Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * NET An implementation of the SOCKET network access protocol. | |
4 | * | |
5 | * Version: @(#)socket.c 1.1.93 18/02/95 | |
6 | * | |
7 | * Authors: Orest Zborowski, <obz@Kodak.COM> | |
02c30a84 | 8 | * Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * | |
11 | * Fixes: | |
12 | * Anonymous : NOTSOCK/BADF cleanup. Error fix in | |
13 | * shutdown() | |
14 | * Alan Cox : verify_area() fixes | |
15 | * Alan Cox : Removed DDI | |
16 | * Jonathan Kamens : SOCK_DGRAM reconnect bug | |
17 | * Alan Cox : Moved a load of checks to the very | |
18 | * top level. | |
19 | * Alan Cox : Move address structures to/from user | |
20 | * mode above the protocol layers. | |
21 | * Rob Janssen : Allow 0 length sends. | |
22 | * Alan Cox : Asynchronous I/O support (cribbed from the | |
23 | * tty drivers). | |
24 | * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style) | |
25 | * Jeff Uphoff : Made max number of sockets command-line | |
26 | * configurable. | |
27 | * Matti Aarnio : Made the number of sockets dynamic, | |
28 | * to be allocated when needed, and mr. | |
29 | * Uphoff's max is used as max to be | |
30 | * allowed to allocate. | |
31 | * Linus : Argh. removed all the socket allocation | |
32 | * altogether: it's in the inode now. | |
33 | * Alan Cox : Made sock_alloc()/sock_release() public | |
34 | * for NetROM and future kernel nfsd type | |
35 | * stuff. | |
36 | * Alan Cox : sendmsg/recvmsg basics. | |
37 | * Tom Dyas : Export net symbols. | |
38 | * Marcin Dalecki : Fixed problems with CONFIG_NET="n". | |
39 | * Alan Cox : Added thread locking to sys_* calls | |
40 | * for sockets. May have errors at the | |
41 | * moment. | |
42 | * Kevin Buhr : Fixed the dumb errors in the above. | |
43 | * Andi Kleen : Some small cleanups, optimizations, | |
44 | * and fixed a copy_from_user() bug. | |
45 | * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0) | |
89bddce5 | 46 | * Tigran Aivazian : Made listen(2) backlog sanity checks |
1da177e4 LT |
47 | * protocol-independent |
48 | * | |
1da177e4 | 49 | * This module is effectively the top level interface to the BSD socket |
89bddce5 | 50 | * paradigm. |
1da177e4 LT |
51 | * |
52 | * Based upon Swansea University Computer Society NET3.039 | |
53 | */ | |
54 | ||
aef2feda | 55 | #include <linux/bpf-cgroup.h> |
cc69837f | 56 | #include <linux/ethtool.h> |
1da177e4 | 57 | #include <linux/mm.h> |
1da177e4 LT |
58 | #include <linux/socket.h> |
59 | #include <linux/file.h> | |
2dc334f1 | 60 | #include <linux/splice.h> |
1da177e4 LT |
61 | #include <linux/net.h> |
62 | #include <linux/interrupt.h> | |
aaca0bdc | 63 | #include <linux/thread_info.h> |
55737fda | 64 | #include <linux/rcupdate.h> |
1da177e4 LT |
65 | #include <linux/netdevice.h> |
66 | #include <linux/proc_fs.h> | |
67 | #include <linux/seq_file.h> | |
4a3e2f71 | 68 | #include <linux/mutex.h> |
1da177e4 | 69 | #include <linux/if_bridge.h> |
20380731 | 70 | #include <linux/if_vlan.h> |
408eccce | 71 | #include <linux/ptp_classify.h> |
1da177e4 LT |
72 | #include <linux/init.h> |
73 | #include <linux/poll.h> | |
74 | #include <linux/cache.h> | |
75 | #include <linux/module.h> | |
76 | #include <linux/highmem.h> | |
1da177e4 | 77 | #include <linux/mount.h> |
fba9be49 | 78 | #include <linux/pseudo_fs.h> |
1da177e4 LT |
79 | #include <linux/security.h> |
80 | #include <linux/syscalls.h> | |
81 | #include <linux/compat.h> | |
82 | #include <linux/kmod.h> | |
3ec3b2fb | 83 | #include <linux/audit.h> |
d86b5e0e | 84 | #include <linux/wireless.h> |
1b8d7ae4 | 85 | #include <linux/nsproxy.h> |
1fd7317d | 86 | #include <linux/magic.h> |
5a0e3ad6 | 87 | #include <linux/slab.h> |
600e1779 | 88 | #include <linux/xattr.h> |
c8e8cd57 | 89 | #include <linux/nospec.h> |
8c3c447b | 90 | #include <linux/indirect_call_wrapper.h> |
8e9fad0e | 91 | #include <linux/io_uring.h> |
1da177e4 | 92 | |
7c0f6ba6 | 93 | #include <linux/uaccess.h> |
1da177e4 LT |
94 | #include <asm/unistd.h> |
95 | ||
96 | #include <net/compat.h> | |
87de87d5 | 97 | #include <net/wext.h> |
f8451725 | 98 | #include <net/cls_cgroup.h> |
1da177e4 LT |
99 | |
100 | #include <net/sock.h> | |
101 | #include <linux/netfilter.h> | |
102 | ||
6b96018b AB |
103 | #include <linux/if_tun.h> |
104 | #include <linux/ipv6_route.h> | |
105 | #include <linux/route.h> | |
c7dc504e | 106 | #include <linux/termios.h> |
6b96018b | 107 | #include <linux/sockios.h> |
076bb0c8 | 108 | #include <net/busy_poll.h> |
f24b9be5 | 109 | #include <linux/errqueue.h> |
d7c08826 | 110 | #include <linux/ptp_clock_kernel.h> |
6e6eda44 | 111 | #include <trace/events/sock.h> |
06021292 | 112 | |
e0d1095a | 113 | #ifdef CONFIG_NET_RX_BUSY_POLL |
64b0dc51 ET |
114 | unsigned int sysctl_net_busy_read __read_mostly; |
115 | unsigned int sysctl_net_busy_poll __read_mostly; | |
06021292 | 116 | #endif |
6b96018b | 117 | |
8ae5e030 AV |
118 | static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to); |
119 | static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from); | |
89bddce5 | 120 | static int sock_mmap(struct file *file, struct vm_area_struct *vma); |
1da177e4 LT |
121 | |
122 | static int sock_close(struct inode *inode, struct file *file); | |
a11e1d43 LT |
123 | static __poll_t sock_poll(struct file *file, |
124 | struct poll_table_struct *wait); | |
89bddce5 | 125 | static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); |
89bbfc95 SP |
126 | #ifdef CONFIG_COMPAT |
127 | static long compat_sock_ioctl(struct file *file, | |
89bddce5 | 128 | unsigned int cmd, unsigned long arg); |
89bbfc95 | 129 | #endif |
1da177e4 | 130 | static int sock_fasync(int fd, struct file *filp, int on); |
9c55e01c | 131 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, |
c6d409cf | 132 | struct pipe_inode_info *pipe, size_t len, |
9c55e01c | 133 | unsigned int flags); |
2bfc6685 | 134 | static void sock_splice_eof(struct file *file); |
542d3065 AB |
135 | |
136 | #ifdef CONFIG_PROC_FS | |
137 | static void sock_show_fdinfo(struct seq_file *m, struct file *f) | |
138 | { | |
139 | struct socket *sock = f->private_data; | |
1ded5e5a | 140 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
542d3065 | 141 | |
1ded5e5a ED |
142 | if (ops->show_fdinfo) |
143 | ops->show_fdinfo(m, sock); | |
542d3065 AB |
144 | } |
145 | #else | |
146 | #define sock_show_fdinfo NULL | |
147 | #endif | |
1da177e4 | 148 | |
1da177e4 LT |
149 | /* |
150 | * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear | |
151 | * in the operation structures but are done directly via the socketcall() multiplexor. | |
152 | */ | |
153 | ||
da7071d7 | 154 | static const struct file_operations socket_file_ops = { |
1da177e4 LT |
155 | .owner = THIS_MODULE, |
156 | .llseek = no_llseek, | |
8ae5e030 AV |
157 | .read_iter = sock_read_iter, |
158 | .write_iter = sock_write_iter, | |
1da177e4 LT |
159 | .poll = sock_poll, |
160 | .unlocked_ioctl = sock_ioctl, | |
89bbfc95 SP |
161 | #ifdef CONFIG_COMPAT |
162 | .compat_ioctl = compat_sock_ioctl, | |
163 | #endif | |
8e9fad0e | 164 | .uring_cmd = io_uring_cmd_sock, |
1da177e4 | 165 | .mmap = sock_mmap, |
1da177e4 LT |
166 | .release = sock_close, |
167 | .fasync = sock_fasync, | |
2dc334f1 | 168 | .splice_write = splice_to_socket, |
9c55e01c | 169 | .splice_read = sock_splice_read, |
2bfc6685 | 170 | .splice_eof = sock_splice_eof, |
b4653342 | 171 | .show_fdinfo = sock_show_fdinfo, |
1da177e4 LT |
172 | }; |
173 | ||
fe0bdbde YD |
174 | static const char * const pf_family_names[] = { |
175 | [PF_UNSPEC] = "PF_UNSPEC", | |
176 | [PF_UNIX] = "PF_UNIX/PF_LOCAL", | |
177 | [PF_INET] = "PF_INET", | |
178 | [PF_AX25] = "PF_AX25", | |
179 | [PF_IPX] = "PF_IPX", | |
180 | [PF_APPLETALK] = "PF_APPLETALK", | |
181 | [PF_NETROM] = "PF_NETROM", | |
182 | [PF_BRIDGE] = "PF_BRIDGE", | |
183 | [PF_ATMPVC] = "PF_ATMPVC", | |
184 | [PF_X25] = "PF_X25", | |
185 | [PF_INET6] = "PF_INET6", | |
186 | [PF_ROSE] = "PF_ROSE", | |
187 | [PF_DECnet] = "PF_DECnet", | |
188 | [PF_NETBEUI] = "PF_NETBEUI", | |
189 | [PF_SECURITY] = "PF_SECURITY", | |
190 | [PF_KEY] = "PF_KEY", | |
191 | [PF_NETLINK] = "PF_NETLINK/PF_ROUTE", | |
192 | [PF_PACKET] = "PF_PACKET", | |
193 | [PF_ASH] = "PF_ASH", | |
194 | [PF_ECONET] = "PF_ECONET", | |
195 | [PF_ATMSVC] = "PF_ATMSVC", | |
196 | [PF_RDS] = "PF_RDS", | |
197 | [PF_SNA] = "PF_SNA", | |
198 | [PF_IRDA] = "PF_IRDA", | |
199 | [PF_PPPOX] = "PF_PPPOX", | |
200 | [PF_WANPIPE] = "PF_WANPIPE", | |
201 | [PF_LLC] = "PF_LLC", | |
202 | [PF_IB] = "PF_IB", | |
203 | [PF_MPLS] = "PF_MPLS", | |
204 | [PF_CAN] = "PF_CAN", | |
205 | [PF_TIPC] = "PF_TIPC", | |
206 | [PF_BLUETOOTH] = "PF_BLUETOOTH", | |
207 | [PF_IUCV] = "PF_IUCV", | |
208 | [PF_RXRPC] = "PF_RXRPC", | |
209 | [PF_ISDN] = "PF_ISDN", | |
210 | [PF_PHONET] = "PF_PHONET", | |
211 | [PF_IEEE802154] = "PF_IEEE802154", | |
212 | [PF_CAIF] = "PF_CAIF", | |
213 | [PF_ALG] = "PF_ALG", | |
214 | [PF_NFC] = "PF_NFC", | |
215 | [PF_VSOCK] = "PF_VSOCK", | |
216 | [PF_KCM] = "PF_KCM", | |
217 | [PF_QIPCRTR] = "PF_QIPCRTR", | |
218 | [PF_SMC] = "PF_SMC", | |
219 | [PF_XDP] = "PF_XDP", | |
bc49d816 | 220 | [PF_MCTP] = "PF_MCTP", |
fe0bdbde YD |
221 | }; |
222 | ||
1da177e4 LT |
223 | /* |
224 | * The protocol list. Each protocol is registered in here. | |
225 | */ | |
226 | ||
1da177e4 | 227 | static DEFINE_SPINLOCK(net_family_lock); |
190683a9 | 228 | static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; |
1da177e4 | 229 | |
1da177e4 | 230 | /* |
89bddce5 SH |
231 | * Support routines. |
232 | * Move socket addresses back and forth across the kernel/user | |
233 | * divide and look after the messy bits. | |
1da177e4 LT |
234 | */ |
235 | ||
1da177e4 LT |
236 | /** |
237 | * move_addr_to_kernel - copy a socket address into kernel space | |
238 | * @uaddr: Address in user space | |
239 | * @kaddr: Address in kernel space | |
240 | * @ulen: Length in user space | |
241 | * | |
242 | * The address is copied into kernel space. If the provided address is | |
243 | * too long an error code of -EINVAL is returned. If the copy gives | |
244 | * invalid addresses -EFAULT is returned. On a success 0 is returned. | |
245 | */ | |
246 | ||
43db362d | 247 | int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr) |
1da177e4 | 248 | { |
230b1839 | 249 | if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) |
1da177e4 | 250 | return -EINVAL; |
89bddce5 | 251 | if (ulen == 0) |
1da177e4 | 252 | return 0; |
89bddce5 | 253 | if (copy_from_user(kaddr, uaddr, ulen)) |
1da177e4 | 254 | return -EFAULT; |
3ec3b2fb | 255 | return audit_sockaddr(ulen, kaddr); |
1da177e4 LT |
256 | } |
257 | ||
258 | /** | |
259 | * move_addr_to_user - copy an address to user space | |
260 | * @kaddr: kernel space address | |
261 | * @klen: length of address in kernel | |
262 | * @uaddr: user space address | |
263 | * @ulen: pointer to user length field | |
264 | * | |
265 | * The value pointed to by ulen on entry is the buffer length available. | |
266 | * This is overwritten with the buffer space used. -EINVAL is returned | |
267 | * if an overlong buffer is specified or a negative buffer size. -EFAULT | |
268 | * is returned if either the buffer or the length field are not | |
269 | * accessible. | |
270 | * After copying the data up to the limit the user specifies, the true | |
271 | * length of the data is written over the length limit the user | |
272 | * specified. Zero is returned for a success. | |
273 | */ | |
89bddce5 | 274 | |
43db362d | 275 | static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen, |
11165f14 | 276 | void __user *uaddr, int __user *ulen) |
1da177e4 LT |
277 | { |
278 | int err; | |
279 | int len; | |
280 | ||
68c6beb3 | 281 | BUG_ON(klen > sizeof(struct sockaddr_storage)); |
89bddce5 SH |
282 | err = get_user(len, ulen); |
283 | if (err) | |
1da177e4 | 284 | return err; |
89bddce5 SH |
285 | if (len > klen) |
286 | len = klen; | |
68c6beb3 | 287 | if (len < 0) |
1da177e4 | 288 | return -EINVAL; |
89bddce5 | 289 | if (len) { |
d6fe3945 SG |
290 | if (audit_sockaddr(klen, kaddr)) |
291 | return -ENOMEM; | |
89bddce5 | 292 | if (copy_to_user(uaddr, kaddr, len)) |
1da177e4 LT |
293 | return -EFAULT; |
294 | } | |
295 | /* | |
89bddce5 SH |
296 | * "fromlen shall refer to the value before truncation.." |
297 | * 1003.1g | |
1da177e4 LT |
298 | */ |
299 | return __put_user(klen, ulen); | |
300 | } | |
301 | ||
08009a76 | 302 | static struct kmem_cache *sock_inode_cachep __ro_after_init; |
1da177e4 LT |
303 | |
304 | static struct inode *sock_alloc_inode(struct super_block *sb) | |
305 | { | |
306 | struct socket_alloc *ei; | |
89bddce5 | 307 | |
fd60b288 | 308 | ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL); |
1da177e4 LT |
309 | if (!ei) |
310 | return NULL; | |
333f7909 AV |
311 | init_waitqueue_head(&ei->socket.wq.wait); |
312 | ei->socket.wq.fasync_list = NULL; | |
313 | ei->socket.wq.flags = 0; | |
89bddce5 | 314 | |
1da177e4 LT |
315 | ei->socket.state = SS_UNCONNECTED; |
316 | ei->socket.flags = 0; | |
317 | ei->socket.ops = NULL; | |
318 | ei->socket.sk = NULL; | |
319 | ei->socket.file = NULL; | |
1da177e4 LT |
320 | |
321 | return &ei->vfs_inode; | |
322 | } | |
323 | ||
6d7855c5 | 324 | static void sock_free_inode(struct inode *inode) |
1da177e4 | 325 | { |
43815482 ED |
326 | struct socket_alloc *ei; |
327 | ||
328 | ei = container_of(inode, struct socket_alloc, vfs_inode); | |
43815482 | 329 | kmem_cache_free(sock_inode_cachep, ei); |
1da177e4 LT |
330 | } |
331 | ||
51cc5068 | 332 | static void init_once(void *foo) |
1da177e4 | 333 | { |
89bddce5 | 334 | struct socket_alloc *ei = (struct socket_alloc *)foo; |
1da177e4 | 335 | |
a35afb83 | 336 | inode_init_once(&ei->vfs_inode); |
1da177e4 | 337 | } |
89bddce5 | 338 | |
1e911632 | 339 | static void init_inodecache(void) |
1da177e4 LT |
340 | { |
341 | sock_inode_cachep = kmem_cache_create("sock_inode_cache", | |
89bddce5 SH |
342 | sizeof(struct socket_alloc), |
343 | 0, | |
344 | (SLAB_HWCACHE_ALIGN | | |
345 | SLAB_RECLAIM_ACCOUNT | | |
5d097056 | 346 | SLAB_MEM_SPREAD | SLAB_ACCOUNT), |
20c2df83 | 347 | init_once); |
1e911632 | 348 | BUG_ON(sock_inode_cachep == NULL); |
1da177e4 LT |
349 | } |
350 | ||
b87221de | 351 | static const struct super_operations sockfs_ops = { |
c6d409cf | 352 | .alloc_inode = sock_alloc_inode, |
6d7855c5 | 353 | .free_inode = sock_free_inode, |
c6d409cf | 354 | .statfs = simple_statfs, |
1da177e4 LT |
355 | }; |
356 | ||
c23fbb6b ED |
357 | /* |
358 | * sockfs_dname() is called from d_path(). | |
359 | */ | |
360 | static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) | |
361 | { | |
0f60d288 | 362 | return dynamic_dname(buffer, buflen, "socket:[%lu]", |
c5ef6035 | 363 | d_inode(dentry)->i_ino); |
c23fbb6b ED |
364 | } |
365 | ||
3ba13d17 | 366 | static const struct dentry_operations sockfs_dentry_operations = { |
c23fbb6b | 367 | .d_dname = sockfs_dname, |
1da177e4 LT |
368 | }; |
369 | ||
bba0bd31 AG |
370 | static int sockfs_xattr_get(const struct xattr_handler *handler, |
371 | struct dentry *dentry, struct inode *inode, | |
372 | const char *suffix, void *value, size_t size) | |
373 | { | |
374 | if (value) { | |
375 | if (dentry->d_name.len + 1 > size) | |
376 | return -ERANGE; | |
377 | memcpy(value, dentry->d_name.name, dentry->d_name.len + 1); | |
378 | } | |
379 | return dentry->d_name.len + 1; | |
380 | } | |
381 | ||
382 | #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname" | |
383 | #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX) | |
384 | #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1) | |
385 | ||
386 | static const struct xattr_handler sockfs_xattr_handler = { | |
387 | .name = XATTR_NAME_SOCKPROTONAME, | |
388 | .get = sockfs_xattr_get, | |
389 | }; | |
390 | ||
4a590153 | 391 | static int sockfs_security_xattr_set(const struct xattr_handler *handler, |
39f60c1c | 392 | struct mnt_idmap *idmap, |
4a590153 AG |
393 | struct dentry *dentry, struct inode *inode, |
394 | const char *suffix, const void *value, | |
395 | size_t size, int flags) | |
396 | { | |
397 | /* Handled by LSM. */ | |
398 | return -EAGAIN; | |
399 | } | |
400 | ||
401 | static const struct xattr_handler sockfs_security_xattr_handler = { | |
402 | .prefix = XATTR_SECURITY_PREFIX, | |
403 | .set = sockfs_security_xattr_set, | |
404 | }; | |
405 | ||
bba0bd31 AG |
406 | static const struct xattr_handler *sockfs_xattr_handlers[] = { |
407 | &sockfs_xattr_handler, | |
4a590153 | 408 | &sockfs_security_xattr_handler, |
bba0bd31 AG |
409 | NULL |
410 | }; | |
411 | ||
fba9be49 | 412 | static int sockfs_init_fs_context(struct fs_context *fc) |
c74a1cbb | 413 | { |
fba9be49 DH |
414 | struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC); |
415 | if (!ctx) | |
416 | return -ENOMEM; | |
417 | ctx->ops = &sockfs_ops; | |
418 | ctx->dops = &sockfs_dentry_operations; | |
419 | ctx->xattr = sockfs_xattr_handlers; | |
420 | return 0; | |
c74a1cbb AV |
421 | } |
422 | ||
423 | static struct vfsmount *sock_mnt __read_mostly; | |
424 | ||
425 | static struct file_system_type sock_fs_type = { | |
426 | .name = "sockfs", | |
fba9be49 | 427 | .init_fs_context = sockfs_init_fs_context, |
c74a1cbb AV |
428 | .kill_sb = kill_anon_super, |
429 | }; | |
430 | ||
1da177e4 LT |
431 | /* |
432 | * Obtains the first available file descriptor and sets it up for use. | |
433 | * | |
39d8c1b6 DM |
434 | * These functions create file structures and maps them to fd space |
435 | * of the current process. On success it returns file descriptor | |
1da177e4 LT |
436 | * and file struct implicitly stored in sock->file. |
