217 }; 218 219 /* 220 * The protocol list. Each protocol is registered in here. 221 */ 222 223 static DEFINE_SPINLOCK(net_family_lock); 224 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; 225 226 /* 227 * Support routines. 228 * Move socket addresses back and forth across the kernel/user 229 * divide and look after the messy bits. 230 */ 231 232 /** 233 * move_addr_to_kernel - copy a socket address into kernel space 234 * @uaddr: Address in user space 235 * @kaddr: Address in kernel space 236 * @ulen: Length in user space 237 * 238 * The address is copied into kernel space. If the provided address is 239 * too long an error code of -EINVAL is returned. If the copy gives 240 * invalid addresses -EFAULT is returned. On a success 0 is returned. 241 */ 242 243 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr) 244 { 245 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) 246 return -EINVAL; 247 if (ulen == 0) 248 return 0; 249 if (copy_from_user(kaddr, uaddr, ulen)) 250 return -EFAULT; 251 return audit_sockaddr(ulen, kaddr); 252 } 253 254 /** 255 * move_addr_to_user - copy an address to user space 256 * @kaddr: kernel space address 257 * @klen: length of address in kernel 258 * @uaddr: user space address 259 * @ulen: pointer to user length field 260 * 261 * The value pointed to by ulen on entry is the buffer length available. 262 * This is overwritten with the buffer space used. -EINVAL is returned 263 * if an overlong buffer is specified or a negative buffer size. -EFAULT 264 * is returned if either the buffer or the length field are not 265 * accessible. 266 * After copying the data up to the limit the user specifies, the true 267 * length of the data is written over the length limit the user 268 * specified. Zero is returned for a success. 269 */ 270 271 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen, 272 void __user *uaddr, int __user *ulen) 273 { 274 int err; 275 int len; 276 277 BUG_ON(klen > sizeof(struct sockaddr_storage)); 278 err = get_user(len, ulen); 279 if (err) 280 return err; 281 if (len > klen) 282 len = klen; 283 if (len < 0) 284 return -EINVAL; 285 if (len) { 286 if (audit_sockaddr(klen, kaddr)) 287 return -ENOMEM; 288 if (copy_to_user(uaddr, kaddr, len)) 289 return -EFAULT; 290 } 291 /* 292 * "fromlen shall refer to the value before truncation.." 293 * 1003.1g 294 */ 295 return __put_user(klen, ulen); 296 } 297 298 static struct kmem_cache *sock_inode_cachep __ro_after_init; 299 300 static struct inode *sock_alloc_inode(struct super_block *sb) 301 { 302 struct socket_alloc *ei; 303 304 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL); 305 if (!ei) 306 return NULL; 307 init_waitqueue_head(&ei->socket.wq.wait); 308 ei->socket.wq.fasync_list = NULL; 309 ei->socket.wq.flags = 0; 310 311 ei->socket.state = SS_UNCONNECTED; 312 ei->socket.flags = 0; 313 ei->socket.ops = NULL; 314 ei->socket.sk = NULL; 315 ei->socket.file = NULL; 316 317 return &ei->vfs_inode; 318 } 319 320 static void sock_free_inode(struct inode *inode) 321 { 322 struct socket_alloc *ei; 323 324 ei = container_of(inode, struct socket_alloc, vfs_inode); 325 kmem_cache_free(sock_inode_cachep, ei); 326 } 327 328 static void init_once(void *foo) 329 { 330 struct socket_alloc *ei = (struct socket_alloc *)foo; 331 332 inode_init_once(&ei->vfs_inode); 333 } 334 335 static void init_inodecache(void) 336 { 337 sock_inode_cachep = kmem_cache_create("sock_inode_cache", 338 sizeof(struct socket_alloc), 339 0, 340 (SLAB_HWCACHE_ALIGN | 341 SLAB_RECLAIM_ACCOUNT | 342 SLAB_MEM_SPREAD | SLAB_ACCOUNT), 343 init_once); 344 BUG_ON(sock_inode_cachep == NULL); 345 } 346 347 static const struct super_operations sockfs_ops = { 348 .alloc_inode = sock_alloc_inode, 349 .free_inode = sock_free_inode, 350 .statfs = simple_statfs, 351 }; 352 353 /* 354 * sockfs_dname() is called from d_path(). 355 */ 356 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) 357 { 358 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", 359 d_inode(dentry)->i_ino); 360 } 361 362 static const struct dentry_operations sockfs_dentry_operations = { 363 .d_dname = sockfs_dname, 364 }; 365 366 static int sockfs_xattr_get(const struct xattr_handler *handler, 367 struct dentry *dentry, struct inode *inode, 368 const char *suffix, void *value, size_t size) 369 { 370 if (value) { 371 if (dentry->d_name.len + 1 > size) 372 return -ERANGE; 373 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1); 374 } 375 return dentry->d_name.len + 1; 376 } 377 378 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname" 379 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX) 380 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1) 381 382 static const struct xattr_handler sockfs_xattr_handler = { 383 .name = XATTR_NAME_SOCKPROTONAME, 384 .get = sockfs_xattr_get, 385 }; 386 387 static int sockfs_security_xattr_set(const struct xattr_handler *handler, 388 struct user_namespace *mnt_userns, 389 struct dentry *dentry, struct inode *inode, 390 const char *suffix, const void *value, 391 size_t size, int flags) 392 { 393 /* Handled by LSM. */ 394 return -EAGAIN; 395 } 396 397 static const struct xattr_handler sockfs_security_xattr_handler = { 398 .prefix = XATTR_SECURITY_PREFIX, 399 .set = sockfs_security_xattr_set, 400 }; 401 402 static const struct xattr_handler *sockfs_xattr_handlers[] = { 403 &sockfs_xattr_handler, 404 &sockfs_security_xattr_handler, 405 NULL 406 }; 407 408 static int sockfs_init_fs_context(struct fs_context *fc) 409 { 410 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC); 411 if (!ctx) 412 return -ENOMEM; 413 ctx->ops = &sockfs_ops; 414 ctx->dops = &sockfs_dentry_operations; 415 ctx->xattr = sockfs_xattr_handlers; 416 return 0; 417 } 418 419 static struct vfsmount *sock_mnt __read_mostly; 420 421 static struct file_system_type sock_fs_type = { 422 .name = "sockfs", 423 .init_fs_context = sockfs_init_fs_context, 424 .kill_sb = kill_anon_super, 425 }; 426 427 /* 428 * Obtains the first available file descriptor and sets it up for use. 429 * 430 * These functions create file structures and maps them to fd space 431 * of the current process. On success it returns file descriptor 432 * and file struct implicitly stored in sock->file. 433 * Note that another thread may close file descriptor before we return 434 * from this function. We use the fact that now we do not refer 435 * to socket after mapping. If one day we will need it, this 436 * function will increment ref. count on file by 1. 437 * 438 * In any case returned fd MAY BE not valid! 439 * This race condition is unavoidable 440 * with shared fd spaces, we cannot solve it inside kernel, 441 * but we take care of internal coherence yet. 442 */ 443 444 /** 445 * sock_alloc_file - Bind a &socket to a &file 446 * @sock: socket 447 * @flags: file status flags 448 * @dname: protocol name 449 * 450 * Returns the &file bound with @sock, implicitly storing it 451 * in sock->file. If dname is %NULL, sets to "". 452 * On failure the return is a ERR pointer (see linux/err.h). 453 * This function uses GFP_KERNEL internally. 454 */ 455 456 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname) 457 { 458 struct file *file; 459 460 if (!dname) 461 dname = sock->sk ? sock->sk->sk_prot_creator->name : ""; 462 463 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname, 464 O_RDWR | (flags & O_NONBLOCK), 465 &socket_file_ops); 466 if (IS_ERR(file)) { 467 sock_release(sock); 468 return file; 469 } 470 471 sock->file = file; 472 file->private_data = sock; 473 stream_open(SOCK_INODE(sock), file); 474 return file; 475 } 476 EXPORT_SYMBOL(sock_alloc_file); 477 478 static int sock_map_fd(struct socket *sock, int flags) 479 { 480 struct file *newfile; 481 int fd = get_unused_fd_flags(flags); 482 if (unlikely(fd < 0)) { 483 sock_release(sock); 484 return fd; 485 } 486 487 newfile = sock_alloc_file(sock, flags, NULL); 488 if (!IS_ERR(newfile)) { 489 fd_install(fd, newfile); 490 return fd; 491 } 492 493 put_unused_fd(fd); 494 return PTR_ERR(newfile); 495 } 496 497 /** 498 * sock_from_file - Return the &socket bounded to @file. 499 * @file: file 500 * 501 * On failure returns %NULL. 502 */ 503 504 struct socket *sock_from_file(struct file *file) 505 { 506 if (file->f_op == &socket_file_ops) 507 return file->private_data; /* set in sock_map_fd */ 508 509 return NULL; 510 } 511 EXPORT_SYMBOL(sock_from_file); 512 513 /** 514 * sockfd_lookup - Go from a file number to its socket slot 515 * @fd: file handle 516 * @err: pointer to an error code return 517 * 518 * The file handle passed in is locked and the socket it is bound 519 * to is returned. If an error occurs the err pointer is overwritten 520 * with a negative errno code and NULL is returned. The function checks 521 * for both invalid handles and passing a handle which is not a socket. 522 * 523 * On a success the socket object pointer is returned. 524 */ 525 526 struct socket *sockfd_lookup(int fd, int *err) 527 { 528 struct file *file; 529 struct socket *sock; 530 531 file = fget(fd); 532 if (!file) { 533 *err = -EBADF; 534 return NULL; 535 } 536 537 sock = sock_from_file(file); 538 if (!sock) { 539 *err = -ENOTSOCK; 540 fput(file); 541 } 542 return sock; 543 } 544 EXPORT_SYMBOL(sockfd_lookup); 545 546 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) 547 { 548 struct fd f = fdget(fd); 549 struct socket *sock; 550 551 *err = -EBADF; 552 if (f.file) { 553 sock = sock_from_file(f.file); 554 if (likely(sock)) { 555 *fput_needed = f.flags & FDPUT_FPUT; 556 return sock; 557 } 558 *err = -ENOTSOCK; 559 fdput(f); 560 } 561 return NULL; 562 } 563 564 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer, 565 size_t size) 566 { 567 ssize_t len; 568 ssize_t used = 0; 569 570 len = security_inode_listsecurity(d_inode(dentry), buffer, size); 571 if (len < 0) 572 return len; 573 used += len; 574 if (buffer) { 575 if (size < used) 576 return -ERANGE; 577 buffer += len; 578 } 579 580 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1); 581 used += len; 582 if (buffer) { 583 if (size < used) 584 return -ERANGE; 585 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len); 586 buffer += len; 587 } 588 589 return used; 590 } 591 592 static int sockfs_setattr(struct user_namespace *mnt_userns, 593 struct dentry *dentry, struct iattr *iattr) 594 { 595 int err = simple_setattr(&init_user_ns, dentry, iattr); 596 597 if (!err && (iattr->ia_valid & ATTR_UID)) { 598 struct socket *sock = SOCKET_I(d_inode(dentry)); 599 600 if (sock->sk) 601 sock->sk->sk_uid = iattr->ia_uid; 602 else 603 err = -ENOENT; 604 } 605 606 return err; 607 } 608 609 static const struct inode_operations sockfs_inode_ops = { 610 .listxattr = sockfs_listxattr, 611 .setattr = sockfs_setattr, 612 }; 613 614 /** 615 * sock_alloc - allocate a socket 616 * 617 * Allocate a new inode and socket object. The two are bound together 618 * and initialised. The socket is then returned. If we are out of inodes 619 * NULL is returned. This functions uses GFP_KERNEL internally. 620 */ 621 622 struct socket *sock_alloc(void) 623 { 624 struct inode *inode; 625 struct socket *sock; 626 627 inode = new_inode_pseudo(sock_mnt->mnt_sb); 628 if (!inode) 629 return NULL; 630 631 sock = SOCKET_I(inode); 632 633 inode->i_ino = get_next_ino(); 634 inode->i_mode = S_IFSOCK | S_IRWXUGO; 635 inode->i_uid = current_fsuid(); 636 inode->i_gid = current_fsgid(); 637 inode->i_op = &sockfs_inode_ops; 638 639 return sock; 640 } 641 EXPORT_SYMBOL(sock_alloc); 642 643 static void __sock_release(struct socket *sock, struct inode *inode) 644 { 645 if (sock->ops) { 646 struct module *owner = sock->ops->owner; 647 648 if (inode) 649 inode_lock(inode); 650 sock->ops->release(sock); 651 sock->sk = NULL; 652 if (inode) 653 inode_unlock(inode); 654 sock->ops = NULL; 655 module_put(owner); 656 } 657 658 if (sock->wq.fasync_list) 659 pr_err("%s: fasync list not empty!