Line data Source code
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 :
108 : #ifdef CONFIG_NET_RX_BUSY_POLL
109 : unsigned int sysctl_net_busy_read __read_mostly;
110 : unsigned int sysctl_net_busy_poll __read_mostly;
111 : #endif
112 :
113 : static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
114 : static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
115 : static int sock_mmap(struct file *file, struct vm_area_struct *vma);
116 :
117 : static int sock_close(struct inode *inode, struct file *file);
118 : static __poll_t sock_poll(struct file *file,
119 : struct poll_table_struct *wait);
120 : static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
121 : #ifdef CONFIG_COMPAT
122 : static long compat_sock_ioctl(struct file *file,
123 : unsigned int cmd, unsigned long arg);
124 : #endif
125 : static int sock_fasync(int fd, struct file *filp, int on);
126 : static ssize_t sock_sendpage(struct file *file, struct page *page,
127 : int offset, size_t size, loff_t *ppos, int more);
128 : static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
129 : struct pipe_inode_info *pipe, size_t len,
130 : unsigned int flags);
131 :
132 : #ifdef CONFIG_PROC_FS
133 0 : static void sock_show_fdinfo(struct seq_file *m, struct file *f)
134 : {
135 0 : struct socket *sock = f->private_data;
136 :
137 0 : if (sock->ops->show_fdinfo)
138 0 : sock->ops->show_fdinfo(m, sock);
139 0 : }
140 : #else
141 : #define sock_show_fdinfo NULL
142 : #endif
143 :
144 : /*
145 : * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
146 : * in the operation structures but are done directly via the socketcall() multiplexor.
147 : */
148 :
149 : static const struct file_operations socket_file_ops = {
150 : .owner = THIS_MODULE,
151 : .llseek = no_llseek,
152 : .read_iter = sock_read_iter,
153 : .write_iter = sock_write_iter,
154 : .poll = sock_poll,
155 : .unlocked_ioctl = sock_ioctl,
156 : #ifdef CONFIG_COMPAT
157 : .compat_ioctl = compat_sock_ioctl,
158 : #endif
159 : .mmap = sock_mmap,
160 : .release = sock_close,
161 : .fasync = sock_fasync,
162 : .sendpage = sock_sendpage,
163 : .splice_write = generic_splice_sendpage,
164 : .splice_read = sock_splice_read,
165 : .show_fdinfo = sock_show_fdinfo,
166 : };
167 :
168 : /*
169 : * The protocol list. Each protocol is registered in here.
170 : */
171 :
172 : static DEFINE_SPINLOCK(net_family_lock);
173 : static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
174 :
175 : /*
176 : * Support routines.
177 : * Move socket addresses back and forth across the kernel/user
178 : * divide and look after the messy bits.
179 : */
180 :
181 : /**
182 : * move_addr_to_kernel - copy a socket address into kernel space
183 : * @uaddr: Address in user space
184 : * @kaddr: Address in kernel space
185 : * @ulen: Length in user space
186 : *
187 : * The address is copied into kernel space. If the provided address is
188 : * too long an error code of -EINVAL is returned. If the copy gives
189 : * invalid addresses -EFAULT is returned. On a success 0 is returned.
190 : */
191 :
192 1168 : int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
193 : {
194 1168 : if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
195 : return -EINVAL;
196 1168 : if (ulen == 0)
197 : return 0;
198 2336 : if (copy_from_user(kaddr, uaddr, ulen))
199 0 : return -EFAULT;
200 1168 : return audit_sockaddr(ulen, kaddr);
201 : }
202 :
203 : /**
204 : * move_addr_to_user - copy an address to user space
205 : * @kaddr: kernel space address
206 : * @klen: length of address in kernel
207 : * @uaddr: user space address
208 : * @ulen: pointer to user length field
209 : *
210 : * The value pointed to by ulen on entry is the buffer length available.
211 : * This is overwritten with the buffer space used. -EINVAL is returned
212 : * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 : * is returned if either the buffer or the length field are not
214 : * accessible.
215 : * After copying the data up to the limit the user specifies, the true
216 : * length of the data is written over the length limit the user
217 : * specified. Zero is returned for a success.
218 : */
219 :
220 952 : static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
221 : void __user *uaddr, int __user *ulen)
222 : {
223 952 : int err;
224 952 : int len;
225 :
226 952 : BUG_ON(klen > sizeof(struct sockaddr_storage));
227 952 : err = get_user(len, ulen);
228 952 : if (err)
229 : return err;
230 952 : if (len > klen)
231 : len = klen;
232 952 : if (len < 0)
233 : return -EINVAL;
234 952 : if (len) {
235 709 : if (audit_sockaddr(klen, kaddr))
236 : return -ENOMEM;
237 1418 : if (copy_to_user(uaddr, kaddr, len))
238 : return -EFAULT;
239 : }
240 : /*
241 : * "fromlen shall refer to the value before truncation.."
242 : * 1003.1g
243 : */
244 952 : return __put_user(klen, ulen);
245 : }
246 :
247 : static struct kmem_cache *sock_inode_cachep __ro_after_init;
248 :
249 689 : static struct inode *sock_alloc_inode(struct super_block *sb)
250 : {
251 689 : struct socket_alloc *ei;
252 :
253 689 : ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
254 689 : if (!ei)
255 : return NULL;
256 689 : init_waitqueue_head(&ei->socket.wq.wait);
257 689 : ei->socket.wq.fasync_list = NULL;
258 689 : ei->socket.wq.flags = 0;
259 :
260 689 : ei->socket.state = SS_UNCONNECTED;
261 689 : ei->socket.flags = 0;
262 689 : ei->socket.ops = NULL;
263 689 : ei->socket.sk = NULL;
264 689 : ei->socket.file = NULL;
265 :
266 689 : return &ei->vfs_inode;
267 : }
268 :
269 590 : static void sock_free_inode(struct inode *inode)
270 : {
271 590 : struct socket_alloc *ei;
272 :
273 590 : ei = container_of(inode, struct socket_alloc, vfs_inode);
274 590 : kmem_cache_free(sock_inode_cachep, ei);
275 590 : }
276 :
277 264 : static void init_once(void *foo)
278 : {
279 264 : struct socket_alloc *ei = (struct socket_alloc *)foo;
280 :
281 264 : inode_init_once(&ei->vfs_inode);
282 264 : }
283 :
284 1 : static void init_inodecache(void)
285 : {
286 1 : sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287 : sizeof(struct socket_alloc),
288 : 0,
289 : (SLAB_HWCACHE_ALIGN |
290 : SLAB_RECLAIM_ACCOUNT |
291 : SLAB_MEM_SPREAD | SLAB_ACCOUNT),
292 : init_once);
293 1 : BUG_ON(sock_inode_cachep == NULL);
294 1 : }
295 :
296 : static const struct super_operations sockfs_ops = {
297 : .alloc_inode = sock_alloc_inode,
298 : .free_inode = sock_free_inode,
299 : .statfs = simple_statfs,
300 : };
301 :
302 : /*
303 : * sockfs_dname() is called from d_path().
304 : */
305 0 : static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
306 : {
307 0 : return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
308 0 : d_inode(dentry)->i_ino);
309 : }
310 :
311 : static const struct dentry_operations sockfs_dentry_operations = {
312 : .d_dname = sockfs_dname,
313 : };
314 :
315 0 : static int sockfs_xattr_get(const struct xattr_handler *handler,
316 : struct dentry *dentry, struct inode *inode,
317 : const char *suffix, void *value, size_t size)
318 : {
319 0 : if (value) {
320 0 : if (dentry->d_name.len + 1 > size)
321 : return -ERANGE;
322 0 : memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
323 : }
324 0 : return dentry->d_name.len + 1;
325 : }
326 :
327 : #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
328 : #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
329 : #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
330 :
331 : static const struct xattr_handler sockfs_xattr_handler = {
332 : .name = XATTR_NAME_SOCKPROTONAME,
333 : .get = sockfs_xattr_get,
334 : };
335 :
336 0 : static int sockfs_security_xattr_set(const struct xattr_handler *handler,
337 : struct user_namespace *mnt_userns,
338 : struct dentry *dentry, struct inode *inode,
339 : const char *suffix, const void *value,
340 : size_t size, int flags)
341 : {
342 : /* Handled by LSM. */
343 0 : return -EAGAIN;
344 : }
345 :
346 : static const struct xattr_handler sockfs_security_xattr_handler = {
347 : .prefix = XATTR_SECURITY_PREFIX,
348 : .set = sockfs_security_xattr_set,
349 : };
350 :
351 : static const struct xattr_handler *sockfs_xattr_handlers[] = {
352 : &sockfs_xattr_handler,
353 : &sockfs_security_xattr_handler,
354 : NULL
355 : };
356 :
357 1 : static int sockfs_init_fs_context(struct fs_context *fc)
358 : {
359 1 : struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
360 1 : if (!ctx)
361 : return -ENOMEM;
362 1 : ctx->ops = &sockfs_ops;
363 1 : ctx->dops = &sockfs_dentry_operations;
364 1 : ctx->xattr = sockfs_xattr_handlers;
365 1 : return 0;
366 : }
367 :
368 : static struct vfsmount *sock_mnt __read_mostly;
369 :
370 : static struct file_system_type sock_fs_type = {
371 : .name = "sockfs",
372 : .init_fs_context = sockfs_init_fs_context,
373 : .kill_sb = kill_anon_super,
374 : };
375 :
376 : /*
377 : * Obtains the first available file descriptor and sets it up for use.
378 : *
379 : * These functions create file structures and maps them to fd space
380 : * of the current process. On success it returns file descriptor
381 : * and file struct implicitly stored in sock->file.
382 : * Note that another thread may close file descriptor before we return
383 : * from this function. We use the fact that now we do not refer
384 : * to socket after mapping. If one day we will need it, this
385 : * function will increment ref. count on file by 1.
386 : *
387 : * In any case returned fd MAY BE not valid!
388 : * This race condition is unavoidable
389 : * with shared fd spaces, we cannot solve it inside kernel,
390 : * but we take care of internal coherence yet.
391 : */
392 :
393 : /**
394 : * sock_alloc_file - Bind a &socket to a &file
395 : * @sock: socket
396 : * @flags: file status flags
397 : * @dname: protocol name
398 : *
399 : * Returns the &file bound with @sock, implicitly storing it
400 : * in sock->file. If dname is %NULL, sets to "".
401 : * On failure the return is a ERR pointer (see linux/err.h).
402 : * This function uses GFP_KERNEL internally.
403 : */
404 :
405 632 : struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
406 : {
407 632 : struct file *file;
408 :
409 632 : if (!dname)
410 541 : dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
411 :
412 1264 : file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
413 632 : O_RDWR | (flags & O_NONBLOCK),
414 : &socket_file_ops);
415 632 : if (IS_ERR(file)) {
416 0 : sock_release(sock);
417 0 : return file;
418 : }
419 :
420 632 : sock->file = file;
421 632 : file->private_data = sock;
422 632 : stream_open(SOCK_INODE(sock), file);
423 632 : return file;
424 : }
425 : EXPORT_SYMBOL(sock_alloc_file);
426 :
427 511 : static int sock_map_fd(struct socket *sock, int flags)
428 : {
429 511 : struct file *newfile;
430 511 : int fd = get_unused_fd_flags(flags);
431 511 : if (unlikely(fd < 0)) {
432 0 : sock_release(sock);
433 0 : return fd;
434 : }
435 :
436 511 : newfile = sock_alloc_file(sock, flags, NULL);
437 511 : if (!IS_ERR(newfile)) {
438 511 : fd_install(fd, newfile);
439 511 : return fd;
440 : }
441 :
442 0 : put_unused_fd(fd);
443 0 : return PTR_ERR(newfile);
444 : }
445 :
446 : /**
447 : * sock_from_file - Return the &socket bounded to @file.
448 : * @file: file
449 : *
450 : * On failure returns %NULL.
451 : */
452 :
453 8852 : struct socket *sock_from_file(struct file *file)
454 : {
455 54 : if (file->f_op == &socket_file_ops)
456 8370 : return file->private_data; /* set in sock_map_fd */
457 :
458 : return NULL;
459 : }
460 : EXPORT_SYMBOL(sock_from_file);
461 :
462 : /**
463 : * sockfd_lookup - Go from a file number to its socket slot
464 : * @fd: file handle
465 : * @err: pointer to an error code return
466 : *
467 : * The file handle passed in is locked and the socket it is bound
468 : * to is returned. If an error occurs the err pointer is overwritten
469 : * with a negative errno code and NULL is returned. The function checks
470 : * for both invalid handles and passing a handle which is not a socket.
471 : *
472 : * On a success the socket object pointer is returned.
473 : */
474 :
475 0 : struct socket *sockfd_lookup(int fd, int *err)
476 : {
477 0 : struct file *file;
478 0 : struct socket *sock;
479 :
480 0 : file = fget(fd);
481 0 : if (!file) {
482 0 : *err = -EBADF;
483 0 : return NULL;
484 : }
485 :
486 0 : sock = sock_from_file(file);
487 0 : if (!sock) {
488 0 : *err = -ENOTSOCK;
489 0 : fput(file);
490 : }
491 : return sock;
492 : }
493 : EXPORT_SYMBOL(sockfd_lookup);
494 :
495 8324 : static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
496 : {
497 8324 : struct fd f = fdget(fd);
498 8325 : struct socket *sock;
499 :
500 8325 : *err = -EBADF;
501 8325 : if (f.file) {
502 8325 : sock = sock_from_file(f.file);
503 8325 : if (likely(sock)) {
504 8325 : *fput_needed = f.flags & FDPUT_FPUT;
505 8325 : return sock;
506 : }
507 0 : *err = -ENOTSOCK;
508 0 : fdput(f);
509 : }
510 : return NULL;
511 : }
512 :
513 0 : static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
514 : size_t size)
515 : {
516 0 : ssize_t len;
517 0 : ssize_t used = 0;
518 :
519 0 : len = security_inode_listsecurity(d_inode(dentry), buffer, size);
520 0 : if (len < 0)
521 : return len;
522 0 : used += len;
523 0 : if (buffer) {
524 0 : if (size < used)
525 : return -ERANGE;
526 0 : buffer += len;
527 : }
528 :
529 0 : len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
530 0 : used += len;
531 0 : if (buffer) {
532 0 : if (size < used)
533 : return -ERANGE;
534 0 : memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
535 0 : buffer += len;
536 : }
537 :
538 : return used;
539 : }
540 :
541 0 : static int sockfs_setattr(struct user_namespace *mnt_userns,
542 : struct dentry *dentry, struct iattr *iattr)
543 : {
544 0 : int err = simple_setattr(&init_user_ns, dentry, iattr);
545 :
546 0 : if (!err && (iattr->ia_valid & ATTR_UID)) {
547 0 : struct socket *sock = SOCKET_I(d_inode(dentry));
548 :
549 0 : if (sock->sk)
550 0 : sock->sk->sk_uid = iattr->ia_uid;
551 : else
552 : err = -ENOENT;
553 : }
554 :
555 0 : return err;
556 : }
557 :
558 : static const struct inode_operations sockfs_inode_ops = {
559 : .listxattr = sockfs_listxattr,
560 : .setattr = sockfs_setattr,
561 : };
562 :
563 : /**
564 : * sock_alloc - allocate a socket
565 : *
566 : * Allocate a new inode and socket object. The two are bound together
567 : * and initialised. The socket is then returned. If we are out of inodes
568 : * NULL is returned. This functions uses GFP_KERNEL internally.
