~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/socket.c

Version: ~ [ linux-5.10-rc5 ] ~ [ linux-5.9.10 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.79 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.159 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.208 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.245 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.245 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

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

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp