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/ethtool.h> 56 #include <linux/mm.h> 57 #include <linux/socket.h> 58 #include <linux/file.h> 59 #include <linux/net.h> 60 #include <linux/interrupt.h> 61 #include <linux/thread_info.h> 62 #include <linux/rcupdate.h> 63 #include <linux/netdevice.h> 64 #include <linux/proc_fs.h> 65 #include <linux/seq_file.h> 66 #include <linux/mutex.h> 67 #include <linux/if_bridge.h> 68 #include <linux/if_vlan.h> 69 #include <linux/ptp_classify.h> 70 #include <linux/init.h> 71 #include <linux/poll.h> 72 #include <linux/cache.h> 73 #include <linux/module.h> 74 #include <linux/highmem.h> 75 #include <linux/mount.h> 76 #include <linux/pseudo_fs.h> 77 #include <linux/security.h> 78 #include <linux/syscalls.h> 79 #include <linux/compat.h> 80 #include <linux/kmod.h> 81 #include <linux/audit.h> 82 #include <linux/wireless.h> 83 #include <linux/nsproxy.h> 84 #include <linux/magic.h> 85 #include <linux/slab.h> 86 #include <linux/xattr.h> 87 #include <linux/nospec.h> 88 #include <linux/indirect_call_wrapper.h> 89 90 #include <linux/uaccess.h> 91 #include <asm/unistd.h> 92 93 #include <net/compat.h> 94 #include <net/wext.h> 95 #include <net/cls_cgroup.h> 96 97 #include <net/sock.h> 98 #include <linux/netfilter.h> 99 100 #include <linux/if_tun.h> 101 #include <linux/ipv6_route.h> 102 #include <linux/route.h> 103 #include <linux/termios.h> 104 #include <linux/sockios.h> 105 #include <net/busy_poll.h> 106 #include <linux/errqueue.h> 107 #include <linux/ptp_clock_kernel.h> 108 109 #ifdef CONFIG_NET_RX_BUSY_POLL 110 unsigned int sysctl_net_busy_read __read_mostly; 111 unsigned int sysctl_net_busy_poll __read_mostly; 112 #endif 113 114 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to); 115 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from); 116 static int sock_mmap(struct file *file, struct vm_area_struct *vma); 117 118 static int sock_close(struct inode *inode, struct file *file); 119 static __poll_t sock_poll(struct file *file, 120 struct poll_table_struct *wait); 121 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 122 #ifdef CONFIG_COMPAT 123 static long compat_sock_ioctl(struct file *file, 124 unsigned int cmd, unsigned long arg); 125 #endif 126 static int sock_fasync(int fd, struct file *filp, int on); 127 static ssize_t sock_sendpage(struct file *file, struct page *page, 128 int offset, size_t size, loff_t *ppos, int more); 129 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 130 struct pipe_inode_info *pipe, size_t len, 131 unsigned int flags); 132 133 #ifdef CONFIG_PROC_FS 134 static void sock_show_fdinfo(struct seq_file *m, struct file *f) 135 { 136 struct socket *sock = f->private_data; 137 138 if (sock->ops->show_fdinfo) 139 sock->ops->show_fdinfo(m, sock); 140 } 141 #else 142 #define sock_show_fdinfo NULL 143 #endif 144 145 /* 146 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear 147 * in the operation structures but are done directly via the socketcall() multiplexor. 148 */ 149 150 static const struct file_operations socket_file_ops = { 151 .owner = THIS_MODULE, 152 .llseek = no_llseek, 153 .read_iter = sock_read_iter, 154 .write_iter = sock_write_iter, 155 .poll = sock_poll, 156 .unlocked_ioctl = sock_ioctl, 157 #ifdef CONFIG_COMPAT 158 .compat_ioctl = compat_sock_ioctl, 159 #endif 160 .mmap = sock_mmap, 161 .release = sock_close, 162 .fasync = sock_fasync, 163 .sendpage = sock_sendpage, 164 .splice_write = generic_splice_sendpage, 165 .splice_read = sock_splice_read, 166 .show_fdinfo = sock_show_fdinfo, 167 }; 168 169 static const char * const pf_family_names[] = { 170 [PF_UNSPEC] = "PF_UNSPEC", 171 [PF_UNIX] = "PF_UNIX/PF_LOCAL", 172 [PF_INET] = "PF_INET", 173 [PF_AX25] = "PF_AX25", 174 [PF_IPX] = "PF_IPX", 175 [PF_APPLETALK] = "PF_APPLETALK", 176 [PF_NETROM] = "PF_NETROM", 177 [PF_BRIDGE] = "PF_BRIDGE", 178 [PF_ATMPVC] = "PF_ATMPVC", 179 [PF_X25] = "PF_X25", 180 [PF_INET6] = "PF_INET6", 181 [PF_ROSE] = "PF_ROSE", 182 [PF_DECnet] = "PF_DECnet", 183 [PF_NETBEUI] = "PF_NETBEUI", 184 [PF_SECURITY] = "PF_SECURITY", 185 [PF_KEY] = "PF_KEY", 186 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE", 187 [PF_PACKET] = "PF_PACKET", 188 [PF_ASH] = "PF_ASH", 189 [PF_ECONET] = "PF_ECONET", 190 [PF_ATMSVC] = "PF_ATMSVC", 191 [PF_RDS] = "PF_RDS", 192 [PF_SNA] = "PF_SNA", 193 [PF_IRDA] = "PF_IRDA", 194 [PF_PPPOX] = "PF_PPPOX", 195 [PF_WANPIPE] = "PF_WANPIPE", 196 [PF_LLC] = "PF_LLC", 197 [PF_IB] = "PF_IB", 198 [PF_MPLS] = "PF_MPLS", 199 [PF_CAN] = "PF_CAN", 200 [PF_TIPC] = "PF_TIPC", 201 [PF_BLUETOOTH] = "PF_BLUETOOTH", 202 [PF_IUCV] = "PF_IUCV", 203 [PF_RXRPC] = "PF_RXRPC", 204 [PF_ISDN] = "PF_ISDN", 205 [PF_PHONET] = "PF_PHONET", 206 [PF_IEEE802154] = "PF_IEEE802154", 207 [PF_CAIF] = "PF_CAIF", 208 [PF_ALG] = "PF_ALG", 209 [PF_NFC] = "PF_NFC", 210 [PF_VSOCK] = "PF_VSOCK", 211 [PF_KCM] = "PF_KCM", 212 [PF_QIPCRTR] = "PF_QIPCRTR", 213 [PF_SMC] = "PF_SMC", 214 [PF_XDP] = "PF_XDP", 215 }; 216 217 /* 218 * The protocol list. Each protocol is registered in here. 219 */ 220 221 static DEFINE_SPINLOCK(net_family_lock); 222 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; 223 224 /* 225 * Support routines. 226 * Move socket addresses back and forth across the kernel/user 227 * divide and look after the messy bits. 228 */ 229 230 /** 231 * move_addr_to_kernel - copy a socket address into kernel space 232 * @uaddr: Address in user space 233 * @kaddr: Address in kernel space 234 * @ulen: Length in user space 235 * 236 * The address is copied into kernel space. If the provided address is 237 * too long an error code of -EINVAL is returned. If the copy gives 238 * invalid addresses -EFAULT is returned. On a success 0 is returned. 239 */ 240 241 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr) 242 { 243 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) 244 return -EINVAL; 245 if (ulen == 0) 246 return 0; 247 if (copy_from_user(kaddr, uaddr, ulen)) 248 return -EFAULT; 249 return audit_sockaddr(ulen, kaddr); 250 } 251 252 /** 253 * move_addr_to_user - copy an address to user space 254 * @kaddr: kernel space address 255 * @klen: length of address in kernel 256 * @uaddr: user space address 257 * @ulen: pointer to user length field 258 * 259 * The value pointed to by ulen on entry is the buffer length available. 260 * This is overwritten with the buffer space used. -EINVAL is returned 261 * if an overlong buffer is specified or a negative buffer size. -EFAULT 262 * is returned if either the buffer or the length field are not 263 * accessible. 264 * After copying the data up to the limit the user specifies, the true 265 * length of the data is written over the length limit the user 266 * specified. Zero is returned for a success. 267 */ 268 269 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen, 270 void __user *uaddr, int __user *ulen) 271 { 272 int err; 273 int len; 274 275 BUG_ON(klen > sizeof(struct sockaddr_storage)); 276 err = get_user(len, ulen); 277 if (err) 278 return err; 279 if (len > klen) 280 len = klen; 281 if (len < 0) 282 return -EINVAL; 283 if (len) { 284 if (audit_sockaddr(klen, kaddr)) 285 return -ENOMEM; 286 if (copy_to_user(uaddr, kaddr, len)) 287 return -EFAULT; 288 } 289 /* 290 * "fromlen shall refer to the value before truncation.." 291 * 1003.1g 292 */ 293 return __put_user(klen, ulen); 294 } 295 296 static struct kmem_cache *sock_inode_cachep __ro_after_init; 297 298 static struct inode *sock_alloc_inode(struct super_block *sb) 299 { 300 struct socket_alloc *ei; 301 302 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); 303 if (!ei) 304 return NULL; 305 init_waitqueue_head(&ei->socket.wq.wait); 306 ei->socket.wq.fasync_list = NULL; 307 ei->socket.wq.flags = 0; 308 309 ei->socket.state = SS_UNCONNECTED; 310 ei->socket.flags = 0; 311 ei->socket.ops = NULL; 312 ei->socket.sk = NULL; 313 ei->socket.file = NULL; 314 315 return &ei->vfs_inode; 316 } 317 318 static void sock_free_inode(struct inode *inode) 319 { 320 struct socket_alloc *ei; 321 322 ei = container_of(inode, struct socket_alloc, vfs_inode); 323 kmem_cache_free(sock_inode_cachep, ei); 324 } 325 326 static void init_once(void *foo) 327 { 328 struct socket_alloc *ei = (struct socket_alloc *)foo; 329 330 inode_init_once(&ei->vfs_inode); 331 } 332 333 static void init_inodecache(void) 334 { 335 sock_inode_cachep = kmem_cache_create("sock_inode_cache", 336 sizeof(struct socket_alloc), 337 0, 338 (SLAB_HWCACHE_ALIGN | 339 SLAB_RECLAIM_ACCOUNT | 340 SLAB_MEM_SPREAD | SLAB_ACCOUNT), 341 init_once); 342 BUG_ON(sock_inode_cachep == NULL); 343 } 344 345 static const struct super_operations sockfs_ops = { 346 .alloc_inode = sock_alloc_inode, 347 .free_inode = sock_free_inode, 348 .statfs = simple_statfs, 349 }; 350 351 /* 352 * sockfs_dname() is called from d_path(). 353 */ 354 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) 355 { 356 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", 357 d_inode(dentry)->i_ino); 358 } 359 360 static const struct dentry_operations sockfs_dentry_operations = { 361 .d_dname = sockfs_dname, 362 }; 363 364 static int sockfs_xattr_get(const struct xattr_handler *handler, 365 struct dentry *dentry, struct inode *inode, 366 const char *suffix, void *value, size_t size) 367 { 368 if (value) { 369 if (dentry->d_name.len + 1 > size) 370 return -ERANGE; 371 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1); 372 } 373 return dentry->d_name.len + 1; 374 } 375 376 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname" 377 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX) 378 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1) 379 380 static const struct xattr_handler sockfs_xattr_handler = { 381 .name = XATTR_NAME_SOCKPROTONAME, 382 .get = sockfs_xattr_get, 383 }; 384 385 static int sockfs_security_xattr_set(const struct xattr_handler *handler, 386 struct user_namespace *mnt_userns, 387 struct dentry *dentry, struct inode *inode, 388 const char *suffix, const void *value, 389 size_t size, int flags) 390 { 391 /* Handled by LSM. */ 392 return -EAGAIN; 393 } 394 395 static const struct xattr_handler sockfs_security_xattr_handler = { 396 .prefix = XATTR_SECURITY_PREFIX, 397 .set = sockfs_security_xattr_set, 398 }; 399 400 static const struct xattr_handler *sockfs_xattr_handlers[] = { 401 &sockfs_xattr_handler, 402 &sockfs_security_xattr_handler, 403 NULL 404 }; 405 406 static int sockfs_init_fs_context(struct fs_context *fc) 407 { 408 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC); 409 if (!ctx) 410 return -ENOMEM; 411 ctx->ops = &sockfs_ops; 412 ctx->dops = &sockfs_dentry_operations; 413 ctx->xattr = sockfs_xattr_handlers; 414 return 0; 415 } 416 417 static struct vfsmount *sock_mnt __read_mostly; 418 419 static struct file_system_type sock_fs_type = { 420 .name = "sockfs", 421 .init_fs_context = sockfs_init_fs_context, 422 .kill_sb = kill_anon_super, 423 }; 424 425 /* 426 * Obtains the first available file descriptor and sets it up for use. 427 * 428 * These functions create file structures and maps them to fd space 429 * of the current process. On success it returns file descriptor 430 * and file struct implicitly stored in sock->file. 431 * Note that another thread may close file descriptor before we return 432 * from this function. We use the fact that now we do not refer 433 * to socket after mapping. If one day we will need it, this 434 * function will increment ref. count on file by 1. 435 * 436 * In any case returned fd MAY BE not valid! 437 * This race condition is unavoidable 438 * with shared fd spaces, we cannot solve it inside kernel, 439 * but we take care of internal coherence yet. 440 */ 441 442 /** 443 * sock_alloc_file - Bind a &socket to a &file 444 * @sock: socket 445 * @flags: file status flags 446 * @dname: protocol name 447 * 448 * Returns the &file bound with @sock, implicitly storing it 449 * in sock->file. If dname is %NULL, sets to "". 450 * On failure the return is a ERR pointer (see linux/err.h). 451 * This function uses GFP_KERNEL internally. 452 */ 453 454 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname) 455 { 456 struct file *file; 457 458 if (!dname) 459 dname = sock->sk ? sock->sk->sk_prot_creator->name : ""; 460 461 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname, 462 O_RDWR | (flags & O_NONBLOCK), 463 &socket_file_ops); 464 if (IS_ERR(file)) { 465 sock_release(sock); 466 return file; 467 } 468 469 sock->file = file; 470 file->private_data = sock; 471 stream_open(SOCK_INODE(sock), file); 472 return file; 473 } 474 EXPORT_SYMBOL(sock_alloc_file); 475 476 static int sock_map_fd(struct socket *sock, int flags) 477 { 478 struct file *newfile; 479 int fd = get_unused_fd_flags(flags); 480 if (unlikely(fd < 0)) { 481 sock_release(sock); 482 return fd; 483 } 484 485 newfile = sock_alloc_file(sock, flags, NULL); 486 if (!IS_ERR(newfile)) { 487 fd_install(fd, newfile); 488 return fd; 489 } 490 491 put_unused_fd(fd); 492 return PTR_ERR(newfile); 493 } 494 495 /** 496 * sock_from_file - Return the &socket bounded to @file. 497 * @file: file 498 * 499 * On failure returns %NULL. 500 */ 501 502 struct socket *sock_from_file(struct file *file) 503 { 504 if (file->f_op == &socket_file_ops) 505 return file->private_data; /* set in sock_map_fd */ 506 507 return NULL; 508 } 509 EXPORT_SYMBOL(sock_from_file); 510 511 /** 512 * sockfd_lookup - Go from a file number to its socket slot 513 * @fd: file handle 514 * @err: pointer to an error code return 515 * 516 * The file handle passed in is locked and the socket it is bound 517 * to is returned. If an error occurs the err pointer is overwritten 518 * with a negative errno code and NULL is returned. The function checks 519 * for both invalid handles and passing a handle which is not a socket. 