569 : */
570 :
571 688 : struct socket *sock_alloc(void)
572 : {
573 688 : struct inode *inode;
574 688 : struct socket *sock;
575 :
576 688 : inode = new_inode_pseudo(sock_mnt->mnt_sb);
577 688 : if (!inode)
578 : return NULL;
579 :
580 688 : sock = SOCKET_I(inode);
581 :
582 688 : inode->i_ino = get_next_ino();
583 688 : inode->i_mode = S_IFSOCK | S_IRWXUGO;
584 688 : inode->i_uid = current_fsuid();
585 688 : inode->i_gid = current_fsgid();
586 688 : inode->i_op = &sockfs_inode_ops;
587 :
588 688 : return sock;
589 : }
590 : EXPORT_SYMBOL(sock_alloc);
591 :
592 589 : static void __sock_release(struct socket *sock, struct inode *inode)
593 : {
594 589 : if (sock->ops) {
595 546 : struct module *owner = sock->ops->owner;
596 :
597 546 : if (inode)
598 546 : inode_lock(inode);
599 547 : sock->ops->release(sock);
600 547 : sock->sk = NULL;
601 547 : if (inode)
602 547 : inode_unlock(inode);
603 547 : sock->ops = NULL;
604 547 : module_put(owner);
605 : }
606 :
607 590 : if (sock->wq.fasync_list)
608 0 : pr_err("%s: fasync list not empty!\n", __func__);
609 :
610 590 : if (!sock->file) {
611 43 : iput(SOCK_INODE(sock));
612 43 : return;
613 : }
614 547 : sock->file = NULL;
615 : }
616 :
617 : /**
618 : * sock_release - close a socket
619 : * @sock: socket to close
620 : *
621 : * The socket is released from the protocol stack if it has a release
622 : * callback, and the inode is then released if the socket is bound to
623 : * an inode not a file.
624 : */
625 43 : void sock_release(struct socket *sock)
626 : {
627 0 : __sock_release(sock, NULL);
628 0 : }
629 : EXPORT_SYMBOL(sock_release);
630 :
631 0 : void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
632 : {
633 0 : u8 flags = *tx_flags;
634 :
635 0 : if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
636 0 : flags |= SKBTX_HW_TSTAMP;
637 :
638 0 : if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
639 0 : flags |= SKBTX_SW_TSTAMP;
640 :
641 0 : if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
642 0 : flags |= SKBTX_SCHED_TSTAMP;
643 :
644 0 : *tx_flags = flags;
645 0 : }
646 : EXPORT_SYMBOL(__sock_tx_timestamp);
647 :
648 : INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
649 : size_t));
650 : INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
651 : size_t));
652 3546 : static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
653 : {
654 3546 : int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
655 : inet_sendmsg, sock, msg,
656 : msg_data_left(msg));
657 3546 : BUG_ON(ret == -EIOCBQUEUED);
658 3546 : return ret;
659 : }
660 :
661 : /**
662 : * sock_sendmsg - send a message through @sock
663 : * @sock: socket
664 : * @msg: message to send
665 : *
666 : * Sends @msg through @sock, passing through LSM.
667 : * Returns the number of bytes sent, or an error code.
668 : */
669 3546 : int sock_sendmsg(struct socket *sock, struct msghdr *msg)
670 : {
671 0 : int err = security_socket_sendmsg(sock, msg,
672 3546 : msg_data_left(msg));
673 :
674 3546 : return err ?: sock_sendmsg_nosec(sock, msg);
675 : }
676 : EXPORT_SYMBOL(sock_sendmsg);
677 :
678 : /**
679 : * kernel_sendmsg - send a message through @sock (kernel-space)
680 : * @sock: socket
681 : * @msg: message header
682 : * @vec: kernel vec
683 : * @num: vec array length
684 : * @size: total message data size
685 : *
686 : * Builds the message data with @vec and sends it through @sock.
687 : * Returns the number of bytes sent, or an error code.
688 : */
689 :
690 0 : int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
691 : struct kvec *vec, size_t num, size_t size)
692 : {
693 0 : iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
694 0 : return sock_sendmsg(sock, msg);
695 : }
696 : EXPORT_SYMBOL(kernel_sendmsg);
697 :
698 : /**
699 : * kernel_sendmsg_locked - send a message through @sock (kernel-space)
700 : * @sk: sock
701 : * @msg: message header
702 : * @vec: output s/g array
703 : * @num: output s/g array length
704 : * @size: total message data size
705 : *
706 : * Builds the message data with @vec and sends it through @sock.
707 : * Returns the number of bytes sent, or an error code.
708 : * Caller must hold @sk.
709 : */
710 :
711 0 : int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
712 : struct kvec *vec, size_t num, size_t size)
713 : {
714 0 : struct socket *sock = sk->sk_socket;
715 :
716 0 : if (!sock->ops->sendmsg_locked)
717 0 : return sock_no_sendmsg_locked(sk, msg, size);
718 :
719 0 : iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
720 :
721 0 : return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
722 : }
723 : EXPORT_SYMBOL(kernel_sendmsg_locked);
724 :
725 0 : static bool skb_is_err_queue(const struct sk_buff *skb)
726 : {
727 : /* pkt_type of skbs enqueued on the error queue are set to
728 : * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
729 : * in recvmsg, since skbs received on a local socket will never
730 : * have a pkt_type of PACKET_OUTGOING.
731 : */
732 0 : return skb->pkt_type == PACKET_OUTGOING;
733 : }
734 :
735 : /* On transmit, software and hardware timestamps are returned independently.
736 : * As the two skb clones share the hardware timestamp, which may be updated
737 : * before the software timestamp is received, a hardware TX timestamp may be
738 : * returned only if there is no software TX timestamp. Ignore false software
739 : * timestamps, which may be made in the __sock_recv_timestamp() call when the
740 : * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
741 : * hardware timestamp.
742 : */
743 0 : static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
744 : {
745 0 : return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
746 : }
747 :
748 0 : static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
749 : {
750 0 : struct scm_ts_pktinfo ts_pktinfo;
751 0 : struct net_device *orig_dev;
752 :
753 0 : if (!skb_mac_header_was_set(skb))
754 0 : return;
755 :
756 0 : memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
757 :
758 0 : rcu_read_lock();
759 0 : orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
760 0 : if (orig_dev)
761 0 : ts_pktinfo.if_index = orig_dev->ifindex;
762 0 : rcu_read_unlock();
763 :
764 0 : ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
765 0 : put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
766 : sizeof(ts_pktinfo), &ts_pktinfo);
767 : }
768 :
769 : /*
770 : * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
771 : */
772 511 : void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
773 : struct sk_buff *skb)
774 : {
775 511 : int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
776 511 : int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
777 511 : struct scm_timestamping_internal tss;
778 :
779 511 : int empty = 1, false_tstamp = 0;
780 511 : struct skb_shared_hwtstamps *shhwtstamps =
781 511 : skb_hwtstamps(skb);
782 :
783 : /* Race occurred between timestamp enabling and packet
784 : receiving. Fill in the current time for now. */
785 511 : if (need_software_tstamp && skb->tstamp == 0) {
786 74 : __net_timestamp(skb);
787 74 : false_tstamp = 1;
788 : }
789 :
790 511 : if (need_software_tstamp) {
791 511 : if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
792 511 : if (new_tstamp) {
793 0 : struct __kernel_sock_timeval tv;
794 :
795 0 : skb_get_new_timestamp(skb, &tv);
796 0 : put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
797 : sizeof(tv), &tv);
798 : } else {
799 511 : struct __kernel_old_timeval tv;
800 :
801 511 : skb_get_timestamp(skb, &tv);
802 511 : put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
803 : sizeof(tv), &tv);
804 : }
805 : } else {
806 0 : if (new_tstamp) {
807 0 : struct __kernel_timespec ts;
808 :
809 0 : skb_get_new_timestampns(skb, &ts);
810 0 : put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
811 : sizeof(ts), &ts);
812 : } else {
813 0 : struct __kernel_old_timespec ts;
814 :
815 0 : skb_get_timestampns(skb, &ts);
816 0 : put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
817 : sizeof(ts), &ts);
818 : }
819 : }
820 : }
821 :
822 511 : memset(&tss, 0, sizeof(tss));
823 511 : if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
824 0 : ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
825 : empty = 0;
826 511 : if (shhwtstamps &&
827 511 : (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
828 0 : !skb_is_swtx_tstamp(skb, false_tstamp) &&
829 0 : ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
830 0 : empty = 0;
831 0 : if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
832 0 : !skb_is_err_queue(skb))
833 0 : put_ts_pktinfo(msg, skb);
834 : }
835 511 : if (!empty) {
836 0 : if (sock_flag(sk, SOCK_TSTAMP_NEW))
837 0 : put_cmsg_scm_timestamping64(msg, &tss);
838 : else
839 0 : put_cmsg_scm_timestamping(msg, &tss);
840 :
841 0 : if (skb_is_err_queue(skb) && skb->len &&
842 : SKB_EXT_ERR(skb)->opt_stats)
843 0 : put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
844 0 : skb->len, skb->data);
845 : }
846 511 : }
847 : EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
848 :
849 0 : void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
850 : struct sk_buff *skb)
851 : {
852 0 : int ack;
853 :
854 0 : if (!sock_flag(sk, SOCK_WIFI_STATUS))
855 0 : return;
856 0 : if (!skb->wifi_acked_valid)
857 : return;
858 :
859 0 : ack = skb->wifi_acked;
860 :
861 0 : put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
862 : }
863 : EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
864 :
865 0 : static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
866 : struct sk_buff *skb)
867 : {
868 0 : if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
869 0 : put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
870 0 : sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
871 0 : }
872 :
873 0 : void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
874 : struct sk_buff *skb)
875 : {
876 0 : sock_recv_timestamp(msg, sk, skb);
877 0 : sock_recv_drops(msg, sk, skb);
878 0 : }
879 : EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
880 :
881 : INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
882 : size_t, int));
883 : INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
884 : size_t, int));
885 4635 : static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
886 : int flags)
887 : {
888 4635 : return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
889 : inet_recvmsg, sock, msg, msg_data_left(msg),
890 : flags);
891 : }
892 :
893 : /**
894 : * sock_recvmsg - receive a message from @sock
895 : * @sock: socket
896 : * @msg: message to receive
897 : * @flags: message flags
898 : *
899 : * Receives @msg from @sock, passing through LSM. Returns the total number
900 : * of bytes received, or an error.
901 : */
902 4635 : int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
903 : {
904 0 : int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
905 :
906 4635 : return err ?: sock_recvmsg_nosec(sock, msg, flags);
907 : }
908 : EXPORT_SYMBOL(sock_recvmsg);
909 :
910 : /**
911 : * kernel_recvmsg - Receive a message from a socket (kernel space)
912 : * @sock: The socket to receive the message from
913 : * @msg: Received message
914 : * @vec: Input s/g array for message data
915 : * @num: Size of input s/g array
916 : * @size: Number of bytes to read
917 : * @flags: Message flags (MSG_DONTWAIT, etc...)
918 : *
919 : * On return the msg structure contains the scatter/gather array passed in the
920 : * vec argument. The array is modified so that it consists of the unfilled
921 : * portion of the original array.
922 : *
923 : * The returned value is the total number of bytes received, or an error.
924 : */
925 :
926 0 : int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
927 : struct kvec *vec, size_t num, size_t size, int flags)
928 : {
929 0 : msg->msg_control_is_user = false;
930 0 : iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
931 0 : return sock_recvmsg(sock, msg, flags);
932 : }
933 : EXPORT_SYMBOL(kernel_recvmsg);
934 :
935 0 : static ssize_t sock_sendpage(struct file *file, struct page *page,
936 : int offset, size_t size, loff_t *ppos, int more)
937 : {
938 0 : struct socket *sock;
939 0 : int flags;
940 :
941 0 : sock = file->private_data;
942 :
943 0 : flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
944 : /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
945 0 : flags |= more;
946 :
947 0 : return kernel_sendpage(sock, page, offset, size, flags);
948 : }
949 :
950 0 : static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
951 : struct pipe_inode_info *pipe, size_t len,
952 : unsigned int flags)
953 : {
954 0 : struct socket *sock = file->private_data;
955 :
956 0 : if (unlikely(!sock->ops->splice_read))
957 0 : return generic_file_splice_read(file, ppos, pipe, len, flags);
958 :
959 0 : return sock->ops->splice_read(sock, ppos, pipe, len, flags);
960 : }
961 :
962 444 : static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
963 : {
964 444 : struct file *file = iocb->ki_filp;
965 444 : struct socket *sock = file->private_data;
966 444 : struct msghdr msg = {.msg_iter = *to,
967 : .msg_iocb = iocb};
968 444 : ssize_t res;
969 :
970 444 : if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
971 188 : msg.msg_flags = MSG_DONTWAIT;
972 :
973 444 : if (iocb->ki_pos != 0)
974 : return -ESPIPE;
975 :
976 444 : if (!iov_iter_count(to)) /* Match SYS5 behaviour */
977 : return 0;
978 :
979 444 : res = sock_recvmsg(sock, &msg, msg.msg_flags);
980 444 : *to = msg.msg_iter;
981 444 : return res;
982 : }
983 :
984 811 : static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
985 : {
986 811 : struct file *file = iocb->ki_filp;
987 811 : struct socket *sock = file->private_data;
988 811 : struct msghdr msg = {.msg_iter = *from,
989 : .msg_iocb = iocb};
990 811 : ssize_t res;
991 :
992 811 : if (iocb->ki_pos != 0)
993 : return -ESPIPE;
994 :
995 811 : if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
996 407 : msg.msg_flags = MSG_DONTWAIT;
997 :
998 811 : if (sock->type == SOCK_SEQPACKET)
999 0 : msg.msg_flags |= MSG_EOR;
1000 :
1001 811 : res = sock_sendmsg(sock, &msg);
1002 811 : *from = msg.msg_iter;
1003 811 : return res;
1004 : }
1005 :
1006 : /*
1007 : * Atomic setting of ioctl hooks to avoid race
1008 : * with module unload.
1009 : */
1010 :
1011 : static DEFINE_MUTEX(br_ioctl_mutex);
1012 : static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1013 :
1014 0 : void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1015 : {
1016 0 : mutex_lock(&br_ioctl_mutex);
1017 0 : br_ioctl_hook = hook;
1018 0 : mutex_unlock(&br_ioctl_mutex);
1019 0 : }
1020 : EXPORT_SYMBOL(brioctl_set);
1021 :
1022 : static DEFINE_MUTEX(vlan_ioctl_mutex);
1023 : static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1024 :
1025 0 : void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1026 : {
1027 0 : mutex_lock(&vlan_ioctl_mutex);
1028 0 : vlan_ioctl_hook = hook;
1029 0 : mutex_unlock(&vlan_ioctl_mutex);
1030 0 : }
1031 : EXPORT_SYMBOL(vlan_ioctl_set);
1032 :
1033 309 : static long sock_do_ioctl(struct net *net, struct socket *sock,
1034 : unsigned int cmd, unsigned long arg)
1035 : {
1036 309 : int err;
1037 309 : void __user *argp = (void __user *)arg;
1038 :
1039 309 : err = sock->ops->ioctl(sock, cmd, arg);
1040 :
1041 : /*
1042 : * If this ioctl is unknown try to hand it down
1043 : * to the NIC driver.
