Version:  2.0.40 2.2.26 2.4.37 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 4.0 4.1

Linux/net/socket.c

  1 /*
  2  * NET          An implementation of the SOCKET network access protocol.
  3  *
  4  * Version:     @(#)socket.c    1.1.93  18/02/95
  5  *
  6  * Authors:     Orest Zborowski, <obz@Kodak.COM>
  7  *              Ross Biro
  8  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  9  *
 10  * Fixes:
 11  *              Anonymous       :       NOTSOCK/BADF cleanup. Error fix in
 12  *                                      shutdown()
 13  *              Alan Cox        :       verify_area() fixes
 14  *              Alan Cox        :       Removed DDI
 15  *              Jonathan Kamens :       SOCK_DGRAM reconnect bug
 16  *              Alan Cox        :       Moved a load of checks to the very
 17  *                                      top level.
 18  *              Alan Cox        :       Move address structures to/from user
 19  *                                      mode above the protocol layers.
 20  *              Rob Janssen     :       Allow 0 length sends.
 21  *              Alan Cox        :       Asynchronous I/O support (cribbed from the
 22  *                                      tty drivers).
 23  *              Niibe Yutaka    :       Asynchronous I/O for writes (4.4BSD style)
 24  *              Jeff Uphoff     :       Made max number of sockets command-line
 25  *                                      configurable.
 26  *              Matti Aarnio    :       Made the number of sockets dynamic,
 27  *                                      to be allocated when needed, and mr.
 28  *                                      Uphoff's max is used as max to be
 29  *                                      allowed to allocate.
 30  *              Linus           :       Argh. removed all the socket allocation
 31  *                                      altogether: it's in the inode now.
 32  *              Alan Cox        :       Made sock_alloc()/sock_release() public
 33  *                                      for NetROM and future kernel nfsd type
 34  *                                      stuff.
 35  *              Alan Cox        :       sendmsg/recvmsg basics.
 36  *              Tom Dyas        :       Export net symbols.
 37  *              Marcin Dalecki  :       Fixed problems with CONFIG_NET="n".
 38  *              Alan Cox        :       Added thread locking to sys_* calls
 39  *                                      for sockets. May have errors at the
 40  *                                      moment.
 41  *              Kevin Buhr      :       Fixed the dumb errors in the above.
 42  *              Andi Kleen      :       Some small cleanups, optimizations,
 43  *                                      and fixed a copy_from_user() bug.
 44  *              Tigran Aivazian :       sys_send(args) calls sys_sendto(args, NULL, 0)
 45  *              Tigran Aivazian :       Made listen(2) backlog sanity checks
 46  *                                      protocol-independent
 47  *
 48  *
 49  *              This program is free software; you can redistribute it and/or
 50  *              modify it under the terms of the GNU General Public License
 51  *              as published by the Free Software Foundation; either version
 52  *              2 of the License, or (at your option) any later version.
 53  *
 54  *
 55  *      This module is effectively the top level interface to the BSD socket
 56  *      paradigm.
 57  *
 58  *      Based upon Swansea University Computer Society NET3.039
 59  */
 60 
 61 #include <linux/mm.h>
 62 #include <linux/socket.h>
 63 #include <linux/file.h>
 64 #include <linux/net.h>
 65 #include <linux/interrupt.h>
 66 #include <linux/thread_info.h>
 67 #include <linux/rcupdate.h>
 68 #include <linux/netdevice.h>
 69 #include <linux/proc_fs.h>
 70 #include <linux/seq_file.h>
 71 #include <linux/mutex.h>
 72 #include <linux/if_bridge.h>
 73 #include <linux/if_frad.h>
 74 #include <linux/if_vlan.h>
 75 #include <linux/ptp_classify.h>
 76 #include <linux/init.h>
 77 #include <linux/poll.h>
 78 #include <linux/cache.h>
 79 #include <linux/module.h>
 80 #include <linux/highmem.h>
 81 #include <linux/mount.h>
 82 #include <linux/security.h>
 83 #include <linux/syscalls.h>
 84 #include <linux/compat.h>
 85 #include <linux/kmod.h>
 86 #include <linux/audit.h>
 87 #include <linux/wireless.h>
 88 #include <linux/nsproxy.h>
 89 #include <linux/magic.h>
 90 #include <linux/slab.h>
 91 #include <linux/xattr.h>
 92 
 93 #include <asm/uaccess.h>
 94 #include <asm/unistd.h>
 95 
 96 #include <net/compat.h>
 97 #include <net/wext.h>
 98 #include <net/cls_cgroup.h>
 99 
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
102 
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
114 #endif
115 
116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119 
120 static int sock_close(struct inode *inode, struct file *file);
121 static unsigned int sock_poll(struct file *file,
122                               struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124 #ifdef CONFIG_COMPAT
125 static long compat_sock_ioctl(struct file *file,
126                               unsigned int cmd, unsigned long arg);
127 #endif
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130                              int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132                                 struct pipe_inode_info *pipe, size_t len,
133                                 unsigned int flags);
134 
135 /*
136  *      Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137  *      in the operation structures but are done directly via the socketcall() multiplexor.
138  */
139 
140 static const struct file_operations socket_file_ops = {
141         .owner =        THIS_MODULE,
142         .llseek =       no_llseek,
143         .read_iter =    sock_read_iter,
144         .write_iter =   sock_write_iter,
145         .poll =         sock_poll,
146         .unlocked_ioctl = sock_ioctl,
147 #ifdef CONFIG_COMPAT
148         .compat_ioctl = compat_sock_ioctl,
149 #endif
150         .mmap =         sock_mmap,
151         .release =      sock_close,
152         .fasync =       sock_fasync,
153         .sendpage =     sock_sendpage,
154         .splice_write = generic_splice_sendpage,
155         .splice_read =  sock_splice_read,
156 };
157 
158 /*
159  *      The protocol list. Each protocol is registered in here.
160  */
161 
162 static DEFINE_SPINLOCK(net_family_lock);
163 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164 
165 /*
166  *      Statistics counters of the socket lists
167  */
168 
169 static DEFINE_PER_CPU(int, sockets_in_use);
170 
171 /*
172  * Support routines.
173  * Move socket addresses back and forth across the kernel/user
174  * divide and look after the messy bits.
175  */
176 
177 /**
178  *      move_addr_to_kernel     -       copy a socket address into kernel space
179  *      @uaddr: Address in user space
180  *      @kaddr: Address in kernel space
181  *      @ulen: Length in user space
182  *
183  *      The address is copied into kernel space. If the provided address is
184  *      too long an error code of -EINVAL is returned. If the copy gives
185  *      invalid addresses -EFAULT is returned. On a success 0 is returned.
186  */
187 
188 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
189 {
190         if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
191                 return -EINVAL;
192         if (ulen == 0)
193                 return 0;
194         if (copy_from_user(kaddr, uaddr, ulen))
195                 return -EFAULT;
196         return audit_sockaddr(ulen, kaddr);
197 }
198 
199 /**
200  *      move_addr_to_user       -       copy an address to user space
201  *      @kaddr: kernel space address
202  *      @klen: length of address in kernel
203  *      @uaddr: user space address
204  *      @ulen: pointer to user length field
205  *
206  *      The value pointed to by ulen on entry is the buffer length available.
207  *      This is overwritten with the buffer space used. -EINVAL is returned
208  *      if an overlong buffer is specified or a negative buffer size. -EFAULT
209  *      is returned if either the buffer or the length field are not
210  *      accessible.
211  *      After copying the data up to the limit the user specifies, the true
212  *      length of the data is written over the length limit the user
213  *      specified. Zero is returned for a success.
214  */
215 
216 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
217                              void __user *uaddr, int __user *ulen)
218 {
219         int err;
220         int len;
221 
222         BUG_ON(klen > sizeof(struct sockaddr_storage));
223         err = get_user(len, ulen);
224         if (err)
225                 return err;
226         if (len > klen)
227                 len = klen;
228         if (len < 0)
229                 return -EINVAL;
230         if (len) {
231                 if (audit_sockaddr(klen, kaddr))
232                         return -ENOMEM;
233                 if (copy_to_user(uaddr, kaddr, len))
234                         return -EFAULT;
235         }
236         /*
237          *      "fromlen shall refer to the value before truncation.."
238          *                      1003.1g
239          */
240         return __put_user(klen, ulen);
241 }
242 
243 static struct kmem_cache *sock_inode_cachep __read_mostly;
244 
245 static struct inode *sock_alloc_inode(struct super_block *sb)
246 {
247         struct socket_alloc *ei;
248         struct socket_wq *wq;
249 
250         ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
251         if (!ei)
252                 return NULL;
253         wq = kmalloc(sizeof(*wq), GFP_KERNEL);
254         if (!wq) {
255                 kmem_cache_free(sock_inode_cachep, ei);
256                 return NULL;
257         }
258         init_waitqueue_head(&wq->wait);
259         wq->fasync_list = NULL;
260         RCU_INIT_POINTER(ei->socket.wq, wq);
261 
262         ei->socket.state = SS_UNCONNECTED;
263         ei->socket.flags = 0;
264         ei->socket.ops = NULL;
265         ei->socket.sk = NULL;
266         ei->socket.file = NULL;
267 
268         return &ei->vfs_inode;
269 }
270 
271 static void sock_destroy_inode(struct inode *inode)
272 {
273         struct socket_alloc *ei;
274         struct socket_wq *wq;
275 
276         ei = container_of(inode, struct socket_alloc, vfs_inode);
277         wq = rcu_dereference_protected(ei->socket.wq, 1);
278         kfree_rcu(wq, rcu);
279         kmem_cache_free(sock_inode_cachep, ei);
280 }
281 
282 static void init_once(void *foo)
283 {
284         struct socket_alloc *ei = (struct socket_alloc *)foo;
285 
286         inode_init_once(&ei->vfs_inode);
287 }
288 
289 static int init_inodecache(void)
290 {
291         sock_inode_cachep = kmem_cache_create("sock_inode_cache",
292                                               sizeof(struct socket_alloc),
293                                               0,
294                                               (SLAB_HWCACHE_ALIGN |
295                                                SLAB_RECLAIM_ACCOUNT |
296                                                SLAB_MEM_SPREAD),
297                                               init_once);
298         if (sock_inode_cachep == NULL)
299                 return -ENOMEM;
300         return 0;
301 }
302 
303 static const struct super_operations sockfs_ops = {
304         .alloc_inode    = sock_alloc_inode,
305         .destroy_inode  = sock_destroy_inode,
306         .statfs         = simple_statfs,
307 };
308 
309 /*
310  * sockfs_dname() is called from d_path().
311  */
312 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
313 {
314         return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
315                                 d_inode(dentry)->i_ino);
316 }
317 
318 static const struct dentry_operations sockfs_dentry_operations = {
319         .d_dname  = sockfs_dname,
320 };
321 
322 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
323                          int flags, const char *dev_name, void *data)
324 {
325         return mount_pseudo(fs_type, "socket:", &sockfs_ops,
326                 &sockfs_dentry_operations, SOCKFS_MAGIC);
327 }
328 
329 static struct vfsmount *sock_mnt __read_mostly;
330 
331 static struct file_system_type sock_fs_type = {
332         .name =         "sockfs",
333         .mount =        sockfs_mount,
334         .kill_sb =      kill_anon_super,
335 };
336 
337 /*
338  *      Obtains the first available file descriptor and sets it up for use.
339  *
340  *      These functions create file structures and maps them to fd space
341  *      of the current process. On success it returns file descriptor
342  *      and file struct implicitly stored in sock->file.
343  *      Note that another thread may close file descriptor before we return
344  *      from this function. We use the fact that now we do not refer
345  *      to socket after mapping. If one day we will need it, this
346  *      function will increment ref. count on file by 1.
347  *
348  *      In any case returned fd MAY BE not valid!
349  *      This race condition is unavoidable
350  *      with shared fd spaces, we cannot solve it inside kernel,
351  *      but we take care of internal coherence yet.