\n", __func__); 660 661 if (!sock->file) { 662 iput(SOCK_INODE(sock)); 663 return; 664 } 665 sock->file = NULL; 666 } 667 668 /** 669 * sock_release - close a socket 670 * @sock: socket to close 671 * 672 * The socket is released from the protocol stack if it has a release 673 * callback, and the inode is then released if the socket is bound to 674 * an inode not a file. 675 */ 676 void sock_release(struct socket *sock) 677 { 678 __sock_release(sock, NULL); 679 } 680 EXPORT_SYMBOL(sock_release); 681 682 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags) 683 { 684 u8 flags = *tx_flags; 685 686 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) { 687 flags |= SKBTX_HW_TSTAMP; 688 689 /* PTP hardware clocks can provide a free running cycle counter 690 * as a time base for virtual clocks. Tell driver to use the 691 * free running cycle counter for timestamp if socket is bound 692 * to virtual clock. 693 */ 694 if (tsflags & SOF_TIMESTAMPING_BIND_PHC) 695 flags |= SKBTX_HW_TSTAMP_USE_CYCLES; 696 } 697 698 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE) 699 flags |= SKBTX_SW_TSTAMP; 700 701 if (tsflags & SOF_TIMESTAMPING_TX_SCHED) 702 flags |= SKBTX_SCHED_TSTAMP; 703 704 *tx_flags = flags; 705 } 706 EXPORT_SYMBOL(__sock_tx_timestamp); 707 708 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *, 709 size_t)); 710 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *, 711 size_t)); 712 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) 713 { 714 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg, 715 inet_sendmsg, sock, msg, 716 msg_data_left(msg)); 717 BUG_ON(ret == -EIOCBQUEUED); 718 return ret; 719 } 720 721 /** 722 * sock_sendmsg - send a message through @sock 723 * @sock: socket 724 * @msg: message to send 725 * 726 * Sends @msg through @sock, passing through LSM. 727 * Returns the number of bytes sent, or an error code. 728 */ 729 int sock_sendmsg(struct socket *sock, struct msghdr *msg) 730 { 731 int err = security_socket_sendmsg(sock, msg, 732 msg_data_left(msg)); 733 734 return err ?: sock_sendmsg_nosec(sock, msg); 735 } 736 EXPORT_SYMBOL(sock_sendmsg); 737 738 /** 739 * kernel_sendmsg - send a message through @sock (kernel-space) 740 * @sock: socket 741 * @msg: message header 742 * @vec: kernel vec 743 * @num: vec array length 744 * @size: total message data size 745 * 746 * Builds the message data with @vec and sends it through @sock. 747 * Returns the number of bytes sent, or an error code. 748 */ 749 750 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, 751 struct kvec *vec, size_t num, size_t size) 752 { 753 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size); 754 return sock_sendmsg(sock, msg); 755 } 756 EXPORT_SYMBOL(kernel_sendmsg); 757 758 /** 759 * kernel_sendmsg_locked - send a message through @sock (kernel-space) 760 * @sk: sock 761 * @msg: message header 762 * @vec: output s/g array 763 * @num: output s/g array length 764 * @size: total message data size 765 * 766 * Builds the message data with @vec and sends it through @sock. 767 * Returns the number of bytes sent, or an error code. 768 * Caller must hold @sk. 769 */ 770 771 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg, 772 struct kvec *vec, size_t num, size_t size) 773 { 774 struct socket *sock = sk->sk_socket; 775 776 if (!sock->ops->sendmsg_locked) 777 return sock_no_sendmsg_locked(sk, msg, size); 778 779 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size); 780 781 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg)); 782 } 783 EXPORT_SYMBOL(kernel_sendmsg_locked); 784 785 static bool skb_is_err_queue(const struct sk_buff *skb) 786 { 787 /* pkt_type of skbs enqueued on the error queue are set to 788 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do 789 * in recvmsg, since skbs received on a local socket will never 790 * have a pkt_type of PACKET_OUTGOING. 791 */ 792 return skb->pkt_type == PACKET_OUTGOING; 793 } 794 795 /* On transmit, software and hardware timestamps are returned independently. 796 * As the two skb clones share the hardware timestamp, which may be updated 797 * before the software timestamp is received, a hardware TX timestamp may be 798 * returned only if there is no software TX timestamp. Ignore false software 799 * timestamps, which may be made in the __sock_recv_timestamp() call when the 800 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a 801 * hardware timestamp. 802 */ 803 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp) 804 { 805 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb); 806 } 807 808 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index) 809 { 810 bool cycles = sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC; 811 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); 812 struct net_device *orig_dev; 813 ktime_t hwtstamp; 814 815 rcu_read_lock(); 816 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 817 if (orig_dev) { 818 *if_index = orig_dev->ifindex; 819 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles); 820 } else { 821 hwtstamp = shhwtstamps->hwtstamp; 822 } 823 rcu_read_unlock(); 824 825 return hwtstamp; 826 } 827 828 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb, 829 int if_index) 830 { 831 struct scm_ts_pktinfo ts_pktinfo; 832 struct net_device *orig_dev; 833 834 if (!skb_mac_header_was_set(skb)) 835 return; 836 837 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo)); 838 839 if (!if_index) { 840 rcu_read_lock(); 841 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 842 if (orig_dev) 843 if_index = orig_dev->ifindex; 844 rcu_read_unlock(); 845 } 846 ts_pktinfo.if_index = if_index; 847 848 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb); 849 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO, 850 sizeof(ts_pktinfo), &ts_pktinfo); 851 } 852 853 /* 854 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) 855 */ 856 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 857 struct sk_buff *skb) 858 { 859 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); 860 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW); 861 struct scm_timestamping_internal tss; 862 863 int empty = 1, false_tstamp = 0; 864 struct skb_shared_hwtstamps *shhwtstamps = 865 skb_hwtstamps(skb); 866 int if_index; 867 ktime_t hwtstamp; 868 869 /* Race occurred between timestamp enabling and packet 870 receiving. Fill in the current time for now. */ 871 if (need_software_tstamp && skb->tstamp == 0) { 872 __net_timestamp(skb); 873 false_tstamp = 1; 874 } 875 876 if (need_software_tstamp) { 877 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { 878 if (new_tstamp) { 879 struct __kernel_sock_timeval tv; 880 881 skb_get_new_timestamp(skb, &tv); 882 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW, 883 sizeof(tv), &tv); 884 } else { 885 struct __kernel_old_timeval tv; 886 887 skb_get_timestamp(skb, &tv); 888 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD, 889 sizeof(tv), &tv); 890 } 891 } else { 892 if (new_tstamp) { 893 struct __kernel_timespec ts; 894 895 skb_get_new_timestampns(skb, &ts); 896 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW, 897 sizeof(ts), &ts); 898 } else { 899 struct __kernel_old_timespec ts; 900 901 skb_get_timestampns(skb, &ts); 902 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD, 903 sizeof(ts), &ts); 904 } 905 } 906 } 907 908 memset(&tss, 0, sizeof(tss)); 909 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) && 910 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0)) 911 empty = 0; 912 if (shhwtstamps && 913 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) && 914 !skb_is_swtx_tstamp(skb, false_tstamp)) { 915 if_index = 0; 916 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV) 917 hwtstamp = get_timestamp(sk, skb, &if_index); 918 else 919 hwtstamp = shhwtstamps->hwtstamp; 920 921 if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC) 922 hwtstamp = ptp_convert_timestamp(&hwtstamp, 923 sk->sk_bind_phc); 924 925 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) { 926 empty = 0; 927 928 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) && 929 !skb_is_err_queue(skb)) 930 put_ts_pktinfo(msg, skb, if_index); 931 } 932 } 933 if (!empty) { 934 if (sock_flag(sk, SOCK_TSTAMP_NEW)) 935 put_cmsg_scm_timestamping64(msg, &tss); 936 else 937 put_cmsg_scm_timestamping(msg, &tss); 938 939 if (skb_is_err_queue(skb) && skb->len && 940 SKB_EXT_ERR(skb)->opt_stats) 941 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS, 942 skb->len, skb->data); 943 } 944 } 945 EXPORT_SYMBOL_GPL(__sock_recv_timestamp); 946 947 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, 948 struct sk_buff *skb) 949 { 950 int ack; 951 952 if (!sock_flag(sk, SOCK_WIFI_STATUS)) 953 return; 954 if (!skb->wifi_acked_valid) 955 return; 956 957 ack = skb->wifi_acked; 958 959 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); 960 } 961 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); 962 963 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, 964 struct sk_buff *skb) 965 { 966 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount) 967 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, 968 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount); 969 } 970 971 static void sock_recv_mark(struct msghdr *msg, struct sock *sk, 972 struct sk_buff *skb) 973 { 974 if (sock_flag(sk, SOCK_RCVMARK) && skb) 975 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), 976 &skb->mark); 977 } 978 979 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk, 980 struct sk_buff *skb) 981 { 982 sock_recv_timestamp(msg, sk, skb); 983 sock_recv_drops(msg, sk, skb); 984 sock_recv_mark(msg, sk, skb); 985 } 986 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs); 987 988 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *, 989 size_t, int)); 990 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *, 991 size_t, int)); 992 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 993 int flags) 994 { 995 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg, 996 inet_recvmsg, sock, msg, msg_data_left(msg), 997 flags); 998 } 999 1000 /** 1001 * sock_recvmsg - receive a message from @sock 1002 * @sock: socket 1003 * @msg: message to receive 1004 * @flags: message flags 1005 * 1006 * Receives @msg from @sock, passing through LSM. Returns the total number 1007 * of bytes received, or an error. 1008 */ 1009 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) 1010 { 1011 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); 1012 1013 return err ?: sock_recvmsg_nosec(sock, msg, flags); 1014 } 1015 EXPORT_SYMBOL(sock_recvmsg); 1016 1017 /** 1018 * kernel_recvmsg - Receive a message from a socket (kernel space) 1019 * @sock: The socket to receive the message from 1020 * @msg: Received message 1021 * @vec: Input s/g array for message data 1022 * @num: Size of input s/g array 1023 * @size: Number of bytes to read 1024 * @flags: Message flags (MSG_DONTWAIT, etc...) 1025 * 1026 * On return the msg structure contains the scatter/gather array passed in the 1027 * vec argument. The array is modified so that it consists of the unfilled 1028 * portion of the original array. 1029 * 1030 * The returned value is the total number of bytes received, or an error. 