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 | ||
295d3c44 | 406 | static const struct xattr_handler * const sockfs_xattr_handlers[] = { |
bba0bd31 | 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 | ||
86a7e0b6 JR |
740 | static int __sock_sendmsg(struct socket *sock, struct msghdr *msg) |
741 | { | |
742 | int err = security_socket_sendmsg(sock, msg, | |
743 | msg_data_left(msg)); | |
744 | ||
745 | return err ?: sock_sendmsg_nosec(sock, msg); | |
746 | } | |
747 | ||
85806af0 RD |
748 | /** |
749 | * sock_sendmsg - send a message through @sock | |
750 | * @sock: socket | |
751 | * @msg: message to send | |
752 | * | |
753 | * Sends @msg through @sock, passing through LSM. | |
754 | * Returns the number of bytes sent, or an error code. | |
755 | */ | |
d8725c86 | 756 | int sock_sendmsg(struct socket *sock, struct msghdr *msg) |
228e548e | 757 | { |
86a7e0b6 JR |
758 | struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name; |
759 | struct sockaddr_storage address; | |
760 | int ret; | |
228e548e | 761 | |
86a7e0b6 JR |
762 | if (msg->msg_name) { |
763 | memcpy(&address, msg->msg_name, msg->msg_namelen); | |
764 | msg->msg_name = &address; | |
765 | } | |
766 | ||
767 | ret = __sock_sendmsg(sock, msg); | |
768 | msg->msg_name = save_addr; | |
769 | ||
770 | return ret; | |
0cf00c6f | 771 | } |
c6d409cf | 772 | EXPORT_SYMBOL(sock_sendmsg); |
1da177e4 | 773 | |
8a3c245c PT |
774 | /** |
775 | * kernel_sendmsg - send a message through @sock (kernel-space) | |
776 | * @sock: socket | |
777 | * @msg: message header | |
778 | * @vec: kernel vec | |
779 | * @num: vec array length | |
780 | * @size: total message data size | |
781 | * | |
782 | * Builds the message data with @vec and sends it through @sock. | |
783 | * Returns the number of bytes sent, or an error code. | |
784 | */ | |
785 | ||
1da177e4 LT |
786 | int kernel_sendmsg(struct socket *sock, struct msghdr *msg, |
787 | struct kvec *vec, size_t num, size_t size) | |
788 | { | |
de4eda9d | 789 | iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size); |
d8725c86 | 790 | return sock_sendmsg(sock, msg); |
1da177e4 | 791 | } |
c6d409cf | 792 | EXPORT_SYMBOL(kernel_sendmsg); |
1da177e4 | 793 | |
8a3c245c PT |
794 | /** |
795 | * kernel_sendmsg_locked - send a message through @sock (kernel-space) | |
796 | * @sk: sock | |
797 | * @msg: message header | |
798 | * @vec: output s/g array | |
799 | * @num: output s/g array length | |
800 | * @size: total message data size | |
801 | * | |
802 | * Builds the message data with @vec and sends it through @sock. | |
803 | * Returns the number of bytes sent, or an error code. | |
804 | * Caller must hold @sk. | |
805 | */ | |
806 | ||
306b13eb TH |
807 | int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg, |
808 | struct kvec *vec, size_t num, size_t size) | |
809 | { | |
810 | struct socket *sock = sk->sk_socket; | |
1ded5e5a | 811 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
306b13eb | 812 | |
1ded5e5a | 813 | if (!ops->sendmsg_locked) |
db5980d8 | 814 | return sock_no_sendmsg_locked(sk, msg, size); |
306b13eb | 815 | |
de4eda9d | 816 | iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size); |
306b13eb | 817 | |
1ded5e5a | 818 | return ops->sendmsg_locked(sk, msg, msg_data_left(msg)); |
306b13eb TH |
819 | } |
820 | EXPORT_SYMBOL(kernel_sendmsg_locked); | |
821 | ||
8605330a SHY |
822 | static bool skb_is_err_queue(const struct sk_buff *skb) |
823 | { | |
824 | /* pkt_type of skbs enqueued on the error queue are set to | |
825 | * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do | |
826 | * in recvmsg, since skbs received on a local socket will never | |
827 | * have a pkt_type of PACKET_OUTGOING. | |
828 | */ | |
829 | return skb->pkt_type == PACKET_OUTGOING; | |
830 | } | |
831 | ||
b50a5c70 ML |
832 | /* On transmit, software and hardware timestamps are returned independently. |
833 | * As the two skb clones share the hardware timestamp, which may be updated | |
834 | * before the software timestamp is received, a hardware TX timestamp may be | |
835 | * returned only if there is no software TX timestamp. Ignore false software | |
836 | * timestamps, which may be made in the __sock_recv_timestamp() call when the | |
7f1bc6e9 | 837 | * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a |
b50a5c70 ML |
838 | * hardware timestamp. |
839 | */ | |
840 | static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp) | |
841 | { | |
842 | return skb->tstamp && !false_tstamp && skb_is_err_queue(skb); | |
843 | } | |
844 | ||
97dc7cd9 GE |
845 | static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index) |
846 | { | |
e3390b30 | 847 | bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC; |
97dc7cd9 GE |
848 | struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); |
849 | struct net_device *orig_dev; | |
850 | ktime_t hwtstamp; | |
851 | ||
852 | rcu_read_lock(); | |
853 | orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); | |
854 | if (orig_dev) { | |
855 | *if_index = orig_dev->ifindex; | |
856 | hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles); | |
857 | } else { | |
858 | hwtstamp = shhwtstamps->hwtstamp; | |
859 | } | |
860 | rcu_read_unlock(); | |
861 | ||
862 | return hwtstamp; | |
863 | } | |
864 | ||
865 | static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb, | |
866 | int if_index) | |
aad9c8c4 ML |
867 | { |
868 | struct scm_ts_pktinfo ts_pktinfo; | |
869 | struct net_device *orig_dev; | |
870 | ||
871 | if (!skb_mac_header_was_set(skb)) | |
872 | return; | |
873 | ||
874 | memset(&ts_pktinfo, 0, sizeof(ts_pktinfo)); | |
875 | ||
97dc7cd9 GE |
876 | if (!if_index) { |
877 | rcu_read_lock(); | |
878 | orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); | |
879 | if (orig_dev) | |
880 | if_index = orig_dev->ifindex; | |
881 | rcu_read_unlock(); | |
882 | } | |
883 | ts_pktinfo.if_index = if_index; | |
aad9c8c4 ML |
884 | |
885 | ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb); | |
886 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO, | |
887 | sizeof(ts_pktinfo), &ts_pktinfo); | |
888 | } | |
889 | ||
92f37fd2 ED |
890 | /* |
891 | * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) | |
892 | */ | |
893 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, | |
894 | struct sk_buff *skb) | |
895 | { | |
20d49473 | 896 | int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); |
887feae3 | 897 | int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW); |
9718475e | 898 | struct scm_timestamping_internal tss; |
b50a5c70 | 899 | int empty = 1, false_tstamp = 0; |
20d49473 PO |
900 | struct skb_shared_hwtstamps *shhwtstamps = |
901 | skb_hwtstamps(skb); | |
97dc7cd9 | 902 | int if_index; |
007747a9 | 903 | ktime_t hwtstamp; |
e3390b30 | 904 | u32 tsflags; |
20d49473 PO |
905 | |
906 | /* Race occurred between timestamp enabling and packet | |
907 | receiving. Fill in the current time for now. */ | |
b50a5c70 | 908 | if (need_software_tstamp && skb->tstamp == 0) { |
20d49473 | 909 | __net_timestamp(skb); |
b50a5c70 ML |
910 | false_tstamp = 1; |
911 | } | |
20d49473 PO |
912 | |
913 | if (need_software_tstamp) { | |
914 | if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { | |
887feae3 DD |
915 | if (new_tstamp) { |
916 | struct __kernel_sock_timeval tv; | |
917 | ||
918 | skb_get_new_timestamp(skb, &tv); | |
919 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW, | |
920 | sizeof(tv), &tv); | |
921 | } else { | |
922 | struct __kernel_old_timeval tv; | |
923 | ||
924 | skb_get_timestamp(skb, &tv); | |
925 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD, | |
926 | sizeof(tv), &tv); | |
927 | } | |
20d49473 | 928 | } else { |
887feae3 DD |
929 | if (new_tstamp) { |
930 | struct __kernel_timespec ts; | |
931 | ||
932 | skb_get_new_timestampns(skb, &ts); | |
933 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW, | |
934 | sizeof(ts), &ts); | |
935 | } else { | |
df1b4ba9 | 936 | struct __kernel_old_timespec ts; |
887feae3 DD |
937 | |
938 | skb_get_timestampns(skb, &ts); | |
939 | put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD, | |
940 | sizeof(ts), &ts); | |
941 | } | |
20d49473 PO |
942 | } |
943 | } | |
944 | ||
f24b9be5 | 945 | memset(&tss, 0, sizeof(tss)); |
e3390b30 ED |
946 | tsflags = READ_ONCE(sk->sk_tsflags); |
947 | if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) && | |
9718475e | 948 | ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0)) |
20d49473 | 949 | empty = 0; |
4d276eb6 | 950 | if (shhwtstamps && |
e3390b30 | 951 | (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) && |
d7c08826 | 952 | !skb_is_swtx_tstamp(skb, false_tstamp)) { |
97dc7cd9 GE |
953 | if_index = 0; |
954 | if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV) | |
955 | hwtstamp = get_timestamp(sk, skb, &if_index); | |
007747a9 ML |
956 | else |
957 | hwtstamp = shhwtstamps->hwtstamp; | |
d7c08826 | 958 | |
e3390b30 | 959 | if (tsflags & SOF_TIMESTAMPING_BIND_PHC) |
97dc7cd9 | 960 | hwtstamp = ptp_convert_timestamp(&hwtstamp, |
251cd405 | 961 | READ_ONCE(sk->sk_bind_phc)); |
97dc7cd9 | 962 | |
007747a9 | 963 | if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) { |
d7c08826 YL |
964 | empty = 0; |
965 | ||
e3390b30 | 966 | if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) && |
d7c08826 | 967 | !skb_is_err_queue(skb)) |
97dc7cd9 | 968 | put_ts_pktinfo(msg, skb, if_index); |
d7c08826 | 969 | } |
aad9c8c4 | 970 | } |
1c885808 | 971 | if (!empty) { |
9718475e DD |
972 | if (sock_flag(sk, SOCK_TSTAMP_NEW)) |
973 | put_cmsg_scm_timestamping64(msg, &tss); | |
974 | else | |
975 | put_cmsg_scm_timestamping(msg, &tss); | |
1c885808 | 976 | |
8605330a | 977 | if (skb_is_err_queue(skb) && skb->len && |
4ef1b286 | 978 | SKB_EXT_ERR(skb)->opt_stats) |
1c885808 FY |
979 | put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS, |
980 | skb->len, skb->data); | |
981 | } | |
92f37fd2 | 982 | } |
7c81fd8b ACM |
983 | EXPORT_SYMBOL_GPL(__sock_recv_timestamp); |
984 | ||
eb6fba75 | 985 | #ifdef CONFIG_WIRELESS |
6e3e939f JB |
986 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, |
987 | struct sk_buff *skb) | |
988 | { | |
989 | int ack; | |
990 | ||
991 | if (!sock_flag(sk, SOCK_WIFI_STATUS)) | |
992 | return; | |
993 | if (!skb->wifi_acked_valid) | |
994 | return; | |
995 | ||
996 | ack = skb->wifi_acked; | |
997 | ||
998 | put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); | |
999 | } | |
1000 | EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); | |
eb6fba75 | 1001 | #endif |
6e3e939f | 1002 | |
11165f14 | 1003 | static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, |
1004 | struct sk_buff *skb) | |
3b885787 | 1005 | { |
744d5a3e | 1006 | if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount) |
3b885787 | 1007 | put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, |
744d5a3e | 1008 | sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount); |
3b885787 NH |
1009 | } |
1010 | ||
6fd1d51c EM |
1011 | static void sock_recv_mark(struct msghdr *msg, struct sock *sk, |
1012 | struct sk_buff *skb) | |
1013 | { | |
2558b803 ED |
1014 | if (sock_flag(sk, SOCK_RCVMARK) && skb) { |
1015 | /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */ | |
1016 | __u32 mark = skb->mark; | |
1017 | ||
1018 | put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark); | |
1019 | } | |
6fd1d51c EM |
1020 | } |
1021 | ||
1022 | void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk, | |
1023 | struct sk_buff *skb) | |
3b885787 NH |
1024 | { |
1025 | sock_recv_timestamp(msg, sk, skb); | |
1026 | sock_recv_drops(msg, sk, skb); | |
6fd1d51c | 1027 | sock_recv_mark(msg, sk, skb); |
3b885787 | 1028 | } |
6fd1d51c | 1029 | EXPORT_SYMBOL_GPL(__sock_recv_cmsgs); |
3b885787 | 1030 | |
8c3c447b | 1031 | INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *, |
a648a592 PA |
1032 | size_t, int)); |
1033 | INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *, | |
1034 | size_t, int)); | |
6e6eda44 YC |
1035 | |
1036 | static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags) | |
1037 | { | |
1038 | trace_sock_recv_length(sk, ret, flags); | |
1039 | } | |
1040 | ||
1b784140 | 1041 | static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, |