352  */
353 
354 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
355 {
356         struct qstr name = { .name = "" };
357         struct path path;
358         struct file *file;
359 
360         if (dname) {
361                 name.name = dname;
362                 name.len = strlen(name.name);
363         } else if (sock->sk) {
364                 name.name = sock->sk->sk_prot_creator->name;
365                 name.len = strlen(name.name);
366         }
367         path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
368         if (unlikely(!path.dentry))
369                 return ERR_PTR(-ENOMEM);
370         path.mnt = mntget(sock_mnt);
371 
372         d_instantiate(path.dentry, SOCK_INODE(sock));
373 
374         file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
375                   &socket_file_ops);
376         if (unlikely(IS_ERR(file))) {
377                 /* drop dentry, keep inode */
378                 ihold(d_inode(path.dentry));
379                 path_put(&path);
380                 return file;
381         }
382 
383         sock->file = file;
384         file->f_flags = O_RDWR | (flags & O_NONBLOCK);
385         file->private_data = sock;
386         return file;
387 }
388 EXPORT_SYMBOL(sock_alloc_file);
389 
390 static int sock_map_fd(struct socket *sock, int flags)
391 {
392         struct file *newfile;
393         int fd = get_unused_fd_flags(flags);
394         if (unlikely(fd < 0))
395                 return fd;
396 
397         newfile = sock_alloc_file(sock, flags, NULL);
398         if (likely(!IS_ERR(newfile))) {
399                 fd_install(fd, newfile);
400                 return fd;
401         }
402 
403         put_unused_fd(fd);
404         return PTR_ERR(newfile);
405 }
406 
407 struct socket *sock_from_file(struct file *file, int *err)
408 {
409         if (file->f_op == &socket_file_ops)
410                 return file->private_data;      /* set in sock_map_fd */
411 
412         *err = -ENOTSOCK;
413         return NULL;
414 }
415 EXPORT_SYMBOL(sock_from_file);
416 
417 /**
418  *      sockfd_lookup - Go from a file number to its socket slot
419  *      @fd: file handle
420  *      @err: pointer to an error code return
421  *
422  *      The file handle passed in is locked and the socket it is bound
423  *      too is returned. If an error occurs the err pointer is overwritten
424  *      with a negative errno code and NULL is returned. The function checks
425  *      for both invalid handles and passing a handle which is not a socket.
426  *
427  *      On a success the socket object pointer is returned.
428  */
429 
430 struct socket *sockfd_lookup(int fd, int *err)
431 {
432         struct file *file;
433         struct socket *sock;
434 
435         file = fget(fd);
436         if (!file) {
437                 *err = -EBADF;
438                 return NULL;
439         }
440 
441         sock = sock_from_file(file, err);
442         if (!sock)
443                 fput(file);
444         return sock;
445 }
446 EXPORT_SYMBOL(sockfd_lookup);
447 
448 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
449 {
450         struct fd f = fdget(fd);
451         struct socket *sock;
452 
453         *err = -EBADF;
454         if (f.file) {
455                 sock = sock_from_file(f.file, err);
456                 if (likely(sock)) {
457                         *fput_needed = f.flags;
458                         return sock;
459                 }
460                 fdput(f);
461         }
462         return NULL;
463 }
464 
465 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
466 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
467 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
468 static ssize_t sockfs_getxattr(struct dentry *dentry,
469                                const char *name, void *value, size_t size)
470 {
471         const char *proto_name;
472         size_t proto_size;
473         int error;
474 
475         error = -ENODATA;
476         if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
477                 proto_name = dentry->d_name.name;
478                 proto_size = strlen(proto_name);
479 
480                 if (value) {
481                         error = -ERANGE;
482                         if (proto_size + 1 > size)
483                                 goto out;
484 
485                         strncpy(value, proto_name, proto_size + 1);
486                 }
487                 error = proto_size + 1;
488         }
489 
490 out:
491         return error;
492 }
493 
494 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
495                                 size_t size)
496 {
497         ssize_t len;
498         ssize_t used = 0;
499 
500         len = security_inode_listsecurity(d_inode(dentry), buffer, size);
501         if (len < 0)
502                 return len;
503         used += len;
504         if (buffer) {
505                 if (size < used)
506                         return -ERANGE;
507                 buffer += len;
508         }
509 
510         len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
511         used += len;
512         if (buffer) {
513                 if (size < used)
514                         return -ERANGE;
515                 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
516                 buffer += len;
517         }
518 
519         return used;
520 }
521 
522 static const struct inode_operations sockfs_inode_ops = {
523         .getxattr = sockfs_getxattr,
524         .listxattr = sockfs_listxattr,
525 };
526 
527 /**
528  *      sock_alloc      -       allocate a socket
529  *
530  *      Allocate a new inode and socket object. The two are bound together
531  *      and initialised. The socket is then returned. If we are out of inodes
532  *      NULL is returned.
533  */
534 
535 static struct socket *sock_alloc(void)
536 {
537         struct inode *inode;
538         struct socket *sock;
539 
540         inode = new_inode_pseudo(sock_mnt->mnt_sb);
541         if (!inode)
542                 return NULL;
543 
544         sock = SOCKET_I(inode);
545 
546         kmemcheck_annotate_bitfield(sock, type);
547         inode->i_ino = get_next_ino();
548         inode->i_mode = S_IFSOCK | S_IRWXUGO;
549         inode->i_uid = current_fsuid();
550         inode->i_gid = current_fsgid();
551         inode->i_op = &sockfs_inode_ops;
552 
553         this_cpu_add(sockets_in_use, 1);
554         return sock;
555 }
556 
557 /**
558  *      sock_release    -       close a socket
559  *      @sock: socket to close
560  *
561  *      The socket is released from the protocol stack if it has a release
562  *      callback, and the inode is then released if the socket is bound to
563  *      an inode not a file.
564  */
565 
566 void sock_release(struct socket *sock)
567 {
568         if (sock->ops) {
569                 struct module *owner = sock->ops->owner;
570 
571                 sock->ops->release(sock);
572                 sock->ops = NULL;
573                 module_put(owner);
574         }
575 
576         if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
577                 pr_err("%s: fasync list not empty!\n", __func__);
578 
579         if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
580                 return;
581 
582         this_cpu_sub(sockets_in_use, 1);
583         if (!sock->file) {
584                 iput(SOCK_INODE(sock));
585                 return;
586         }
587         sock->file = NULL;
588 }
589 EXPORT_SYMBOL(sock_release);
590 
591 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
592 {
593         u8 flags = *tx_flags;
594 
595         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
596                 flags |= SKBTX_HW_TSTAMP;
597 
598         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
599                 flags |= SKBTX_SW_TSTAMP;
600 
601         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
602                 flags |= SKBTX_SCHED_TSTAMP;
603 
604         if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
605                 flags |= SKBTX_ACK_TSTAMP;
606 
607         *tx_flags = flags;
608 }
609 EXPORT_SYMBOL(__sock_tx_timestamp);
610 
611 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
612 {
613         int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
614         BUG_ON(ret == -EIOCBQUEUED);
615         return ret;
616 }
617 
618 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
619 {
620         int err = security_socket_sendmsg(sock, msg,
621                                           msg_data_left(msg));
622 
623         return err ?: sock_sendmsg_nosec(sock, msg);
624 }
625 EXPORT_SYMBOL(sock_sendmsg);
626 
627 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
628                    struct kvec *vec, size_t num, size_t size)
629 {
630         iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
631         return sock_sendmsg(sock, msg);
632 }
633 EXPORT_SYMBOL(kernel_sendmsg);
634 
635 /*
636  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
637  */
638 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
639         struct sk_buff *skb)
640 {
641         int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
642         struct scm_timestamping tss;
643         int empty = 1;
644         struct skb_shared_hwtstamps *shhwtstamps =
645                 skb_hwtstamps(skb);
646 
647         /* Race occurred between timestamp enabling and packet
648            receiving.  Fill in the current time for now. */
649         if (need_software_tstamp && skb->tstamp.tv64 == 0)
650                 __net_timestamp(skb);
651 
652         if (need_software_tstamp) {
653                 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
654                         struct timeval tv;
655                         skb_get_timestamp(skb, &tv);
656                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
657                                  sizeof(tv), &tv);
658                 } else {
659                         struct timespec ts;
660                         skb_get_timestampns(skb, &ts);
661                         put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
662                                  sizeof(ts), &ts);
663                 }
664         }
665 
666         memset(&tss, 0, sizeof(tss));
667         if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
668             ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
669                 empty = 0;
670         if (shhwtstamps &&
671             (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
672             ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
673                 empty = 0;
674         if (!empty)
675                 put_cmsg(msg, SOL_SOCKET,
676                          SCM_TIMESTAMPING, sizeof(tss), &tss);
677 }
678 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
679 
680 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
681         struct sk_buff *skb)
682 {
683         int ack;
684 
685         if (!sock_flag(sk, SOCK_WIFI_STATUS))
686                 return;
687         if (!skb->wifi_acked_valid)
688                 return;
689 
690         ack = skb->wifi_acked;
691 
692         put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
693 }
694 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
695 
696 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
697                                    struct sk_buff *skb)
698 {
699         if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
700                 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
701                         sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
702 }
703 
704 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
705         struct sk_buff *skb)
706 {
707         sock_recv_timestamp(msg, sk, skb);
708         sock_recv_drops(msg, sk, skb);
709 }
710 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
711 
712 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
713                                      size_t size, int flags)
714 {
715         return sock->ops->recvmsg(sock, msg, size, flags);
716 }
717 
718 int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
719                  int flags)
720 {
721         int err = security_socket_recvmsg(sock, msg, size, flags);
722 
723         return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
724 }
725 EXPORT_SYMBOL(sock_recvmsg);
726 
727 /**
728  * kernel_recvmsg - Receive a message from a socket (kernel space)
729  * @sock:       The socket to receive the message from
730  * @msg:        Received message
731  * @vec:        Input s/g array for message data
732  * @num:        Size of input s/g array
733  * @size:       Number of bytes to read
734  * @flags:      Message flags (MSG_DONTWAIT, etc...)
735  *
736  * On return the msg structure contains the scatter/gather array passed in the
737  * vec argument. The array is modified so that it consists of the unfilled
738  * portion of the original array.
739  *
740  * The returned value is the total number of bytes received, or an error.
741  */
742 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
743                    struct kvec *vec, size_t num, size_t size, int flags)
744 {
745         mm_segment_t oldfs = get_fs();
746         int result;
747 
748         iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
749         set_fs(KERNEL_DS);
750         result = sock_recvmsg(sock, msg, size, flags);
751         set_fs(oldfs);
752         return result;
753 }
754 EXPORT_SYMBOL(kernel_recvmsg);
755 
756 static ssize_t sock_sendpage(struct file *file, struct page *page,
757                              int offset, size_t size, loff_t *ppos, int more)
758 {
759         struct socket *sock;
760         int flags;
761 
762         sock = file->private_data;
763 
764         flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
765         /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
766         flags |= more;
767 
768         return kernel_sendpage(sock, page, offset, size, flags);
769 }
770 
771 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
772                                 struct pipe_inode_info *pipe, size_t len,
773                                 unsigned int flags)
774 {
775         struct socket *sock = file->private_data;
776 
777         if (unlikely(!sock->ops->splice_read))
778                 return -EINVAL;
779 
780         return sock->ops->splice_read(sock, ppos, pipe, len, flags);
781 }
782 
783 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
784 {
785         struct file *file = iocb->ki_filp;
786         struct socket *sock = file->private_data;
787         struct msghdr msg = {.msg_iter = *to,
788                              .msg_iocb = iocb};
789         ssize_t res;
790 
791         if (file->f_flags & O_NONBLOCK)
792                 msg.msg_flags = MSG_DONTWAIT;
793 
794         if (iocb->ki_pos != 0)
795                 return -ESPIPE;
796 
797         if (!iov_iter_count(to))        /* Match SYS5 behaviour */
798                 return 0;
799 
800         res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
801         *to = msg.msg_iter;
802         return res;
803 }
804 
805 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
806 {
807         struct file *file = iocb->ki_filp;
808         struct socket *sock = file->private_data;
809         struct msghdr msg = {.msg_iter = *from,
810                              .msg_iocb = iocb};
811         ssize_t res;
812 
813         if (iocb->ki_pos != 0)
814                 return -ESPIPE;
815 
816         if (file->f_flags & O_NONBLOCK)
817                 msg.msg_flags = MSG_DONTWAIT;
818 
819         if (sock->type == SOCK_SEQPACKET)
820                 msg.msg_flags |= MSG_EOR;
821 
822         res = sock_sendmsg(sock, &msg);
823         *from = msg.msg_iter;
824         return res;
825 }
826 
827 /*
828  * Atomic setting of ioctl hooks to avoid race
829  * with module unload.