520 * 521 * On a success the socket object pointer is returned. 522 */ 523 524 struct socket *sockfd_lookup(int fd, int *err) 525 { 526 struct file *file; 527 struct socket *sock; 528 529 file = fget(fd); 530 if (!file) { 531 *err = -EBADF; 532 return NULL; 533 } 534 535 sock = sock_from_file(file); 536 if (!sock) { 537 *err = -ENOTSOCK; 538 fput(file); 539 } 540 return sock; 541 } 542 EXPORT_SYMBOL(sockfd_lookup); 543 544 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) 545 { 546 struct fd f = fdget(fd); 547 struct socket *sock; 548 549 *err = -EBADF; 550 if (f.file) { 551 sock = sock_from_file(f.file); 552 if (likely(sock)) { 553 *fput_needed = f.flags & FDPUT_FPUT; 554 return sock; 555 } 556 *err = -ENOTSOCK; 557 fdput(f); 558 } 559 return NULL; 560 } 561 562 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer, 563 size_t size) 564 { 565 ssize_t len; 566 ssize_t used = 0; 567 568 len = security_inode_listsecurity(d_inode(dentry), buffer, size); 569 if (len < 0) 570 return len; 571 used += len; 572 if (buffer) { 573 if (size < used) 574 return -ERANGE; 575 buffer += len; 576 } 577 578 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1); 579 used += len; 580 if (buffer) { 581 if (size < used) 582 return -ERANGE; 583 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len); 584 buffer += len; 585 } 586 587 return used; 588 } 589 590 static int sockfs_setattr(struct user_namespace *mnt_userns, 591 struct dentry *dentry, struct iattr *iattr) 592 { 593 int err = simple_setattr(&init_user_ns, dentry, iattr); 594 595 if (!err && (iattr->ia_valid & ATTR_UID)) { 596 struct socket *sock = SOCKET_I(d_inode(dentry)); 597 598 if (sock->sk) 599 sock->sk->sk_uid = iattr->ia_uid; 600 else 601 err = -ENOENT; 602 } 603 604 return err; 605 } 606 607 static const struct inode_operations sockfs_inode_ops = { 608 .listxattr = sockfs_listxattr, 609 .setattr = sockfs_setattr, 610 }; 611 612 /** 613 * sock_alloc - allocate a socket 614 * 615 * Allocate a new inode and socket object. The two are bound together 616 * and initialised. The socket is then returned. If we are out of inodes 617 * NULL is returned. This functions uses GFP_KERNEL internally. 618 */ 619 620 struct socket *sock_alloc(void) 621 { 622 struct inode *inode; 623 struct socket *sock; 624 625 inode = new_inode_pseudo(sock_mnt->mnt_sb); 626 if (!inode) 627 return NULL; 628 629 sock = SOCKET_I(inode); 630 631 inode->i_ino = get_next_ino(); 632 inode->i_mode = S_IFSOCK | S_IRWXUGO; 633 inode->i_uid = current_fsuid(); 634 inode->i_gid = current_fsgid(); 635 inode->i_op = &sockfs_inode_ops; 636 637 return sock; 638 } 639 EXPORT_SYMBOL(sock_alloc); 640 641 static void __sock_release(struct socket *sock, struct inode *inode) 642 { 643 if (sock->ops) { 644 struct module *owner = sock->ops->owner; 645 646 if (inode) 647 inode_lock(inode); 648 sock->ops->release(sock); 649 sock->sk = NULL; 650 if (inode) 651 inode_unlock(inode); 652 sock->ops = NULL; 653 module_put(owner); 654 } 655 656 if (sock->wq.fasync_list) 657 pr_err("%s: fasync list not empty!\n", __func__); 658 659 if (!sock->file) { 660 iput(SOCK_INODE(sock)); 661 return; 662 } 663 sock->file = NULL; 664 } 665 666 /** 667 * sock_release - close a socket 668 * @sock: socket to close 669 * 670 * The socket is released from the protocol stack if it has a release 671 * callback, and the inode is then released if the socket is bound to 672 * an inode not a file. 673 */ 674 void sock_release(struct socket *sock) 675 { 676 __sock_release(sock, NULL); 677 } 678 EXPORT_SYMBOL(sock_release); 679 680 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags) 681 { 682 u8 flags = *tx_flags; 683 684 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) 685 flags |= SKBTX_HW_TSTAMP; 686 687 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE) 688 flags |= SKBTX_SW_TSTAMP; 689 690 if (tsflags & SOF_TIMESTAMPING_TX_SCHED) 691 flags |= SKBTX_SCHED_TSTAMP; 692 693 *tx_flags = flags; 694 } 695 EXPORT_SYMBOL(__sock_tx_timestamp); 696 697 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *, 698 size_t)); 699 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *, 700 size_t)); 701 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) 702 { 703 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg, 704 inet_sendmsg, sock, msg, 705 msg_data_left(msg)); 706 BUG_ON(ret == -EIOCBQUEUED); 707 return ret; 708 } 709 710 /** 711 * sock_sendmsg - send a message through @sock 712 * @sock: socket 713 * @msg: message to send 714 * 715 * Sends @msg through @sock, passing through LSM. 716 * Returns the number of bytes sent, or an error code. 717 */ 718 int sock_sendmsg(struct socket *sock, struct msghdr *msg) 719 { 720 int err = security_socket_sendmsg(sock, msg, 721 msg_data_left(msg)); 722 723 return err ?: sock_sendmsg_nosec(sock, msg); 724 } 725 EXPORT_SYMBOL(sock_sendmsg); 726 727 /** 728 * kernel_sendmsg - send a message through @sock (kernel-space) 729 * @sock: socket 730 * @msg: message header 731 * @vec: kernel vec 732 * @num: vec array length 733 * @size: total message data size 734 * 735 * Builds the message data with @vec and sends it through @sock. 736 * Returns the number of bytes sent, or an error code. 737 */ 738 739 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, 740 struct kvec *vec, size_t num, size_t size) 741 { 742 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size); 743 return sock_sendmsg(sock, msg); 744 } 745 EXPORT_SYMBOL(kernel_sendmsg); 746 747 /** 748 * kernel_sendmsg_locked - send a message through @sock (kernel-space) 749 * @sk: sock 750 * @msg: message header 751 * @vec: output s/g array 752 * @num: output s/g array length 753 * @size: total message data size 754 * 755 * Builds the message data with @vec and sends it through @sock. 756 * Returns the number of bytes sent, or an error code. 757 * Caller must hold @sk. 758 */ 759 760 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg, 761 struct kvec *vec, size_t num, size_t size) 762 { 763 struct socket *sock = sk->sk_socket; 764 765 if (!sock->ops->sendmsg_locked) 766 return sock_no_sendmsg_locked(sk, msg, size); 767 768 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size); 769 770 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg)); 771 } 772 EXPORT_SYMBOL(kernel_sendmsg_locked); 773 774 static bool skb_is_err_queue(const struct sk_buff *skb) 775 { 776 /* pkt_type of skbs enqueued on the error queue are set to 777 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do 778 * in recvmsg, since skbs received on a local socket will never 779 * have a pkt_type of PACKET_OUTGOING. 780 */ 781 return skb->pkt_type == PACKET_OUTGOING; 782 } 783 784 /* On transmit, software and hardware timestamps are returned independently. 785 * As the two skb clones share the hardware timestamp, which may be updated 786 * before the software timestamp is received, a hardware TX timestamp may be 787 * returned only if there is no software TX timestamp. Ignore false software 788 * timestamps, which may be made in the __sock_recv_timestamp() call when the 789 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a 790 * hardware timestamp. 791 */ 792 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp) 793 { 794 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb); 795 } 796 797 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb) 798 { 799 struct scm_ts_pktinfo ts_pktinfo; 800 struct net_device *orig_dev; 801 802 if (!skb_mac_header_was_set(skb)) 803 return; 804 805 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo)); 806 807 rcu_read_lock(); 808 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 809 if (orig_dev) 810 ts_pktinfo.if_index = orig_dev->ifindex; 811 rcu_read_unlock(); 812 813 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb); 814 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO, 815 sizeof(ts_pktinfo), &ts_pktinfo); 816 } 817 818 /* 819 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) 820 */ 821 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 822 struct sk_buff *skb) 823 { 824 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); 825 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW); 826 struct scm_timestamping_internal tss; 827 828 int empty = 1, false_tstamp = 0; 829 struct skb_shared_hwtstamps *shhwtstamps = 830 skb_hwtstamps(skb); 831 832 /* Race occurred between timestamp enabling and packet 833 receiving. Fill in the current time for now. */ 834 if (need_software_tstamp && skb->tstamp == 0) { 835 __net_timestamp(skb); 836 false_tstamp = 1; 837 } 838 839 if (need_software_tstamp) { 840 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { 841 if (new_tstamp) { 842 struct __kernel_sock_timeval tv; 843 844 skb_get_new_timestamp(skb, &tv); 845 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW, 846 sizeof(tv), &tv); 847 } else { 848 struct __kernel_old_timeval tv; 849 850 skb_get_timestamp(skb, &tv); 851 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD, 852 sizeof(tv), &tv); 853 } 854 } else { 855 if (new_tstamp) { 856 struct __kernel_timespec ts; 857 858 skb_get_new_timestampns(skb, &ts); 859 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW, 860 sizeof(ts), &ts); 861 } else { 862 struct __kernel_old_timespec ts; 863 864 skb_get_timestampns(skb, &ts); 865 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD, 866 sizeof(ts), &ts); 867 } 868 } 869 } 870 871 memset(&tss, 0, sizeof(tss)); 872 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) && 873 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0)) 874 empty = 0; 875 if (shhwtstamps && 876 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) && 877 !skb_is_swtx_tstamp(skb, false_tstamp)) { 878 if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC) 879 ptp_convert_timestamp(shhwtstamps, sk->sk_bind_phc); 880 881 if (ktime_to_timespec64_cond(shhwtstamps->hwtstamp, 882 tss.ts + 2)) { 883 empty = 0; 884 885 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) && 886 !skb_is_err_queue(skb)) 887 put_ts_pktinfo(msg, skb); 888 } 889 } 890 if (!empty) { 891 if (sock_flag(sk, SOCK_TSTAMP_NEW)) 892 put_cmsg_scm_timestamping64(msg, &tss); 893 else 894 put_cmsg_scm_timestamping(msg, &tss); 895 896 if (skb_is_err_queue(skb) && skb->len && 897 SKB_EXT_ERR(skb)->opt_stats) 898 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS, 899 skb->len, skb->data); 900 } 901 } 902 EXPORT_SYMBOL_GPL(__sock_recv_timestamp); 903 904 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, 905 struct sk_buff *skb) 906 { 907 int ack; 908 909 if (!sock_flag(sk, SOCK_WIFI_STATUS)) 910 return; 911 if (!skb->wifi_acked_valid) 912 return; 913 914 ack = skb->wifi_acked; 915 916 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); 917 } 918 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); 919 920 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, 921 struct sk_buff *skb) 922 { 923 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount) 924 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, 925 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount); 926 } 927 928 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 929 struct sk_buff *skb) 930 { 931 sock_recv_timestamp(msg, sk, skb); 932 sock_recv_drops(msg, sk, skb); 933 } 934 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops); 935 936 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *, 937 size_t, int)); 938 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *, 939 size_t, int)); 940 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 941 int flags) 942 { 943 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg, 944 inet_recvmsg, sock, msg, msg_data_left(msg), 945 flags); 946 } 947 948 /** 949 * sock_recvmsg - receive a message from @sock 950 * @sock: socket 951 * @msg: message to receive 952 * @flags: message flags 953 * 954 * Receives @msg from @sock, passing through LSM. Returns the total number 955 * of bytes received, or an error. 