437 | * Note that another thread may close file descriptor before we return | |
438 | * from this function. We use the fact that now we do not refer | |
439 | * to socket after mapping. If one day we will need it, this | |
440 | * function will increment ref. count on file by 1. | |
441 | * | |
442 | * In any case returned fd MAY BE not valid! | |
443 | * This race condition is unavoidable | |
444 | * with shared fd spaces, we cannot solve it inside kernel, | |
445 | * but we take care of internal coherence yet. | |
446 | */ | |
447 | ||
8a3c245c PT |
448 | /** |
449 | * sock_alloc_file - Bind a &socket to a &file | |
450 | * @sock: socket | |
451 | * @flags: file status flags | |
452 | * @dname: protocol name | |
453 | * | |
454 | * Returns the &file bound with @sock, implicitly storing it | |
455 | * in sock->file. If dname is %NULL, sets to "". | |
649c15c7 TLSC |
456 | * |
457 | * On failure @sock is released, and an ERR pointer is returned. | |
458 | * | |
8a3c245c PT |
459 | * This function uses GFP_KERNEL internally. |
460 | */ | |
461 | ||
aab174f0 | 462 | struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname) |
1da177e4 | 463 | { |
7cbe66b6 | 464 | struct file *file; |
1da177e4 | 465 | |
d93aa9d8 AV |
466 | if (!dname) |
467 | dname = sock->sk ? sock->sk->sk_prot_creator->name : ""; | |
39d8c1b6 | 468 | |
d93aa9d8 AV |
469 | file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname, |
470 | O_RDWR | (flags & O_NONBLOCK), | |
471 | &socket_file_ops); | |
b5ffe634 | 472 | if (IS_ERR(file)) { |
8e1611e2 | 473 | sock_release(sock); |
39b65252 | 474 | return file; |
cc3808f8 AV |
475 | } |
476 | ||
fe34db06 | 477 | file->f_mode |= FMODE_NOWAIT; |
cc3808f8 | 478 | sock->file = file; |
39d8c1b6 | 479 | file->private_data = sock; |
d8e464ec | 480 | stream_open(SOCK_INODE(sock), file); |
28407630 | 481 | return file; |
39d8c1b6 | 482 | } |
56b31d1c | 483 | EXPORT_SYMBOL(sock_alloc_file); |
39d8c1b6 | 484 | |
56b31d1c | 485 | static int sock_map_fd(struct socket *sock, int flags) |
39d8c1b6 DM |
486 | { |
487 | struct file *newfile; | |
28407630 | 488 | int fd = get_unused_fd_flags(flags); |
ce4bb04c AV |
489 | if (unlikely(fd < 0)) { |
490 | sock_release(sock); | |
28407630 | 491 | return fd; |
ce4bb04c | 492 | } |
39d8c1b6 | 493 | |
aab174f0 | 494 | newfile = sock_alloc_file(sock, flags, NULL); |
4546e44c | 495 | if (!IS_ERR(newfile)) { |
39d8c1b6 | 496 | fd_install(fd, newfile); |
28407630 AV |
497 | return fd; |
498 | } | |
7cbe66b6 | 499 | |
28407630 AV |
500 | put_unused_fd(fd); |
501 | return PTR_ERR(newfile); | |
1da177e4 LT |
502 | } |
503 | ||
8a3c245c PT |
504 | /** |
505 | * sock_from_file - Return the &socket bounded to @file. | |
506 | * @file: file | |
8a3c245c | 507 | * |
dba4a925 | 508 | * On failure returns %NULL. |
8a3c245c PT |
509 | */ |
510 | ||
dba4a925 | 511 | struct socket *sock_from_file(struct file *file) |
6cb153ca | 512 | { |
6cb153ca | 513 | if (file->f_op == &socket_file_ops) |
da214a47 | 514 | return file->private_data; /* set in sock_alloc_file */ |
6cb153ca | 515 | |
23bb80d2 | 516 | return NULL; |
6cb153ca | 517 | } |
406a3c63 | 518 | EXPORT_SYMBOL(sock_from_file); |
6cb153ca | 519 | |
1da177e4 | 520 | /** |
c6d409cf | 521 | * sockfd_lookup - Go from a file number to its socket slot |
1da177e4 LT |
522 | * @fd: file handle |
523 | * @err: pointer to an error code return | |
524 | * | |
525 | * The file handle passed in is locked and the socket it is bound | |
241c4667 | 526 | * to is returned. If an error occurs the err pointer is overwritten |
1da177e4 LT |
527 | * with a negative errno code and NULL is returned. The function checks |
528 | * for both invalid handles and passing a handle which is not a socket. | |
529 | * | |
530 | * On a success the socket object pointer is returned. | |
531 | */ | |
532 | ||
533 | struct socket *sockfd_lookup(int fd, int *err) | |
534 | { | |
535 | struct file *file; | |
1da177e4 LT |
536 | struct socket *sock; |
537 | ||
89bddce5 SH |
538 | file = fget(fd); |
539 | if (!file) { | |
1da177e4 LT |
540 | *err = -EBADF; |
541 | return NULL; | |
542 | } | |
89bddce5 | 543 | |
dba4a925 FR |
544 | sock = sock_from_file(file); |
545 | if (!sock) { | |
546 | *err = -ENOTSOCK; | |
1da177e4 | 547 | fput(file); |
dba4a925 | 548 | } |
6cb153ca BL |
549 | return sock; |
550 | } | |
c6d409cf | 551 | EXPORT_SYMBOL(sockfd_lookup); |
1da177e4 | 552 | |
6cb153ca BL |
553 | static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) |
554 | { | |
00e188ef | 555 | struct fd f = fdget(fd); |
6cb153ca BL |
556 | struct socket *sock; |
557 | ||
3672558c | 558 | *err = -EBADF; |
00e188ef | 559 | if (f.file) { |
dba4a925 | 560 | sock = sock_from_file(f.file); |
00e188ef | 561 | if (likely(sock)) { |
ce787a5a | 562 | *fput_needed = f.flags & FDPUT_FPUT; |
6cb153ca | 563 | return sock; |
00e188ef | 564 | } |
dba4a925 | 565 | *err = -ENOTSOCK; |
00e188ef | 566 | fdput(f); |
1da177e4 | 567 | } |
6cb153ca | 568 | return NULL; |
1da177e4 LT |
569 | } |
570 | ||
600e1779 MY |
571 | static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer, |
572 | size_t size) | |
573 | { | |
574 | ssize_t len; | |
575 | ssize_t used = 0; | |
576 | ||
c5ef6035 | 577 | len = security_inode_listsecurity(d_inode(dentry), buffer, size); |
600e1779 MY |
578 | if (len < 0) |
579 | return len; | |
580 | used += len; | |
581 | if (buffer) { | |
582 | if (size < used) | |
583 | return -ERANGE; | |
584 | buffer += len; | |
585 | } | |
586 | ||
587 | len = (XATTR_NAME_SOCKPROTONAME_LEN + 1); | |
588 | used += len; | |
589 | if (buffer) { | |
590 | if (size < used) | |
591 | return -ERANGE; | |
592 | memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len); | |
593 | buffer += len; | |
594 | } | |
595 | ||
596 | return used; | |
597 | } | |
598 | ||
c1632a0f | 599 | static int sockfs_setattr(struct mnt_idmap *idmap, |
549c7297 | 600 | struct dentry *dentry, struct iattr *iattr) |
86741ec2 | 601 | { |
c1632a0f | 602 | int err = simple_setattr(&nop_mnt_idmap, dentry, iattr); |
86741ec2 | 603 | |
e1a3a60a | 604 | if (!err && (iattr->ia_valid & ATTR_UID)) { |
86741ec2 LC |
605 | struct socket *sock = SOCKET_I(d_inode(dentry)); |
606 | ||
6d8c50dc CW |
607 | if (sock->sk) |
608 | sock->sk->sk_uid = iattr->ia_uid; | |
609 | else | |
610 | err = -ENOENT; | |
86741ec2 LC |
611 | } |
612 | ||
613 | return err; | |
614 | } | |
615 | ||
600e1779 | 616 | static const struct inode_operations sockfs_inode_ops = { |
600e1779 | 617 | .listxattr = sockfs_listxattr, |
86741ec2 | 618 | .setattr = sockfs_setattr, |
600e1779 MY |
619 | }; |
620 | ||
1da177e4 | 621 | /** |
8a3c245c | 622 | * sock_alloc - allocate a socket |
89bddce5 | 623 | * |
1da177e4 LT |
624 | * Allocate a new inode and socket object. The two are bound together |
625 | * and initialised. The socket is then returned. If we are out of inodes | |
8a3c245c | 626 | * NULL is returned. This functions uses GFP_KERNEL internally. |
1da177e4 LT |
627 | */ |
628 | ||
f4a00aac | 629 | struct socket *sock_alloc(void) |
1da177e4 | 630 | { |
89bddce5 SH |
631 | struct inode *inode; |
632 | struct socket *sock; | |
1da177e4 | 633 | |
a209dfc7 | 634 | inode = new_inode_pseudo(sock_mnt->mnt_sb); |
1da177e4 LT |
635 | if (!inode) |
636 | return NULL; | |
637 | ||
638 | sock = SOCKET_I(inode); | |
639 | ||
85fe4025 | 640 | inode->i_ino = get_next_ino(); |
89bddce5 | 641 | inode->i_mode = S_IFSOCK | S_IRWXUGO; |
8192b0c4 DH |
642 | inode->i_uid = current_fsuid(); |
643 | inode->i_gid = current_fsgid(); | |
600e1779 | 644 | inode->i_op = &sockfs_inode_ops; |
1da177e4 | 645 | |
1da177e4 LT |
646 | return sock; |
647 | } | |
f4a00aac | 648 | EXPORT_SYMBOL(sock_alloc); |
1da177e4 | 649 | |
6d8c50dc | 650 | static void __sock_release(struct socket *sock, struct inode *inode) |
1da177e4 | 651 | { |
1ded5e5a ED |
652 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
653 | ||
654 | if (ops) { | |
655 | struct module *owner = ops->owner; | |
1da177e4 | 656 | |
6d8c50dc CW |
657 | if (inode) |
658 | inode_lock(inode); | |
1ded5e5a | 659 | ops->release(sock); |
ff7b11aa | 660 | sock->sk = NULL; |
6d8c50dc CW |
661 | if (inode) |
662 | inode_unlock(inode); | |
1da177e4 LT |
663 | sock->ops = NULL; |
664 | module_put(owner); | |
665 | } | |
666 | ||
333f7909 | 667 | if (sock->wq.fasync_list) |
3410f22e | 668 | pr_err("%s: fasync list not empty!\n", __func__); |
1da177e4 | 669 | |
1da177e4 LT |
670 | if (!sock->file) { |
671 | iput(SOCK_INODE(sock)); | |
672 | return; | |
673 | } | |
89bddce5 | 674 | sock->file = NULL; |
1da177e4 | 675 | } |
6d8c50dc | 676 | |
9a8ad9ac AL |
677 | /** |
678 | * sock_release - close a socket | |
679 | * @sock: socket to close | |
680 | * | |
681 | * The socket is released from the protocol stack if it has a release | |
682 | * callback, and the inode is then released if the socket is bound to | |
683 | * an inode not a file. | |
684 | */ | |
6d8c50dc CW |
685 | void sock_release(struct socket *sock) |
686 | { | |
687 | __sock_release(sock, NULL); | |
688 | } | |
c6d409cf | 689 | EXPORT_SYMBOL(sock_release); |
1da177e4 | 690 | |
c14ac945 | 691 | void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags) |
20d49473 | 692 | { |
140c55d4 ED |
693 | u8 flags = *tx_flags; |
694 | ||
51eb7492 | 695 | if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) { |
140c55d4 ED |
696 | flags |= SKBTX_HW_TSTAMP; |
697 | ||
51eb7492 GE |
698 | /* PTP hardware clocks can provide a free running cycle counter |
699 | * as a time base for virtual clocks. Tell driver to use the | |
700 | * free running cycle counter for timestamp if socket is bound | |
701 | * to virtual clock. | |
702 | */ | |
703 | if (tsflags & SOF_TIMESTAMPING_BIND_PHC) | |
704 | flags |= SKBTX_HW_TSTAMP_USE_CYCLES; | |
705 | } | |
706 | ||
c14ac945 | 707 | if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE) |
140c55d4 ED |
708 | flags |= SKBTX_SW_TSTAMP; |
709 | ||
c14ac945 | 710 | if (tsflags & SOF_TIMESTAMPING_TX_SCHED) |
140c55d4 ED |
711 | flags |= SKBTX_SCHED_TSTAMP; |
712 | ||
140c55d4 | 713 | *tx_flags = flags; |
20d49473 | 714 | } |
67cc0d40 | 715 | EXPORT_SYMBOL(__sock_tx_timestamp); |
20d49473 | 716 | |
8c3c447b PA |
717 | INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *, |
718 | size_t)); | |
a648a592 PA |
719 | INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *, |
720 | size_t)); | |
6e6eda44 YC |
721 | |
722 | static noinline void call_trace_sock_send_length(struct sock *sk, int ret, | |
723 | int flags) | |
724 | { | |
725 | trace_sock_send_length(sk, ret, 0); | |
726 | } | |
727 | ||
d8725c86 | 728 | static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) |
1da177e4 | 729 | { |
1ded5e5a | 730 | int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg, |
a648a592 PA |
731 | inet_sendmsg, sock, msg, |
732 | msg_data_left(msg)); | |
d8725c86 | 733 | BUG_ON(ret == -EIOCBQUEUED); |
6e6eda44 YC |
734 | |
735 | if (trace_sock_send_length_enabled()) | |
736 | call_trace_sock_send_length(sock->sk, ret, 0); | |
d8725c86 | 737 | return ret; |
1da177e4 LT |
738 | } |
739 | ||
85806af0 RD |
740 | /** |
741 | * sock_sendmsg - send a message through @sock | |
742 | * @sock: socket | |
743 | * @msg: message to send | |
744 | * | |
745 | * Sends @msg through @sock, passing through LSM. | |
746 | * Returns the number of bytes sent, or an error code. | |
747 | */ | |
d8725c86 | 748 | int sock_sendmsg(struct socket *sock, struct msghdr *msg) |
228e548e | 749 | { |
d8725c86 | 750 | int err = security_socket_sendmsg(sock, msg, |
01e97e65 | 751 | msg_data_left(msg)); |
228e548e | 752 | |
d8725c86 | 753 | return err ?: sock_sendmsg_nosec(sock, msg); |
0cf00c6f | 754 | } |
c6d409cf | 755 | EXPORT_SYMBOL(sock_sendmsg); |
1da177e4 | 756 | |
8a3c245c PT |
757 | /** |
758 | * kernel_sendmsg - send a message through @sock (kernel-space) | |
759 | * @sock: socket | |
760 | * @msg: message header | |
761 | * @vec: kernel vec | |
762 | * @num: vec array length | |
763 | * @size: total message data size | |
764 | * | |
765 | * Builds the message data with @vec and sends it through @sock. | |
766 | * Returns the number of bytes sent, or an error code. | |
767 | */ | |
768 | ||
1da177e4 LT |
769 | int kernel_sendmsg(struct socket *sock, struct msghdr *msg, |
770 | struct kvec *vec, size_t num, size_t size) | |
771 | { | |
de4eda9d | 772 | iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size); |
d8725c86 | 773 | return sock_sendmsg(sock, msg); |
1da177e4 | 774 | } |
c6d409cf | 775 | EXPORT_SYMBOL(kernel_sendmsg); |
1da177e4 | 776 | |
8a3c245c PT |
777 | /** |
778 | * kernel_sendmsg_locked - send a message through @sock (kernel-space) | |
779 | * @sk: sock | |
780 | * @msg: message header | |
781 | * @vec: output s/g array | |
782 | * @num: output s/g array length | |
783 | * @size: total message data size | |
784 | * | |
785 | * Builds the message data with @vec and sends it through @sock. | |
786 | * Returns the number of bytes sent, or an error code. | |
787 | * Caller must hold @sk. | |
788 | */ | |
789 | ||
306b13eb TH |
790 | int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg, |
791 | struct kvec *vec, size_t num, size_t size) | |
792 | { | |
793 | struct socket *sock = sk->sk_socket; | |
1ded5e5a | 794 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
306b13eb | 795 | |
1ded5e5a | 796 | if (!ops->sendmsg_locked) |
db5980d8 | 797 | return sock_no_sendmsg_locked(sk, msg, size); |
306b13eb | 798 | |
de4eda9d | 799 | iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size); |
306b13eb | 800 | |
1ded5e5a | 801 | return ops->sendmsg_locked(sk, msg, msg_data_left(msg)); |
306b13eb TH |
802 | } |
803 | EXPORT_SYMBOL(kernel_sendmsg_locked); | |
804 | ||
8605330a SHY |
805 | static bool skb_is_err_queue(const struct sk_buff *skb) |
806 | { | |
807 | /* pkt_type of skbs enqueued on the error queue are set to | |
808 | * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do | |
809 | * in recvmsg, since skbs received on a local socket will never | |
810 | * have a pkt_type of PACKET_OUTGOING. | |
811 | */ | |
812 | return skb->pkt_type == PACKET_OUTGOING; | |
813 | } | |
814 | ||
b50a5c70 ML |
815 | /* On transmit, software and hardware timestamps are returned independently. |
816 | * As the two skb clones share the hardware timestamp, which may be updated | |
817 | * before the software timestamp is received, a hardware TX timestamp may be | |
818 | * returned only if there is no software TX timestamp. Ignore false software | |
819 | * timestamps, which may be made in the __sock_recv_timestamp() call when the | |
7f1bc6e9 | 820 | * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a |
b50a5c70 ML |
821 | * hardware timestamp. |
822 | */ | |
823 | static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp) | |
824 | { | |
825 | return skb->tstamp && !false_tstamp && skb_is_err_queue(skb); | |
826 | } | |
827 | ||
97dc7cd9 GE |
828 | static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index) |
829 | { | |
830 | bool cycles = sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC; | |
831 | struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); | |
832 | struct net_device *orig_dev; | |
833 | ktime_t hwtstamp; | |
834 | ||
835 | rcu_read_lock(); | |
836 | orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); | |
837 | if (orig_dev) { | |
838 | *if_index = orig_dev->ifindex; | |
839 | hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles); | |
840 | } else { | |
841 | hwtstamp = shhwtstamps->hwtstamp; | |
842 | } | |
843 | rcu_read_unlock(); | |
844 | ||
845 | return hwtstamp; | |
846 | } | |
847 | ||
848 | static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb, | |
849 | int if_index) | |
aad9c8c4 ML |
850 | { |
851 | struct scm_ts_pktinfo ts_pktinfo; | |
852 | struct net_device *orig_dev; | |
853 | ||
854 | if (!skb_mac_header_was_set(skb)) | |
855 | return; | |
856 | ||
857 | memset(&ts_pktinfo, 0, sizeof(ts_pktinfo)); | |
858 | ||
97dc7cd9 GE |
859 | if (!if_index) { |
860 | rcu_read_lock(); | |
861 | orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); | |
862 | if (orig_dev) | |
863 | if_index = orig_dev->ifindex; | |
864 | rcu_read_unlock(); | |
865 | } | |
866 | ts_pktinfo.if_index = if_index; | |
aad9c8c4 ML |
867 | |
868 | ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb); | |
869 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO, | |
870 | sizeof(ts_pktinfo), &ts_pktinfo); | |
871 | } | |
872 | ||
92f37fd2 ED |
873 | /* |
874 | * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) | |
875 | */ | |
876 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, | |
877 | struct sk_buff *skb) | |
878 | { | |
20d49473 | 879 | int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); |
887feae3 | 880 | int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW); |
9718475e DD |
881 | struct scm_timestamping_internal tss; |
882 | ||
b50a5c70 | 883 | int empty = 1, false_tstamp = 0; |
20d49473 PO |
884 | struct skb_shared_hwtstamps *shhwtstamps = |
885 | skb_hwtstamps(skb); | |
97dc7cd9 | 886 | int if_index; |
007747a9 | 887 | ktime_t hwtstamp; |
20d49473 PO |
888 | |
889 | /* Race occurred between timestamp enabling and packet | |
890 | receiving. Fill in the current time for now. */ | |
b50a5c70 | 891 | if (need_software_tstamp && skb->tstamp == 0) { |
20d49473 | 892 | __net_timestamp(skb); |
b50a5c70 ML |
893 | false_tstamp = 1; |
894 | } | |
20d49473 PO |
895 | |
896 | if (need_software_tstamp) { | |
897 | if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { | |
887feae3 DD |
898 | if (new_tstamp) { |
899 | struct __kernel_sock_timeval tv; | |
900 | ||
901 | skb_get_new_timestamp(skb, &tv); | |