1031 */ 1032 1033 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 1034 struct kvec *vec, size_t num, size_t size, int flags) 1035 { 1036 msg->msg_control_is_user = false; 1037 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size); 1038 return sock_recvmsg(sock, msg, flags); 1039 } 1040 EXPORT_SYMBOL(kernel_recvmsg); 1041 1042 static ssize_t sock_sendpage(struct file *file, struct page *page, 1043 int offset, size_t size, loff_t *ppos, int more) 1044 { 1045 struct socket *sock; 1046 int flags; 1047 1048 sock = file->private_data; 1049 1050 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 1051 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */ 1052 flags |= more; 1053 1054 return kernel_sendpage(sock, page, offset, size, flags); 1055 } 1056 1057 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 1058 struct pipe_inode_info *pipe, size_t len, 1059 unsigned int flags) 1060 { 1061 struct socket *sock = file->private_data; 1062 1063 if (unlikely(!sock->ops->splice_read)) 1064 return generic_file_splice_read(file, ppos, pipe, len, flags); 1065 1066 return sock->ops->splice_read(sock, ppos, pipe, len, flags); 1067 } 1068 1069 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) 1070 { 1071 struct file *file = iocb->ki_filp; 1072 struct socket *sock = file->private_data; 1073 struct msghdr msg = {.msg_iter = *to, 1074 .msg_iocb = iocb}; 1075 ssize_t res; 1076 1077 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1078 msg.msg_flags = MSG_DONTWAIT; 1079 1080 if (iocb->ki_pos != 0) 1081 return -ESPIPE; 1082 1083 if (!iov_iter_count(to)) /* Match SYS5 behaviour */ 1084 return 0; 1085 1086 res = sock_recvmsg(sock, &msg, msg.msg_flags); 1087 *to = msg.msg_iter; 1088 return res; 1089 } 1090 1091 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) 1092 { 1093 struct file *file = iocb->ki_filp; 1094 struct socket *sock = file->private_data; 1095 struct msghdr msg = {.msg_iter = *from, 1096 .msg_iocb = iocb}; 1097 ssize_t res; 1098 1099 if (iocb->ki_pos != 0) 1100 return -ESPIPE; 1101 1102 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1103 msg.msg_flags = MSG_DONTWAIT; 1104 1105 if (sock->type == SOCK_SEQPACKET) 1106 msg.msg_flags |= MSG_EOR; 1107 1108 res = sock_sendmsg(sock, &msg); 1109 *from = msg.msg_iter; 1110 return res; 1111 } 1112 1113 /* 1114 * Atomic setting of ioctl hooks to avoid race 1115 * with module unload. 1116 */ 1117 1118 static DEFINE_MUTEX(br_ioctl_mutex); 1119 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br, 1120 unsigned int cmd, struct ifreq *ifr, 1121 void __user *uarg); 1122 1123 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br, 1124 unsigned int cmd, struct ifreq *ifr, 1125 void __user *uarg)) 1126 { 1127 mutex_lock(&br_ioctl_mutex); 1128 br_ioctl_hook = hook; 1129 mutex_unlock(&br_ioctl_mutex); 1130 } 1131 EXPORT_SYMBOL(brioctl_set); 1132 1133 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd, 1134 struct ifreq *ifr, void __user *uarg) 1135 { 1136 int err = -ENOPKG; 1137 1138 if (!br_ioctl_hook) 1139 request_module("bridge"); 1140 1141 mutex_lock(&br_ioctl_mutex); 1142 if (br_ioctl_hook) 1143 err = br_ioctl_hook(net, br, cmd, ifr, uarg); 1144 mutex_unlock(&br_ioctl_mutex); 1145 1146 return err; 1147 } 1148 1149 static DEFINE_MUTEX(vlan_ioctl_mutex); 1150 static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 1151 1152 void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 1153 { 1154 mutex_lock(&vlan_ioctl_mutex); 1155 vlan_ioctl_hook = hook; 1156 mutex_unlock(&vlan_ioctl_mutex); 1157 } 1158 EXPORT_SYMBOL(vlan_ioctl_set); 1159 1160 static long sock_do_ioctl(struct net *net, struct socket *sock, 1161 unsigned int cmd, unsigned long arg) 1162 { 1163 struct ifreq ifr; 1164 bool need_copyout; 1165 int err; 1166 void __user *argp = (void __user *)arg; 1167 void __user *data; 1168 1169 err = sock->ops->ioctl(sock, cmd, arg); 1170 1171 /* 1172 * If this ioctl is unknown try to hand it down 1173 * to the NIC driver. 1174 */ 1175 if (err != -ENOIOCTLCMD) 1176 return err; 1177 1178 if (!is_socket_ioctl_cmd(cmd)) 1179 return -ENOTTY; 1180 1181 if (get_user_ifreq(&ifr, &data, argp)) 1182 return -EFAULT; 1183 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); 1184 if (!err && need_copyout) 1185 if (put_user_ifreq(&ifr, argp)) 1186 return -EFAULT; 1187 1188 return err; 1189 } 1190 1191 /* 1192 * With an ioctl, arg may well be a user mode pointer, but we don't know 1193 * what to do with it - that's up to the protocol still. 1194 */ 1195 1196 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 1197 { 1198 struct socket *sock; 1199 struct sock *sk; 1200 void __user *argp = (void __user *)arg; 1201 int pid, err; 1202 struct net *net; 1203 1204 sock = file->private_data; 1205 sk = sock->sk; 1206 net = sock_net(sk); 1207 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { 1208 struct ifreq ifr; 1209 void __user *data; 1210 bool need_copyout; 1211 if (get_user_ifreq(&ifr, &data, argp)) 1212 return -EFAULT; 1213 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); 1214 if (!err && need_copyout) 1215 if (put_user_ifreq(&ifr, argp)) 1216 return -EFAULT; 1217 } else 1218 #ifdef CONFIG_WEXT_CORE 1219 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 1220 err = wext_handle_ioctl(net, cmd, argp); 1221 } else 1222 #endif 1223 switch (cmd) { 1224 case FIOSETOWN: 1225 case SIOCSPGRP: 1226 err = -EFAULT; 1227 if (get_user(pid, (int __user *)argp)) 1228 break; 1229 err = f_setown(sock->file, pid, 1); 1230 break; 1231 case FIOGETOWN: 1232 case SIOCGPGRP: 1233 err = put_user(f_getown(sock->file), 1234 (int __user *)argp); 1235 break; 1236 case SIOCGIFBR: 1237 case SIOCSIFBR: 1238 case SIOCBRADDBR: 1239 case SIOCBRDELBR: 1240 err = br_ioctl_call(net, NULL, cmd, NULL, argp); 1241 break; 1242 case SIOCGIFVLAN: 1243 case SIOCSIFVLAN: 1244 err = -ENOPKG; 1245 if (!vlan_ioctl_hook) 1246 request_module("8021q"); 1247 1248 mutex_lock(&vlan_ioctl_mutex); 1249 if (vlan_ioctl_hook) 1250 err = vlan_ioctl_hook(net, argp); 1251 mutex_unlock(&vlan_ioctl_mutex); 1252 break; 1253 case SIOCGSKNS: 1254 err = -EPERM; 1255 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1256 break; 1257 1258 err = open_related_ns(&net->ns, get_net_ns); 1259 break; 1260 case SIOCGSTAMP_OLD: 1261 case SIOCGSTAMPNS_OLD: 1262 if (!sock->ops->gettstamp) { 1263 err = -ENOIOCTLCMD; 1264 break; 1265 } 1266 err = sock->ops->gettstamp(sock, argp, 1267 cmd == SIOCGSTAMP_OLD, 1268 !IS_ENABLED(CONFIG_64BIT)); 1269 break; 1270 case SIOCGSTAMP_NEW: 1271 case SIOCGSTAMPNS_NEW: 1272 if (!sock->ops->gettstamp) { 1273 err = -ENOIOCTLCMD; 1274 break; 1275 } 1276 err = sock->ops->gettstamp(sock, argp, 1277 cmd == SIOCGSTAMP_NEW, 1278 false); 1279 break; 1280 1281 case SIOCGIFCONF: 1282 err = dev_ifconf(net, argp); 1283 break; 1284 1285 default: 1286 err = sock_do_ioctl(net, sock, cmd, arg); 1287 break; 1288 } 1289 return err; 1290 } 1291 1292 /** 1293 * sock_create_lite - creates a socket 1294 * @family: protocol family (AF_INET, ...) 1295 * @type: communication type (SOCK_STREAM, ...) 1296 * @protocol: protocol (0, ...) 1297 * @res: new socket 1298 * 1299 * Creates a new socket and assigns it to @res, passing through LSM. 1300 * The new socket initialization is not complete, see kernel_accept(). 1301 * Returns 0 or an error. On failure @res is set to %NULL. 1302 * This function internally uses GFP_KERNEL. 1303 */ 1304 1305 int sock_create_lite(int family, int type, int protocol, struct socket **res) 1306 { 1307 int err; 1308 struct socket *sock = NULL; 1309 1310 err = security_socket_create(family, type, protocol, 1); 1311 if (err) 1312 goto out; 1313 1314 sock = sock_alloc(); 1315 if (!sock) { 1316 err = -ENOMEM; 1317 goto out; 1318 } 1319 1320 sock->type = type; 1321 err = security_socket_post_create(sock, family, type, protocol, 1); 1322 if (err) 1323 goto out_release; 1324 1325 out: 1326 *res = sock; 1327 return err; 1328 out_release: 1329 sock_release(sock); 1330 sock = NULL; 1331 goto out; 1332 } 1333 EXPORT_SYMBOL(sock_create_lite); 1334 1335 /* No kernel lock held - perfect */ 1336 static __poll_t sock_poll(struct file *file, poll_table *wait) 1337 { 1338 struct socket *sock = file->private_data; 1339 __poll_t events = poll_requested_events(wait), flag = 0; 1340 1341 if (!sock->ops->poll) 1342 return 0; 1343 1344 if (sk_can_busy_loop(sock->sk)) { 1345 /* poll once if requested by the syscall */ 1346 if (events & POLL_BUSY_LOOP) 1347 sk_busy_loop(sock->sk, 1); 1348 1349 /* if this socket can poll_ll, tell the system call */ 1350 flag = POLL_BUSY_LOOP; 1351 } 1352 1353 return sock->ops->poll(file, sock, wait) | flag; 1354 } 1355 1356 static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1357 { 1358 struct socket *sock = file->private_data; 1359 1360 return sock->ops->mmap(file, sock, vma); 1361 } 1362 1363 static int sock_close(struct inode *inode, struct file *filp) 1364 { 1365 __sock_release(SOCKET_I(inode), inode); 1366 return 0; 1367 } 1368 1369 /* 1370 * Update the socket async list 1371 * 1372 * Fasync_list locking strategy. 1373 * 1374 * 1. fasync_list is modified only under process context socket lock 1375 * i.e. under semaphore. 1376 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1377 * or under socket lock 1378 */ 1379 1380 static int sock_fasync(int fd, struct file *filp, int on) 1381 { 1382 struct socket *sock = filp->private_data; 1383 struct sock *sk = sock->sk; 1384 struct socket_wq *wq = &sock->wq; 1385 1386 if (sk == NULL) 1387 return -EINVAL; 1388 1389 lock_sock(sk); 1390 fasync_helper(fd, filp, on, &wq->fasync_list); 1391 1392 if (!wq->fasync_list) 1393 sock_reset_flag(sk, SOCK_FASYNC); 1394 else 1395 sock_set_flag(sk, SOCK_FASYNC); 1396 1397 release_sock(sk); 1398 return 0; 1399 } 1400 1401 /* This function may be called only under rcu_lock */ 1402 1403 int sock_wake_async(struct socket_wq *wq, int how, int band) 1404 { 1405 if (!wq || !wq->fasync_list) 1406 return -1; 1407 1408 switch (how) { 1409 case SOCK_WAKE_WAITD: 1410 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) 1411 break; 1412 goto call_kill; 1413 case SOCK_WAKE_SPACE: 1414 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) 1415 break; 1416 fallthrough; 1417 case SOCK_WAKE_IO: 1418 call_kill: 1419 kill_fasync(&wq->fasync_list, SIGIO, band); 1420 break; 1421 case SOCK_WAKE_URG: 1422 kill_fasync(&wq->fasync_list, SIGURG, band); 1423 } 1424 1425 return 0; 1426 } 1427 EXPORT_SYMBOL(sock_wake_async); 1428 1429 /** 1430 * __sock_create - creates a socket 1431 * @net: net namespace 1432 * @family: protocol family (AF_INET, ...) 1433 * @type: communication type (SOCK_STREAM, ...) 1434 * @protocol: protocol (0, ...) 1435 * @res: new socket 1436 * @kern: boolean for kernel space sockets 1437 * 1438 * Creates a new socket and assigns it to @res, passing through LSM. 1439 * Returns 0 or an error. On failure @res is set to %NULL. @kern must 1440 * be set to true if the socket resides in kernel space. 1441 * This function internally uses GFP_KERNEL. 1442 */ 1443 1444 int __sock_create(struct net *net, int family, int type, int protocol, 1445 struct socket **res, int kern) 1446 { 1447 int err; 1448 struct socket *sock; 1449 const struct net_proto_family *pf; 1450 1451 /* 1452 * Check protocol is in range 1453 */ 1454 if (family < 0 || family >= NPROTO) 1455 return -EAFNOSUPPORT; 1456 if (type < 0 || type >= SOCK_MAX) 1457 return -EINVAL; 1458 1459 /* Compatibility. 1460 1461 This uglymoron is moved from INET layer to here to avoid 1462 deadlock in module load. 1463 */ 1464 if (family == PF_INET && type == SOCK_PACKET) { 1465 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1466 current->comm); 1467 family = PF_PACKET; 1468 } 1469 1470 err = security_socket_create(family, type, protocol, kern); 1471 if (err) 1472 return err; 1473 1474 /* 1475 * Allocate the socket and allow the family to set things up. if 1476 * the protocol is 0, the family is instructed to select an appropriate 1477 * default. 