956 */ 957 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) 958 { 959 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); 960 961 return err ?: sock_recvmsg_nosec(sock, msg, flags); 962 } 963 EXPORT_SYMBOL(sock_recvmsg); 964 965 /** 966 * kernel_recvmsg - Receive a message from a socket (kernel space) 967 * @sock: The socket to receive the message from 968 * @msg: Received message 969 * @vec: Input s/g array for message data 970 * @num: Size of input s/g array 971 * @size: Number of bytes to read 972 * @flags: Message flags (MSG_DONTWAIT, etc...) 973 * 974 * On return the msg structure contains the scatter/gather array passed in the 975 * vec argument. The array is modified so that it consists of the unfilled 976 * portion of the original array. 977 * 978 * The returned value is the total number of bytes received, or an error. 979 */ 980 981 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 982 struct kvec *vec, size_t num, size_t size, int flags) 983 { 984 msg->msg_control_is_user = false; 985 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size); 986 return sock_recvmsg(sock, msg, flags); 987 } 988 EXPORT_SYMBOL(kernel_recvmsg); 989 990 static ssize_t sock_sendpage(struct file *file, struct page *page, 991 int offset, size_t size, loff_t *ppos, int more) 992 { 993 struct socket *sock; 994 int flags; 995 996 sock = file->private_data; 997 998 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 999 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */ 1000 flags |= more; 1001 1002 return kernel_sendpage(sock, page, offset, size, flags); 1003 } 1004 1005 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 1006 struct pipe_inode_info *pipe, size_t len, 1007 unsigned int flags) 1008 { 1009 struct socket *sock = file->private_data; 1010 1011 if (unlikely(!sock->ops->splice_read)) 1012 return generic_file_splice_read(file, ppos, pipe, len, flags); 1013 1014 return sock->ops->splice_read(sock, ppos, pipe, len, flags); 1015 } 1016 1017 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) 1018 { 1019 struct file *file = iocb->ki_filp; 1020 struct socket *sock = file->private_data; 1021 struct msghdr msg = {.msg_iter = *to, 1022 .msg_iocb = iocb}; 1023 ssize_t res; 1024 1025 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1026 msg.msg_flags = MSG_DONTWAIT; 1027 1028 if (iocb->ki_pos != 0) 1029 return -ESPIPE; 1030 1031 if (!iov_iter_count(to)) /* Match SYS5 behaviour */ 1032 return 0; 1033 1034 res = sock_recvmsg(sock, &msg, msg.msg_flags); 1035 *to = msg.msg_iter; 1036 return res; 1037 } 1038 1039 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) 1040 { 1041 struct file *file = iocb->ki_filp; 1042 struct socket *sock = file->private_data; 1043 struct msghdr msg = {.msg_iter = *from, 1044 .msg_iocb = iocb}; 1045 ssize_t res; 1046 1047 if (iocb->ki_pos != 0) 1048 return -ESPIPE; 1049 1050 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1051 msg.msg_flags = MSG_DONTWAIT; 1052 1053 if (sock->type == SOCK_SEQPACKET) 1054 msg.msg_flags |= MSG_EOR; 1055 1056 res = sock_sendmsg(sock, &msg); 1057 *from = msg.msg_iter; 1058 return res; 1059 } 1060 1061 /* 1062 * Atomic setting of ioctl hooks to avoid race 1063 * with module unload. 1064 */ 1065 1066 static DEFINE_MUTEX(br_ioctl_mutex); 1067 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg); 1068 1069 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) 1070 { 1071 mutex_lock(&br_ioctl_mutex); 1072 br_ioctl_hook = hook; 1073 mutex_unlock(&br_ioctl_mutex); 1074 } 1075 EXPORT_SYMBOL(brioctl_set); 1076 1077 static DEFINE_MUTEX(vlan_ioctl_mutex); 1078 static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 1079 1080 void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 1081 { 1082 mutex_lock(&vlan_ioctl_mutex); 1083 vlan_ioctl_hook = hook; 1084 mutex_unlock(&vlan_ioctl_mutex); 1085 } 1086 EXPORT_SYMBOL(vlan_ioctl_set); 1087 1088 static long sock_do_ioctl(struct net *net, struct socket *sock, 1089 unsigned int cmd, unsigned long arg) 1090 { 1091 int err; 1092 void __user *argp = (void __user *)arg; 1093 1094 err = sock->ops->ioctl(sock, cmd, arg); 1095 1096 /* 1097 * If this ioctl is unknown try to hand it down 1098 * to the NIC driver. 1099 */ 1100 if (err != -ENOIOCTLCMD) 1101 return err; 1102 1103 if (cmd == SIOCGIFCONF) { 1104 struct ifconf ifc; 1105 if (copy_from_user(&ifc, argp, sizeof(struct ifconf))) 1106 return -EFAULT; 1107 rtnl_lock(); 1108 err = dev_ifconf(net, &ifc, sizeof(struct ifreq)); 1109 rtnl_unlock(); 1110 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf))) 1111 err = -EFAULT; 1112 } else if (is_socket_ioctl_cmd(cmd)) { 1113 struct ifreq ifr; 1114 bool need_copyout; 1115 if (copy_from_user(&ifr, argp, sizeof(struct ifreq))) 1116 return -EFAULT; 1117 err = dev_ioctl(net, cmd, &ifr, &need_copyout); 1118 if (!err && need_copyout) 1119 if (copy_to_user(argp, &ifr, sizeof(struct ifreq))) 1120 return -EFAULT; 1121 } else { 1122 err = -ENOTTY; 1123 } 1124 return err; 1125 } 1126 1127 /* 1128 * With an ioctl, arg may well be a user mode pointer, but we don't know 1129 * what to do with it - that's up to the protocol still. 1130 */ 1131 1132 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 1133 { 1134 struct socket *sock; 1135 struct sock *sk; 1136 void __user *argp = (void __user *)arg; 1137 int pid, err; 1138 struct net *net; 1139 1140 sock = file->private_data; 1141 sk = sock->sk; 1142 net = sock_net(sk); 1143 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { 1144 struct ifreq ifr; 1145 bool need_copyout; 1146 if (copy_from_user(&ifr, argp, sizeof(struct ifreq))) 1147 return -EFAULT; 1148 err = dev_ioctl(net, cmd, &ifr, &need_copyout); 1149 if (!err && need_copyout) 1150 if (copy_to_user(argp, &ifr, sizeof(struct ifreq))) 1151 return -EFAULT; 1152 } else 1153 #ifdef CONFIG_WEXT_CORE 1154 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 1155 err = wext_handle_ioctl(net, cmd, argp); 1156 } else 1157 #endif 1158 switch (cmd) { 1159 case FIOSETOWN: 1160 case SIOCSPGRP: 1161 err = -EFAULT; 1162 if (get_user(pid, (int __user *)argp)) 1163 break; 1164 err = f_setown(sock->file, pid, 1); 1165 break; 1166 case FIOGETOWN: 1167 case SIOCGPGRP: 1168 err = put_user(f_getown(sock->file), 1169 (int __user *)argp); 1170 break; 1171 case SIOCGIFBR: 1172 case SIOCSIFBR: 1173 case SIOCBRADDBR: 1174 case SIOCBRDELBR: 1175 err = -ENOPKG; 1176 if (!br_ioctl_hook) 1177 request_module("bridge"); 1178 1179 mutex_lock(&br_ioctl_mutex); 1180 if (br_ioctl_hook) 1181 err = br_ioctl_hook(net, cmd, argp); 1182 mutex_unlock(&br_ioctl_mutex); 1183 break; 1184 case SIOCGIFVLAN: 1185 case SIOCSIFVLAN: 1186 err = -ENOPKG; 1187 if (!vlan_ioctl_hook) 1188 request_module("8021q"); 1189 1190 mutex_lock(&vlan_ioctl_mutex); 1191 if (vlan_ioctl_hook) 1192 err = vlan_ioctl_hook(net, argp); 1193 mutex_unlock(&vlan_ioctl_mutex); 1194 break; 1195 case SIOCGSKNS: 1196 err = -EPERM; 1197 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1198 break; 1199 1200 err = open_related_ns(&net->ns, get_net_ns); 1201 break; 1202 case SIOCGSTAMP_OLD: 1203 case SIOCGSTAMPNS_OLD: 1204 if (!sock->ops->gettstamp) { 1205 err = -ENOIOCTLCMD; 1206 break; 1207 } 1208 err = sock->ops->gettstamp(sock, argp, 1209 cmd == SIOCGSTAMP_OLD, 1210 !IS_ENABLED(CONFIG_64BIT)); 1211 break; 1212 case SIOCGSTAMP_NEW: 1213 case SIOCGSTAMPNS_NEW: 1214 if (!sock->ops->gettstamp) { 1215 err = -ENOIOCTLCMD; 1216 break; 1217 } 1218 err = sock->ops->gettstamp(sock, argp, 1219 cmd == SIOCGSTAMP_NEW, 1220 false); 1221 break; 1222 default: 1223 err = sock_do_ioctl(net, sock, cmd, arg); 1224 break; 1225 } 1226 return err; 1227 } 1228 1229 /** 1230 * sock_create_lite - creates a socket 1231 * @family: protocol family (AF_INET, ...) 1232 * @type: communication type (SOCK_STREAM, ...) 1233 * @protocol: protocol (0, ...) 1234 * @res: new socket 1235 * 1236 * Creates a new socket and assigns it to @res, passing through LSM. 1237 * The new socket initialization is not complete, see kernel_accept(). 1238 * Returns 0 or an error. On failure @res is set to %NULL. 1239 * This function internally uses GFP_KERNEL. 1240 */ 1241 1242 int sock_create_lite(int family, int type, int protocol, struct socket **res) 1243 { 1244 int err; 1245 struct socket *sock = NULL; 1246 1247 err = security_socket_create(family, type, protocol, 1); 1248 if (err) 1249 goto out; 1250 1251 sock = sock_alloc(); 1252 if (!sock) { 1253 err = -ENOMEM; 1254 goto out; 1255 } 1256 1257 sock->type = type; 1258 err = security_socket_post_create(sock, family, type, protocol, 1); 1259 if (err) 1260 goto out_release; 1261 1262 out: 1263 *res = sock; 1264 return err; 1265 out_release: 1266 sock_release(sock); 1267 sock = NULL; 1268 goto out; 1269 } 1270 EXPORT_SYMBOL(sock_create_lite); 1271 1272 /* No kernel lock held - perfect */ 1273 static __poll_t sock_poll(struct file *file, poll_table *wait) 1274 { 1275 struct socket *sock = file->private_data; 1276 __poll_t events = poll_requested_events(wait), flag = 0; 1277 1278 if (!sock->ops->poll) 1279 return 0; 1280 1281 if (sk_can_busy_loop(sock->sk)) { 1282 /* poll once if requested by the syscall */ 1283 if (events & POLL_BUSY_LOOP) 1284 sk_busy_loop(sock->sk, 1); 1285 1286 /* if this socket can poll_ll, tell the system call */ 1287 flag = POLL_BUSY_LOOP; 1288 } 1289 1290 return sock->ops->poll(file, sock, wait) | flag; 1291 } 1292 1293 static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1294 { 1295 struct socket *sock = file->private_data; 1296 1297 return sock->ops->mmap(file, sock, vma); 1298 } 1299 1300 static int sock_close(struct inode *inode, struct file *filp) 1301 { 1302 __sock_release(SOCKET_I(inode), inode); 1303 return 0; 1304 } 1305 1306 /* 1307 * Update the socket async list 1308 * 1309 * Fasync_list locking strategy. 1310 * 1311 * 1. fasync_list is modified only under process context socket lock 1312 * i.e. under semaphore. 1313 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1314 * or under socket lock 1315 */ 1316 1317 static int sock_fasync(int fd, struct file *filp, int on) 1318 { 1319 struct socket *sock = filp->private_data; 1320 struct sock *sk = sock->sk; 1321 struct socket_wq *wq = &sock->wq; 1322 1323 if (sk == NULL) 1324 return -EINVAL; 1325 1326 lock_sock(sk); 1327 fasync_helper(fd, filp, on, &wq->fasync_list); 1328 1329 if (!wq->fasync_list) 1330 sock_reset_flag(sk, SOCK_FASYNC); 1331 else 1332 sock_set_flag(sk, SOCK_FASYNC); 1333 1334 release_sock(sk); 1335 return 0; 1336 } 1337 1338 /* This function may be called only under rcu_lock */ 1339 1340 int sock_wake_async(struct socket_wq *wq, int how, int band) 1341 { 1342 if (!wq || !wq->fasync_list) 1343 return -1; 1344 1345 switch (how) { 1346 case SOCK_WAKE_WAITD: 1347 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) 1348 break; 1349 goto call_kill; 1350 case SOCK_WAKE_SPACE: 1351 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) 1352 break; 1353 fallthrough; 1354 case SOCK_WAKE_IO: 1355 call_kill: 1356 kill_fasync(&wq->fasync_list, SIGIO, band); 1357 break; 1358 case SOCK_WAKE_URG: 1359 kill_fasync(&wq->fasync_list, SIGURG, band); 1360 } 1361 1362 return 0; 1363 } 1364 EXPORT_SYMBOL(sock_wake_async); 1365 1366 /** 1367 * __sock_create - creates a socket 1368 * @net: net namespace 1369 * @family: protocol family (AF_INET, ...) 1370 * @type: communication type (SOCK_STREAM, ...) 1371 * @protocol: protocol (0, ...) 1372 * @res: new socket 1373 * @kern: boolean for kernel space sockets 1374 * 1375 * Creates a new socket and assigns it to @res, passing through LSM. 1376 * Returns 0 or an error. On failure @res is set to %NULL. @kern must 1377 * be set to true if the socket resides in kernel space. 1378 * This function internally uses GFP_KERNEL. 1379 */ 1380 1381 int __sock_create(struct net *net, int family, int type, int protocol, 1382 struct socket **res, int kern) 1383 { 1384 int err; 1385 struct socket *sock; 1386 const struct net_proto_family *pf; 1387 1388 /* 1389 * Check protocol is in range 1390 */ 1391 if (family < 0 || family >= NPROTO) 1392 return -EAFNOSUPPORT; 1393 if (type < 0 || type >= SOCK_MAX) 1394 return -EINVAL; 1395 1396 /* Compatibility. 1397 1398 This uglymoron is moved from INET layer to here to avoid 1399 deadlock in module load. 1400 */ 1401 if (family == PF_INET && type == SOCK_PACKET) { 1402 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1403 current->comm); 1404 family = PF_PACKET; 1405 } 1406 1407 err = security_socket_create(family, type, protocol, kern); 1408 if (err) 1409 return err; 1410 1411 /* 1412 * Allocate the socket and allow the family to set things up. if 1413 * the protocol is 0, the family is instructed to select an appropriate 1414 * default. 1415 */ 1416 sock = sock_alloc(); 1417 if (!sock) { 1418 net_warn_ratelimited("socket: no more sockets\n"); 1419 return -ENFILE; /* Not exactly a match, but its the 1420 closest posix thing */ 1421 } 1422 1423 sock->type = type; 1424 1425 #ifdef CONFIG_MODULES 1426 /* Attempt to load a protocol module if the find failed. 