1044 : */
1045 309 : if (err != -ENOIOCTLCMD)
1046 243 : return err;
1047 :
1048 66 : if (cmd == SIOCGIFCONF) {
1049 0 : struct ifconf ifc;
1050 0 : if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1051 0 : return -EFAULT;
1052 0 : rtnl_lock();
1053 0 : err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1054 0 : rtnl_unlock();
1055 0 : if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1056 0 : err = -EFAULT;
1057 : } else {
1058 66 : struct ifreq ifr;
1059 66 : bool need_copyout;
1060 66 : if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1061 0 : return -EFAULT;
1062 66 : err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1063 66 : if (!err && need_copyout)
1064 19 : if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1065 : return -EFAULT;
1066 : }
1067 66 : return err;
1068 : }
1069 :
1070 : /*
1071 : * With an ioctl, arg may well be a user mode pointer, but we don't know
1072 : * what to do with it - that's up to the protocol still.
1073 : */
1074 :
1075 : /**
1076 : * get_net_ns - increment the refcount of the network namespace
1077 : * @ns: common namespace (net)
1078 : *
1079 : * Returns the net's common namespace.
1080 : */
1081 :
1082 0 : struct ns_common *get_net_ns(struct ns_common *ns)
1083 : {
1084 0 : return &get_net(container_of(ns, struct net, ns))->ns;
1085 : }
1086 : EXPORT_SYMBOL_GPL(get_net_ns);
1087 :
1088 309 : static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1089 : {
1090 309 : struct socket *sock;
1091 309 : struct sock *sk;
1092 309 : void __user *argp = (void __user *)arg;
1093 309 : int pid, err;
1094 309 : struct net *net;
1095 :
1096 309 : sock = file->private_data;
1097 309 : sk = sock->sk;
1098 309 : net = sock_net(sk);
1099 309 : if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1100 0 : struct ifreq ifr;
1101 0 : bool need_copyout;
1102 0 : if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1103 0 : return -EFAULT;
1104 0 : err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1105 0 : if (!err && need_copyout)
1106 0 : if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1107 : return -EFAULT;
1108 : } else
1109 : #ifdef CONFIG_WEXT_CORE
1110 : if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1111 : err = wext_handle_ioctl(net, cmd, argp);
1112 : } else
1113 : #endif
1114 309 : switch (cmd) {
1115 0 : case FIOSETOWN:
1116 : case SIOCSPGRP:
1117 0 : err = -EFAULT;
1118 0 : if (get_user(pid, (int __user *)argp))
1119 : break;
1120 0 : err = f_setown(sock->file, pid, 1);
1121 0 : break;
1122 0 : case FIOGETOWN:
1123 : case SIOCGPGRP:
1124 0 : err = put_user(f_getown(sock->file),
1125 : (int __user *)argp);
1126 0 : break;
1127 0 : case SIOCGIFBR:
1128 : case SIOCSIFBR:
1129 : case SIOCBRADDBR:
1130 : case SIOCBRDELBR:
1131 0 : err = -ENOPKG;
1132 0 : if (!br_ioctl_hook)
1133 0 : request_module("bridge");
1134 :
1135 0 : mutex_lock(&br_ioctl_mutex);
1136 0 : if (br_ioctl_hook)
1137 0 : err = br_ioctl_hook(net, cmd, argp);
1138 0 : mutex_unlock(&br_ioctl_mutex);
1139 0 : break;
1140 0 : case SIOCGIFVLAN:
1141 : case SIOCSIFVLAN:
1142 0 : err = -ENOPKG;
1143 0 : if (!vlan_ioctl_hook)
1144 0 : request_module("8021q");
1145 :
1146 0 : mutex_lock(&vlan_ioctl_mutex);
1147 0 : if (vlan_ioctl_hook)
1148 0 : err = vlan_ioctl_hook(net, argp);
1149 0 : mutex_unlock(&vlan_ioctl_mutex);
1150 0 : break;
1151 0 : case SIOCGSKNS:
1152 0 : err = -EPERM;
1153 0 : if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1154 : break;
1155 :
1156 0 : err = open_related_ns(&net->ns, get_net_ns);
1157 0 : break;
1158 0 : case SIOCGSTAMP_OLD:
1159 : case SIOCGSTAMPNS_OLD:
1160 0 : if (!sock->ops->gettstamp) {
1161 : err = -ENOIOCTLCMD;
1162 : break;
1163 : }
1164 0 : err = sock->ops->gettstamp(sock, argp,
1165 : cmd == SIOCGSTAMP_OLD,
1166 : !IS_ENABLED(CONFIG_64BIT));
1167 0 : break;
1168 0 : case SIOCGSTAMP_NEW:
1169 : case SIOCGSTAMPNS_NEW:
1170 0 : if (!sock->ops->gettstamp) {
1171 : err = -ENOIOCTLCMD;
1172 : break;
1173 : }
1174 0 : err = sock->ops->gettstamp(sock, argp,
1175 : cmd == SIOCGSTAMP_NEW,
1176 : false);
1177 0 : break;
1178 309 : default:
1179 309 : err = sock_do_ioctl(net, sock, cmd, arg);
1180 309 : break;
1181 : }
1182 309 : return err;
1183 : }
1184 :
1185 : /**
1186 : * sock_create_lite - creates a socket
1187 : * @family: protocol family (AF_INET, ...)
1188 : * @type: communication type (SOCK_STREAM, ...)
1189 : * @protocol: protocol (0, ...)
1190 : * @res: new socket
1191 : *
1192 : * Creates a new socket and assigns it to @res, passing through LSM.
1193 : * The new socket initialization is not complete, see kernel_accept().
1194 : * Returns 0 or an error. On failure @res is set to %NULL.
1195 : * This function internally uses GFP_KERNEL.
1196 : */
1197 :
1198 5 : int sock_create_lite(int family, int type, int protocol, struct socket **res)
1199 : {
1200 5 : int err;
1201 5 : struct socket *sock = NULL;
1202 :
1203 5 : err = security_socket_create(family, type, protocol, 1);
1204 5 : if (err)
1205 : goto out;
1206 :
1207 5 : sock = sock_alloc();
1208 5 : if (!sock) {
1209 0 : err = -ENOMEM;
1210 0 : goto out;
1211 : }
1212 :
1213 5 : sock->type = type;
1214 5 : err = security_socket_post_create(sock, family, type, protocol, 1);
1215 5 : if (err)
1216 : goto out_release;
1217 :
1218 5 : out:
1219 5 : *res = sock;
1220 5 : return err;
1221 : out_release:
1222 : sock_release(sock);
1223 : sock = NULL;
1224 : goto out;
1225 : }
1226 : EXPORT_SYMBOL(sock_create_lite);
1227 :
1228 : /* No kernel lock held - perfect */
1229 6852 : static __poll_t sock_poll(struct file *file, poll_table *wait)
1230 : {
1231 6852 : struct socket *sock = file->private_data;
1232 6852 : __poll_t events = poll_requested_events(wait), flag = 0;
1233 :
1234 6852 : if (!sock->ops->poll)
1235 : return 0;
1236 :
1237 6852 : if (sk_can_busy_loop(sock->sk)) {
1238 : /* poll once if requested by the syscall */
1239 0 : if (events & POLL_BUSY_LOOP)
1240 0 : sk_busy_loop(sock->sk, 1);
1241 :
1242 : /* if this socket can poll_ll, tell the system call */
1243 : flag = POLL_BUSY_LOOP;
1244 : }
1245 :
1246 6852 : return sock->ops->poll(file, sock, wait) | flag;
1247 : }
1248 :
1249 0 : static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1250 : {
1251 0 : struct socket *sock = file->private_data;
1252 :
1253 0 : return sock->ops->mmap(file, sock, vma);
1254 : }
1255 :
1256 546 : static int sock_close(struct inode *inode, struct file *filp)
1257 : {
1258 546 : __sock_release(SOCKET_I(inode), inode);
1259 547 : return 0;
1260 : }
1261 :
1262 : /*
1263 : * Update the socket async list
1264 : *
1265 : * Fasync_list locking strategy.
1266 : *
1267 : * 1. fasync_list is modified only under process context socket lock
1268 : * i.e. under semaphore.
1269 : * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1270 : * or under socket lock
1271 : */
1272 :
1273 0 : static int sock_fasync(int fd, struct file *filp, int on)
1274 : {
1275 0 : struct socket *sock = filp->private_data;
1276 0 : struct sock *sk = sock->sk;
1277 0 : struct socket_wq *wq = &sock->wq;
1278 :
1279 0 : if (sk == NULL)
1280 : return -EINVAL;
1281 :
1282 0 : lock_sock(sk);
1283 0 : fasync_helper(fd, filp, on, &wq->fasync_list);
1284 :
1285 0 : if (!wq->fasync_list)
1286 0 : sock_reset_flag(sk, SOCK_FASYNC);
1287 : else
1288 0 : sock_set_flag(sk, SOCK_FASYNC);
1289 :
1290 0 : release_sock(sk);
1291 0 : return 0;
1292 : }
1293 :
1294 : /* This function may be called only under rcu_lock */
1295 :
1296 0 : int sock_wake_async(struct socket_wq *wq, int how, int band)
1297 : {
1298 0 : if (!wq || !wq->fasync_list)
1299 : return -1;
1300 :
1301 0 : switch (how) {
1302 0 : case SOCK_WAKE_WAITD:
1303 0 : if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1304 : break;
1305 0 : goto call_kill;
1306 0 : case SOCK_WAKE_SPACE:
1307 0 : if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1308 : break;
1309 0 : fallthrough;
1310 : case SOCK_WAKE_IO:
1311 0 : call_kill:
1312 0 : kill_fasync(&wq->fasync_list, SIGIO, band);
1313 0 : break;
1314 0 : case SOCK_WAKE_URG:
1315 0 : kill_fasync(&wq->fasync_list, SIGURG, band);
1316 : }
1317 :
1318 0 : return 0;
1319 : }
1320 : EXPORT_SYMBOL(sock_wake_async);
1321 :
1322 : /**
1323 : * __sock_create - creates a socket
1324 : * @net: net namespace
1325 : * @family: protocol family (AF_INET, ...)
1326 : * @type: communication type (SOCK_STREAM, ...)
1327 : * @protocol: protocol (0, ...)
1328 : * @res: new socket
1329 : * @kern: boolean for kernel space sockets
1330 : *
1331 : * Creates a new socket and assigns it to @res, passing through LSM.
1332 : * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1333 : * be set to true if the socket resides in kernel space.
1334 : * This function internally uses GFP_KERNEL.
1335 : */
1336 :
1337 592 : int __sock_create(struct net *net, int family, int type, int protocol,
1338 : struct socket **res, int kern)
1339 : {
1340 592 : int err;
1341 592 : struct socket *sock;
1342 592 : const struct net_proto_family *pf;
1343 :
1344 : /*
1345 : * Check protocol is in range
1346 : */
1347 592 : if (family < 0 || family >= NPROTO)
1348 : return -EAFNOSUPPORT;
1349 592 : if (type < 0 || type >= SOCK_MAX)
1350 : return -EINVAL;
1351 :
1352 : /* Compatibility.
1353 :
1354 : This uglymoron is moved from INET layer to here to avoid
1355 : deadlock in module load.
1356 : */
1357 592 : if (family == PF_INET && type == SOCK_PACKET) {
1358 0 : pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1359 : current->comm);
1360 0 : family = PF_PACKET;
1361 : }
1362 :
1363 592 : err = security_socket_create(family, type, protocol, kern);
1364 592 : if (err)
1365 : return err;
1366 :
1367 : /*
1368 : * Allocate the socket and allow the family to set things up. if
1369 : * the protocol is 0, the family is instructed to select an appropriate
1370 : * default.
1371 : */
1372 592 : sock = sock_alloc();
1373 592 : if (!sock) {
1374 0 : net_warn_ratelimited("socket: no more sockets\n");
1375 0 : return -ENFILE; /* Not exactly a match, but its the
1376 : closest posix thing */
1377 : }
1378 :
1379 592 : sock->type = type;
1380 :
1381 : #ifdef CONFIG_MODULES
1382 : /* Attempt to load a protocol module if the find failed.
1383 : *
1384 : * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1385 : * requested real, full-featured networking support upon configuration.
1386 : * Otherwise module support will break!
1387 : */
1388 : if (rcu_access_pointer(net_families[family]) == NULL)
1389 : request_module("net-pf-%d", family);
1390 : #endif
1391 :
1392 592 : rcu_read_lock();
1393 592 : pf = rcu_dereference(net_families[family]);
1394 592 : err = -EAFNOSUPPORT;
1395 592 : if (!pf)
1396 11 : goto out_release;
1397 :
1398 : /*
1399 : * We will call the ->create function, that possibly is in a loadable
1400 : * module, so we have to bump that loadable module refcnt first.
1401 : */
1402 581 : if (!try_module_get(pf->owner))
1403 : goto out_release;
1404 :
1405 : /* Now protected by module ref count */
1406 581 : rcu_read_unlock();
1407 :
1408 581 : err = pf->create(net, sock, protocol, kern);
1409 581 : if (err < 0)
1410 32 : goto out_module_put;
1411 :
1412 : /*
1413 : * Now to bump the refcnt of the [loadable] module that owns this
1414 : * socket at sock_release time we decrement its refcnt.
1415 : */
1416 549 : if (!try_module_get(sock->ops->owner))
1417 : goto out_module_busy;
1418 :
1419 : /*
1420 : * Now that we're done with the ->create function, the [loadable]
1421 : * module can have its refcnt decremented
1422 : */
1423 549 : module_put(pf->owner);
1424 549 : err = security_socket_post_create(sock, family, type, protocol, kern);
1425 549 : if (err)
1426 : goto out_sock_release;
1427 549 : *res = sock;
1428 :
1429 549 : return 0;
1430 :
1431 : out_module_busy:
1432 : err = -EAFNOSUPPORT;
1433 32 : out_module_put:
1434 32 : sock->ops = NULL;
1435 32 : module_put(pf->owner);
1436 43 : out_sock_release:
1437 43 : sock_release(sock);
1438 43 : return err;
1439 :
1440 11 : out_release:
1441 11 : rcu_read_unlock();
1442 11 : goto out_sock_release;
1443 : }
1444 : EXPORT_SYMBOL(__sock_create);
1445 :
1446 : /**
1447 : * sock_create - creates a socket
1448 : * @family: protocol family (AF_INET, ...)
1449 : * @type: communication type (SOCK_STREAM, ...)
1450 : * @protocol: protocol (0, ...)
1451 : * @res: new socket
1452 : *
1453 : * A wrapper around __sock_create().
1454 : * Returns 0 or an error. This function internally uses GFP_KERNEL.
1455 : */
1456 :
1457 584 : int sock_create(int family, int type, int protocol, struct socket **res)
1458 : {
1459 584 : return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1460 : }
1461 : EXPORT_SYMBOL(sock_create);
1462 :
1463 : /**
1464 : * sock_create_kern - creates a socket (kernel space)
1465 : * @net: net namespace
1466 : * @family: protocol family (AF_INET, ...)
1467 : * @type: communication type (SOCK_STREAM, ...)