2da62906 | 1042 | int flags) |
1da177e4 | 1043 | { |
1ded5e5a ED |
1044 | int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg, |
1045 | inet6_recvmsg, | |
6e6eda44 YC |
1046 | inet_recvmsg, sock, msg, |
1047 | msg_data_left(msg), flags); | |
1048 | if (trace_sock_recv_length_enabled()) | |
1049 | call_trace_sock_recv_length(sock->sk, ret, flags); | |
1050 | return ret; | |
1da177e4 LT |
1051 | } |
1052 | ||
85806af0 RD |
1053 | /** |
1054 | * sock_recvmsg - receive a message from @sock | |
1055 | * @sock: socket | |
1056 | * @msg: message to receive | |
1057 | * @flags: message flags | |
1058 | * | |
1059 | * Receives @msg from @sock, passing through LSM. Returns the total number | |
1060 | * of bytes received, or an error. | |
1061 | */ | |
2da62906 | 1062 | int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) |
a2e27255 | 1063 | { |
2da62906 | 1064 | int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); |
a2e27255 | 1065 | |
2da62906 | 1066 | return err ?: sock_recvmsg_nosec(sock, msg, flags); |
1da177e4 | 1067 | } |
c6d409cf | 1068 | EXPORT_SYMBOL(sock_recvmsg); |
1da177e4 | 1069 | |
c1249c0a | 1070 | /** |
8a3c245c PT |
1071 | * kernel_recvmsg - Receive a message from a socket (kernel space) |
1072 | * @sock: The socket to receive the message from | |
1073 | * @msg: Received message | |
1074 | * @vec: Input s/g array for message data | |
1075 | * @num: Size of input s/g array | |
1076 | * @size: Number of bytes to read | |
1077 | * @flags: Message flags (MSG_DONTWAIT, etc...) | |
c1249c0a | 1078 | * |
8a3c245c PT |
1079 | * On return the msg structure contains the scatter/gather array passed in the |
1080 | * vec argument. The array is modified so that it consists of the unfilled | |
1081 | * portion of the original array. | |
c1249c0a | 1082 | * |
8a3c245c | 1083 | * The returned value is the total number of bytes received, or an error. |
c1249c0a | 1084 | */ |
8a3c245c | 1085 | |
89bddce5 SH |
1086 | int kernel_recvmsg(struct socket *sock, struct msghdr *msg, |
1087 | struct kvec *vec, size_t num, size_t size, int flags) | |
1da177e4 | 1088 | { |
1f466e1f | 1089 | msg->msg_control_is_user = false; |
de4eda9d | 1090 | iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size); |
1f466e1f | 1091 | return sock_recvmsg(sock, msg, flags); |
1da177e4 | 1092 | } |
c6d409cf | 1093 | EXPORT_SYMBOL(kernel_recvmsg); |
1da177e4 | 1094 | |
9c55e01c | 1095 | static ssize_t sock_splice_read(struct file *file, loff_t *ppos, |
c6d409cf | 1096 | struct pipe_inode_info *pipe, size_t len, |
9c55e01c JA |
1097 | unsigned int flags) |
1098 | { | |
1099 | struct socket *sock = file->private_data; | |
1ded5e5a | 1100 | const struct proto_ops *ops; |
9c55e01c | 1101 | |
1ded5e5a ED |
1102 | ops = READ_ONCE(sock->ops); |
1103 | if (unlikely(!ops->splice_read)) | |
67178fd0 | 1104 | return copy_splice_read(file, ppos, pipe, len, flags); |
997b37da | 1105 | |
1ded5e5a | 1106 | return ops->splice_read(sock, ppos, pipe, len, flags); |
9c55e01c JA |
1107 | } |
1108 | ||
2bfc6685 DH |
1109 | static void sock_splice_eof(struct file *file) |
1110 | { | |
1111 | struct socket *sock = file->private_data; | |
1ded5e5a | 1112 | const struct proto_ops *ops; |
2bfc6685 | 1113 | |
1ded5e5a ED |
1114 | ops = READ_ONCE(sock->ops); |
1115 | if (ops->splice_eof) | |
1116 | ops->splice_eof(sock); | |
2bfc6685 DH |
1117 | } |
1118 | ||
8ae5e030 | 1119 | static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) |
ce1d4d3e | 1120 | { |
6d652330 AV |
1121 | struct file *file = iocb->ki_filp; |
1122 | struct socket *sock = file->private_data; | |
0345f931 | 1123 | struct msghdr msg = {.msg_iter = *to, |
1124 | .msg_iocb = iocb}; | |
8ae5e030 | 1125 | ssize_t res; |
ce1d4d3e | 1126 | |
ebfcd895 | 1127 | if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) |
8ae5e030 AV |
1128 | msg.msg_flags = MSG_DONTWAIT; |
1129 | ||
1130 | if (iocb->ki_pos != 0) | |
1da177e4 | 1131 | return -ESPIPE; |
027445c3 | 1132 | |
66ee59af | 1133 | if (!iov_iter_count(to)) /* Match SYS5 behaviour */ |
1da177e4 LT |
1134 | return 0; |
1135 | ||
2da62906 | 1136 | res = sock_recvmsg(sock, &msg, msg.msg_flags); |
8ae5e030 AV |
1137 | *to = msg.msg_iter; |
1138 | return res; | |
1da177e4 LT |
1139 | } |
1140 | ||
8ae5e030 | 1141 | static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) |
ce1d4d3e | 1142 | { |
6d652330 AV |
1143 | struct file *file = iocb->ki_filp; |
1144 | struct socket *sock = file->private_data; | |
0345f931 | 1145 | struct msghdr msg = {.msg_iter = *from, |
1146 | .msg_iocb = iocb}; | |
8ae5e030 | 1147 | ssize_t res; |
1da177e4 | 1148 | |
8ae5e030 | 1149 | if (iocb->ki_pos != 0) |
ce1d4d3e | 1150 | return -ESPIPE; |
027445c3 | 1151 | |
ebfcd895 | 1152 | if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) |
8ae5e030 AV |
1153 | msg.msg_flags = MSG_DONTWAIT; |
1154 | ||
6d652330 AV |
1155 | if (sock->type == SOCK_SEQPACKET) |
1156 | msg.msg_flags |= MSG_EOR; | |
1157 | ||
86a7e0b6 | 1158 | res = __sock_sendmsg(sock, &msg); |
8ae5e030 AV |
1159 | *from = msg.msg_iter; |
1160 | return res; | |
1da177e4 LT |
1161 | } |
1162 | ||
1da177e4 LT |
1163 | /* |
1164 | * Atomic setting of ioctl hooks to avoid race | |
1165 | * with module unload. | |
1166 | */ | |
1167 | ||
4a3e2f71 | 1168 | static DEFINE_MUTEX(br_ioctl_mutex); |
ad2f99ae AB |
1169 | static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br, |
1170 | unsigned int cmd, struct ifreq *ifr, | |
1171 | void __user *uarg); | |
1da177e4 | 1172 | |
ad2f99ae AB |
1173 | void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br, |
1174 | unsigned int cmd, struct ifreq *ifr, | |
1175 | void __user *uarg)) | |
1da177e4 | 1176 | { |
4a3e2f71 | 1177 | mutex_lock(&br_ioctl_mutex); |
1da177e4 | 1178 | br_ioctl_hook = hook; |
4a3e2f71 | 1179 | mutex_unlock(&br_ioctl_mutex); |
1da177e4 LT |
1180 | } |
1181 | EXPORT_SYMBOL(brioctl_set); | |
1182 | ||
ad2f99ae AB |
1183 | int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd, |
1184 | struct ifreq *ifr, void __user *uarg) | |
1185 | { | |
1186 | int err = -ENOPKG; | |
1187 | ||
1188 | if (!br_ioctl_hook) | |
1189 | request_module("bridge"); | |
1190 | ||
1191 | mutex_lock(&br_ioctl_mutex); | |
1192 | if (br_ioctl_hook) | |
1193 | err = br_ioctl_hook(net, br, cmd, ifr, uarg); | |
1194 | mutex_unlock(&br_ioctl_mutex); | |
1195 | ||
1196 | return err; | |
1197 | } | |
1198 | ||
4a3e2f71 | 1199 | static DEFINE_MUTEX(vlan_ioctl_mutex); |
881d966b | 1200 | static int (*vlan_ioctl_hook) (struct net *, void __user *arg); |
1da177e4 | 1201 | |
881d966b | 1202 | void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) |
1da177e4 | 1203 | { |
4a3e2f71 | 1204 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 1205 | vlan_ioctl_hook = hook; |
4a3e2f71 | 1206 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 LT |
1207 | } |
1208 | EXPORT_SYMBOL(vlan_ioctl_set); | |
1209 | ||
6b96018b | 1210 | static long sock_do_ioctl(struct net *net, struct socket *sock, |
63ff03ab | 1211 | unsigned int cmd, unsigned long arg) |
6b96018b | 1212 | { |
1ded5e5a | 1213 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
876f0bf9 AB |
1214 | struct ifreq ifr; |
1215 | bool need_copyout; | |
6b96018b AB |
1216 | int err; |
1217 | void __user *argp = (void __user *)arg; | |
a554bf96 | 1218 | void __user *data; |
6b96018b | 1219 | |
1ded5e5a | 1220 | err = ops->ioctl(sock, cmd, arg); |
6b96018b AB |
1221 | |
1222 | /* | |
1223 | * If this ioctl is unknown try to hand it down | |
1224 | * to the NIC driver. | |
1225 | */ | |
36fd633e AV |
1226 | if (err != -ENOIOCTLCMD) |
1227 | return err; | |
6b96018b | 1228 | |
29ce8f97 JK |
1229 | if (!is_socket_ioctl_cmd(cmd)) |
1230 | return -ENOTTY; | |
1231 | ||
a554bf96 | 1232 | if (get_user_ifreq(&ifr, &data, argp)) |
876f0bf9 | 1233 | return -EFAULT; |
a554bf96 | 1234 | err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); |
876f0bf9 | 1235 | if (!err && need_copyout) |
a554bf96 | 1236 | if (put_user_ifreq(&ifr, argp)) |
44c02a2c | 1237 | return -EFAULT; |
876f0bf9 | 1238 | |
6b96018b AB |
1239 | return err; |
1240 | } | |
1241 | ||
1da177e4 LT |
1242 | /* |
1243 | * With an ioctl, arg may well be a user mode pointer, but we don't know | |
1244 | * what to do with it - that's up to the protocol still. | |
1245 | */ | |
1246 | ||
1247 | static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) | |
1248 | { | |
1ded5e5a | 1249 | const struct proto_ops *ops; |
1da177e4 | 1250 | struct socket *sock; |
881d966b | 1251 | struct sock *sk; |
1da177e4 LT |
1252 | void __user *argp = (void __user *)arg; |
1253 | int pid, err; | |
881d966b | 1254 | struct net *net; |
1da177e4 | 1255 | |
b69aee04 | 1256 | sock = file->private_data; |
1ded5e5a | 1257 | ops = READ_ONCE(sock->ops); |
881d966b | 1258 | sk = sock->sk; |
3b1e0a65 | 1259 | net = sock_net(sk); |
44c02a2c AV |
1260 | if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { |
1261 | struct ifreq ifr; | |
a554bf96 | 1262 | void __user *data; |
44c02a2c | 1263 | bool need_copyout; |
a554bf96 | 1264 | if (get_user_ifreq(&ifr, &data, argp)) |
44c02a2c | 1265 | return -EFAULT; |
a554bf96 | 1266 | err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); |
44c02a2c | 1267 | if (!err && need_copyout) |
a554bf96 | 1268 | if (put_user_ifreq(&ifr, argp)) |
44c02a2c | 1269 | return -EFAULT; |
1da177e4 | 1270 | } else |
3d23e349 | 1271 | #ifdef CONFIG_WEXT_CORE |
1da177e4 | 1272 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { |
b1b0c245 | 1273 | err = wext_handle_ioctl(net, cmd, argp); |
1da177e4 | 1274 | } else |
3d23e349 | 1275 | #endif |
89bddce5 | 1276 | switch (cmd) { |
1da177e4 LT |
1277 | case FIOSETOWN: |
1278 | case SIOCSPGRP: | |
1279 | err = -EFAULT; | |
1280 | if (get_user(pid, (int __user *)argp)) | |
1281 | break; | |
393cc3f5 | 1282 | err = f_setown(sock->file, pid, 1); |
1da177e4 LT |
1283 | break; |
1284 | case FIOGETOWN: | |
1285 | case SIOCGPGRP: | |
609d7fa9 | 1286 | err = put_user(f_getown(sock->file), |
89bddce5 | 1287 | (int __user *)argp); |
1da177e4 LT |
1288 | break; |
1289 | case SIOCGIFBR: | |
1290 | case SIOCSIFBR: | |
1291 | case SIOCBRADDBR: | |
1292 | case SIOCBRDELBR: | |
ad2f99ae | 1293 | err = br_ioctl_call(net, NULL, cmd, NULL, argp); |
1da177e4 LT |
1294 | break; |
1295 | case SIOCGIFVLAN: | |
1296 | case SIOCSIFVLAN: | |
1297 | err = -ENOPKG; | |
1298 | if (!vlan_ioctl_hook) | |
1299 | request_module("8021q"); | |
1300 | ||
4a3e2f71 | 1301 | mutex_lock(&vlan_ioctl_mutex); |
1da177e4 | 1302 | if (vlan_ioctl_hook) |
881d966b | 1303 | err = vlan_ioctl_hook(net, argp); |
4a3e2f71 | 1304 | mutex_unlock(&vlan_ioctl_mutex); |
1da177e4 | 1305 | break; |
c62cce2c AV |
1306 | case SIOCGSKNS: |
1307 | err = -EPERM; | |
1308 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1309 | break; | |
1310 | ||
1311 | err = open_related_ns(&net->ns, get_net_ns); | |
1312 | break; | |
0768e170 AB |
1313 | case SIOCGSTAMP_OLD: |
1314 | case SIOCGSTAMPNS_OLD: | |
1ded5e5a | 1315 | if (!ops->gettstamp) { |
c7cbdbf2 AB |
1316 | err = -ENOIOCTLCMD; |
1317 | break; | |
1318 | } | |
1ded5e5a ED |
1319 | err = ops->gettstamp(sock, argp, |
1320 | cmd == SIOCGSTAMP_OLD, | |
1321 | !IS_ENABLED(CONFIG_64BIT)); | |
60747828 | 1322 | break; |
0768e170 AB |
1323 | case SIOCGSTAMP_NEW: |
1324 | case SIOCGSTAMPNS_NEW: | |
1ded5e5a | 1325 | if (!ops->gettstamp) { |
0768e170 AB |
1326 | err = -ENOIOCTLCMD; |
1327 | break; | |
1328 | } | |
1ded5e5a ED |
1329 | err = ops->gettstamp(sock, argp, |
1330 | cmd == SIOCGSTAMP_NEW, | |
1331 | false); | |
c7cbdbf2 | 1332 | break; |
876f0bf9 AB |
1333 | |
1334 | case SIOCGIFCONF: | |
1335 | err = dev_ifconf(net, argp); | |
1336 | break; | |
1337 | ||
1da177e4 | 1338 | default: |
63ff03ab | 1339 | err = sock_do_ioctl(net, sock, cmd, arg); |
1da177e4 | 1340 | break; |
89bddce5 | 1341 | } |
1da177e4 LT |
1342 | return err; |
1343 | } | |
1344 | ||
8a3c245c PT |
1345 | /** |
1346 | * sock_create_lite - creates a socket | |
1347 | * @family: protocol family (AF_INET, ...) | |
1348 | * @type: communication type (SOCK_STREAM, ...) | |
1349 | * @protocol: protocol (0, ...) | |
1350 | * @res: new socket | |
1351 | * | |