1478 */ 1479 sock = sock_alloc(); 1480 if (!sock) { 1481 net_warn_ratelimited("socket: no more sockets\n"); 1482 return -ENFILE; /* Not exactly a match, but its the 1483 closest posix thing */ 1484 } 1485 1486 sock->type = type; 1487 1488 #ifdef CONFIG_MODULES 1489 /* Attempt to load a protocol module if the find failed. 1490 * 1491 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1492 * requested real, full-featured networking support upon configuration. 1493 * Otherwise module support will break! 1494 */ 1495 if (rcu_access_pointer(net_families[family]) == NULL) 1496 request_module("net-pf-%d", family); 1497 #endif 1498 1499 rcu_read_lock(); 1500 pf = rcu_dereference(net_families[family]); 1501 err = -EAFNOSUPPORT; 1502 if (!pf) 1503 goto out_release; 1504 1505 /* 1506 * We will call the ->create function, that possibly is in a loadable 1507 * module, so we have to bump that loadable module refcnt first. 1508 */ 1509 if (!try_module_get(pf->owner)) 1510 goto out_release; 1511 1512 /* Now protected by module ref count */ 1513 rcu_read_unlock(); 1514 1515 err = pf->create(net, sock, protocol, kern); 1516 if (err < 0) 1517 goto out_module_put; 1518 1519 /* 1520 * Now to bump the refcnt of the [loadable] module that owns this 1521 * socket at sock_release time we decrement its refcnt. 1522 */ 1523 if (!try_module_get(sock->ops->owner)) 1524 goto out_module_busy; 1525 1526 /* 1527 * Now that we're done with the ->create function, the [loadable] 1528 * module can have its refcnt decremented 1529 */ 1530 module_put(pf->owner); 1531 err = security_socket_post_create(sock, family, type, protocol, kern); 1532 if (err) 1533 goto out_sock_release; 1534 *res = sock; 1535 1536 return 0; 1537 1538 out_module_busy: 1539 err = -EAFNOSUPPORT; 1540 out_module_put: 1541 sock->ops = NULL; 1542 module_put(pf->owner); 1543 out_sock_release: 1544 sock_release(sock); 1545 return err; 1546 1547 out_release: 1548 rcu_read_unlock(); 1549 goto out_sock_release; 1550 } 1551 EXPORT_SYMBOL(__sock_create); 1552 1553 /** 1554 * sock_create - creates a socket 1555 * @family: protocol family (AF_INET, ...) 1556 * @type: communication type (SOCK_STREAM, ...) 1557 * @protocol: protocol (0, ...) 1558 * @res: new socket 1559 * 1560 * A wrapper around __sock_create(). 1561 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1562 */ 1563 1564 int sock_create(int family, int type, int protocol, struct socket **res) 1565 { 1566 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1567 } 1568 EXPORT_SYMBOL(sock_create); 1569 1570 /** 1571 * sock_create_kern - creates a socket (kernel space) 1572 * @net: net namespace 1573 * @family: protocol family (AF_INET, ...) 1574 * @type: communication type (SOCK_STREAM, ...) 1575 * @protocol: protocol (0, ...) 1576 * @res: new socket 1577 * 1578 * A wrapper around __sock_create(). 1579 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1580 */ 1581 1582 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) 1583 { 1584 return __sock_create(net, family, type, protocol, res, 1); 1585 } 1586 EXPORT_SYMBOL(sock_create_kern); 1587 1588 int __sys_socket(int family, int type, int protocol) 1589 { 1590 int retval; 1591 struct socket *sock; 1592 int flags; 1593 1594 /* Check the SOCK_* constants for consistency. */ 1595 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1596 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1597 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1598 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1599 1600 flags = type & ~SOCK_TYPE_MASK; 1601 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1602 return -EINVAL; 1603 type &= SOCK_TYPE_MASK; 1604 1605 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1606 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1607 1608 retval = sock_create(family, type, protocol, &sock); 1609 if (retval < 0) 1610 return retval; 1611 1612 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1613 } 1614 1615 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1616 { 1617 return __sys_socket(family, type, protocol); 1618 } 1619 1620 /* 1621 * Create a pair of connected sockets. 1622 */ 1623 1624 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) 1625 { 1626 struct socket *sock1, *sock2; 1627 int fd1, fd2, err; 1628 struct file *newfile1, *newfile2; 1629 int flags; 1630 1631 flags = type & ~SOCK_TYPE_MASK; 1632 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1633 return -EINVAL; 1634 type &= SOCK_TYPE_MASK; 1635 1636 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1637 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1638 1639 /* 1640 * reserve descriptors and make sure we won't fail 1641 * to return them to userland. 1642 */ 1643 fd1 = get_unused_fd_flags(flags); 1644 if (unlikely(fd1 < 0)) 1645 return fd1; 1646 1647 fd2 = get_unused_fd_flags(flags); 1648 if (unlikely(fd2 < 0)) { 1649 put_unused_fd(fd1); 1650 return fd2; 1651 } 1652 1653 err = put_user(fd1, &usockvec[0]); 1654 if (err) 1655 goto out; 1656 1657 err = put_user(fd2, &usockvec[1]); 1658 if (err) 1659 goto out; 1660 1661 /* 1662 * Obtain the first socket and check if the underlying protocol 1663 * supports the socketpair call. 1664 */ 1665 1666 err = sock_create(family, type, protocol, &sock1); 1667 if (unlikely(err < 0)) 1668 goto out; 1669 1670 err = sock_create(family, type, protocol, &sock2); 1671 if (unlikely(err < 0)) { 1672 sock_release(sock1); 1673 goto out; 1674 } 1675 1676 err = security_socket_socketpair(sock1, sock2); 1677 if (unlikely(err)) { 1678 sock_release(sock2); 1679 sock_release(sock1); 1680 goto out; 1681 } 1682 1683 err = sock1->ops->socketpair(sock1, sock2); 1684 if (unlikely(err < 0)) { 1685 sock_release(sock2); 1686 sock_release(sock1); 1687 goto out; 1688 } 1689 1690 newfile1 = sock_alloc_file(sock1, flags, NULL); 1691 if (IS_ERR(newfile1)) { 1692 err = PTR_ERR(newfile1); 1693 sock_release(sock2); 1694 goto out; 1695 } 1696 1697 newfile2 = sock_alloc_file(sock2, flags, NULL); 1698 if (IS_ERR(newfile2)) { 1699 err = PTR_ERR(newfile2); 1700 fput(newfile1); 1701 goto out; 1702 } 1703 1704 audit_fd_pair(fd1, fd2); 1705 1706 fd_install(fd1, newfile1); 1707 fd_install(fd2, newfile2); 1708 return 0; 1709 1710 out: 1711 put_unused_fd(fd2); 1712 put_unused_fd(fd1); 1713 return err; 1714 } 1715 1716 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1717 int __user *, usockvec) 1718 { 1719 return __sys_socketpair(family, type, protocol, usockvec); 1720 } 1721 1722 /* 1723 * Bind a name to a socket. Nothing much to do here since it's 1724 * the protocol's responsibility to handle the local address. 1725 * 1726 * We move the socket address to kernel space before we call 1727 * the protocol layer (having also checked the address is ok). 1728 */ 1729 1730 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) 1731 { 1732 struct socket *sock; 1733 struct sockaddr_storage address; 1734 int err, fput_needed; 1735 1736 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1737 if (sock) { 1738 err = move_addr_to_kernel(umyaddr, addrlen, &address); 1739 if (!err) { 1740 err = security_socket_bind(sock, 1741 (struct sockaddr *)&address, 1742 addrlen); 1743 if (!err) 1744 err = sock->ops->bind(sock, 1745 (struct sockaddr *) 1746 &address, addrlen); 1747 } 1748 fput_light(sock->file, fput_needed); 1749 } 1750 return err; 1751 } 1752 1753 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1754 { 1755 return __sys_bind(fd, umyaddr, addrlen); 1756 } 1757 1758 /* 1759 * Perform a listen. Basically, we allow the protocol to do anything 1760 * necessary for a listen, and if that works, we mark the socket as 1761 * ready for listening. 1762 */ 1763 1764 int __sys_listen(int fd, int backlog) 1765 { 1766 struct socket *sock; 1767 int err, fput_needed; 1768 int somaxconn; 1769 1770 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1771 if (sock) { 1772 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; 1773 if ((unsigned int)backlog > somaxconn) 1774 backlog = somaxconn; 1775 1776 err = security_socket_listen(sock, backlog); 1777 if (!err) 1778 err = sock->ops->listen(sock, backlog); 1779 1780 fput_light(sock->file, fput_needed); 1781 } 1782 return err; 1783 } 1784 1785 SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1786 { 1787 return __sys_listen(fd, backlog); 1788 } 1789 1790 struct file *do_accept(struct file *file, unsigned file_flags, 1791 struct sockaddr __user *upeer_sockaddr, 1792 int __user *upeer_addrlen, int flags) 1793 { 1794 struct socket *sock, *newsock; 1795 struct file *newfile; 1796 int err, len; 1797 struct sockaddr_storage address; 1798 1799 sock = sock_from_file(file); 1800 if (!sock) 1801 return ERR_PTR(-ENOTSOCK); 1802 1803 newsock = sock_alloc(); 1804 if (!newsock) 1805 return ERR_PTR(-ENFILE); 1806 1807 newsock->type = sock->type; 1808 newsock->ops = sock->ops; 1809 1810 /* 1811 * We don't need try_module_get here, as the listening socket (sock) 1812 * has the protocol module (sock->ops->owner) held. 1813 */ 1814 __module_get(newsock->ops->owner); 1815 1816 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); 1817 if (IS_ERR(newfile)) 1818 return newfile; 1819 1820 err = security_socket_accept(sock, newsock); 1821 if (err) 1822 goto out_fd; 1823 1824 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags, 1825 false); 1826 if (err < 0) 1827 goto out_fd; 1828 1829 if (upeer_sockaddr) { 1830 len = newsock->ops->getname(newsock, 1831 (struct sockaddr *)&address, 2); 1832 if (len < 0) { 1833 err = -ECONNABORTED; 1834 goto out_fd; 1835 } 1836 err = move_addr_to_user(&address, 1837 len, upeer_sockaddr, upeer_addrlen); 1838 if (err < 0) 1839 goto out_fd; 1840 } 1841 1842 /* File flags are not inherited via accept() unlike another OSes. */ 1843 return newfile; 1844 out_fd: 1845 fput(newfile); 1846 return ERR_PTR(err); 1847 } 1848 1849 int __sys_accept4_file(struct file *file, unsigned file_flags, 1850 struct sockaddr __user *upeer_sockaddr, 1851 int __user *upeer_addrlen, int flags, 1852 unsigned long nofile) 1853 { 1854 struct file *newfile; 1855 int newfd; 1856 1857 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1858 return -EINVAL; 1859 1860 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1861 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1862 1863 newfd = __get_unused_fd_flags(flags, nofile); 1864 if (unlikely(newfd < 0)) 1865 return newfd; 1866 1867 newfile = do_accept(file, file_flags, upeer_sockaddr, upeer_addrlen, 1868 flags); 1869 if (IS_ERR(newfile)) { 1870 put_unused_fd(newfd); 1871 return PTR_ERR(newfile); 1872 } 1873 fd_install(newfd, newfile); 1874 return newfd; 1875 } 1876 1877 /* 1878 * For accept, we attempt to create a new socket, set up the link 1879 * with the client, wake up the client, then return the new 1880 * connected fd. We collect the address of the connector in kernel 1881 * space and move it to user at the very end. This is unclean because 1882 * we open the socket then return an error. 1883 * 1884 * 1003.1g adds the ability to recvmsg() to query connection pending 1885 * status to recvmsg. We need to add that support in a way thats 1886 * clean when we restructure accept also. 1887 */ 1888 1889 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, 1890 int __user *upeer_addrlen, int flags) 1891 { 1892 int ret = -EBADF; 1893 struct fd f; 1894 1895 f = fdget(fd); 1896 if (f.file) { 1897 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr, 1898 upeer_addrlen, flags, 1899 rlimit(RLIMIT_NOFILE)); 1900 fdput(f); 1901 } 1902 1903 return ret; 1904 } 1905 1906 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 1907 int __user *, upeer_addrlen, int, flags) 1908 { 1909 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); 1910 } 1911 1912 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 1913 int __user *, upeer_addrlen) 1914 { 1915 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 1916 } 1917 1918 /* 1919 * Attempt to connect to a socket with the server address. The address 1920 * is in user space so we verify it is OK and move it to kernel space. 