902 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW, | |
903 | sizeof(tv), &tv); | |
904 | } else { | |
905 | struct __kernel_old_timeval tv; | |
906 | ||
907 | skb_get_timestamp(skb, &tv); | |
908 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD, | |
909 | sizeof(tv), &tv); | |
910 | } | |
20d49473 | 911 | } else { |
887feae3 DD |
912 | if (new_tstamp) { |
913 | struct __kernel_timespec ts; | |
914 | ||
915 | skb_get_new_timestampns(skb, &ts); | |
916 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW, | |
917 | sizeof(ts), &ts); | |
918 | } else { | |
df1b4ba9 | 919 | struct __kernel_old_timespec ts; |
887feae3 DD |
920 | |
921 | skb_get_timestampns(skb, &ts); | |
922 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD, | |
923 | sizeof(ts), &ts); | |
924 | } | |
20d49473 PO |
925 | } |
926 | } | |
927 | ||
f24b9be5 | 928 | memset(&tss, 0, sizeof(tss)); |
c199105d | 929 | if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) && |
9718475e | 930 | ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0)) |
20d49473 | 931 | empty = 0; |
4d276eb6 | 932 | if (shhwtstamps && |
b9f40e21 | 933 | (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) && |
d7c08826 | 934 | !skb_is_swtx_tstamp(skb, false_tstamp)) { |
97dc7cd9 GE |
935 | if_index = 0; |
936 | if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV) | |
937 | hwtstamp = get_timestamp(sk, skb, &if_index); | |
007747a9 ML |
938 | else |
939 | hwtstamp = shhwtstamps->hwtstamp; | |
d7c08826 | 940 | |
97dc7cd9 GE |
941 | if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC) |
942 | hwtstamp = ptp_convert_timestamp(&hwtstamp, | |
943 | sk->sk_bind_phc); | |
944 | ||
007747a9 | 945 | if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) { |
d7c08826 YL |
946 | empty = 0; |
947 | ||
948 | if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) && | |
949 | !skb_is_err_queue(skb)) | |
97dc7cd9 | 950 | put_ts_pktinfo(msg, skb, if_index); |
d7c08826 | 951 | } |
aad9c8c4 | 952 | } |
1c885808 | 953 | if (!empty) { |
9718475e DD |
954 | if (sock_flag(sk, SOCK_TSTAMP_NEW)) |
955 | put_cmsg_scm_timestamping64(msg, &tss); | |
956 | else | |
957 | put_cmsg_scm_timestamping(msg, &tss); | |
1c885808 | 958 | |
8605330a | 959 | if (skb_is_err_queue(skb) && skb->len && |
4ef1b286 | 960 | SKB_EXT_ERR(skb)->opt_stats) |
1c885808 FY |
961 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS, |
962 | skb->len, skb->data); | |
963 | } | |
92f37fd2 | 964 | } |
7c81fd8b ACM |
965 | EXPORT_SYMBOL_GPL(__sock_recv_timestamp); |
966 | ||
eb6fba75 | 967 | #ifdef CONFIG_WIRELESS |
6e3e939f JB |
968 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, |
969 | struct sk_buff *skb) | |
970 | { | |
971 | int ack; | |
972 | ||
973 | if (!sock_flag(sk, SOCK_WIFI_STATUS)) | |
974 | return; | |
975 | if (!skb->wifi_acked_valid) | |
976 | return; | |
977 | ||
978 | ack = skb->wifi_acked; | |
979 | ||
980 | put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); | |
981 | } | |
982 | EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); | |
eb6fba75 | 983 | #endif |
6e3e939f | 984 | |
11165f14 | 985 | static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, |
986 | struct sk_buff *skb) | |
3b885787 | 987 | { |
744d5a3e | 988 | if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount) |
3b885787 | 989 | put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, |
744d5a3e | 990 | sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount); |
3b885787 NH |
991 | } |
992 | ||
6fd1d51c EM |
993 | static void sock_recv_mark(struct msghdr *msg, struct sock *sk, |
994 | struct sk_buff *skb) | |
995 | { | |
2558b803 ED |
996 | if (sock_flag(sk, SOCK_RCVMARK) && skb) { |
997 | /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */ | |
998 | __u32 mark = skb->mark; | |
999 | ||
1000 | put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark); | |
1001 | } | |
6fd1d51c EM |
1002 | } |
1003 | ||
1004 | void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk, | |
1005 | struct sk_buff *skb) | |
3b885787 NH |
1006 | { |
1007 | sock_recv_timestamp(msg, sk, skb); | |
1008 | sock_recv_drops(msg, sk, skb); | |
6fd1d51c | 1009 | sock_recv_mark(msg, sk, skb); |
3b885787 | 1010 | } |
6fd1d51c | 1011 | EXPORT_SYMBOL_GPL(__sock_recv_cmsgs); |
3b885787 | 1012 | |
8c3c447b | 1013 | INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *, |
a648a592 PA |
1014 | size_t, int)); |
1015 | INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *, | |
1016 | size_t, int)); | |
6e6eda44 YC |
1017 | |
1018 | static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags) | |
1019 | { | |
1020 | trace_sock_recv_length(sk, ret, flags); | |
1021 | } | |
1022 | ||
1b784140 | 1023 | static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, |
2da62906 | 1024 | int flags) |
1da177e4 | 1025 | { |
1ded5e5a ED |
1026 | int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg, |
1027 | inet6_recvmsg, | |
6e6eda44 YC |
1028 | inet_recvmsg, sock, msg, |
1029 | msg_data_left(msg), flags); | |
1030 | if (trace_sock_recv_length_enabled()) | |
1031 | call_trace_sock_recv_length(sock->sk, ret, flags); | |
1032 | return ret; | |
1da177e4 LT |
1033 | } |
1034 | ||
85806af0 RD |
1035 | /** |
1036 | * sock_recvmsg - receive a message from @sock | |
1037 | * @sock: socket | |
1038 | * @msg: message to receive | |
1039 | * @flags: message flags | |
1040 | * | |
1041 | * Receives @msg from @sock, passing through LSM. Returns the total number | |
1042 | * of bytes received, or an error. | |
1043 | */ | |
2da62906 | 1044 | int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) |
a2e27255 | 1045 | { |
2da62906 | 1046 | int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); |
a2e27255 | 1047 | |
2da62906 | 1048 | return err ?: sock_recvmsg_nosec(sock, msg, flags); |
1da177e4 | 1049 | } |
c6d409cf | 1050 | EXPORT_SYMBOL(sock_recvmsg); |
1da177e4 | 1051 | |
c1249c0a | 1052 | /** |
8a3c245c PT |
1053 | * kernel_recvmsg - Receive a message from a socket (kernel space) |
1054 | * @sock: The socket to receive the message from | |
1055 | * @msg: Received message | |
1056 | * @vec: Input s/g array for message data | |
1057 | * @num: Size of input s/g array | |
1058 | * @size: Number of bytes to read | |
1059 | * @flags: Message flags (MSG_DONTWAIT, etc...) | |
c1249c0a | 1060 | * |
8a3c245c PT |
1061 | * On return the msg structure contains the scatter/gather array passed in the |
1062 | * vec argument. The array is modified so that it consists of the unfilled | |
1063 | * portion of the original array. | |
c1249c0a | 1064 | * |
8a3c245c | 1065 | * The returned value is the total number of bytes received, or an error. |
c1249c0a | 1066 | */ |
8a3c245c | 1067 | |
89bddce5 SH |
1068 | int kernel_recvmsg(struct socket *sock, struct msghdr *msg, |
1069 | struct kvec *vec, size_t num, size_t size, int flags) | |
1da177e4 | 1070 | { |
1f466e1f | 1071 | msg->msg_control_is_user = false; |
de4eda9d | 1072 | iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size); |
1f466e1f | 1073 | return sock_recvmsg(sock, msg, flags); |
1da177e4 | 1074 | } |
c6d409cf | 1075 | EXPORT_SYMBOL(kernel_recvmsg); |
1da177e4 | 1076 | |
9c55e01c | 1077 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, |
c6d409cf | 1078 | struct pipe_inode_info *pipe, size_t len, |
9c55e01c JA |
1079 | unsigned int flags) |
1080 | { | |
1081 | struct socket *sock = file->private_data; | |
1ded5e5a | 1082 | const struct proto_ops *ops; |
9c55e01c | 1083 | |
1ded5e5a ED |
1084 | ops = READ_ONCE(sock->ops); |
1085 | if (unlikely(!ops->splice_read)) | |
67178fd0 | 1086 | return copy_splice_read(file, ppos, pipe, len, flags); |
997b37da | 1087 | |
1ded5e5a | 1088 | return ops->splice_read(sock, ppos, pipe, len, flags); |
9c55e01c JA |
1089 | } |
1090 | ||
2bfc6685 DH |
1091 | static void sock_splice_eof(struct file *file) |
1092 | { | |
1093 | struct socket *sock = file->private_data; | |
1ded5e5a | 1094 | const struct proto_ops *ops; |
2bfc6685 | 1095 | |
1ded5e5a ED |
1096 | ops = READ_ONCE(sock->ops); |
1097 | if (ops->splice_eof) | |
1098 | ops->splice_eof(sock); | |
2bfc6685 DH |
1099 | } |
1100 | ||
8ae5e030 | 1101 | static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) |
ce1d4d3e | 1102 | { |
6d652330 AV |
1103 | struct file *file = iocb->ki_filp; |
1104 | struct socket *sock = file->private_data; | |
0345f931 | 1105 | struct msghdr msg = {.msg_iter = *to, |
1106 | .msg_iocb = iocb}; | |
8ae5e030 | 1107 | ssize_t res; |
ce1d4d3e | 1108 | |
ebfcd895 | 1109 | if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) |
8ae5e030 AV |
1110 | msg.msg_flags = MSG_DONTWAIT; |
1111 | ||
1112 | if (iocb->ki_pos != 0) | |
1da177e4 | 1113 | return -ESPIPE; |
027445c3 | 1114 | |
66ee59af | 1115 | if (!iov_iter_count(to)) /* Match SYS5 behaviour */ |
1da177e4 LT |
1116 | return 0; |
1117 | ||
2da62906 | 1118 | res = sock_recvmsg(sock, &msg, msg.msg_flags); |
8ae5e030 AV |
1119 | *to = msg.msg_iter; |
1120 | return res; | |
1da177e4 LT |
1121 | } |
1122 | ||
8ae5e030 | 1123 | static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) |
ce1d4d3e | 1124 | { |
6d652330 AV |
1125 | struct file *file = iocb->ki_filp; |
1126 | struct socket *sock = file->private_data; | |
0345f931 | 1127 | struct msghdr msg = {.msg_iter = *from, |
1128 | .msg_iocb = iocb}; | |
8ae5e030 | 1129 | ssize_t res; |
1da177e4 | 1130 | |
8ae5e030 | 1131 | if (iocb->ki_pos != 0) |
ce1d4d3e | 1132 | return -ESPIPE; |
027445c3 | 1133 | |
ebfcd895 | 1134 | if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) |
8ae5e030 AV |
1135 | msg.msg_flags = MSG_DONTWAIT; |
1136 | ||
6d652330 AV |
1137 | if (sock->type == SOCK_SEQPACKET) |
1138 | msg.msg_flags |= MSG_EOR; | |
1139 | ||
d8725c86 | 1140 | res = sock_sendmsg(sock, &msg); |
8ae5e030 AV |
1141 | *from = msg.msg_iter; |
1142 | return res; | |
1da177e4 LT |
1143 | } |
1144 | ||
1da177e4 LT |
1145 | /* |
1146 | * Atomic setting of ioctl hooks to avoid race | |
1147 | * with module unload. | |
1148 | */ | |
1149 | ||
4a3e2f71 | 1150 | static DEFINE_MUTEX(br_ioctl_mutex); |
ad2f99ae AB |
1151 | static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br, |
1152 | unsigned int cmd, struct ifreq *ifr, | |
1153 | void __user *uarg); | |
1da177e4 | 1154 | |
ad2f99ae AB |
1155 | void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br, |
1156 | unsigned int cmd, struct ifreq *ifr, | |
1157 | void __user *uarg)) | |
1da177e4 | 1158 | { |
4a3e2f71 | 1159 | mutex_lock(&br_ioctl_mutex); |
1da177e4 | 1160 | br_ioctl_hook = hook; |
4a3e2f71 | 1161 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 LT |
1162 | } |
1163 | EXPORT_SYMBOL(brioctl_set); | |
1164 | ||
ad2f99ae AB |
1165 | int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd, |
1166 | struct ifreq *ifr, void __user *uarg) | |
1167 | { | |
1168 | int err = -ENOPKG; | |
1169 | ||
1170 | if (!br_ioctl_hook) | |
1171 | request_module("bridge"); | |
1172 | ||
1173 | mutex_lock(&br_ioctl_mutex); | |
1174 | if (br_ioctl_hook) | |
1175 | err = br_ioctl_hook(net, br, cmd, ifr, uarg); | |
1176 | mutex_unlock(&br_ioctl_mutex); | |
1177 | ||
1178 | return err; | |
1179 | } | |
1180 | ||
4a3e2f71 | 1181 | static DEFINE_MUTEX(vlan_ioctl_mutex); |
881d966b | 1182 | static int (*vlan_ioctl_hook) (struct net *, void __user *arg); |
1da177e4 | 1183 | |
881d966b | 1184 | void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) |
1da177e4 | 1185 | { |
4a3e2f71 | 1186 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 1187 | vlan_ioctl_hook = hook; |
4a3e2f71 | 1188 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 LT |
1189 | } |
1190 | EXPORT_SYMBOL(vlan_ioctl_set); | |
1191 | ||
6b96018b | 1192 | static long sock_do_ioctl(struct net *net, struct socket *sock, |
63ff03ab | 1193 | unsigned int cmd, unsigned long arg) |
6b96018b | 1194 | { |
1ded5e5a | 1195 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
876f0bf9 AB |
1196 | struct ifreq ifr; |
1197 | bool need_copyout; | |
6b96018b AB |
1198 | int err; |
1199 | void __user *argp = (void __user *)arg; | |
a554bf96 | 1200 | void __user *data; |
6b96018b | 1201 | |
1ded5e5a | 1202 | err = ops->ioctl(sock, cmd, arg); |
6b96018b AB |
1203 | |
1204 | /* | |
1205 | * If this ioctl is unknown try to hand it down | |
1206 | * to the NIC driver. | |
1207 | */ | |
36fd633e AV |
1208 | if (err != -ENOIOCTLCMD) |
1209 | return err; | |
6b96018b | 1210 | |
29ce8f97 JK |
1211 | if (!is_socket_ioctl_cmd(cmd)) |
1212 | return -ENOTTY; | |
1213 | ||
a554bf96 | 1214 | if (get_user_ifreq(&ifr, &data, argp)) |
876f0bf9 | 1215 | return -EFAULT; |
a554bf96 | 1216 | err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); |
876f0bf9 | 1217 | if (!err && need_copyout) |
a554bf96 | 1218 | if (put_user_ifreq(&ifr, argp)) |
44c02a2c | 1219 | return -EFAULT; |
876f0bf9 | 1220 | |
6b96018b AB |
1221 | return err; |
1222 | } | |
1223 | ||
1da177e4 LT |
1224 | /* |
1225 | * With an ioctl, arg may well be a user mode pointer, but we don't know | |
1226 | * what to do with it - that's up to the protocol still. | |
1227 | */ | |
1228 | ||
1229 | static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) | |
1230 | { | |
1ded5e5a | 1231 | const struct proto_ops *ops; |
1da177e4 | 1232 | struct socket *sock; |
881d966b | 1233 | struct sock *sk; |
1da177e4 LT |
1234 | void __user *argp = (void __user *)arg; |
1235 | int pid, err; | |
881d966b | 1236 | struct net *net; |
1da177e4 | 1237 | |
b69aee04 | 1238 | sock = file->private_data; |
1ded5e5a | 1239 | ops = READ_ONCE(sock->ops); |
881d966b | 1240 | sk = sock->sk; |
3b1e0a65 | 1241 | net = sock_net(sk); |
44c02a2c AV |
1242 | if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { |
1243 | struct ifreq ifr; | |
a554bf96 | 1244 | void __user *data; |
44c02a2c | 1245 | bool need_copyout; |
a554bf96 | 1246 | if (get_user_ifreq(&ifr, &data, argp)) |
44c02a2c | 1247 | return -EFAULT; |
a554bf96 | 1248 | err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); |
44c02a2c | 1249 | if (!err && need_copyout) |
a554bf96 | 1250 | if (put_user_ifreq(&ifr, argp)) |
44c02a2c | 1251 | return -EFAULT; |
1da177e4 | 1252 | } else |
3d23e349 | 1253 | #ifdef CONFIG_WEXT_CORE |
1da177e4 | 1254 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { |
b1b0c245 | 1255 | err = wext_handle_ioctl(net, cmd, argp); |
1da177e4 | 1256 | } else |
3d23e349 | 1257 | #endif |
89bddce5 | 1258 | switch (cmd) { |
1da177e4 LT |
1259 | case FIOSETOWN: |
1260 | case SIOCSPGRP: | |
1261 | err = -EFAULT; | |
1262 | if (get_user(pid, (int __user *)argp)) | |
1263 | break; | |
393cc3f5 | 1264 | err = f_setown(sock->file, pid, 1); |
1da177e4 LT |
1265 | break; |
1266 | case FIOGETOWN: | |
1267 | case SIOCGPGRP: | |
609d7fa9 | 1268 | err = put_user(f_getown(sock->file), |
89bddce5 | 1269 | (int __user *)argp); |
1da177e4 LT |
1270 | break; |
1271 | case SIOCGIFBR: | |
1272 | case SIOCSIFBR: | |
1273 | case SIOCBRADDBR: | |
1274 | case SIOCBRDELBR: | |
ad2f99ae | 1275 | err = br_ioctl_call(net, NULL, cmd, NULL, argp); |
1da177e4 LT |
1276 | break; |
1277 | case SIOCGIFVLAN: | |
1278 | case SIOCSIFVLAN: | |
1279 | err = -ENOPKG; | |
1280 | if (!vlan_ioctl_hook) | |
1281 | request_module("8021q"); | |
1282 | ||
4a3e2f71 | 1283 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 1284 | if (vlan_ioctl_hook) |
881d966b | 1285 | err = vlan_ioctl_hook(net, argp); |
4a3e2f71 | 1286 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 1287 | break; |
c62cce2c AV |
1288 | case SIOCGSKNS: |
1289 | err = -EPERM; | |
1290 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1291 | break; | |
1292 | ||
1293 | err = open_related_ns(&net->ns, get_net_ns); | |
1294 | break; | |
0768e170 AB |
1295 | case SIOCGSTAMP_OLD: |
1296 | case SIOCGSTAMPNS_OLD: | |
1ded5e5a | 1297 | if (!ops->gettstamp) { |
c7cbdbf2 AB |
1298 | err = -ENOIOCTLCMD; |
1299 | break; | |
1300 | } | |
1ded5e5a ED |
1301 | err = ops->gettstamp(sock, argp, |
1302 | cmd == SIOCGSTAMP_OLD, | |
1303 | !IS_ENABLED(CONFIG_64BIT)); | |
60747828 | 1304 | break; |
0768e170 AB |
1305 | case SIOCGSTAMP_NEW: |
1306 | case SIOCGSTAMPNS_NEW: | |
1ded5e5a | 1307 | if (!ops->gettstamp) { |
0768e170 AB |
1308 | err = -ENOIOCTLCMD; |
1309 | break; | |
1310 | } | |
1ded5e5a ED |
1311 | err = ops->gettstamp(sock, argp, |
1312 | cmd == SIOCGSTAMP_NEW, | |
1313 | false); | |
c7cbdbf2 | 1314 | break; |
876f0bf9 AB |
1315 | |
1316 | case SIOCGIFCONF: | |
1317 | err = dev_ifconf(net, argp); | |
1318 | break; | |
1319 | ||
1da177e4 | 1320 | default: |
63ff03ab | 1321 | err = sock_do_ioctl(net, sock, cmd, arg); |
1da177e4 | 1322 | break; |
89bddce5 | 1323 | } |
1da177e4 LT |
1324 | return err; |
1325 | } | |
1326 | ||
8a3c245c PT |
1327 | /** |
1328 | * sock_create_lite - creates a socket | |
1329 | * @family: protocol family (AF_INET, ...) | |
1330 | * @type: communication type (SOCK_STREAM, ...) | |
1331 | * @protocol: protocol (0, ...) | |
1332 | * @res: new socket | |
1333 | * | |
1334 | * Creates a new socket and assigns it to @res, passing through LSM. | |
1335 | * The new socket initialization is not complete, see kernel_accept(). | |
1336 | * Returns 0 or an error. On failure @res is set to %NULL. | |
1337 | * This function internally uses GFP_KERNEL. | |
1338 | */ | |
1339 | ||
1da177e4 LT |
1340 | int sock_create_lite(int family, int type, int protocol, struct socket **res) |
1341 | { | |
1342 | int err; | |
1343 | struct socket *sock = NULL; | |
89bddce5 | 1344 | |
1da177e4 LT |
1345 | err = security_socket_create(family, type, protocol, 1); |
1346 | if (err) | |
1347 | goto out; | |
1348 | ||
1349 | sock = sock_alloc(); | |
1350 | if (!sock) { | |
1351 | err = -ENOMEM; | |