1352 | * Creates a new socket and assigns it to @res, passing through LSM. | |
1353 | * The new socket initialization is not complete, see kernel_accept(). | |
1354 | * Returns 0 or an error. On failure @res is set to %NULL. | |
1355 | * This function internally uses GFP_KERNEL. | |
1356 | */ | |
1357 | ||
1da177e4 LT |
1358 | int sock_create_lite(int family, int type, int protocol, struct socket **res) |
1359 | { | |
1360 | int err; | |
1361 | struct socket *sock = NULL; | |
89bddce5 | 1362 | |
1da177e4 LT |
1363 | err = security_socket_create(family, type, protocol, 1); |
1364 | if (err) | |
1365 | goto out; | |
1366 | ||
1367 | sock = sock_alloc(); | |
1368 | if (!sock) { | |
1369 | err = -ENOMEM; | |
1370 | goto out; | |
1371 | } | |
1372 | ||
1da177e4 | 1373 | sock->type = type; |
7420ed23 VY |
1374 | err = security_socket_post_create(sock, family, type, protocol, 1); |
1375 | if (err) | |
1376 | goto out_release; | |
1377 | ||
1da177e4 LT |
1378 | out: |
1379 | *res = sock; | |
1380 | return err; | |
7420ed23 VY |
1381 | out_release: |
1382 | sock_release(sock); | |
1383 | sock = NULL; | |
1384 | goto out; | |
1da177e4 | 1385 | } |
c6d409cf | 1386 | EXPORT_SYMBOL(sock_create_lite); |
1da177e4 LT |
1387 | |
1388 | /* No kernel lock held - perfect */ | |
ade994f4 | 1389 | static __poll_t sock_poll(struct file *file, poll_table *wait) |
1da177e4 | 1390 | { |
3cafb376 | 1391 | struct socket *sock = file->private_data; |
1ded5e5a | 1392 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
a331de3b | 1393 | __poll_t events = poll_requested_events(wait), flag = 0; |
2d48d67f | 1394 | |
1ded5e5a | 1395 | if (!ops->poll) |
e88958e6 | 1396 | return 0; |
f641f13b | 1397 | |
a331de3b CH |
1398 | if (sk_can_busy_loop(sock->sk)) { |
1399 | /* poll once if requested by the syscall */ | |
1400 | if (events & POLL_BUSY_LOOP) | |
1401 | sk_busy_loop(sock->sk, 1); | |
1402 | ||
1403 | /* if this socket can poll_ll, tell the system call */ | |
1404 | flag = POLL_BUSY_LOOP; | |
1405 | } | |
1406 | ||
1ded5e5a | 1407 | return ops->poll(file, sock, wait) | flag; |
1da177e4 LT |
1408 | } |
1409 | ||
89bddce5 | 1410 | static int sock_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 1411 | { |
b69aee04 | 1412 | struct socket *sock = file->private_data; |
1da177e4 | 1413 | |
1ded5e5a | 1414 | return READ_ONCE(sock->ops)->mmap(file, sock, vma); |
1da177e4 LT |
1415 | } |
1416 | ||
20380731 | 1417 | static int sock_close(struct inode *inode, struct file *filp) |
1da177e4 | 1418 | { |
6d8c50dc | 1419 | __sock_release(SOCKET_I(inode), inode); |
1da177e4 LT |
1420 | return 0; |
1421 | } | |
1422 | ||
1423 | /* | |
1424 | * Update the socket async list | |
1425 | * | |
1426 | * Fasync_list locking strategy. | |
1427 | * | |
1428 | * 1. fasync_list is modified only under process context socket lock | |
1429 | * i.e. under semaphore. | |
1430 | * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) | |
989a2979 | 1431 | * or under socket lock |
1da177e4 LT |
1432 | */ |
1433 | ||
1434 | static int sock_fasync(int fd, struct file *filp, int on) | |
1435 | { | |
989a2979 ED |
1436 | struct socket *sock = filp->private_data; |
1437 | struct sock *sk = sock->sk; | |
333f7909 | 1438 | struct socket_wq *wq = &sock->wq; |
1da177e4 | 1439 | |
989a2979 | 1440 | if (sk == NULL) |
1da177e4 | 1441 | return -EINVAL; |
1da177e4 LT |
1442 | |
1443 | lock_sock(sk); | |
eaefd110 | 1444 | fasync_helper(fd, filp, on, &wq->fasync_list); |
1da177e4 | 1445 | |
eaefd110 | 1446 | if (!wq->fasync_list) |
989a2979 ED |
1447 | sock_reset_flag(sk, SOCK_FASYNC); |
1448 | else | |
bcdce719 | 1449 | sock_set_flag(sk, SOCK_FASYNC); |
1da177e4 | 1450 | |
989a2979 | 1451 | release_sock(sk); |
1da177e4 LT |
1452 | return 0; |
1453 | } | |
1454 | ||
ceb5d58b | 1455 | /* This function may be called only under rcu_lock */ |
1da177e4 | 1456 | |
ceb5d58b | 1457 | int sock_wake_async(struct socket_wq *wq, int how, int band) |
1da177e4 | 1458 | { |
ceb5d58b | 1459 | if (!wq || !wq->fasync_list) |
1da177e4 | 1460 | return -1; |
ceb5d58b | 1461 | |
89bddce5 | 1462 | switch (how) { |
8d8ad9d7 | 1463 | case SOCK_WAKE_WAITD: |
ceb5d58b | 1464 | if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) |
1da177e4 LT |
1465 | break; |
1466 | goto call_kill; | |
8d8ad9d7 | 1467 | case SOCK_WAKE_SPACE: |
ceb5d58b | 1468 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) |
1da177e4 | 1469 | break; |
7c7ab580 | 1470 | fallthrough; |
8d8ad9d7 | 1471 | case SOCK_WAKE_IO: |
89bddce5 | 1472 | call_kill: |
43815482 | 1473 | kill_fasync(&wq->fasync_list, SIGIO, band); |
1da177e4 | 1474 | break; |
8d8ad9d7 | 1475 | case SOCK_WAKE_URG: |
43815482 | 1476 | kill_fasync(&wq->fasync_list, SIGURG, band); |
1da177e4 | 1477 | } |
ceb5d58b | 1478 | |
1da177e4 LT |
1479 | return 0; |
1480 | } | |
c6d409cf | 1481 | EXPORT_SYMBOL(sock_wake_async); |
1da177e4 | 1482 | |
8a3c245c PT |
1483 | /** |
1484 | * __sock_create - creates a socket | |
1485 | * @net: net namespace | |
1486 | * @family: protocol family (AF_INET, ...) | |
1487 | * @type: communication type (SOCK_STREAM, ...) | |
1488 | * @protocol: protocol (0, ...) | |
1489 | * @res: new socket | |
1490 | * @kern: boolean for kernel space sockets | |
1491 | * | |
1492 | * Creates a new socket and assigns it to @res, passing through LSM. | |
1493 | * Returns 0 or an error. On failure @res is set to %NULL. @kern must | |
1494 | * be set to true if the socket resides in kernel space. | |
1495 | * This function internally uses GFP_KERNEL. | |
1496 | */ | |
1497 | ||
721db93a | 1498 | int __sock_create(struct net *net, int family, int type, int protocol, |
89bddce5 | 1499 | struct socket **res, int kern) |
1da177e4 LT |
1500 | { |
1501 | int err; | |
1502 | struct socket *sock; | |
55737fda | 1503 | const struct net_proto_family *pf; |
1da177e4 LT |
1504 | |
1505 | /* | |
89bddce5 | 1506 | * Check protocol is in range |
1da177e4 LT |
1507 | */ |
1508 | if (family < 0 || family >= NPROTO) | |
1509 | return -EAFNOSUPPORT; | |
1510 | if (type < 0 || type >= SOCK_MAX) | |
1511 | return -EINVAL; | |
1512 | ||
1513 | /* Compatibility. | |
1514 | ||
1515 | This uglymoron is moved from INET layer to here to avoid | |
1516 | deadlock in module load. | |
1517 | */ | |
1518 | if (family == PF_INET && type == SOCK_PACKET) { | |
f3c98690 | 1519 | pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", |
1520 | current->comm); | |
1da177e4 LT |
1521 | family = PF_PACKET; |
1522 | } | |
1523 | ||
1524 | err = security_socket_create(family, type, protocol, kern); | |
1525 | if (err) | |
1526 | return err; | |
89bddce5 | 1527 | |
55737fda SH |
1528 | /* |
1529 | * Allocate the socket and allow the family to set things up. if | |
1530 | * the protocol is 0, the family is instructed to select an appropriate | |
1531 | * default. | |
1532 | */ | |
1533 | sock = sock_alloc(); | |
1534 | if (!sock) { | |
e87cc472 | 1535 | net_warn_ratelimited("socket: no more sockets\n"); |
55737fda SH |
1536 | return -ENFILE; /* Not exactly a match, but its the |
1537 | closest posix thing */ | |
1538 | } | |
1539 | ||
1540 | sock->type = type; | |
1541 | ||
95a5afca | 1542 | #ifdef CONFIG_MODULES |
89bddce5 SH |
1543 | /* Attempt to load a protocol module if the find failed. |
1544 | * | |
1545 | * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user | |
1da177e4 LT |
1546 | * requested real, full-featured networking support upon configuration. |
1547 | * Otherwise module support will break! | |
1548 | */ | |
190683a9 | 1549 | if (rcu_access_pointer(net_families[family]) == NULL) |
89bddce5 | 1550 | request_module("net-pf-%d", family); |
1da177e4 LT |
1551 | #endif |
1552 | ||
55737fda SH |
1553 | rcu_read_lock(); |
1554 | pf = rcu_dereference(net_families[family]); | |
1555 | err = -EAFNOSUPPORT; | |
1556 | if (!pf) | |
1557 | goto out_release; | |
1da177e4 LT |
1558 | |
1559 | /* | |
1560 | * We will call the ->create function, that possibly is in a loadable | |
1561 | * module, so we have to bump that loadable module refcnt first. | |
1562 | */ | |
55737fda | 1563 | if (!try_module_get(pf->owner)) |
1da177e4 LT |
1564 | goto out_release; |
1565 | ||
55737fda SH |
1566 | /* Now protected by module ref count */ |
1567 | rcu_read_unlock(); | |
1568 | ||
3f378b68 | 1569 | err = pf->create(net, sock, protocol, kern); |
55737fda | 1570 | if (err < 0) |
1da177e4 | 1571 | goto out_module_put; |
a79af59e | 1572 | |
1da177e4 LT |
1573 | /* |
1574 | * Now to bump the refcnt of the [loadable] module that owns this | |
1575 | * socket at sock_release time we decrement its refcnt. | |
1576 | */ | |
55737fda SH |
1577 | if (!try_module_get(sock->ops->owner)) |
1578 | goto out_module_busy; | |
1579 | ||
1da177e4 LT |
1580 | /* |
1581 | * Now that we're done with the ->create function, the [loadable] | |
1582 | * module can have its refcnt decremented | |
1583 | */ | |
55737fda | 1584 | module_put(pf->owner); |
7420ed23 VY |
1585 | err = security_socket_post_create(sock, family, type, protocol, kern); |
1586 | if (err) | |
3b185525 | 1587 | goto out_sock_release; |
55737fda | 1588 | *res = sock; |
1da177e4 | 1589 | |
55737fda SH |
1590 | return 0; |
1591 | ||
1592 | out_module_busy: | |
1593 | err = -EAFNOSUPPORT; | |
1da177e4 | 1594 | out_module_put: |
55737fda SH |
1595 | sock->ops = NULL; |
1596 | module_put(pf->owner); | |
1597 | out_sock_release: | |
1da177e4 | 1598 | sock_release(sock); |
55737fda SH |
1599 | return err; |
1600 | ||
1601 | out_release: | |
1602 | rcu_read_unlock(); | |
1603 | goto out_sock_release; | |
1da177e4 | 1604 | } |
721db93a | 1605 | EXPORT_SYMBOL(__sock_create); |
1da177e4 | 1606 | |
8a3c245c PT |
1607 | /** |
1608 | * sock_create - creates a socket | |
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 | ||
1da177e4 LT |
1618 | int sock_create(int family, int type, int protocol, struct socket **res) |
1619 | { | |
1b8d7ae4 | 1620 | return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); |
1da177e4 | 1621 | } |
c6d409cf | 1622 | EXPORT_SYMBOL(sock_create); |
1da177e4 | 1623 | |
8a3c245c PT |
1624 | /** |
1625 | * sock_create_kern - creates a socket (kernel space) | |
1626 | * @net: net namespace | |
1627 | * @family: protocol family (AF_INET, ...) | |
1628 | * @type: communication type (SOCK_STREAM, ...) | |
1629 | * @protocol: protocol (0, ...) | |
1630 | * @res: new socket | |
1631 | * | |
1632 | * A wrapper around __sock_create(). | |
1633 | * Returns 0 or an error. This function internally uses GFP_KERNEL. | |
1634 | */ | |
1635 | ||
eeb1bd5c | 1636 | int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) |
1da177e4 | 1637 | { |
eeb1bd5c | 1638 | return __sock_create(net, family, type, protocol, res, 1); |
1da177e4 | 1639 | } |
c6d409cf | 1640 | EXPORT_SYMBOL(sock_create_kern); |
1da177e4 | 1641 | |
da214a47 | 1642 | static struct socket *__sys_socket_create(int family, int type, int protocol) |
1da177e4 | 1643 | { |
1da177e4 | 1644 | struct socket *sock; |
da214a47 | 1645 | int retval; |
a677a039 | 1646 | |
e38b36f3 UD |
1647 | /* Check the SOCK_* constants for consistency. */ |
1648 | BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); | |
1649 | BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); | |
1650 | BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); | |
1651 | BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); | |
1652 | ||
da214a47 JA |
1653 | if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
1654 | return ERR_PTR(-EINVAL); | |
a677a039 | 1655 | type &= SOCK_TYPE_MASK; |
1da177e4 | 1656 | |
da214a47 JA |
1657 | retval = sock_create(family, type, protocol, &sock); |
1658 | if (retval < 0) | |
1659 | return ERR_PTR(retval); | |
1660 | ||
1661 | return sock; | |
1662 | } | |
1663 | ||
1664 | struct file *__sys_socket_file(int family, int type, int protocol) | |
1665 | { | |
1666 | struct socket *sock; | |
da214a47 JA |
1667 | int flags; |
1668 | ||
1669 | sock = __sys_socket_create(family, type, protocol); | |
1670 | if (IS_ERR(sock)) | |
1671 | return ERR_CAST(sock); | |
1672 | ||
1673 | flags = type & ~SOCK_TYPE_MASK; | |
aaca0bdc UD |
1674 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1675 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1676 | ||
649c15c7 | 1677 | return sock_alloc_file(sock, flags, NULL); |