830  */
831 
832 static DEFINE_MUTEX(br_ioctl_mutex);
833 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
834 
835 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
836 {
837         mutex_lock(&br_ioctl_mutex);
838         br_ioctl_hook = hook;
839         mutex_unlock(&br_ioctl_mutex);
840 }
841 EXPORT_SYMBOL(brioctl_set);
842 
843 static DEFINE_MUTEX(vlan_ioctl_mutex);
844 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
845 
846 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
847 {
848         mutex_lock(&vlan_ioctl_mutex);
849         vlan_ioctl_hook = hook;
850         mutex_unlock(&vlan_ioctl_mutex);
851 }
852 EXPORT_SYMBOL(vlan_ioctl_set);
853 
854 static DEFINE_MUTEX(dlci_ioctl_mutex);
855 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
856 
857 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
858 {
859         mutex_lock(&dlci_ioctl_mutex);
860         dlci_ioctl_hook = hook;
861         mutex_unlock(&dlci_ioctl_mutex);
862 }
863 EXPORT_SYMBOL(dlci_ioctl_set);
864 
865 static long sock_do_ioctl(struct net *net, struct socket *sock,
866                                  unsigned int cmd, unsigned long arg)
867 {
868         int err;
869         void __user *argp = (void __user *)arg;
870 
871         err = sock->ops->ioctl(sock, cmd, arg);
872 
873         /*
874          * If this ioctl is unknown try to hand it down
875          * to the NIC driver.
876          */
877         if (err == -ENOIOCTLCMD)
878                 err = dev_ioctl(net, cmd, argp);
879 
880         return err;
881 }
882 
883 /*
884  *      With an ioctl, arg may well be a user mode pointer, but we don't know
885  *      what to do with it - that's up to the protocol still.
886  */
887 
888 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
889 {
890         struct socket *sock;
891         struct sock *sk;
892         void __user *argp = (void __user *)arg;
893         int pid, err;
894         struct net *net;
895 
896         sock = file->private_data;
897         sk = sock->sk;
898         net = sock_net(sk);
899         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
900                 err = dev_ioctl(net, cmd, argp);
901         } else
902 #ifdef CONFIG_WEXT_CORE
903         if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
904                 err = dev_ioctl(net, cmd, argp);
905         } else
906 #endif
907                 switch (cmd) {
908                 case FIOSETOWN:
909                 case SIOCSPGRP:
910                         err = -EFAULT;
911                         if (get_user(pid, (int __user *)argp))
912                                 break;
913                         f_setown(sock->file, pid, 1);
914                         err = 0;
915                         break;
916                 case FIOGETOWN:
917                 case SIOCGPGRP:
918                         err = put_user(f_getown(sock->file),
919                                        (int __user *)argp);
920                         break;
921                 case SIOCGIFBR:
922                 case SIOCSIFBR:
923                 case SIOCBRADDBR:
924                 case SIOCBRDELBR:
925                         err = -ENOPKG;
926                         if (!br_ioctl_hook)
927                                 request_module("bridge");
928 
929                         mutex_lock(&br_ioctl_mutex);
930                         if (br_ioctl_hook)
931                                 err = br_ioctl_hook(net, cmd, argp);
932                         mutex_unlock(&br_ioctl_mutex);
933                         break;
934                 case SIOCGIFVLAN:
935                 case SIOCSIFVLAN:
936                         err = -ENOPKG;
937                         if (!vlan_ioctl_hook)
938                                 request_module("8021q");
939 
940                         mutex_lock(&vlan_ioctl_mutex);
941                         if (vlan_ioctl_hook)
942                                 err = vlan_ioctl_hook(net, argp);
943                         mutex_unlock(&vlan_ioctl_mutex);
944                         break;
945                 case SIOCADDDLCI:
946                 case SIOCDELDLCI:
947                         err = -ENOPKG;
948                         if (!dlci_ioctl_hook)
949                                 request_module("dlci");
950 
951                         mutex_lock(&dlci_ioctl_mutex);
952                         if (dlci_ioctl_hook)
953                                 err = dlci_ioctl_hook(cmd, argp);
954                         mutex_unlock(&dlci_ioctl_mutex);
955                         break;
956                 default:
957                         err = sock_do_ioctl(net, sock, cmd, arg);
958                         break;
959                 }
960         return err;
961 }
962 
963 int sock_create_lite(int family, int type, int protocol, struct socket **res)
964 {
965         int err;
966         struct socket *sock = NULL;
967 
968         err = security_socket_create(family, type, protocol, 1);
969         if (err)
970                 goto out;
971 
972         sock = sock_alloc();
973         if (!sock) {
974                 err = -ENOMEM;
975                 goto out;
976         }
977 
978         sock->type = type;
979         err = security_socket_post_create(sock, family, type, protocol, 1);
980         if (err)
981                 goto out_release;
982 
983 out:
984         *res = sock;
985         return err;
986 out_release:
987         sock_release(sock);
988         sock = NULL;
989         goto out;
990 }
991 EXPORT_SYMBOL(sock_create_lite);
992 
993 /* No kernel lock held - perfect */
994 static unsigned int sock_poll(struct file *file, poll_table *wait)
995 {
996         unsigned int busy_flag = 0;
997         struct socket *sock;
998 
999         /*
1000          *      We can't return errors to poll, so it's either yes or no.
1001          */
1002         sock = file->private_data;
1003 
1004         if (sk_can_busy_loop(sock->sk)) {
1005                 /* this socket can poll_ll so tell the system call */
1006                 busy_flag = POLL_BUSY_LOOP;
1007 
1008                 /* once, only if requested by syscall */
1009                 if (wait && (wait->_key & POLL_BUSY_LOOP))
1010                         sk_busy_loop(sock->sk, 1);
1011         }
1012 
1013         return busy_flag | sock->ops->poll(file, sock, wait);
1014 }
1015 
1016 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1017 {
1018         struct socket *sock = file->private_data;
1019 
1020         return sock->ops->mmap(file, sock, vma);
1021 }
1022 
1023 static int sock_close(struct inode *inode, struct file *filp)
1024 {
1025         sock_release(SOCKET_I(inode));
1026         return 0;
1027 }
1028 
1029 /*
1030  *      Update the socket async list
1031  *
1032  *      Fasync_list locking strategy.
1033  *
1034  *      1. fasync_list is modified only under process context socket lock
1035  *         i.e. under semaphore.
1036  *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1037  *         or under socket lock
1038  */
1039 
1040 static int sock_fasync(int fd, struct file *filp, int on)
1041 {
1042         struct socket *sock = filp->private_data;
1043         struct sock *sk = sock->sk;
1044         struct socket_wq *wq;
1045 
1046         if (sk == NULL)
1047                 return -EINVAL;
1048 
1049         lock_sock(sk);
1050         wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1051         fasync_helper(fd, filp, on, &wq->fasync_list);
1052 
1053         if (!wq->fasync_list)
1054                 sock_reset_flag(sk, SOCK_FASYNC);
1055         else
1056                 sock_set_flag(sk, SOCK_FASYNC);
1057 
1058         release_sock(sk);
1059         return 0;
1060 }
1061 
1062 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1063 
1064 int sock_wake_async(struct socket *sock, int how, int band)
1065 {
1066         struct socket_wq *wq;
1067 
1068         if (!sock)
1069                 return -1;
1070         rcu_read_lock();
1071         wq = rcu_dereference(sock->wq);
1072         if (!wq || !wq->fasync_list) {
1073                 rcu_read_unlock();
1074                 return -1;
1075         }
1076         switch (how) {
1077         case SOCK_WAKE_WAITD:
1078                 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1079                         break;
1080                 goto call_kill;
1081         case SOCK_WAKE_SPACE:
1082                 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1083                         break;
1084                 /* fall through */
1085         case SOCK_WAKE_IO:
1086 call_kill:
1087                 kill_fasync(&wq->fasync_list, SIGIO, band);
1088                 break;
1089         case SOCK_WAKE_URG:
1090                 kill_fasync(&wq->fasync_list, SIGURG, band);
1091         }
1092         rcu_read_unlock();
1093         return 0;
1094 }
1095 EXPORT_SYMBOL(sock_wake_async);
1096 
1097 int __sock_create(struct net *net, int family, int type, int protocol,
1098                          struct socket **res, int kern)
1099 {
1100         int err;
1101         struct socket *sock;
1102         const struct net_proto_family *pf;
1103 
1104         /*
1105          *      Check protocol is in range
1106          */
1107         if (family < 0 || family >= NPROTO)
1108                 return -EAFNOSUPPORT;
1109         if (type < 0 || type >= SOCK_MAX)
1110                 return -EINVAL;
1111 
1112         /* Compatibility.
1113 
1114            This uglymoron is moved from INET layer to here to avoid
1115            deadlock in module load.
1116          */
1117         if (family == PF_INET && type == SOCK_PACKET) {
1118                 static int warned;
1119                 if (!warned) {
1120                         warned = 1;
1121                         pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1122                                 current->comm);
1123                 }
1124                 family = PF_PACKET;
1125         }
1126 
1127         err = security_socket_create(family, type, protocol, kern);
1128         if (err)
1129                 return err;
1130 
1131         /*
1132          *      Allocate the socket and allow the family to set things up. if
1133          *      the protocol is 0, the family is instructed to select an appropriate
1134          *      default.
1135          */
1136         sock = sock_alloc();
1137         if (!sock) {
1138                 net_warn_ratelimited("socket: no more sockets\n");
1139                 return -ENFILE; /* Not exactly a match, but its the
1140                                    closest posix thing */
1141         }
1142 
1143         sock->type = type;
1144 
1145 #ifdef CONFIG_MODULES
1146         /* Attempt to load a protocol module if the find failed.
1147          *
1148          * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1149          * requested real, full-featured networking support upon configuration.
1150          * Otherwise module support will break!
1151          */
1152         if (rcu_access_pointer(net_families[family]) == NULL)
1153                 request_module("net-pf-%d", family);
1154 #endif
1155 
1156         rcu_read_lock();
1157         pf = rcu_dereference(net_families[family]);
1158         err = -EAFNOSUPPORT;
1159         if (!pf)
1160                 goto out_release;
1161 
1162         /*
1163          * We will call the ->create function, that possibly is in a loadable
1164          * module, so we have to bump that loadable module refcnt first.
1165          */
1166         if (!try_module_get(pf->owner))
1167                 goto out_release;
1168 
1169         /* Now protected by module ref count */
1170         rcu_read_unlock();
1171 
1172         err = pf->create(net, sock, protocol, kern);
1173         if (err < 0)
1174                 goto out_module_put;
1175 
1176         /*
1177          * Now to bump the refcnt of the [loadable] module that owns this
1178          * socket at sock_release time we decrement its refcnt.
1179          */
1180         if (!try_module_get(sock->ops->owner))
1181                 goto out_module_busy;
1182 
1183         /*
1184          * Now that we're done with the ->create function, the [loadable]
1185          * module can have its refcnt decremented
1186          */
1187         module_put(pf->owner);
1188         err = security_socket_post_create(sock, family, type, protocol, kern);
1189         if (err)
1190                 goto out_sock_release;
1191         *res = sock;
1192 
1193         return 0;
1194 
1195 out_module_busy:
1196         err = -EAFNOSUPPORT;
1197 out_module_put:
1198         sock->ops = NULL;
1199         module_put(pf->owner);
1200 out_sock_release:
1201         sock_release(sock);
1202         return err;
1203 
1204 out_release:
1205         rcu_read_unlock();
1206         goto out_sock_release;
1207 }
1208 EXPORT_SYMBOL(__sock_create);
1209 
1210 int sock_create(int family, int type, int protocol, struct socket **res)
1211 {
1212         return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1213 }
1214 EXPORT_SYMBOL(sock_create);
1215 
1216 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1217 {
1218         return __sock_create(&init_net, family, type, protocol, res, 1);
1219 }
1220 EXPORT_SYMBOL(sock_create_kern);
1221 
1222 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1223 {
1224         int retval;
1225         struct socket *sock;
1226         int flags;
1227 
1228         /* Check the SOCK_* constants for consistency.  */
1229         BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1230         BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1231         BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1232         BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1233 
1234         flags = type & ~SOCK_TYPE_MASK;
1235         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1236                 return -EINVAL;
1237         type &= SOCK_TYPE_MASK;
1238 
1239         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1240                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1241 
1242         retval = sock_create(family, type, protocol, &sock);
1243         if (retval < 0)
1244                 goto out;
1245 
1246         retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1247         if (retval < 0)
1248                 goto out_release;
1249 
1250 out:
1251         /* It may be already another descriptor 8) Not kernel problem. */
1252         return retval;
1253 
1254 out_release:
1255         sock_release(sock);
1256         return retval;
1257 }
1258 
1259 /*
1260  *      Create a pair of connected sockets.
