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