da214a47 JA |
1678 | } |
1679 | ||
0dd061a6 GT |
1680 | /* A hook for bpf progs to attach to and update socket protocol. |
1681 | * | |
1682 | * A static noinline declaration here could cause the compiler to | |
1683 | * optimize away the function. A global noinline declaration will | |
1684 | * keep the definition, but may optimize away the callsite. | |
1685 | * Therefore, __weak is needed to ensure that the call is still | |
1686 | * emitted, by telling the compiler that we don't know what the | |
1687 | * function might eventually be. | |
0dd061a6 GT |
1688 | */ |
1689 | ||
15fb6f2b | 1690 | __bpf_hook_start(); |
0dd061a6 GT |
1691 | |
1692 | __weak noinline int update_socket_protocol(int family, int type, int protocol) | |
1693 | { | |
1694 | return protocol; | |
1695 | } | |
1696 | ||
15fb6f2b | 1697 | __bpf_hook_end(); |
0dd061a6 | 1698 | |
da214a47 JA |
1699 | int __sys_socket(int family, int type, int protocol) |
1700 | { | |
1701 | struct socket *sock; | |
1702 | int flags; | |
1703 | ||
0dd061a6 GT |
1704 | sock = __sys_socket_create(family, type, |
1705 | update_socket_protocol(family, type, protocol)); | |
da214a47 JA |
1706 | if (IS_ERR(sock)) |
1707 | return PTR_ERR(sock); | |
1708 | ||
1709 | flags = type & ~SOCK_TYPE_MASK; | |
1710 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) | |
1711 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1da177e4 | 1712 | |
8e1611e2 | 1713 | return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); |
1da177e4 LT |
1714 | } |
1715 | ||
9d6a15c3 DB |
1716 | SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) |
1717 | { | |
1718 | return __sys_socket(family, type, protocol); | |
1719 | } | |
1720 | ||
1da177e4 LT |
1721 | /* |
1722 | * Create a pair of connected sockets. | |
1723 | */ | |
1724 | ||
6debc8d8 | 1725 | int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) |
1da177e4 LT |
1726 | { |
1727 | struct socket *sock1, *sock2; | |
1728 | int fd1, fd2, err; | |
db349509 | 1729 | struct file *newfile1, *newfile2; |
a677a039 UD |
1730 | int flags; |
1731 | ||
1732 | flags = type & ~SOCK_TYPE_MASK; | |
77d27200 | 1733 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) |
a677a039 UD |
1734 | return -EINVAL; |
1735 | type &= SOCK_TYPE_MASK; | |
1da177e4 | 1736 | |
aaca0bdc UD |
1737 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) |
1738 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
1739 | ||
016a266b AV |
1740 | /* |
1741 | * reserve descriptors and make sure we won't fail | |
1742 | * to return them to userland. | |
1743 | */ | |
1744 | fd1 = get_unused_fd_flags(flags); | |
1745 | if (unlikely(fd1 < 0)) | |
1746 | return fd1; | |
1747 | ||
1748 | fd2 = get_unused_fd_flags(flags); | |
1749 | if (unlikely(fd2 < 0)) { | |
1750 | put_unused_fd(fd1); | |
1751 | return fd2; | |
1752 | } | |
1753 | ||
1754 | err = put_user(fd1, &usockvec[0]); | |
1755 | if (err) | |
1756 | goto out; | |
1757 | ||
1758 | err = put_user(fd2, &usockvec[1]); | |
1759 | if (err) | |
1760 | goto out; | |
1761 | ||
1da177e4 LT |
1762 | /* |
1763 | * Obtain the first socket and check if the underlying protocol | |
1764 | * supports the socketpair call. | |
1765 | */ | |
1766 | ||
1767 | err = sock_create(family, type, protocol, &sock1); | |
016a266b | 1768 | if (unlikely(err < 0)) |
1da177e4 LT |
1769 | goto out; |
1770 | ||
1771 | err = sock_create(family, type, protocol, &sock2); | |
016a266b AV |
1772 | if (unlikely(err < 0)) { |
1773 | sock_release(sock1); | |
1774 | goto out; | |
bf3c23d1 | 1775 | } |
d73aa286 | 1776 | |
d47cd945 DH |
1777 | err = security_socket_socketpair(sock1, sock2); |
1778 | if (unlikely(err)) { | |
1779 | sock_release(sock2); | |
1780 | sock_release(sock1); | |
1781 | goto out; | |
1782 | } | |
1783 | ||
1ded5e5a | 1784 | err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2); |
016a266b AV |
1785 | if (unlikely(err < 0)) { |
1786 | sock_release(sock2); | |
1787 | sock_release(sock1); | |
1788 | goto out; | |
28407630 AV |
1789 | } |
1790 | ||
aab174f0 | 1791 | newfile1 = sock_alloc_file(sock1, flags, NULL); |
b5ffe634 | 1792 | if (IS_ERR(newfile1)) { |
28407630 | 1793 | err = PTR_ERR(newfile1); |
016a266b AV |
1794 | sock_release(sock2); |
1795 | goto out; | |
28407630 AV |
1796 | } |
1797 | ||
aab174f0 | 1798 | newfile2 = sock_alloc_file(sock2, flags, NULL); |
28407630 AV |
1799 | if (IS_ERR(newfile2)) { |
1800 | err = PTR_ERR(newfile2); | |
016a266b AV |
1801 | fput(newfile1); |
1802 | goto out; | |
db349509 AV |
1803 | } |
1804 | ||
157cf649 | 1805 | audit_fd_pair(fd1, fd2); |
d73aa286 | 1806 | |
db349509 AV |
1807 | fd_install(fd1, newfile1); |
1808 | fd_install(fd2, newfile2); | |
d73aa286 | 1809 | return 0; |
1da177e4 | 1810 | |
016a266b | 1811 | out: |
d73aa286 | 1812 | put_unused_fd(fd2); |
d73aa286 | 1813 | put_unused_fd(fd1); |
1da177e4 LT |
1814 | return err; |
1815 | } | |
1816 | ||
6debc8d8 DB |
1817 | SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, |
1818 | int __user *, usockvec) | |
1819 | { | |
1820 | return __sys_socketpair(family, type, protocol, usockvec); | |
1821 | } | |
1822 | ||
1da177e4 LT |
1823 | /* |
1824 | * Bind a name to a socket. Nothing much to do here since it's | |
1825 | * the protocol's responsibility to handle the local address. | |
1826 | * | |
1827 | * We move the socket address to kernel space before we call | |
1828 | * the protocol layer (having also checked the address is ok). | |
1829 | */ | |
1830 | ||
a87d35d8 | 1831 | int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) |
1da177e4 LT |
1832 | { |
1833 | struct socket *sock; | |
230b1839 | 1834 | struct sockaddr_storage address; |
6cb153ca | 1835 | int err, fput_needed; |
1da177e4 | 1836 | |
89bddce5 | 1837 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
e71a4783 | 1838 | if (sock) { |
43db362d | 1839 | err = move_addr_to_kernel(umyaddr, addrlen, &address); |
068b88cc | 1840 | if (!err) { |
89bddce5 | 1841 | err = security_socket_bind(sock, |
230b1839 | 1842 | (struct sockaddr *)&address, |
89bddce5 | 1843 | addrlen); |
6cb153ca | 1844 | if (!err) |
1ded5e5a | 1845 | err = READ_ONCE(sock->ops)->bind(sock, |
89bddce5 | 1846 | (struct sockaddr *) |
230b1839 | 1847 | &address, addrlen); |
1da177e4 | 1848 | } |
6cb153ca | 1849 | fput_light(sock->file, fput_needed); |
89bddce5 | 1850 | } |
1da177e4 LT |
1851 | return err; |
1852 | } | |
1853 | ||
a87d35d8 DB |
1854 | SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) |
1855 | { | |
1856 | return __sys_bind(fd, umyaddr, addrlen); | |
1857 | } | |
1858 | ||
1da177e4 LT |
1859 | /* |
1860 | * Perform a listen. Basically, we allow the protocol to do anything | |
1861 | * necessary for a listen, and if that works, we mark the socket as | |
1862 | * ready for listening. | |
1863 | */ | |
1864 | ||
25e290ee | 1865 | int __sys_listen(int fd, int backlog) |
1da177e4 LT |
1866 | { |
1867 | struct socket *sock; | |
6cb153ca | 1868 | int err, fput_needed; |
b8e1f9b5 | 1869 | int somaxconn; |
89bddce5 SH |
1870 | |
1871 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
1872 | if (sock) { | |
3c9ba81d | 1873 | somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn); |
95c96174 | 1874 | if ((unsigned int)backlog > somaxconn) |
b8e1f9b5 | 1875 | backlog = somaxconn; |
1da177e4 LT |
1876 | |
1877 | err = security_socket_listen(sock, backlog); | |
6cb153ca | 1878 | if (!err) |
1ded5e5a | 1879 | err = READ_ONCE(sock->ops)->listen(sock, backlog); |
1da177e4 | 1880 | |
6cb153ca | 1881 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
1882 | } |
1883 | return err; | |
1884 | } | |
1885 | ||
25e290ee DB |
1886 | SYSCALL_DEFINE2(listen, int, fd, int, backlog) |
1887 | { | |
1888 | return __sys_listen(fd, backlog); | |
1889 | } | |
1890 | ||
d32f89da | 1891 | struct file *do_accept(struct file *file, unsigned file_flags, |
de2ea4b6 | 1892 | struct sockaddr __user *upeer_sockaddr, |
d32f89da | 1893 | int __user *upeer_addrlen, int flags) |
1da177e4 LT |
1894 | { |
1895 | struct socket *sock, *newsock; | |
39d8c1b6 | 1896 | struct file *newfile; |
d32f89da | 1897 | int err, len; |
230b1839 | 1898 | struct sockaddr_storage address; |
1ded5e5a | 1899 | const struct proto_ops *ops; |
1da177e4 | 1900 | |
dba4a925 | 1901 | sock = sock_from_file(file); |
d32f89da PB |
1902 | if (!sock) |
1903 | return ERR_PTR(-ENOTSOCK); | |
1da177e4 | 1904 | |
c6d409cf ED |
1905 | newsock = sock_alloc(); |
1906 | if (!newsock) | |
d32f89da | 1907 | return ERR_PTR(-ENFILE); |
1ded5e5a | 1908 | ops = READ_ONCE(sock->ops); |
1da177e4 LT |
1909 | |
1910 | newsock->type = sock->type; | |
1ded5e5a | 1911 | newsock->ops = ops; |
1da177e4 | 1912 | |
1da177e4 LT |
1913 | /* |
1914 | * We don't need try_module_get here, as the listening socket (sock) | |
1915 | * has the protocol module (sock->ops->owner) held. | |
1916 | */ | |
1ded5e5a | 1917 | __module_get(ops->owner); |
1da177e4 | 1918 | |
aab174f0 | 1919 | newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); |
d32f89da PB |
1920 | if (IS_ERR(newfile)) |
1921 | return newfile; | |
39d8c1b6 | 1922 | |
a79af59e FF |
1923 | err = security_socket_accept(sock, newsock); |
1924 | if (err) | |
39d8c1b6 | 1925 | goto out_fd; |
a79af59e | 1926 | |
1ded5e5a | 1927 | err = ops->accept(sock, newsock, sock->file->f_flags | file_flags, |
de2ea4b6 | 1928 | false); |
1da177e4 | 1929 | if (err < 0) |
39d8c1b6 | 1930 | goto out_fd; |
1da177e4 LT |
1931 | |
1932 | if (upeer_sockaddr) { | |
1ded5e5a | 1933 | len = ops->getname(newsock, (struct sockaddr *)&address, 2); |
9b2c45d4 | 1934 | if (len < 0) { |
1da177e4 | 1935 | err = -ECONNABORTED; |
39d8c1b6 | 1936 | goto out_fd; |
1da177e4 | 1937 | } |
43db362d | 1938 | err = move_addr_to_user(&address, |
230b1839 | 1939 | len, upeer_sockaddr, upeer_addrlen); |
1da177e4 | 1940 | if (err < 0) |
39d8c1b6 | 1941 | goto out_fd; |
1da177e4 LT |
1942 | } |
1943 | ||
1944 | /* File flags are not inherited via accept() unlike another OSes. */ | |
d32f89da | 1945 | return newfile; |
39d8c1b6 | 1946 | out_fd: |
9606a216 | 1947 | fput(newfile); |
d32f89da PB |
1948 | return ERR_PTR(err); |
1949 | } | |
1950 | ||
c0424532 YD |
1951 | static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr, |
1952 | int __user *upeer_addrlen, int flags) | |
d32f89da PB |
1953 | { |
1954 | struct file *newfile; | |
1955 | int newfd; | |
1956 | ||
1957 | if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) | |
1958 | return -EINVAL; | |
1959 | ||
1960 | if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) | |
1961 | flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; | |
de2ea4b6 | 1962 | |
c0424532 | 1963 | newfd = get_unused_fd_flags(flags); |
d32f89da PB |
1964 | if (unlikely(newfd < 0)) |
1965 | return newfd; | |
1966 | ||
c0424532 | 1967 | newfile = do_accept(file, 0, upeer_sockaddr, upeer_addrlen, |
d32f89da PB |
1968 | flags); |
1969 | if (IS_ERR(newfile)) { | |
1970 | put_unused_fd(newfd); | |
1971 | return PTR_ERR(newfile); | |
1972 | } | |
1973 | fd_install(newfd, newfile); | |
1974 | return newfd; | |
de2ea4b6 JA |
1975 | } |
1976 | ||
1977 | /* | |
1978 | * For accept, we attempt to create a new socket, set up the link | |
1979 | * with the client, wake up the client, then return the new | |
1980 | * connected fd. We collect the address of the connector in kernel | |
1981 | * space and move it to user at the very end. This is unclean because | |
1982 | * we open the socket then return an error. | |
1983 | * | |
1984 | * 1003.1g adds the ability to recvmsg() to query connection pending | |
1985 | * status to recvmsg. We need to add that support in a way thats | |
1986 | * clean when we restructure accept also. | |
1987 | */ | |
1988 | ||
1989 | int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, | |
1990 | int __user *upeer_addrlen, int flags) | |
1991 | { | |
1992 | int ret = -EBADF; | |
1993 | struct fd f; | |
1994 | ||
1995 | f = fdget(fd); | |
1996 | if (f.file) { | |
c0424532 YD |
1997 | ret = __sys_accept4_file(f.file, upeer_sockaddr, |
1998 | upeer_addrlen, flags); | |
6b07edeb | 1999 | fdput(f); |
de2ea4b6 JA |
2000 | } |
2001 | ||
2002 | return ret; | |
1da177e4 LT |
2003 | } |
2004 | ||
4541e805 DB |
2005 | SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, |