1921 * 1922 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 1923 * break bindings 1924 * 1925 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 1926 * other SEQPACKET protocols that take time to connect() as it doesn't 1927 * include the -EINPROGRESS status for such sockets. 1928 */ 1929 1930 int __sys_connect_file(struct file *file, struct sockaddr_storage *address, 1931 int addrlen, int file_flags) 1932 { 1933 struct socket *sock; 1934 int err; 1935 1936 sock = sock_from_file(file); 1937 if (!sock) { 1938 err = -ENOTSOCK; 1939 goto out; 1940 } 1941 1942 err = 1943 security_socket_connect(sock, (struct sockaddr *)address, addrlen); 1944 if (err) 1945 goto out; 1946 1947 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen, 1948 sock->file->f_flags | file_flags); 1949 out: 1950 return err; 1951 } 1952 1953 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) 1954 { 1955 int ret = -EBADF; 1956 struct fd f; 1957 1958 f = fdget(fd); 1959 if (f.file) { 1960 struct sockaddr_storage address; 1961 1962 ret = move_addr_to_kernel(uservaddr, addrlen, &address); 1963 if (!ret) 1964 ret = __sys_connect_file(f.file, &address, addrlen, 0); 1965 fdput(f); 1966 } 1967 1968 return ret; 1969 } 1970 1971 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 1972 int, addrlen) 1973 { 1974 return __sys_connect(fd, uservaddr, addrlen); 1975 } 1976 1977 /* 1978 * Get the local address ('name') of a socket object. Move the obtained 1979 * name to user space. 1980 */ 1981 1982 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, 1983 int __user *usockaddr_len) 1984 { 1985 struct socket *sock; 1986 struct sockaddr_storage address; 1987 int err, fput_needed; 1988 1989 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1990 if (!sock) 1991 goto out; 1992 1993 err = security_socket_getsockname(sock); 1994 if (err) 1995 goto out_put; 1996 1997 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0); 1998 if (err < 0) 1999 goto out_put; 2000 /* "err" is actually length in this case */ 2001 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); 2002 2003 out_put: 2004 fput_light(sock->file, fput_needed); 2005 out: 2006 return err; 2007 } 2008 2009 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 2010 int __user *, usockaddr_len) 2011 { 2012 return __sys_getsockname(fd, usockaddr, usockaddr_len); 2013 } 2014 2015 /* 2016 * Get the remote address ('name') of a socket object. Move the obtained 2017 * name to user space. 2018 */ 2019 2020 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, 2021 int __user *usockaddr_len) 2022 { 2023 struct socket *sock; 2024 struct sockaddr_storage address; 2025 int err, fput_needed; 2026 2027 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2028 if (sock != NULL) { 2029 err = security_socket_getpeername(sock); 2030 if (err) { 2031 fput_light(sock->file, fput_needed); 2032 return err; 2033 } 2034 2035 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1); 2036 if (err >= 0) 2037 /* "err" is actually length in this case */ 2038 err = move_addr_to_user(&address, err, usockaddr, 2039 usockaddr_len); 2040 fput_light(sock->file, fput_needed); 2041 } 2042 return err; 2043 } 2044 2045 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 2046 int __user *, usockaddr_len) 2047 { 2048 return __sys_getpeername(fd, usockaddr, usockaddr_len); 2049 } 2050 2051 /* 2052 * Send a datagram to a given address. We move the address into kernel 2053 * space and check the user space data area is readable before invoking 2054 * the protocol. 2055 */ 2056 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, 2057 struct sockaddr __user *addr, int addr_len) 2058 { 2059 struct socket *sock; 2060 struct sockaddr_storage address; 2061 int err; 2062 struct msghdr msg; 2063 struct iovec iov; 2064 int fput_needed; 2065 2066 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter); 2067 if (unlikely(err)) 2068 return err; 2069 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2070 if (!sock) 2071 goto out; 2072 2073 msg.msg_name = NULL; 2074 msg.msg_control = NULL; 2075 msg.msg_controllen = 0; 2076 msg.msg_namelen = 0; 2077 if (addr) { 2078 err = move_addr_to_kernel(addr, addr_len, &address); 2079 if (err < 0) 2080 goto out_put; 2081 msg.msg_name = (struct sockaddr *)&address; 2082 msg.msg_namelen = addr_len; 2083 } 2084 if (sock->file->f_flags & O_NONBLOCK) 2085 flags |= MSG_DONTWAIT; 2086 msg.msg_flags = flags; 2087 err = sock_sendmsg(sock, &msg); 2088 2089 out_put: 2090 fput_light(sock->file, fput_needed); 2091 out: 2092 return err; 2093 } 2094 2095 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 2096 unsigned int, flags, struct sockaddr __user *, addr, 2097 int, addr_len) 2098 { 2099 return __sys_sendto(fd, buff, len, flags, addr, addr_len); 2100 } 2101 2102 /* 2103 * Send a datagram down a socket. 2104 */ 2105 2106 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 2107 unsigned int, flags) 2108 { 2109 return __sys_sendto(fd, buff, len, flags, NULL, 0); 2110 } 2111 2112 /* 2113 * Receive a frame from the socket and optionally record the address of the 2114 * sender. We verify the buffers are writable and if needed move the 2115 * sender address from kernel to user space. 2116 */ 2117 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, 2118 struct sockaddr __user *addr, int __user *addr_len) 2119 { 2120 struct socket *sock; 2121 struct iovec iov; 2122 struct msghdr msg; 2123 struct sockaddr_storage address; 2124 int err, err2; 2125 int fput_needed; 2126 2127 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter); 2128 if (unlikely(err)) 2129 return err; 2130 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2131 if (!sock) 2132 goto out; 2133 2134 msg.msg_control = NULL; 2135 msg.msg_controllen = 0; 2136 /* Save some cycles and don't copy the address if not needed */ 2137 msg.msg_name = addr ? (struct sockaddr *)&address : NULL; 2138 /* We assume all kernel code knows the size of sockaddr_storage */ 2139 msg.msg_namelen = 0; 2140 msg.msg_iocb = NULL; 2141 msg.msg_flags = 0; 2142 if (sock->file->f_flags & O_NONBLOCK) 2143 flags |= MSG_DONTWAIT; 2144 err = sock_recvmsg(sock, &msg, flags); 2145 2146 if (err >= 0 && addr != NULL) { 2147 err2 = move_addr_to_user(&address, 2148 msg.msg_namelen, addr, addr_len); 2149 if (err2 < 0) 2150 err = err2; 2151 } 2152 2153 fput_light(sock->file, fput_needed); 2154 out: 2155 return err; 2156 } 2157 2158 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 2159 unsigned int, flags, struct sockaddr __user *, addr, 2160 int __user *, addr_len) 2161 { 2162 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); 2163 } 2164 2165 /* 2166 * Receive a datagram from a socket. 2167 */ 2168 2169 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, 2170 unsigned int, flags) 2171 { 2172 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 2173 } 2174 2175 static bool sock_use_custom_sol_socket(const struct socket *sock) 2176 { 2177 const struct sock *sk = sock->sk; 2178 2179 /* Use sock->ops->setsockopt() for MPTCP */ 2180 return IS_ENABLED(CONFIG_MPTCP) && 2181 sk->sk_protocol == IPPROTO_MPTCP && 2182 sk->sk_type == SOCK_STREAM && 2183 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6); 2184 } 2185 2186 /* 2187 * Set a socket option. Because we don't know the option lengths we have 2188 * to pass the user mode parameter for the protocols to sort out. 2189 */ 2190 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval, 2191 int optlen) 2192 { 2193 sockptr_t optval = USER_SOCKPTR(user_optval); 2194 char *kernel_optval = NULL; 2195 int err, fput_needed; 2196 struct socket *sock; 2197 2198 if (optlen < 0) 2199 return -EINVAL; 2200 2201 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2202 if (!sock) 2203 return err; 2204 2205 err = security_socket_setsockopt(sock, level, optname); 2206 if (err) 2207 goto out_put; 2208 2209 if (!in_compat_syscall()) 2210 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname, 2211 user_optval, &optlen, 2212 &kernel_optval); 2213 if (err < 0) 2214 goto out_put; 2215 if (err > 0) { 2216 err = 0; 2217 goto out_put; 2218 } 2219 2220 if (kernel_optval) 2221 optval = KERNEL_SOCKPTR(kernel_optval); 2222 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock)) 2223 err = sock_setsockopt(sock, level, optname, optval, optlen); 2224 else if (unlikely(!sock->ops->setsockopt)) 2225 err = -EOPNOTSUPP; 2226 else 2227 err = sock->ops->setsockopt(sock, level, optname, optval, 2228 optlen); 2229 kfree(kernel_optval); 2230 out_put: 2231 fput_light(sock->file, fput_needed); 2232 return err; 2233 } 2234 2235 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 2236 char __user *, optval, int, optlen) 2237 { 2238 return __sys_setsockopt(fd, level, optname, optval, optlen); 2239 } 2240 2241 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level, 2242 int optname)); 2243 2244 /* 2245 * Get a socket option. Because we don't know the option lengths we have 2246 * to pass a user mode parameter for the protocols to sort out. 2247 */ 2248 int __sys_getsockopt(int fd, int level, int optname, char __user *optval, 2249 int __user *optlen) 2250 { 2251 int err, fput_needed; 2252 struct socket *sock; 2253 int max_optlen; 2254 2255 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2256 if (!sock) 2257 return err; 2258 2259 err = security_socket_getsockopt(sock, level, optname); 2260 if (err) 2261 goto out_put; 2262 2263 if (!in_compat_syscall()) 2264 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen); 2265 2266 if (level == SOL_SOCKET) 2267 err = sock_getsockopt(sock, level, optname, optval, optlen); 2268 else if (unlikely(!sock->ops->getsockopt)) 2269 err = -EOPNOTSUPP; 2270 else 2271 err = sock->ops->getsockopt(sock, level, optname, optval, 2272 optlen); 2273 2274 if (!in_compat_syscall()) 2275 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname, 2276 optval, optlen, max_optlen, 2277 err); 2278 out_put: 2279 fput_light(sock->file, fput_needed); 2280 return err; 2281 } 2282 2283 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 2284 char __user *, optval, int __user *, optlen) 2285 { 2286 return __sys_getsockopt(fd, level, optname, optval, optlen); 2287 } 2288 2289 /* 2290 * Shutdown a socket. 2291 */ 2292 2293 int __sys_shutdown_sock(struct socket *sock, int how) 2294 { 2295 int err; 2296 2297 err = security_socket_shutdown(sock, how); 2298 if (!err) 2299 err = sock->ops->shutdown(sock, how); 2300 2301 return err; 2302 } 2303 2304 int __sys_shutdown(int fd, int how) 2305 { 2306 int err, fput_needed; 2307 struct socket *sock; 2308 2309 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2310 if (sock != NULL) { 2311 err = __sys_shutdown_sock(sock, how); 2312 fput_light(sock->file, fput_needed); 2313 } 2314 return err; 2315 } 2316 2317 SYSCALL_DEFINE2(shutdown, int, fd, int, how) 2318 { 2319 return __sys_shutdown(fd, how); 2320 } 2321 2322 /* A couple of helpful macros for getting the address of the 32/64 bit 2323 * fields which are the same type (int / unsigned) on our platforms. 