1352 | goto out; | |
1353 | } | |
1354 | ||
1da177e4 | 1355 | sock->type = type; |
7420ed23 VY |
1356 | err = security_socket_post_create(sock, family, type, protocol, 1); |
1357 | if (err) | |
1358 | goto out_release; | |
1359 | ||
1da177e4 LT |
1360 | out: |
1361 | *res = sock; | |
1362 | return err; | |
7420ed23 VY |
1363 | out_release: |
1364 | sock_release(sock); | |
1365 | sock = NULL; | |
1366 | goto out; | |
1da177e4 | 1367 | } |
c6d409cf | 1368 | EXPORT_SYMBOL(sock_create_lite); |
1da177e4 LT |
1369 | |
1370 | /* No kernel lock held - perfect */ | |
ade994f4 | 1371 | static __poll_t sock_poll(struct file *file, poll_table *wait) |
1da177e4 | 1372 | { |
3cafb376 | 1373 | struct socket *sock = file->private_data; |
1ded5e5a | 1374 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
a331de3b | 1375 | __poll_t events = poll_requested_events(wait), flag = 0; |
2d48d67f | 1376 | |
1ded5e5a | 1377 | if (!ops->poll) |
e88958e6 | 1378 | return 0; |
f641f13b | 1379 | |
a331de3b CH |
1380 | if (sk_can_busy_loop(sock->sk)) { |
1381 | /* poll once if requested by the syscall */ | |
1382 | if (events & POLL_BUSY_LOOP) | |
1383 | sk_busy_loop(sock->sk, 1); | |
1384 | ||
1385 | /* if this socket can poll_ll, tell the system call */ | |
1386 | flag = POLL_BUSY_LOOP; | |
1387 | } | |
1388 | ||
1ded5e5a | 1389 | return ops->poll(file, sock, wait) | flag; |
1da177e4 LT |
1390 | } |
1391 | ||
89bddce5 | 1392 | static int sock_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 1393 | { |
b69aee04 | 1394 | struct socket *sock = file->private_data; |
1da177e4 | 1395 | |
1ded5e5a | 1396 | return READ_ONCE(sock->ops)->mmap(file, sock, vma); |
1da177e4 LT |
1397 | } |
1398 | ||
20380731 | 1399 | static int sock_close(struct inode *inode, struct file *filp) |
1da177e4 | 1400 | { |
6d8c50dc | 1401 | __sock_release(SOCKET_I(inode), inode); |
1da177e4 LT |
1402 | return 0; |
1403 | } | |
1404 | ||
1405 | /* | |
1406 | * Update the socket async list | |
1407 | * | |
1408 | * Fasync_list locking strategy. | |
1409 | * | |
1410 | * 1. fasync_list is modified only under process context socket lock | |
1411 | * i.e. under semaphore. | |
1412 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) | |
989a2979 | 1413 | * or under socket lock |
1da177e4 LT |
1414 | */ |
1415 | ||
1416 | static int sock_fasync(int fd, struct file *filp, int on) | |
1417 | { | |
989a2979 ED |
1418 | struct socket *sock = filp->private_data; |
1419 | struct sock *sk = sock->sk; | |
333f7909 | 1420 | struct socket_wq *wq = &sock->wq; |
1da177e4 | 1421 | |
989a2979 | 1422 | if (sk == NULL) |
1da177e4 | 1423 | return -EINVAL; |
1da177e4 LT |
1424 | |
1425 | lock_sock(sk); | |
eaefd110 | 1426 | fasync_helper(fd, filp, on, &wq->fasync_list); |
1da177e4 | 1427 | |
eaefd110 | 1428 | if (!wq->fasync_list) |
989a2979 ED |
1429 | sock_reset_flag(sk, SOCK_FASYNC); |
1430 | else | |
bcdce719 | 1431 | sock_set_flag(sk, SOCK_FASYNC); |
1da177e4 | 1432 | |
989a2979 | 1433 | release_sock(sk); |
1da177e4 LT |
1434 | return 0; |
1435 | } | |
1436 | ||
ceb5d58b | 1437 | /* This function may be called only under rcu_lock */ |
1da177e4 | 1438 | |
ceb5d58b | 1439 | int sock_wake_async(struct socket_wq *wq, int how, int band) |
1da177e4 | 1440 | { |
ceb5d58b | 1441 | if (!wq || !wq->fasync_list) |
1da177e4 | 1442 | return -1; |
ceb5d58b | 1443 | |
89bddce5 | 1444 | switch (how) { |
8d8ad9d7 | 1445 | case SOCK_WAKE_WAITD: |
ceb5d58b | 1446 | if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) |
1da177e4 LT |
1447 | break; |
1448 | goto call_kill; | |
8d8ad9d7 | 1449 | case SOCK_WAKE_SPACE: |
ceb5d58b | 1450 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) |
1da177e4 | 1451 | break; |
7c7ab580 | 1452 | fallthrough; |
8d8ad9d7 | 1453 | case SOCK_WAKE_IO: |
89bddce5 | 1454 | call_kill: |
43815482 | 1455 | kill_fasync(&wq->fasync_list, SIGIO, band); |
1da177e4 | 1456 | break; |
8d8ad9d7 | 1457 | case SOCK_WAKE_URG: |
43815482 | 1458 | kill_fasync(&wq->fasync_list, SIGURG, band); |
1da177e4 | 1459 | } |
ceb5d58b | 1460 | |
1da177e4 LT |
1461 | return 0; |
1462 | } | |
c6d409cf | 1463 | EXPORT_SYMBOL(sock_wake_async); |
1da177e4 | 1464 | |
8a3c245c PT |
1465 | /** |
1466 | * __sock_create - creates a socket | |
1467 | * @net: net namespace | |
1468 | * @family: protocol family (AF_INET, ...) | |
1469 | * @type: communication type (SOCK_STREAM, ...) | |
1470 | * @protocol: protocol (0, ...) | |
1471 | * @res: new socket | |
1472 | * @kern: boolean for kernel space sockets | |
1473 | * | |
1474 | * Creates a new socket and assigns it to @res, passing through LSM. | |
1475 | * Returns 0 or an error. On failure @res is set to %NULL. @kern must | |
1476 | * be set to true if the socket resides in kernel space. | |
1477 | * This function internally uses GFP_KERNEL. | |
1478 | */ | |
1479 | ||
721db93a | 1480 | int __sock_create(struct net *net, int family, int type, int protocol, |
89bddce5 | 1481 | struct socket **res, int kern) |
1da177e4 LT |
1482 | { |
1483 | int err; | |
1484 | struct socket *sock; | |
55737fda | 1485 | const struct net_proto_family *pf; |
1da177e4 LT |
1486 | |
1487 | /* | |
89bddce5 | 1488 | * Check protocol is in range |
1da177e4 LT |
1489 | */ |
1490 | if (family < 0 || family >= NPROTO) | |
1491 | return -EAFNOSUPPORT; | |
1492 | if (type < 0 || type >= SOCK_MAX) | |
1493 | return -EINVAL; | |
1494 | ||
1495 | /* Compatibility. | |
1496 | ||
1497 | This uglymoron is moved from INET layer to here to avoid | |
1498 | deadlock in module load. | |
1499 | */ | |
1500 | if (family == PF_INET && type == SOCK_PACKET) { | |
f3c98690 | 1501 | pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", |
1502 | current->comm); | |
1da177e4 LT |
1503 | family = PF_PACKET; |
1504 | } | |
1505 | ||
1506 | err = security_socket_create(family, type, protocol, kern); | |
1507 | if (err) | |
1508 | return err; | |
89bddce5 | 1509 | |
55737fda SH |
1510 | /* |
1511 | * Allocate the socket and allow the family to set things up. if | |
1512 | * the protocol is 0, the family is instructed to select an appropriate | |
1513 | * default. | |
1514 | */ | |
1515 | sock = sock_alloc(); | |
1516 | if (!sock) { | |
e87cc472 | 1517 | net_warn_ratelimited("socket: no more sockets\n"); |
55737fda SH |
1518 | return -ENFILE; /* Not exactly a match, but its the |
1519 | closest posix thing */ | |
1520 | } | |
1521 | ||
1522 | sock->type = type; | |
1523 | ||
95a5afca | 1524 | #ifdef CONFIG_MODULES |
89bddce5 SH |
1525 | /* Attempt to load a protocol module if the find failed. |
1526 | * | |
1527 | * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user | |
1da177e4 LT |
1528 | * requested real, full-featured networking support upon configuration. |
1529 | * Otherwise module support will break! | |
1530 | */ | |
190683a9 | 1531 | if (rcu_access_pointer(net_families[family]) == NULL) |
89bddce5 | 1532 | request_module("net-pf-%d", family); |
1da177e4 LT |
1533 | #endif |
1534 | ||
55737fda SH |
1535 | rcu_read_lock(); |
1536 | pf = rcu_dereference(net_families[family]); | |
1537 | err = -EAFNOSUPPORT; | |
1538 | if (!pf) | |
1539 | goto out_release; | |
1da177e4 LT |
1540 | |
1541 | /* | |
1542 | * We will call the ->create function, that possibly is in a loadable | |
1543 | * module, so we have to bump that loadable module refcnt first. | |
1544 | */ | |
55737fda | 1545 | if (!try_module_get(pf->owner)) |
1da177e4 LT |
1546 | goto out_release; |
1547 | ||
55737fda SH |
1548 | /* Now protected by module ref count */ |
1549 | rcu_read_unlock(); | |
1550 | ||
3f378b68 | 1551 | err = pf->create(net, sock, protocol, kern); |
55737fda | 1552 | if (err < 0) |
1da177e4 | 1553 | goto out_module_put; |
a79af59e | 1554 | |
1da177e4 LT |
1555 | /* |
1556 | * Now to bump the refcnt of the [loadable] module that owns this | |
1557 | * socket at sock_release time we decrement its refcnt. | |
1558 | */ | |
55737fda SH |
1559 | if (!try_module_get(sock->ops->owner)) |
1560 | goto out_module_busy; | |
1561 | ||
1da177e4 LT |
1562 | /* |
1563 | * Now that we're done with the ->create function, the [loadable] | |
1564 | * module can have its refcnt decremented | |
1565 | */ | |
55737fda | 1566 | module_put(pf->owner); |
7420ed23 VY |
1567 | err = security_socket_post_create(sock, family, type, protocol, kern); |
1568 | if (err) | |
3b185525 | 1569 | goto out_sock_release; |
55737fda | 1570 | *res = sock; |
1da177e4 | 1571 | |
55737fda SH |
1572 | return 0; |
1573 | ||
1574 | out_module_busy: | |
1575 | err = -EAFNOSUPPORT; | |
1da177e4 | 1576 | out_module_put: |
55737fda SH |
1577 | sock->ops = NULL; |
1578 | module_put(pf->owner); | |
1579 | out_sock_release: | |
1da177e4 | 1580 | sock_release(sock); |
55737fda SH |
1581 | return err; |
1582 | ||
1583 | out_release: | |
1584 | rcu_read_unlock(); | |
1585 | goto out_sock_release; | |
1da177e4 | 1586 | } |
721db93a | 1587 | EXPORT_SYMBOL(__sock_create); |
1da177e4 | 1588 | |
8a3c245c PT |
1589 | /** |
1590 | * sock_create - creates a socket | |
1591 | * @family: protocol family (AF_INET, ...) | |
1592 | * @type: communication type (SOCK_STREAM, ...) | |
1593 | * @protocol: protocol (0, ...) | |
1594 | * @res: new socket | |
1595 | * | |
1596 | * A wrapper around __sock_create(). | |
1597 | * Returns 0 or an error. This function internally uses GFP_KERNEL. | |
1598 | */ | |
1599 | ||
1da177e4 LT |
1600 | int sock_create(int family, int type, int protocol, struct socket **res) |
1601 | { | |
1b8d7ae4 | 1602 | return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); |
1da177e4 | 1603 | } |
c6d409cf | 1604 | EXPORT_SYMBOL(sock_create); |
1da177e4 | 1605 | |
8a3c245c PT |
1606 | /** |
1607 | * sock_create_kern - creates a socket (kernel space) | |
1608 | * @net: net namespace | |
1609 | * @family: protocol family (AF_INET, ...) | |
1610 | * @type: communication type (SOCK_STREAM, ...) | |
1611 | * @protocol: protocol (0, ...) | |
1612 | * @res: new socket | |
1613 | * | |
1614 | * A wrapper around __sock_create(). | |
1615 | * Returns 0 or an error. This function internally uses GFP_KERNEL. | |
1616 | */ | |
1617 | ||
eeb1bd5c | 1618 | int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) |
1da177e4 | 1619 | { |
eeb1bd5c | 1620 | return __sock_create(net, family, type, protocol, res, 1); |
1da177e4 | 1621 | } |
c6d409cf | 1622 | EXPORT_SYMBOL(sock_create_kern); |
1da177e4 | 1623 | |
da214a47 | 1624 | static struct socket *__sys_socket_create(int family, int type, int protocol) |
1da177e4 | 1625 | { |
1da177e4 | 1626 | struct socket *sock; |
da214a47 | 1627 | int retval; |
a677a039 | 1628 | |
e38b36f3 UD |
1629 | /* Check the SOCK_* constants for consistency. */ |
1630 | BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); | |
1631 | BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); | |
1632 | BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); | |
1633 | BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); | |
1634 | ||
da214a47 JA |
1635 | if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
1636 | return ERR_PTR(-EINVAL); | |
a677a039 | 1637 | type &= SOCK_TYPE_MASK; |
1da177e4 | 1638 | |
da214a47 JA |
1639 | retval = sock_create(family, type, protocol, &sock); |
1640 | if (retval < 0) | |
1641 | return ERR_PTR(retval); | |
1642 | ||
1643 | return sock; | |
1644 | } | |
1645 | ||
1646 | struct file *__sys_socket_file(int family, int type, int protocol) | |
1647 | { | |
1648 | struct socket *sock; | |
da214a47 JA |
1649 | int flags; |
1650 | ||
1651 | sock = __sys_socket_create(family, type, protocol); | |
1652 | if (IS_ERR(sock)) | |
1653 | return ERR_CAST(sock); | |
1654 | ||
1655 | flags = type & ~SOCK_TYPE_MASK; | |
aaca0bdc UD |
1656 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1657 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1658 | ||
649c15c7 | 1659 | return sock_alloc_file(sock, flags, NULL); |
da214a47 JA |
1660 | } |
1661 | ||
0dd061a6 GT |
1662 | /* A hook for bpf progs to attach to and update socket protocol. |
1663 | * | |
1664 | * A static noinline declaration here could cause the compiler to | |
1665 | * optimize away the function. A global noinline declaration will | |
1666 | * keep the definition, but may optimize away the callsite. | |
1667 | * Therefore, __weak is needed to ensure that the call is still | |
1668 | * emitted, by telling the compiler that we don't know what the | |
1669 | * function might eventually be. | |
1670 | * | |
1671 | * __diag_* below are needed to dismiss the missing prototype warning. | |
1672 | */ | |
1673 | ||
1674 | __diag_push(); | |
1675 | __diag_ignore_all("-Wmissing-prototypes", | |
1676 | "A fmod_ret entry point for BPF programs"); | |
1677 | ||
1678 | __weak noinline int update_socket_protocol(int family, int type, int protocol) | |
1679 | { | |
1680 | return protocol; | |
1681 | } | |
1682 | ||
1683 | __diag_pop(); | |
1684 | ||
da214a47 JA |
1685 | int __sys_socket(int family, int type, int protocol) |
1686 | { | |
1687 | struct socket *sock; | |
1688 | int flags; | |
1689 | ||
0dd061a6 GT |
1690 | sock = __sys_socket_create(family, type, |
1691 | update_socket_protocol(family, type, protocol)); | |
da214a47 JA |
1692 | if (IS_ERR(sock)) |
1693 | return PTR_ERR(sock); | |
1694 | ||
1695 | flags = type & ~SOCK_TYPE_MASK; | |
1696 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) | |
1697 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1da177e4 | 1698 | |
8e1611e2 | 1699 | return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); |
1da177e4 LT |
1700 | } |
1701 | ||
9d6a15c3 DB |
1702 | SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) |
1703 | { | |
1704 | return __sys_socket(family, type, protocol); | |
1705 | } | |
1706 | ||
1da177e4 LT |
1707 | /* |
1708 | * Create a pair of connected sockets. | |
1709 | */ | |
1710 | ||
6debc8d8 | 1711 | int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) |
1da177e4 LT |
1712 | { |
1713 | struct socket *sock1, *sock2; | |
1714 | int fd1, fd2, err; | |
db349509 | 1715 | struct file *newfile1, *newfile2; |
a677a039 UD |
1716 | int flags; |
1717 | ||
1718 | flags = type & ~SOCK_TYPE_MASK; | |
77d27200 | 1719 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
a677a039 UD |
1720 | return -EINVAL; |
1721 | type &= SOCK_TYPE_MASK; | |
1da177e4 | 1722 | |
aaca0bdc UD |
1723 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1724 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1725 | ||
016a266b AV |
1726 | /* |
1727 | * reserve descriptors and make sure we won't fail | |
1728 | * to return them to userland. | |
1729 | */ | |
1730 | fd1 = get_unused_fd_flags(flags); | |
1731 | if (unlikely(fd1 < 0)) | |
1732 | return fd1; | |
1733 | ||
1734 | fd2 = get_unused_fd_flags(flags); | |
1735 | if (unlikely(fd2 < 0)) { | |
1736 | put_unused_fd(fd1); | |
1737 | return fd2; | |
1738 | } | |
1739 | ||
1740 | err = put_user(fd1, &usockvec[0]); | |
1741 | if (err) | |
1742 | goto out; | |
1743 | ||
1744 | err = put_user(fd2, &usockvec[1]); | |
1745 | if (err) | |
1746 | goto out; | |
1747 | ||
1da177e4 LT |
1748 | /* |
1749 | * Obtain the first socket and check if the underlying protocol | |
1750 | * supports the socketpair call. | |
1751 | */ | |
1752 | ||
1753 | err = sock_create(family, type, protocol, &sock1); | |
016a266b | 1754 | if (unlikely(err < 0)) |
1da177e4 LT |
1755 | goto out; |
1756 | ||
1757 | err = sock_create(family, type, protocol, &sock2); | |
016a266b AV |
1758 | if (unlikely(err < 0)) { |
1759 | sock_release(sock1); | |
1760 | goto out; | |
bf3c23d1 | 1761 | } |
d73aa286 | 1762 | |
d47cd945 DH |
1763 | err = security_socket_socketpair(sock1, sock2); |
1764 | if (unlikely(err)) { | |
1765 | sock_release(sock2); | |
1766 | sock_release(sock1); | |
1767 | goto out; | |
1768 | } | |
1769 | ||
1ded5e5a | 1770 | err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2); |
016a266b AV |
1771 | if (unlikely(err < 0)) { |
1772 | sock_release(sock2); | |
1773 | sock_release(sock1); | |
1774 | goto out; | |
28407630 AV |
1775 | } |
1776 | ||
aab174f0 | 1777 | newfile1 = sock_alloc_file(sock1, flags, NULL); |
b5ffe634 | 1778 | if (IS_ERR(newfile1)) { |
28407630 | 1779 | err = PTR_ERR(newfile1); |
016a266b AV |
1780 | sock_release(sock2); |
1781 | goto out; | |
28407630 AV |
1782 | } |
1783 | ||
aab174f0 | 1784 | newfile2 = sock_alloc_file(sock2, flags, NULL); |
28407630 AV |
1785 | if (IS_ERR(newfile2)) { |
1786 | err = PTR_ERR(newfile2); | |
016a266b AV |
1787 | fput(newfile1); |
1788 | goto out; | |
db349509 AV |
1789 | } |
1790 | ||
157cf649 | 1791 | audit_fd_pair(fd1, fd2); |
d73aa286 | 1792 | |
db349509 AV |
1793 | fd_install(fd1, newfile1); |
1794 | fd_install(fd2, newfile2); | |
d73aa286 | 1795 | return 0; |
1da177e4 | 1796 | |
016a266b | 1797 | out: |
d73aa286 | 1798 | put_unused_fd(fd2); |
d73aa286 | 1799 | put_unused_fd(fd1); |
1da177e4 LT |
1800 | return err; |
1801 | } | |
1802 | ||
6debc8d8 DB |
1803 | SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, |
1804 | int __user *, usockvec) | |
1805 | { | |
1806 | return __sys_socketpair(family, type, protocol, usockvec); | |
1807 | } | |
1808 | ||
1da177e4 LT |
1809 | /* |
1810 | * Bind a name to a socket. Nothing much to do here since it's | |
1811 | * the protocol's responsibility to handle the local address. | |
1812 | * | |
1813 | * We move the socket address to kernel space before we call | |
1814 | * the protocol layer (having also checked the address is ok). | |
1815 | */ | |
1816 | ||
a87d35d8 | 1817 | int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) |
1da177e4 LT |
1818 | { |
1819 | struct socket *sock; | |
230b1839 | 1820 | struct sockaddr_storage address; |