1468 : * @protocol: protocol (0, ...)
1469 : * @res: new socket
1470 : *
1471 : * A wrapper around __sock_create().
1472 : * Returns 0 or an error. This function internally uses GFP_KERNEL.
1473 : */
1474 :
1475 8 : int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1476 : {
1477 8 : return __sock_create(net, family, type, protocol, res, 1);
1478 : }
1479 : EXPORT_SYMBOL(sock_create_kern);
1480 :
1481 554 : int __sys_socket(int family, int type, int protocol)
1482 : {
1483 554 : int retval;
1484 554 : struct socket *sock;
1485 554 : int flags;
1486 :
1487 : /* Check the SOCK_* constants for consistency. */
1488 554 : BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1489 554 : BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1490 554 : BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1491 554 : BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1492 :
1493 554 : flags = type & ~SOCK_TYPE_MASK;
1494 554 : if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1495 : return -EINVAL;
1496 554 : type &= SOCK_TYPE_MASK;
1497 :
1498 554 : if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1499 : flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1500 :
1501 554 : retval = sock_create(family, type, protocol, &sock);
1502 554 : if (retval < 0)
1503 : return retval;
1504 :
1505 511 : return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1506 : }
1507 :
1508 1108 : SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1509 : {
1510 554 : return __sys_socket(family, type, protocol);
1511 : }
1512 :
1513 : /*
1514 : * Create a pair of connected sockets.
1515 : */
1516 :
1517 15 : int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1518 : {
1519 15 : struct socket *sock1, *sock2;
1520 15 : int fd1, fd2, err;
1521 15 : struct file *newfile1, *newfile2;
1522 15 : int flags;
1523 :
1524 15 : flags = type & ~SOCK_TYPE_MASK;
1525 15 : if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1526 : return -EINVAL;
1527 15 : type &= SOCK_TYPE_MASK;
1528 :
1529 15 : if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1530 : flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1531 :
1532 : /*
1533 : * reserve descriptors and make sure we won't fail
1534 : * to return them to userland.
1535 : */
1536 15 : fd1 = get_unused_fd_flags(flags);
1537 15 : if (unlikely(fd1 < 0))
1538 : return fd1;
1539 :
1540 15 : fd2 = get_unused_fd_flags(flags);
1541 15 : if (unlikely(fd2 < 0)) {
1542 0 : put_unused_fd(fd1);
1543 0 : return fd2;
1544 : }
1545 :
1546 15 : err = put_user(fd1, &usockvec[0]);
1547 15 : if (err)
1548 0 : goto out;
1549 :
1550 15 : err = put_user(fd2, &usockvec[1]);
1551 15 : if (err)
1552 0 : goto out;
1553 :
1554 : /*
1555 : * Obtain the first socket and check if the underlying protocol
1556 : * supports the socketpair call.
1557 : */
1558 :
1559 15 : err = sock_create(family, type, protocol, &sock1);
1560 15 : if (unlikely(err < 0))
1561 0 : goto out;
1562 :
1563 15 : err = sock_create(family, type, protocol, &sock2);
1564 15 : if (unlikely(err < 0)) {
1565 0 : sock_release(sock1);
1566 0 : goto out;
1567 : }
1568 :
1569 15 : err = security_socket_socketpair(sock1, sock2);
1570 15 : if (unlikely(err)) {
1571 : sock_release(sock2);
1572 : sock_release(sock1);
1573 : goto out;
1574 : }
1575 :
1576 15 : err = sock1->ops->socketpair(sock1, sock2);
1577 15 : if (unlikely(err < 0)) {
1578 0 : sock_release(sock2);
1579 0 : sock_release(sock1);
1580 0 : goto out;
1581 : }
1582 :
1583 15 : newfile1 = sock_alloc_file(sock1, flags, NULL);
1584 15 : if (IS_ERR(newfile1)) {
1585 0 : err = PTR_ERR(newfile1);
1586 0 : sock_release(sock2);
1587 0 : goto out;
1588 : }
1589 :
1590 15 : newfile2 = sock_alloc_file(sock2, flags, NULL);
1591 15 : if (IS_ERR(newfile2)) {
1592 0 : err = PTR_ERR(newfile2);
1593 0 : fput(newfile1);
1594 0 : goto out;
1595 : }
1596 :
1597 15 : audit_fd_pair(fd1, fd2);
1598 :
1599 15 : fd_install(fd1, newfile1);
1600 15 : fd_install(fd2, newfile2);
1601 15 : return 0;
1602 :
1603 0 : out:
1604 0 : put_unused_fd(fd2);
1605 0 : put_unused_fd(fd1);
1606 0 : return err;
1607 : }
1608 :
1609 30 : SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1610 : int __user *, usockvec)
1611 : {
1612 15 : return __sys_socketpair(family, type, protocol, usockvec);
1613 : }
1614 :
1615 : /*
1616 : * Bind a name to a socket. Nothing much to do here since it's
1617 : * the protocol's responsibility to handle the local address.
1618 : *
1619 : * We move the socket address to kernel space before we call
1620 : * the protocol layer (having also checked the address is ok).
1621 : */
1622 :
1623 74 : int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1624 : {
1625 74 : struct socket *sock;
1626 74 : struct sockaddr_storage address;
1627 74 : int err, fput_needed;
1628 :
1629 74 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
1630 74 : if (sock) {
1631 74 : err = move_addr_to_kernel(umyaddr, addrlen, &address);
1632 74 : if (!err) {
1633 74 : err = security_socket_bind(sock,
1634 : (struct sockaddr *)&address,
1635 : addrlen);
1636 74 : if (!err)
1637 74 : err = sock->ops->bind(sock,
1638 : (struct sockaddr *)
1639 : &address, addrlen);
1640 : }
1641 74 : fput_light(sock->file, fput_needed);
1642 : }
1643 74 : return err;
1644 : }
1645 :
1646 148 : SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1647 : {
1648 74 : return __sys_bind(fd, umyaddr, addrlen);
1649 : }
1650 :
1651 : /*
1652 : * Perform a listen. Basically, we allow the protocol to do anything
1653 : * necessary for a listen, and if that works, we mark the socket as
1654 : * ready for listening.
1655 : */
1656 :
1657 16 : int __sys_listen(int fd, int backlog)
1658 : {
1659 16 : struct socket *sock;
1660 16 : int err, fput_needed;
1661 16 : int somaxconn;
1662 :
1663 16 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
1664 16 : if (sock) {
1665 16 : somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1666 16 : if ((unsigned int)backlog > somaxconn)
1667 0 : backlog = somaxconn;
1668 :
1669 16 : err = security_socket_listen(sock, backlog);
1670 16 : if (!err)
1671 16 : err = sock->ops->listen(sock, backlog);
1672 :
1673 16 : fput_light(sock->file, fput_needed);
1674 : }
1675 16 : return err;
1676 : }
1677 :
1678 32 : SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1679 : {
1680 16 : return __sys_listen(fd, backlog);
1681 : }
1682 :
1683 91 : int __sys_accept4_file(struct file *file, unsigned file_flags,
1684 : struct sockaddr __user *upeer_sockaddr,
1685 : int __user *upeer_addrlen, int flags,
1686 : unsigned long nofile)
1687 : {
1688 91 : struct socket *sock, *newsock;
1689 91 : struct file *newfile;
1690 91 : int err, len, newfd;
1691 91 : struct sockaddr_storage address;
1692 :
1693 91 : if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1694 : return -EINVAL;
1695 :
1696 91 : if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1697 : flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1698 :
1699 91 : sock = sock_from_file(file);
1700 91 : if (!sock) {
1701 0 : err = -ENOTSOCK;
1702 0 : goto out;
1703 : }
1704 :
1705 91 : err = -ENFILE;
1706 91 : newsock = sock_alloc();
1707 91 : if (!newsock)
1708 0 : goto out;
1709 :
1710 91 : newsock->type = sock->type;
1711 91 : newsock->ops = sock->ops;
1712 :
1713 : /*
1714 : * We don't need try_module_get here, as the listening socket (sock)
1715 : * has the protocol module (sock->ops->owner) held.
1716 : */
1717 91 : __module_get(newsock->ops->owner);
1718 :
1719 91 : newfd = __get_unused_fd_flags(flags, nofile);
1720 91 : if (unlikely(newfd < 0)) {
1721 0 : err = newfd;
1722 0 : sock_release(newsock);
1723 0 : goto out;
1724 : }
1725 91 : newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1726 91 : if (IS_ERR(newfile)) {
1727 0 : err = PTR_ERR(newfile);
1728 0 : put_unused_fd(newfd);
1729 0 : goto out;
1730 : }
1731 :
1732 91 : err = security_socket_accept(sock, newsock);
1733 91 : if (err)
1734 : goto out_fd;
1735 :
1736 91 : err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1737 : false);
1738 91 : if (err < 0)
1739 0 : goto out_fd;
1740 :
1741 91 : if (upeer_sockaddr) {
1742 29 : len = newsock->ops->getname(newsock,
1743 : (struct sockaddr *)&address, 2);
1744 29 : if (len < 0) {
1745 0 : err = -ECONNABORTED;
1746 0 : goto out_fd;
1747 : }
1748 29 : err = move_addr_to_user(&address,
1749 : len, upeer_sockaddr, upeer_addrlen);
1750 29 : if (err < 0)
1751 0 : goto out_fd;
1752 : }
1753 :
1754 : /* File flags are not inherited via accept() unlike another OSes. */
1755 :
1756 91 : fd_install(newfd, newfile);
1757 91 : err = newfd;
1758 : out:
1759 : return err;
1760 0 : out_fd:
1761 0 : fput(newfile);
1762 0 : put_unused_fd(newfd);
1763 0 : goto out;
1764 :
1765 : }
1766 :
1767 : /*
1768 : * For accept, we attempt to create a new socket, set up the link
1769 : * with the client, wake up the client, then return the new
1770 : * connected fd. We collect the address of the connector in kernel
1771 : * space and move it to user at the very end. This is unclean because
1772 : * we open the socket then return an error.
1773 : *
1774 : * 1003.1g adds the ability to recvmsg() to query connection pending
1775 : * status to recvmsg. We need to add that support in a way thats
1776 : * clean when we restructure accept also.
1777 : */
1778 :
1779 91 : int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1780 : int __user *upeer_addrlen, int flags)
1781 : {
1782 91 : int ret = -EBADF;
1783 91 : struct fd f;
1784 :
1785 91 : f = fdget(fd);
1786 91 : if (f.file) {
1787 91 : ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1788 : upeer_addrlen, flags,
1789 : rlimit(RLIMIT_NOFILE));
1790 91 : fdput(f);
1791 : }
1792 :
1793 91 : return ret;
1794 : }
1795 :
1796 174 : SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1797 : int __user *, upeer_addrlen, int, flags)
1798 : {
1799 87 : return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1800 : }
1801 :
1802 8 : SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1803 : int __user *, upeer_addrlen)
1804 : {
1805 4 : return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1806 : }
1807 :
1808 : /*
1809 : * Attempt to connect to a socket with the server address. The address
1810 : * is in user space so we verify it is OK and move it to kernel space.
1811 : *
1812 : * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1813 : * break bindings
1814 : *
1815 : * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1816 : * other SEQPACKET protocols that take time to connect() as it doesn't
1817 : * include the -EINPROGRESS status for such sockets.
1818 : */
1819 :
1820 382 : int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1821 : int addrlen, int file_flags)
1822 : {
1823 382 : struct socket *sock;
1824 382 : int err;
1825 :
1826 382 : sock = sock_from_file(file);
1827 382 : if (!sock) {
1828 0 : err = -ENOTSOCK;
1829 0 : goto out;
1830 : }
1831 :
1832 382 : err =
1833 382 : security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1834 382 : if (err)
1835 : goto out;
1836 :
1837 382 : err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1838 382 : sock->file->f_flags | file_flags);
1839 381 : out:
1840 381 : return err;
1841 : }
1842 :
1843 382 : int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1844 : {
1845 382 : int ret = -EBADF;
1846 382 : struct fd f;
1847 :
1848 382 : f = fdget(fd);
1849 382 : if (f.file) {
1850 382 : struct sockaddr_storage address;
1851 :
1852 382 : ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1853 382 : if (!ret)
1854 382 : ret = __sys_connect_file(f.file, &address, addrlen, 0);
1855 412 : fdput(f);
1856 : }
1857 :
1858 381 : return ret;
1859 : }
1860 :
1861 763 : SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1862 : int, addrlen)
1863 : {
1864 382 : return __sys_connect(fd, uservaddr, addrlen);
1865 : }
1866 :
1867 : /*
1868 : * Get the local address ('name') of a socket object. Move the obtained
1869 : * name to user space.
1870 : */
1871 :
1872 199 : int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1873 : int __user *usockaddr_len)
1874 : {
1875 199 : struct socket *sock;
1876 199 : struct sockaddr_storage address;
1877 199 : int err, fput_needed;
1878 :
1879 199 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
1880 199 : if (!sock)
1881 0 : goto out;
1882 :
1883 199 : err = security_socket_getsockname(sock);
1884 199 : if (err)
1885 : goto out_put;
1886 :
1887 199 : err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1888 199 : if (err < 0)
1889 0 : goto out_put;
1890 : /* "err" is actually length in this case */
1891 199 : err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1892 :
1893 199 : out_put:
1894 199 : fput_light(sock->file, fput_needed);
1895 199 : out:
1896 199 : return err;
1897 : }
1898 :
1899 398 : SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1900 : int __user *, usockaddr_len)
1901 : {
1902 199 : return __sys_getsockname(fd, usockaddr, usockaddr_len);
1903 : }
1904 :
1905 : /*
1906 : * Get the remote address ('name') of a socket object. Move the obtained
1907 : * name to user space.
1908 : */
1909 :
1910 18 : int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1911 : int __user *usockaddr_len)
1912 : {
1913 18 : struct socket *sock;
1914 18 : struct sockaddr_storage address;
1915 18 : int err, fput_needed;
1916 :
1917 18 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
1918 18 : if (sock != NULL) {
1919 18 : err = security_socket_getpeername(sock);
1920 18 : if (err) {
1921 : fput_light(sock->file, fput_needed);
1922 : return err;
1923 : }
1924 :
1925 18 : err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1926 18 : if (err >= 0)
1927 : /* "err" is actually length in this case */
1928 16 : err = move_addr_to_user(&address, err, usockaddr,
1929 : usockaddr_len);
1930 18 : fput_light(sock->file, fput_needed);
1931 : }
1932 18 : return err;
1933 : }
1934 :
1935 36 : SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1936 : int __user *, usockaddr_len)
1937 : {
1938 18 : return __sys_getpeername(fd, usockaddr, usockaddr_len);
1939 : }
1940 :
1941 : /*
1942 : * Send a datagram to a given address. We move the address into kernel
1943 : * space and check the user space data area is readable before invoking
1944 : * the protocol.