2324 */ 2325 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 2326 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 2327 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 2328 2329 struct used_address { 2330 struct sockaddr_storage name; 2331 unsigned int name_len; 2332 }; 2333 2334 int __copy_msghdr_from_user(struct msghdr *kmsg, 2335 struct user_msghdr __user *umsg, 2336 struct sockaddr __user **save_addr, 2337 struct iovec __user **uiov, size_t *nsegs) 2338 { 2339 struct user_msghdr msg; 2340 ssize_t err; 2341 2342 if (copy_from_user(&msg, umsg, sizeof(*umsg))) 2343 return -EFAULT; 2344 2345 kmsg->msg_control_is_user = true; 2346 kmsg->msg_control_user = msg.msg_control; 2347 kmsg->msg_controllen = msg.msg_controllen; 2348 kmsg->msg_flags = msg.msg_flags; 2349 2350 kmsg->msg_namelen = msg.msg_namelen; 2351 if (!msg.msg_name) 2352 kmsg->msg_namelen = 0; 2353 2354 if (kmsg->msg_namelen < 0) 2355 return -EINVAL; 2356 2357 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) 2358 kmsg->msg_namelen = sizeof(struct sockaddr_storage); 2359 2360 if (save_addr) 2361 *save_addr = msg.msg_name; 2362 2363 if (msg.msg_name && kmsg->msg_namelen) { 2364 if (!save_addr) { 2365 err = move_addr_to_kernel(msg.msg_name, 2366 kmsg->msg_namelen, 2367 kmsg->msg_name); 2368 if (err < 0) 2369 return err; 2370 } 2371 } else { 2372 kmsg->msg_name = NULL; 2373 kmsg->msg_namelen = 0; 2374 } 2375 2376 if (msg.msg_iovlen > UIO_MAXIOV) 2377 return -EMSGSIZE; 2378 2379 kmsg->msg_iocb = NULL; 2380 *uiov = msg.msg_iov; 2381 *nsegs = msg.msg_iovlen; 2382 return 0; 2383 } 2384 2385 static int copy_msghdr_from_user(struct msghdr *kmsg, 2386 struct user_msghdr __user *umsg, 2387 struct sockaddr __user **save_addr, 2388 struct iovec **iov) 2389 { 2390 struct user_msghdr msg; 2391 ssize_t err; 2392 2393 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov, 2394 &msg.msg_iovlen); 2395 if (err) 2396 return err; 2397 2398 err = import_iovec(save_addr ? READ : WRITE, 2399 msg.msg_iov, msg.msg_iovlen, 2400 UIO_FASTIOV, iov, &kmsg->msg_iter); 2401 return err < 0 ? err : 0; 2402 } 2403 2404 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, 2405 unsigned int flags, struct used_address *used_address, 2406 unsigned int allowed_msghdr_flags) 2407 { 2408 unsigned char ctl[sizeof(struct cmsghdr) + 20] 2409 __aligned(sizeof(__kernel_size_t)); 2410 /* 20 is size of ipv6_pktinfo */ 2411 unsigned char *ctl_buf = ctl; 2412 int ctl_len; 2413 ssize_t err; 2414 2415 err = -ENOBUFS; 2416 2417 if (msg_sys->msg_controllen > INT_MAX) 2418 goto out; 2419 flags |= (msg_sys->msg_flags & allowed_msghdr_flags); 2420 ctl_len = msg_sys->msg_controllen; 2421 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 2422 err = 2423 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, 2424 sizeof(ctl)); 2425 if (err) 2426 goto out; 2427 ctl_buf = msg_sys->msg_control; 2428 ctl_len = msg_sys->msg_controllen; 2429 } else if (ctl_len) { 2430 BUILD_BUG_ON(sizeof(struct cmsghdr) != 2431 CMSG_ALIGN(sizeof(struct cmsghdr))); 2432 if (ctl_len > sizeof(ctl)) { 2433 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 2434 if (ctl_buf == NULL) 2435 goto out; 2436 } 2437 err = -EFAULT; 2438 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len)) 2439 goto out_freectl; 2440 msg_sys->msg_control = ctl_buf; 2441 msg_sys->msg_control_is_user = false; 2442 } 2443 msg_sys->msg_flags = flags; 2444 2445 if (sock->file->f_flags & O_NONBLOCK) 2446 msg_sys->msg_flags |= MSG_DONTWAIT; 2447 /* 2448 * If this is sendmmsg() and current destination address is same as 2449 * previously succeeded address, omit asking LSM's decision. 2450 * used_address->name_len is initialized to UINT_MAX so that the first 2451 * destination address never matches. 2452 */ 2453 if (used_address && msg_sys->msg_name && 2454 used_address->name_len == msg_sys->msg_namelen && 2455 !memcmp(&used_address->name, msg_sys->msg_name, 2456 used_address->name_len)) { 2457 err = sock_sendmsg_nosec(sock, msg_sys); 2458 goto out_freectl; 2459 } 2460 err = sock_sendmsg(sock, msg_sys); 2461 /* 2462 * If this is sendmmsg() and sending to current destination address was 2463 * successful, remember it. 2464 */ 2465 if (used_address && err >= 0) { 2466 used_address->name_len = msg_sys->msg_namelen; 2467 if (msg_sys->msg_name) 2468 memcpy(&used_address->name, msg_sys->msg_name, 2469 used_address->name_len); 2470 } 2471 2472 out_freectl: 2473 if (ctl_buf != ctl) 2474 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 2475 out: 2476 return err; 2477 } 2478 2479 int sendmsg_copy_msghdr(struct msghdr *msg, 2480 struct user_msghdr __user *umsg, unsigned flags, 2481 struct iovec **iov) 2482 { 2483 int err; 2484 2485 if (flags & MSG_CMSG_COMPAT) { 2486 struct compat_msghdr __user *msg_compat; 2487 2488 msg_compat = (struct compat_msghdr __user *) umsg; 2489 err = get_compat_msghdr(msg, msg_compat, NULL, iov); 2490 } else { 2491 err = copy_msghdr_from_user(msg, umsg, NULL, iov); 2492 } 2493 if (err < 0) 2494 return err; 2495 2496 return 0; 2497 } 2498 2499 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, 2500 struct msghdr *msg_sys, unsigned int flags, 2501 struct used_address *used_address, 2502 unsigned int allowed_msghdr_flags) 2503 { 2504 struct sockaddr_storage address; 2505 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2506 ssize_t err; 2507 2508 msg_sys->msg_name = &address; 2509 2510 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov); 2511 if (err < 0) 2512 return err; 2513 2514 err = ____sys_sendmsg(sock, msg_sys, flags, used_address, 2515 allowed_msghdr_flags); 2516 kfree(iov); 2517 return err; 2518 } 2519 2520 /* 2521 * BSD sendmsg interface 2522 */ 2523 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg, 2524 unsigned int flags) 2525 { 2526 return ____sys_sendmsg(sock, msg, flags, NULL, 0); 2527 } 2528 2529 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2530 bool forbid_cmsg_compat) 2531 { 2532 int fput_needed, err; 2533 struct msghdr msg_sys; 2534 struct socket *sock; 2535 2536 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2537 return -EINVAL; 2538 2539 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2540 if (!sock) 2541 goto out; 2542 2543 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); 2544 2545 fput_light(sock->file, fput_needed); 2546 out: 2547 return err; 2548 } 2549 2550 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) 2551 { 2552 return __sys_sendmsg(fd, msg, flags, true); 2553 } 2554 2555 /* 2556 * Linux sendmmsg interface 2557 */ 2558 2559 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2560 unsigned int flags, bool forbid_cmsg_compat) 2561 { 2562 int fput_needed, err, datagrams; 2563 struct socket *sock; 2564 struct mmsghdr __user *entry; 2565 struct compat_mmsghdr __user *compat_entry; 2566 struct msghdr msg_sys; 2567 struct used_address used_address; 2568 unsigned int oflags = flags; 2569 2570 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2571 return -EINVAL; 2572 2573 if (vlen > UIO_MAXIOV) 2574 vlen = UIO_MAXIOV; 2575 2576 datagrams = 0; 2577 2578 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2579 if (!sock) 2580 return err; 2581 2582 used_address.name_len = UINT_MAX; 2583 entry = mmsg; 2584 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2585 err = 0; 2586 flags |= MSG_BATCH; 2587 2588 while (datagrams < vlen) { 2589 if (datagrams == vlen - 1) 2590 flags = oflags; 2591 2592 if (MSG_CMSG_COMPAT & flags) { 2593 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, 2594 &msg_sys, flags, &used_address, MSG_EOR); 2595 if (err < 0) 2596 break; 2597 err = __put_user(err, &compat_entry->msg_len); 2598 ++compat_entry; 2599 } else { 2600 err = ___sys_sendmsg(sock, 2601 (struct user_msghdr __user *)entry, 2602 &msg_sys, flags, &used_address, MSG_EOR); 2603 if (err < 0) 2604 break; 2605 err = put_user(err, &entry->msg_len); 2606 ++entry; 2607 } 2608 2609 if (err) 2610 break; 2611 ++datagrams; 2612 if (msg_data_left(&msg_sys)) 2613 break; 2614 cond_resched(); 2615 } 2616 2617 fput_light(sock->file, fput_needed); 2618 2619 /* We only return an error if no datagrams were able to be sent */ 2620 if (datagrams != 0) 2621 return datagrams; 2622 2623 return err; 2624 } 2625 2626 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, 2627 unsigned int, vlen, unsigned int, flags) 2628 { 2629 return __sys_sendmmsg(fd, mmsg, vlen, flags, true); 2630 } 2631 2632 int recvmsg_copy_msghdr(struct msghdr *msg, 2633 struct user_msghdr __user *umsg, unsigned flags, 2634 struct sockaddr __user **uaddr, 2635 struct iovec **iov) 2636 { 2637 ssize_t err; 2638 2639 if (MSG_CMSG_COMPAT & flags) { 2640 struct compat_msghdr __user *msg_compat; 2641 2642 msg_compat = (struct compat_msghdr __user *) umsg; 2643 err = get_compat_msghdr(msg, msg_compat, uaddr, iov); 2644 } else { 2645 err = copy_msghdr_from_user(msg, umsg, uaddr, iov); 2646 } 2647 if (err < 0) 2648 return err; 2649 2650 return 0; 2651 } 2652 2653 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys, 2654 struct user_msghdr __user *msg, 2655 struct sockaddr __user *uaddr, 2656 unsigned int flags, int nosec) 2657 { 2658 struct compat_msghdr __user *msg_compat = 2659 (struct compat_msghdr __user *) msg; 2660 int __user *uaddr_len = COMPAT_NAMELEN(msg); 2661 struct sockaddr_storage addr; 2662 unsigned long cmsg_ptr; 2663 int len; 2664 ssize_t err; 2665 2666 msg_sys->msg_name = &addr; 2667 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2668 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2669 2670 /* We assume all kernel code knows the size of sockaddr_storage */ 2671 msg_sys->msg_namelen = 0; 2672 2673 if (sock->file->f_flags & O_NONBLOCK) 2674 flags |= MSG_DONTWAIT; 2675 2676 if (unlikely(nosec)) 2677 err = sock_recvmsg_nosec(sock, msg_sys, flags); 2678 else 2679 err = sock_recvmsg(sock, msg_sys, flags); 2680 2681 if (err < 0) 2682 goto out; 2683 len = err; 2684 2685 if (uaddr != NULL) { 2686 err = move_addr_to_user(&addr, 2687 msg_sys->msg_namelen, uaddr, 2688 uaddr_len); 2689 if (err < 0) 2690 goto out; 2691 } 2692 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2693 COMPAT_FLAGS(msg)); 2694 if (err) 2695 goto out; 2696 if (MSG_CMSG_COMPAT & flags) 2697 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2698 &msg_compat->msg_controllen); 2699 else 2700 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2701 &msg->msg_controllen); 2702 if (err) 2703 goto out; 2704 err = len; 2705 out: 2706 return err; 2707 } 2708 2709 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, 2710 struct msghdr *msg_sys, unsigned int flags, int nosec) 2711 { 2712 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2713 /* user mode address pointers */ 2714 struct sockaddr __user *uaddr; 2715 ssize_t err; 2716 2717 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov); 2718 if (err < 0) 2719 return err; 2720 2721 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec); 2722 kfree(iov); 2723 return err; 2724 } 2725 2726 /* 2727 * BSD recvmsg interface 2728 */ 2729 2730 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg, 2731 struct user_msghdr __user *umsg, 2732 struct sockaddr __user *uaddr, unsigned int flags) 2733 { 2734 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0); 2735 } 2736 2737 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2738 bool forbid_cmsg_compat) 2739 { 2740 int fput_needed, err; 2741 struct msghdr msg_sys; 2742 struct socket *sock; 2743 2744 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2745 return -EINVAL; 2746 2747 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2748 if (!sock) 2749 goto out; 2750 2751 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2752 2753 fput_light(sock->file, fput_needed); 2754 out: 2755 return err; 2756 } 2757 2758 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, 2759 unsigned int, flags) 2760 { 2761 return __sys_recvmsg(fd, msg, flags, true); 2762 } 2763 2764 /* 2765 * Linux recvmmsg interface 2766 */ 2767 2768 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2769 unsigned int vlen, unsigned int flags, 2770 struct timespec64 *timeout) 2771 { 2772 int fput_needed, err, datagrams; 2773 struct socket *sock; 2774 struct mmsghdr __user *entry; 2775 struct compat_mmsghdr __user *compat_entry; 2776 struct msghdr msg_sys; 2777 struct timespec64 end_time; 2778 struct timespec64 timeout64; 2779 2780 if (timeout && 2781 poll_select_set_timeout(&end_time, timeout->tv_sec, 2782 timeout->tv_nsec)) 2783 return -EINVAL; 2784 2785 datagrams = 0; 2786 2787 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2788 if (!sock) 2789 return err; 2790 2791 if (likely(!