1427 * 1428 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1429 * requested real, full-featured networking support upon configuration. 1430 * Otherwise module support will break! 1431 */ 1432 if (rcu_access_pointer(net_families[family]) == NULL) 1433 request_module("net-pf-%d", family); 1434 #endif 1435 1436 rcu_read_lock(); 1437 pf = rcu_dereference(net_families[family]); 1438 err = -EAFNOSUPPORT; 1439 if (!pf) 1440 goto out_release; 1441 1442 /* 1443 * We will call the ->create function, that possibly is in a loadable 1444 * module, so we have to bump that loadable module refcnt first. 1445 */ 1446 if (!try_module_get(pf->owner)) 1447 goto out_release; 1448 1449 /* Now protected by module ref count */ 1450 rcu_read_unlock(); 1451 1452 err = pf->create(net, sock, protocol, kern); 1453 if (err < 0) 1454 goto out_module_put; 1455 1456 /* 1457 * Now to bump the refcnt of the [loadable] module that owns this 1458 * socket at sock_release time we decrement its refcnt. 1459 */ 1460 if (!try_module_get(sock->ops->owner)) 1461 goto out_module_busy; 1462 1463 /* 1464 * Now that we're done with the ->create function, the [loadable] 1465 * module can have its refcnt decremented 1466 */ 1467 module_put(pf->owner); 1468 err = security_socket_post_create(sock, family, type, protocol, kern); 1469 if (err) 1470 goto out_sock_release; 1471 *res = sock; 1472 1473 return 0; 1474 1475 out_module_busy: 1476 err = -EAFNOSUPPORT; 1477 out_module_put: 1478 sock->ops = NULL; 1479 module_put(pf->owner); 1480 out_sock_release: 1481 sock_release(sock); 1482 return err; 1483 1484 out_release: 1485 rcu_read_unlock(); 1486 goto out_sock_release; 1487 } 1488 EXPORT_SYMBOL(__sock_create); 1489 1490 /** 1491 * sock_create - creates a socket 1492 * @family: protocol family (AF_INET, ...) 1493 * @type: communication type (SOCK_STREAM, ...) 1494 * @protocol: protocol (0, ...) 1495 * @res: new socket 1496 * 1497 * A wrapper around __sock_create(). 1498 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1499 */ 1500 1501 int sock_create(int family, int type, int protocol, struct socket **res) 1502 { 1503 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1504 } 1505 EXPORT_SYMBOL(sock_create); 1506 1507 /** 1508 * sock_create_kern - creates a socket (kernel space) 1509 * @net: net namespace 1510 * @family: protocol family (AF_INET, ...) 1511 * @type: communication type (SOCK_STREAM, ...) 1512 * @protocol: protocol (0, ...) 1513 * @res: new socket 1514 * 1515 * A wrapper around __sock_create(). 1516 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1517 */ 1518 1519 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) 1520 { 1521 return __sock_create(net, family, type, protocol, res, 1); 1522 } 1523 EXPORT_SYMBOL(sock_create_kern); 1524 1525 int __sys_socket(int family, int type, int protocol) 1526 { 1527 int retval; 1528 struct socket *sock; 1529 int flags; 1530 1531 /* Check the SOCK_* constants for consistency. */ 1532 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1533 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1534 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1535 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1536 1537 flags = type & ~SOCK_TYPE_MASK; 1538 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1539 return -EINVAL; 1540 type &= SOCK_TYPE_MASK; 1541 1542 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1543 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1544 1545 retval = sock_create(family, type, protocol, &sock); 1546 if (retval < 0) 1547 return retval; 1548 1549 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1550 } 1551 1552 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1553 { 1554 return __sys_socket(family, type, protocol); 1555 } 1556 1557 /* 1558 * Create a pair of connected sockets. 1559 */ 1560 1561 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) 1562 { 1563 struct socket *sock1, *sock2; 1564 int fd1, fd2, err; 1565 struct file *newfile1, *newfile2; 1566 int flags; 1567 1568 flags = type & ~SOCK_TYPE_MASK; 1569 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1570 return -EINVAL; 1571 type &= SOCK_TYPE_MASK; 1572 1573 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1574 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1575 1576 /* 1577 * reserve descriptors and make sure we won't fail 1578 * to return them to userland. 1579 */ 1580 fd1 = get_unused_fd_flags(flags); 1581 if (unlikely(fd1 < 0)) 1582 return fd1; 1583 1584 fd2 = get_unused_fd_flags(flags); 1585 if (unlikely(fd2 < 0)) { 1586 put_unused_fd(fd1); 1587 return fd2; 1588 } 1589 1590 err = put_user(fd1, &usockvec[0]); 1591 if (err) 1592 goto out; 1593 1594 err = put_user(fd2, &usockvec[1]); 1595 if (err) 1596 goto out; 1597 1598 /* 1599 * Obtain the first socket and check if the underlying protocol 1600 * supports the socketpair call. 1601 */ 1602 1603 err = sock_create(family, type, protocol, &sock1); 1604 if (unlikely(err < 0)) 1605 goto out; 1606 1607 err = sock_create(family, type, protocol, &sock2); 1608 if (unlikely(err < 0)) { 1609 sock_release(sock1); 1610 goto out; 1611 } 1612 1613 err = security_socket_socketpair(sock1, sock2); 1614 if (unlikely(err)) { 1615 sock_release(sock2); 1616 sock_release(sock1); 1617 goto out; 1618 } 1619 1620 err = sock1->ops->socketpair(sock1, sock2); 1621 if (unlikely(err < 0)) { 1622 sock_release(sock2); 1623 sock_release(sock1); 1624 goto out; 1625 } 1626 1627 newfile1 = sock_alloc_file(sock1, flags, NULL); 1628 if (IS_ERR(newfile1)) { 1629 err = PTR_ERR(newfile1); 1630 sock_release(sock2); 1631 goto out; 1632 } 1633 1634 newfile2 = sock_alloc_file(sock2, flags, NULL); 1635 if (IS_ERR(newfile2)) { 1636 err = PTR_ERR(newfile2); 1637 fput(newfile1); 1638 goto out; 1639 } 1640 1641 audit_fd_pair(fd1, fd2); 1642 1643 fd_install(fd1, newfile1); 1644 fd_install(fd2, newfile2); 1645 return 0; 1646 1647 out: 1648 put_unused_fd(fd2); 1649 put_unused_fd(fd1); 1650 return err; 1651 } 1652 1653 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1654 int __user *, usockvec) 1655 { 1656 return __sys_socketpair(family, type, protocol, usockvec); 1657 } 1658 1659 /* 1660 * Bind a name to a socket. Nothing much to do here since it's 1661 * the protocol's responsibility to handle the local address. 1662 * 1663 * We move the socket address to kernel space before we call 1664 * the protocol layer (having also checked the address is ok). 1665 */ 1666 1667 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) 1668 { 1669 struct socket *sock; 1670 struct sockaddr_storage address; 1671 int err, fput_needed; 1672 1673 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1674 if (sock) { 1675 err = move_addr_to_kernel(umyaddr, addrlen, &address); 1676 if (!err) { 1677 err = security_socket_bind(sock, 1678 (struct sockaddr *)&address, 1679 addrlen); 1680 if (!err) 1681 err = sock->ops->bind(sock, 1682 (struct sockaddr *) 1683 &address, addrlen); 1684 } 1685 fput_light(sock->file, fput_needed); 1686 } 1687 return err; 1688 } 1689 1690 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1691 { 1692 return __sys_bind(fd, umyaddr, addrlen); 1693 } 1694 1695 /* 1696 * Perform a listen. Basically, we allow the protocol to do anything 1697 * necessary for a listen, and if that works, we mark the socket as 1698 * ready for listening. 1699 */ 1700 1701 int __sys_listen(int fd, int backlog) 1702 { 1703 struct socket *sock; 1704 int err, fput_needed; 1705 int somaxconn; 1706 1707 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1708 if (sock) { 1709 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; 1710 if ((unsigned int)backlog > somaxconn) 1711 backlog = somaxconn; 1712 1713 err = security_socket_listen(sock, backlog); 1714 if (!err) 1715 err = sock->ops->listen(sock, backlog); 1716 1717 fput_light(sock->file, fput_needed); 1718 } 1719 return err; 1720 } 1721 1722 SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1723 { 1724 return __sys_listen(fd, backlog); 1725 } 1726 1727 int __sys_accept4_file(struct file *file, unsigned file_flags, 1728 struct sockaddr __user *upeer_sockaddr, 1729 int __user *upeer_addrlen, int flags, 1730 unsigned long nofile) 1731 { 1732 struct socket *sock, *newsock; 1733 struct file *newfile; 1734 int err, len, newfd; 1735 struct sockaddr_storage address; 1736 1737 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1738 return -EINVAL; 1739 1740 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1741 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1742 1743 sock = sock_from_file(file); 1744 if (!sock) { 1745 err = -ENOTSOCK; 1746 goto out; 1747 } 1748 1749 err = -ENFILE; 1750 newsock = sock_alloc(); 1751 if (!newsock) 1752 goto out; 1753 1754 newsock->type = sock->type; 1755 newsock->ops = sock->ops; 1756 1757 /* 1758 * We don't need try_module_get here, as the listening socket (sock) 1759 * has the protocol module (sock->ops->owner) held. 1760 */ 1761 __module_get(newsock->ops->owner); 1762 1763 newfd = __get_unused_fd_flags(flags, nofile); 1764 if (unlikely(newfd < 0)) { 1765 err = newfd; 1766 sock_release(newsock); 1767 goto out; 1768 } 1769 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); 1770 if (IS_ERR(newfile)) { 1771 err = PTR_ERR(newfile); 1772 put_unused_fd(newfd); 1773 goto out; 1774 } 1775 1776 err = security_socket_accept(sock, newsock); 1777 if (err) 1778 goto out_fd; 1779 1780 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags, 1781 false); 1782 if (err < 0) 1783 goto out_fd; 1784 1785 if (upeer_sockaddr) { 1786 len = newsock->ops->getname(newsock, 1787 (struct sockaddr *)&address, 2); 1788 if (len < 0) { 1789 err = -ECONNABORTED; 1790 goto out_fd; 1791 } 1792 err = move_addr_to_user(&address, 1793 len, upeer_sockaddr, upeer_addrlen); 1794 if (err < 0) 1795 goto out_fd; 1796 } 1797 1798 /* File flags are not inherited via accept() unlike another OSes. */ 1799 1800 fd_install(newfd, newfile); 1801 err = newfd; 1802 out: 1803 return err; 1804 out_fd: 1805 fput(newfile); 1806 put_unused_fd(newfd); 1807 goto out; 1808 1809 } 1810 1811 /* 1812 * For accept, we attempt to create a new socket, set up the link 1813 * with the client, wake up the client, then return the new 1814 * connected fd. We collect the address of the connector in kernel 1815 * space and move it to user at the very end. This is unclean because 1816 * we open the socket then return an error. 1817 * 1818 * 1003.1g adds the ability to recvmsg() to query connection pending 1819 * status to recvmsg. We need to add that support in a way thats 1820 * clean when we restructure accept also. 1821 */ 1822 1823 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, 1824 int __user *upeer_addrlen, int flags) 1825 { 1826 int ret = -EBADF; 1827 struct fd f; 1828 1829 f = fdget(fd); 1830 if (f.file) { 1831 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr, 1832 upeer_addrlen, flags, 1833 rlimit(RLIMIT_NOFILE)); 1834 fdput(f); 1835 } 1836 1837 return ret; 1838 } 1839 1840 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 1841 int __user *, upeer_addrlen, int, flags) 1842 { 1843 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); 1844 } 1845 1846 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 1847 int __user *, upeer_addrlen) 1848 { 1849 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 1850 } 1851 1852 /* 1853 * Attempt to connect to a socket with the server address. The address 1854 * is in user space so we verify it is OK and move it to kernel space. 1855 * 1856 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 1857 * break bindings 1858 * 1859 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 1860 * other SEQPACKET protocols that take time to connect() as it doesn't 1861 * include the -EINPROGRESS status for such sockets. 1862 */ 1863 1864 int __sys_connect_file(struct file *file, struct sockaddr_storage *address, 1865 int addrlen, int file_flags) 1866 { 1867 struct socket *sock; 1868 int err; 1869 1870 sock = sock_from_file(file); 1871 if (!sock) { 1872 err = -ENOTSOCK; 1873 goto out; 1874 } 1875 1876 err = 1877 security_socket_connect(sock, (struct sockaddr *)address, addrlen); 1878 if (err) 1879 goto out; 1880 1881 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen, 1882 sock->file->f_flags | file_flags); 1883 out: 1884 return err; 1885 } 1886 1887 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) 1888 { 1889 int ret = -EBADF; 1890 struct fd f; 1891 1892 f = fdget(fd); 1893 if (f.file) { 1894 struct sockaddr_storage address; 1895 1896 ret = move_addr_to_kernel(uservaddr, addrlen, &address); 1897 if (!ret) 1898 ret = __sys_connect_file(f.file, &address, addrlen, 0); 1899 fdput(f); 1900 } 1901 1902 return ret; 1903 } 1904 1905 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 1906 int, addrlen) 1907 { 1908 return __sys_connect(fd, uservaddr, addrlen); 1909 } 1910 1911 /* 1912 * Get the local address ('name') of a socket object. Move the obtained 1913 * name to user space. 