6cb153ca | 1821 | int err, fput_needed; |
1da177e4 | 1822 | |
89bddce5 | 1823 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
e71a4783 | 1824 | if (sock) { |
43db362d | 1825 | err = move_addr_to_kernel(umyaddr, addrlen, &address); |
068b88cc | 1826 | if (!err) { |
89bddce5 | 1827 | err = security_socket_bind(sock, |
230b1839 | 1828 | (struct sockaddr *)&address, |
89bddce5 | 1829 | addrlen); |
6cb153ca | 1830 | if (!err) |
1ded5e5a | 1831 | err = READ_ONCE(sock->ops)->bind(sock, |
89bddce5 | 1832 | (struct sockaddr *) |
230b1839 | 1833 | &address, addrlen); |
1da177e4 | 1834 | } |
6cb153ca | 1835 | fput_light(sock->file, fput_needed); |
89bddce5 | 1836 | } |
1da177e4 LT |
1837 | return err; |
1838 | } | |
1839 | ||
a87d35d8 DB |
1840 | SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) |
1841 | { | |
1842 | return __sys_bind(fd, umyaddr, addrlen); | |
1843 | } | |
1844 | ||
1da177e4 LT |
1845 | /* |
1846 | * Perform a listen. Basically, we allow the protocol to do anything | |
1847 | * necessary for a listen, and if that works, we mark the socket as | |
1848 | * ready for listening. | |
1849 | */ | |
1850 | ||
25e290ee | 1851 | int __sys_listen(int fd, int backlog) |
1da177e4 LT |
1852 | { |
1853 | struct socket *sock; | |
6cb153ca | 1854 | int err, fput_needed; |
b8e1f9b5 | 1855 | int somaxconn; |
89bddce5 SH |
1856 | |
1857 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1858 | if (sock) { | |
3c9ba81d | 1859 | somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn); |
95c96174 | 1860 | if ((unsigned int)backlog > somaxconn) |
b8e1f9b5 | 1861 | backlog = somaxconn; |
1da177e4 LT |
1862 | |
1863 | err = security_socket_listen(sock, backlog); | |
6cb153ca | 1864 | if (!err) |
1ded5e5a | 1865 | err = READ_ONCE(sock->ops)->listen(sock, backlog); |
1da177e4 | 1866 | |
6cb153ca | 1867 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1868 | } |
1869 | return err; | |
1870 | } | |
1871 | ||
25e290ee DB |
1872 | SYSCALL_DEFINE2(listen, int, fd, int, backlog) |
1873 | { | |
1874 | return __sys_listen(fd, backlog); | |
1875 | } | |
1876 | ||
d32f89da | 1877 | struct file *do_accept(struct file *file, unsigned file_flags, |
de2ea4b6 | 1878 | struct sockaddr __user *upeer_sockaddr, |
d32f89da | 1879 | int __user *upeer_addrlen, int flags) |
1da177e4 LT |
1880 | { |
1881 | struct socket *sock, *newsock; | |
39d8c1b6 | 1882 | struct file *newfile; |
d32f89da | 1883 | int err, len; |
230b1839 | 1884 | struct sockaddr_storage address; |
1ded5e5a | 1885 | const struct proto_ops *ops; |
1da177e4 | 1886 | |
dba4a925 | 1887 | sock = sock_from_file(file); |
d32f89da PB |
1888 | if (!sock) |
1889 | return ERR_PTR(-ENOTSOCK); | |
1da177e4 | 1890 | |
c6d409cf ED |
1891 | newsock = sock_alloc(); |
1892 | if (!newsock) | |
d32f89da | 1893 | return ERR_PTR(-ENFILE); |
1ded5e5a | 1894 | ops = READ_ONCE(sock->ops); |
1da177e4 LT |
1895 | |
1896 | newsock->type = sock->type; | |
1ded5e5a | 1897 | newsock->ops = ops; |
1da177e4 | 1898 | |
1da177e4 LT |
1899 | /* |
1900 | * We don't need try_module_get here, as the listening socket (sock) | |
1901 | * has the protocol module (sock->ops->owner) held. | |
1902 | */ | |
1ded5e5a | 1903 | __module_get(ops->owner); |
1da177e4 | 1904 | |
aab174f0 | 1905 | newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); |
d32f89da PB |
1906 | if (IS_ERR(newfile)) |
1907 | return newfile; | |
39d8c1b6 | 1908 | |
a79af59e FF |
1909 | err = security_socket_accept(sock, newsock); |
1910 | if (err) | |
39d8c1b6 | 1911 | goto out_fd; |
a79af59e | 1912 | |
1ded5e5a | 1913 | err = ops->accept(sock, newsock, sock->file->f_flags | file_flags, |
de2ea4b6 | 1914 | false); |
1da177e4 | 1915 | if (err < 0) |
39d8c1b6 | 1916 | goto out_fd; |
1da177e4 LT |
1917 | |
1918 | if (upeer_sockaddr) { | |
1ded5e5a | 1919 | len = ops->getname(newsock, (struct sockaddr *)&address, 2); |
9b2c45d4 | 1920 | if (len < 0) { |
1da177e4 | 1921 | err = -ECONNABORTED; |
39d8c1b6 | 1922 | goto out_fd; |
1da177e4 | 1923 | } |
43db362d | 1924 | err = move_addr_to_user(&address, |
230b1839 | 1925 | len, upeer_sockaddr, upeer_addrlen); |
1da177e4 | 1926 | if (err < 0) |
39d8c1b6 | 1927 | goto out_fd; |
1da177e4 LT |
1928 | } |
1929 | ||
1930 | /* File flags are not inherited via accept() unlike another OSes. */ | |
d32f89da | 1931 | return newfile; |
39d8c1b6 | 1932 | out_fd: |
9606a216 | 1933 | fput(newfile); |
d32f89da PB |
1934 | return ERR_PTR(err); |
1935 | } | |
1936 | ||
c0424532 YD |
1937 | static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr, |
1938 | int __user *upeer_addrlen, int flags) | |
d32f89da PB |
1939 | { |
1940 | struct file *newfile; | |
1941 | int newfd; | |
1942 | ||
1943 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) | |
1944 | return -EINVAL; | |
1945 | ||
1946 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) | |
1947 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
de2ea4b6 | 1948 | |
c0424532 | 1949 | newfd = get_unused_fd_flags(flags); |
d32f89da PB |
1950 | if (unlikely(newfd < 0)) |
1951 | return newfd; | |
1952 | ||
c0424532 | 1953 | newfile = do_accept(file, 0, upeer_sockaddr, upeer_addrlen, |
d32f89da PB |
1954 | flags); |
1955 | if (IS_ERR(newfile)) { | |
1956 | put_unused_fd(newfd); | |
1957 | return PTR_ERR(newfile); | |
1958 | } | |
1959 | fd_install(newfd, newfile); | |
1960 | return newfd; | |
de2ea4b6 JA |
1961 | } |
1962 | ||
1963 | /* | |
1964 | * For accept, we attempt to create a new socket, set up the link | |
1965 | * with the client, wake up the client, then return the new | |
1966 | * connected fd. We collect the address of the connector in kernel | |
1967 | * space and move it to user at the very end. This is unclean because | |
1968 | * we open the socket then return an error. | |
1969 | * | |
1970 | * 1003.1g adds the ability to recvmsg() to query connection pending | |
1971 | * status to recvmsg. We need to add that support in a way thats | |
1972 | * clean when we restructure accept also. | |
1973 | */ | |
1974 | ||
1975 | int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, | |
1976 | int __user *upeer_addrlen, int flags) | |
1977 | { | |
1978 | int ret = -EBADF; | |
1979 | struct fd f; | |
1980 | ||
1981 | f = fdget(fd); | |
1982 | if (f.file) { | |
c0424532 YD |
1983 | ret = __sys_accept4_file(f.file, upeer_sockaddr, |
1984 | upeer_addrlen, flags); | |
6b07edeb | 1985 | fdput(f); |
de2ea4b6 JA |
1986 | } |
1987 | ||
1988 | return ret; | |
1da177e4 LT |
1989 | } |
1990 | ||
4541e805 DB |
1991 | SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, |
1992 | int __user *, upeer_addrlen, int, flags) | |
1993 | { | |
1994 | return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); | |
1995 | } | |
1996 | ||
20f37034 HC |
1997 | SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, |
1998 | int __user *, upeer_addrlen) | |
aaca0bdc | 1999 | { |
4541e805 | 2000 | return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); |
aaca0bdc UD |
2001 | } |
2002 | ||
1da177e4 LT |
2003 | /* |
2004 | * Attempt to connect to a socket with the server address. The address | |
2005 | * is in user space so we verify it is OK and move it to kernel space. | |
2006 | * | |
2007 | * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to | |
2008 | * break bindings | |
2009 | * | |
2010 | * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and | |
2011 | * other SEQPACKET protocols that take time to connect() as it doesn't | |
2012 | * include the -EINPROGRESS status for such sockets. | |
2013 | */ | |
2014 | ||
f499a021 | 2015 | int __sys_connect_file(struct file *file, struct sockaddr_storage *address, |
bd3ded31 | 2016 | int addrlen, int file_flags) |
1da177e4 LT |
2017 | { |
2018 | struct socket *sock; | |
bd3ded31 | 2019 | int err; |
1da177e4 | 2020 | |
dba4a925 FR |
2021 | sock = sock_from_file(file); |
2022 | if (!sock) { | |
2023 | err = -ENOTSOCK; | |
1da177e4 | 2024 | goto out; |
dba4a925 | 2025 | } |
1da177e4 | 2026 | |
89bddce5 | 2027 | err = |
f499a021 | 2028 | security_socket_connect(sock, (struct sockaddr *)address, addrlen); |
1da177e4 | 2029 | if (err) |
bd3ded31 | 2030 | goto out; |
1da177e4 | 2031 | |
1ded5e5a ED |
2032 | err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address, |
2033 | addrlen, sock->file->f_flags | file_flags); | |
1da177e4 LT |
2034 | out: |
2035 | return err; | |
2036 | } | |
2037 | ||
bd3ded31 JA |
2038 | int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) |
2039 | { | |
2040 | int ret = -EBADF; | |
2041 | struct fd f; | |
2042 | ||
2043 | f = fdget(fd); | |
2044 | if (f.file) { | |
f499a021 JA |
2045 | struct sockaddr_storage address; |
2046 | ||
2047 | ret = move_addr_to_kernel(uservaddr, addrlen, &address); | |
2048 | if (!ret) | |
2049 | ret = __sys_connect_file(f.file, &address, addrlen, 0); | |
6b07edeb | 2050 | fdput(f); |
bd3ded31 JA |
2051 | } |
2052 | ||
2053 | return ret; | |
2054 | } | |
2055 | ||
1387c2c2 DB |
2056 | SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, |
2057 | int, addrlen) | |
2058 | { | |
2059 | return __sys_connect(fd, uservaddr, addrlen); | |
2060 | } | |
2061 | ||
1da177e4 LT |
2062 | /* |
2063 | * Get the local address ('name') of a socket object. Move the obtained | |
2064 | * name to user space. | |
2065 | */ | |
2066 | ||
8882a107 DB |
2067 | int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, |
2068 | int __user *usockaddr_len) | |
1da177e4 LT |
2069 | { |
2070 | struct socket *sock; | |
230b1839 | 2071 | struct sockaddr_storage address; |
9b2c45d4 | 2072 | int err, fput_needed; |
89bddce5 | 2073 | |
6cb153ca | 2074 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
2075 | if (!sock) |
2076 | goto out; | |
2077 | ||
2078 | err = security_socket_getsockname(sock); | |
2079 | if (err) | |
2080 | goto out_put; | |
2081 | ||
1ded5e5a | 2082 | err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0); |
9b2c45d4 | 2083 | if (err < 0) |
1da177e4 | 2084 | goto out_put; |
e44ef1d4 | 2085 | /* "err" is actually length in this case */ |
9b2c45d4 | 2086 | err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); |
1da177e4 LT |
2087 | |
2088 | out_put: | |
6cb153ca | 2089 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2090 | out: |
2091 | return err; | |
2092 | } | |
2093 | ||
8882a107 DB |
2094 | SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, |
2095 | int __user *, usockaddr_len) | |
2096 | { | |
2097 | return __sys_getsockname(fd, usockaddr, usockaddr_len); | |
2098 | } | |
2099 | ||
1da177e4 LT |
2100 | /* |
2101 | * Get the remote address ('name') of a socket object. Move the obtained | |
2102 | * name to user space. | |
2103 | */ | |
2104 | ||
b21c8f83 DB |
2105 | int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, |
2106 | int __user *usockaddr_len) | |
1da177e4 LT |
2107 | { |
2108 | struct socket *sock; | |
230b1839 | 2109 | struct sockaddr_storage address; |
9b2c45d4 | 2110 | int err, fput_needed; |
1da177e4 | 2111 | |
89bddce5 SH |
2112 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2113 | if (sock != NULL) { | |
1ded5e5a ED |
2114 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
2115 | ||
1da177e4 LT |
2116 | err = security_socket_getpeername(sock); |
2117 | if (err) { | |
6cb153ca | 2118 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2119 | return err; |
2120 | } | |
2121 | ||
1ded5e5a | 2122 | err = ops->getname(sock, (struct sockaddr *)&address, 1); |
9b2c45d4 DV |
2123 | if (err >= 0) |
2124 | /* "err" is actually length in this case */ | |
2125 | err = move_addr_to_user(&address, err, usockaddr, | |
89bddce5 | 2126 | usockaddr_len); |
6cb153ca | 2127 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2128 | } |
2129 | return err; | |
2130 | } | |
2131 | ||
b21c8f83 DB |
2132 | SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, |
2133 | int __user *, usockaddr_len) | |
2134 | { | |
2135 | return __sys_getpeername(fd, usockaddr, usockaddr_len); | |
2136 | } | |
2137 | ||
1da177e4 LT |
2138 | /* |
2139 | * Send a datagram to a given address. We move the address into kernel | |
2140 | * space and check the user space data area is readable before invoking | |
2141 | * the protocol. | |
2142 | */ | |
211b634b DB |
2143 | int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, |
2144 | struct sockaddr __user *addr, int addr_len) | |
1da177e4 LT |
2145 | { |
2146 | struct socket *sock; | |
230b1839 | 2147 | struct sockaddr_storage address; |
1da177e4 LT |
2148 | int err; |
2149 | struct msghdr msg; | |
2150 | struct iovec iov; | |
6cb153ca | 2151 | int fput_needed; |
6cb153ca | 2152 | |
de4eda9d | 2153 | err = import_single_range(ITER_SOURCE, buff, len, &iov, &msg.msg_iter); |
602bd0e9 AV |
2154 | if (unlikely(err)) |
2155 | return err; | |
de0fa95c PE |
2156 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2157 | if (!sock) | |
4387ff75 | 2158 | goto out; |
6cb153ca | 2159 | |
89bddce5 | 2160 | msg.msg_name = NULL; |
89bddce5 SH |
2161 | msg.msg_control = NULL; |
2162 | msg.msg_controllen = 0; | |
2163 | msg.msg_namelen = 0; | |
7c701d92 | 2164 | msg.msg_ubuf = NULL; |
6cb153ca | 2165 | if (addr) { |
43db362d | 2166 | err = move_addr_to_kernel(addr, addr_len, &address); |
1da177e4 LT |
2167 | if (err < 0) |
2168 | goto out_put; | |
230b1839 | 2169 | msg.msg_name = (struct sockaddr *)&address; |
89bddce5 | 2170 | msg.msg_namelen = addr_len; |
1da177e4 | 2171 | } |
b841b901 | 2172 | flags &= ~MSG_INTERNAL_SENDMSG_FLAGS; |
1da177e4 LT |
2173 | if (sock->file->f_flags & O_NONBLOCK) |
2174 | flags |= MSG_DONTWAIT; | |
2175 | msg.msg_flags = flags; | |
d8725c86 | 2176 | err = sock_sendmsg(sock, &msg); |
1da177e4 | 2177 | |
89bddce5 | 2178 | out_put: |
de0fa95c | 2179 | fput_light(sock->file, fput_needed); |
4387ff75 | 2180 | out: |
1da177e4 LT |
2181 | return err; |
2182 | } | |
2183 | ||
211b634b DB |
2184 | SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, |
2185 | unsigned int, flags, struct sockaddr __user *, addr, | |
2186 | int, addr_len) | |
2187 | { | |
2188 | return __sys_sendto(fd, buff, len, flags, addr, addr_len); | |
2189 | } | |
2190 | ||
1da177e4 | 2191 | /* |
89bddce5 | 2192 | * Send a datagram down a socket. |
1da177e4 LT |
2193 | */ |
2194 | ||
3e0fa65f | 2195 | SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, |
95c96174 | 2196 | unsigned int, flags) |
1da177e4 | 2197 | { |
211b634b | 2198 | return __sys_sendto(fd, buff, len, flags, NULL, 0); |
1da177e4 LT |
2199 | } |
2200 | ||
2201 | /* | |
89bddce5 | 2202 | * Receive a frame from the socket and optionally record the address of the |
1da177e4 LT |
2203 | * sender. We verify the buffers are writable and if needed move the |
2204 | * sender address from kernel to user space. | |
2205 | */ | |
7a09e1eb DB |
2206 | int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, |
2207 | struct sockaddr __user *addr, int __user *addr_len) | |
1da177e4 | 2208 | { |
1228b34c ED |
2209 | struct sockaddr_storage address; |
2210 | struct msghdr msg = { | |
2211 | /* Save some cycles and don't copy the address if not needed */ | |
2212 | .msg_name = addr ? (struct sockaddr *)&address : NULL, | |
2213 | }; | |
1da177e4 LT |
2214 | struct socket *sock; |
2215 | struct iovec iov; | |
89bddce5 | 2216 | int err, err2; |
6cb153ca BL |
2217 | int fput_needed; |
2218 | ||
de4eda9d | 2219 | err = import_single_range(ITER_DEST, ubuf, size, &iov, &msg.msg_iter); |
602bd0e9 AV |
2220 | if (unlikely(err)) |
2221 | return err; | |
de0fa95c | 2222 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 | 2223 | if (!sock) |
de0fa95c | 2224 | goto out; |
1da177e4 | 2225 | |
1da177e4 LT |
2226 | if (sock->file->f_flags & O_NONBLOCK) |
2227 | flags |= MSG_DONTWAIT; | |
2da62906 | 2228 | err = sock_recvmsg(sock, &msg, flags); |
1da177e4 | 2229 | |
89bddce5 | 2230 | if (err >= 0 && addr != NULL) { |
43db362d | 2231 | err2 = move_addr_to_user(&address, |
230b1839 | 2232 | msg.msg_namelen, addr, addr_len); |
89bddce5 SH |
2233 | if (err2 < 0) |
2234 | err = err2; | |
1da177e4 | 2235 | } |
de0fa95c PE |
2236 | |
2237 | fput_light(sock->file, fput_needed); | |
4387ff75 | 2238 | out: |
1da177e4 LT |
2239 | return err; |
2240 | } | |
2241 | ||
7a09e1eb DB |
2242 | SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, |
2243 | unsigned int, flags, struct sockaddr __user *, addr, | |
2244 | int __user *, addr_len) | |
2245 | { | |
2246 | return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); | |
2247 | } | |
2248 | ||
1da177e4 | 2249 | /* |
89bddce5 | 2250 | * Receive a datagram from a socket. |
1da177e4 LT |
2251 | */ |
2252 | ||
b7c0ddf5 JG |
2253 | SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, |
2254 | unsigned int, flags) | |
1da177e4 | 2255 | { |
7a09e1eb | 2256 | return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); |
1da177e4 LT |
2257 | } |
2258 | ||
83f0c10b FW |
2259 | static bool sock_use_custom_sol_socket(const struct socket *sock) |
2260 | { | |
a5ef058d | 2261 | return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags); |
83f0c10b FW |
2262 | } |
2263 | ||
1da177e4 LT |
2264 | /* |
2265 | * Set a socket option. Because we don't know the option lengths we have | |
2266 | * to pass the user mode parameter for the protocols to sort out. | |
2267 | */ | |
a7b75c5a | 2268 | int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval, |
55db9c0e | 2269 | int optlen) |
1da177e4 | 2270 | { |
519a8a6c | 2271 | sockptr_t optval = USER_SOCKPTR(user_optval); |
1ded5e5a | 2272 | const struct proto_ops *ops; |
0d01da6a | 2273 | char *kernel_optval = NULL; |
6cb153ca | 2274 | int err, fput_needed; |
1da177e4 LT |
2275 | struct socket *sock; |
2276 | ||
2277 | if (optlen < 0) | |
2278 | return -EINVAL; | |