2006 | int __user *, upeer_addrlen, int, flags) | |
2007 | { | |
2008 | return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); | |
2009 | } | |
2010 | ||
20f37034 HC |
2011 | SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, |
2012 | int __user *, upeer_addrlen) | |
aaca0bdc | 2013 | { |
4541e805 | 2014 | return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); |
aaca0bdc UD |
2015 | } |
2016 | ||
1da177e4 LT |
2017 | /* |
2018 | * Attempt to connect to a socket with the server address. The address | |
2019 | * is in user space so we verify it is OK and move it to kernel space. | |
2020 | * | |
2021 | * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to | |
2022 | * break bindings | |
2023 | * | |
2024 | * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and | |
2025 | * other SEQPACKET protocols that take time to connect() as it doesn't | |
2026 | * include the -EINPROGRESS status for such sockets. | |
2027 | */ | |
2028 | ||
f499a021 | 2029 | int __sys_connect_file(struct file *file, struct sockaddr_storage *address, |
bd3ded31 | 2030 | int addrlen, int file_flags) |
1da177e4 LT |
2031 | { |
2032 | struct socket *sock; | |
bd3ded31 | 2033 | int err; |
1da177e4 | 2034 | |
dba4a925 FR |
2035 | sock = sock_from_file(file); |
2036 | if (!sock) { | |
2037 | err = -ENOTSOCK; | |
1da177e4 | 2038 | goto out; |
dba4a925 | 2039 | } |
1da177e4 | 2040 | |
89bddce5 | 2041 | err = |
f499a021 | 2042 | security_socket_connect(sock, (struct sockaddr *)address, addrlen); |
1da177e4 | 2043 | if (err) |
bd3ded31 | 2044 | goto out; |
1da177e4 | 2045 | |
1ded5e5a ED |
2046 | err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address, |
2047 | addrlen, sock->file->f_flags | file_flags); | |
1da177e4 LT |
2048 | out: |
2049 | return err; | |
2050 | } | |
2051 | ||
bd3ded31 JA |
2052 | int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) |
2053 | { | |
2054 | int ret = -EBADF; | |
2055 | struct fd f; | |
2056 | ||
2057 | f = fdget(fd); | |
2058 | if (f.file) { | |
f499a021 JA |
2059 | struct sockaddr_storage address; |
2060 | ||
2061 | ret = move_addr_to_kernel(uservaddr, addrlen, &address); | |
2062 | if (!ret) | |
2063 | ret = __sys_connect_file(f.file, &address, addrlen, 0); | |
6b07edeb | 2064 | fdput(f); |
bd3ded31 JA |
2065 | } |
2066 | ||
2067 | return ret; | |
2068 | } | |
2069 | ||
1387c2c2 DB |
2070 | SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, |
2071 | int, addrlen) | |
2072 | { | |
2073 | return __sys_connect(fd, uservaddr, addrlen); | |
2074 | } | |
2075 | ||
1da177e4 LT |
2076 | /* |
2077 | * Get the local address ('name') of a socket object. Move the obtained | |
2078 | * name to user space. | |
2079 | */ | |
2080 | ||
8882a107 DB |
2081 | int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, |
2082 | int __user *usockaddr_len) | |
1da177e4 LT |
2083 | { |
2084 | struct socket *sock; | |
230b1839 | 2085 | struct sockaddr_storage address; |
9b2c45d4 | 2086 | int err, fput_needed; |
89bddce5 | 2087 | |
6cb153ca | 2088 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 LT |
2089 | if (!sock) |
2090 | goto out; | |
2091 | ||
2092 | err = security_socket_getsockname(sock); | |
2093 | if (err) | |
2094 | goto out_put; | |
2095 | ||
1ded5e5a | 2096 | err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0); |
9b2c45d4 | 2097 | if (err < 0) |
1da177e4 | 2098 | goto out_put; |
e44ef1d4 | 2099 | /* "err" is actually length in this case */ |
9b2c45d4 | 2100 | err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); |
1da177e4 LT |
2101 | |
2102 | out_put: | |
6cb153ca | 2103 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2104 | out: |
2105 | return err; | |
2106 | } | |
2107 | ||
8882a107 DB |
2108 | SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, |
2109 | int __user *, usockaddr_len) | |
2110 | { | |
2111 | return __sys_getsockname(fd, usockaddr, usockaddr_len); | |
2112 | } | |
2113 | ||
1da177e4 LT |
2114 | /* |
2115 | * Get the remote address ('name') of a socket object. Move the obtained | |
2116 | * name to user space. | |
2117 | */ | |
2118 | ||
b21c8f83 DB |
2119 | int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, |
2120 | int __user *usockaddr_len) | |
1da177e4 LT |
2121 | { |
2122 | struct socket *sock; | |
230b1839 | 2123 | struct sockaddr_storage address; |
9b2c45d4 | 2124 | int err, fput_needed; |
1da177e4 | 2125 | |
89bddce5 SH |
2126 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2127 | if (sock != NULL) { | |
1ded5e5a ED |
2128 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
2129 | ||
1da177e4 LT |
2130 | err = security_socket_getpeername(sock); |
2131 | if (err) { | |
6cb153ca | 2132 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2133 | return err; |
2134 | } | |
2135 | ||
1ded5e5a | 2136 | err = ops->getname(sock, (struct sockaddr *)&address, 1); |
9b2c45d4 DV |
2137 | if (err >= 0) |
2138 | /* "err" is actually length in this case */ | |
2139 | err = move_addr_to_user(&address, err, usockaddr, | |
89bddce5 | 2140 | usockaddr_len); |
6cb153ca | 2141 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2142 | } |
2143 | return err; | |
2144 | } | |
2145 | ||
b21c8f83 DB |
2146 | SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, |
2147 | int __user *, usockaddr_len) | |
2148 | { | |
2149 | return __sys_getpeername(fd, usockaddr, usockaddr_len); | |
2150 | } | |
2151 | ||
1da177e4 LT |
2152 | /* |
2153 | * Send a datagram to a given address. We move the address into kernel | |
2154 | * space and check the user space data area is readable before invoking | |
2155 | * the protocol. | |
2156 | */ | |
211b634b DB |
2157 | int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, |
2158 | struct sockaddr __user *addr, int addr_len) | |
1da177e4 LT |
2159 | { |
2160 | struct socket *sock; | |
230b1839 | 2161 | struct sockaddr_storage address; |
1da177e4 LT |
2162 | int err; |
2163 | struct msghdr msg; | |
2164 | struct iovec iov; | |
6cb153ca | 2165 | int fput_needed; |
6cb153ca | 2166 | |
de4eda9d | 2167 | err = import_single_range(ITER_SOURCE, buff, len, &iov, &msg.msg_iter); |
602bd0e9 AV |
2168 | if (unlikely(err)) |
2169 | return err; | |
de0fa95c PE |
2170 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2171 | if (!sock) | |
4387ff75 | 2172 | goto out; |
6cb153ca | 2173 | |
89bddce5 | 2174 | msg.msg_name = NULL; |
89bddce5 SH |
2175 | msg.msg_control = NULL; |
2176 | msg.msg_controllen = 0; | |
2177 | msg.msg_namelen = 0; | |
7c701d92 | 2178 | msg.msg_ubuf = NULL; |
6cb153ca | 2179 | if (addr) { |
43db362d | 2180 | err = move_addr_to_kernel(addr, addr_len, &address); |
1da177e4 LT |
2181 | if (err < 0) |
2182 | goto out_put; | |
230b1839 | 2183 | msg.msg_name = (struct sockaddr *)&address; |
89bddce5 | 2184 | msg.msg_namelen = addr_len; |
1da177e4 | 2185 | } |
b841b901 | 2186 | flags &= ~MSG_INTERNAL_SENDMSG_FLAGS; |
1da177e4 LT |
2187 | if (sock->file->f_flags & O_NONBLOCK) |
2188 | flags |= MSG_DONTWAIT; | |
2189 | msg.msg_flags = flags; | |
86a7e0b6 | 2190 | err = __sock_sendmsg(sock, &msg); |
1da177e4 | 2191 | |
89bddce5 | 2192 | out_put: |
de0fa95c | 2193 | fput_light(sock->file, fput_needed); |
4387ff75 | 2194 | out: |
1da177e4 LT |
2195 | return err; |
2196 | } | |
2197 | ||
211b634b DB |
2198 | SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, |
2199 | unsigned int, flags, struct sockaddr __user *, addr, | |
2200 | int, addr_len) | |
2201 | { | |
2202 | return __sys_sendto(fd, buff, len, flags, addr, addr_len); | |
2203 | } | |
2204 | ||
1da177e4 | 2205 | /* |
89bddce5 | 2206 | * Send a datagram down a socket. |
1da177e4 LT |
2207 | */ |
2208 | ||
3e0fa65f | 2209 | SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, |
95c96174 | 2210 | unsigned int, flags) |
1da177e4 | 2211 | { |
211b634b | 2212 | return __sys_sendto(fd, buff, len, flags, NULL, 0); |
1da177e4 LT |
2213 | } |
2214 | ||
2215 | /* | |
89bddce5 | 2216 | * Receive a frame from the socket and optionally record the address of the |
1da177e4 LT |
2217 | * sender. We verify the buffers are writable and if needed move the |
2218 | * sender address from kernel to user space. | |
2219 | */ | |
7a09e1eb DB |
2220 | int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, |
2221 | struct sockaddr __user *addr, int __user *addr_len) | |
1da177e4 | 2222 | { |
1228b34c ED |
2223 | struct sockaddr_storage address; |
2224 | struct msghdr msg = { | |
2225 | /* Save some cycles and don't copy the address if not needed */ | |
2226 | .msg_name = addr ? (struct sockaddr *)&address : NULL, | |
2227 | }; | |
1da177e4 LT |
2228 | struct socket *sock; |
2229 | struct iovec iov; | |
89bddce5 | 2230 | int err, err2; |
6cb153ca BL |
2231 | int fput_needed; |
2232 | ||
de4eda9d | 2233 | err = import_single_range(ITER_DEST, ubuf, size, &iov, &msg.msg_iter); |
602bd0e9 AV |
2234 | if (unlikely(err)) |
2235 | return err; | |
de0fa95c | 2236 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
1da177e4 | 2237 | if (!sock) |
de0fa95c | 2238 | goto out; |
1da177e4 | 2239 | |
1da177e4 LT |
2240 | if (sock->file->f_flags & O_NONBLOCK) |
2241 | flags |= MSG_DONTWAIT; | |
2da62906 | 2242 | err = sock_recvmsg(sock, &msg, flags); |
1da177e4 | 2243 | |
89bddce5 | 2244 | if (err >= 0 && addr != NULL) { |
43db362d | 2245 | err2 = move_addr_to_user(&address, |
230b1839 | 2246 | msg.msg_namelen, addr, addr_len); |
89bddce5 SH |
2247 | if (err2 < 0) |
2248 | err = err2; | |
1da177e4 | 2249 | } |
de0fa95c PE |
2250 | |
2251 | fput_light(sock->file, fput_needed); | |
4387ff75 | 2252 | out: |
1da177e4 LT |
2253 | return err; |
2254 | } | |
2255 | ||
7a09e1eb DB |
2256 | SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, |
2257 | unsigned int, flags, struct sockaddr __user *, addr, | |
2258 | int __user *, addr_len) | |
2259 | { | |
2260 | return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); | |
2261 | } | |
2262 | ||
1da177e4 | 2263 | /* |
89bddce5 | 2264 | * Receive a datagram from a socket. |
1da177e4 LT |
2265 | */ |
2266 | ||
b7c0ddf5 JG |
2267 | SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, |
2268 | unsigned int, flags) | |
1da177e4 | 2269 | { |
7a09e1eb | 2270 | return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); |
1da177e4 LT |
2271 | } |
2272 | ||
83f0c10b FW |
2273 | static bool sock_use_custom_sol_socket(const struct socket *sock) |
2274 | { | |
a5ef058d | 2275 | return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags); |
83f0c10b FW |
2276 | } |
2277 | ||
1406245c BL |
2278 | int do_sock_setsockopt(struct socket *sock, bool compat, int level, |
2279 | int optname, sockptr_t optval, int optlen) | |
1da177e4 | 2280 | { |
1ded5e5a | 2281 | const struct proto_ops *ops; |
0d01da6a | 2282 | char *kernel_optval = NULL; |
1406245c | 2283 | int err; |
1da177e4 LT |
2284 | |
2285 | if (optlen < 0) | |
2286 | return -EINVAL; | |
89bddce5 | 2287 | |
4a367299 CH |
2288 | err = security_socket_setsockopt(sock, level, optname); |
2289 | if (err) | |
2290 | goto out_put; | |
0d01da6a | 2291 | |
1406245c | 2292 | if (!compat) |
55db9c0e | 2293 | err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname, |
3f31e0d1 | 2294 | optval, &optlen, |
55db9c0e | 2295 | &kernel_optval); |
4a367299 CH |
2296 | if (err < 0) |
2297 | goto out_put; | |
2298 | if (err > 0) { | |
2299 | err = 0; | |
2300 | goto out_put; | |
2301 | } | |
0d01da6a | 2302 | |
a7b75c5a CH |
2303 | if (kernel_optval) |
2304 | optval = KERNEL_SOCKPTR(kernel_optval); | |
1ded5e5a | 2305 | ops = READ_ONCE(sock->ops); |
4a367299 | 2306 | if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock)) |
a7b75c5a | 2307 | err = sock_setsockopt(sock, level, optname, optval, optlen); |
1ded5e5a | 2308 | else if (unlikely(!ops->setsockopt)) |
a44d9e72 | 2309 | err = -EOPNOTSUPP; |
4a367299 | 2310 | else |
1ded5e5a | 2311 | err = ops->setsockopt(sock, level, optname, optval, |
89bddce5 | 2312 | optlen); |
a7b75c5a | 2313 | kfree(kernel_optval); |
4a367299 | 2314 | out_put: |
1406245c BL |
2315 | return err; |
2316 | } | |
2317 | EXPORT_SYMBOL(do_sock_setsockopt); | |
2318 | ||
2319 | /* Set a socket option. Because we don't know the option lengths we have | |
2320 | * to pass the user mode parameter for the protocols to sort out. | |