1261  */
1262 
1263 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1264                 int __user *, usockvec)
1265 {
1266         struct socket *sock1, *sock2;
1267         int fd1, fd2, err;
1268         struct file *newfile1, *newfile2;
1269         int flags;
1270 
1271         flags = type & ~SOCK_TYPE_MASK;
1272         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1273                 return -EINVAL;
1274         type &= SOCK_TYPE_MASK;
1275 
1276         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1277                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1278 
1279         /*
1280          * Obtain the first socket and check if the underlying protocol
1281          * supports the socketpair call.
1282          */
1283 
1284         err = sock_create(family, type, protocol, &sock1);
1285         if (err < 0)
1286                 goto out;
1287 
1288         err = sock_create(family, type, protocol, &sock2);
1289         if (err < 0)
1290                 goto out_release_1;
1291 
1292         err = sock1->ops->socketpair(sock1, sock2);
1293         if (err < 0)
1294                 goto out_release_both;
1295 
1296         fd1 = get_unused_fd_flags(flags);
1297         if (unlikely(fd1 < 0)) {
1298                 err = fd1;
1299                 goto out_release_both;
1300         }
1301 
1302         fd2 = get_unused_fd_flags(flags);
1303         if (unlikely(fd2 < 0)) {
1304                 err = fd2;
1305                 goto out_put_unused_1;
1306         }
1307 
1308         newfile1 = sock_alloc_file(sock1, flags, NULL);
1309         if (unlikely(IS_ERR(newfile1))) {
1310                 err = PTR_ERR(newfile1);
1311                 goto out_put_unused_both;
1312         }
1313 
1314         newfile2 = sock_alloc_file(sock2, flags, NULL);
1315         if (IS_ERR(newfile2)) {
1316                 err = PTR_ERR(newfile2);
1317                 goto out_fput_1;
1318         }
1319 
1320         err = put_user(fd1, &usockvec[0]);
1321         if (err)
1322                 goto out_fput_both;
1323 
1324         err = put_user(fd2, &usockvec[1]);
1325         if (err)
1326                 goto out_fput_both;
1327 
1328         audit_fd_pair(fd1, fd2);
1329 
1330         fd_install(fd1, newfile1);
1331         fd_install(fd2, newfile2);
1332         /* fd1 and fd2 may be already another descriptors.
1333          * Not kernel problem.
1334          */
1335 
1336         return 0;
1337 
1338 out_fput_both:
1339         fput(newfile2);
1340         fput(newfile1);
1341         put_unused_fd(fd2);
1342         put_unused_fd(fd1);
1343         goto out;
1344 
1345 out_fput_1:
1346         fput(newfile1);
1347         put_unused_fd(fd2);
1348         put_unused_fd(fd1);
1349         sock_release(sock2);
1350         goto out;
1351 
1352 out_put_unused_both:
1353         put_unused_fd(fd2);
1354 out_put_unused_1:
1355         put_unused_fd(fd1);
1356 out_release_both:
1357         sock_release(sock2);
1358 out_release_1:
1359         sock_release(sock1);
1360 out:
1361         return err;
1362 }
1363 
1364 /*
1365  *      Bind a name to a socket. Nothing much to do here since it's
1366  *      the protocol's responsibility to handle the local address.
1367  *
1368  *      We move the socket address to kernel space before we call
1369  *      the protocol layer (having also checked the address is ok).
1370  */
1371 
1372 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1373 {
1374         struct socket *sock;
1375         struct sockaddr_storage address;
1376         int err, fput_needed;
1377 
1378         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1379         if (sock) {
1380                 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1381                 if (err >= 0) {
1382                         err = security_socket_bind(sock,
1383                                                    (struct sockaddr *)&address,
1384                                                    addrlen);
1385                         if (!err)
1386                                 err = sock->ops->bind(sock,
1387                                                       (struct sockaddr *)
1388                                                       &address, addrlen);
1389                 }
1390                 fput_light(sock->file, fput_needed);
1391         }
1392         return err;
1393 }
1394 
1395 /*
1396  *      Perform a listen. Basically, we allow the protocol to do anything
1397  *      necessary for a listen, and if that works, we mark the socket as
1398  *      ready for listening.
1399  */
1400 
1401 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1402 {
1403         struct socket *sock;
1404         int err, fput_needed;
1405         int somaxconn;
1406 
1407         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1408         if (sock) {
1409                 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1410                 if ((unsigned int)backlog > somaxconn)
1411                         backlog = somaxconn;
1412 
1413                 err = security_socket_listen(sock, backlog);
1414                 if (!err)
1415                         err = sock->ops->listen(sock, backlog);
1416 
1417                 fput_light(sock->file, fput_needed);
1418         }
1419         return err;
1420 }
1421 
1422 /*
1423  *      For accept, we attempt to create a new socket, set up the link
1424  *      with the client, wake up the client, then return the new
1425  *      connected fd. We collect the address of the connector in kernel
1426  *      space and move it to user at the very end. This is unclean because
1427  *      we open the socket then return an error.
1428  *
1429  *      1003.1g adds the ability to recvmsg() to query connection pending
1430  *      status to recvmsg. We need to add that support in a way thats
1431  *      clean when we restucture accept also.
1432  */
1433 
1434 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1435                 int __user *, upeer_addrlen, int, flags)
1436 {
1437         struct socket *sock, *newsock;
1438         struct file *newfile;
1439         int err, len, newfd, fput_needed;
1440         struct sockaddr_storage address;
1441 
1442         if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1443                 return -EINVAL;
1444 
1445         if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1446                 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1447 
1448         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1449         if (!sock)
1450                 goto out;
1451 
1452         err = -ENFILE;
1453         newsock = sock_alloc();
1454         if (!newsock)
1455                 goto out_put;
1456 
1457         newsock->type = sock->type;
1458         newsock->ops = sock->ops;
1459 
1460         /*
1461          * We don't need try_module_get here, as the listening socket (sock)
1462          * has the protocol module (sock->ops->owner) held.
1463          */
1464         __module_get(newsock->ops->owner);
1465 
1466         newfd = get_unused_fd_flags(flags);
1467         if (unlikely(newfd < 0)) {
1468                 err = newfd;
1469                 sock_release(newsock);
1470                 goto out_put;
1471         }
1472         newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1473         if (unlikely(IS_ERR(newfile))) {
1474                 err = PTR_ERR(newfile);
1475                 put_unused_fd(newfd);
1476                 sock_release(newsock);
1477                 goto out_put;
1478         }
1479 
1480         err = security_socket_accept(sock, newsock);
1481         if (err)
1482                 goto out_fd;
1483 
1484         err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1485         if (err < 0)
1486                 goto out_fd;
1487 
1488         if (upeer_sockaddr) {
1489                 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1490                                           &len, 2) < 0) {
1491                         err = -ECONNABORTED;
1492                         goto out_fd;
1493                 }
1494                 err = move_addr_to_user(&address,
1495                                         len, upeer_sockaddr, upeer_addrlen);
1496                 if (err < 0)
1497                         goto out_fd;
1498         }
1499 
1500         /* File flags are not inherited via accept() unlike another OSes. */
1501 
1502         fd_install(newfd, newfile);
1503         err = newfd;
1504 
1505 out_put:
1506         fput_light(sock->file, fput_needed);
1507 out:
1508         return err;
1509 out_fd:
1510         fput(newfile);
1511         put_unused_fd(newfd);
1512         goto out_put;
1513 }
1514 
1515 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1516                 int __user *, upeer_addrlen)
1517 {
1518         return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1519 }
1520 
1521 /*
1522  *      Attempt to connect to a socket with the server address.  The address
1523  *      is in user space so we verify it is OK and move it to kernel space.
1524  *
1525  *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1526  *      break bindings
1527  *
1528  *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1529  *      other SEQPACKET protocols that take time to connect() as it doesn't
1530  *      include the -EINPROGRESS status for such sockets.
1531  */
1532 
1533 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1534                 int, addrlen)
1535 {
1536         struct socket *sock;
1537         struct sockaddr_storage address;
1538         int err, fput_needed;
1539 
1540         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1541         if (!sock)
1542                 goto out;
1543         err = move_addr_to_kernel(uservaddr, addrlen, &address);
1544         if (err < 0)
1545                 goto out_put;
1546 
1547         err =
1548             security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1549         if (err)
1550                 goto out_put;
1551 
1552         err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1553                                  sock->file->f_flags);
1554 out_put:
1555         fput_light(sock->file, fput_needed);
1556 out:
1557         return err;
1558 }
1559 
1560 /*
1561  *      Get the local address ('name') of a socket object. Move the obtained
1562  *      name to user space.
1563  */
1564 
1565 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1566                 int __user *, usockaddr_len)
1567 {
1568         struct socket *sock;
1569         struct sockaddr_storage address;
1570         int len, err, fput_needed;
1571 
1572         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1573         if (!sock)
1574                 goto out;
1575 
1576         err = security_socket_getsockname(sock);
1577         if (err)
1578                 goto out_put;
1579 
1580         err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1581         if (err)
1582                 goto out_put;
1583         err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1584 
1585 out_put:
1586         fput_light(sock->file, fput_needed);
1587 out:
1588         return err;
1589 }
1590 
1591 /*
1592  *      Get the remote address ('name') of a socket object. Move the obtained
1593  *      name to user space.
1594  */
1595 
1596 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1597                 int __user *, usockaddr_len)
1598 {
1599         struct socket *sock;
1600         struct sockaddr_storage address;
1601         int len, err, fput_needed;
1602 
1603         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1604         if (sock != NULL) {
1605                 err = security_socket_getpeername(sock);
1606                 if (err) {
1607                         fput_light(sock->file, fput_needed);
1608                         return err;
1609                 }
1610 
1611                 err =
1612                     sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1613                                        1);
1614                 if (!err)
1615                         err = move_addr_to_user(&address, len, usockaddr,
1616                                                 usockaddr_len);
1617                 fput_light(sock->file, fput_needed);
1618         }
1619         return err;
1620 }
1621 
1622 /*
1623  *      Send a datagram to a given address. We move the address into kernel
1624  *      space and check the user space data area is readable before invoking
1625  *      the protocol.
1626  */
1627 
1628 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1629                 unsigned int, flags, struct sockaddr __user *, addr,
1630                 int, addr_len)
1631 {
1632         struct socket *sock;
1633         struct sockaddr_storage address;
1634         int err;
1635         struct msghdr msg;
1636         struct iovec iov;
1637         int fput_needed;
1638 
1639         err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1640         if (unlikely(err))
1641                 return err;
1642         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1643         if (!sock)
1644                 goto out;
1645 
1646         msg.msg_name = NULL;
1647         msg.msg_control = NULL;
1648         msg.msg_controllen = 0;
1649         msg.msg_namelen = 0;
1650         if (addr) {
1651                 err = move_addr_to_kernel(addr, addr_len, &address);
1652                 if (err < 0)
1653                         goto out_put;
1654                 msg.msg_name = (struct sockaddr *)&address;
1655                 msg.msg_namelen = addr_len;
1656         }
1657         if (sock->file->f_flags & O_NONBLOCK)
1658                 flags |= MSG_DONTWAIT;
1659         msg.msg_flags = flags;
1660         err = sock_sendmsg(sock, &msg);
1661 
1662 out_put:
1663         fput_light(sock->file, fput_needed);
1664 out:
1665         return err;
1666 }
1667 
1668 /*
1669  *      Send a datagram down a socket.
1670  */
1671 
1672 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1673                 unsigned int, flags)
1674 {
1675         return sys_sendto(fd, buff, len, flags, NULL, 0);
1676 }
1677 
1678 /*
1679  *      Receive a frame from the socket and optionally record the address of the
1680  *      sender. We verify the buffers are writable and if needed move the
1681  *      sender address from kernel to user space.