1914 */ 1915 1916 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, 1917 int __user *usockaddr_len) 1918 { 1919 struct socket *sock; 1920 struct sockaddr_storage address; 1921 int err, fput_needed; 1922 1923 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1924 if (!sock) 1925 goto out; 1926 1927 err = security_socket_getsockname(sock); 1928 if (err) 1929 goto out_put; 1930 1931 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0); 1932 if (err < 0) 1933 goto out_put; 1934 /* "err" is actually length in this case */ 1935 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); 1936 1937 out_put: 1938 fput_light(sock->file, fput_needed); 1939 out: 1940 return err; 1941 } 1942 1943 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 1944 int __user *, usockaddr_len) 1945 { 1946 return __sys_getsockname(fd, usockaddr, usockaddr_len); 1947 } 1948 1949 /* 1950 * Get the remote address ('name') of a socket object. Move the obtained 1951 * name to user space. 1952 */ 1953 1954 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, 1955 int __user *usockaddr_len) 1956 { 1957 struct socket *sock; 1958 struct sockaddr_storage address; 1959 int err, fput_needed; 1960 1961 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1962 if (sock != NULL) { 1963 err = security_socket_getpeername(sock); 1964 if (err) { 1965 fput_light(sock->file, fput_needed); 1966 return err; 1967 } 1968 1969 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1); 1970 if (err >= 0) 1971 /* "err" is actually length in this case */ 1972 err = move_addr_to_user(&address, err, usockaddr, 1973 usockaddr_len); 1974 fput_light(sock->file, fput_needed); 1975 } 1976 return err; 1977 } 1978 1979 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 1980 int __user *, usockaddr_len) 1981 { 1982 return __sys_getpeername(fd, usockaddr, usockaddr_len); 1983 } 1984 1985 /* 1986 * Send a datagram to a given address. We move the address into kernel 1987 * space and check the user space data area is readable before invoking 1988 * the protocol. 1989 */ 1990 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, 1991 struct sockaddr __user *addr, int addr_len) 1992 { 1993 struct socket *sock; 1994 struct sockaddr_storage address; 1995 int err; 1996 struct msghdr msg; 1997 struct iovec iov; 1998 int fput_needed; 1999 2000 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter); 2001 if (unlikely(err)) 2002 return err; 2003 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2004 if (!sock) 2005 goto out; 2006 2007 msg.msg_name = NULL; 2008 msg.msg_control = NULL; 2009 msg.msg_controllen = 0; 2010 msg.msg_namelen = 0; 2011 if (addr) { 2012 err = move_addr_to_kernel(addr, addr_len, &address); 2013 if (err < 0) 2014 goto out_put; 2015 msg.msg_name = (struct sockaddr *)&address; 2016 msg.msg_namelen = addr_len; 2017 } 2018 if (sock->file->f_flags & O_NONBLOCK) 2019 flags |= MSG_DONTWAIT; 2020 msg.msg_flags = flags; 2021 err = sock_sendmsg(sock, &msg); 2022 2023 out_put: 2024 fput_light(sock->file, fput_needed); 2025 out: 2026 return err; 2027 } 2028 2029 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 2030 unsigned int, flags, struct sockaddr __user *, addr, 2031 int, addr_len) 2032 { 2033 return __sys_sendto(fd, buff, len, flags, addr, addr_len); 2034 } 2035 2036 /* 2037 * Send a datagram down a socket. 2038 */ 2039 2040 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 2041 unsigned int, flags) 2042 { 2043 return __sys_sendto(fd, buff, len, flags, NULL, 0); 2044 } 2045 2046 /* 2047 * Receive a frame from the socket and optionally record the address of the 2048 * sender. We verify the buffers are writable and if needed move the 2049 * sender address from kernel to user space. 2050 */ 2051 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, 2052 struct sockaddr __user *addr, int __user *addr_len) 2053 { 2054 struct socket *sock; 2055 struct iovec iov; 2056 struct msghdr msg; 2057 struct sockaddr_storage address; 2058 int err, err2; 2059 int fput_needed; 2060 2061 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter); 2062 if (unlikely(err)) 2063 return err; 2064 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2065 if (!sock) 2066 goto out; 2067 2068 msg.msg_control = NULL; 2069 msg.msg_controllen = 0; 2070 /* Save some cycles and don't copy the address if not needed */ 2071 msg.msg_name = addr ? (struct sockaddr *)&address : NULL; 2072 /* We assume all kernel code knows the size of sockaddr_storage */ 2073 msg.msg_namelen = 0; 2074 msg.msg_iocb = NULL; 2075 msg.msg_flags = 0; 2076 if (sock->file->f_flags & O_NONBLOCK) 2077 flags |= MSG_DONTWAIT; 2078 err = sock_recvmsg(sock, &msg, flags); 2079 2080 if (err >= 0 && addr != NULL) { 2081 err2 = move_addr_to_user(&address, 2082 msg.msg_namelen, addr, addr_len); 2083 if (err2 < 0) 2084 err = err2; 2085 } 2086 2087 fput_light(sock->file, fput_needed); 2088 out: 2089 return err; 2090 } 2091 2092 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 2093 unsigned int, flags, struct sockaddr __user *, addr, 2094 int __user *, addr_len) 2095 { 2096 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); 2097 } 2098 2099 /* 2100 * Receive a datagram from a socket. 2101 */ 2102 2103 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, 2104 unsigned int, flags) 2105 { 2106 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 2107 } 2108 2109 static bool sock_use_custom_sol_socket(const struct socket *sock) 2110 { 2111 const struct sock *sk = sock->sk; 2112 2113 /* Use sock->ops->setsockopt() for MPTCP */ 2114 return IS_ENABLED(CONFIG_MPTCP) && 2115 sk->sk_protocol == IPPROTO_MPTCP && 2116 sk->sk_type == SOCK_STREAM && 2117 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6); 2118 } 2119 2120 /* 2121 * Set a socket option. Because we don't know the option lengths we have 2122 * to pass the user mode parameter for the protocols to sort out. 2123 */ 2124 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval, 2125 int optlen) 2126 { 2127 sockptr_t optval = USER_SOCKPTR(user_optval); 2128 char *kernel_optval = NULL; 2129 int err, fput_needed; 2130 struct socket *sock; 2131 2132 if (optlen < 0) 2133 return -EINVAL; 2134 2135 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2136 if (!sock) 2137 return err; 2138 2139 err = security_socket_setsockopt(sock, level, optname); 2140 if (err) 2141 goto out_put; 2142 2143 if (!in_compat_syscall()) 2144 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname, 2145 user_optval, &optlen, 2146 &kernel_optval); 2147 if (err < 0) 2148 goto out_put; 2149 if (err > 0) { 2150 err = 0; 2151 goto out_put; 2152 } 2153 2154 if (kernel_optval) 2155 optval = KERNEL_SOCKPTR(kernel_optval); 2156 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock)) 2157 err = sock_setsockopt(sock, level, optname, optval, optlen); 2158 else if (unlikely(!sock->ops->setsockopt)) 2159 err = -EOPNOTSUPP; 2160 else 2161 err = sock->ops->setsockopt(sock, level, optname, optval, 2162 optlen); 2163 kfree(kernel_optval); 2164 out_put: 2165 fput_light(sock->file, fput_needed); 2166 return err; 2167 } 2168 2169 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 2170 char __user *, optval, int, optlen) 2171 { 2172 return __sys_setsockopt(fd, level, optname, optval, optlen); 2173 } 2174 2175 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level, 2176 int optname)); 2177 2178 /* 2179 * Get a socket option. Because we don't know the option lengths we have 2180 * to pass a user mode parameter for the protocols to sort out. 2181 */ 2182 int __sys_getsockopt(int fd, int level, int optname, char __user *optval, 2183 int __user *optlen) 2184 { 2185 int err, fput_needed; 2186 struct socket *sock; 2187 int max_optlen; 2188 2189 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2190 if (!sock) 2191 return err; 2192 2193 err = security_socket_getsockopt(sock, level, optname); 2194 if (err) 2195 goto out_put; 2196 2197 if (!in_compat_syscall()) 2198 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen); 2199 2200 if (level == SOL_SOCKET) 2201 err = sock_getsockopt(sock, level, optname, optval, optlen); 2202 else if (unlikely(!sock->ops->getsockopt)) 2203 err = -EOPNOTSUPP; 2204 else 2205 err = sock->ops->getsockopt(sock, level, optname, optval, 2206 optlen); 2207 2208 if (!in_compat_syscall()) 2209 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname, 2210 optval, optlen, max_optlen, 2211 err); 2212 out_put: 2213 fput_light(sock->file, fput_needed); 2214 return err; 2215 } 2216 2217 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 2218 char __user *, optval, int __user *, optlen) 2219 { 2220 return __sys_getsockopt(fd, level, optname, optval, optlen); 2221 } 2222 2223 /* 2224 * Shutdown a socket. 2225 */ 2226 2227 int __sys_shutdown_sock(struct socket *sock, int how) 2228 { 2229 int err; 2230 2231 err = security_socket_shutdown(sock, how); 2232 if (!err) 2233 err = sock->ops->shutdown(sock, how); 2234 2235 return err; 2236 } 2237 2238 int __sys_shutdown(int fd, int how) 2239 { 2240 int err, fput_needed; 2241 struct socket *sock; 2242 2243 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2244 if (sock != NULL) { 2245 err = __sys_shutdown_sock(sock, how); 2246 fput_light(sock->file, fput_needed); 2247 } 2248 return err; 2249 } 2250 2251 SYSCALL_DEFINE2(shutdown, int, fd, int, how) 2252 { 2253 return __sys_shutdown(fd, how); 2254 } 2255 2256 /* A couple of helpful macros for getting the address of the 32/64 bit 2257 * fields which are the same type (int / unsigned) on our platforms. 2258 */ 2259 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 2260 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 2261 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 2262 2263 struct used_address { 2264 struct sockaddr_storage name; 2265 unsigned int name_len; 2266 }; 2267 2268 int __copy_msghdr_from_user(struct msghdr *kmsg, 2269 struct user_msghdr __user *umsg, 2270 struct sockaddr __user **save_addr, 2271 struct iovec __user **uiov, size_t *nsegs) 2272 { 2273 struct user_msghdr msg; 2274 ssize_t err; 2275 2276 if (copy_from_user(&msg, umsg, sizeof(*umsg))) 2277 return -EFAULT; 2278 2279 kmsg->msg_control_is_user = true; 2280 kmsg->msg_control_user = msg.msg_control; 2281 kmsg->msg_controllen = msg.msg_controllen; 2282 kmsg->msg_flags = msg.msg_flags; 2283 2284 kmsg->msg_namelen = msg.msg_namelen; 2285 if (!msg.msg_name) 2286 kmsg->msg_namelen = 0; 2287 2288 if (kmsg->msg_namelen < 0) 2289 return -EINVAL; 2290 2291 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) 2292 kmsg->msg_namelen = sizeof(struct sockaddr_storage); 2293 2294 if (save_addr) 2295 *save_addr = msg.msg_name; 2296 2297 if (msg.msg_name && kmsg->msg_namelen) { 2298 if (!save_addr) { 2299 err = move_addr_to_kernel(msg.msg_name, 2300 kmsg->msg_namelen, 2301 kmsg->msg_name); 2302 if (err < 0) 2303 return err; 2304 } 2305 } else { 2306 kmsg->msg_name = NULL; 2307 kmsg->msg_namelen = 0; 2308 } 2309 2310 if (msg.msg_iovlen > UIO_MAXIOV) 2311 return -EMSGSIZE; 2312 2313 kmsg->msg_iocb = NULL; 2314 *uiov = msg.msg_iov; 2315 *nsegs = msg.msg_iovlen; 2316 return 0; 2317 } 2318 2319 static int copy_msghdr_from_user(struct msghdr *kmsg, 2320 struct user_msghdr __user *umsg, 2321 struct sockaddr __user **save_addr, 2322 struct iovec **iov) 2323 { 2324 struct user_msghdr msg; 2325 ssize_t err; 2326 2327 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov, 2328 &msg.msg_iovlen); 2329 if (err) 2330 return err; 2331 2332 err = import_iovec(save_addr ? READ : WRITE, 2333 msg.msg_iov, msg.msg_iovlen, 2334 UIO_FASTIOV, iov, &kmsg->msg_iter); 2335 return err < 0 ? err : 0; 2336 } 2337 2338 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, 2339 unsigned int flags, struct used_address *used_address, 2340 unsigned int allowed_msghdr_flags) 2341 { 2342 unsigned char ctl[sizeof(struct cmsghdr) + 20] 2343 __aligned(sizeof(__kernel_size_t)); 2344 /* 20 is size of ipv6_pktinfo */ 2345 unsigned char *ctl_buf = ctl; 2346 int ctl_len; 2347 ssize_t err; 2348 2349 err = -ENOBUFS; 2350 2351 if (msg_sys->msg_controllen > INT_MAX) 2352 goto out; 2353 flags |= (msg_sys->msg_flags & allowed_msghdr_flags); 2354 ctl_len = msg_sys->msg_controllen; 2355 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 2356 err = 2357 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, 2358 sizeof(ctl)); 2359 if (err) 2360 goto out; 2361 ctl_buf = msg_sys->msg_control; 2362 ctl_len = msg_sys->msg_controllen; 2363 } else if (ctl_len) { 2364 BUILD_BUG_ON(sizeof(struct cmsghdr) != 2365 CMSG_ALIGN(sizeof(struct cmsghdr))); 2366 if (ctl_len > sizeof(ctl)) { 2367 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 2368 if (ctl_buf == NULL) 2369 goto out; 2370 } 2371 err = -EFAULT; 2372 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len)) 2373 goto out_freectl; 2374 msg_sys->msg_control = ctl_buf; 2375 msg_sys->msg_control_is_user = false; 2376 } 2377 msg_sys->msg_flags = flags; 2378 2379 if (sock->file->f_flags & O_NONBLOCK) 2380 msg_sys->msg_flags |= MSG_DONTWAIT; 2381 /* 2382 * If this is sendmmsg() and current destination address is same as 2383 * previously succeeded address, omit asking LSM's decision. 