2321 | */ | |
2322 | int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval, | |
2323 | int optlen) | |
2324 | { | |
2325 | sockptr_t optval = USER_SOCKPTR(user_optval); | |
2326 | bool compat = in_compat_syscall(); | |
2327 | int err, fput_needed; | |
2328 | struct socket *sock; | |
2329 | ||
2330 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
2331 | if (!sock) | |
2332 | return err; | |
2333 | ||
2334 | err = do_sock_setsockopt(sock, compat, level, optname, optval, optlen); | |
2335 | ||
4a367299 | 2336 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2337 | return err; |
2338 | } | |
2339 | ||
cc36dca0 DB |
2340 | SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, |
2341 | char __user *, optval, int, optlen) | |
2342 | { | |
2343 | return __sys_setsockopt(fd, level, optname, optval, optlen); | |
2344 | } | |
2345 | ||
9cacf81f SF |
2346 | INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level, |
2347 | int optname)); | |
2348 | ||
0b05b0cd BL |
2349 | int do_sock_getsockopt(struct socket *sock, bool compat, int level, |
2350 | int optname, sockptr_t optval, sockptr_t optlen) | |
1da177e4 | 2351 | { |
ad4bf5f2 | 2352 | int max_optlen __maybe_unused; |
1ded5e5a | 2353 | const struct proto_ops *ops; |
0b05b0cd | 2354 | int err; |
d8a9b38f CH |
2355 | |
2356 | err = security_socket_getsockopt(sock, level, optname); | |
2357 | if (err) | |
0b05b0cd | 2358 | return err; |
1da177e4 | 2359 | |
0b05b0cd | 2360 | if (!compat) |
55db9c0e | 2361 | max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen); |
0d01da6a | 2362 | |
1ded5e5a | 2363 | ops = READ_ONCE(sock->ops); |
0b05b0cd BL |
2364 | if (level == SOL_SOCKET) { |
2365 | err = sk_getsockopt(sock->sk, level, optname, optval, optlen); | |
2366 | } else if (unlikely(!ops->getsockopt)) { | |
a44d9e72 | 2367 | err = -EOPNOTSUPP; |
0b05b0cd BL |
2368 | } else { |
2369 | if (WARN_ONCE(optval.is_kernel || optlen.is_kernel, | |
2370 | "Invalid argument type")) | |
2371 | return -EOPNOTSUPP; | |
2372 | ||
2373 | err = ops->getsockopt(sock, level, optname, optval.user, | |
2374 | optlen.user); | |
2375 | } | |
0d01da6a | 2376 | |
0b05b0cd | 2377 | if (!compat) |
55db9c0e CH |
2378 | err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname, |
2379 | optval, optlen, max_optlen, | |
2380 | err); | |
0b05b0cd BL |
2381 | |
2382 | return err; | |
2383 | } | |
2384 | EXPORT_SYMBOL(do_sock_getsockopt); | |
2385 | ||
1da177e4 LT |
2386 | /* |
2387 | * Get a socket option. Because we don't know the option lengths we have | |
2388 | * to pass a user mode parameter for the protocols to sort out. | |
2389 | */ | |
55db9c0e CH |
2390 | int __sys_getsockopt(int fd, int level, int optname, char __user *optval, |
2391 | int __user *optlen) | |
1da177e4 | 2392 | { |
6cb153ca | 2393 | int err, fput_needed; |
1da177e4 | 2394 | struct socket *sock; |
0b05b0cd | 2395 | bool compat; |
1da177e4 | 2396 | |
89bddce5 | 2397 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
d8a9b38f CH |
2398 | if (!sock) |
2399 | return err; | |
2400 | ||
0b05b0cd BL |
2401 | compat = in_compat_syscall(); |
2402 | err = do_sock_getsockopt(sock, compat, level, optname, | |
2403 | USER_SOCKPTR(optval), USER_SOCKPTR(optlen)); | |
1da177e4 | 2404 | |
d8a9b38f | 2405 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2406 | return err; |
2407 | } | |
2408 | ||
13a2d70e DB |
2409 | SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, |
2410 | char __user *, optval, int __user *, optlen) | |
2411 | { | |
2412 | return __sys_getsockopt(fd, level, optname, optval, optlen); | |
2413 | } | |
2414 | ||
1da177e4 LT |
2415 | /* |
2416 | * Shutdown a socket. | |
2417 | */ | |
2418 | ||
b713c195 JA |
2419 | int __sys_shutdown_sock(struct socket *sock, int how) |
2420 | { | |
2421 | int err; | |
2422 | ||
2423 | err = security_socket_shutdown(sock, how); | |
2424 | if (!err) | |
1ded5e5a | 2425 | err = READ_ONCE(sock->ops)->shutdown(sock, how); |
b713c195 JA |
2426 | |
2427 | return err; | |
2428 | } | |
2429 | ||
005a1aea | 2430 | int __sys_shutdown(int fd, int how) |
1da177e4 | 2431 | { |
6cb153ca | 2432 | int err, fput_needed; |
1da177e4 LT |
2433 | struct socket *sock; |
2434 | ||
89bddce5 SH |
2435 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
2436 | if (sock != NULL) { | |
b713c195 | 2437 | err = __sys_shutdown_sock(sock, how); |
6cb153ca | 2438 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2439 | } |
2440 | return err; | |
2441 | } | |
2442 | ||
005a1aea DB |
2443 | SYSCALL_DEFINE2(shutdown, int, fd, int, how) |
2444 | { | |
2445 | return __sys_shutdown(fd, how); | |
2446 | } | |
2447 | ||
89bddce5 | 2448 | /* A couple of helpful macros for getting the address of the 32/64 bit |
1da177e4 LT |
2449 | * fields which are the same type (int / unsigned) on our platforms. |
2450 | */ | |
2451 | #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) | |
2452 | #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) | |
2453 | #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) | |
2454 | ||
c71d8ebe TH |
2455 | struct used_address { |
2456 | struct sockaddr_storage name; | |
2457 | unsigned int name_len; | |
2458 | }; | |
2459 | ||
7fa875b8 DY |
2460 | int __copy_msghdr(struct msghdr *kmsg, |
2461 | struct user_msghdr *msg, | |
2462 | struct sockaddr __user **save_addr) | |
1661bf36 | 2463 | { |
08adb7da AV |
2464 | ssize_t err; |
2465 | ||
1f466e1f | 2466 | kmsg->msg_control_is_user = true; |
1228b34c | 2467 | kmsg->msg_get_inq = 0; |
7fa875b8 DY |
2468 | kmsg->msg_control_user = msg->msg_control; |
2469 | kmsg->msg_controllen = msg->msg_controllen; | |
2470 | kmsg->msg_flags = msg->msg_flags; | |
ffb07550 | 2471 | |
7fa875b8 DY |
2472 | kmsg->msg_namelen = msg->msg_namelen; |
2473 | if (!msg->msg_name) | |
6a2a2b3a AS |
2474 | kmsg->msg_namelen = 0; |
2475 | ||
dbb490b9 ML |
2476 | if (kmsg->msg_namelen < 0) |
2477 | return -EINVAL; | |
2478 | ||
1661bf36 | 2479 | if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) |
db31c55a | 2480 | kmsg->msg_namelen = sizeof(struct sockaddr_storage); |
08adb7da AV |
2481 | |
2482 | if (save_addr) | |
7fa875b8 | 2483 | *save_addr = msg->msg_name; |
08adb7da | 2484 | |
7fa875b8 | 2485 | if (msg->msg_name && kmsg->msg_namelen) { |
08adb7da | 2486 | if (!save_addr) { |
7fa875b8 | 2487 | err = move_addr_to_kernel(msg->msg_name, |
864d9664 | 2488 | kmsg->msg_namelen, |
08adb7da AV |
2489 | kmsg->msg_name); |
2490 | if (err < 0) | |
2491 | return err; | |
2492 | } | |
2493 | } else { | |
2494 | kmsg->msg_name = NULL; | |
2495 | kmsg->msg_namelen = 0; | |
2496 | } | |
2497 | ||
7fa875b8 | 2498 | if (msg->msg_iovlen > UIO_MAXIOV) |
08adb7da AV |
2499 | return -EMSGSIZE; |
2500 | ||
0345f931 | 2501 | kmsg->msg_iocb = NULL; |
7c701d92 | 2502 | kmsg->msg_ubuf = NULL; |
0a384abf JA |
2503 | return 0; |
2504 | } | |
2505 | ||
2506 | static int copy_msghdr_from_user(struct msghdr *kmsg, | |
2507 | struct user_msghdr __user *umsg, | |
2508 | struct sockaddr __user **save_addr, | |
2509 | struct iovec **iov) | |
2510 | { | |
2511 | struct user_msghdr msg; | |
2512 | ssize_t err; | |
2513 | ||
7fa875b8 DY |
2514 | if (copy_from_user(&msg, umsg, sizeof(*umsg))) |
2515 | return -EFAULT; | |
2516 | ||
2517 | err = __copy_msghdr(kmsg, &msg, save_addr); | |
0a384abf JA |
2518 | if (err) |
2519 | return err; | |
0345f931 | 2520 | |
de4eda9d | 2521 | err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE, |
ffb07550 | 2522 | msg.msg_iov, msg.msg_iovlen, |
da184284 | 2523 | UIO_FASTIOV, iov, &kmsg->msg_iter); |
87e5e6da | 2524 | return err < 0 ? err : 0; |
1661bf36 DC |
2525 | } |
2526 | ||
4257c8ca JA |
2527 | static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, |
2528 | unsigned int flags, struct used_address *used_address, | |
2529 | unsigned int allowed_msghdr_flags) | |
1da177e4 | 2530 | { |
b9d717a7 | 2531 | unsigned char ctl[sizeof(struct cmsghdr) + 20] |
846cc123 | 2532 | __aligned(sizeof(__kernel_size_t)); |
89bddce5 | 2533 | /* 20 is size of ipv6_pktinfo */ |
1da177e4 | 2534 | unsigned char *ctl_buf = ctl; |
d8725c86 | 2535 | int ctl_len; |
08adb7da | 2536 | ssize_t err; |
89bddce5 | 2537 | |
1da177e4 LT |
2538 | err = -ENOBUFS; |
2539 | ||
228e548e | 2540 | if (msg_sys->msg_controllen > INT_MAX) |
4257c8ca | 2541 | goto out; |
28a94d8f | 2542 | flags |= (msg_sys->msg_flags & allowed_msghdr_flags); |
228e548e | 2543 | ctl_len = msg_sys->msg_controllen; |
1da177e4 | 2544 | if ((MSG_CMSG_COMPAT & flags) && ctl_len) { |
89bddce5 | 2545 | err = |
228e548e | 2546 | cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, |
89bddce5 | 2547 | sizeof(ctl)); |
1da177e4 | 2548 | if (err) |
4257c8ca | 2549 | goto out; |
228e548e AB |
2550 | ctl_buf = msg_sys->msg_control; |
2551 | ctl_len = msg_sys->msg_controllen; | |
1da177e4 | 2552 | } else if (ctl_len) { |
ac4340fc DM |
2553 | BUILD_BUG_ON(sizeof(struct cmsghdr) != |
2554 | CMSG_ALIGN(sizeof(struct cmsghdr))); | |
89bddce5 | 2555 | if (ctl_len > sizeof(ctl)) { |
1da177e4 | 2556 | ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); |
89bddce5 | 2557 | if (ctl_buf == NULL) |
4257c8ca | 2558 | goto out; |
1da177e4 LT |
2559 | } |
2560 | err = -EFAULT; | |
1f466e1f | 2561 | if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len)) |
1da177e4 | 2562 | goto out_freectl; |
228e548e | 2563 | msg_sys->msg_control = ctl_buf; |
1f466e1f | 2564 | msg_sys->msg_control_is_user = false; |
1da177e4 | 2565 | } |
b841b901 | 2566 | flags &= ~MSG_INTERNAL_SENDMSG_FLAGS; |
228e548e | 2567 | msg_sys->msg_flags = flags; |
1da177e4 LT |
2568 | |
2569 | if (sock->file->f_flags & O_NONBLOCK) | |
228e548e | 2570 | msg_sys->msg_flags |= MSG_DONTWAIT; |
c71d8ebe TH |
2571 | /* |
2572 | * If this is sendmmsg() and current destination address is same as | |
2573 | * previously succeeded address, omit asking LSM's decision. | |
2574 | * used_address->name_len is initialized to UINT_MAX so that the first | |
2575 | * destination address never matches. | |
2576 | */ | |
bc909d9d MD |
2577 | if (used_address && msg_sys->msg_name && |
2578 | used_address->name_len == msg_sys->msg_namelen && | |
2579 | !memcmp(&used_address->name, msg_sys->msg_name, | |
c71d8ebe | 2580 | used_address->name_len)) { |
d8725c86 | 2581 | err = sock_sendmsg_nosec(sock, msg_sys); |
c71d8ebe TH |
2582 | goto out_freectl; |
2583 | } | |
86a7e0b6 | 2584 | err = __sock_sendmsg(sock, msg_sys); |
c71d8ebe TH |
2585 | /* |
2586 | * If this is sendmmsg() and sending to current destination address was | |
2587 | * successful, remember it. | |
2588 | */ | |
2589 | if (used_address && err >= 0) { | |
2590 | used_address->name_len = msg_sys->msg_namelen; | |
bc909d9d MD |
2591 | if (msg_sys->msg_name) |
2592 | memcpy(&used_address->name, msg_sys->msg_name, | |
2593 | used_address->name_len); | |
c71d8ebe | 2594 | } |
1da177e4 LT |
2595 | |
2596 | out_freectl: | |
89bddce5 | 2597 | if (ctl_buf != ctl) |
1da177e4 | 2598 | sock_kfree_s(sock->sk, ctl_buf, ctl_len); |
4257c8ca JA |
2599 | out: |
2600 | return err; | |
2601 | } | |
2602 | ||
03b1230c JA |
2603 | int sendmsg_copy_msghdr(struct msghdr *msg, |
2604 | struct user_msghdr __user *umsg, unsigned flags, | |
2605 | struct iovec **iov) | |
4257c8ca JA |
2606 | { |
2607 | int err; | |
2608 | ||
2609 | if (flags & MSG_CMSG_COMPAT) { | |
2610 | struct compat_msghdr __user *msg_compat; | |
2611 | ||
2612 | msg_compat = (struct compat_msghdr __user *) umsg; | |
2613 | err = get_compat_msghdr(msg, msg_compat, NULL, iov); | |
2614 | } else { | |
2615 | err = copy_msghdr_from_user(msg, umsg, NULL, iov); | |
2616 | } | |
2617 | if (err < 0) | |
2618 | return err; | |
2619 | ||
2620 | return 0; | |
2621 | } | |
2622 | ||
2623 | static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, | |
2624 | struct msghdr *msg_sys, unsigned int flags, | |
2625 | struct used_address *used_address, | |
2626 | unsigned int allowed_msghdr_flags) | |
2627 | { | |
2628 | struct sockaddr_storage address; | |
2629 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; | |
2630 | ssize_t err; | |
2631 | ||
2632 | msg_sys->msg_name = &address; | |
2633 | ||
2634 | err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov); | |
2635 | if (err < 0) | |