89bddce5 SH |
2279 | |
2280 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
4a367299 CH |
2281 | if (!sock) |
2282 | return err; | |
1da177e4 | 2283 | |
4a367299 CH |
2284 | err = security_socket_setsockopt(sock, level, optname); |
2285 | if (err) | |
2286 | goto out_put; | |
0d01da6a | 2287 | |
55db9c0e CH |
2288 | if (!in_compat_syscall()) |
2289 | err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname, | |
a7b75c5a | 2290 | user_optval, &optlen, |
55db9c0e | 2291 | &kernel_optval); |
4a367299 CH |
2292 | if (err < 0) |
2293 | goto out_put; | |
2294 | if (err > 0) { | |
2295 | err = 0; | |
2296 | goto out_put; | |
2297 | } | |
0d01da6a | 2298 | |
a7b75c5a CH |
2299 | if (kernel_optval) |
2300 | optval = KERNEL_SOCKPTR(kernel_optval); | |
1ded5e5a | 2301 | ops = READ_ONCE(sock->ops); |
4a367299 | 2302 | if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock)) |
a7b75c5a | 2303 | err = sock_setsockopt(sock, level, optname, optval, optlen); |
1ded5e5a | 2304 | else if (unlikely(!ops->setsockopt)) |
a44d9e72 | 2305 | err = -EOPNOTSUPP; |
4a367299 | 2306 | else |
1ded5e5a | 2307 | err = ops->setsockopt(sock, level, optname, optval, |
89bddce5 | 2308 | optlen); |
a7b75c5a | 2309 | kfree(kernel_optval); |
4a367299 CH |
2310 | out_put: |
2311 | fput_light(sock->file, fput_needed); | |
1da177e4 LT |
2312 | return err; |
2313 | } | |
2314 | ||
cc36dca0 DB |
2315 | SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, |
2316 | char __user *, optval, int, optlen) | |
2317 | { | |
2318 | return __sys_setsockopt(fd, level, optname, optval, optlen); | |
2319 | } | |
2320 | ||
9cacf81f SF |
2321 | INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level, |
2322 | int optname)); | |
2323 | ||
1da177e4 LT |
2324 | /* |
2325 | * Get a socket option. Because we don't know the option lengths we have | |
2326 | * to pass a user mode parameter for the protocols to sort out. | |
2327 | */ | |
55db9c0e CH |
2328 | int __sys_getsockopt(int fd, int level, int optname, char __user *optval, |
2329 | int __user *optlen) | |
1da177e4 | 2330 | { |
ad4bf5f2 | 2331 | int max_optlen __maybe_unused; |
1ded5e5a | 2332 | const struct proto_ops *ops; |
6cb153ca | 2333 | int err, fput_needed; |
1da177e4 LT |
2334 | struct socket *sock; |
2335 | ||
89bddce5 | 2336 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
d8a9b38f CH |
2337 | if (!sock) |
2338 | return err; | |
2339 | ||
2340 | err = security_socket_getsockopt(sock, level, optname); | |
2341 | if (err) | |
2342 | goto out_put; | |
1da177e4 | 2343 | |
55db9c0e CH |
2344 | if (!in_compat_syscall()) |
2345 | max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen); | |
0d01da6a | 2346 | |
1ded5e5a | 2347 | ops = READ_ONCE(sock->ops); |
d8a9b38f CH |
2348 | if (level == SOL_SOCKET) |
2349 | err = sock_getsockopt(sock, level, optname, optval, optlen); | |
1ded5e5a | 2350 | else if (unlikely(!ops->getsockopt)) |
a44d9e72 | 2351 | err = -EOPNOTSUPP; |
d8a9b38f | 2352 | else |
1ded5e5a | 2353 | err = ops->getsockopt(sock, level, optname, optval, |
89bddce5 | 2354 | optlen); |
0d01da6a | 2355 | |
55db9c0e CH |
2356 | if (!in_compat_syscall()) |
2357 | err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname, | |
2358 | optval, optlen, max_optlen, | |
2359 | err); | |
6cb153ca | 2360 | out_put: |
d8a9b38f | 2361 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2362 | return err; |
2363 | } | |
2364 | ||
13a2d70e DB |
2365 | SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, |
2366 | char __user *, optval, int __user *, optlen) | |
2367 | { | |
2368 | return __sys_getsockopt(fd, level, optname, optval, optlen); | |
2369 | } | |
2370 | ||
1da177e4 LT |
2371 | /* |
2372 | * Shutdown a socket. | |
2373 | */ | |
2374 | ||
b713c195 JA |
2375 | int __sys_shutdown_sock(struct socket *sock, int how) |
2376 | { | |
2377 | int err; | |
2378 | ||
2379 | err = security_socket_shutdown(sock, how); | |
2380 | if (!err) | |
1ded5e5a | 2381 | err = READ_ONCE(sock->ops)->shutdown(sock, how); |
b713c195 JA |
2382 | |
2383 | return err; | |
2384 | } | |
2385 | ||
005a1aea | 2386 | int __sys_shutdown(int fd, int how) |
1da177e4 | 2387 | { |
6cb153ca | 2388 | int err, fput_needed; |
1da177e4 LT |
2389 | struct socket *sock; |
2390 | ||
89bddce5 SH |
2391 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2392 | if (sock != NULL) { | |
b713c195 | 2393 | err = __sys_shutdown_sock(sock, how); |
6cb153ca | 2394 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2395 | } |
2396 | return err; | |
2397 | } | |
2398 | ||
005a1aea DB |
2399 | SYSCALL_DEFINE2(shutdown, int, fd, int, how) |
2400 | { | |
2401 | return __sys_shutdown(fd, how); | |
2402 | } | |
2403 | ||
89bddce5 | 2404 | /* A couple of helpful macros for getting the address of the 32/64 bit |
1da177e4 LT |
2405 | * fields which are the same type (int / unsigned) on our platforms. |
2406 | */ | |
2407 | #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) | |
2408 | #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) | |
2409 | #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) | |
2410 | ||
c71d8ebe TH |
2411 | struct used_address { |
2412 | struct sockaddr_storage name; | |
2413 | unsigned int name_len; | |
2414 | }; | |
2415 | ||
7fa875b8 DY |
2416 | int __copy_msghdr(struct msghdr *kmsg, |
2417 | struct user_msghdr *msg, | |
2418 | struct sockaddr __user **save_addr) | |
1661bf36 | 2419 | { |
08adb7da AV |
2420 | ssize_t err; |
2421 | ||
1f466e1f | 2422 | kmsg->msg_control_is_user = true; |
1228b34c | 2423 | kmsg->msg_get_inq = 0; |
7fa875b8 DY |
2424 | kmsg->msg_control_user = msg->msg_control; |
2425 | kmsg->msg_controllen = msg->msg_controllen; | |
2426 | kmsg->msg_flags = msg->msg_flags; | |
ffb07550 | 2427 | |
7fa875b8 DY |
2428 | kmsg->msg_namelen = msg->msg_namelen; |
2429 | if (!msg->msg_name) | |
6a2a2b3a AS |
2430 | kmsg->msg_namelen = 0; |
2431 | ||
dbb490b9 ML |
2432 | if (kmsg->msg_namelen < 0) |
2433 | return -EINVAL; | |
2434 | ||
1661bf36 | 2435 | if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) |
db31c55a | 2436 | kmsg->msg_namelen = sizeof(struct sockaddr_storage); |
08adb7da AV |
2437 | |
2438 | if (save_addr) | |
7fa875b8 | 2439 | *save_addr = msg->msg_name; |
08adb7da | 2440 | |
7fa875b8 | 2441 | if (msg->msg_name && kmsg->msg_namelen) { |
08adb7da | 2442 | if (!save_addr) { |
7fa875b8 | 2443 | err = move_addr_to_kernel(msg->msg_name, |
864d9664 | 2444 | kmsg->msg_namelen, |
08adb7da AV |
2445 | kmsg->msg_name); |
2446 | if (err < 0) | |
2447 | return err; | |
2448 | } | |
2449 | } else { | |
2450 | kmsg->msg_name = NULL; | |
2451 | kmsg->msg_namelen = 0; | |
2452 | } | |
2453 | ||
7fa875b8 | 2454 | if (msg->msg_iovlen > UIO_MAXIOV) |
08adb7da AV |
2455 | return -EMSGSIZE; |
2456 | ||
0345f931 | 2457 | kmsg->msg_iocb = NULL; |
7c701d92 | 2458 | kmsg->msg_ubuf = NULL; |
0a384abf JA |
2459 | return 0; |
2460 | } | |
2461 | ||
2462 | static int copy_msghdr_from_user(struct msghdr *kmsg, | |
2463 | struct user_msghdr __user *umsg, | |
2464 | struct sockaddr __user **save_addr, | |
2465 | struct iovec **iov) | |
2466 | { | |
2467 | struct user_msghdr msg; | |
2468 | ssize_t err; | |
2469 | ||
7fa875b8 DY |
2470 | if (copy_from_user(&msg, umsg, sizeof(*umsg))) |
2471 | return -EFAULT; | |
2472 | ||
2473 | err = __copy_msghdr(kmsg, &msg, save_addr); | |
0a384abf JA |
2474 | if (err) |
2475 | return err; | |
0345f931 | 2476 | |
de4eda9d | 2477 | err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE, |
ffb07550 | 2478 | msg.msg_iov, msg.msg_iovlen, |
da184284 | 2479 | UIO_FASTIOV, iov, &kmsg->msg_iter); |
87e5e6da | 2480 | return err < 0 ? err : 0; |
1661bf36 DC |
2481 | } |
2482 | ||
4257c8ca JA |
2483 | static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, |
2484 | unsigned int flags, struct used_address *used_address, | |
2485 | unsigned int allowed_msghdr_flags) | |
1da177e4 | 2486 | { |
b9d717a7 | 2487 | unsigned char ctl[sizeof(struct cmsghdr) + 20] |
846cc123 | 2488 | __aligned(sizeof(__kernel_size_t)); |
89bddce5 | 2489 | /* 20 is size of ipv6_pktinfo */ |
1da177e4 | 2490 | unsigned char *ctl_buf = ctl; |
d8725c86 | 2491 | int ctl_len; |
08adb7da | 2492 | ssize_t err; |
89bddce5 | 2493 | |
1da177e4 LT |
2494 | err = -ENOBUFS; |
2495 | ||
228e548e | 2496 | if (msg_sys->msg_controllen > INT_MAX) |
4257c8ca | 2497 | goto out; |
28a94d8f | 2498 | flags |= (msg_sys->msg_flags & allowed_msghdr_flags); |
228e548e | 2499 | ctl_len = msg_sys->msg_controllen; |
1da177e4 | 2500 | if ((MSG_CMSG_COMPAT & flags) && ctl_len) { |
89bddce5 | 2501 | err = |
228e548e | 2502 | cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, |
89bddce5 | 2503 | sizeof(ctl)); |
1da177e4 | 2504 | if (err) |
4257c8ca | 2505 | goto out; |
228e548e AB |
2506 | ctl_buf = msg_sys->msg_control; |
2507 | ctl_len = msg_sys->msg_controllen; | |
1da177e4 | 2508 | } else if (ctl_len) { |
ac4340fc DM |
2509 | BUILD_BUG_ON(sizeof(struct cmsghdr) != |
2510 | CMSG_ALIGN(sizeof(struct cmsghdr))); | |
89bddce5 | 2511 | if (ctl_len > sizeof(ctl)) { |
1da177e4 | 2512 | ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); |
89bddce5 | 2513 | if (ctl_buf == NULL) |
4257c8ca | 2514 | goto out; |
1da177e4 LT |
2515 | } |
2516 | err = -EFAULT; | |
1f466e1f | 2517 | if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len)) |
1da177e4 | 2518 | goto out_freectl; |
228e548e | 2519 | msg_sys->msg_control = ctl_buf; |
1f466e1f | 2520 | msg_sys->msg_control_is_user = false; |
1da177e4 | 2521 | } |
b841b901 | 2522 | flags &= ~MSG_INTERNAL_SENDMSG_FLAGS; |
228e548e | 2523 | msg_sys->msg_flags = flags; |
1da177e4 LT |
2524 | |
2525 | if (sock->file->f_flags & O_NONBLOCK) | |
228e548e | 2526 | msg_sys->msg_flags |= MSG_DONTWAIT; |
c71d8ebe TH |
2527 | /* |
2528 | * If this is sendmmsg() and current destination address is same as | |
2529 | * previously succeeded address, omit asking LSM's decision. | |
2530 | * used_address->name_len is initialized to UINT_MAX so that the first | |
2531 | * destination address never matches. | |
2532 | */ | |
bc909d9d MD |
2533 | if (used_address && msg_sys->msg_name && |
2534 | used_address->name_len == msg_sys->msg_namelen && | |
2535 | !memcmp(&used_address->name, msg_sys->msg_name, | |
c71d8ebe | 2536 | used_address->name_len)) { |
d8725c86 | 2537 | err = sock_sendmsg_nosec(sock, msg_sys); |
c71d8ebe TH |
2538 | goto out_freectl; |
2539 | } | |
d8725c86 | 2540 | err = sock_sendmsg(sock, msg_sys); |
c71d8ebe TH |
2541 | /* |
2542 | * If this is sendmmsg() and sending to current destination address was | |
2543 | * successful, remember it. | |
2544 | */ | |
2545 | if (used_address && err >= 0) { | |
2546 | used_address->name_len = msg_sys->msg_namelen; | |
bc909d9d MD |
2547 | if (msg_sys->msg_name) |
2548 | memcpy(&used_address->name, msg_sys->msg_name, | |
2549 | used_address->name_len); | |
c71d8ebe | 2550 | } |
1da177e4 LT |
2551 | |
2552 | out_freectl: | |
89bddce5 | 2553 | if (ctl_buf != ctl) |
1da177e4 | 2554 | sock_kfree_s(sock->sk, ctl_buf, ctl_len); |
4257c8ca JA |
2555 | out: |
2556 | return err; | |
2557 | } | |
2558 | ||
03b1230c JA |
2559 | int sendmsg_copy_msghdr(struct msghdr *msg, |
2560 | struct user_msghdr __user *umsg, unsigned flags, | |
2561 | struct iovec **iov) | |
4257c8ca JA |
2562 | { |
2563 | int err; | |
2564 | ||
2565 | if (flags & MSG_CMSG_COMPAT) { | |
2566 | struct compat_msghdr __user *msg_compat; | |
2567 | ||
2568 | msg_compat = (struct compat_msghdr __user *) umsg; | |
2569 | err = get_compat_msghdr(msg, msg_compat, NULL, iov); | |
2570 | } else { | |
2571 | err = copy_msghdr_from_user(msg, umsg, NULL, iov); | |
2572 | } | |
2573 | if (err < 0) | |
2574 | return err; | |
2575 | ||
2576 | return 0; | |
2577 | } | |
2578 | ||
2579 | static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, | |
2580 | struct msghdr *msg_sys, unsigned int flags, | |
2581 | struct used_address *used_address, | |
2582 | unsigned int allowed_msghdr_flags) | |
2583 | { | |
2584 | struct sockaddr_storage address; | |
2585 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; | |
2586 | ssize_t err; | |
2587 | ||
2588 | msg_sys->msg_name = &address; | |
2589 | ||
2590 | err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov); | |
2591 | if (err < 0) | |
2592 | return err; | |
2593 | ||
2594 | err = ____sys_sendmsg(sock, msg_sys, flags, used_address, | |
2595 | allowed_msghdr_flags); | |
da184284 | 2596 | kfree(iov); |
228e548e AB |
2597 | return err; |
2598 | } | |
2599 | ||
2600 | /* | |
2601 | * BSD sendmsg interface | |
2602 | */ | |
03b1230c | 2603 | long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg, |
0fa03c62 JA |
2604 | unsigned int flags) |
2605 | { | |
03b1230c | 2606 | return ____sys_sendmsg(sock, msg, flags, NULL, 0); |
0fa03c62 | 2607 | } |
228e548e | 2608 | |
e1834a32 DB |
2609 | long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, |
2610 | bool forbid_cmsg_compat) | |
228e548e AB |
2611 | { |
2612 | int fput_needed, err; | |
2613 | struct msghdr msg_sys; | |
1be374a0 AL |
2614 | struct socket *sock; |
2615 | ||
e1834a32 DB |
2616 | if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) |
2617 | return -EINVAL; | |
2618 | ||
1be374a0 | 2619 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
228e548e AB |
2620 | if (!sock) |
2621 | goto out; | |
2622 | ||
28a94d8f | 2623 | err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); |
228e548e | 2624 | |
6cb153ca | 2625 | fput_light(sock->file, fput_needed); |
89bddce5 | 2626 | out: |
1da177e4 LT |
2627 | return err; |
2628 | } | |
2629 | ||
666547ff | 2630 | SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) |
a7526eb5 | 2631 | { |
e1834a32 | 2632 | return __sys_sendmsg(fd, msg, flags, true); |
a7526eb5 AL |
2633 | } |
2634 | ||
228e548e AB |
2635 | /* |
2636 | * Linux sendmmsg interface | |
2637 | */ | |
2638 | ||
2639 | int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, | |
e1834a32 | 2640 | unsigned int flags, bool forbid_cmsg_compat) |
228e548e AB |
2641 | { |
2642 | int fput_needed, err, datagrams; | |
2643 | struct socket *sock; | |
2644 | struct mmsghdr __user *entry; | |
2645 | struct compat_mmsghdr __user *compat_entry; | |
2646 | struct msghdr msg_sys; | |
c71d8ebe | 2647 | struct used_address used_address; |
f092276d | 2648 | unsigned int oflags = flags; |
228e548e | 2649 | |
e1834a32 DB |
2650 | if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) |
2651 | return -EINVAL; | |
2652 | ||
98382f41 AB |
2653 | if (vlen > UIO_MAXIOV) |
2654 | vlen = UIO_MAXIOV; | |
228e548e AB |
2655 | |
2656 | datagrams = 0; | |
2657 | ||
2658 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
2659 | if (!sock) | |
2660 | return err; | |
2661 | ||
c71d8ebe | 2662 | used_address.name_len = UINT_MAX; |
228e548e AB |
2663 | entry = mmsg; |
2664 | compat_entry = (struct compat_mmsghdr __user *)mmsg; | |
728ffb86 | 2665 | err = 0; |
f092276d | 2666 | flags |= MSG_BATCH; |
228e548e AB |
2667 | |
2668 | while (datagrams < vlen) { | |
f092276d TH |
2669 | if (datagrams == vlen - 1) |
2670 | flags = oflags; | |
2671 | ||
228e548e | 2672 | if (MSG_CMSG_COMPAT & flags) { |
666547ff | 2673 | err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, |
28a94d8f | 2674 | &msg_sys, flags, &used_address, MSG_EOR); |
228e548e AB |
2675 | if (err < 0) |
2676 | break; | |
2677 | err = __put_user(err, &compat_entry->msg_len); | |
2678 | ++compat_entry; | |
2679 | } else { | |
a7526eb5 | 2680 | err = ___sys_sendmsg(sock, |
666547ff | 2681 | (struct user_msghdr __user *)entry, |
28a94d8f | 2682 | &msg_sys, flags, &used_address, MSG_EOR); |
228e548e AB |
2683 | if (err < 0) |
2684 | break; | |
2685 | err = put_user(err, &entry->msg_len); | |
2686 | ++entry; | |
2687 | } | |
2688 | ||
2689 | if (err) | |
2690 | break; | |
2691 | ++datagrams; | |
3023898b SHY |
2692 | if (msg_data_left(&msg_sys)) |
2693 | break; | |
a78cb84c | 2694 | cond_resched(); |
228e548e AB |
2695 | } |
2696 | ||
228e548e AB |
2697 | fput_light(sock->file, fput_needed); |
2698 | ||
728ffb86 AB |
2699 | /* We only return an error if no datagrams were able to be sent */ |
2700 | if (datagrams != 0) | |
228e548e AB |
2701 | return datagrams; |
2702 | ||
228e548e AB |
2703 | return err; |
2704 | } | |
2705 | ||
2706 | SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, | |
2707 | unsigned int, vlen, unsigned int, flags) | |
2708 | { | |
e1834a32 | 2709 | return __sys_sendmmsg(fd, mmsg, vlen, flags, true); |
228e548e AB |
2710 | } |
2711 | ||
03b1230c JA |
2712 | int recvmsg_copy_msghdr(struct msghdr *msg, |
2713 | struct user_msghdr __user *umsg, unsigned flags, | |
2714 | struct sockaddr __user **uaddr, | |
2715 | struct iovec **iov) | |
1da177e4 | 2716 | { |
08adb7da | 2717 | ssize_t err; |
1da177e4 | 2718 | |
4257c8ca JA |
2719 | if (MSG_CMSG_COMPAT & flags) { |
2720 | struct compat_msghdr __user *msg_compat; | |
1da177e4 | 2721 | |
4257c8ca JA |
2722 | msg_compat = (struct compat_msghdr __user *) umsg; |
2723 | err = get_compat_msghdr(msg, msg_compat, uaddr, iov); | |
2724 | } else { | |
2725 | err = copy_msghdr_from_user(msg, umsg, uaddr, iov); | |
2726 | } | |
1da177e4 | 2727 | if (err < 0) |
da184284 | 2728 | return err; |
1da177e4 | 2729 | |
4257c8ca JA |
2730 | return 0; |
2731 | } | |
2732 | ||
2733 | static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys, | |
2734 | struct user_msghdr __user *msg, | |
2735 | struct sockaddr __user *uaddr, | |
2736 | unsigned int flags, int nosec) | |
2737 | { | |
2738 | struct compat_msghdr __user *msg_compat = | |
2739 | (struct compat_msghdr __user *) msg; | |
2740 | int __user *uaddr_len = COMPAT_NAMELEN(msg); | |
2741 | struct sockaddr_storage addr; | |