1945 : */
1946 152 : int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1947 : struct sockaddr __user *addr, int addr_len)
1948 : {
1949 152 : struct socket *sock;
1950 152 : struct sockaddr_storage address;
1951 152 : int err;
1952 152 : struct msghdr msg;
1953 152 : struct iovec iov;
1954 152 : int fput_needed;
1955 :
1956 152 : err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1957 152 : if (unlikely(err))
1958 : return err;
1959 152 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
1960 152 : if (!sock)
1961 0 : goto out;
1962 :
1963 152 : msg.msg_name = NULL;
1964 152 : msg.msg_control = NULL;
1965 152 : msg.msg_controllen = 0;
1966 152 : msg.msg_namelen = 0;
1967 152 : if (addr) {
1968 35 : err = move_addr_to_kernel(addr, addr_len, &address);
1969 35 : if (err < 0)
1970 0 : goto out_put;
1971 35 : msg.msg_name = (struct sockaddr *)&address;
1972 35 : msg.msg_namelen = addr_len;
1973 : }
1974 152 : if (sock->file->f_flags & O_NONBLOCK)
1975 44 : flags |= MSG_DONTWAIT;
1976 152 : msg.msg_flags = flags;
1977 152 : err = sock_sendmsg(sock, &msg);
1978 :
1979 152 : out_put:
1980 152 : fput_light(sock->file, fput_needed);
1981 152 : out:
1982 152 : return err;
1983 : }
1984 :
1985 304 : SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1986 : unsigned int, flags, struct sockaddr __user *, addr,
1987 : int, addr_len)
1988 : {
1989 152 : return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1990 : }
1991 :
1992 : /*
1993 : * Send a datagram down a socket.
1994 : */
1995 :
1996 0 : SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1997 : unsigned int, flags)
1998 : {
1999 0 : return __sys_sendto(fd, buff, len, flags, NULL, 0);
2000 : }
2001 :
2002 : /*
2003 : * Receive a frame from the socket and optionally record the address of the
2004 : * sender. We verify the buffers are writable and if needed move the
2005 : * sender address from kernel to user space.
2006 : */
2007 34 : int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2008 : struct sockaddr __user *addr, int __user *addr_len)
2009 : {
2010 34 : struct socket *sock;
2011 34 : struct iovec iov;
2012 34 : struct msghdr msg;
2013 34 : struct sockaddr_storage address;
2014 34 : int err, err2;
2015 34 : int fput_needed;
2016 :
2017 34 : err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2018 34 : if (unlikely(err))
2019 : return err;
2020 34 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
2021 34 : if (!sock)
2022 0 : goto out;
2023 :
2024 34 : msg.msg_control = NULL;
2025 34 : msg.msg_controllen = 0;
2026 : /* Save some cycles and don't copy the address if not needed */
2027 34 : msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2028 : /* We assume all kernel code knows the size of sockaddr_storage */
2029 34 : msg.msg_namelen = 0;
2030 34 : msg.msg_iocb = NULL;
2031 34 : msg.msg_flags = 0;
2032 34 : if (sock->file->f_flags & O_NONBLOCK)
2033 13 : flags |= MSG_DONTWAIT;
2034 34 : err = sock_recvmsg(sock, &msg, flags);
2035 :
2036 34 : if (err >= 0 && addr != NULL) {
2037 2 : err2 = move_addr_to_user(&address,
2038 : msg.msg_namelen, addr, addr_len);
2039 2 : if (err2 < 0)
2040 0 : err = err2;
2041 : }
2042 :
2043 34 : fput_light(sock->file, fput_needed);
2044 34 : out:
2045 34 : return err;
2046 : }
2047 :
2048 68 : SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2049 : unsigned int, flags, struct sockaddr __user *, addr,
2050 : int __user *, addr_len)
2051 : {
2052 34 : return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2053 : }
2054 :
2055 : /*
2056 : * Receive a datagram from a socket.
2057 : */
2058 :
2059 0 : SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2060 : unsigned int, flags)
2061 : {
2062 0 : return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2063 : }
2064 :
2065 351 : static bool sock_use_custom_sol_socket(const struct socket *sock)
2066 : {
2067 351 : const struct sock *sk = sock->sk;
2068 :
2069 : /* Use sock->ops->setsockopt() for MPTCP */
2070 351 : return IS_ENABLED(CONFIG_MPTCP) &&
2071 : sk->sk_protocol == IPPROTO_MPTCP &&
2072 : sk->sk_type == SOCK_STREAM &&
2073 : (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2074 : }
2075 :
2076 : /*
2077 : * Set a socket option. Because we don't know the option lengths we have
2078 : * to pass the user mode parameter for the protocols to sort out.
2079 : */
2080 370 : int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2081 : int optlen)
2082 : {
2083 370 : sockptr_t optval = USER_SOCKPTR(user_optval);
2084 370 : char *kernel_optval = NULL;
2085 370 : int err, fput_needed;
2086 370 : struct socket *sock;
2087 :
2088 370 : if (optlen < 0)
2089 : return -EINVAL;
2090 :
2091 370 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
2092 370 : if (!sock)
2093 0 : return err;
2094 :
2095 370 : err = security_socket_setsockopt(sock, level, optname);
2096 370 : if (err)
2097 : goto out_put;
2098 :
2099 370 : if (!in_compat_syscall())
2100 : err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2101 : user_optval, &optlen,
2102 : &kernel_optval);
2103 370 : if (err < 0)
2104 : goto out_put;
2105 370 : if (err > 0) {
2106 : err = 0;
2107 : goto out_put;
2108 : }
2109 :
2110 370 : if (kernel_optval)
2111 370 : optval = KERNEL_SOCKPTR(kernel_optval);
2112 370 : if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2113 351 : err = sock_setsockopt(sock, level, optname, optval, optlen);
2114 19 : else if (unlikely(!sock->ops->setsockopt))
2115 0 : err = -EOPNOTSUPP;
2116 : else
2117 19 : err = sock->ops->setsockopt(sock, level, optname, optval,
2118 : optlen);
2119 370 : kfree(kernel_optval);
2120 370 : out_put:
2121 370 : fput_light(sock->file, fput_needed);
2122 370 : return err;
2123 : }
2124 :
2125 740 : SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2126 : char __user *, optval, int, optlen)
2127 : {
2128 370 : return __sys_setsockopt(fd, level, optname, optval, optlen);
2129 : }
2130 :
2131 : INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2132 : int optname));
2133 :
2134 : /*
2135 : * Get a socket option. Because we don't know the option lengths we have
2136 : * to pass a user mode parameter for the protocols to sort out.
2137 : */
2138 632 : int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2139 : int __user *optlen)
2140 : {
2141 632 : int err, fput_needed;
2142 632 : struct socket *sock;
2143 632 : int max_optlen;
2144 :
2145 632 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
2146 632 : if (!sock)
2147 0 : return err;
2148 :
2149 632 : err = security_socket_getsockopt(sock, level, optname);
2150 632 : if (err)
2151 : goto out_put;
2152 :
2153 632 : if (!in_compat_syscall())
2154 632 : max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2155 :
2156 632 : if (level == SOL_SOCKET)
2157 623 : err = sock_getsockopt(sock, level, optname, optval, optlen);
2158 9 : else if (unlikely(!sock->ops->getsockopt))
2159 0 : err = -EOPNOTSUPP;
2160 : else
2161 9 : err = sock->ops->getsockopt(sock, level, optname, optval,
2162 : optlen);
2163 :
2164 632 : if (!in_compat_syscall())
2165 : err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2166 : optval, optlen, max_optlen,
2167 : err);
2168 632 : out_put:
2169 632 : fput_light(sock->file, fput_needed);
2170 632 : return err;
2171 : }
2172 :
2173 1264 : SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2174 : char __user *, optval, int __user *, optlen)
2175 : {
2176 632 : return __sys_getsockopt(fd, level, optname, optval, optlen);
2177 : }
2178 :
2179 : /*
2180 : * Shutdown a socket.
2181 : */
2182 :
2183 90 : int __sys_shutdown_sock(struct socket *sock, int how)
2184 : {
2185 90 : int err;
2186 :
2187 0 : err = security_socket_shutdown(sock, how);
2188 90 : if (!err)
2189 90 : err = sock->ops->shutdown(sock, how);
2190 :
2191 90 : return err;
2192 : }
2193 :
2194 90 : int __sys_shutdown(int fd, int how)
2195 : {
2196 90 : int err, fput_needed;
2197 90 : struct socket *sock;
2198 :
2199 90 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
2200 90 : if (sock != NULL) {
2201 90 : err = __sys_shutdown_sock(sock, how);
2202 90 : fput_light(sock->file, fput_needed);
2203 : }
2204 90 : return err;
2205 : }
2206 :
2207 180 : SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2208 : {
2209 90 : return __sys_shutdown(fd, how);
2210 : }
2211 :
2212 : /* A couple of helpful macros for getting the address of the 32/64 bit
2213 : * fields which are the same type (int / unsigned) on our platforms.
2214 : */
2215 : #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2216 : #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2217 : #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2218 :
2219 : struct used_address {
2220 : struct sockaddr_storage name;
2221 : unsigned int name_len;
2222 : };
2223 :
2224 6740 : int __copy_msghdr_from_user(struct msghdr *kmsg,
2225 : struct user_msghdr __user *umsg,
2226 : struct sockaddr __user **save_addr,
2227 : struct iovec __user **uiov, size_t *nsegs)
2228 : {
2229 6740 : struct user_msghdr msg;
2230 6740 : ssize_t err;
2231 :
2232 6740 : if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2233 : return -EFAULT;
2234 :
2235 6740 : kmsg->msg_control_is_user = true;
2236 6740 : kmsg->msg_control_user = msg.msg_control;
2237 6740 : kmsg->msg_controllen = msg.msg_controllen;
2238 6740 : kmsg->msg_flags = msg.msg_flags;
2239 :
2240 6740 : kmsg->msg_namelen = msg.msg_namelen;
2241 6740 : if (!msg.msg_name)
2242 5357 : kmsg->msg_namelen = 0;
2243 :
2244 6740 : if (kmsg->msg_namelen < 0)
2245 : return -EINVAL;
2246 :
2247 6740 : if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2248 0 : kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2249 :
2250 6740 : if (save_addr)
2251 4157 : *save_addr = msg.msg_name;
2252 :
2253 6740 : if (msg.msg_name && kmsg->msg_namelen) {
2254 1383 : if (!save_addr) {
2255 1354 : err = move_addr_to_kernel(msg.msg_name,
2256 : kmsg->msg_namelen,
2257 677 : kmsg->msg_name);
2258 677 : if (err < 0)
2259 : return err;
2260 : }
2261 : } else {
2262 5357 : kmsg->msg_name = NULL;
2263 5357 : kmsg->msg_namelen = 0;
2264 : }
2265 :
2266 6740 : if (msg.msg_iovlen > UIO_MAXIOV)
2267 : return -EMSGSIZE;
2268 :
2269 6740 : kmsg->msg_iocb = NULL;
2270 6740 : *uiov = msg.msg_iov;
2271 6740 : *nsegs = msg.msg_iovlen;
2272 6740 : return 0;
2273 : }
2274 :
2275 6740 : static int copy_msghdr_from_user(struct msghdr *kmsg,
2276 : struct user_msghdr __user *umsg,
2277 : struct sockaddr __user **save_addr,
2278 : struct iovec **iov)
2279 : {
2280 6740 : struct user_msghdr msg;
2281 6740 : ssize_t err;
2282 :
2283 6740 : err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2284 : &msg.msg_iovlen);
2285 6740 : if (err)
2286 : return err;
2287 :
2288 13480 : err = import_iovec(save_addr ? READ : WRITE,
2289 6740 : msg.msg_iov, msg.msg_iovlen,
2290 : UIO_FASTIOV, iov, &kmsg->msg_iter);
2291 6740 : return err < 0 ? err : 0;
2292 : }
2293 :
2294 2583 : static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2295 : unsigned int flags, struct used_address *used_address,
2296 : unsigned int allowed_msghdr_flags)
2297 : {
2298 2583 : unsigned char ctl[sizeof(struct cmsghdr) + 20]
2299 : __aligned(sizeof(__kernel_size_t));
2300 : /* 20 is size of ipv6_pktinfo */
2301 2583 : unsigned char *ctl_buf = ctl;
2302 2583 : int ctl_len;
2303 2583 : ssize_t err;
2304 :
2305 2583 : err = -ENOBUFS;
2306 :
2307 2583 : if (msg_sys->msg_controllen > INT_MAX)
2308 0 : goto out;
2309 2583 : flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2310 2583 : ctl_len = msg_sys->msg_controllen;
2311 2583 : if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2312 0 : err =
2313 0 : cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2314 : sizeof(ctl));
2315 0 : if (err)
2316 0 : goto out;
2317 0 : ctl_buf = msg_sys->msg_control;
2318 0 : ctl_len = msg_sys->msg_controllen;
2319 2583 : } else if (ctl_len) {
2320 323 : BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2321 : CMSG_ALIGN(sizeof(struct cmsghdr)));
2322 323 : if (ctl_len > sizeof(ctl)) {
2323 0 : ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2324 0 : if (ctl_buf == NULL)
2325 0 : goto out;
2326 : }
2327 323 : err = -EFAULT;
2328 646 : if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2329 0 : goto out_freectl;
2330 323 : msg_sys->msg_control = ctl_buf;
2331 323 : msg_sys->msg_control_is_user = false;
2332 : }
2333 2583 : msg_sys->msg_flags = flags;
2334 :
2335 2583 : if (sock->file->f_flags & O_NONBLOCK)
2336 2163 : msg_sys->msg_flags |= MSG_DONTWAIT;
2337 : /*
2338 : * If this is sendmmsg() and current destination address is same as
2339 : * previously succeeded address, omit asking LSM's decision.
2340 : * used_address->name_len is initialized to UINT_MAX so that the first
2341 : * destination address never matches.
2342 : */
2343 2583 : if (used_address && msg_sys->msg_name &&
2344 0 : used_address->name_len == msg_sys->msg_namelen &&
2345 0 : !memcmp(&used_address->name, msg_sys->msg_name,
2346 : used_address->name_len)) {
2347 0 : err = sock_sendmsg_nosec(sock, msg_sys);
2348 0 : goto out_freectl;
2349 : }
2350 2583 : err = sock_sendmsg(sock, msg_sys);
2351 : /*
2352 : * If this is sendmmsg() and sending to current destination address was
2353 : * successful, remember it.