2636 | return err; | |
2637 | ||
2638 | err = ____sys_sendmsg(sock, msg_sys, flags, used_address, | |
2639 | allowed_msghdr_flags); | |
da184284 | 2640 | kfree(iov); |
228e548e AB |
2641 | return err; |
2642 | } | |
2643 | ||
2644 | /* | |
2645 | * BSD sendmsg interface | |
2646 | */ | |
03b1230c | 2647 | long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg, |
0fa03c62 JA |
2648 | unsigned int flags) |
2649 | { | |
03b1230c | 2650 | return ____sys_sendmsg(sock, msg, flags, NULL, 0); |
0fa03c62 | 2651 | } |
228e548e | 2652 | |
e1834a32 DB |
2653 | long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, |
2654 | bool forbid_cmsg_compat) | |
228e548e AB |
2655 | { |
2656 | int fput_needed, err; | |
2657 | struct msghdr msg_sys; | |
1be374a0 AL |
2658 | struct socket *sock; |
2659 | ||
e1834a32 DB |
2660 | if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) |
2661 | return -EINVAL; | |
2662 | ||
1be374a0 | 2663 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
228e548e AB |
2664 | if (!sock) |
2665 | goto out; | |
2666 | ||
28a94d8f | 2667 | err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); |
228e548e | 2668 | |
6cb153ca | 2669 | fput_light(sock->file, fput_needed); |
89bddce5 | 2670 | out: |
1da177e4 LT |
2671 | return err; |
2672 | } | |
2673 | ||
666547ff | 2674 | SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) |
a7526eb5 | 2675 | { |
e1834a32 | 2676 | return __sys_sendmsg(fd, msg, flags, true); |
a7526eb5 AL |
2677 | } |
2678 | ||
228e548e AB |
2679 | /* |
2680 | * Linux sendmmsg interface | |
2681 | */ | |
2682 | ||
2683 | int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, | |
e1834a32 | 2684 | unsigned int flags, bool forbid_cmsg_compat) |
228e548e AB |
2685 | { |
2686 | int fput_needed, err, datagrams; | |
2687 | struct socket *sock; | |
2688 | struct mmsghdr __user *entry; | |
2689 | struct compat_mmsghdr __user *compat_entry; | |
2690 | struct msghdr msg_sys; | |
c71d8ebe | 2691 | struct used_address used_address; |
f092276d | 2692 | unsigned int oflags = flags; |
228e548e | 2693 | |
e1834a32 DB |
2694 | if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) |
2695 | return -EINVAL; | |
2696 | ||
98382f41 AB |
2697 | if (vlen > UIO_MAXIOV) |
2698 | vlen = UIO_MAXIOV; | |
228e548e AB |
2699 | |
2700 | datagrams = 0; | |
2701 | ||
2702 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
2703 | if (!sock) | |
2704 | return err; | |
2705 | ||
c71d8ebe | 2706 | used_address.name_len = UINT_MAX; |
228e548e AB |
2707 | entry = mmsg; |
2708 | compat_entry = (struct compat_mmsghdr __user *)mmsg; | |
728ffb86 | 2709 | err = 0; |
f092276d | 2710 | flags |= MSG_BATCH; |
228e548e AB |
2711 | |
2712 | while (datagrams < vlen) { | |
f092276d TH |
2713 | if (datagrams == vlen - 1) |
2714 | flags = oflags; | |
2715 | ||
228e548e | 2716 | if (MSG_CMSG_COMPAT & flags) { |
666547ff | 2717 | err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, |
28a94d8f | 2718 | &msg_sys, flags, &used_address, MSG_EOR); |
228e548e AB |
2719 | if (err < 0) |
2720 | break; | |
2721 | err = __put_user(err, &compat_entry->msg_len); | |
2722 | ++compat_entry; | |
2723 | } else { | |
a7526eb5 | 2724 | err = ___sys_sendmsg(sock, |
666547ff | 2725 | (struct user_msghdr __user *)entry, |
28a94d8f | 2726 | &msg_sys, flags, &used_address, MSG_EOR); |
228e548e AB |
2727 | if (err < 0) |
2728 | break; | |
2729 | err = put_user(err, &entry->msg_len); | |
2730 | ++entry; | |
2731 | } | |
2732 | ||
2733 | if (err) | |
2734 | break; | |
2735 | ++datagrams; | |
3023898b SHY |
2736 | if (msg_data_left(&msg_sys)) |
2737 | break; | |
a78cb84c | 2738 | cond_resched(); |
228e548e AB |
2739 | } |
2740 | ||
228e548e AB |
2741 | fput_light(sock->file, fput_needed); |
2742 | ||
728ffb86 AB |
2743 | /* We only return an error if no datagrams were able to be sent */ |
2744 | if (datagrams != 0) | |
228e548e AB |
2745 | return datagrams; |
2746 | ||
228e548e AB |
2747 | return err; |
2748 | } | |
2749 | ||
2750 | SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, | |
2751 | unsigned int, vlen, unsigned int, flags) | |
2752 | { | |
e1834a32 | 2753 | return __sys_sendmmsg(fd, mmsg, vlen, flags, true); |
228e548e AB |
2754 | } |
2755 | ||
03b1230c JA |
2756 | int recvmsg_copy_msghdr(struct msghdr *msg, |
2757 | struct user_msghdr __user *umsg, unsigned flags, | |
2758 | struct sockaddr __user **uaddr, | |
2759 | struct iovec **iov) | |
1da177e4 | 2760 | { |
08adb7da | 2761 | ssize_t err; |
1da177e4 | 2762 | |
4257c8ca JA |
2763 | if (MSG_CMSG_COMPAT & flags) { |
2764 | struct compat_msghdr __user *msg_compat; | |
1da177e4 | 2765 | |
4257c8ca JA |
2766 | msg_compat = (struct compat_msghdr __user *) umsg; |
2767 | err = get_compat_msghdr(msg, msg_compat, uaddr, iov); | |
2768 | } else { | |
2769 | err = copy_msghdr_from_user(msg, umsg, uaddr, iov); | |
2770 | } | |
1da177e4 | 2771 | if (err < 0) |
da184284 | 2772 | return err; |
1da177e4 | 2773 | |
4257c8ca JA |
2774 | return 0; |
2775 | } | |
2776 | ||
2777 | static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys, | |
2778 | struct user_msghdr __user *msg, | |
2779 | struct sockaddr __user *uaddr, | |
2780 | unsigned int flags, int nosec) | |
2781 | { | |
2782 | struct compat_msghdr __user *msg_compat = | |
2783 | (struct compat_msghdr __user *) msg; | |
2784 | int __user *uaddr_len = COMPAT_NAMELEN(msg); | |
2785 | struct sockaddr_storage addr; | |
2786 | unsigned long cmsg_ptr; | |
2787 | int len; | |
2788 | ssize_t err; | |
2789 | ||
2790 | msg_sys->msg_name = &addr; | |
a2e27255 ACM |
2791 | cmsg_ptr = (unsigned long)msg_sys->msg_control; |
2792 | msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); | |
89bddce5 | 2793 | |
f3d33426 HFS |
2794 | /* We assume all kernel code knows the size of sockaddr_storage */ |
2795 | msg_sys->msg_namelen = 0; | |
2796 | ||
1da177e4 LT |
2797 | if (sock->file->f_flags & O_NONBLOCK) |
2798 | flags |= MSG_DONTWAIT; | |
1af66221 ED |
2799 | |
2800 | if (unlikely(nosec)) | |
2801 | err = sock_recvmsg_nosec(sock, msg_sys, flags); | |
2802 | else | |
2803 | err = sock_recvmsg(sock, msg_sys, flags); | |
2804 | ||
1da177e4 | 2805 | if (err < 0) |
4257c8ca | 2806 | goto out; |
1da177e4 LT |
2807 | len = err; |
2808 | ||
2809 | if (uaddr != NULL) { | |
43db362d | 2810 | err = move_addr_to_user(&addr, |
a2e27255 | 2811 | msg_sys->msg_namelen, uaddr, |
89bddce5 | 2812 | uaddr_len); |
1da177e4 | 2813 | if (err < 0) |
4257c8ca | 2814 | goto out; |
1da177e4 | 2815 | } |
a2e27255 | 2816 | err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), |
37f7f421 | 2817 | COMPAT_FLAGS(msg)); |
1da177e4 | 2818 | if (err) |
4257c8ca | 2819 | goto out; |
1da177e4 | 2820 | if (MSG_CMSG_COMPAT & flags) |
a2e27255 | 2821 | err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, |
1da177e4 LT |
2822 | &msg_compat->msg_controllen); |
2823 | else | |
a2e27255 | 2824 | err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, |
1da177e4 LT |
2825 | &msg->msg_controllen); |
2826 | if (err) | |
4257c8ca | 2827 | goto out; |
1da177e4 | 2828 | err = len; |
4257c8ca JA |
2829 | out: |
2830 | return err; | |
2831 | } | |
2832 | ||
2833 | static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, | |
2834 | struct msghdr *msg_sys, unsigned int flags, int nosec) | |
2835 | { | |
2836 | struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; | |
2837 | /* user mode address pointers */ | |
2838 | struct sockaddr __user *uaddr; | |
2839 | ssize_t err; | |
2840 | ||
2841 | err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov); | |
2842 | if (err < 0) | |
2843 | return err; | |
1da177e4 | 2844 | |
4257c8ca | 2845 | err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec); |
da184284 | 2846 | kfree(iov); |
a2e27255 ACM |
2847 | return err; |
2848 | } | |
2849 | ||
2850 | /* | |
2851 | * BSD recvmsg interface | |
2852 | */ | |
2853 | ||
03b1230c JA |
2854 | long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg, |
2855 | struct user_msghdr __user *umsg, | |
2856 | struct sockaddr __user *uaddr, unsigned int flags) | |
aa1fa28f | 2857 | { |
03b1230c | 2858 | return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0); |
aa1fa28f JA |
2859 | } |
2860 | ||
e1834a32 DB |
2861 | long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, |
2862 | bool forbid_cmsg_compat) | |
a2e27255 ACM |
2863 | { |
2864 | int fput_needed, err; | |
2865 | struct msghdr msg_sys; | |
1be374a0 AL |
2866 | struct socket *sock; |
2867 | ||
e1834a32 DB |
2868 | if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) |
2869 | return -EINVAL; | |
2870 | ||
1be374a0 | 2871 | sock = sockfd_lookup_light(fd, &err, &fput_needed); |
a2e27255 ACM |
2872 | if (!sock) |
2873 | goto out; | |
2874 | ||
a7526eb5 | 2875 | err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); |
a2e27255 | 2876 | |
6cb153ca | 2877 | fput_light(sock->file, fput_needed); |
1da177e4 LT |
2878 | out: |
2879 | return err; | |
2880 | } | |
2881 | ||
666547ff | 2882 | SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, |
a7526eb5 AL |
2883 | unsigned int, flags) |
2884 | { | |
e1834a32 | 2885 | return __sys_recvmsg(fd, msg, flags, true); |
a7526eb5 AL |
2886 | } |
2887 | ||
a2e27255 ACM |
2888 | /* |
2889 | * Linux recvmmsg interface | |
2890 | */ | |
2891 | ||
e11d4284 AB |
2892 | static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg, |
2893 | unsigned int vlen, unsigned int flags, | |
2894 | struct timespec64 *timeout) | |
a2e27255 ACM |
2895 | { |
2896 | int fput_needed, err, datagrams; | |
2897 | struct socket *sock; | |
2898 | struct mmsghdr __user *entry; | |
d7256d0e | 2899 | struct compat_mmsghdr __user *compat_entry; |
a2e27255 | 2900 | struct msghdr msg_sys; |
766b9f92 DD |
2901 | struct timespec64 end_time; |
2902 | struct timespec64 timeout64; | |
a2e27255 ACM |
2903 | |
2904 | if (timeout && | |
2905 | poll_select_set_timeout(&end_time, timeout->tv_sec, | |
2906 | timeout->tv_nsec)) | |
2907 | return -EINVAL; | |
2908 | ||
2909 | datagrams = 0; | |
2910 | ||
2911 | sock = sockfd_lookup_light(fd, &err, &fput_needed); | |
2912 | if (!sock) | |
2913 | return err; | |
2914 | ||
7797dc41 SHY |
2915 | if (likely(!(flags & MSG_ERRQUEUE))) { |
2916 | err = sock_error(sock->sk); | |
2917 | if (err) { | |
2918 | datagrams = err; | |
2919 | goto out_put; | |
2920 | } | |
e623a9e9 | 2921 | } |
a2e27255 ACM |
2922 | |
2923 | entry = mmsg; | |
d7256d0e | 2924 | compat_entry = (struct compat_mmsghdr __user *)mmsg; |
a2e27255 ACM |
2925 | |
2926 | while (datagrams < vlen) { | |
2927 | /* | |
2928 | * No need to ask LSM for more than the first datagram. | |
2929 | */ | |
d7256d0e | 2930 | if (MSG_CMSG_COMPAT & flags) { |
666547ff | 2931 | err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, |
a7526eb5 AL |
2932 | &msg_sys, flags & ~MSG_WAITFORONE, |
2933 | datagrams); | |
d7256d0e JMG |
2934 | if (err < 0) |
2935 | break; | |
2936 | err = __put_user(err, &compat_entry->msg_len); | |
2937 | ++compat_entry; | |
2938 | } else { | |
a7526eb5 | 2939 | err = ___sys_recvmsg(sock, |
666547ff | 2940 | (struct user_msghdr __user *)entry, |
a7526eb5 AL |
2941 | &msg_sys, flags & ~MSG_WAITFORONE, |
2942 | datagrams); | |
d7256d0e JMG |
2943 | if (err < 0) |
2944 | break; | |
2945 | err = put_user(err, &entry->msg_len); | |
2946 | ++entry; | |
2947 | } | |
2948 | ||
a2e27255 ACM |
2949 | if (err) |
2950 | break; | |
a2e27255 ACM |
2951 | ++datagrams; |
2952 | ||
71c5c159 BB |
2953 | /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ |
2954 | if (flags & MSG_WAITFORONE) | |
2955 | flags |= MSG_DONTWAIT; | |
2956 | ||
a2e27255 | 2957 | if (timeout) { |
766b9f92 | 2958 | ktime_get_ts64(&timeout64); |
c2e6c856 | 2959 | *timeout = timespec64_sub(end_time, timeout64); |
a2e27255 ACM |
2960 | if (timeout->tv_sec < 0) { |
2961 | timeout->tv_sec = timeout->tv_nsec = 0; | |
2962 | break; | |
2963 | } | |
2964 | ||
2965 | /* Timeout, return less than vlen datagrams */ | |
2966 | if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) | |
2967 | break; | |
2968 | } | |
2969 | ||
2970 | /* Out of band data, return right away */ | |
2971 | if (msg_sys.msg_flags & MSG_OOB) | |
2972 | break; | |
a78cb84c | 2973 | cond_resched(); |
a2e27255 ACM |
2974 | } |
2975 | ||