(flags & MSG_ERRQUEUE))) { 2792 err = sock_error(sock->sk); 2793 if (err) { 2794 datagrams = err; 2795 goto out_put; 2796 } 2797 } 2798 2799 entry = mmsg; 2800 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2801 2802 while (datagrams < vlen) { 2803 /* 2804 * No need to ask LSM for more than the first datagram. 2805 */ 2806 if (MSG_CMSG_COMPAT & flags) { 2807 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, 2808 &msg_sys, flags & ~MSG_WAITFORONE, 2809 datagrams); 2810 if (err < 0) 2811 break; 2812 err = __put_user(err, &compat_entry->msg_len); 2813 ++compat_entry; 2814 } else { 2815 err = ___sys_recvmsg(sock, 2816 (struct user_msghdr __user *)entry, 2817 &msg_sys, flags & ~MSG_WAITFORONE, 2818 datagrams); 2819 if (err < 0) 2820 break; 2821 err = put_user(err, &entry->msg_len); 2822 ++entry; 2823 } 2824 2825 if (err) 2826 break; 2827 ++datagrams; 2828 2829 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2830 if (flags & MSG_WAITFORONE) 2831 flags |= MSG_DONTWAIT; 2832 2833 if (timeout) { 2834 ktime_get_ts64(&timeout64); 2835 *timeout = timespec64_sub(end_time, timeout64); 2836 if (timeout->tv_sec < 0) { 2837 timeout->tv_sec = timeout->tv_nsec = 0; 2838 break; 2839 } 2840 2841 /* Timeout, return less than vlen datagrams */ 2842 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2843 break; 2844 } 2845 2846 /* Out of band data, return right away */ 2847 if (msg_sys.msg_flags & MSG_OOB) 2848 break; 2849 cond_resched(); 2850 } 2851 2852 if (err == 0) 2853 goto out_put; 2854 2855 if (datagrams == 0) { 2856 datagrams = err; 2857 goto out_put; 2858 } 2859 2860 /* 2861 * We may return less entries than requested (vlen) if the 2862 * sock is non block and there aren't enough datagrams... 2863 */ 2864 if (err != -EAGAIN) { 2865 /* 2866 * ... or if recvmsg returns an error after we 2867 * received some datagrams, where we record the 2868 * error to return on the next call or if the 2869 * app asks about it using getsockopt(SO_ERROR). 2870 */ 2871 sock->sk->sk_err = -err; 2872 } 2873 out_put: 2874 fput_light(sock->file, fput_needed); 2875 2876 return datagrams; 2877 } 2878 2879 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2880 unsigned int vlen, unsigned int flags, 2881 struct __kernel_timespec __user *timeout, 2882 struct old_timespec32 __user *timeout32) 2883 { 2884 int datagrams; 2885 struct timespec64 timeout_sys; 2886 2887 if (timeout && get_timespec64(&timeout_sys, timeout)) 2888 return -EFAULT; 2889 2890 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32)) 2891 return -EFAULT; 2892 2893 if (!timeout && !timeout32) 2894 return do_recvmmsg(fd, mmsg, vlen, flags, NULL); 2895 2896 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 2897 2898 if (datagrams <= 0) 2899 return datagrams; 2900 2901 if (timeout && put_timespec64(&timeout_sys, timeout)) 2902 datagrams = -EFAULT; 2903 2904 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32)) 2905 datagrams = -EFAULT; 2906 2907 return datagrams; 2908 } 2909 2910 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 2911 unsigned int, vlen, unsigned int, flags, 2912 struct __kernel_timespec __user *, timeout) 2913 { 2914 if (flags & MSG_CMSG_COMPAT) 2915 return -EINVAL; 2916 2917 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL); 2918 } 2919 2920 #ifdef CONFIG_COMPAT_32BIT_TIME 2921 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg, 2922 unsigned int, vlen, unsigned int, flags, 2923 struct old_timespec32 __user *, timeout) 2924 { 2925 if (flags & MSG_CMSG_COMPAT) 2926 return -EINVAL; 2927 2928 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout); 2929 } 2930 #endif 2931 2932 #ifdef __ARCH_WANT_SYS_SOCKETCALL 2933 /* Argument list sizes for sys_socketcall */ 2934 #define AL(x) ((x) * sizeof(unsigned long)) 2935 static const unsigned char nargs[21] = { 2936 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 2937 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 2938 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 2939 AL(4), AL(5), AL(4) 2940 }; 2941 2942 #undef AL 2943 2944 /* 2945 * System call vectors. 2946 * 2947 * Argument checking cleaned up. Saved 20% in size. 2948 * This function doesn't need to set the kernel lock because 2949 * it is set by the callees. 2950 */ 2951 2952 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 2953 { 2954 unsigned long a[AUDITSC_ARGS]; 2955 unsigned long a0, a1; 2956 int err; 2957 unsigned int len; 2958 2959 if (call < 1 || call > SYS_SENDMMSG) 2960 return -EINVAL; 2961 call = array_index_nospec(call, SYS_SENDMMSG + 1); 2962 2963 len = nargs[call]; 2964 if (len > sizeof(a)) 2965 return -EINVAL; 2966 2967 /* copy_from_user should be SMP safe. */ 2968 if (copy_from_user(a, args, len)) 2969 return -EFAULT; 2970 2971 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); 2972 if (err) 2973 return err; 2974 2975 a0 = a[0]; 2976 a1 = a[1]; 2977 2978 switch (call) { 2979 case SYS_SOCKET: 2980 err = __sys_socket(a0, a1, a[2]); 2981 break; 2982 case SYS_BIND: 2983 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 2984 break; 2985 case SYS_CONNECT: 2986 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 2987 break; 2988 case SYS_LISTEN: 2989 err = __sys_listen(a0, a1); 2990 break; 2991 case SYS_ACCEPT: 2992 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 2993 (int __user *)a[2], 0); 2994 break; 2995 case SYS_GETSOCKNAME: 2996 err = 2997 __sys_getsockname(a0, (struct sockaddr __user *)a1, 2998 (int __user *)a[2]); 2999 break; 3000 case SYS_GETPEERNAME: 3001 err = 3002 __sys_getpeername(a0, (struct sockaddr __user *)a1, 3003 (int __user *)a[2]); 3004 break; 3005 case SYS_SOCKETPAIR: 3006 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 3007 break; 3008 case SYS_SEND: 3009 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 3010 NULL, 0); 3011 break; 3012 case SYS_SENDTO: 3013 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 3014 (struct sockaddr __user *)a[4], a[5]); 3015 break; 3016 case SYS_RECV: 3017 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 3018 NULL, NULL); 3019 break; 3020 case SYS_RECVFROM: 3021 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 3022 (struct sockaddr __user *)a[4], 3023 (int __user *)a[5]); 3024 break; 3025 case SYS_SHUTDOWN: 3026 err = __sys_shutdown(a0, a1); 3027 break; 3028 case SYS_SETSOCKOPT: 3029 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], 3030 a[4]); 3031 break; 3032 case SYS_GETSOCKOPT: 3033 err = 3034 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 3035 (int __user *)a[4]); 3036 break; 3037 case SYS_SENDMSG: 3038 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, 3039 a[2], true); 3040 break; 3041 case SYS_SENDMMSG: 3042 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], 3043 a[3], true); 3044 break; 3045 case SYS_RECVMSG: 3046 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, 3047 a[2], true); 3048 break; 3049 case SYS_RECVMMSG: 3050 if (IS_ENABLED(CONFIG_64BIT)) 3051 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 3052 a[2], a[3], 3053 (struct __kernel_timespec __user *)a[4], 3054 NULL); 3055 else 3056 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 3057 a[2], a[3], NULL, 3058 (struct old_timespec32 __user *)a[4]); 3059 break; 3060 case SYS_ACCEPT4: 3061 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 3062 (int __user *)a[2], a[3]); 3063 break; 3064 default: 3065 err = -EINVAL; 3066 break; 3067 } 3068 return err; 3069 } 3070 3071 #endif /* __ARCH_WANT_SYS_SOCKETCALL */ 3072 3073 /** 3074 * sock_register - add a socket protocol handler 3075 * @ops: description of protocol 3076 * 3077 * This function is called by a protocol handler that wants to 3078 * advertise its address family, and have it linked into the 3079 * socket interface. The value ops->family corresponds to the 3080 * socket system call protocol family. 3081 */ 3082 int sock_register(const struct net_proto_family *ops) 3083 { 3084 int err; 3085 3086 if (ops->family >= NPROTO) { 3087 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); 3088 return -ENOBUFS; 3089 } 3090 3091 spin_lock(&net_family_lock); 3092 if (rcu_dereference_protected(net_families[ops->family], 3093 lockdep_is_held(&net_family_lock))) 3094 err = -EEXIST; 3095 else { 3096 rcu_assign_pointer(net_families[ops->family], ops); 3097 err = 0; 3098 } 3099 spin_unlock(&net_family_lock); 3100 3101 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]); 3102 return err; 3103 } 3104 EXPORT_SYMBOL(sock_register); 3105 3106 /** 3107 * sock_unregister - remove a protocol handler 3108 * @family: protocol family to remove 3109 * 3110 * This function is called by a protocol handler that wants to 3111 * remove its address family, and have it unlinked from the 3112 * new socket creation. 3113 * 3114 * If protocol handler is a module, then it can use module reference 3115 * counts to protect against new references. If protocol handler is not 3116 * a module then it needs to provide its own protection in 3117 * the ops->create routine. 3118 */ 3119 void sock_unregister(int family) 3120 { 3121 BUG_ON(family < 0 || family >= NPROTO); 3122 3123 spin_lock(&net_family_lock); 3124 RCU_INIT_POINTER(net_families[family], NULL); 3125 spin_unlock(&net_family_lock); 3126 3127 synchronize_rcu(); 3128 3129 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]); 3130 } 3131 EXPORT_SYMBOL(sock_unregister); 3132 3133 bool sock_is_registered(int family) 3134 { 3135 return family < NPROTO && rcu_access_pointer(net_families[family]); 3136 } 3137 3138 static int __init sock_init(void) 3139 { 3140 int err; 3141 /* 3142 * Initialize the network sysctl infrastructure. 3143 */ 3144 err = net_sysctl_init(); 3145 if (err) 3146 goto out; 3147 3148 /* 3149 * Initialize skbuff SLAB cache 3150 */ 3151 skb_init(); 3152 3153 /* 3154 * Initialize the protocols module. 3155 */ 3156 3157 init_inodecache(); 3158 3159 err = register_filesystem(&sock_fs_type); 3160 if (err) 3161 goto out; 3162 sock_mnt = kern_mount(&sock_fs_type); 3163 if (IS_ERR(sock_mnt)) { 3164 err = PTR_ERR(sock_mnt); 3165 goto out_mount; 3166 } 3167 3168 /* The real protocol initialization is performed in later initcalls. 3169 */ 3170 3171 #ifdef CONFIG_NETFILTER 3172 err = netfilter_init(); 3173 if (err) 3174 goto out; 3175 #endif 3176 3177 ptp_classifier_init(); 3178 3179 out: 3180 return err; 3181 3182 out_mount: 3183 unregister_filesystem(&sock_fs_type); 3184 goto out; 3185 } 3186 3187 core_initcall(sock_init); /* early initcall */ 3188 3189 #ifdef CONFIG_PROC_FS 3190 void socket_seq_show(struct seq_file *seq) 3191 { 3192 seq_printf(seq, "sockets: used %d\n", 3193 sock_inuse_get(seq->private)); 3194 } 3195 #endif /* CONFIG_PROC_FS */ 3196 3197 /* Handle the fact that while struct ifreq has the same *layout* on 3198 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data, 3199 * which are handled elsewhere, it still has different *size* due to 3200 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit, 3201 * resulting in struct ifreq being 32 and 40 bytes respectively). 