2384 * used_address->name_len is initialized to UINT_MAX so that the first 2385 * destination address never matches. 2386 */ 2387 if (used_address && msg_sys->msg_name && 2388 used_address->name_len == msg_sys->msg_namelen && 2389 !memcmp(&used_address->name, msg_sys->msg_name, 2390 used_address->name_len)) { 2391 err = sock_sendmsg_nosec(sock, msg_sys); 2392 goto out_freectl; 2393 } 2394 err = sock_sendmsg(sock, msg_sys); 2395 /* 2396 * If this is sendmmsg() and sending to current destination address was 2397 * successful, remember it. 2398 */ 2399 if (used_address && err >= 0) { 2400 used_address->name_len = msg_sys->msg_namelen; 2401 if (msg_sys->msg_name) 2402 memcpy(&used_address->name, msg_sys->msg_name, 2403 used_address->name_len); 2404 } 2405 2406 out_freectl: 2407 if (ctl_buf != ctl) 2408 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 2409 out: 2410 return err; 2411 } 2412 2413 int sendmsg_copy_msghdr(struct msghdr *msg, 2414 struct user_msghdr __user *umsg, unsigned flags, 2415 struct iovec **iov) 2416 { 2417 int err; 2418 2419 if (flags & MSG_CMSG_COMPAT) { 2420 struct compat_msghdr __user *msg_compat; 2421 2422 msg_compat = (struct compat_msghdr __user *) umsg; 2423 err = get_compat_msghdr(msg, msg_compat, NULL, iov); 2424 } else { 2425 err = copy_msghdr_from_user(msg, umsg, NULL, iov); 2426 } 2427 if (err < 0) 2428 return err; 2429 2430 return 0; 2431 } 2432 2433 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, 2434 struct msghdr *msg_sys, unsigned int flags, 2435 struct used_address *used_address, 2436 unsigned int allowed_msghdr_flags) 2437 { 2438 struct sockaddr_storage address; 2439 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2440 ssize_t err; 2441 2442 msg_sys->msg_name = &address; 2443 2444 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov); 2445 if (err < 0) 2446 return err; 2447 2448 err = ____sys_sendmsg(sock, msg_sys, flags, used_address, 2449 allowed_msghdr_flags); 2450 kfree(iov); 2451 return err; 2452 } 2453 2454 /* 2455 * BSD sendmsg interface 2456 */ 2457 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg, 2458 unsigned int flags) 2459 { 2460 return ____sys_sendmsg(sock, msg, flags, NULL, 0); 2461 } 2462 2463 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2464 bool forbid_cmsg_compat) 2465 { 2466 int fput_needed, err; 2467 struct msghdr msg_sys; 2468 struct socket *sock; 2469 2470 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2471 return -EINVAL; 2472 2473 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2474 if (!sock) 2475 goto out; 2476 2477 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); 2478 2479 fput_light(sock->file, fput_needed); 2480 out: 2481 return err; 2482 } 2483 2484 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) 2485 { 2486 return __sys_sendmsg(fd, msg, flags, true); 2487 } 2488 2489 /* 2490 * Linux sendmmsg interface 2491 */ 2492 2493 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2494 unsigned int flags, bool forbid_cmsg_compat) 2495 { 2496 int fput_needed, err, datagrams; 2497 struct socket *sock; 2498 struct mmsghdr __user *entry; 2499 struct compat_mmsghdr __user *compat_entry; 2500 struct msghdr msg_sys; 2501 struct used_address used_address; 2502 unsigned int oflags = flags; 2503 2504 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2505 return -EINVAL; 2506 2507 if (vlen > UIO_MAXIOV) 2508 vlen = UIO_MAXIOV; 2509 2510 datagrams = 0; 2511 2512 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2513 if (!sock) 2514 return err; 2515 2516 used_address.name_len = UINT_MAX; 2517 entry = mmsg; 2518 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2519 err = 0; 2520 flags |= MSG_BATCH; 2521 2522 while (datagrams < vlen) { 2523 if (datagrams == vlen - 1) 2524 flags = oflags; 2525 2526 if (MSG_CMSG_COMPAT & flags) { 2527 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, 2528 &msg_sys, flags, &used_address, MSG_EOR); 2529 if (err < 0) 2530 break; 2531 err = __put_user(err, &compat_entry->msg_len); 2532 ++compat_entry; 2533 } else { 2534 err = ___sys_sendmsg(sock, 2535 (struct user_msghdr __user *)entry, 2536 &msg_sys, flags, &used_address, MSG_EOR); 2537 if (err < 0) 2538 break; 2539 err = put_user(err, &entry->msg_len); 2540 ++entry; 2541 } 2542 2543 if (err) 2544 break; 2545 ++datagrams; 2546 if (msg_data_left(&msg_sys)) 2547 break; 2548 cond_resched(); 2549 } 2550 2551 fput_light(sock->file, fput_needed); 2552 2553 /* We only return an error if no datagrams were able to be sent */ 2554 if (datagrams != 0) 2555 return datagrams; 2556 2557 return err; 2558 } 2559 2560 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, 2561 unsigned int, vlen, unsigned int, flags) 2562 { 2563 return __sys_sendmmsg(fd, mmsg, vlen, flags, true); 2564 } 2565 2566 int recvmsg_copy_msghdr(struct msghdr *msg, 2567 struct user_msghdr __user *umsg, unsigned flags, 2568 struct sockaddr __user **uaddr, 2569 struct iovec **iov) 2570 { 2571 ssize_t err; 2572 2573 if (MSG_CMSG_COMPAT & flags) { 2574 struct compat_msghdr __user *msg_compat; 2575 2576 msg_compat = (struct compat_msghdr __user *) umsg; 2577 err = get_compat_msghdr(msg, msg_compat, uaddr, iov); 2578 } else { 2579 err = copy_msghdr_from_user(msg, umsg, uaddr, iov); 2580 } 2581 if (err < 0) 2582 return err; 2583 2584 return 0; 2585 } 2586 2587 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys, 2588 struct user_msghdr __user *msg, 2589 struct sockaddr __user *uaddr, 2590 unsigned int flags, int nosec) 2591 { 2592 struct compat_msghdr __user *msg_compat = 2593 (struct compat_msghdr __user *) msg; 2594 int __user *uaddr_len = COMPAT_NAMELEN(msg); 2595 struct sockaddr_storage addr; 2596 unsigned long cmsg_ptr; 2597 int len; 2598 ssize_t err; 2599 2600 msg_sys->msg_name = &addr; 2601 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2602 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2603 2604 /* We assume all kernel code knows the size of sockaddr_storage */ 2605 msg_sys->msg_namelen = 0; 2606 2607 if (sock->file->f_flags & O_NONBLOCK) 2608 flags |= MSG_DONTWAIT; 2609 2610 if (unlikely(nosec)) 2611 err = sock_recvmsg_nosec(sock, msg_sys, flags); 2612 else 2613 err = sock_recvmsg(sock, msg_sys, flags); 2614 2615 if (err < 0) 2616 goto out; 2617 len = err; 2618 2619 if (uaddr != NULL) { 2620 err = move_addr_to_user(&addr, 2621 msg_sys->msg_namelen, uaddr, 2622 uaddr_len); 2623 if (err < 0) 2624 goto out; 2625 } 2626 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2627 COMPAT_FLAGS(msg)); 2628 if (err) 2629 goto out; 2630 if (MSG_CMSG_COMPAT & flags) 2631 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2632 &msg_compat->msg_controllen); 2633 else 2634 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2635 &msg->msg_controllen); 2636 if (err) 2637 goto out; 2638 err = len; 2639 out: 2640 return err; 2641 } 2642 2643 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, 2644 struct msghdr *msg_sys, unsigned int flags, int nosec) 2645 { 2646 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2647 /* user mode address pointers */ 2648 struct sockaddr __user *uaddr; 2649 ssize_t err; 2650 2651 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov); 2652 if (err < 0) 2653 return err; 2654 2655 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec); 2656 kfree(iov); 2657 return err; 2658 } 2659 2660 /* 2661 * BSD recvmsg interface 2662 */ 2663 2664 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg, 2665 struct user_msghdr __user *umsg, 2666 struct sockaddr __user *uaddr, unsigned int flags) 2667 { 2668 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0); 2669 } 2670 2671 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2672 bool forbid_cmsg_compat) 2673 { 2674 int fput_needed, err; 2675 struct msghdr msg_sys; 2676 struct socket *sock; 2677 2678 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2679 return -EINVAL; 2680 2681 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2682 if (!sock) 2683 goto out; 2684 2685 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2686 2687 fput_light(sock->file, fput_needed); 2688 out: 2689 return err; 2690 } 2691 2692 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, 2693 unsigned int, flags) 2694 { 2695 return __sys_recvmsg(fd, msg, flags, true); 2696 } 2697 2698 /* 2699 * Linux recvmmsg interface 2700 */ 2701 2702 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2703 unsigned int vlen, unsigned int flags, 2704 struct timespec64 *timeout) 2705 { 2706 int fput_needed, err, datagrams; 2707 struct socket *sock; 2708 struct mmsghdr __user *entry; 2709 struct compat_mmsghdr __user *compat_entry; 2710 struct msghdr msg_sys; 2711 struct timespec64 end_time; 2712 struct timespec64 timeout64; 2713 2714 if (timeout && 2715 poll_select_set_timeout(&end_time, timeout->tv_sec, 2716 timeout->tv_nsec)) 2717 return -EINVAL; 2718 2719 datagrams = 0; 2720 2721 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2722 if (!sock) 2723 return err; 2724 2725 if (likely(!(flags & MSG_ERRQUEUE))) { 2726 err = sock_error(sock->sk); 2727 if (err) { 2728 datagrams = err; 2729 goto out_put; 2730 } 2731 } 2732 2733 entry = mmsg; 2734 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2735 2736 while (datagrams < vlen) { 2737 /* 2738 * No need to ask LSM for more than the first datagram. 2739 */ 2740 if (MSG_CMSG_COMPAT & flags) { 2741 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, 2742 &msg_sys, flags & ~MSG_WAITFORONE, 2743 datagrams); 2744 if (err < 0) 2745 break; 2746 err = __put_user(err, &compat_entry->msg_len); 2747 ++compat_entry; 2748 } else { 2749 err = ___sys_recvmsg(sock, 2750 (struct user_msghdr __user *)entry, 2751 &msg_sys, flags & ~MSG_WAITFORONE, 2752 datagrams); 2753 if (err < 0) 2754 break; 2755 err = put_user(err, &entry->msg_len); 2756 ++entry; 2757 } 2758 2759 if (err) 2760 break; 2761 ++datagrams; 2762 2763 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2764 if (flags & MSG_WAITFORONE) 2765 flags |= MSG_DONTWAIT; 2766 2767 if (timeout) { 2768 ktime_get_ts64(&timeout64); 2769 *timeout = timespec64_sub(end_time, timeout64); 2770 if (timeout->tv_sec < 0) { 2771 timeout->tv_sec = timeout->tv_nsec = 0; 2772 break; 2773 } 2774 2775 /* Timeout, return less than vlen datagrams */ 2776 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2777 break; 2778 } 2779 2780 /* Out of band data, return right away */ 2781 if (msg_sys.msg_flags & MSG_OOB) 2782 break; 2783 cond_resched(); 2784 } 2785 2786 if (err == 0) 2787 goto out_put; 2788 2789 if (datagrams == 0) { 2790 datagrams = err; 2791 goto out_put; 2792 } 2793 2794 /* 2795 * We may return less entries than requested (vlen) if the 2796 * sock is non block and there aren't enough datagrams... 2797 */ 2798 if (err != -EAGAIN) { 2799 /* 2800 * ... or if recvmsg returns an error after we 2801 * received some datagrams, where we record the 2802 * error to return on the next call or if the 2803 * app asks about it using getsockopt(SO_ERROR). 2804 */ 2805 sock->sk->sk_err = -err; 2806 } 2807 out_put: 2808 fput_light(sock->file, fput_needed); 2809 2810 return datagrams; 2811 } 2812 2813 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2814 unsigned int vlen, unsigned int flags, 2815 struct __kernel_timespec __user *timeout, 2816 struct old_timespec32 __user *timeout32) 2817 { 2818 int datagrams; 2819 struct timespec64 timeout_sys; 2820 2821 if (timeout && get_timespec64(&timeout_sys, timeout)) 2822 return -EFAULT; 2823 2824 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32)) 2825 return -EFAULT; 2826 2827 if (!timeout && !timeout32) 2828 return do_recvmmsg(fd, mmsg, vlen, flags, NULL); 2829 2830 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 2831 2832 if (datagrams <= 0) 2833 return datagrams; 2834 2835 if (timeout && put_timespec64(&timeout_sys, timeout)) 2836 datagrams = -EFAULT; 2837 2838 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32)) 2839 datagrams = -EFAULT; 2840 2841 return datagrams; 2842 } 2843 2844 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 2845 unsigned int, vlen, unsigned int, flags, 2846 struct __kernel_timespec __user *, timeout) 2847 { 2848 if (flags & MSG_CMSG_COMPAT) 2849 return -EINVAL; 2850 2851 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL); 2852 } 2853 2854 #ifdef CONFIG_COMPAT_32BIT_TIME 2855 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg, 2856 unsigned int, vlen, unsigned int, flags, 2857 struct old_timespec32 __user *, timeout) 2858 { 2859 if (flags & MSG_CMSG_COMPAT) 2860 return -EINVAL; 2861 2862 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout); 2863 } 2864 #endif 2865 2866 #ifdef __ARCH_WANT_SYS_SOCKETCALL 2867 /* Argument list sizes for sys_socketcall */ 2868 #define AL(x) ((x) * sizeof(unsigned long)) 2869 static const unsigned char nargs[21] = { 2870 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 2871 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 2872 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 2873 AL(4), AL(5), AL(4) 2874 }; 2875 2876 #undef AL 2877 2878 /* 2879 * System call vectors. 