2354 : */
2355 2583 : if (used_address && err >= 0) {
2356 0 : used_address->name_len = msg_sys->msg_namelen;
2357 0 : if (msg_sys->msg_name)
2358 0 : memcpy(&used_address->name, msg_sys->msg_name,
2359 : used_address->name_len);
2360 : }
2361 :
2362 2583 : out_freectl:
2363 2583 : if (ctl_buf != ctl)
2364 0 : sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2365 2583 : out:
2366 2583 : return err;
2367 : }
2368 :
2369 2583 : int sendmsg_copy_msghdr(struct msghdr *msg,
2370 : struct user_msghdr __user *umsg, unsigned flags,
2371 : struct iovec **iov)
2372 : {
2373 2583 : int err;
2374 :
2375 2583 : if (flags & MSG_CMSG_COMPAT) {
2376 0 : struct compat_msghdr __user *msg_compat;
2377 :
2378 0 : msg_compat = (struct compat_msghdr __user *) umsg;
2379 0 : err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2380 : } else {
2381 2583 : err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2382 : }
2383 2583 : if (err < 0)
2384 : return err;
2385 :
2386 : return 0;
2387 : }
2388 :
2389 2583 : static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2390 : struct msghdr *msg_sys, unsigned int flags,
2391 : struct used_address *used_address,
2392 : unsigned int allowed_msghdr_flags)
2393 : {
2394 2583 : struct sockaddr_storage address;
2395 2583 : struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2396 2583 : ssize_t err;
2397 :
2398 2583 : msg_sys->msg_name = &address;
2399 :
2400 2583 : err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2401 2583 : if (err < 0)
2402 : return err;
2403 :
2404 2583 : err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2405 : allowed_msghdr_flags);
2406 2583 : kfree(iov);
2407 2583 : return err;
2408 : }
2409 :
2410 : /*
2411 : * BSD sendmsg interface
2412 : */
2413 0 : long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2414 : unsigned int flags)
2415 : {
2416 0 : return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2417 : }
2418 :
2419 2583 : long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2420 : bool forbid_cmsg_compat)
2421 : {
2422 2583 : int fput_needed, err;
2423 2583 : struct msghdr msg_sys;
2424 2583 : struct socket *sock;
2425 :
2426 2583 : if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2427 : return -EINVAL;
2428 :
2429 2583 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
2430 2583 : if (!sock)
2431 0 : goto out;
2432 :
2433 2583 : err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2434 :
2435 2582 : fput_light(sock->file, fput_needed);
2436 2582 : out:
2437 2582 : return err;
2438 : }
2439 :
2440 5165 : SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2441 : {
2442 2583 : return __sys_sendmsg(fd, msg, flags, true);
2443 : }
2444 :
2445 : /*
2446 : * Linux sendmmsg interface
2447 : */
2448 :
2449 0 : int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2450 : unsigned int flags, bool forbid_cmsg_compat)
2451 : {
2452 0 : int fput_needed, err, datagrams;
2453 0 : struct socket *sock;
2454 0 : struct mmsghdr __user *entry;
2455 0 : struct compat_mmsghdr __user *compat_entry;
2456 0 : struct msghdr msg_sys;
2457 0 : struct used_address used_address;
2458 0 : unsigned int oflags = flags;
2459 :
2460 0 : if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2461 : return -EINVAL;
2462 :
2463 0 : if (vlen > UIO_MAXIOV)
2464 : vlen = UIO_MAXIOV;
2465 :
2466 0 : datagrams = 0;
2467 :
2468 0 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
2469 0 : if (!sock)
2470 0 : return err;
2471 :
2472 0 : used_address.name_len = UINT_MAX;
2473 0 : entry = mmsg;
2474 0 : compat_entry = (struct compat_mmsghdr __user *)mmsg;
2475 0 : err = 0;
2476 0 : flags |= MSG_BATCH;
2477 :
2478 0 : while (datagrams < vlen) {
2479 0 : if (datagrams == vlen - 1)
2480 0 : flags = oflags;
2481 :
2482 0 : if (MSG_CMSG_COMPAT & flags) {
2483 0 : err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2484 : &msg_sys, flags, &used_address, MSG_EOR);
2485 0 : if (err < 0)
2486 : break;
2487 0 : err = __put_user(err, &compat_entry->msg_len);
2488 0 : ++compat_entry;
2489 : } else {
2490 0 : err = ___sys_sendmsg(sock,
2491 : (struct user_msghdr __user *)entry,
2492 : &msg_sys, flags, &used_address, MSG_EOR);
2493 0 : if (err < 0)
2494 : break;
2495 0 : err = put_user(err, &entry->msg_len);
2496 0 : ++entry;
2497 : }
2498 :
2499 0 : if (err)
2500 : break;
2501 0 : ++datagrams;
2502 0 : if (msg_data_left(&msg_sys))
2503 : break;
2504 0 : cond_resched();
2505 : }
2506 :
2507 0 : fput_light(sock->file, fput_needed);
2508 :
2509 : /* We only return an error if no datagrams were able to be sent */
2510 0 : if (datagrams != 0)
2511 : return datagrams;
2512 :
2513 0 : return err;
2514 : }
2515 :
2516 0 : SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2517 : unsigned int, vlen, unsigned int, flags)
2518 : {
2519 0 : return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2520 : }
2521 :
2522 4157 : int recvmsg_copy_msghdr(struct msghdr *msg,
2523 : struct user_msghdr __user *umsg, unsigned flags,
2524 : struct sockaddr __user **uaddr,
2525 : struct iovec **iov)
2526 : {
2527 4157 : ssize_t err;
2528 :
2529 4157 : if (MSG_CMSG_COMPAT & flags) {
2530 0 : struct compat_msghdr __user *msg_compat;
2531 :
2532 0 : msg_compat = (struct compat_msghdr __user *) umsg;
2533 0 : err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2534 : } else {
2535 4157 : err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2536 : }
2537 4157 : if (err < 0)
2538 : return err;
2539 :
2540 : return 0;
2541 : }
2542 :
2543 4157 : static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2544 : struct user_msghdr __user *msg,
2545 : struct sockaddr __user *uaddr,
2546 : unsigned int flags, int nosec)
2547 : {
2548 4157 : struct compat_msghdr __user *msg_compat =
2549 : (struct compat_msghdr __user *) msg;
2550 4157 : int __user *uaddr_len = COMPAT_NAMELEN(msg);
2551 4157 : struct sockaddr_storage addr;
2552 4157 : unsigned long cmsg_ptr;
2553 4157 : int len;
2554 4157 : ssize_t err;
2555 :
2556 4157 : msg_sys->msg_name = &addr;
2557 4157 : cmsg_ptr = (unsigned long)msg_sys->msg_control;
2558 4157 : msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2559 :
2560 : /* We assume all kernel code knows the size of sockaddr_storage */
2561 4157 : msg_sys->msg_namelen = 0;
2562 :
2563 4157 : if (sock->file->f_flags & O_NONBLOCK)
2564 3309 : flags |= MSG_DONTWAIT;
2565 :
2566 4157 : if (unlikely(nosec))
2567 0 : err = sock_recvmsg_nosec(sock, msg_sys, flags);
2568 : else
2569 4157 : err = sock_recvmsg(sock, msg_sys, flags);
2570 :
2571 4154 : if (err < 0)
2572 729 : goto out;
2573 3425 : len = err;
2574 :
2575 3425 : if (uaddr != NULL) {
2576 706 : err = move_addr_to_user(&addr,
2577 : msg_sys->msg_namelen, uaddr,
2578 : uaddr_len);
2579 706 : if (err < 0)
2580 0 : goto out;
2581 : }
2582 3425 : err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2583 : COMPAT_FLAGS(msg));
2584 3425 : if (err)
2585 0 : goto out;
2586 3425 : if (MSG_CMSG_COMPAT & flags)
2587 0 : err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2588 : &msg_compat->msg_controllen);
2589 : else
2590 3425 : err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2591 : &msg->msg_controllen);
2592 3425 : if (err)
2593 0 : goto out;
2594 3425 : err = len;
2595 4154 : out:
2596 4154 : return err;
2597 : }
2598 :
2599 4157 : static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2600 : struct msghdr *msg_sys, unsigned int flags, int nosec)
2601 : {
2602 4157 : struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2603 : /* user mode address pointers */
2604 4157 : struct sockaddr __user *uaddr;
2605 4157 : ssize_t err;
2606 :
2607 4157 : err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2608 4157 : if (err < 0)
2609 : return err;
2610 :
2611 4157 : err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2612 4156 : kfree(iov);
2613 4156 : return err;
2614 : }
2615 :
2616 : /*
2617 : * BSD recvmsg interface
2618 : */
2619 :
2620 0 : long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2621 : struct user_msghdr __user *umsg,
2622 : struct sockaddr __user *uaddr, unsigned int flags)
2623 : {
2624 0 : return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2625 : }
2626 :
2627 4155 : long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2628 : bool forbid_cmsg_compat)
2629 : {
2630 4155 : int fput_needed, err;
2631 4155 : struct msghdr msg_sys;
2632 4155 : struct socket *sock;
2633 :
2634 4155 : if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2635 : return -EINVAL;
2636 :
2637 4155 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
2638 4157 : if (!sock)
2639 0 : goto out;
2640 :
2641 4157 : err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2642 :
2643 4157 : fput_light(sock->file, fput_needed);
2644 4157 : out:
2645 4157 : return err;
2646 : }
2647 :
2648 8311 : SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2649 : unsigned int, flags)
2650 : {
2651 4155 : return __sys_recvmsg(fd, msg, flags, true);
2652 : }
2653 :
2654 : /*
2655 : * Linux recvmmsg interface
2656 : */
2657 :
2658 0 : static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2659 : unsigned int vlen, unsigned int flags,
2660 : struct timespec64 *timeout)
2661 : {
2662 0 : int fput_needed, err, datagrams;
2663 0 : struct socket *sock;
2664 0 : struct mmsghdr __user *entry;
2665 0 : struct compat_mmsghdr __user *compat_entry;
2666 0 : struct msghdr msg_sys;
2667 0 : struct timespec64 end_time;
2668 0 : struct timespec64 timeout64;
2669 :
2670 0 : if (timeout &&
2671 0 : poll_select_set_timeout(&end_time, timeout->tv_sec,
2672 : timeout->tv_nsec))
2673 : return -EINVAL;
2674 :
2675 0 : datagrams = 0;
2676 :
2677 0 : sock = sockfd_lookup_light(fd, &err, &fput_needed);
2678 0 : if (!sock)
2679 0 : return err;
2680 :
2681 0 : if (likely(!(flags & MSG_ERRQUEUE))) {
2682 0 : err = sock_error(sock->sk);
2683 0 : if (err) {
2684 0 : datagrams = err;
2685 0 : goto out_put;
2686 : }
2687 : }
2688 :
2689 : entry = mmsg;
2690 : compat_entry = (struct compat_mmsghdr __user *)mmsg;
2691 :
2692 0 : while (datagrams < vlen) {
2693 : /*
2694 : * No need to ask LSM for more than the first datagram.
2695 : */
2696 0 : if (MSG_CMSG_COMPAT & flags) {
2697 0 : err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2698 : &msg_sys, flags & ~MSG_WAITFORONE,
2699 : datagrams);
2700 0 : if (err < 0)
2701 : break;
2702 0 : err = __put_user(err, &compat_entry->msg_len);
2703 0 : ++compat_entry;
2704 : } else {
2705 0 : err = ___sys_recvmsg(sock,
2706 : (struct user_msghdr __user *)entry,
2707 : &msg_sys, flags & ~MSG_WAITFORONE,
2708 : datagrams);
2709 0 : if (err < 0)
2710 : break;
2711 0 : err = put_user(err, &entry->msg_len);
2712 0 : ++entry;
2713 : }
2714 :
2715 0 : if (err)
2716 : break;
2717 0 : ++datagrams;
2718 :
2719 : /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2720 0 : if (flags & MSG_WAITFORONE)
2721 0 : flags |= MSG_DONTWAIT;
2722 :
2723 0 : if (timeout) {
2724 0 : ktime_get_ts64(&timeout64);
2725 0 : *timeout = timespec64_sub(end_time, timeout64);
2726 0 : if (timeout->tv_sec < 0) {
2727 0 : timeout->tv_sec = timeout->tv_nsec = 0;
2728 0 : break;
2729 : }
2730 :
2731 : /* Timeout, return less than vlen datagrams */
2732 0 : if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2733 : break;
2734 : }
2735 :
2736 : /* Out of band data, return right away */
2737 0 : if (msg_sys.msg_flags & MSG_OOB)
2738 : break;
2739 0 : cond_resched();
2740 : }
2741 :
2742 0 : if (err == 0)
2743 0 : goto out_put;
2744 :
2745 0 : if (datagrams == 0) {
2746 0 : datagrams = err;
2747 0 : goto out_put;
2748 : }
2749 :
2750 : /*
2751 : * We may return less entries than requested (vlen) if the
2752 : * sock is non block and there aren't enough datagrams...
2753 : */
2754 0 : if (err != -EAGAIN) {
2755 : /*
2756 : * ... or if recvmsg returns an error after we
2757 : * received some datagrams, where we record the
2758 : * error to return on the next call or if the
2759 : * app asks about it using getsockopt(SO_ERROR).
2760 : */
2761 0 : sock->sk->sk_err = -err;
2762 : }
2763 0 : out_put:
2764 0 : fput_light(sock->file, fput_needed);
2765 :
2766 : return datagrams;
2767 : }
2768 :
2769 0 : int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2770 : unsigned int vlen, unsigned int flags,
2771 : struct __kernel_timespec __user *timeout,
2772 : struct old_timespec32 __user *timeout32)
2773 : {
2774 0 : int datagrams;
2775 0 : struct timespec64 timeout_sys;
2776 :
2777 0 : if (timeout && get_timespec64(&timeout_sys, timeout))
2778 : return -EFAULT;
2779 :
2780 0 : if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2781 : return -EFAULT;
2782 :
2783 0 : if (!timeout && !timeout32)
2784 0 : return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2785 :
2786 0 : datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2787 :
2788 0 : if (datagrams <= 0)
2789 : return datagrams;
2790 :
2791 0 : if (timeout && put_timespec64(&timeout_sys, timeout))
2792 0 : datagrams = -EFAULT;
2793 :
2794 0 : if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2795 0 : datagrams = -EFAULT;
2796 :
2797 : return datagrams;
2798 : }
2799 :
2800 0 : SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2801 : unsigned int, vlen, unsigned int, flags,
2802 : struct __kernel_timespec __user *, timeout)
2803 : {
2804 0 : if (flags & MSG_CMSG_COMPAT)
2805 : return -EINVAL;
2806 :
2807 0 : return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2808 : }
2809 :
2810 : #ifdef CONFIG_COMPAT_32BIT_TIME
2811 : SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2812 : unsigned int, vlen, unsigned int, flags,
2813 : struct old_timespec32 __user *, timeout)
2814 : {
2815 : if (flags & MSG_CMSG_COMPAT)
2816 : return -EINVAL;
2817 :
2818 : return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2819 : }
2820 : #endif
2821 :
2822 : #ifdef __ARCH_WANT_SYS_SOCKETCALL
2823 : /* Argument list sizes for sys_socketcall */
2824 : #define AL(x) ((x) * sizeof(unsigned long))
2825 : static const unsigned char nargs[21] = {
2826 : AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2827 : AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2828 : AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2829 : AL(4), AL(5), AL(4)
2830 : };
2831 :
2832 : #undef AL
2833 :
2834 : /*
2835 : * System call vectors.
2836 : *
2837 : * Argument checking cleaned up. Saved 20% in size.
2838 : * This function doesn't need to set the kernel lock because
2839 : * it is set by the callees.