2742 | unsigned long cmsg_ptr; | |
2743 | int len; | |
2744 | ssize_t err; | |
2745 | ||
2746 | msg_sys->msg_name = &addr; | |
a2e27255 ACM |
2747 | cmsg_ptr = (unsigned long)msg_sys->msg_control; |
2748 | msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); | |
89bddce5 | 2749 | |
f3d33426 HFS |
2750 | /* We assume all kernel code knows the size of sockaddr_storage */ |
2751 | msg_sys->msg_namelen = 0; | |
2752 | ||
1da177e4 LT |
2753 | if (sock->file->f_flags & O_NONBLOCK) |
2754 | flags |= MSG_DONTWAIT; | |
1af66221 ED |
2755 | |
2756 | if (unlikely(nosec)) | |
2757 | err = sock_recvmsg_nosec(sock, msg_sys, flags); | |
2758 | else | |
2759 | err = sock_recvmsg(sock, msg_sys, flags); | |
2760 | ||
1da177e4 | 2761 | if (err < 0) |
4257c8ca | 2762 | goto out; |
1da177e4 LT |
2763 | len = err; |
2764 | ||
2765 | if (uaddr != NULL) { | |
43db362d | 2766 | err = move_addr_to_user(&addr, |
a2e27255 | 2767 | msg_sys->msg_namelen, uaddr, |
89bddce5 | 2768 | uaddr_len); |
1da177e4 | 2769 | if (err < 0) |
4257c8ca | 2770 | goto out; |
1da177e4 | 2771 | } |
a2e27255 | 2772 | err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), |
37f7f421 | 2773 | COMPAT_FLAGS(msg)); |
1da177e4 | 2774 | if (err) |
4257c8ca | 2775 | goto out; |
1da177e4 | 2776 | if (MSG_CMSG_COMPAT & flags) |
a2e27255 | 2777 | err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, |
1da177e4 LT |
2778 | &msg_compat->msg_controllen); |
2779 | else | |
a2e27255 | 2780 | err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, |
1da177e4 LT |
2781 | &msg->msg_controllen); |
2782 | if (err) | |
4257c8ca | 2783 | goto out; |
1da177e4 | 2784 | err = len; |
4257c8ca JA |
2785 | out: |
2786 | return err; | |
2787 | } | |
2788 | ||
2789 | static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, | |
2790 | struct msghdr *msg_sys, unsigned int flags, int nosec) | |
2791 | { | |
2792 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; | |
2793 | /* user mode address pointers */ | |
2794 | struct sockaddr __user *uaddr; | |
2795 | ssize_t err; | |
2796 | ||
2797 | err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov); | |
2798 | if (err < 0) | |
2799 | return err; | |
1da177e4 | 2800 | |
4257c8ca | 2801 | err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec); |
da184284 | 2802 | kfree(iov); |
a2e27255 ACM |
2803 | return err; |
2804 | } | |
2805 | ||
2806 | /* | |
2807 | * BSD recvmsg interface | |
2808 | */ | |
2809 | ||
03b1230c JA |
2810 | long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg, |
2811 | struct user_msghdr __user *umsg, | |
2812 | struct sockaddr __user *uaddr, unsigned int flags) | |
aa1fa28f | 2813 | { |
03b1230c | 2814 | return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0); |
aa1fa28f JA |
2815 | } |
2816 | ||
e1834a32 DB |
2817 | long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, |
2818 | bool forbid_cmsg_compat) | |
a2e27255 ACM |
2819 | { |
2820 | int fput_needed, err; | |
2821 | struct msghdr msg_sys; | |
1be374a0 AL |
2822 | struct socket *sock; |
2823 | ||
e1834a32 DB |
2824 | if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) |
2825 | return -EINVAL; | |
2826 | ||
1be374a0 | 2827 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
a2e27255 ACM |
2828 | if (!sock) |
2829 | goto out; | |
2830 | ||
a7526eb5 | 2831 | err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); |
a2e27255 | 2832 | |
6cb153ca | 2833 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2834 | out: |
2835 | return err; | |
2836 | } | |
2837 | ||
666547ff | 2838 | SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, |
a7526eb5 AL |
2839 | unsigned int, flags) |
2840 | { | |
e1834a32 | 2841 | return __sys_recvmsg(fd, msg, flags, true); |
a7526eb5 AL |
2842 | } |
2843 | ||
a2e27255 ACM |
2844 | /* |
2845 | * Linux recvmmsg interface | |
2846 | */ | |
2847 | ||
e11d4284 AB |
2848 | static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg, |
2849 | unsigned int vlen, unsigned int flags, | |
2850 | struct timespec64 *timeout) | |
a2e27255 ACM |
2851 | { |
2852 | int fput_needed, err, datagrams; | |
2853 | struct socket *sock; | |
2854 | struct mmsghdr __user *entry; | |
d7256d0e | 2855 | struct compat_mmsghdr __user *compat_entry; |
a2e27255 | 2856 | struct msghdr msg_sys; |
766b9f92 DD |
2857 | struct timespec64 end_time; |
2858 | struct timespec64 timeout64; | |
a2e27255 ACM |
2859 | |
2860 | if (timeout && | |
2861 | poll_select_set_timeout(&end_time, timeout->tv_sec, | |
2862 | timeout->tv_nsec)) | |
2863 | return -EINVAL; | |
2864 | ||
2865 | datagrams = 0; | |
2866 | ||
2867 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
2868 | if (!sock) | |
2869 | return err; | |
2870 | ||
7797dc41 SHY |
2871 | if (likely(!(flags & MSG_ERRQUEUE))) { |
2872 | err = sock_error(sock->sk); | |
2873 | if (err) { | |
2874 | datagrams = err; | |
2875 | goto out_put; | |
2876 | } | |
e623a9e9 | 2877 | } |
a2e27255 ACM |
2878 | |
2879 | entry = mmsg; | |
d7256d0e | 2880 | compat_entry = (struct compat_mmsghdr __user *)mmsg; |
a2e27255 ACM |
2881 | |
2882 | while (datagrams < vlen) { | |
2883 | /* | |
2884 | * No need to ask LSM for more than the first datagram. | |
2885 | */ | |
d7256d0e | 2886 | if (MSG_CMSG_COMPAT & flags) { |
666547ff | 2887 | err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, |
a7526eb5 AL |
2888 | &msg_sys, flags & ~MSG_WAITFORONE, |
2889 | datagrams); | |
d7256d0e JMG |
2890 | if (err < 0) |
2891 | break; | |
2892 | err = __put_user(err, &compat_entry->msg_len); | |
2893 | ++compat_entry; | |
2894 | } else { | |
a7526eb5 | 2895 | err = ___sys_recvmsg(sock, |
666547ff | 2896 | (struct user_msghdr __user *)entry, |
a7526eb5 AL |
2897 | &msg_sys, flags & ~MSG_WAITFORONE, |
2898 | datagrams); | |
d7256d0e JMG |
2899 | if (err < 0) |
2900 | break; | |
2901 | err = put_user(err, &entry->msg_len); | |
2902 | ++entry; | |
2903 | } | |
2904 | ||
a2e27255 ACM |
2905 | if (err) |
2906 | break; | |
a2e27255 ACM |
2907 | ++datagrams; |
2908 | ||
71c5c159 BB |
2909 | /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ |
2910 | if (flags & MSG_WAITFORONE) | |
2911 | flags |= MSG_DONTWAIT; | |
2912 | ||
a2e27255 | 2913 | if (timeout) { |
766b9f92 | 2914 | ktime_get_ts64(&timeout64); |
c2e6c856 | 2915 | *timeout = timespec64_sub(end_time, timeout64); |
a2e27255 ACM |
2916 | if (timeout->tv_sec < 0) { |
2917 | timeout->tv_sec = timeout->tv_nsec = 0; | |
2918 | break; | |
2919 | } | |
2920 | ||
2921 | /* Timeout, return less than vlen datagrams */ | |
2922 | if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) | |
2923 | break; | |
2924 | } | |
2925 | ||
2926 | /* Out of band data, return right away */ | |
2927 | if (msg_sys.msg_flags & MSG_OOB) | |
2928 | break; | |
a78cb84c | 2929 | cond_resched(); |
a2e27255 ACM |
2930 | } |
2931 | ||
a2e27255 | 2932 | if (err == 0) |
34b88a68 ACM |
2933 | goto out_put; |
2934 | ||
2935 | if (datagrams == 0) { | |
2936 | datagrams = err; | |
2937 | goto out_put; | |
2938 | } | |
a2e27255 | 2939 | |
34b88a68 ACM |
2940 | /* |
2941 | * We may return less entries than requested (vlen) if the | |
2942 | * sock is non block and there aren't enough datagrams... | |
2943 | */ | |
2944 | if (err != -EAGAIN) { | |
a2e27255 | 2945 | /* |
34b88a68 ACM |
2946 | * ... or if recvmsg returns an error after we |
2947 | * received some datagrams, where we record the | |
2948 | * error to return on the next call or if the | |
2949 | * app asks about it using getsockopt(SO_ERROR). | |
a2e27255 | 2950 | */ |
e05a5f51 | 2951 | WRITE_ONCE(sock->sk->sk_err, -err); |
a2e27255 | 2952 | } |
34b88a68 ACM |
2953 | out_put: |
2954 | fput_light(sock->file, fput_needed); | |
a2e27255 | 2955 | |
34b88a68 | 2956 | return datagrams; |
a2e27255 ACM |
2957 | } |
2958 | ||
e11d4284 AB |
2959 | int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, |
2960 | unsigned int vlen, unsigned int flags, | |
2961 | struct __kernel_timespec __user *timeout, | |
2962 | struct old_timespec32 __user *timeout32) | |
a2e27255 ACM |
2963 | { |
2964 | int datagrams; | |
c2e6c856 | 2965 | struct timespec64 timeout_sys; |
a2e27255 | 2966 | |
e11d4284 AB |
2967 | if (timeout && get_timespec64(&timeout_sys, timeout)) |
2968 | return -EFAULT; | |
a2e27255 | 2969 | |
e11d4284 | 2970 | if (timeout32 && get_old_timespec32(&timeout_sys, timeout32)) |
a2e27255 ACM |
2971 | return -EFAULT; |
2972 | ||
e11d4284 AB |
2973 | if (!timeout && !timeout32) |
2974 | return do_recvmmsg(fd, mmsg, vlen, flags, NULL); | |
2975 | ||
2976 | datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); | |
a2e27255 | 2977 | |
e11d4284 AB |
2978 | if (datagrams <= 0) |
2979 | return datagrams; | |
2980 | ||
2981 | if (timeout && put_timespec64(&timeout_sys, timeout)) | |
2982 | datagrams = -EFAULT; | |
2983 | ||
2984 | if (timeout32 && put_old_timespec32(&timeout_sys, timeout32)) | |
a2e27255 ACM |
2985 | datagrams = -EFAULT; |
2986 | ||
2987 | return datagrams; | |
2988 | } | |
2989 | ||
1255e269 DB |
2990 | SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, |
2991 | unsigned int, vlen, unsigned int, flags, | |
c2e6c856 | 2992 | struct __kernel_timespec __user *, timeout) |
1255e269 | 2993 | { |
e11d4284 AB |
2994 | if (flags & MSG_CMSG_COMPAT) |
2995 | return -EINVAL; | |
2996 | ||
2997 | return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL); | |
2998 | } | |
2999 | ||
3000 | #ifdef CONFIG_COMPAT_32BIT_TIME | |
3001 | SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg, | |
3002 | unsigned int, vlen, unsigned int, flags, | |
3003 | struct old_timespec32 __user *, timeout) | |
3004 | { | |
3005 | if (flags & MSG_CMSG_COMPAT) | |
3006 | return -EINVAL; | |
3007 | ||
3008 | return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout); | |
1255e269 | 3009 | } |
e11d4284 | 3010 | #endif |
1255e269 | 3011 | |
a2e27255 | 3012 | #ifdef __ARCH_WANT_SYS_SOCKETCALL |
1da177e4 LT |
3013 | /* Argument list sizes for sys_socketcall */ |
3014 | #define AL(x) ((x) * sizeof(unsigned long)) | |
228e548e | 3015 | static const unsigned char nargs[21] = { |
c6d409cf ED |
3016 | AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), |
3017 | AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), | |
3018 | AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), | |
228e548e | 3019 | AL(4), AL(5), AL(4) |
89bddce5 SH |
3020 | }; |
3021 | ||
1da177e4 LT |
3022 | #undef AL |
3023 | ||
3024 | /* | |
89bddce5 | 3025 | * System call vectors. |
1da177e4 LT |
3026 | * |
3027 | * Argument checking cleaned up. Saved 20% in size. | |
3028 | * This function doesn't need to set the kernel lock because | |
89bddce5 | 3029 | * it is set by the callees. |
1da177e4 LT |
3030 | */ |
3031 | ||
3e0fa65f | 3032 | SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) |
1da177e4 | 3033 | { |
2950fa9d | 3034 | unsigned long a[AUDITSC_ARGS]; |
89bddce5 | 3035 | unsigned long a0, a1; |
1da177e4 | 3036 | int err; |
47379052 | 3037 | unsigned int len; |
1da177e4 | 3038 | |
228e548e | 3039 | if (call < 1 || call > SYS_SENDMMSG) |
1da177e4 | 3040 | return -EINVAL; |
c8e8cd57 | 3041 | call = array_index_nospec(call, SYS_SENDMMSG + 1); |
1da177e4 | 3042 | |
47379052 AV |
3043 | len = nargs[call]; |
3044 | if (len > sizeof(a)) | |
3045 | return -EINVAL; | |
3046 | ||
1da177e4 | 3047 | /* copy_from_user should be SMP safe. */ |
47379052 | 3048 | if (copy_from_user(a, args, len)) |
1da177e4 | 3049 | return -EFAULT; |
3ec3b2fb | 3050 | |
2950fa9d CG |
3051 | err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); |
3052 | if (err) | |
3053 | return err; | |
3ec3b2fb | 3054 | |
89bddce5 SH |
3055 | a0 = a[0]; |
3056 | a1 = a[1]; | |
3057 | ||
3058 | switch (call) { | |
3059 | case SYS_SOCKET: | |
9d6a15c3 | 3060 | err = __sys_socket(a0, a1, a[2]); |
89bddce5 SH |
3061 | break; |
3062 | case SYS_BIND: | |
a87d35d8 | 3063 | err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); |
89bddce5 SH |
3064 | break; |
3065 | case SYS_CONNECT: | |
1387c2c2 | 3066 | err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); |
89bddce5 SH |
3067 | break; |
3068 | case SYS_LISTEN: | |
25e290ee | 3069 | err = __sys_listen(a0, a1); |
89bddce5 SH |
3070 | break; |
3071 | case SYS_ACCEPT: | |
4541e805 DB |
3072 | err = __sys_accept4(a0, (struct sockaddr __user *)a1, |
3073 | (int __user *)a[2], 0); | |
89bddce5 SH |
3074 | break; |
3075 | case SYS_GETSOCKNAME: | |
3076 | err = | |
8882a107 DB |
3077 | __sys_getsockname(a0, (struct sockaddr __user *)a1, |
3078 | (int __user *)a[2]); | |
89bddce5 SH |
3079 | break; |
3080 | case SYS_GETPEERNAME: | |
3081 | err = | |
b21c8f83 DB |
3082 | __sys_getpeername(a0, (struct sockaddr __user *)a1, |
3083 | (int __user *)a[2]); | |
89bddce5 SH |
3084 | break; |
3085 | case SYS_SOCKETPAIR: | |
6debc8d8 | 3086 | err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); |
89bddce5 SH |
3087 | break; |
3088 | case SYS_SEND: | |
f3bf896b DB |
3089 | err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], |
3090 | NULL, 0); | |
89bddce5 SH |
3091 | break; |
3092 | case SYS_SENDTO: | |
211b634b DB |
3093 | err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], |
3094 | (struct sockaddr __user *)a[4], a[5]); | |
89bddce5 SH |
3095 | break; |
3096 | case SYS_RECV: | |
d27e9afc DB |
3097 | err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], |
3098 | NULL, NULL); | |
89bddce5 SH |
3099 | break; |
3100 | case SYS_RECVFROM: | |
7a09e1eb DB |
3101 | err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], |
3102 | (struct sockaddr __user *)a[4], | |
3103 | (int __user *)a[5]); | |
89bddce5 SH |
3104 | break; |
3105 | case SYS_SHUTDOWN: | |
005a1aea | 3106 | err = __sys_shutdown(a0, a1); |
89bddce5 SH |
3107 | break; |
3108 | case SYS_SETSOCKOPT: | |
cc36dca0 DB |
3109 | err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], |
3110 | a[4]); | |
89bddce5 SH |
3111 | break; |
3112 | case SYS_GETSOCKOPT: | |
3113 | err = | |
13a2d70e DB |
3114 | __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], |
3115 | (int __user *)a[4]); | |
89bddce5 SH |
3116 | break; |
3117 | case SYS_SENDMSG: | |
e1834a32 DB |
3118 | err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, |
3119 | a[2], true); | |
89bddce5 | 3120 | break; |
228e548e | 3121 | case SYS_SENDMMSG: |
e1834a32 DB |
3122 | err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], |
3123 | a[3], true); | |
228e548e | 3124 | break; |
89bddce5 | 3125 | case SYS_RECVMSG: |
e1834a32 DB |
3126 | err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, |
3127 | a[2], true); | |
89bddce5 | 3128 | break; |
a2e27255 | 3129 | case SYS_RECVMMSG: |
3ca47e95 | 3130 | if (IS_ENABLED(CONFIG_64BIT)) |
e11d4284 AB |
3131 | err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, |
3132 | a[2], a[3], | |
3133 | (struct __kernel_timespec __user *)a[4], | |
3134 | NULL); | |
3135 | else | |
3136 | err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, | |
3137 | a[2], a[3], NULL, | |
3138 | (struct old_timespec32 __user *)a[4]); | |
a2e27255 | 3139 | break; |
de11defe | 3140 | case SYS_ACCEPT4: |
4541e805 DB |
3141 | err = __sys_accept4(a0, (struct sockaddr __user *)a1, |
3142 | (int __user *)a[2], a[3]); | |
aaca0bdc | 3143 | break; |
89bddce5 SH |
3144 | default: |
3145 | err = -EINVAL; | |
3146 | break; | |
1da177e4 LT |
3147 | } |
3148 | return err; | |
3149 | } | |
3150 | ||
89bddce5 | 3151 | #endif /* __ARCH_WANT_SYS_SOCKETCALL */ |
1da177e4 | 3152 | |
55737fda SH |
3153 | /** |
3154 | * sock_register - add a socket protocol handler | |
3155 | * @ops: description of protocol | |
3156 | * | |
1da177e4 LT |
3157 | * This function is called by a protocol handler that wants to |
3158 | * advertise its address family, and have it linked into the | |
e793c0f7 | 3159 | * socket interface. The value ops->family corresponds to the |
55737fda | 3160 | * socket system call protocol family. |
1da177e4 | 3161 | */ |
f0fd27d4 | 3162 | int sock_register(const struct net_proto_family *ops) |
1da177e4 LT |
3163 | { |
3164 | int err; | |
3165 | ||
3166 | if (ops->family >= NPROTO) { | |
3410f22e | 3167 | pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); |
1da177e4 LT |
3168 | return -ENOBUFS; |
3169 | } | |
55737fda SH |
3170 | |
3171 | spin_lock(&net_family_lock); | |
190683a9 ED |
3172 | if (rcu_dereference_protected(net_families[ops->family], |
3173 | lockdep_is_held(&net_family_lock))) | |
55737fda SH |
3174 | err = -EEXIST; |
3175 | else { | |
cf778b00 | 3176 | rcu_assign_pointer(net_families[ops->family], ops); |
1da177e4 LT |
3177 | err = 0; |
3178 | } | |
55737fda SH |
3179 | spin_unlock(&net_family_lock); |
3180 | ||
fe0bdbde | 3181 | pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]); |
1da177e4 LT |
3182 | return err; |
3183 | } | |
c6d409cf | 3184 | EXPORT_SYMBOL(sock_register); |
1da177e4 | 3185 | |
55737fda SH |
3186 | /** |
3187 | * sock_unregister - remove a protocol handler | |
3188 | * @family: protocol family to remove | |
3189 | * | |
1da177e4 LT |
3190 | * This function is called by a protocol handler that wants to |
3191 | * remove its address family, and have it unlinked from the | |
55737fda SH |
3192 | * new socket creation. |
3193 | * | |
3194 | * If protocol handler is a module, then it can use module reference | |
3195 | * counts to protect against new references. If protocol handler is not | |
3196 | * a module then it needs to provide its own protection in | |
3197 | * the ops->create routine. | |
1da177e4 | 3198 | */ |
f0fd27d4 | 3199 | void sock_unregister(int family) |
1da177e4 | 3200 | { |
f0fd27d4 | 3201 | BUG_ON(family < 0 || family >= NPROTO); |
1da177e4 | 3202 | |
55737fda | 3203 | spin_lock(&net_family_lock); |