1682  */
1683 
1684 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1685                 unsigned int, flags, struct sockaddr __user *, addr,
1686                 int __user *, addr_len)
1687 {
1688         struct socket *sock;
1689         struct iovec iov;
1690         struct msghdr msg;
1691         struct sockaddr_storage address;
1692         int err, err2;
1693         int fput_needed;
1694 
1695         err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1696         if (unlikely(err))
1697                 return err;
1698         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1699         if (!sock)
1700                 goto out;
1701 
1702         msg.msg_control = NULL;
1703         msg.msg_controllen = 0;
1704         /* Save some cycles and don't copy the address if not needed */
1705         msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1706         /* We assume all kernel code knows the size of sockaddr_storage */
1707         msg.msg_namelen = 0;
1708         if (sock->file->f_flags & O_NONBLOCK)
1709                 flags |= MSG_DONTWAIT;
1710         err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1711 
1712         if (err >= 0 && addr != NULL) {
1713                 err2 = move_addr_to_user(&address,
1714                                          msg.msg_namelen, addr, addr_len);
1715                 if (err2 < 0)
1716                         err = err2;
1717         }
1718 
1719         fput_light(sock->file, fput_needed);
1720 out:
1721         return err;
1722 }
1723 
1724 /*
1725  *      Receive a datagram from a socket.
1726  */
1727 
1728 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1729                 unsigned int, flags)
1730 {
1731         return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1732 }
1733 
1734 /*
1735  *      Set a socket option. Because we don't know the option lengths we have
1736  *      to pass the user mode parameter for the protocols to sort out.
1737  */
1738 
1739 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1740                 char __user *, optval, int, optlen)
1741 {
1742         int err, fput_needed;
1743         struct socket *sock;
1744 
1745         if (optlen < 0)
1746                 return -EINVAL;
1747 
1748         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1749         if (sock != NULL) {
1750                 err = security_socket_setsockopt(sock, level, optname);
1751                 if (err)
1752                         goto out_put;
1753 
1754                 if (level == SOL_SOCKET)
1755                         err =
1756                             sock_setsockopt(sock, level, optname, optval,
1757                                             optlen);
1758                 else
1759                         err =
1760                             sock->ops->setsockopt(sock, level, optname, optval,
1761                                                   optlen);
1762 out_put:
1763                 fput_light(sock->file, fput_needed);
1764         }
1765         return err;
1766 }
1767 
1768 /*
1769  *      Get a socket option. Because we don't know the option lengths we have
1770  *      to pass a user mode parameter for the protocols to sort out.
1771  */
1772 
1773 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1774                 char __user *, optval, int __user *, optlen)
1775 {
1776         int err, fput_needed;
1777         struct socket *sock;
1778 
1779         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1780         if (sock != NULL) {
1781                 err = security_socket_getsockopt(sock, level, optname);
1782                 if (err)
1783                         goto out_put;
1784 
1785                 if (level == SOL_SOCKET)
1786                         err =
1787                             sock_getsockopt(sock, level, optname, optval,
1788                                             optlen);
1789                 else
1790                         err =
1791                             sock->ops->getsockopt(sock, level, optname, optval,
1792                                                   optlen);
1793 out_put:
1794                 fput_light(sock->file, fput_needed);
1795         }
1796         return err;
1797 }
1798 
1799 /*
1800  *      Shutdown a socket.
1801  */
1802 
1803 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1804 {
1805         int err, fput_needed;
1806         struct socket *sock;
1807 
1808         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1809         if (sock != NULL) {
1810                 err = security_socket_shutdown(sock, how);
1811                 if (!err)
1812                         err = sock->ops->shutdown(sock, how);
1813                 fput_light(sock->file, fput_needed);
1814         }
1815         return err;
1816 }
1817 
1818 /* A couple of helpful macros for getting the address of the 32/64 bit
1819  * fields which are the same type (int / unsigned) on our platforms.
1820  */
1821 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1822 #define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1823 #define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1824 
1825 struct used_address {
1826         struct sockaddr_storage name;
1827         unsigned int name_len;
1828 };
1829 
1830 static int copy_msghdr_from_user(struct msghdr *kmsg,
1831                                  struct user_msghdr __user *umsg,
1832                                  struct sockaddr __user **save_addr,
1833                                  struct iovec **iov)
1834 {
1835         struct sockaddr __user *uaddr;
1836         struct iovec __user *uiov;
1837         size_t nr_segs;
1838         ssize_t err;
1839 
1840         if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1841             __get_user(uaddr, &umsg->msg_name) ||
1842             __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1843             __get_user(uiov, &umsg->msg_iov) ||
1844             __get_user(nr_segs, &umsg->msg_iovlen) ||
1845             __get_user(kmsg->msg_control, &umsg->msg_control) ||
1846             __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1847             __get_user(kmsg->msg_flags, &umsg->msg_flags))
1848                 return -EFAULT;
1849 
1850         if (!uaddr)
1851                 kmsg->msg_namelen = 0;
1852 
1853         if (kmsg->msg_namelen < 0)
1854                 return -EINVAL;
1855 
1856         if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1857                 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1858 
1859         if (save_addr)
1860                 *save_addr = uaddr;
1861 
1862         if (uaddr && kmsg->msg_namelen) {
1863                 if (!save_addr) {
1864                         err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1865                                                   kmsg->msg_name);
1866                         if (err < 0)
1867                                 return err;
1868                 }
1869         } else {
1870                 kmsg->msg_name = NULL;
1871                 kmsg->msg_namelen = 0;
1872         }
1873 
1874         if (nr_segs > UIO_MAXIOV)
1875                 return -EMSGSIZE;
1876 
1877         kmsg->msg_iocb = NULL;
1878 
1879         return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1880                             UIO_FASTIOV, iov, &kmsg->msg_iter);
1881 }
1882 
1883 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1884                          struct msghdr *msg_sys, unsigned int flags,
1885                          struct used_address *used_address)
1886 {
1887         struct compat_msghdr __user *msg_compat =
1888             (struct compat_msghdr __user *)msg;
1889         struct sockaddr_storage address;
1890         struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1891         unsigned char ctl[sizeof(struct cmsghdr) + 20]
1892             __attribute__ ((aligned(sizeof(__kernel_size_t))));
1893         /* 20 is size of ipv6_pktinfo */
1894         unsigned char *ctl_buf = ctl;
1895         int ctl_len;
1896         ssize_t err;
1897 
1898         msg_sys->msg_name = &address;
1899 
1900         if (MSG_CMSG_COMPAT & flags)
1901                 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1902         else
1903                 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1904         if (err < 0)
1905                 return err;
1906 
1907         err = -ENOBUFS;
1908 
1909         if (msg_sys->msg_controllen > INT_MAX)
1910                 goto out_freeiov;
1911         ctl_len = msg_sys->msg_controllen;
1912         if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1913                 err =
1914                     cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1915                                                      sizeof(ctl));
1916                 if (err)
1917                         goto out_freeiov;
1918                 ctl_buf = msg_sys->msg_control;
1919                 ctl_len = msg_sys->msg_controllen;
1920         } else if (ctl_len) {
1921                 if (ctl_len > sizeof(ctl)) {
1922                         ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1923                         if (ctl_buf == NULL)
1924                                 goto out_freeiov;
1925                 }
1926                 err = -EFAULT;
1927                 /*
1928                  * Careful! Before this, msg_sys->msg_control contains a user pointer.
1929                  * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1930                  * checking falls down on this.
1931                  */
1932                 if (copy_from_user(ctl_buf,
1933                                    (void __user __force *)msg_sys->msg_control,
1934                                    ctl_len))
1935                         goto out_freectl;
1936                 msg_sys->msg_control = ctl_buf;
1937         }
1938         msg_sys->msg_flags = flags;
1939 
1940         if (sock->file->f_flags & O_NONBLOCK)
1941                 msg_sys->msg_flags |= MSG_DONTWAIT;
1942         /*
1943          * If this is sendmmsg() and current destination address is same as
1944          * previously succeeded address, omit asking LSM's decision.
1945          * used_address->name_len is initialized to UINT_MAX so that the first
1946          * destination address never matches.
1947          */
1948         if (used_address && msg_sys->msg_name &&
1949             used_address->name_len == msg_sys->msg_namelen &&
1950             !memcmp(&used_address->name, msg_sys->msg_name,
1951                     used_address->name_len)) {
1952                 err = sock_sendmsg_nosec(sock, msg_sys);
1953                 goto out_freectl;
1954         }
1955         err = sock_sendmsg(sock, msg_sys);
1956         /*
1957          * If this is sendmmsg() and sending to current destination address was
1958          * successful, remember it.
1959          */
1960         if (used_address && err >= 0) {
1961                 used_address->name_len = msg_sys->msg_namelen;
1962                 if (msg_sys->msg_name)
1963                         memcpy(&used_address->name, msg_sys->msg_name,
1964                                used_address->name_len);
1965         }
1966 
1967 out_freectl:
1968         if (ctl_buf != ctl)
1969                 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1970 out_freeiov:
1971         kfree(iov);
1972         return err;
1973 }
1974 
1975 /*
1976  *      BSD sendmsg interface
1977  */
1978 
1979 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1980 {
1981         int fput_needed, err;
1982         struct msghdr msg_sys;
1983         struct socket *sock;
1984 
1985         sock = sockfd_lookup_light(fd, &err, &fput_needed);
1986         if (!sock)
1987                 goto out;
1988 
1989         err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
1990 
1991         fput_light(sock->file, fput_needed);
1992 out:
1993         return err;
1994 }
1995 
1996 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1997 {
1998         if (flags & MSG_CMSG_COMPAT)
1999                 return -EINVAL;
2000         return __sys_sendmsg(fd, msg, flags);
2001 }
2002 
2003 /*
2004  *      Linux sendmmsg interface
2005  */
2006 
2007 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2008                    unsigned int flags)
2009 {
2010         int fput_needed, err, datagrams;
2011         struct socket *sock;
2012         struct mmsghdr __user *entry;
2013         struct compat_mmsghdr __user *compat_entry;
2014         struct msghdr msg_sys;
2015         struct used_address used_address;
2016 
2017         if (vlen > UIO_MAXIOV)
2018                 vlen = UIO_MAXIOV;
2019 
2020         datagrams = 0;
2021 
2022         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2023         if (!sock)
2024                 return err;
2025 
2026         used_address.name_len = UINT_MAX;
2027         entry = mmsg;
2028         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2029         err = 0;
2030 
2031         while (datagrams < vlen) {
2032                 if (MSG_CMSG_COMPAT & flags) {
2033                         err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2034                                              &msg_sys, flags, &used_address);
2035                         if (err < 0)
2036                                 break;
2037                         err = __put_user(err, &compat_entry->msg_len);
2038                         ++compat_entry;
2039                 } else {
2040                         err = ___sys_sendmsg(sock,
2041                                              (struct user_msghdr __user *)entry,
2042                                              &msg_sys, flags, &used_address);
2043                         if (err < 0)
2044                                 break;
2045                         err = put_user(err, &entry->msg_len);
2046                         ++entry;
2047                 }
2048 
2049                 if (err)
2050                         break;
2051                 ++datagrams;
2052         }
2053 
2054         fput_light(sock->file, fput_needed);
2055 
2056         /* We only return an error if no datagrams were able to be sent */
2057         if (datagrams != 0)
2058                 return datagrams;
2059 
2060         return err;
2061 }
2062 
2063 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2064                 unsigned int, vlen, unsigned int, flags)
2065 {
2066         if (flags & MSG_CMSG_COMPAT)
2067                 return -EINVAL;
2068         return __sys_sendmmsg(fd, mmsg, vlen, flags);
2069 }
2070 
2071 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2072                          struct msghdr *msg_sys, unsigned int flags, int nosec)
2073 {
2074         struct compat_msghdr __user *msg_compat =
2075             (struct compat_msghdr __user *)msg;
2076         struct iovec iovstack[UIO_FASTIOV];
2077         struct iovec *iov = iovstack;
2078         unsigned long cmsg_ptr;
2079         int total_len, len;
2080         ssize_t err;
2081 
2082         /* kernel mode address */
2083         struct sockaddr_storage addr;
2084 
2085         /* user mode address pointers */
2086         struct sockaddr __user *uaddr;
2087         int __user *uaddr_len = COMPAT_NAMELEN(msg);
2088 
2089         msg_sys->msg_name = &addr;
2090 
2091         if (MSG_CMSG_COMPAT & flags)
2092                 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2093         else
2094                 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2095         if (err < 0)
2096                 return err;
2097         total_len = iov_iter_count(&msg_sys->msg_iter);
2098 
2099         cmsg_ptr = (unsigned long)msg_sys->msg_control;
2100         msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2101 
2102         /* We assume all kernel code knows the size of sockaddr_storage */
2103         msg_sys->msg_namelen = 0;
2104 
2105         if (sock->file->f_flags & O_NONBLOCK)
2106                 flags |= MSG_DONTWAIT;
2107         err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2108                                                           total_len, flags);
2109         if (err < 0)
2110                 goto out_freeiov;
2111         len = err;
2112 
2113         if (uaddr != NULL) {
2114                 err = move_addr_to_user(&addr,
2115                                         msg_sys->msg_namelen, uaddr,
2116                                         uaddr_len);
2117                 if (err < 0)
2118                         goto out_freeiov;
2119         }
2120         err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2121                          COMPAT_FLAGS(msg));
2122         if (err)
2123                 goto out_freeiov;
2124         if (MSG_CMSG_COMPAT & flags)
2125                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2126                                  &msg_compat->msg_controllen);
2127         else
2128                 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2129                                  &msg->msg_controllen);
2130         if (err)
2131                 goto out_freeiov;
2132         err = len;
2133 
2134 out_freeiov:
2135         kfree(iov);
2136         return err;
2137 }
2138 
2139 /*
2140  *      BSD recvmsg interface
2141  */
2142 
2143 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2144 {
2145         int fput_needed, err;
2146         struct msghdr msg_sys;
2147         struct socket *sock;
2148 
2149         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2150         if (!sock)
2151                 goto out;
2152 
2153         err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2154 
2155         fput_light(sock->file, fput_needed);
2156 out:
2157         return err;
2158 }
2159 
2160 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2161                 unsigned int, flags)
2162 {
2163         if (flags & MSG_CMSG_COMPAT)
2164                 return -EINVAL;
2165         return __sys_recvmsg(fd, msg, flags);
2166 }
2167 
2168 /*
2169  *     Linux recvmmsg interface
2170  */
2171 
2172 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2173                    unsigned int flags, struct timespec *timeout)
2174 {
2175         int fput_needed, err, datagrams;
2176         struct socket *sock;
2177         struct mmsghdr __user *entry;
2178         struct compat_mmsghdr __user *compat_entry;
2179         struct msghdr msg_sys;
2180         struct timespec end_time;
2181 
2182         if (timeout &&
2183             poll_select_set_timeout(&end_time, timeout->tv_sec,
2184                                     timeout->tv_nsec))
2185                 return -EINVAL;
2186 
2187         datagrams = 0;
2188 
2189         sock = sockfd_lookup_light(fd, &err, &fput_needed);
2190         if (!sock)
2191                 return err;
2192 
2193         err = sock_error(sock->sk);
2194         if (err)
2195                 goto out_put;
2196 
2197         entry = mmsg;
2198         compat_entry = (struct compat_mmsghdr __user *)mmsg;
2199 
2200         while (datagrams < vlen) {
2201                 /*
2202                  * No need to ask LSM for more than the first datagram.