2840 : */
2841 :
2842 0 : SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2843 : {
2844 0 : unsigned long a[AUDITSC_ARGS];
2845 0 : unsigned long a0, a1;
2846 0 : int err;
2847 0 : unsigned int len;
2848 :
2849 0 : if (call < 1 || call > SYS_SENDMMSG)
2850 : return -EINVAL;
2851 0 : call = array_index_nospec(call, SYS_SENDMMSG + 1);
2852 :
2853 0 : len = nargs[call];
2854 0 : if (len > sizeof(a))
2855 : return -EINVAL;
2856 :
2857 : /* copy_from_user should be SMP safe. */
2858 0 : if (copy_from_user(a, args, len))
2859 : return -EFAULT;
2860 :
2861 0 : err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2862 0 : if (err)
2863 : return err;
2864 :
2865 0 : a0 = a[0];
2866 0 : a1 = a[1];
2867 :
2868 0 : switch (call) {
2869 0 : case SYS_SOCKET:
2870 0 : err = __sys_socket(a0, a1, a[2]);
2871 0 : break;
2872 0 : case SYS_BIND:
2873 0 : err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2874 0 : break;
2875 0 : case SYS_CONNECT:
2876 0 : err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2877 0 : break;
2878 0 : case SYS_LISTEN:
2879 0 : err = __sys_listen(a0, a1);
2880 0 : break;
2881 0 : case SYS_ACCEPT:
2882 0 : err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2883 0 : (int __user *)a[2], 0);
2884 0 : break;
2885 0 : case SYS_GETSOCKNAME:
2886 0 : err =
2887 0 : __sys_getsockname(a0, (struct sockaddr __user *)a1,
2888 0 : (int __user *)a[2]);
2889 0 : break;
2890 0 : case SYS_GETPEERNAME:
2891 0 : err =
2892 0 : __sys_getpeername(a0, (struct sockaddr __user *)a1,
2893 0 : (int __user *)a[2]);
2894 0 : break;
2895 0 : case SYS_SOCKETPAIR:
2896 0 : err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2897 0 : break;
2898 0 : case SYS_SEND:
2899 0 : err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2900 : NULL, 0);
2901 0 : break;
2902 0 : case SYS_SENDTO:
2903 0 : err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2904 0 : (struct sockaddr __user *)a[4], a[5]);
2905 0 : break;
2906 0 : case SYS_RECV:
2907 0 : err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2908 : NULL, NULL);
2909 0 : break;
2910 0 : case SYS_RECVFROM:
2911 0 : err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2912 0 : (struct sockaddr __user *)a[4],
2913 0 : (int __user *)a[5]);
2914 0 : break;
2915 0 : case SYS_SHUTDOWN:
2916 0 : err = __sys_shutdown(a0, a1);
2917 0 : break;
2918 0 : case SYS_SETSOCKOPT:
2919 0 : err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2920 0 : a[4]);
2921 0 : break;
2922 0 : case SYS_GETSOCKOPT:
2923 0 : err =
2924 0 : __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2925 0 : (int __user *)a[4]);
2926 0 : break;
2927 0 : case SYS_SENDMSG:
2928 0 : err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2929 0 : a[2], true);
2930 0 : break;
2931 0 : case SYS_SENDMMSG:
2932 0 : err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2933 0 : a[3], true);
2934 0 : break;
2935 0 : case SYS_RECVMSG:
2936 0 : err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2937 0 : a[2], true);
2938 0 : break;
2939 : case SYS_RECVMMSG:
2940 0 : if (IS_ENABLED(CONFIG_64BIT))
2941 0 : err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2942 0 : a[2], a[3],
2943 0 : (struct __kernel_timespec __user *)a[4],
2944 : NULL);
2945 : else
2946 : err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2947 : a[2], a[3], NULL,
2948 : (struct old_timespec32 __user *)a[4]);
2949 0 : break;
2950 0 : case SYS_ACCEPT4:
2951 0 : err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2952 0 : (int __user *)a[2], a[3]);
2953 0 : break;
2954 : default:
2955 : err = -EINVAL;
2956 : break;
2957 : }
2958 0 : return err;
2959 : }
2960 :
2961 : #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2962 :
2963 : /**
2964 : * sock_register - add a socket protocol handler
2965 : * @ops: description of protocol
2966 : *
2967 : * This function is called by a protocol handler that wants to
2968 : * advertise its address family, and have it linked into the
2969 : * socket interface. The value ops->family corresponds to the
2970 : * socket system call protocol family.
2971 : */
2972 4 : int sock_register(const struct net_proto_family *ops)
2973 : {
2974 4 : int err;
2975 :
2976 4 : if (ops->family >= NPROTO) {
2977 0 : pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2978 0 : return -ENOBUFS;
2979 : }
2980 :
2981 4 : spin_lock(&net_family_lock);
2982 8 : if (rcu_dereference_protected(net_families[ops->family],
2983 : lockdep_is_held(&net_family_lock)))
2984 : err = -EEXIST;
2985 : else {
2986 4 : rcu_assign_pointer(net_families[ops->family], ops);
2987 4 : err = 0;
2988 : }
2989 4 : spin_unlock(&net_family_lock);
2990 :
2991 4 : pr_info("NET: Registered protocol family %d\n", ops->family);
2992 4 : return err;
2993 : }
2994 : EXPORT_SYMBOL(sock_register);
2995 :
2996 : /**
2997 : * sock_unregister - remove a protocol handler
2998 : * @family: protocol family to remove
2999 : *
3000 : * This function is called by a protocol handler that wants to
3001 : * remove its address family, and have it unlinked from the
3002 : * new socket creation.
3003 : *
3004 : * If protocol handler is a module, then it can use module reference
3005 : * counts to protect against new references. If protocol handler is not
3006 : * a module then it needs to provide its own protection in
3007 : * the ops->create routine.
3008 : */
3009 0 : void sock_unregister(int family)
3010 : {
3011 0 : BUG_ON(family < 0 || family >= NPROTO);
3012 :
3013 0 : spin_lock(&net_family_lock);
3014 0 : RCU_INIT_POINTER(net_families[family], NULL);
3015 0 : spin_unlock(&net_family_lock);
3016 :
3017 0 : synchronize_rcu();
3018 :
3019 0 : pr_info("NET: Unregistered protocol family %d\n", family);
3020 0 : }
3021 : EXPORT_SYMBOL(sock_unregister);
3022 :
3023 0 : bool sock_is_registered(int family)
3024 : {
3025 0 : return family < NPROTO && rcu_access_pointer(net_families[family]);
3026 : }
3027 :
3028 1 : static int __init sock_init(void)
3029 : {
3030 1 : int err;
3031 : /*
3032 : * Initialize the network sysctl infrastructure.
3033 : */
3034 1 : err = net_sysctl_init();
3035 1 : if (err)
3036 0 : goto out;
3037 :
3038 : /*
3039 : * Initialize skbuff SLAB cache
3040 : */
3041 1 : skb_init();
3042 :
3043 : /*
3044 : * Initialize the protocols module.
3045 : */
3046 :
3047 1 : init_inodecache();
3048 :
3049 1 : err = register_filesystem(&sock_fs_type);
3050 1 : if (err)
3051 0 : goto out;
3052 1 : sock_mnt = kern_mount(&sock_fs_type);
3053 1 : if (IS_ERR(sock_mnt)) {
3054 0 : err = PTR_ERR(sock_mnt);
3055 0 : goto out_mount;
3056 : }
3057 :
3058 : /* The real protocol initialization is performed in later initcalls.
3059 : */
3060 :
3061 : #ifdef CONFIG_NETFILTER
3062 1 : err = netfilter_init();
3063 1 : if (err)
3064 0 : goto out;
3065 : #endif
3066 :
3067 : ptp_classifier_init();
3068 :
3069 1 : out:
3070 1 : return err;
3071 :
3072 0 : out_mount:
3073 0 : unregister_filesystem(&sock_fs_type);
3074 0 : goto out;
3075 : }
3076 :
3077 : core_initcall(sock_init); /* early initcall */
3078 :
3079 : #ifdef CONFIG_PROC_FS
3080 0 : void socket_seq_show(struct seq_file *seq)
3081 : {
3082 0 : seq_printf(seq, "sockets: used %d\n",
3083 0 : sock_inuse_get(seq->private));
3084 0 : }
3085 : #endif /* CONFIG_PROC_FS */
3086 :
3087 : #ifdef CONFIG_COMPAT
3088 0 : static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3089 : {
3090 0 : struct compat_ifconf ifc32;
3091 0 : struct ifconf ifc;
3092 0 : int err;
3093 :
3094 0 : if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3095 : return -EFAULT;
3096 :
3097 0 : ifc.ifc_len = ifc32.ifc_len;
3098 0 : ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3099 :
3100 0 : rtnl_lock();
3101 0 : err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3102 0 : rtnl_unlock();
3103 0 : if (err)
3104 : return err;
3105 :
3106 0 : ifc32.ifc_len = ifc.ifc_len;
3107 0 : if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3108 0 : return -EFAULT;
3109 :
3110 : return 0;
3111 : }
3112 :
3113 0 : static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
3114 : {
3115 0 : struct compat_ethtool_rxnfc __user *compat_rxnfc;
3116 0 : bool convert_in = false, convert_out = false;
3117 0 : size_t buf_size = 0;
3118 0 : struct ethtool_rxnfc __user *rxnfc = NULL;
3119 0 : struct ifreq ifr;
3120 0 : u32 rule_cnt = 0, actual_rule_cnt;
3121 0 : u32 ethcmd;
3122 0 : u32 data;
3123 0 : int ret;
3124 :
3125 0 : if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3126 : return -EFAULT;
3127 :
3128 0 : compat_rxnfc = compat_ptr(data);
3129 :
3130 0 : if (get_user(ethcmd, &compat_rxnfc->cmd))
3131 : return -EFAULT;
3132 :
3133 : /* Most ethtool structures are defined without padding.
3134 : * Unfortunately struct ethtool_rxnfc is an exception.
3135 : */
3136 0 : switch (ethcmd) {
3137 : default:
3138 : break;
3139 0 : case ETHTOOL_GRXCLSRLALL:
3140 : /* Buffer size is variable */
3141 0 : if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3142 : return -EFAULT;
3143 0 : if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3144 : return -ENOMEM;
3145 0 : buf_size += rule_cnt * sizeof(u32);
3146 : fallthrough;
3147 : case ETHTOOL_GRXRINGS:
3148 : case ETHTOOL_GRXCLSRLCNT:
3149 : case ETHTOOL_GRXCLSRULE:
3150 : case ETHTOOL_SRXCLSRLINS:
3151 : convert_out = true;
3152 0 : fallthrough;
3153 0 : case ETHTOOL_SRXCLSRLDEL:
3154 0 : buf_size += sizeof(struct ethtool_rxnfc);
3155 0 : convert_in = true;
3156 0 : rxnfc = compat_alloc_user_space(buf_size);
3157 0 : break;
3158 : }
3159 :
3160 0 : if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3161 : return -EFAULT;
3162 :
3163 0 : ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3164 :
3165 0 : if (convert_in) {
3166 : /* We expect there to be holes between fs.m_ext and
3167 : * fs.ring_cookie and at the end of fs, but nowhere else.
3168 : */
3169 0 : BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3170 : sizeof(compat_rxnfc->fs.m_ext) !=
3171 : offsetof(struct ethtool_rxnfc, fs.m_ext) +
3172 : sizeof(rxnfc->fs.m_ext));
3173 0 : BUILD_BUG_ON(
3174 : offsetof(struct compat_ethtool_rxnfc, fs.location) -
3175 : offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3176 : offsetof(struct ethtool_rxnfc, fs.location) -
3177 : offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3178 :
3179 0 : if (copy_in_user(rxnfc, compat_rxnfc,
3180 0 : (void __user *)(&rxnfc->fs.m_ext + 1) -
3181 0 : (void __user *)rxnfc) ||
3182 0 : copy_in_user(&rxnfc->fs.ring_cookie,
3183 0 : &compat_rxnfc->fs.ring_cookie,
3184 0 : (void __user *)(&rxnfc->fs.location + 1) -
3185 0 : (void __user *)&rxnfc->fs.ring_cookie))
3186 0 : return -EFAULT;
3187 0 : if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3188 0 : if (put_user(rule_cnt, &rxnfc->rule_cnt))
3189 : return -EFAULT;
3190 0 : } else if (copy_in_user(&rxnfc->rule_cnt,
3191 0 : &compat_rxnfc->rule_cnt,
3192 : sizeof(rxnfc->rule_cnt)))
3193 0 : return -EFAULT;
3194 : }
3195 :
3196 0 : ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3197 0 : if (ret)
3198 : return ret;
3199 :
3200 0 : if (convert_out) {
3201 0 : if (copy_in_user(compat_rxnfc, rxnfc,
3202 0 : (const void __user *)(&rxnfc->fs.m_ext + 1) -
3203 0 : (const void __user *)rxnfc) ||
3204 0 : copy_in_user(&compat_rxnfc->fs.ring_cookie,
3205 : &rxnfc->fs.ring_cookie,
3206 0 : (const void __user *)(&rxnfc->fs.location + 1) -
3207 0 : (const void __user *)&rxnfc->fs.ring_cookie) ||
3208 0 : copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3209 : sizeof(rxnfc->rule_cnt)))
3210 0 : return -EFAULT;
3211 :
3212 0 : if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3213 : /* As an optimisation, we only copy the actual
3214 : * number of rules that the underlying
3215 : * function returned. Since Mallory might
3216 : * change the rule count in user memory, we
3217 : * check that it is less than the rule count
3218 : * originally given (as the user buffer size),
3219 : * which has been range-checked.
3220 : */
3221 0 : if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3222 : return -EFAULT;
3223 0 : if (actual_rule_cnt < rule_cnt)
3224 : rule_cnt = actual_rule_cnt;
3225 0 : if (copy_in_user(&compat_rxnfc->rule_locs[0],
3226 0 : &rxnfc->rule_locs[0],
3227 : rule_cnt * sizeof(u32)))
3228 0 : return -EFAULT;
3229 : }
3230 : }
3231 :
3232 : return 0;
3233 : }
3234 :
3235 0 : static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3236 : {
3237 0 : compat_uptr_t uptr32;
3238 0 : struct ifreq ifr;
3239 0 : void __user *saved;
3240 0 : int err;
3241 :
3242 0 : if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3243 : return -EFAULT;
3244 :
3245 0 : if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3246 : return -EFAULT;
3247 :
3248 0 : saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3249 0 : ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3250 :
3251 0 : err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3252 0 : if (!err) {
3253 0 : ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3254 0 : if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3255 0 : err = -EFAULT;
3256 : }
3257 : return err;
3258 : }
3259 :
3260 : /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3261 0 : static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3262 : struct compat_ifreq __user *u_ifreq32)
3263 : {
3264 0 : struct ifreq ifreq;
3265 0 : u32 data32;
3266 :
3267 0 : if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3268 : return -EFAULT;
3269 0 : if (get_user(data32, &u_ifreq32->ifr_data))
3270 : return -EFAULT;
3271 0 : ifreq.ifr_data = compat_ptr(data32);
3272 :
3273 0 : return dev_ioctl(net, cmd, &ifreq, NULL);
3274 : }
3275 :
3276 0 : static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3277 : unsigned int cmd,
3278 : struct compat_ifreq __user *uifr32)
3279 : {
3280 0 : struct ifreq __user *uifr;
3281 0 : int err;
3282 :
3283 : /* Handle the fact that while struct ifreq has the same *layout* on
3284 : * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3285 : * which are handled elsewhere, it still has different *size* due to
3286 : * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3287 : * resulting in struct ifreq being 32 and 40 bytes respectively).
3288 : * As a result, if the struct happens to be at the end of a page and
3289 : * the next page isn't readable/writable, we get a fault. To prevent
3290 : * that, copy back and forth to the full size.
