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