a9b3cd7f | 3204 | RCU_INIT_POINTER(net_families[family], NULL); |
55737fda SH |
3205 | spin_unlock(&net_family_lock); |
3206 | ||
3207 | synchronize_rcu(); | |
3208 | ||
fe0bdbde | 3209 | pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]); |
1da177e4 | 3210 | } |
c6d409cf | 3211 | EXPORT_SYMBOL(sock_unregister); |
1da177e4 | 3212 | |
bf2ae2e4 XL |
3213 | bool sock_is_registered(int family) |
3214 | { | |
66b51b0a | 3215 | return family < NPROTO && rcu_access_pointer(net_families[family]); |
bf2ae2e4 XL |
3216 | } |
3217 | ||
77d76ea3 | 3218 | static int __init sock_init(void) |
1da177e4 | 3219 | { |
b3e19d92 | 3220 | int err; |
2ca794e5 EB |
3221 | /* |
3222 | * Initialize the network sysctl infrastructure. | |
3223 | */ | |
3224 | err = net_sysctl_init(); | |
3225 | if (err) | |
3226 | goto out; | |
b3e19d92 | 3227 | |
1da177e4 | 3228 | /* |
89bddce5 | 3229 | * Initialize skbuff SLAB cache |
1da177e4 LT |
3230 | */ |
3231 | skb_init(); | |
1da177e4 LT |
3232 | |
3233 | /* | |
89bddce5 | 3234 | * Initialize the protocols module. |
1da177e4 LT |
3235 | */ |
3236 | ||
3237 | init_inodecache(); | |
b3e19d92 NP |
3238 | |
3239 | err = register_filesystem(&sock_fs_type); | |
3240 | if (err) | |
47260ba9 | 3241 | goto out; |
1da177e4 | 3242 | sock_mnt = kern_mount(&sock_fs_type); |
b3e19d92 NP |
3243 | if (IS_ERR(sock_mnt)) { |
3244 | err = PTR_ERR(sock_mnt); | |
3245 | goto out_mount; | |
3246 | } | |
77d76ea3 AK |
3247 | |
3248 | /* The real protocol initialization is performed in later initcalls. | |
1da177e4 LT |
3249 | */ |
3250 | ||
3251 | #ifdef CONFIG_NETFILTER | |
6d11cfdb PNA |
3252 | err = netfilter_init(); |
3253 | if (err) | |
3254 | goto out; | |
1da177e4 | 3255 | #endif |
cbeb321a | 3256 | |
408eccce | 3257 | ptp_classifier_init(); |
c1f19b51 | 3258 | |
b3e19d92 NP |
3259 | out: |
3260 | return err; | |
3261 | ||
3262 | out_mount: | |
3263 | unregister_filesystem(&sock_fs_type); | |
b3e19d92 | 3264 | goto out; |
1da177e4 LT |
3265 | } |
3266 | ||
77d76ea3 AK |
3267 | core_initcall(sock_init); /* early initcall */ |
3268 | ||
1da177e4 LT |
3269 | #ifdef CONFIG_PROC_FS |
3270 | void socket_seq_show(struct seq_file *seq) | |
3271 | { | |
648845ab TZ |
3272 | seq_printf(seq, "sockets: used %d\n", |
3273 | sock_inuse_get(seq->private)); | |
1da177e4 | 3274 | } |
89bddce5 | 3275 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 3276 | |
29c49648 AB |
3277 | /* Handle the fact that while struct ifreq has the same *layout* on |
3278 | * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data, | |
3279 | * which are handled elsewhere, it still has different *size* due to | |
3280 | * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit, | |
3281 | * resulting in struct ifreq being 32 and 40 bytes respectively). | |
3282 | * As a result, if the struct happens to be at the end of a page and | |
3283 | * the next page isn't readable/writable, we get a fault. To prevent | |
3284 | * that, copy back and forth to the full size. | |
3285 | */ | |
3286 | int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg) | |
7a229387 | 3287 | { |
29c49648 AB |
3288 | if (in_compat_syscall()) { |
3289 | struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr; | |
7a229387 | 3290 | |
29c49648 AB |
3291 | memset(ifr, 0, sizeof(*ifr)); |
3292 | if (copy_from_user(ifr32, arg, sizeof(*ifr32))) | |
3293 | return -EFAULT; | |
7a229387 | 3294 | |
29c49648 AB |
3295 | if (ifrdata) |
3296 | *ifrdata = compat_ptr(ifr32->ifr_data); | |
7a229387 | 3297 | |
29c49648 AB |
3298 | return 0; |
3299 | } | |
7a229387 | 3300 | |
29c49648 | 3301 | if (copy_from_user(ifr, arg, sizeof(*ifr))) |
7a229387 AB |
3302 | return -EFAULT; |
3303 | ||
29c49648 AB |
3304 | if (ifrdata) |
3305 | *ifrdata = ifr->ifr_data; | |
3306 | ||
7a229387 AB |
3307 | return 0; |
3308 | } | |
29c49648 | 3309 | EXPORT_SYMBOL(get_user_ifreq); |
7a229387 | 3310 | |
29c49648 | 3311 | int put_user_ifreq(struct ifreq *ifr, void __user *arg) |
7a229387 | 3312 | { |
29c49648 | 3313 | size_t size = sizeof(*ifr); |
7a229387 | 3314 | |
29c49648 AB |
3315 | if (in_compat_syscall()) |
3316 | size = sizeof(struct compat_ifreq); | |
7a229387 | 3317 | |
29c49648 | 3318 | if (copy_to_user(arg, ifr, size)) |
7a229387 AB |
3319 | return -EFAULT; |
3320 | ||
3a7da39d | 3321 | return 0; |
7a229387 | 3322 | } |
29c49648 | 3323 | EXPORT_SYMBOL(put_user_ifreq); |
7a229387 | 3324 | |
89bbfc95 | 3325 | #ifdef CONFIG_COMPAT |
7a50a240 AB |
3326 | static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) |
3327 | { | |
7a50a240 | 3328 | compat_uptr_t uptr32; |
44c02a2c AV |
3329 | struct ifreq ifr; |
3330 | void __user *saved; | |
3331 | int err; | |
7a50a240 | 3332 | |
29c49648 | 3333 | if (get_user_ifreq(&ifr, NULL, uifr32)) |
7a50a240 AB |
3334 | return -EFAULT; |
3335 | ||
3336 | if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) | |
3337 | return -EFAULT; | |
3338 | ||
44c02a2c AV |
3339 | saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; |
3340 | ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); | |
7a229387 | 3341 | |
a554bf96 | 3342 | err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL); |
44c02a2c AV |
3343 | if (!err) { |
3344 | ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; | |
29c49648 | 3345 | if (put_user_ifreq(&ifr, uifr32)) |
44c02a2c | 3346 | err = -EFAULT; |
ccbd6a5a | 3347 | } |
44c02a2c | 3348 | return err; |
7a229387 AB |
3349 | } |
3350 | ||
590d4693 BH |
3351 | /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ |
3352 | static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, | |
6b96018b | 3353 | struct compat_ifreq __user *u_ifreq32) |
7a229387 | 3354 | { |
44c02a2c | 3355 | struct ifreq ifreq; |
a554bf96 | 3356 | void __user *data; |
7a229387 | 3357 | |
d0efb162 PC |
3358 | if (!is_socket_ioctl_cmd(cmd)) |
3359 | return -ENOTTY; | |
a554bf96 | 3360 | if (get_user_ifreq(&ifreq, &data, u_ifreq32)) |
7a229387 | 3361 | return -EFAULT; |
a554bf96 | 3362 | ifreq.ifr_data = data; |
7a229387 | 3363 | |
a554bf96 | 3364 | return dev_ioctl(net, cmd, &ifreq, data, NULL); |
a2116ed2 AB |
3365 | } |
3366 | ||
6b96018b AB |
3367 | static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, |
3368 | unsigned int cmd, unsigned long arg) | |
3369 | { | |
3370 | void __user *argp = compat_ptr(arg); | |
3371 | struct sock *sk = sock->sk; | |
3372 | struct net *net = sock_net(sk); | |
1ded5e5a | 3373 | const struct proto_ops *ops; |
7a229387 | 3374 | |
6b96018b | 3375 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) |
88fc023f | 3376 | return sock_ioctl(file, cmd, (unsigned long)argp); |
6b96018b AB |
3377 | |
3378 | switch (cmd) { | |
7a50a240 AB |
3379 | case SIOCWANDEV: |
3380 | return compat_siocwandev(net, argp); | |
0768e170 AB |
3381 | case SIOCGSTAMP_OLD: |
3382 | case SIOCGSTAMPNS_OLD: | |
1ded5e5a ED |
3383 | ops = READ_ONCE(sock->ops); |
3384 | if (!ops->gettstamp) | |
c7cbdbf2 | 3385 | return -ENOIOCTLCMD; |
1ded5e5a ED |
3386 | return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD, |
3387 | !COMPAT_USE_64BIT_TIME); | |
c7cbdbf2 | 3388 | |
dd98d289 | 3389 | case SIOCETHTOOL: |
590d4693 BH |
3390 | case SIOCBONDSLAVEINFOQUERY: |
3391 | case SIOCBONDINFOQUERY: | |
a2116ed2 | 3392 | case SIOCSHWTSTAMP: |
fd468c74 | 3393 | case SIOCGHWTSTAMP: |
590d4693 | 3394 | return compat_ifr_data_ioctl(net, cmd, argp); |
6b96018b AB |
3395 | |
3396 | case FIOSETOWN: | |
3397 | case SIOCSPGRP: | |
3398 | case FIOGETOWN: | |
3399 | case SIOCGPGRP: | |
3400 | case SIOCBRADDBR: | |
3401 | case SIOCBRDELBR: | |
3402 | case SIOCGIFVLAN: | |
3403 | case SIOCSIFVLAN: | |
c62cce2c | 3404 | case SIOCGSKNS: |
0768e170 AB |
3405 | case SIOCGSTAMP_NEW: |
3406 | case SIOCGSTAMPNS_NEW: | |
876f0bf9 | 3407 | case SIOCGIFCONF: |
fd3a4590 RP |
3408 | case SIOCSIFBR: |
3409 | case SIOCGIFBR: | |
6b96018b AB |
3410 | return sock_ioctl(file, cmd, arg); |
3411 | ||
3412 | case SIOCGIFFLAGS: | |
3413 | case SIOCSIFFLAGS: | |
709566d7 AB |
3414 | case SIOCGIFMAP: |
3415 | case SIOCSIFMAP: | |
6b96018b AB |
3416 | case SIOCGIFMETRIC: |
3417 | case SIOCSIFMETRIC: | |
3418 | case SIOCGIFMTU: | |
3419 | case SIOCSIFMTU: | |
3420 | case SIOCGIFMEM: | |
3421 | case SIOCSIFMEM: | |
3422 | case SIOCGIFHWADDR: | |
3423 | case SIOCSIFHWADDR: | |
3424 | case SIOCADDMULTI: | |
3425 | case SIOCDELMULTI: | |
3426 | case SIOCGIFINDEX: | |
6b96018b AB |
3427 | case SIOCGIFADDR: |
3428 | case SIOCSIFADDR: | |
3429 | case SIOCSIFHWBROADCAST: | |
6b96018b | 3430 | case SIOCDIFADDR: |
6b96018b AB |
3431 | case SIOCGIFBRDADDR: |
3432 | case SIOCSIFBRDADDR: | |
3433 | case SIOCGIFDSTADDR: | |
3434 | case SIOCSIFDSTADDR: | |
3435 | case SIOCGIFNETMASK: | |
3436 | case SIOCSIFNETMASK: | |
3437 | case SIOCSIFPFLAGS: | |
3438 | case SIOCGIFPFLAGS: | |
3439 | case SIOCGIFTXQLEN: | |
3440 | case SIOCSIFTXQLEN: | |
3441 | case SIOCBRADDIF: | |
3442 | case SIOCBRDELIF: | |
c6c9fee3 | 3443 | case SIOCGIFNAME: |
9177efd3 AB |
3444 | case SIOCSIFNAME: |
3445 | case SIOCGMIIPHY: | |
3446 | case SIOCGMIIREG: | |
3447 | case SIOCSMIIREG: | |
f92d4fc9 AV |
3448 | case SIOCBONDENSLAVE: |
3449 | case SIOCBONDRELEASE: | |
3450 | case SIOCBONDSETHWADDR: | |
3451 | case SIOCBONDCHANGEACTIVE: | |
6b96018b AB |
3452 | case SIOCSARP: |
3453 | case SIOCGARP: | |
3454 | case SIOCDARP: | |
c7dc504e | 3455 | case SIOCOUTQ: |
9d7bf41f | 3456 | case SIOCOUTQNSD: |
6b96018b | 3457 | case SIOCATMARK: |
63ff03ab | 3458 | return sock_do_ioctl(net, sock, cmd, arg); |
9177efd3 AB |
3459 | } |
3460 | ||
6b96018b AB |
3461 | return -ENOIOCTLCMD; |
3462 | } | |
7a229387 | 3463 | |
95c96174 | 3464 | static long compat_sock_ioctl(struct file *file, unsigned int cmd, |
89bddce5 | 3465 | unsigned long arg) |
89bbfc95 SP |
3466 | { |
3467 | struct socket *sock = file->private_data; | |
1ded5e5a | 3468 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
89bbfc95 | 3469 | int ret = -ENOIOCTLCMD; |
87de87d5 DM |
3470 | struct sock *sk; |
3471 | struct net *net; | |
3472 | ||
3473 | sk = sock->sk; | |
3474 | net = sock_net(sk); | |
89bbfc95 | 3475 | |
1ded5e5a ED |
3476 | if (ops->compat_ioctl) |
3477 | ret = ops->compat_ioctl(sock, cmd, arg); | |
89bbfc95 | 3478 | |
87de87d5 DM |
3479 | if (ret == -ENOIOCTLCMD && |
3480 | (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) | |
3481 | ret = compat_wext_handle_ioctl(net, cmd, arg); | |
3482 | ||
6b96018b AB |
3483 | if (ret == -ENOIOCTLCMD) |
3484 | ret = compat_sock_ioctl_trans(file, sock, cmd, arg); | |
3485 | ||
89bbfc95 SP |
3486 | return ret; |
3487 | } | |
3488 | #endif | |
3489 | ||
8a3c245c PT |
3490 | /** |
3491 | * kernel_bind - bind an address to a socket (kernel space) | |
3492 | * @sock: socket | |
3493 | * @addr: address | |
3494 | * @addrlen: length of address | |
3495 | * | |
3496 | * Returns 0 or an error. | |
3497 | */ | |
3498 | ||
ac5a488e SS |
3499 | int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) |
3500 | { | |
1ded5e5a | 3501 | return READ_ONCE(sock->ops)->bind(sock, addr, addrlen); |
ac5a488e | 3502 | } |
c6d409cf | 3503 | EXPORT_SYMBOL(kernel_bind); |
ac5a488e | 3504 | |
8a3c245c PT |
3505 | /** |
3506 | * kernel_listen - move socket to listening state (kernel space) | |
3507 | * @sock: socket | |
3508 | * @backlog: pending connections queue size | |
3509 | * | |
3510 | * Returns 0 or an error. | |
3511 | */ | |
3512 | ||
ac5a488e SS |
3513 | int kernel_listen(struct socket *sock, int backlog) |
3514 | { | |
1ded5e5a | 3515 | return READ_ONCE(sock->ops)->listen(sock, backlog); |
ac5a488e | 3516 | } |
c6d409cf | 3517 | EXPORT_SYMBOL(kernel_listen); |
ac5a488e | 3518 | |
8a3c245c PT |
3519 | /** |
3520 | * kernel_accept - accept a connection (kernel space) | |
3521 | * @sock: listening socket | |
3522 | * @newsock: new connected socket | |
3523 | * @flags: flags | |
3524 | * | |
3525 | * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0. | |
3526 | * If it fails, @newsock is guaranteed to be %NULL. | |
3527 | * Returns 0 or an error. | |
3528 | */ | |
3529 | ||
ac5a488e SS |
3530 | int kernel_accept(struct socket *sock, struct socket **newsock, int flags) |
3531 | { | |
3532 | struct sock *sk = sock->sk; | |
1ded5e5a | 3533 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
ac5a488e SS |
3534 | int err; |
3535 | ||
3536 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
3537 | newsock); | |
3538 | if (err < 0) | |
3539 | goto done; | |
3540 | ||
1ded5e5a | 3541 | err = ops->accept(sock, *newsock, flags, true); |
ac5a488e SS |
3542 | if (err < 0) { |
3543 | sock_release(*newsock); | |
fa8705b0 | 3544 | *newsock = NULL; |
ac5a488e SS |
3545 | goto done; |
3546 | } | |
3547 | ||
1ded5e5a ED |
3548 | (*newsock)->ops = ops; |
3549 | __module_get(ops->owner); | |
ac5a488e SS |
3550 | |
3551 | done: | |
3552 | return err; | |
3553 | } | |
c6d409cf | 3554 | EXPORT_SYMBOL(kernel_accept); |
ac5a488e | 3555 | |
8a3c245c PT |
3556 | /** |
3557 | * kernel_connect - connect a socket (kernel space) | |
3558 | * @sock: socket | |
3559 | * @addr: address | |
3560 | * @addrlen: address length | |
3561 | * @flags: flags (O_NONBLOCK, ...) | |
3562 | * | |
f1dcffcc | 3563 | * For datagram sockets, @addr is the address to which datagrams are sent |
8a3c245c PT |
3564 | * by default, and the only address from which datagrams are received. |
3565 | * For stream sockets, attempts to connect to @addr. | |
3566 | * Returns 0 or an error code. | |
3567 | */ | |
3568 | ||
ac5a488e | 3569 | int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, |
4768fbcb | 3570 | int flags) |
ac5a488e | 3571 | { |
0bdf3993 JR |
3572 | struct sockaddr_storage address; |
3573 | ||
3574 | memcpy(&address, addr, addrlen); | |
3575 | ||
3576 | return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address, | |
3577 | addrlen, flags); | |
ac5a488e | 3578 | } |
c6d409cf | 3579 | EXPORT_SYMBOL(kernel_connect); |
ac5a488e | 3580 | |
8a3c245c PT |
3581 | /** |
3582 | * kernel_getsockname - get the address which the socket is bound (kernel space) | |
3583 | * @sock: socket | |
3584 | * @addr: address holder | |
3585 | * | |
3586 | * Fills the @addr pointer with the address which the socket is bound. | |
0fc95dec | 3587 | * Returns the length of the address in bytes or an error code. |
8a3c245c PT |
3588 | */ |
3589 | ||
9b2c45d4 | 3590 | int kernel_getsockname(struct socket *sock, struct sockaddr *addr) |
ac5a488e | 3591 | { |
1ded5e5a | 3592 | return READ_ONCE(sock->ops)->getname(sock, addr, 0); |
ac5a488e | 3593 | } |
c6d409cf | 3594 | EXPORT_SYMBOL(kernel_getsockname); |
ac5a488e | 3595 | |
8a3c245c | 3596 | /** |
645f0897 | 3597 | * kernel_getpeername - get the address which the socket is connected (kernel space) |
8a3c245c PT |
3598 | * @sock: socket |
3599 | * @addr: address holder | |
3600 | * | |
3601 | * Fills the @addr pointer with the address which the socket is connected. | |
0fc95dec | 3602 | * Returns the length of the address in bytes or an error code. |
8a3c245c PT |
3603 | */ |
3604 | ||
9b2c45d4 | 3605 | int kernel_getpeername(struct socket *sock, struct sockaddr *addr) |
ac5a488e | 3606 | { |
1ded5e5a | 3607 | return READ_ONCE(sock->ops)->getname(sock, addr, 1); |
ac5a488e | 3608 | } |
c6d409cf | 3609 | EXPORT_SYMBOL(kernel_getpeername); |
ac5a488e | 3610 | |
8a3c245c | 3611 | /** |
645f0897 | 3612 | * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space) |
8a3c245c PT |
3613 | * @sock: socket |
3614 | * @how: connection part | |
3615 | * | |
3616 | * Returns 0 or an error. | |
3617 | */ | |
3618 | ||
91cf45f0 TM |
3619 | int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) |
3620 | { | |
1ded5e5a | 3621 | return READ_ONCE(sock->ops)->shutdown(sock, how); |
91cf45f0 | 3622 | } |
91cf45f0 | 3623 | EXPORT_SYMBOL(kernel_sock_shutdown); |
113c3075 | 3624 | |
8a3c245c PT |
3625 | /** |
3626 | * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket | |
3627 | * @sk: socket | |
3628 | * | |
3629 | * This routine returns the IP overhead imposed by a socket i.e. | |
3630 | * the length of the underlying IP header, depending on whether | |
3631 | * this is an IPv4 or IPv6 socket and the length from IP options turned | |
3632 | * on at the socket. Assumes that the caller has a lock on the socket. | |
113c3075 | 3633 | */ |
8a3c245c | 3634 | |
113c3075 P |
3635 | u32 kernel_sock_ip_overhead(struct sock *sk) |
3636 | { | |
3637 | struct inet_sock *inet; | |
3638 | struct ip_options_rcu *opt; | |
3639 | u32 overhead = 0; | |
113c3075 P |
3640 | #if IS_ENABLED(CONFIG_IPV6) |
3641 | struct ipv6_pinfo *np; | |
3642 | struct ipv6_txoptions *optv6 = NULL; | |
3643 | #endif /* IS_ENABLED(CONFIG_IPV6) */ | |
3644 | ||
3645 | if (!sk) | |
3646 | return overhead; | |
3647 | ||
113c3075 P |
3648 | switch (sk->sk_family) { |
3649 | case AF_INET: | |
3650 | inet = inet_sk(sk); | |
3651 | overhead += sizeof(struct iphdr); | |
3652 | opt = rcu_dereference_protected(inet->inet_opt, | |
614d79c0 | 3653 | sock_owned_by_user(sk)); |
113c3075 P |
3654 | if (opt) |
3655 | overhead += opt->opt.optlen; | |
3656 | return overhead; | |
3657 | #if IS_ENABLED(CONFIG_IPV6) | |
3658 | case AF_INET6: | |
3659 | np = inet6_sk(sk); | |
3660 | overhead += sizeof(struct ipv6hdr); | |
3661 | if (np) | |
3662 | optv6 = rcu_dereference_protected(np->opt, | |
614d79c0 | 3663 | sock_owned_by_user(sk)); |
113c3075 P |
3664 | if (optv6) |
3665 | overhead += (optv6->opt_flen + optv6->opt_nflen); | |
3666 | return overhead; | |
3667 | #endif /* IS_ENABLED(CONFIG_IPV6) */ | |
3668 | default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ | |
3669 | return overhead; | |
3670 | } | |
3671 | } | |
3672 | EXPORT_SYMBOL(kernel_sock_ip_overhead); |