3291 : */
3292 :
3293 0 : uifr = compat_alloc_user_space(sizeof(*uifr));
3294 0 : if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3295 0 : return -EFAULT;
3296 :
3297 0 : err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3298 :
3299 0 : if (!err) {
3300 0 : switch (cmd) {
3301 : case SIOCGIFFLAGS:
3302 : case SIOCGIFMETRIC:
3303 : case SIOCGIFMTU:
3304 : case SIOCGIFMEM:
3305 : case SIOCGIFHWADDR:
3306 : case SIOCGIFINDEX:
3307 : case SIOCGIFADDR:
3308 : case SIOCGIFBRDADDR:
3309 : case SIOCGIFDSTADDR:
3310 : case SIOCGIFNETMASK:
3311 : case SIOCGIFPFLAGS:
3312 : case SIOCGIFTXQLEN:
3313 : case SIOCGMIIPHY:
3314 : case SIOCGMIIREG:
3315 : case SIOCGIFNAME:
3316 0 : if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3317 : err = -EFAULT;
3318 : break;
3319 : }
3320 0 : }
3321 : return err;
3322 : }
3323 :
3324 0 : static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3325 : struct compat_ifreq __user *uifr32)
3326 : {
3327 0 : struct ifreq ifr;
3328 0 : struct compat_ifmap __user *uifmap32;
3329 0 : int err;
3330 :
3331 0 : uifmap32 = &uifr32->ifr_ifru.ifru_map;
3332 0 : err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3333 0 : err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3334 0 : err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3335 0 : err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3336 0 : err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3337 0 : err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3338 0 : err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3339 0 : if (err)
3340 : return -EFAULT;
3341 :
3342 0 : err = dev_ioctl(net, cmd, &ifr, NULL);
3343 :
3344 0 : if (cmd == SIOCGIFMAP && !err) {
3345 0 : err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3346 0 : err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3347 0 : err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3348 0 : err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3349 0 : err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3350 0 : err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3351 0 : err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3352 0 : if (err)
3353 0 : err = -EFAULT;
3354 : }
3355 : return err;
3356 : }
3357 :
3358 : /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3359 : * for some operations; this forces use of the newer bridge-utils that
3360 : * use compatible ioctls
3361 : */
3362 0 : static int old_bridge_ioctl(compat_ulong_t __user *argp)
3363 : {
3364 0 : compat_ulong_t tmp;
3365 :
3366 0 : if (get_user(tmp, argp))
3367 : return -EFAULT;
3368 0 : if (tmp == BRCTL_GET_VERSION)
3369 0 : return BRCTL_VERSION + 1;
3370 : return -EINVAL;
3371 : }
3372 :
3373 0 : static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3374 : unsigned int cmd, unsigned long arg)
3375 : {
3376 0 : void __user *argp = compat_ptr(arg);
3377 0 : struct sock *sk = sock->sk;
3378 0 : struct net *net = sock_net(sk);
3379 :
3380 0 : if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3381 0 : return compat_ifr_data_ioctl(net, cmd, argp);
3382 :
3383 0 : switch (cmd) {
3384 0 : case SIOCSIFBR:
3385 : case SIOCGIFBR:
3386 0 : return old_bridge_ioctl(argp);
3387 0 : case SIOCGIFCONF:
3388 0 : return compat_dev_ifconf(net, argp);
3389 0 : case SIOCETHTOOL:
3390 0 : return ethtool_ioctl(net, argp);
3391 0 : case SIOCWANDEV:
3392 0 : return compat_siocwandev(net, argp);
3393 0 : case SIOCGIFMAP:
3394 : case SIOCSIFMAP:
3395 0 : return compat_sioc_ifmap(net, cmd, argp);
3396 0 : case SIOCGSTAMP_OLD:
3397 : case SIOCGSTAMPNS_OLD:
3398 0 : if (!sock->ops->gettstamp)
3399 : return -ENOIOCTLCMD;
3400 0 : return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3401 : !COMPAT_USE_64BIT_TIME);
3402 :
3403 0 : case SIOCBONDSLAVEINFOQUERY:
3404 : case SIOCBONDINFOQUERY:
3405 : case SIOCSHWTSTAMP:
3406 : case SIOCGHWTSTAMP:
3407 0 : return compat_ifr_data_ioctl(net, cmd, argp);
3408 :
3409 0 : case FIOSETOWN:
3410 : case SIOCSPGRP:
3411 : case FIOGETOWN:
3412 : case SIOCGPGRP:
3413 : case SIOCBRADDBR:
3414 : case SIOCBRDELBR:
3415 : case SIOCGIFVLAN:
3416 : case SIOCSIFVLAN:
3417 : case SIOCGSKNS:
3418 : case SIOCGSTAMP_NEW:
3419 : case SIOCGSTAMPNS_NEW:
3420 0 : return sock_ioctl(file, cmd, arg);
3421 :
3422 0 : case SIOCGIFFLAGS:
3423 : case SIOCSIFFLAGS:
3424 : case SIOCGIFMETRIC:
3425 : case SIOCSIFMETRIC:
3426 : case SIOCGIFMTU:
3427 : case SIOCSIFMTU:
3428 : case SIOCGIFMEM:
3429 : case SIOCSIFMEM:
3430 : case SIOCGIFHWADDR:
3431 : case SIOCSIFHWADDR:
3432 : case SIOCADDMULTI:
3433 : case SIOCDELMULTI:
3434 : case SIOCGIFINDEX:
3435 : case SIOCGIFADDR:
3436 : case SIOCSIFADDR:
3437 : case SIOCSIFHWBROADCAST:
3438 : case SIOCDIFADDR:
3439 : case SIOCGIFBRDADDR:
3440 : case SIOCSIFBRDADDR:
3441 : case SIOCGIFDSTADDR:
3442 : case SIOCSIFDSTADDR:
3443 : case SIOCGIFNETMASK:
3444 : case SIOCSIFNETMASK:
3445 : case SIOCSIFPFLAGS:
3446 : case SIOCGIFPFLAGS:
3447 : case SIOCGIFTXQLEN:
3448 : case SIOCSIFTXQLEN:
3449 : case SIOCBRADDIF:
3450 : case SIOCBRDELIF:
3451 : case SIOCGIFNAME:
3452 : case SIOCSIFNAME:
3453 : case SIOCGMIIPHY:
3454 : case SIOCGMIIREG:
3455 : case SIOCSMIIREG:
3456 : case SIOCBONDENSLAVE:
3457 : case SIOCBONDRELEASE:
3458 : case SIOCBONDSETHWADDR:
3459 : case SIOCBONDCHANGEACTIVE:
3460 0 : return compat_ifreq_ioctl(net, sock, cmd, argp);
3461 :
3462 0 : case SIOCSARP:
3463 : case SIOCGARP:
3464 : case SIOCDARP:
3465 : case SIOCOUTQ:
3466 : case SIOCOUTQNSD:
3467 : case SIOCATMARK:
3468 0 : return sock_do_ioctl(net, sock, cmd, arg);
3469 : }
3470 :
3471 : return -ENOIOCTLCMD;
3472 : }
3473 :
3474 0 : static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3475 : unsigned long arg)
3476 : {
3477 0 : struct socket *sock = file->private_data;
3478 0 : int ret = -ENOIOCTLCMD;
3479 0 : struct sock *sk;
3480 0 : struct net *net;
3481 :
3482 0 : sk = sock->sk;
3483 0 : net = sock_net(sk);
3484 :
3485 0 : if (sock->ops->compat_ioctl)
3486 0 : ret = sock->ops->compat_ioctl(sock, cmd, arg);
3487 :
3488 0 : if (ret == -ENOIOCTLCMD &&
3489 0 : (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3490 0 : ret = compat_wext_handle_ioctl(net, cmd, arg);
3491 :
3492 0 : if (ret == -ENOIOCTLCMD)
3493 0 : ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3494 :
3495 0 : return ret;
3496 : }
3497 : #endif
3498 :
3499 : /**
3500 : * kernel_bind - bind an address to a socket (kernel space)
3501 : * @sock: socket
3502 : * @addr: address
3503 : * @addrlen: length of address
3504 : *
3505 : * Returns 0 or an error.
3506 : */
3507 :
3508 0 : int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3509 : {
3510 0 : return sock->ops->bind(sock, addr, addrlen);
3511 : }
3512 : EXPORT_SYMBOL(kernel_bind);
3513 :
3514 : /**
3515 : * kernel_listen - move socket to listening state (kernel space)
3516 : * @sock: socket
3517 : * @backlog: pending connections queue size
3518 : *
3519 : * Returns 0 or an error.
3520 : */
3521 :
3522 0 : int kernel_listen(struct socket *sock, int backlog)
3523 : {
3524 0 : return sock->ops->listen(sock, backlog);
3525 : }
3526 : EXPORT_SYMBOL(kernel_listen);
3527 :
3528 : /**
3529 : * kernel_accept - accept a connection (kernel space)
3530 : * @sock: listening socket
3531 : * @newsock: new connected socket
3532 : * @flags: flags
3533 : *
3534 : * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3535 : * If it fails, @newsock is guaranteed to be %NULL.
3536 : * Returns 0 or an error.
3537 : */
3538 :
3539 0 : int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3540 : {
3541 0 : struct sock *sk = sock->sk;
3542 0 : int err;
3543 :
3544 0 : err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3545 : newsock);
3546 0 : if (err < 0)
3547 0 : goto done;
3548 :
3549 0 : err = sock->ops->accept(sock, *newsock, flags, true);
3550 0 : if (err < 0) {
3551 0 : sock_release(*newsock);
3552 0 : *newsock = NULL;
3553 0 : goto done;
3554 : }
3555 :
3556 0 : (*newsock)->ops = sock->ops;
3557 0 : __module_get((*newsock)->ops->owner);
3558 :
3559 0 : done:
3560 0 : return err;
3561 : }
3562 : EXPORT_SYMBOL(kernel_accept);
3563 :
3564 : /**
3565 : * kernel_connect - connect a socket (kernel space)
3566 : * @sock: socket
3567 : * @addr: address
3568 : * @addrlen: address length
3569 : * @flags: flags (O_NONBLOCK, ...)
3570 : *
3571 : * For datagram sockets, @addr is the addres to which datagrams are sent
3572 : * by default, and the only address from which datagrams are received.
3573 : * For stream sockets, attempts to connect to @addr.
3574 : * Returns 0 or an error code.
3575 : */
3576 :
3577 0 : int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3578 : int flags)
3579 : {
3580 0 : return sock->ops->connect(sock, addr, addrlen, flags);
3581 : }
3582 : EXPORT_SYMBOL(kernel_connect);
3583 :
3584 : /**
3585 : * kernel_getsockname - get the address which the socket is bound (kernel space)
3586 : * @sock: socket
3587 : * @addr: address holder
3588 : *
3589 : * Fills the @addr pointer with the address which the socket is bound.
3590 : * Returns 0 or an error code.
3591 : */
3592 :
3593 0 : int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3594 : {
3595 0 : return sock->ops->getname(sock, addr, 0);
3596 : }
3597 : EXPORT_SYMBOL(kernel_getsockname);
3598 :
3599 : /**
3600 : * kernel_getpeername - get the address which the socket is connected (kernel space)
3601 : * @sock: socket
3602 : * @addr: address holder
3603 : *
3604 : * Fills the @addr pointer with the address which the socket is connected.
3605 : * Returns 0 or an error code.
3606 : */
3607 :
3608 0 : int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3609 : {
3610 0 : return sock->ops->getname(sock, addr, 1);
3611 : }
3612 : EXPORT_SYMBOL(kernel_getpeername);
3613 :
3614 : /**
3615 : * kernel_sendpage - send a &page through a socket (kernel space)
3616 : * @sock: socket
3617 : * @page: page
3618 : * @offset: page offset
3619 : * @size: total size in bytes
3620 : * @flags: flags (MSG_DONTWAIT, ...)
3621 : *
3622 : * Returns the total amount sent in bytes or an error.
3623 : */
3624 :
3625 0 : int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3626 : size_t size, int flags)
3627 : {
3628 0 : if (sock->ops->sendpage) {
3629 : /* Warn in case the improper page to zero-copy send */
3630 0 : WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3631 0 : return sock->ops->sendpage(sock, page, offset, size, flags);
3632 : }
3633 0 : return sock_no_sendpage(sock, page, offset, size, flags);
3634 : }
3635 : EXPORT_SYMBOL(kernel_sendpage);
3636 :
3637 : /**
3638 : * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3639 : * @sk: sock
3640 : * @page: page
3641 : * @offset: page offset
3642 : * @size: total size in bytes
3643 : * @flags: flags (MSG_DONTWAIT, ...)
3644 : *
3645 : * Returns the total amount sent in bytes or an error.
3646 : * Caller must hold @sk.
3647 : */
3648 :
3649 0 : int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3650 : size_t size, int flags)
3651 : {
3652 0 : struct socket *sock = sk->sk_socket;
3653 :
3654 0 : if (sock->ops->sendpage_locked)
3655 0 : return sock->ops->sendpage_locked(sk, page, offset, size,
3656 : flags);
3657 :
3658 0 : return sock_no_sendpage_locked(sk, page, offset, size, flags);
3659 : }
3660 : EXPORT_SYMBOL(kernel_sendpage_locked);
3661 :
3662 : /**
3663 : * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3664 : * @sock: socket
3665 : * @how: connection part
3666 : *
3667 : * Returns 0 or an error.
3668 : */
3669 :
3670 0 : int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3671 : {
3672 0 : return sock->ops->shutdown(sock, how);
3673 : }
3674 : EXPORT_SYMBOL(kernel_sock_shutdown);
3675 :
3676 : /**
3677 : * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3678 : * @sk: socket
3679 : *
3680 : * This routine returns the IP overhead imposed by a socket i.e.
3681 : * the length of the underlying IP header, depending on whether
3682 : * this is an IPv4 or IPv6 socket and the length from IP options turned
3683 : * on at the socket. Assumes that the caller has a lock on the socket.
3684 : */
3685 :
3686 0 : u32 kernel_sock_ip_overhead(struct sock *sk)
3687 : {
3688 0 : struct inet_sock *inet;
3689 0 : struct ip_options_rcu *opt;
3690 0 : u32 overhead = 0;
3691 : #if IS_ENABLED(CONFIG_IPV6)
3692 : struct ipv6_pinfo *np;
3693 : struct ipv6_txoptions *optv6 = NULL;
3694 : #endif /* IS_ENABLED(CONFIG_IPV6) */
3695 :
3696 0 : if (!sk)
3697 : return overhead;
3698 :
3699 0 : switch (sk->sk_family) {
3700 : case AF_INET:
3701 0 : inet = inet_sk(sk);
3702 0 : overhead += sizeof(struct iphdr);
3703 0 : opt = rcu_dereference_protected(inet->inet_opt,
3704 : sock_owned_by_user(sk));
3705 0 : if (opt)
3706 0 : overhead += opt->opt.optlen;
3707 : return overhead;
3708 : #if IS_ENABLED(CONFIG_IPV6)
3709 : case AF_INET6:
3710 : np = inet6_sk(sk);
3711 : overhead += sizeof(struct ipv6hdr);
3712 : if (np)
3713 : optv6 = rcu_dereference_protected(np->opt,
3714 : sock_owned_by_user(sk));
3715 : if (optv6)
3716 : overhead += (optv6->opt_flen + optv6->opt_nflen);
3717 : return overhead;
3718 : #endif /* IS_ENABLED(CONFIG_IPV6) */
3719 : default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3720 : return overhead;
3721 : }
3722 : }
3723 : EXPORT_SYMBOL(kernel_sock_ip_overhead);
|