3202 * As a result, if the struct happens to be at the end of a page and 3203 * the next page isn't readable/writable, we get a fault. To prevent 3204 * that, copy back and forth to the full size. 3205 */ 3206 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg) 3207 { 3208 if (in_compat_syscall()) { 3209 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr; 3210 3211 memset(ifr, 0, sizeof(*ifr)); 3212 if (copy_from_user(ifr32, arg, sizeof(*ifr32))) 3213 return -EFAULT; 3214 3215 if (ifrdata) 3216 *ifrdata = compat_ptr(ifr32->ifr_data); 3217 3218 return 0; 3219 } 3220 3221 if (copy_from_user(ifr, arg, sizeof(*ifr))) 3222 return -EFAULT; 3223 3224 if (ifrdata) 3225 *ifrdata = ifr->ifr_data; 3226 3227 return 0; 3228 } 3229 EXPORT_SYMBOL(get_user_ifreq); 3230 3231 int put_user_ifreq(struct ifreq *ifr, void __user *arg) 3232 { 3233 size_t size = sizeof(*ifr); 3234 3235 if (in_compat_syscall()) 3236 size = sizeof(struct compat_ifreq); 3237 3238 if (copy_to_user(arg, ifr, size)) 3239 return -EFAULT; 3240 3241 return 0; 3242 } 3243 EXPORT_SYMBOL(put_user_ifreq); 3244 3245 #ifdef CONFIG_COMPAT 3246 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 3247 { 3248 compat_uptr_t uptr32; 3249 struct ifreq ifr; 3250 void __user *saved; 3251 int err; 3252 3253 if (get_user_ifreq(&ifr, NULL, uifr32)) 3254 return -EFAULT; 3255 3256 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 3257 return -EFAULT; 3258 3259 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; 3260 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); 3261 3262 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL); 3263 if (!err) { 3264 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; 3265 if (put_user_ifreq(&ifr, uifr32)) 3266 err = -EFAULT; 3267 } 3268 return err; 3269 } 3270 3271 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ 3272 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, 3273 struct compat_ifreq __user *u_ifreq32) 3274 { 3275 struct ifreq ifreq; 3276 void __user *data; 3277 3278 if (!is_socket_ioctl_cmd(cmd)) 3279 return -ENOTTY; 3280 if (get_user_ifreq(&ifreq, &data, u_ifreq32)) 3281 return -EFAULT; 3282 ifreq.ifr_data = data; 3283 3284 return dev_ioctl(net, cmd, &ifreq, data, NULL); 3285 } 3286 3287 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 3288 unsigned int cmd, unsigned long arg) 3289 { 3290 void __user *argp = compat_ptr(arg); 3291 struct sock *sk = sock->sk; 3292 struct net *net = sock_net(sk); 3293 3294 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 3295 return sock_ioctl(file, cmd, (unsigned long)argp); 3296 3297 switch (cmd) { 3298 case SIOCWANDEV: 3299 return compat_siocwandev(net, argp); 3300 case SIOCGSTAMP_OLD: 3301 case SIOCGSTAMPNS_OLD: 3302 if (!sock->ops->gettstamp) 3303 return -ENOIOCTLCMD; 3304 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD, 3305 !COMPAT_USE_64BIT_TIME); 3306 3307 case SIOCETHTOOL: 3308 case SIOCBONDSLAVEINFOQUERY: 3309 case SIOCBONDINFOQUERY: 3310 case SIOCSHWTSTAMP: 3311 case SIOCGHWTSTAMP: 3312 return compat_ifr_data_ioctl(net, cmd, argp); 3313 3314 case FIOSETOWN: 3315 case SIOCSPGRP: 3316 case FIOGETOWN: 3317 case SIOCGPGRP: 3318 case SIOCBRADDBR: 3319 case SIOCBRDELBR: 3320 case SIOCGIFVLAN: 3321 case SIOCSIFVLAN: 3322 case SIOCGSKNS: 3323 case SIOCGSTAMP_NEW: 3324 case SIOCGSTAMPNS_NEW: 3325 case SIOCGIFCONF: 3326 case SIOCSIFBR: 3327 case SIOCGIFBR: 3328 return sock_ioctl(file, cmd, arg); 3329 3330 case SIOCGIFFLAGS: 3331 case SIOCSIFFLAGS: 3332 case SIOCGIFMAP: 3333 case SIOCSIFMAP: 3334 case SIOCGIFMETRIC: 3335 case SIOCSIFMETRIC: 3336 case SIOCGIFMTU: 3337 case SIOCSIFMTU: 3338 case SIOCGIFMEM: 3339 case SIOCSIFMEM: 3340 case SIOCGIFHWADDR: 3341 case SIOCSIFHWADDR: 3342 case SIOCADDMULTI: 3343 case SIOCDELMULTI: 3344 case SIOCGIFINDEX: 3345 case SIOCGIFADDR: 3346 case SIOCSIFADDR: 3347 case SIOCSIFHWBROADCAST: 3348 case SIOCDIFADDR: 3349 case SIOCGIFBRDADDR: 3350 case SIOCSIFBRDADDR: 3351 case SIOCGIFDSTADDR: 3352 case SIOCSIFDSTADDR: 3353 case SIOCGIFNETMASK: 3354 case SIOCSIFNETMASK: 3355 case SIOCSIFPFLAGS: 3356 case SIOCGIFPFLAGS: 3357 case SIOCGIFTXQLEN: 3358 case SIOCSIFTXQLEN: 3359 case SIOCBRADDIF: 3360 case SIOCBRDELIF: 3361 case SIOCGIFNAME: 3362 case SIOCSIFNAME: 3363 case SIOCGMIIPHY: 3364 case SIOCGMIIREG: 3365 case SIOCSMIIREG: 3366 case SIOCBONDENSLAVE: 3367 case SIOCBONDRELEASE: 3368 case SIOCBONDSETHWADDR: 3369 case SIOCBONDCHANGEACTIVE: 3370 case SIOCSARP: 3371 case SIOCGARP: 3372 case SIOCDARP: 3373 case SIOCOUTQ: 3374 case SIOCOUTQNSD: 3375 case SIOCATMARK: 3376 return sock_do_ioctl(net, sock, cmd, arg); 3377 } 3378 3379 return -ENOIOCTLCMD; 3380 } 3381 3382 static long compat_sock_ioctl(struct file *file, unsigned int cmd, 3383 unsigned long arg) 3384 { 3385 struct socket *sock = file->private_data; 3386 int ret = -ENOIOCTLCMD; 3387 struct sock *sk; 3388 struct net *net; 3389 3390 sk = sock->sk; 3391 net = sock_net(sk); 3392 3393 if (sock->ops->compat_ioctl) 3394 ret = sock->ops->compat_ioctl(sock, cmd, arg); 3395 3396 if (ret == -ENOIOCTLCMD && 3397 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 3398 ret = compat_wext_handle_ioctl(net, cmd, arg); 3399 3400 if (ret == -ENOIOCTLCMD) 3401 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 3402 3403 return ret; 3404 } 3405 #endif 3406 3407 /** 3408 * kernel_bind - bind an address to a socket (kernel space) 3409 * @sock: socket 3410 * @addr: address 3411 * @addrlen: length of address 3412 * 3413 * Returns 0 or an error. 3414 */ 3415 3416 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3417 { 3418 return sock->ops->bind(sock, addr, addrlen); 3419 } 3420 EXPORT_SYMBOL(kernel_bind); 3421 3422 /** 3423 * kernel_listen - move socket to listening state (kernel space) 3424 * @sock: socket 3425 * @backlog: pending connections queue size 3426 * 3427 * Returns 0 or an error. 3428 */ 3429 3430 int kernel_listen(struct socket *sock, int backlog) 3431 { 3432 return sock->ops->listen(sock, backlog); 3433 } 3434 EXPORT_SYMBOL(kernel_listen); 3435 3436 /** 3437 * kernel_accept - accept a connection (kernel space) 3438 * @sock: listening socket 3439 * @newsock: new connected socket 3440 * @flags: flags 3441 * 3442 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0. 3443 * If it fails, @newsock is guaranteed to be %NULL. 3444 * Returns 0 or an error. 3445 */ 3446 3447 int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3448 { 3449 struct sock *sk = sock->sk; 3450 int err; 3451 3452 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3453 newsock); 3454 if (err < 0) 3455 goto done; 3456 3457 err = sock->ops->accept(sock, *newsock, flags, true); 3458 if (err < 0) { 3459 sock_release(*newsock); 3460 *newsock = NULL; 3461 goto done; 3462 } 3463 3464 (*newsock)->ops = sock->ops; 3465 __module_get((*newsock)->ops->owner); 3466 3467 done: 3468 return err; 3469 } 3470 EXPORT_SYMBOL(kernel_accept); 3471 3472 /** 3473 * kernel_connect - connect a socket (kernel space) 3474 * @sock: socket 3475 * @addr: address 3476 * @addrlen: address length 3477 * @flags: flags (O_NONBLOCK, ...) 3478 * 3479 * For datagram sockets, @addr is the address to which datagrams are sent 3480 * by default, and the only address from which datagrams are received. 3481 * For stream sockets, attempts to connect to @addr. 3482 * Returns 0 or an error code. 3483 */ 3484 3485 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3486 int flags) 3487 { 3488 return sock->ops->connect(sock, addr, addrlen, flags); 3489 } 3490 EXPORT_SYMBOL(kernel_connect); 3491 3492 /** 3493 * kernel_getsockname - get the address which the socket is bound (kernel space) 3494 * @sock: socket 3495 * @addr: address holder 3496 * 3497 * Fills the @addr pointer with the address which the socket is bound. 3498 * Returns the length of the address in bytes or an error code. 3499 */ 3500 3501 int kernel_getsockname(struct socket *sock, struct sockaddr *addr) 3502 { 3503 return sock->ops->getname(sock, addr, 0); 3504 } 3505 EXPORT_SYMBOL(kernel_getsockname); 3506 3507 /** 3508 * kernel_getpeername - get the address which the socket is connected (kernel space) 3509 * @sock: socket 3510 * @addr: address holder 3511 * 3512 * Fills the @addr pointer with the address which the socket is connected. 3513 * Returns the length of the address in bytes or an error code. 3514 */ 3515 3516 int kernel_getpeername(struct socket *sock, struct sockaddr *addr) 3517 { 3518 return sock->ops->getname(sock, addr, 1); 3519 } 3520 EXPORT_SYMBOL(kernel_getpeername); 3521 3522 /** 3523 * kernel_sendpage - send a &page through a socket (kernel space) 3524 * @sock: socket 3525 * @page: page 3526 * @offset: page offset 3527 * @size: total size in bytes 3528 * @flags: flags (MSG_DONTWAIT, ...) 3529 * 3530 * Returns the total amount sent in bytes or an error. 3531 */ 3532 3533 int kernel_sendpage(struct socket *sock, struct page *page, int offset, 3534 size_t size, int flags) 3535 { 3536 if (sock->ops->sendpage) { 3537 /* Warn in case the improper page to zero-copy send */ 3538 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send"); 3539 return sock->ops->sendpage(sock, page, offset, size, flags); 3540 } 3541 return sock_no_sendpage(sock, page, offset, size, flags); 3542 } 3543 EXPORT_SYMBOL(kernel_sendpage); 3544 3545 /** 3546 * kernel_sendpage_locked - send a &page through the locked sock (kernel space) 3547 * @sk: sock 3548 * @page: page 3549 * @offset: page offset 3550 * @size: total size in bytes 3551 * @flags: flags (MSG_DONTWAIT, ...) 3552 * 3553 * Returns the total amount sent in bytes or an error. 3554 * Caller must hold @sk. 3555 */ 3556 3557 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset, 3558 size_t size, int flags) 3559 { 3560 struct socket *sock = sk->sk_socket; 3561 3562 if (sock->ops->sendpage_locked) 3563 return sock->ops->sendpage_locked(sk, page, offset, size, 3564 flags); 3565 3566 return sock_no_sendpage_locked(sk, page, offset, size, flags); 3567 } 3568 EXPORT_SYMBOL(kernel_sendpage_locked); 3569 3570 /** 3571 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space) 3572 * @sock: socket 3573 * @how: connection part 3574 * 3575 * Returns 0 or an error. 3576 */ 3577 3578 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3579 { 3580 return sock->ops->shutdown(sock, how); 3581 } 3582 EXPORT_SYMBOL(kernel_sock_shutdown); 3583 3584 /** 3585 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket 3586 * @sk: socket 3587 * 3588 * This routine returns the IP overhead imposed by a socket i.e. 3589 * the length of the underlying IP header, depending on whether 3590 * this is an IPv4 or IPv6 socket and the length from IP options turned 3591 * on at the socket. Assumes that the caller has a lock on the socket. 3592 */ 3593 3594 u32 kernel_sock_ip_overhead(struct sock *sk) 3595 { 3596 struct inet_sock *inet; 3597 struct ip_options_rcu *opt; 3598 u32 overhead = 0; 3599 #if IS_ENABLED(CONFIG_IPV6) 3600 struct ipv6_pinfo *np; 3601 struct ipv6_txoptions *optv6 = NULL; 3602 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3603 3604 if (!sk) 3605 return overhead; 3606 3607 switch (sk->sk_family) { 3608 case AF_INET: 3609 inet = inet_sk(sk); 3610 overhead += sizeof(struct iphdr); 3611 opt = rcu_dereference_protected(inet->inet_opt, 3612 sock_owned_by_user(sk)); 3613 if (opt) 3614 overhead += opt->opt.optlen; 3615 return overhead; 3616 #if IS_ENABLED(CONFIG_IPV6) 3617 case AF_INET6: 3618 np = inet6_sk(sk); 3619 overhead += sizeof(struct ipv6hdr); 3620 if (np) 3621 optv6 = rcu_dereference_protected(np->opt, 3622 sock_owned_by_user(sk)); 3623 if (optv6) 3624 overhead += (optv6->opt_flen + optv6->opt_nflen); 3625 return overhead; 3626 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3627 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ 3628 return overhead; 3629 } 3630 } 3631 EXPORT_SYMBOL(kernel_sock_ip_overhead); 3632