2203                  */
2204                 if (MSG_CMSG_COMPAT & flags) {
2205                         err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2206                                              &msg_sys, flags & ~MSG_WAITFORONE,
2207                                              datagrams);
2208                         if (err < 0)
2209                                 break;
2210                         err = __put_user(err, &compat_entry->msg_len);
2211                         ++compat_entry;
2212                 } else {
2213                         err = ___sys_recvmsg(sock,
2214                                              (struct user_msghdr __user *)entry,
2215                                              &msg_sys, flags & ~MSG_WAITFORONE,
2216                                              datagrams);
2217                         if (err < 0)
2218                                 break;
2219                         err = put_user(err, &entry->msg_len);
2220                         ++entry;
2221                 }
2222 
2223                 if (err)
2224                         break;
2225                 ++datagrams;
2226 
2227                 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2228                 if (flags & MSG_WAITFORONE)
2229                         flags |= MSG_DONTWAIT;
2230 
2231                 if (timeout) {
2232                         ktime_get_ts(timeout);
2233                         *timeout = timespec_sub(end_time, *timeout);
2234                         if (timeout->tv_sec < 0) {
2235                                 timeout->tv_sec = timeout->tv_nsec = 0;
2236                                 break;
2237                         }
2238 
2239                         /* Timeout, return less than vlen datagrams */
2240                         if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2241                                 break;
2242                 }
2243 
2244                 /* Out of band data, return right away */
2245                 if (msg_sys.msg_flags & MSG_OOB)
2246                         break;
2247         }
2248 
2249 out_put:
2250         fput_light(sock->file, fput_needed);
2251 
2252         if (err == 0)
2253                 return datagrams;
2254 
2255         if (datagrams != 0) {
2256                 /*
2257                  * We may return less entries than requested (vlen) if the
2258                  * sock is non block and there aren't enough datagrams...
2259                  */
2260                 if (err != -EAGAIN) {
2261                         /*
2262                          * ... or  if recvmsg returns an error after we
2263                          * received some datagrams, where we record the
2264                          * error to return on the next call or if the
2265                          * app asks about it using getsockopt(SO_ERROR).
2266                          */
2267                         sock->sk->sk_err = -err;
2268                 }
2269 
2270                 return datagrams;
2271         }
2272 
2273         return err;
2274 }
2275 
2276 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2277                 unsigned int, vlen, unsigned int, flags,
2278                 struct timespec __user *, timeout)
2279 {
2280         int datagrams;
2281         struct timespec timeout_sys;
2282 
2283         if (flags & MSG_CMSG_COMPAT)
2284                 return -EINVAL;
2285 
2286         if (!timeout)
2287                 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2288 
2289         if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2290                 return -EFAULT;
2291 
2292         datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2293 
2294         if (datagrams > 0 &&
2295             copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2296                 datagrams = -EFAULT;
2297 
2298         return datagrams;
2299 }
2300 
2301 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2302 /* Argument list sizes for sys_socketcall */
2303 #define AL(x) ((x) * sizeof(unsigned long))
2304 static const unsigned char nargs[21] = {
2305         AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2306         AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2307         AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2308         AL(4), AL(5), AL(4)
2309 };
2310 
2311 #undef AL
2312 
2313 /*
2314  *      System call vectors.
2315  *
2316  *      Argument checking cleaned up. Saved 20% in size.
2317  *  This function doesn't need to set the kernel lock because
2318  *  it is set by the callees.
2319  */
2320 
2321 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2322 {
2323         unsigned long a[AUDITSC_ARGS];
2324         unsigned long a0, a1;
2325         int err;
2326         unsigned int len;
2327 
2328         if (call < 1 || call > SYS_SENDMMSG)
2329                 return -EINVAL;
2330 
2331         len = nargs[call];
2332         if (len > sizeof(a))
2333                 return -EINVAL;
2334 
2335         /* copy_from_user should be SMP safe. */
2336         if (copy_from_user(a, args, len))
2337                 return -EFAULT;
2338 
2339         err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2340         if (err)
2341                 return err;
2342 
2343         a0 = a[0];
2344         a1 = a[1];
2345 
2346         switch (call) {
2347         case SYS_SOCKET:
2348                 err = sys_socket(a0, a1, a[2]);
2349                 break;
2350         case SYS_BIND:
2351                 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2352                 break;
2353         case SYS_CONNECT:
2354                 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2355                 break;
2356         case SYS_LISTEN:
2357                 err = sys_listen(a0, a1);
2358                 break;
2359         case SYS_ACCEPT:
2360                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2361                                   (int __user *)a[2], 0);
2362                 break;
2363         case SYS_GETSOCKNAME:
2364                 err =
2365                     sys_getsockname(a0, (struct sockaddr __user *)a1,
2366                                     (int __user *)a[2]);
2367                 break;
2368         case SYS_GETPEERNAME:
2369                 err =
2370                     sys_getpeername(a0, (struct sockaddr __user *)a1,
2371                                     (int __user *)a[2]);
2372                 break;
2373         case SYS_SOCKETPAIR:
2374                 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2375                 break;
2376         case SYS_SEND:
2377                 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2378                 break;
2379         case SYS_SENDTO:
2380                 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2381                                  (struct sockaddr __user *)a[4], a[5]);
2382                 break;
2383         case SYS_RECV:
2384                 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2385                 break;
2386         case SYS_RECVFROM:
2387                 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2388                                    (struct sockaddr __user *)a[4],
2389                                    (int __user *)a[5]);
2390                 break;
2391         case SYS_SHUTDOWN:
2392                 err = sys_shutdown(a0, a1);
2393                 break;
2394         case SYS_SETSOCKOPT:
2395                 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2396                 break;
2397         case SYS_GETSOCKOPT:
2398                 err =
2399                     sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2400                                    (int __user *)a[4]);
2401                 break;
2402         case SYS_SENDMSG:
2403                 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2404                 break;
2405         case SYS_SENDMMSG:
2406                 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2407                 break;
2408         case SYS_RECVMSG:
2409                 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2410                 break;
2411         case SYS_RECVMMSG:
2412                 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2413                                    (struct timespec __user *)a[4]);
2414                 break;
2415         case SYS_ACCEPT4:
2416                 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2417                                   (int __user *)a[2], a[3]);
2418                 break;
2419         default:
2420                 err = -EINVAL;
2421                 break;
2422         }
2423         return err;
2424 }
2425 
2426 #endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2427 
2428 /**
2429  *      sock_register - add a socket protocol handler
2430  *      @ops: description of protocol
2431  *
2432  *      This function is called by a protocol handler that wants to
2433  *      advertise its address family, and have it linked into the
2434  *      socket interface. The value ops->family corresponds to the
2435  *      socket system call protocol family.
2436  */
2437 int sock_register(const struct net_proto_family *ops)
2438 {
2439         int err;
2440 
2441         if (ops->family >= NPROTO) {
2442                 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2443                 return -ENOBUFS;
2444         }
2445 
2446         spin_lock(&net_family_lock);
2447         if (rcu_dereference_protected(net_families[ops->family],
2448                                       lockdep_is_held(&net_family_lock)))
2449                 err = -EEXIST;
2450         else {
2451                 rcu_assign_pointer(net_families[ops->family], ops);
2452                 err = 0;
2453         }
2454         spin_unlock(&net_family_lock);
2455 
2456         pr_info("NET: Registered protocol family %d\n", ops->family);
2457         return err;
2458 }
2459 EXPORT_SYMBOL(sock_register);
2460 
2461 /**
2462  *      sock_unregister - remove a protocol handler
2463  *      @family: protocol family to remove
2464  *
2465  *      This function is called by a protocol handler that wants to
2466  *      remove its address family, and have it unlinked from the
2467  *      new socket creation.
2468  *
2469  *      If protocol handler is a module, then it can use module reference
2470  *      counts to protect against new references. If protocol handler is not
2471  *      a module then it needs to provide its own protection in
2472  *      the ops->create routine.
2473  */
2474 void sock_unregister(int family)
2475 {
2476         BUG_ON(family < 0 || family >= NPROTO);
2477 
2478         spin_lock(&net_family_lock);
2479         RCU_INIT_POINTER(net_families[family], NULL);
2480         spin_unlock(&net_family_lock);
2481 
2482         synchronize_rcu();
2483 
2484         pr_info("NET: Unregistered protocol family %d\n", family);
2485 }
2486 EXPORT_SYMBOL(sock_unregister);
2487 
2488 static int __init sock_init(void)
2489 {
2490         int err;
2491         /*
2492          *      Initialize the network sysctl infrastructure.
2493          */
2494         err = net_sysctl_init();
2495         if (err)
2496                 goto out;
2497 
2498         /*
2499          *      Initialize skbuff SLAB cache
2500          */
2501         skb_init();
2502 
2503         /*
2504          *      Initialize the protocols module.
2505          */
2506 
2507         init_inodecache();
2508 
2509         err = register_filesystem(&sock_fs_type);
2510         if (err)
2511                 goto out_fs;
2512         sock_mnt = kern_mount(&sock_fs_type);
2513         if (IS_ERR(sock_mnt)) {
2514                 err = PTR_ERR(sock_mnt);
2515                 goto out_mount;
2516         }
2517 
2518         /* The real protocol initialization is performed in later initcalls.