2880 * 2881 * Argument checking cleaned up. Saved 20% in size. 2882 * This function doesn't need to set the kernel lock because 2883 * it is set by the callees. 2884 */ 2885 2886 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 2887 { 2888 unsigned long a[AUDITSC_ARGS]; 2889 unsigned long a0, a1; 2890 int err; 2891 unsigned int len; 2892 2893 if (call < 1 || call > SYS_SENDMMSG) 2894 return -EINVAL; 2895 call = array_index_nospec(call, SYS_SENDMMSG + 1); 2896 2897 len = nargs[call]; 2898 if (len > sizeof(a)) 2899 return -EINVAL; 2900 2901 /* copy_from_user should be SMP safe. */ 2902 if (copy_from_user(a, args, len)) 2903 return -EFAULT; 2904 2905 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); 2906 if (err) 2907 return err; 2908 2909 a0 = a[0]; 2910 a1 = a[1]; 2911 2912 switch (call) { 2913 case SYS_SOCKET: 2914 err = __sys_socket(a0, a1, a[2]); 2915 break; 2916 case SYS_BIND: 2917 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 2918 break; 2919 case SYS_CONNECT: 2920 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 2921 break; 2922 case SYS_LISTEN: 2923 err = __sys_listen(a0, a1); 2924 break; 2925 case SYS_ACCEPT: 2926 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 2927 (int __user *)a[2], 0); 2928 break; 2929 case SYS_GETSOCKNAME: 2930 err = 2931 __sys_getsockname(a0, (struct sockaddr __user *)a1, 2932 (int __user *)a[2]); 2933 break; 2934 case SYS_GETPEERNAME: 2935 err = 2936 __sys_getpeername(a0, (struct sockaddr __user *)a1, 2937 (int __user *)a[2]); 2938 break; 2939 case SYS_SOCKETPAIR: 2940 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 2941 break; 2942 case SYS_SEND: 2943 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 2944 NULL, 0); 2945 break; 2946 case SYS_SENDTO: 2947 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 2948 (struct sockaddr __user *)a[4], a[5]); 2949 break; 2950 case SYS_RECV: 2951 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 2952 NULL, NULL); 2953 break; 2954 case SYS_RECVFROM: 2955 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 2956 (struct sockaddr __user *)a[4], 2957 (int __user *)a[5]); 2958 break; 2959 case SYS_SHUTDOWN: 2960 err = __sys_shutdown(a0, a1); 2961 break; 2962 case SYS_SETSOCKOPT: 2963 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], 2964 a[4]); 2965 break; 2966 case SYS_GETSOCKOPT: 2967 err = 2968 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 2969 (int __user *)a[4]); 2970 break; 2971 case SYS_SENDMSG: 2972 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, 2973 a[2], true); 2974 break; 2975 case SYS_SENDMMSG: 2976 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], 2977 a[3], true); 2978 break; 2979 case SYS_RECVMSG: 2980 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, 2981 a[2], true); 2982 break; 2983 case SYS_RECVMMSG: 2984 if (IS_ENABLED(CONFIG_64BIT)) 2985 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 2986 a[2], a[3], 2987 (struct __kernel_timespec __user *)a[4], 2988 NULL); 2989 else 2990 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 2991 a[2], a[3], NULL, 2992 (struct old_timespec32 __user *)a[4]); 2993 break; 2994 case SYS_ACCEPT4: 2995 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 2996 (int __user *)a[2], a[3]); 2997 break; 2998 default: 2999 err = -EINVAL; 3000 break; 3001 } 3002 return err; 3003 } 3004 3005 #endif /* __ARCH_WANT_SYS_SOCKETCALL */ 3006 3007 /** 3008 * sock_register - add a socket protocol handler 3009 * @ops: description of protocol 3010 * 3011 * This function is called by a protocol handler that wants to 3012 * advertise its address family, and have it linked into the 3013 * socket interface. The value ops->family corresponds to the 3014 * socket system call protocol family. 3015 */ 3016 int sock_register(const struct net_proto_family *ops) 3017 { 3018 int err; 3019 3020 if (ops->family >= NPROTO) { 3021 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); 3022 return -ENOBUFS; 3023 } 3024 3025 spin_lock(&net_family_lock); 3026 if (rcu_dereference_protected(net_families[ops->family], 3027 lockdep_is_held(&net_family_lock))) 3028 err = -EEXIST; 3029 else { 3030 rcu_assign_pointer(net_families[ops->family], ops); 3031 err = 0; 3032 } 3033 spin_unlock(&net_family_lock); 3034 3035 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]); 3036 return err; 3037 } 3038 EXPORT_SYMBOL(sock_register); 3039 3040 /** 3041 * sock_unregister - remove a protocol handler 3042 * @family: protocol family to remove 3043 * 3044 * This function is called by a protocol handler that wants to 3045 * remove its address family, and have it unlinked from the 3046 * new socket creation. 3047 * 3048 * If protocol handler is a module, then it can use module reference 3049 * counts to protect against new references. If protocol handler is not 3050 * a module then it needs to provide its own protection in 3051 * the ops->create routine. 3052 */ 3053 void sock_unregister(int family) 3054 { 3055 BUG_ON(family < 0 || family >= NPROTO); 3056 3057 spin_lock(&net_family_lock); 3058 RCU_INIT_POINTER(net_families[family], NULL); 3059 spin_unlock(&net_family_lock); 3060 3061 synchronize_rcu(); 3062 3063 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]); 3064 } 3065 EXPORT_SYMBOL(sock_unregister); 3066 3067 bool sock_is_registered(int family) 3068 { 3069 return family < NPROTO && rcu_access_pointer(net_families[family]); 3070 } 3071 3072 static int __init sock_init(void) 3073 { 3074 int err; 3075 /* 3076 * Initialize the network sysctl infrastructure. 3077 */ 3078 err = net_sysctl_init(); 3079 if (err) 3080 goto out; 3081 3082 /* 3083 * Initialize skbuff SLAB cache 3084 */ 3085 skb_init(); 3086 3087 /* 3088 * Initialize the protocols module. 3089 */ 3090 3091 init_inodecache(); 3092 3093 err = register_filesystem(&sock_fs_type); 3094 if (err) 3095 goto out; 3096 sock_mnt = kern_mount(&sock_fs_type); 3097 if (IS_ERR(sock_mnt)) { 3098 err = PTR_ERR(sock_mnt); 3099 goto out_mount; 3100 } 3101 3102 /* The real protocol initialization is performed in later initcalls. 3103 */ 3104 3105 #ifdef CONFIG_NETFILTER 3106 err = netfilter_init(); 3107 if (err) 3108 goto out; 3109 #endif 3110 3111 ptp_classifier_init(); 3112 3113 out: 3114 return err; 3115 3116 out_mount: 3117 unregister_filesystem(&sock_fs_type); 3118 goto out; 3119 } 3120 3121 core_initcall(sock_init); /* early initcall */ 3122 3123 #ifdef CONFIG_PROC_FS 3124 void socket_seq_show(struct seq_file *seq) 3125 { 3126 seq_printf(seq, "sockets: used %d\n", 3127 sock_inuse_get(seq->private)); 3128 } 3129 #endif /* CONFIG_PROC_FS */ 3130 3131 #ifdef CONFIG_COMPAT 3132 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32) 3133 { 3134 struct compat_ifconf ifc32; 3135 struct ifconf ifc; 3136 int err; 3137 3138 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf))) 3139 return -EFAULT; 3140 3141 ifc.ifc_len = ifc32.ifc_len; 3142 ifc.ifc_req = compat_ptr(ifc32.ifcbuf); 3143 3144 rtnl_lock(); 3145 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq)); 3146 rtnl_unlock(); 3147 if (err) 3148 return err; 3149 3150 ifc32.ifc_len = ifc.ifc_len; 3151 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf))) 3152 return -EFAULT; 3153 3154 return 0; 3155 } 3156 3157 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32) 3158 { 3159 struct compat_ethtool_rxnfc __user *compat_rxnfc; 3160 bool convert_in = false, convert_out = false; 3161 size_t buf_size = 0; 3162 struct ethtool_rxnfc __user *rxnfc = NULL; 3163 struct ifreq ifr; 3164 u32 rule_cnt = 0, actual_rule_cnt; 3165 u32 ethcmd; 3166 u32 data; 3167 int ret; 3168 3169 if (get_user(data, &ifr32->ifr_ifru.ifru_data)) 3170 return -EFAULT; 3171 3172 compat_rxnfc = compat_ptr(data); 3173 3174 if (get_user(ethcmd, &compat_rxnfc->cmd)) 3175 return -EFAULT; 3176 3177 /* Most ethtool structures are defined without padding. 3178 * Unfortunately struct ethtool_rxnfc is an exception. 3179 */ 3180 switch (ethcmd) { 3181 default: 3182 break; 3183 case ETHTOOL_GRXCLSRLALL: 3184 /* Buffer size is variable */ 3185 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt)) 3186 return -EFAULT; 3187 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32)) 3188 return -ENOMEM; 3189 buf_size += rule_cnt * sizeof(u32); 3190 fallthrough; 3191 case ETHTOOL_GRXRINGS: 3192 case ETHTOOL_GRXCLSRLCNT: 3193 case ETHTOOL_GRXCLSRULE: 3194 case ETHTOOL_SRXCLSRLINS: 3195 convert_out = true; 3196 fallthrough; 3197 case ETHTOOL_SRXCLSRLDEL: 3198 buf_size += sizeof(struct ethtool_rxnfc); 3199 convert_in = true; 3200 rxnfc = compat_alloc_user_space(buf_size); 3201 break; 3202 } 3203 3204 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ)) 3205 return -EFAULT; 3206 3207 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc; 3208 3209 if (convert_in) { 3210 /* We expect there to be holes between fs.m_ext and 3211 * fs.ring_cookie and at the end of fs, but nowhere else. 3212 */ 3213 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) + 3214 sizeof(compat_rxnfc->fs.m_ext) != 3215 offsetof(struct ethtool_rxnfc, fs.m_ext) + 3216 sizeof(rxnfc->fs.m_ext)); 3217 BUILD_BUG_ON( 3218 offsetof(struct compat_ethtool_rxnfc, fs.location) - 3219 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) != 3220 offsetof(struct ethtool_rxnfc, fs.location) - 3221 offsetof(struct ethtool_rxnfc, fs.ring_cookie)); 3222 3223 if (copy_in_user(rxnfc, compat_rxnfc, 3224 (void __user *)(&rxnfc->fs.m_ext + 1) - 3225 (void __user *)rxnfc) || 3226 copy_in_user(&rxnfc->fs.ring_cookie, 3227 &compat_rxnfc->fs.ring_cookie, 3228 (void __user *)(&rxnfc->fs.location + 1) - 3229 (void __user *)&rxnfc->fs.ring_cookie)) 3230 return -EFAULT; 3231 if (ethcmd == ETHTOOL_GRXCLSRLALL) { 3232 if (put_user(rule_cnt, &rxnfc->rule_cnt)) 3233 return -EFAULT; 3234 } else if (copy_in_user(&rxnfc->rule_cnt, 3235 &compat_rxnfc->rule_cnt, 3236 sizeof(rxnfc->rule_cnt))) 3237 return -EFAULT; 3238 } 3239 3240 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL); 3241 if (ret) 3242 return ret; 3243 3244 if (convert_out) { 3245 if (copy_in_user(compat_rxnfc, rxnfc, 3246 (const void __user *)(&rxnfc->fs.m_ext + 1) - 3247 (const void __user *)rxnfc) || 3248 copy_in_user(&compat_rxnfc->fs.ring_cookie, 3249 &rxnfc->fs.ring_cookie, 3250 (const void __user *)(&rxnfc->fs.location + 1) - 3251 (const void __user *)&rxnfc->fs.ring_cookie) || 3252 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt, 3253 sizeof(rxnfc->rule_cnt))) 3254 return -EFAULT; 3255 3256 if (ethcmd == ETHTOOL_GRXCLSRLALL) { 3257 /* As an optimisation, we only copy the actual 3258 * number of rules that the underlying 3259 * function returned. Since Mallory might 3260 * change the rule count in user memory, we 3261 * check that it is less than the rule count 3262 * originally given (as the user buffer size), 3263 * which has been range-checked. 3264 */ 3265 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt)) 3266 return -EFAULT; 3267 if (actual_rule_cnt < rule_cnt) 3268 rule_cnt = actual_rule_cnt; 3269 if (copy_in_user(&compat_rxnfc->rule_locs[0], 3270 &rxnfc->rule_locs[0], 3271 rule_cnt * sizeof(u32))) 3272 return -EFAULT; 3273 } 3274 } 3275 3276 return 0; 3277 } 3278 3279 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 3280 { 3281 compat_uptr_t uptr32; 3282 struct ifreq ifr; 3283 void __user *saved; 3284 int err; 3285 3286 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq))) 3287 return -EFAULT; 3288 3289 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 3290 return -EFAULT; 3291 3292 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; 3293 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); 3294 3295 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL); 3296 if (!err) { 3297 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; 3298 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq))) 3299 err = -EFAULT; 3300 } 3301 return err; 3302 } 3303 3304 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ 3305 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, 3306 struct compat_ifreq __user *u_ifreq32) 3307 { 3308 struct ifreq ifreq; 3309 u32 data32; 3310 3311 if (!is_socket_ioctl_cmd(cmd)) 3312 return -ENOTTY; 3313 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ)) 3314 return -EFAULT; 3315 if (get_user(data32, &u_ifreq32->ifr_data)) 3316 return -EFAULT; 3317 ifreq.ifr_data = compat_ptr(data32); 3318 3319 return dev_ioctl(net, cmd, &ifreq, NULL); 3320 } 3321 3322 static int compat_ifreq_ioctl(struct net *net, struct socket *sock, 3323 unsigned int cmd, 3324 struct compat_ifreq __user *uifr32) 3325 { 3326 struct ifreq __user *uifr; 3327 int err; 3328 3329 /* Handle the fact that while struct ifreq has the same *layout* on 3330 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data, 3331 * which are handled elsewhere, it still has different *size* due to 3332 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit, 3333 * resulting in struct ifreq being 32 and 40 bytes respectively). 3334 * As a result, if the struct happens to be at the end of a page and 3335 * the next page isn't readable/writable, we get a fault. To prevent 3336 * that, copy back and forth to the full size. 3337 */ 3338 3339 uifr = compat_alloc_user_space(sizeof(*uifr)); 3340 if (copy_in_user(uifr, uifr32, sizeof(*uifr32))) 3341 return -EFAULT; 3342 3343 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr); 3344 3345 if (!err) { 3346 switch (cmd) { 3347 case SIOCGIFFLAGS: 3348 case SIOCGIFMETRIC: 3349 case SIOCGIFMTU: 3350 case SIOCGIFMEM: 3351 case SIOCGIFHWADDR: 3352 case SIOCGIFINDEX: 3353 case SIOCGIFADDR: 3354 case SIOCGIFBRDADDR: 3355 case SIOCGIFDSTADDR: 3356 case SIOCGIFNETMASK: 3357 case SIOCGIFPFLAGS: 3358 case SIOCGIFTXQLEN: 3359 case SIOCGMIIPHY: 3360 case SIOCGMIIREG: 3361 case SIOCGIFNAME: 3362 if (copy_in_user(uifr32, uifr, sizeof(*uifr32))) 3363 err = -EFAULT; 3364 break; 3365 } 3366 } 3367 return err; 3368 } 3369 3370 static int compat_sioc_ifmap(struct net *net, unsigned int cmd, 3371 struct compat_ifreq __user *uifr32) 3372 { 3373 struct ifreq ifr; 3374 struct compat_ifmap __user *uifmap32; 3375 int err; 3376 3377 uifmap32 = &uifr32->ifr_ifru.ifru_map; 3378 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name)); 3379 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 3380 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 3381 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 3382 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq); 3383 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma); 3384 err |= get_user(ifr.ifr_map.port, &uifmap32->port); 3385 if (err) 3386 return -EFAULT; 3387 3388 err = dev_ioctl(net, cmd, &ifr, NULL); 3389 3390 if (cmd == SIOCGIFMAP && !err) { 3391 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name)); 3392 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 3393 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 3394 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 3395 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq); 3396 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma); 3397 err |= put_user(ifr.ifr_map.port, &uifmap32->port); 3398 if (err) 3399 err = -EFAULT; 3400 } 3401 return err; 3402 } 3403 3404 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE 3405 * for some operations; this forces use of the newer bridge-utils that 3406 * use compatible ioctls 3407 */ 3408 static int old_bridge_ioctl(compat_ulong_t __user *argp) 3409 { 3410 compat_ulong_t tmp; 3411 3412 if (get_user(tmp, argp)) 3413 return -EFAULT; 3414 if (tmp == BRCTL_GET_VERSION) 3415 return BRCTL_VERSION + 1; 3416 return -EINVAL; 3417 } 3418 3419 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 3420 unsigned int cmd, unsigned long arg) 3421 { 3422 void __user *argp = compat_ptr(arg); 3423 struct sock *sk = sock->sk; 3424 struct net *net = sock_net(sk); 3425 3426 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 3427 return compat_ifr_data_ioctl(net, cmd, argp); 3428 3429 switch (cmd) { 3430 case SIOCSIFBR: 3431 case SIOCGIFBR: 3432 return old_bridge_ioctl(argp); 3433 case SIOCGIFCONF: 3434 return compat_dev_ifconf(net, argp); 3435 case SIOCETHTOOL: 3436 return ethtool_ioctl(net, argp); 3437 case SIOCWANDEV: 3438 return compat_siocwandev(net, argp); 3439 case SIOCGIFMAP: 3440 case SIOCSIFMAP: 3441 return compat_sioc_ifmap(net, cmd, argp); 3442 case SIOCGSTAMP_OLD: 3443 case SIOCGSTAMPNS_OLD: 3444 if (!sock->ops->gettstamp) 3445 return -ENOIOCTLCMD; 3446 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD, 3447 !COMPAT_USE_64BIT_TIME); 3448 3449 case SIOCBONDSLAVEINFOQUERY: 3450 case SIOCBONDINFOQUERY: 3451 case SIOCSHWTSTAMP: 3452 case SIOCGHWTSTAMP: 3453 return compat_ifr_data_ioctl(net, cmd, argp); 3454 3455 case FIOSETOWN: 3456 case SIOCSPGRP: 3457 case FIOGETOWN: 3458 case SIOCGPGRP: 3459 case SIOCBRADDBR: 3460 case SIOCBRDELBR: 3461 case SIOCGIFVLAN: 3462 case SIOCSIFVLAN: 3463 case SIOCGSKNS: 3464 case SIOCGSTAMP_NEW: 3465 case SIOCGSTAMPNS_NEW: 3466 return sock_ioctl(file, cmd, arg); 3467 3468 case SIOCGIFFLAGS: 3469 case SIOCSIFFLAGS: 3470 case SIOCGIFMETRIC: 3471 case SIOCSIFMETRIC: 3472 case SIOCGIFMTU: 3473 case SIOCSIFMTU: 3474 case SIOCGIFMEM: 3475 case SIOCSIFMEM: 3476 case SIOCGIFHWADDR: 3477 case SIOCSIFHWADDR: 3478 case SIOCADDMULTI: 3479 case SIOCDELMULTI: 3480 case SIOCGIFINDEX: 3481 case SIOCGIFADDR: 3482 case SIOCSIFADDR: 3483 case SIOCSIFHWBROADCAST: 3484 case SIOCDIFADDR: 3485 case SIOCGIFBRDADDR: 3486 case SIOCSIFBRDADDR: 3487 case SIOCGIFDSTADDR: 3488 case SIOCSIFDSTADDR: 3489 case SIOCGIFNETMASK: 3490 case SIOCSIFNETMASK: 3491 case SIOCSIFPFLAGS: 3492 case SIOCGIFPFLAGS: 3493 case SIOCGIFTXQLEN: 3494 case SIOCSIFTXQLEN: 3495 case SIOCBRADDIF: 3496 case SIOCBRDELIF: 3497 case SIOCGIFNAME: 3498 case SIOCSIFNAME: 3499 case SIOCGMIIPHY: 3500 case SIOCGMIIREG: 3501 case SIOCSMIIREG: 3502 case SIOCBONDENSLAVE: 3503 case SIOCBONDRELEASE: 3504 case SIOCBONDSETHWADDR: 3505 case SIOCBONDCHANGEACTIVE: 3506 return compat_ifreq_ioctl(net, sock, cmd, argp); 3507 3508 case SIOCSARP: 3509 case SIOCGARP: 3510 case SIOCDARP: 3511 case SIOCOUTQ: 3512 case SIOCOUTQNSD: 3513 case SIOCATMARK: 3514 return sock_do_ioctl(net, sock, cmd, arg); 3515 } 3516 3517 return -ENOIOCTLCMD; 3518 } 3519 3520 static long compat_sock_ioctl(struct file *file, unsigned int cmd, 3521 unsigned long arg) 3522 { 3523 struct socket *sock = file->private_data; 3524 int ret = -ENOIOCTLCMD; 3525 struct sock *sk; 3526 struct net *net; 3527 3528 sk = sock->sk; 3529 net = sock_net(sk); 3530 3531 if (sock->ops->compat_ioctl) 3532 ret = sock->ops->compat_ioctl(sock, cmd, arg); 3533 3534 if (ret == -ENOIOCTLCMD && 3535 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 3536 ret = compat_wext_handle_ioctl(net, cmd, arg); 3537 3538 if (ret == -ENOIOCTLCMD) 3539 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 3540 3541 return ret; 3542 } 3543 #endif 3544 3545 /** 3546 * kernel_bind - bind an address to a socket (kernel space) 3547 * @sock: socket 3548 * @addr: address 3549 * @addrlen: length of address 3550 * 3551 * Returns 0 or an error. 3552 */ 3553 3554 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3555 { 3556 return sock->ops->bind(sock, addr, addrlen); 3557 } 3558 EXPORT_SYMBOL(kernel_bind); 3559 3560 /** 3561 * kernel_listen - move socket to listening state (kernel space) 3562 * @sock: socket 3563 * @backlog: pending connections queue size 3564 * 3565 * Returns 0 or an error. 3566 */ 3567 3568 int kernel_listen(struct socket *sock, int backlog) 3569 { 3570 return sock->ops->listen(sock, backlog); 3571 } 3572 EXPORT_SYMBOL(kernel_listen); 3573 3574 /** 3575 * kernel_accept - accept a connection (kernel space) 3576 * @sock: listening socket 3577 * @newsock: new connected socket 3578 * @flags: flags 3579 * 3580 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0. 3581 * If it fails, @newsock is guaranteed to be %NULL. 3582 * Returns 0 or an error. 3583 */ 3584 3585 int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3586 { 3587 struct sock *sk = sock->sk; 3588 int err; 3589 3590 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3591 newsock); 3592 if (err < 0) 3593 goto done; 3594 3595 err = sock->ops->accept(sock, *newsock, flags, true); 3596 if (err < 0) { 3597 sock_release(*newsock); 3598 *newsock = NULL; 3599 goto done; 3600 } 3601 3602 (*newsock)->ops = sock->ops; 3603 __module_get((*newsock)->ops->owner); 3604 3605 done: 3606 return err; 3607 } 3608 EXPORT_SYMBOL(kernel_accept); 3609 3610 /** 3611 * kernel_connect - connect a socket (kernel space) 3612 * @sock: socket 3613 * @addr: address 3614 * @addrlen: address length 3615 * @flags: flags (O_NONBLOCK, ...) 3616 * 3617 * For datagram sockets, @addr is the address to which datagrams are sent 3618 * by default, and the only address from which datagrams are received. 3619 * For stream sockets, attempts to connect to @addr. 3620 * Returns 0 or an error code. 3621 */ 3622 3623 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3624 int flags) 3625 { 3626 return sock->ops->connect(sock, addr, addrlen, flags); 3627 } 3628 EXPORT_SYMBOL(kernel_connect); 3629 3630 /** 3631 * kernel_getsockname - get the address which the socket is bound (kernel space) 3632 * @sock: socket 3633 * @addr: address holder 3634 * 3635 * Fills the @addr pointer with the address which the socket is bound. 3636 * Returns 0 or an error code. 3637 */ 3638 3639 int kernel_getsockname(struct socket *sock, struct sockaddr *addr) 3640 { 3641 return sock->ops->getname(sock, addr, 0); 3642 } 3643 EXPORT_SYMBOL(kernel_getsockname); 3644 3645 /** 3646 * kernel_getpeername - get the address which the socket is connected (kernel space) 3647 * @sock: socket 3648 * @addr: address holder 3649 * 3650 * Fills the @addr pointer with the address which the socket is connected. 3651 * Returns 0 or an error code. 3652 */ 3653 3654 int kernel_getpeername(struct socket *sock, struct sockaddr *addr) 3655 { 3656 return sock->ops->getname(sock, addr, 1); 3657 } 3658 EXPORT_SYMBOL(kernel_getpeername); 3659 3660 /** 3661 * kernel_sendpage - send a &page through a socket (kernel space) 3662 * @sock: socket 3663 * @page: page 3664 * @offset: page offset 3665 * @size: total size in bytes 3666 * @flags: flags (MSG_DONTWAIT, ...) 3667 * 3668 * Returns the total amount sent in bytes or an error. 3669 */ 3670 3671 int kernel_sendpage(struct socket *sock, struct page *page, int offset, 3672 size_t size, int flags) 3673 { 3674 if (sock->ops->sendpage) { 3675 /* Warn in case the improper page to zero-copy send */ 3676 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send"); 3677 return sock->ops->sendpage(sock, page, offset, size, flags); 3678 } 3679 return sock_no_sendpage(sock, page, offset, size, flags); 3680 } 3681 EXPORT_SYMBOL(kernel_sendpage); 3682 3683 /** 3684 * kernel_sendpage_locked - send a &page through the locked sock (kernel space) 3685 * @sk: sock 3686 * @page: page 3687 * @offset: page offset 3688 * @size: total size in bytes 3689 * @flags: flags (MSG_DONTWAIT, ...) 3690 * 3691 * Returns the total amount sent in bytes or an error. 3692 * Caller must hold @sk. 3693 */ 3694 3695 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset, 3696 size_t size, int flags) 3697 { 3698 struct socket *sock = sk->sk_socket; 3699 3700 if (sock->ops->sendpage_locked) 3701 return sock->ops->sendpage_locked(sk, page, offset, size, 3702 flags); 3703 3704 return sock_no_sendpage_locked(sk, page, offset, size, flags); 3705 } 3706 EXPORT_SYMBOL(kernel_sendpage_locked); 3707 3708 /** 3709 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space) 3710 * @sock: socket 3711 * @how: connection part 3712 * 3713 * Returns 0 or an error. 3714 */ 3715 3716 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3717 { 3718 return sock->ops->shutdown(sock, how); 3719 } 3720 EXPORT_SYMBOL(kernel_sock_shutdown); 3721 3722 /** 3723 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket 3724 * @sk: socket 3725 * 3726 * This routine returns the IP overhead imposed by a socket i.e. 3727 * the length of the underlying IP header, depending on whether 3728 * this is an IPv4 or IPv6 socket and the length from IP options turned 3729 * on at the socket. Assumes that the caller has a lock on the socket. 3730 */ 3731 3732 u32 kernel_sock_ip_overhead(struct sock *sk) 3733 { 3734 struct inet_sock *inet; 3735 struct ip_options_rcu *opt; 3736 u32 overhead = 0; 3737 #if IS_ENABLED(CONFIG_IPV6) 3738 struct ipv6_pinfo *np; 3739 struct ipv6_txoptions *optv6 = NULL; 3740 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3741 3742 if (!sk) 3743 return overhead; 3744 3745 switch (sk->sk_family) { 3746 case AF_INET: 3747 inet = inet_sk(sk); 3748 overhead += sizeof(struct iphdr); 3749 opt = rcu_dereference_protected(inet->inet_opt, 3750 sock_owned_by_user(sk)); 3751 if (opt) 3752 overhead += opt->opt.optlen; 3753 return overhead; 3754 #if IS_ENABLED(CONFIG_IPV6) 3755 case AF_INET6: 3756 np = inet6_sk(sk); 3757 overhead += sizeof(struct ipv6hdr); 3758 if (np) 3759 optv6 = rcu_dereference_protected(np->opt, 3760 sock_owned_by_user(sk)); 3761 if (optv6) 3762 overhead += (optv6->opt_flen + optv6->opt_nflen); 3763 return overhead; 3764 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3765 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ 3766 return overhead; 3767 } 3768 } 3769 EXPORT_SYMBOL(kernel_sock_ip_overhead); 3770