2519          */
2520 
2521 #ifdef CONFIG_NETFILTER
2522         err = netfilter_init();
2523         if (err)
2524                 goto out;
2525 #endif
2526 
2527         ptp_classifier_init();
2528 
2529 out:
2530         return err;
2531 
2532 out_mount:
2533         unregister_filesystem(&sock_fs_type);
2534 out_fs:
2535         goto out;
2536 }
2537 
2538 core_initcall(sock_init);       /* early initcall */
2539 
2540 #ifdef CONFIG_PROC_FS
2541 void socket_seq_show(struct seq_file *seq)
2542 {
2543         int cpu;
2544         int counter = 0;
2545 
2546         for_each_possible_cpu(cpu)
2547             counter += per_cpu(sockets_in_use, cpu);
2548 
2549         /* It can be negative, by the way. 8) */
2550         if (counter < 0)
2551                 counter = 0;
2552 
2553         seq_printf(seq, "sockets: used %d\n", counter);
2554 }
2555 #endif                          /* CONFIG_PROC_FS */
2556 
2557 #ifdef CONFIG_COMPAT
2558 static int do_siocgstamp(struct net *net, struct socket *sock,
2559                          unsigned int cmd, void __user *up)
2560 {
2561         mm_segment_t old_fs = get_fs();
2562         struct timeval ktv;
2563         int err;
2564 
2565         set_fs(KERNEL_DS);
2566         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2567         set_fs(old_fs);
2568         if (!err)
2569                 err = compat_put_timeval(&ktv, up);
2570 
2571         return err;
2572 }
2573 
2574 static int do_siocgstampns(struct net *net, struct socket *sock,
2575                            unsigned int cmd, void __user *up)
2576 {
2577         mm_segment_t old_fs = get_fs();
2578         struct timespec kts;
2579         int err;
2580 
2581         set_fs(KERNEL_DS);
2582         err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2583         set_fs(old_fs);
2584         if (!err)
2585                 err = compat_put_timespec(&kts, up);
2586 
2587         return err;
2588 }
2589 
2590 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2591 {
2592         struct ifreq __user *uifr;
2593         int err;
2594 
2595         uifr = compat_alloc_user_space(sizeof(struct ifreq));
2596         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2597                 return -EFAULT;
2598 
2599         err = dev_ioctl(net, SIOCGIFNAME, uifr);
2600         if (err)
2601                 return err;
2602 
2603         if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2604                 return -EFAULT;
2605 
2606         return 0;
2607 }
2608 
2609 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2610 {
2611         struct compat_ifconf ifc32;
2612         struct ifconf ifc;
2613         struct ifconf __user *uifc;
2614         struct compat_ifreq __user *ifr32;
2615         struct ifreq __user *ifr;
2616         unsigned int i, j;
2617         int err;
2618 
2619         if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2620                 return -EFAULT;
2621 
2622         memset(&ifc, 0, sizeof(ifc));
2623         if (ifc32.ifcbuf == 0) {
2624                 ifc32.ifc_len = 0;
2625                 ifc.ifc_len = 0;
2626                 ifc.ifc_req = NULL;
2627                 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2628         } else {
2629                 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2630                         sizeof(struct ifreq);
2631                 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2632                 ifc.ifc_len = len;
2633                 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2634                 ifr32 = compat_ptr(ifc32.ifcbuf);
2635                 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2636                         if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2637                                 return -EFAULT;
2638                         ifr++;
2639                         ifr32++;
2640                 }
2641         }
2642         if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2643                 return -EFAULT;
2644 
2645         err = dev_ioctl(net, SIOCGIFCONF, uifc);
2646         if (err)
2647                 return err;
2648 
2649         if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2650                 return -EFAULT;
2651 
2652         ifr = ifc.ifc_req;
2653         ifr32 = compat_ptr(ifc32.ifcbuf);
2654         for (i = 0, j = 0;
2655              i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2656              i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2657                 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2658                         return -EFAULT;
2659                 ifr32++;
2660                 ifr++;
2661         }
2662 
2663         if (ifc32.ifcbuf == 0) {
2664                 /* Translate from 64-bit structure multiple to
2665                  * a 32-bit one.
2666                  */
2667                 i = ifc.ifc_len;
2668                 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2669                 ifc32.ifc_len = i;
2670         } else {
2671                 ifc32.ifc_len = i;
2672         }
2673         if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2674                 return -EFAULT;
2675 
2676         return 0;
2677 }
2678 
2679 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2680 {
2681         struct compat_ethtool_rxnfc __user *compat_rxnfc;
2682         bool convert_in = false, convert_out = false;
2683         size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2684         struct ethtool_rxnfc __user *rxnfc;
2685         struct ifreq __user *ifr;
2686         u32 rule_cnt = 0, actual_rule_cnt;
2687         u32 ethcmd;
2688         u32 data;
2689         int ret;
2690 
2691         if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2692                 return -EFAULT;
2693 
2694         compat_rxnfc = compat_ptr(data);
2695 
2696         if (get_user(ethcmd, &compat_rxnfc->cmd))
2697                 return -EFAULT;
2698 
2699         /* Most ethtool structures are defined without padding.
2700          * Unfortunately struct ethtool_rxnfc is an exception.
2701          */
2702         switch (ethcmd) {
2703         default:
2704                 break;
2705         case ETHTOOL_GRXCLSRLALL:
2706                 /* Buffer size is variable */
2707                 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2708                         return -EFAULT;
2709                 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2710                         return -ENOMEM;
2711                 buf_size += rule_cnt * sizeof(u32);
2712                 /* fall through */
2713         case ETHTOOL_GRXRINGS:
2714         case ETHTOOL_GRXCLSRLCNT:
2715         case ETHTOOL_GRXCLSRULE:
2716         case ETHTOOL_SRXCLSRLINS:
2717                 convert_out = true;
2718                 /* fall through */
2719         case ETHTOOL_SRXCLSRLDEL:
2720                 buf_size += sizeof(struct ethtool_rxnfc);
2721                 convert_in = true;
2722                 break;
2723         }
2724 
2725         ifr = compat_alloc_user_space(buf_size);
2726         rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2727 
2728         if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2729                 return -EFAULT;
2730 
2731         if (put_user(convert_in ? rxnfc : compat_ptr(data),
2732                      &ifr->ifr_ifru.ifru_data))
2733                 return -EFAULT;
2734 
2735         if (convert_in) {
2736                 /* We expect there to be holes between fs.m_ext and
2737                  * fs.ring_cookie and at the end of fs, but nowhere else.
2738                  */
2739                 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2740                              sizeof(compat_rxnfc->fs.m_ext) !=
2741                              offsetof(struct ethtool_rxnfc, fs.m_ext) +
2742                              sizeof(rxnfc->fs.m_ext));
2743                 BUILD_BUG_ON(
2744                         offsetof(struct compat_ethtool_rxnfc, fs.location) -
2745                         offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2746                         offsetof(struct ethtool_rxnfc, fs.location) -
2747                         offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2748 
2749                 if (copy_in_user(rxnfc, compat_rxnfc,
2750                                  (void __user *)(&rxnfc->fs.m_ext + 1) -
2751                                  (void __user *)rxnfc) ||
2752                     copy_in_user(&rxnfc->fs.ring_cookie,
2753                                  &compat_rxnfc->fs.ring_cookie,
2754                                  (void __user *)(&rxnfc->fs.location + 1) -
2755                                  (void __user *)&rxnfc->fs.ring_cookie) ||
2756                     copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2757                                  sizeof(rxnfc->rule_cnt)))
2758                         return -EFAULT;
2759         }
2760 
2761         ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2762         if (ret)
2763                 return ret;
2764 
2765         if (convert_out) {
2766                 if (copy_in_user(compat_rxnfc, rxnfc,
2767                                  (const void __user *)(&rxnfc->fs.m_ext + 1) -
2768                                  (const void __user *)rxnfc) ||
2769                     copy_in_user(&compat_rxnfc->fs.ring_cookie,
2770                                  &rxnfc->fs.ring_cookie,
2771                                  (const void __user *)(&rxnfc->fs.location + 1) -
2772                                  (const void __user *)&rxnfc->fs.ring_cookie) ||
2773                     copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2774                                  sizeof(rxnfc->rule_cnt)))
2775                         return -EFAULT;
2776 
2777                 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2778                         /* As an optimisation, we only copy the actual
2779                          * number of rules that the underlying
2780                          * function returned.  Since Mallory might
2781                          * change the rule count in user memory, we
2782                          * check that it is less than the rule count
2783                          * originally given (as the user buffer size),
2784                          * which has been range-checked.
2785                          */
2786                         if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2787                                 return -EFAULT;
2788                         if (actual_rule_cnt < rule_cnt)
2789                                 rule_cnt = actual_rule_cnt;
2790                         if (copy_in_user(&compat_rxnfc->rule_locs[0],
2791                                          &rxnfc->rule_locs[0],
2792                                          rule_cnt * sizeof(u32)))
2793                                 return -EFAULT;
2794                 }
2795         }
2796 
2797         return 0;
2798 }
2799 
2800 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2801 {
2802         void __user *uptr;
2803         compat_uptr_t uptr32;
2804         struct ifreq __user *uifr;
2805 
2806         uifr = compat_alloc_user_space(sizeof(*uifr));
2807         if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2808                 return -EFAULT;
2809 
2810         if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2811                 return -EFAULT;
2812 
2813         uptr = compat_ptr(uptr32);
2814 
2815         if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2816                 return -EFAULT;
2817 
2818         return dev_ioctl(net, SIOCWANDEV, uifr);
2819 }
2820 
2821 static int bond_ioctl(struct net *net, unsigned int cmd,
2822                          struct compat_ifreq __user *ifr32)
2823 {
2824         struct ifreq kifr;
2825         mm_segment_t old_fs;
2826         int err;
2827 
2828         switch (cmd) {
2829         case SIOCBONDENSLAVE:
2830         case SIOCBONDRELEASE:
2831         case SIOCBONDSETHWADDR:
2832         case SIOCBONDCHANGEACTIVE:
2833                 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2834                         return -EFAULT;
2835 
2836                 old_fs = get_fs();
2837                 set_fs(KERNEL_DS);
2838                 err = dev_ioctl(net, cmd,
2839                                 (struct ifreq __user __force *) &kifr);
2840                 set_fs(old_fs);
2841 
2842                 return err;
2843         default:
2844                 return -ENOIOCTLCMD;
2845         }
2846 }
2847 
2848 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2849 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2850                                  struct compat_ifreq __user *u_ifreq32)
2851 {
2852         struct ifreq __user *u_ifreq64;
2853         char tmp_buf[IFNAMSIZ];
2854         void __user *data64;
2855         u32 data32;
2856 
2857         if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2858                            IFNAMSIZ))
2859                 return -EFAULT;
2860         if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2861                 return -EFAULT;
2862         data64 = compat_ptr(data32);
2863 
2864         u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2865 
2866         if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2867                          IFNAMSIZ))
2868                 return -EFAULT;
2869         if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2870                 return -EFAULT;
2871 
2872         return dev_ioctl(net, cmd, u_ifreq64);
2873 }
2874 
2875 static int dev_ifsioc(struct net *net, struct socket *sock,
2876                          unsigned int cmd, struct compat_ifreq __user *uifr32)
2877 {
2878         struct ifreq __user *uifr;
2879         int err;
2880 
2881         uifr = compat_alloc_user_space(sizeof(*uifr));
2882         if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2883                 return -EFAULT;
2884 
2885         err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2886 
2887         if (!err) {
2888                 switch (cmd) {
2889                 case SIOCGIFFLAGS:
2890                 case SIOCGIFMETRIC:
2891                 case SIOCGIFMTU:
2892                 case SIOCGIFMEM:
2893                 case SIOCGIFHWADDR:
2894                 case SIOCGIFINDEX:
2895                 case SIOCGIFADDR:
2896                 case SIOCGIFBRDADDR:
2897                 case SIOCGIFDSTADDR:
2898                 case SIOCGIFNETMASK:
2899                 case SIOCGIFPFLAGS:
2900                 case SIOCGIFTXQLEN:
2901                 case SIOCGMIIPHY:
2902                 case SIOCGMIIREG:
2903                         if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2904                                 err = -EFAULT;
2905                         break;
2906                 }
2907         }
2908         return err;
2909 }
2910 
2911 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2912                         struct compat_ifreq __user *uifr32)
2913 {
2914         struct ifreq ifr;
2915         struct compat_ifmap __user *uifmap32;
2916         mm_segment_t old_fs;
2917         int err;
2918 
2919         uifmap32 = &uifr32->ifr_ifru.ifru_map;
2920         err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2921         err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2922         err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2923         err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2924         err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2925         err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2926         err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2927         if (err)
2928                 return -EFAULT;
2929 
2930         old_fs = get_fs();
2931         set_fs(KERNEL_DS);
2932         err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
2933         set_fs(old_fs);
2934 
2935         if (cmd == SIOCGIFMAP && !err) {
2936                 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2937                 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2938                 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2939                 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2940                 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2941                 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2942                 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2943                 if (err)
2944                         err = -EFAULT;
2945         }
2946         return err;
2947 }
2948 
2949 struct rtentry32 {
2950         u32             rt_pad1;
2951         struct sockaddr rt_dst;         /* target address               */
2952         struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
2953         struct sockaddr rt_genmask;     /* target network mask (IP)     */
2954         unsigned short  rt_flags;
2955         short           rt_pad2;
2956         u32             rt_pad3;
2957         unsigned char   rt_tos;
2958         unsigned char   rt_class;
2959         short           rt_pad4;
2960         short           rt_metric;      /* +1 for binary compatibility! */
2961         /* char * */ u32 rt_dev;        /* forcing the device at add    */
2962         u32             rt_mtu;         /* per route MTU/Window         */
2963         u32             rt_window;      /* Window clamping              */
2964         unsigned short  rt_irtt;        /* Initial RTT                  */
2965 };
2966 
2967 struct in6_rtmsg32 {
2968         struct in6_addr         rtmsg_dst;
2969         struct in6_addr         rtmsg_src;
2970         struct in6_addr         rtmsg_gateway;
2971         u32                     rtmsg_type;
2972         u16                     rtmsg_dst_len;
2973         u16                     rtmsg_src_len;
2974         u32                     rtmsg_metric;
2975         u32                     rtmsg_info;
2976         u32                     rtmsg_flags;
2977         s32                     rtmsg_ifindex;
2978 };
2979 
2980 static int routing_ioctl(struct net *net, struct socket *sock,
2981                          unsigned int cmd, void __user *argp)
2982 {
2983         int ret;
2984         void *r = NULL;
2985         struct in6_rtmsg r6;
2986         struct rtentry r4;
2987         char devname[16];
2988         u32 rtdev;
2989         mm_segment_t old_fs = get_fs();
2990 
2991         if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2992                 struct in6_rtmsg32 __user *ur6 = argp;
2993                 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2994                         3 * sizeof(struct in6_addr));
2995                 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2996                 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2997                 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2998                 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2999                 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3000                 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3001                 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3002 
3003                 r = (void *) &r6;
3004         } else { /* ipv4 */
3005                 struct rtentry32 __user *ur4 = argp;
3006                 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3007                                         3 * sizeof(struct sockaddr));
3008                 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3009                 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3010                 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3011                 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3012                 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3013                 ret |= get_user(rtdev, &(ur4->rt_dev));
3014                 if (rtdev) {
3015                         ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3016                         r4.rt_dev = (char __user __force *)devname;
3017                         devname[15] = 0;
3018                 } else
3019                         r4.rt_dev = NULL;
3020 
3021                 r = (void *) &r4;
3022         }
3023 
3024         if (ret) {
3025                 ret = -EFAULT;
3026                 goto out;
3027         }
3028 
3029         set_fs(KERNEL_DS);
3030         ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3031         set_fs(old_fs);
3032 
3033 out:
3034         return ret;
3035 }
3036 
3037 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3038  * for some operations; this forces use of the newer bridge-utils that
3039  * use compatible ioctls
3040  */
3041 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3042 {
3043         compat_ulong_t tmp;
3044 
3045         if (get_user(tmp, argp))
3046                 return -EFAULT;
3047         if (tmp == BRCTL_GET_VERSION)
3048                 return BRCTL_VERSION + 1;
3049         return -EINVAL;
3050 }
3051 
3052 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3053                          unsigned int cmd, unsigned long arg)
3054 {
3055         void __user *argp = compat_ptr(arg);
3056         struct sock *sk = sock->sk;
3057         struct net *net = sock_net(sk);
3058 
3059         if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3060                 return compat_ifr_data_ioctl(net, cmd, argp);
3061 
3062         switch (cmd) {
3063         case SIOCSIFBR:
3064         case SIOCGIFBR:
3065                 return old_bridge_ioctl(argp);
3066         case SIOCGIFNAME:
3067                 return dev_ifname32(net, argp);
3068         case SIOCGIFCONF:
3069                 return dev_ifconf(net, argp);
3070         case SIOCETHTOOL:
3071                 return ethtool_ioctl(net, argp);
3072         case SIOCWANDEV:
3073                 return compat_siocwandev(net, argp);
3074         case SIOCGIFMAP:
3075         case SIOCSIFMAP:
3076                 return compat_sioc_ifmap(net, cmd, argp);
3077         case SIOCBONDENSLAVE:
3078         case SIOCBONDRELEASE:
3079         case SIOCBONDSETHWADDR:
3080         case SIOCBONDCHANGEACTIVE:
3081                 return bond_ioctl(net, cmd, argp);
3082         case SIOCADDRT:
3083         case SIOCDELRT:
3084                 return routing_ioctl(net, sock, cmd, argp);
3085         case SIOCGSTAMP:
3086                 return do_siocgstamp(net, sock, cmd, argp);
3087         case SIOCGSTAMPNS:
3088                 return do_siocgstampns(net, sock, cmd, argp);
3089         case SIOCBONDSLAVEINFOQUERY:
3090         case SIOCBONDINFOQUERY:
3091         case SIOCSHWTSTAMP:
3092         case SIOCGHWTSTAMP:
3093                 return compat_ifr_data_ioctl(net, cmd, argp);
3094 
3095         case FIOSETOWN:
3096         case SIOCSPGRP:
3097         case FIOGETOWN:
3098         case SIOCGPGRP:
3099         case SIOCBRADDBR:
3100         case SIOCBRDELBR:
3101         case SIOCGIFVLAN:
3102         case SIOCSIFVLAN:
3103         case SIOCADDDLCI:
3104         case SIOCDELDLCI:
3105                 return sock_ioctl(file, cmd, arg);
3106 
3107         case SIOCGIFFLAGS:
3108         case SIOCSIFFLAGS:
3109         case SIOCGIFMETRIC:
3110         case SIOCSIFMETRIC:
3111         case SIOCGIFMTU:
3112         case SIOCSIFMTU:
3113         case SIOCGIFMEM:
3114         case SIOCSIFMEM:
3115         case SIOCGIFHWADDR:
3116         case SIOCSIFHWADDR:
3117         case SIOCADDMULTI:
3118         case SIOCDELMULTI:
3119         case SIOCGIFINDEX:
3120         case SIOCGIFADDR:
3121         case SIOCSIFADDR:
3122         case SIOCSIFHWBROADCAST:
3123         case SIOCDIFADDR:
3124         case SIOCGIFBRDADDR:
3125         case SIOCSIFBRDADDR:
3126         case SIOCGIFDSTADDR:
3127         case SIOCSIFDSTADDR:
3128         case SIOCGIFNETMASK:
3129         case SIOCSIFNETMASK:
3130         case SIOCSIFPFLAGS:
3131         case SIOCGIFPFLAGS:
3132         case SIOCGIFTXQLEN:
3133         case SIOCSIFTXQLEN:
3134         case SIOCBRADDIF:
3135         case SIOCBRDELIF:
3136         case SIOCSIFNAME:
3137         case SIOCGMIIPHY:
3138         case SIOCGMIIREG:
3139         case SIOCSMIIREG:
3140                 return dev_ifsioc(net, sock, cmd, argp);
3141 
3142         case SIOCSARP:
3143         case SIOCGARP:
3144         case SIOCDARP:
3145         case SIOCATMARK:
3146                 return sock_do_ioctl(net, sock, cmd, arg);
3147         }
3148 
3149         return -ENOIOCTLCMD;
3150 }
3151 
3152 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3153                               unsigned long arg)
3154 {
3155         struct socket *sock = file->private_data;
3156         int ret = -ENOIOCTLCMD;
3157         struct sock *sk;
3158         struct net *net;
3159 
3160         sk = sock->sk;
3161         net = sock_net(sk);
3162 
3163         if (sock->ops->compat_ioctl)
3164                 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3165 
3166         if (ret == -ENOIOCTLCMD &&
3167             (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3168                 ret = compat_wext_handle_ioctl(net, cmd, arg);
3169 
3170         if (ret == -ENOIOCTLCMD)
3171                 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3172 
3173         return ret;
3174 }
3175 #endif
3176 
3177 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3178 {
3179         return sock->ops->bind(sock, addr, addrlen);
3180 }
3181 EXPORT_SYMBOL(kernel_bind);
3182 
3183 int kernel_listen(struct socket *sock, int backlog)
3184 {
3185         return sock->ops->listen(sock, backlog);
3186 }
3187 EXPORT_SYMBOL(kernel_listen);
3188 
3189 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3190 {
3191         struct sock *sk = sock->sk;
3192         int err;
3193 
3194         err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3195                                newsock);
3196         if (err < 0)
3197                 goto done;
3198 
3199         err = sock->ops->accept(sock, *newsock, flags);
3200         if (err < 0) {
3201                 sock_release(*newsock);
3202                 *newsock = NULL;
3203                 goto done;
3204         }
3205 
3206         (*newsock)->ops = sock->ops;
3207         __module_get((*newsock)->ops->owner);
3208 
3209 done:
3210         return err;
3211 }
3212 EXPORT_SYMBOL(kernel_accept);
3213 
3214 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3215                    int flags)
3216 {
3217         return sock->ops->connect(sock, addr, addrlen, flags);
3218 }
3219 EXPORT_SYMBOL(kernel_connect);
3220 
3221 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3222                          int *addrlen)
3223 {
3224         return sock->ops->getname(sock, addr, addrlen, 0);
3225 }
3226 EXPORT_SYMBOL(kernel_getsockname);
3227 
3228 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3229                          int *addrlen)
3230 {
3231         return sock->ops->getname(sock, addr, addrlen, 1);
3232 }
3233 EXPORT_SYMBOL(kernel_getpeername);
3234 
3235 int kernel_getsockopt(struct socket *sock, int level, int optname,
3236                         char *optval, int *optlen)
3237 {
3238         mm_segment_t oldfs = get_fs();
3239         char __user *uoptval;
3240         int __user *uoptlen;
3241         int err;
3242 
3243         uoptval = (char __user __force *) optval;
3244         uoptlen = (int __user __force *) optlen;
3245 
3246         set_fs(KERNEL_DS);
3247         if (level == SOL_SOCKET)
3248                 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3249         else
3250                 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3251                                             uoptlen);
3252         set_fs(oldfs);
3253         return err;
3254 }
3255 EXPORT_SYMBOL(kernel_getsockopt);
3256 
3257 int kernel_setsockopt(struct socket *sock, int level, int optname,
3258                         char *optval, unsigned int optlen)
3259 {
3260         mm_segment_t oldfs = get_fs();
3261         char __user *uoptval;
3262         int err;
3263 
3264         uoptval = (char __user __force *) optval;
3265 
3266         set_fs(KERNEL_DS);
3267         if (level == SOL_SOCKET)
3268                 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3269         else
3270                 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3271                                             optlen);
3272         set_fs(oldfs);
3273         return err;
3274 }
3275 EXPORT_SYMBOL(kernel_setsockopt);
3276 
3277 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3278                     size_t size, int flags)
3279 {
3280         if (sock->ops->sendpage)
3281                 return sock->ops->sendpage(sock, page, offset, size, flags);
3282 
3283         return sock_no_sendpage(sock, page, offset, size, flags);
3284 }
3285 EXPORT_SYMBOL(kernel_sendpage);
3286 
3287 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3288 {
3289         mm_segment_t oldfs = get_fs();
3290         int err;
3291 
3292         set_fs(KERNEL_DS);
3293         err = sock->ops->ioctl(sock, cmd, arg);
3294         set_fs(oldfs);
3295 
3296         return err;
3297 }
3298 EXPORT_SYMBOL(kernel_sock_ioctl);
3299 
3300 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3301 {
3302         return sock->ops->shutdown(sock, how);
3303 }
3304 EXPORT_SYMBOL(kernel_sock_shutdown);
3305 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us