a2e27255 | 2976 | if (err == 0) |
34b88a68 ACM |
2977 | goto out_put; |
2978 | ||
2979 | if (datagrams == 0) { | |
2980 | datagrams = err; | |
2981 | goto out_put; | |
2982 | } | |
a2e27255 | 2983 | |
34b88a68 ACM |
2984 | /* |
2985 | * We may return less entries than requested (vlen) if the | |
2986 | * sock is non block and there aren't enough datagrams... | |
2987 | */ | |
2988 | if (err != -EAGAIN) { | |
a2e27255 | 2989 | /* |
34b88a68 ACM |
2990 | * ... or if recvmsg returns an error after we |
2991 | * received some datagrams, where we record the | |
2992 | * error to return on the next call or if the | |
2993 | * app asks about it using getsockopt(SO_ERROR). | |
a2e27255 | 2994 | */ |
e05a5f51 | 2995 | WRITE_ONCE(sock->sk->sk_err, -err); |
a2e27255 | 2996 | } |
34b88a68 ACM |
2997 | out_put: |
2998 | fput_light(sock->file, fput_needed); | |
a2e27255 | 2999 | |
34b88a68 | 3000 | return datagrams; |
a2e27255 ACM |
3001 | } |
3002 | ||
e11d4284 AB |
3003 | int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, |
3004 | unsigned int vlen, unsigned int flags, | |
3005 | struct __kernel_timespec __user *timeout, | |
3006 | struct old_timespec32 __user *timeout32) | |
a2e27255 ACM |
3007 | { |
3008 | int datagrams; | |
c2e6c856 | 3009 | struct timespec64 timeout_sys; |
a2e27255 | 3010 | |
e11d4284 AB |
3011 | if (timeout && get_timespec64(&timeout_sys, timeout)) |
3012 | return -EFAULT; | |
a2e27255 | 3013 | |
e11d4284 | 3014 | if (timeout32 && get_old_timespec32(&timeout_sys, timeout32)) |
a2e27255 ACM |
3015 | return -EFAULT; |
3016 | ||
e11d4284 AB |
3017 | if (!timeout && !timeout32) |
3018 | return do_recvmmsg(fd, mmsg, vlen, flags, NULL); | |
3019 | ||
3020 | datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); | |
a2e27255 | 3021 | |
e11d4284 AB |
3022 | if (datagrams <= 0) |
3023 | return datagrams; | |
3024 | ||
3025 | if (timeout && put_timespec64(&timeout_sys, timeout)) | |
3026 | datagrams = -EFAULT; | |
3027 | ||
3028 | if (timeout32 && put_old_timespec32(&timeout_sys, timeout32)) | |
a2e27255 ACM |
3029 | datagrams = -EFAULT; |
3030 | ||
3031 | return datagrams; | |
3032 | } | |
3033 | ||
1255e269 DB |
3034 | SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, |
3035 | unsigned int, vlen, unsigned int, flags, | |
c2e6c856 | 3036 | struct __kernel_timespec __user *, timeout) |
1255e269 | 3037 | { |
e11d4284 AB |
3038 | if (flags & MSG_CMSG_COMPAT) |
3039 | return -EINVAL; | |
3040 | ||
3041 | return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL); | |
3042 | } | |
3043 | ||
3044 | #ifdef CONFIG_COMPAT_32BIT_TIME | |
3045 | SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg, | |
3046 | unsigned int, vlen, unsigned int, flags, | |
3047 | struct old_timespec32 __user *, timeout) | |
3048 | { | |
3049 | if (flags & MSG_CMSG_COMPAT) | |
3050 | return -EINVAL; | |
3051 | ||
3052 | return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout); | |
1255e269 | 3053 | } |
e11d4284 | 3054 | #endif |
1255e269 | 3055 | |
a2e27255 | 3056 | #ifdef __ARCH_WANT_SYS_SOCKETCALL |
1da177e4 LT |
3057 | /* Argument list sizes for sys_socketcall */ |
3058 | #define AL(x) ((x) * sizeof(unsigned long)) | |
228e548e | 3059 | static const unsigned char nargs[21] = { |
c6d409cf ED |
3060 | AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), |
3061 | AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), | |
3062 | AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), | |
228e548e | 3063 | AL(4), AL(5), AL(4) |
89bddce5 SH |
3064 | }; |
3065 | ||
1da177e4 LT |
3066 | #undef AL |
3067 | ||
3068 | /* | |
89bddce5 | 3069 | * System call vectors. |
1da177e4 LT |
3070 | * |
3071 | * Argument checking cleaned up. Saved 20% in size. | |
3072 | * This function doesn't need to set the kernel lock because | |
89bddce5 | 3073 | * it is set by the callees. |
1da177e4 LT |
3074 | */ |
3075 | ||
3e0fa65f | 3076 | SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) |
1da177e4 | 3077 | { |
2950fa9d | 3078 | unsigned long a[AUDITSC_ARGS]; |
89bddce5 | 3079 | unsigned long a0, a1; |
1da177e4 | 3080 | int err; |
47379052 | 3081 | unsigned int len; |
1da177e4 | 3082 | |
228e548e | 3083 | if (call < 1 || call > SYS_SENDMMSG) |
1da177e4 | 3084 | return -EINVAL; |
c8e8cd57 | 3085 | call = array_index_nospec(call, SYS_SENDMMSG + 1); |
1da177e4 | 3086 | |
47379052 AV |
3087 | len = nargs[call]; |
3088 | if (len > sizeof(a)) | |
3089 | return -EINVAL; | |
3090 | ||
1da177e4 | 3091 | /* copy_from_user should be SMP safe. */ |
47379052 | 3092 | if (copy_from_user(a, args, len)) |
1da177e4 | 3093 | return -EFAULT; |
3ec3b2fb | 3094 | |
2950fa9d CG |
3095 | err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); |
3096 | if (err) | |
3097 | return err; | |
3ec3b2fb | 3098 | |
89bddce5 SH |
3099 | a0 = a[0]; |
3100 | a1 = a[1]; | |
3101 | ||
3102 | switch (call) { | |
3103 | case SYS_SOCKET: | |
9d6a15c3 | 3104 | err = __sys_socket(a0, a1, a[2]); |
89bddce5 SH |
3105 | break; |
3106 | case SYS_BIND: | |
a87d35d8 | 3107 | err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); |
89bddce5 SH |
3108 | break; |
3109 | case SYS_CONNECT: | |
1387c2c2 | 3110 | err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); |
89bddce5 SH |
3111 | break; |
3112 | case SYS_LISTEN: | |
25e290ee | 3113 | err = __sys_listen(a0, a1); |
89bddce5 SH |
3114 | break; |
3115 | case SYS_ACCEPT: | |
4541e805 DB |
3116 | err = __sys_accept4(a0, (struct sockaddr __user *)a1, |
3117 | (int __user *)a[2], 0); | |
89bddce5 SH |
3118 | break; |
3119 | case SYS_GETSOCKNAME: | |
3120 | err = | |
8882a107 DB |
3121 | __sys_getsockname(a0, (struct sockaddr __user *)a1, |
3122 | (int __user *)a[2]); | |
89bddce5 SH |
3123 | break; |
3124 | case SYS_GETPEERNAME: | |
3125 | err = | |
b21c8f83 DB |
3126 | __sys_getpeername(a0, (struct sockaddr __user *)a1, |
3127 | (int __user *)a[2]); | |
89bddce5 SH |
3128 | break; |
3129 | case SYS_SOCKETPAIR: | |
6debc8d8 | 3130 | err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); |
89bddce5 SH |
3131 | break; |
3132 | case SYS_SEND: | |
f3bf896b DB |
3133 | err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], |
3134 | NULL, 0); | |
89bddce5 SH |
3135 | break; |
3136 | case SYS_SENDTO: | |
211b634b DB |
3137 | err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], |
3138 | (struct sockaddr __user *)a[4], a[5]); | |
89bddce5 SH |
3139 | break; |
3140 | case SYS_RECV: | |
d27e9afc DB |
3141 | err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], |
3142 | NULL, NULL); | |
89bddce5 SH |
3143 | break; |
3144 | case SYS_RECVFROM: | |
7a09e1eb DB |
3145 | err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], |
3146 | (struct sockaddr __user *)a[4], | |
3147 | (int __user *)a[5]); | |
89bddce5 SH |
3148 | break; |
3149 | case SYS_SHUTDOWN: | |
005a1aea | 3150 | err = __sys_shutdown(a0, a1); |
89bddce5 SH |
3151 | break; |
3152 | case SYS_SETSOCKOPT: | |
cc36dca0 DB |
3153 | err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], |
3154 | a[4]); | |
89bddce5 SH |
3155 | break; |
3156 | case SYS_GETSOCKOPT: | |
3157 | err = | |
13a2d70e DB |
3158 | __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], |
3159 | (int __user *)a[4]); | |
89bddce5 SH |
3160 | break; |
3161 | case SYS_SENDMSG: | |
e1834a32 DB |
3162 | err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, |
3163 | a[2], true); | |
89bddce5 | 3164 | break; |
228e548e | 3165 | case SYS_SENDMMSG: |
e1834a32 DB |
3166 | err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], |
3167 | a[3], true); | |
228e548e | 3168 | break; |
89bddce5 | 3169 | case SYS_RECVMSG: |
e1834a32 DB |
3170 | err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, |
3171 | a[2], true); | |
89bddce5 | 3172 | break; |
a2e27255 | 3173 | case SYS_RECVMMSG: |
3ca47e95 | 3174 | if (IS_ENABLED(CONFIG_64BIT)) |
e11d4284 AB |
3175 | err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, |
3176 | a[2], a[3], | |
3177 | (struct __kernel_timespec __user *)a[4], | |
3178 | NULL); | |
3179 | else | |
3180 | err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, | |
3181 | a[2], a[3], NULL, | |
3182 | (struct old_timespec32 __user *)a[4]); | |
a2e27255 | 3183 | break; |
de11defe | 3184 | case SYS_ACCEPT4: |
4541e805 DB |
3185 | err = __sys_accept4(a0, (struct sockaddr __user *)a1, |
3186 | (int __user *)a[2], a[3]); | |
aaca0bdc | 3187 | break; |
89bddce5 SH |
3188 | default: |
3189 | err = -EINVAL; | |
3190 | break; | |
1da177e4 LT |
3191 | } |
3192 | return err; | |
3193 | } | |
3194 | ||
89bddce5 | 3195 | #endif /* __ARCH_WANT_SYS_SOCKETCALL */ |
1da177e4 | 3196 | |
55737fda SH |
3197 | /** |
3198 | * sock_register - add a socket protocol handler | |
3199 | * @ops: description of protocol | |
3200 | * | |
1da177e4 LT |
3201 | * This function is called by a protocol handler that wants to |
3202 | * advertise its address family, and have it linked into the | |
e793c0f7 | 3203 | * socket interface. The value ops->family corresponds to the |
55737fda | 3204 | * socket system call protocol family. |
1da177e4 | 3205 | */ |
f0fd27d4 | 3206 | int sock_register(const struct net_proto_family *ops) |
1da177e4 LT |
3207 | { |
3208 | int err; | |
3209 | ||
3210 | if (ops->family >= NPROTO) { | |
3410f22e | 3211 | pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); |
1da177e4 LT |
3212 | return -ENOBUFS; |
3213 | } | |
55737fda SH |
3214 | |
3215 | spin_lock(&net_family_lock); | |
190683a9 ED |
3216 | if (rcu_dereference_protected(net_families[ops->family], |
3217 | lockdep_is_held(&net_family_lock))) | |
55737fda SH |
3218 | err = -EEXIST; |
3219 | else { | |
cf778b00 | 3220 | rcu_assign_pointer(net_families[ops->family], ops); |
1da177e4 LT |
3221 | err = 0; |
3222 | } | |
55737fda SH |
3223 | spin_unlock(&net_family_lock); |
3224 | ||
fe0bdbde | 3225 | pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]); |
1da177e4 LT |
3226 | return err; |
3227 | } | |
c6d409cf | 3228 | EXPORT_SYMBOL(sock_register); |
1da177e4 | 3229 | |
55737fda SH |
3230 | /** |
3231 | * sock_unregister - remove a protocol handler | |
3232 | * @family: protocol family to remove | |
3233 | * | |
1da177e4 LT |
3234 | * This function is called by a protocol handler that wants to |
3235 | * remove its address family, and have it unlinked from the | |
55737fda SH |
3236 | * new socket creation. |
3237 | * | |
3238 | * If protocol handler is a module, then it can use module reference | |
3239 | * counts to protect against new references. If protocol handler is not | |
3240 | * a module then it needs to provide its own protection in | |
3241 | * the ops->create routine. | |
1da177e4 | 3242 | */ |
f0fd27d4 | 3243 | void sock_unregister(int family) |
1da177e4 | 3244 | { |
f0fd27d4 | 3245 | BUG_ON(family < 0 || family >= NPROTO); |
1da177e4 | 3246 | |
55737fda | 3247 | spin_lock(&net_family_lock); |
a9b3cd7f | 3248 | RCU_INIT_POINTER(net_families[family], NULL); |
55737fda SH |
3249 | spin_unlock(&net_family_lock); |
3250 | ||
3251 | synchronize_rcu(); | |
3252 | ||
fe0bdbde | 3253 | pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]); |
1da177e4 | 3254 | } |
c6d409cf | 3255 | EXPORT_SYMBOL(sock_unregister); |
1da177e4 | 3256 | |
bf2ae2e4 XL |
3257 | bool sock_is_registered(int family) |
3258 | { | |
66b51b0a | 3259 | return family < NPROTO && rcu_access_pointer(net_families[family]); |
bf2ae2e4 XL |
3260 | } |
3261 | ||
77d76ea3 | 3262 | static int __init sock_init(void) |
1da177e4 | 3263 | { |
b3e19d92 | 3264 | int err; |
2ca794e5 EB |
3265 | /* |
3266 | * Initialize the network sysctl infrastructure. | |
3267 | */ | |
3268 | err = net_sysctl_init(); | |
3269 | if (err) | |
3270 | goto out; | |
b3e19d92 | 3271 | |
1da177e4 | 3272 | /* |
89bddce5 | 3273 | * Initialize skbuff SLAB cache |
1da177e4 LT |
3274 | */ |
3275 | skb_init(); | |
1da177e4 LT |
3276 | |
3277 | /* | |
89bddce5 | 3278 | * Initialize the protocols module. |
1da177e4 LT |
3279 | */ |
3280 | ||
3281 | init_inodecache(); | |
b3e19d92 NP |
3282 | |
3283 | err = register_filesystem(&sock_fs_type); | |
3284 | if (err) | |
47260ba9 | 3285 | goto out; |
1da177e4 | 3286 | sock_mnt = kern_mount(&sock_fs_type); |
b3e19d92 NP |
3287 | if (IS_ERR(sock_mnt)) { |
3288 | err = PTR_ERR(sock_mnt); | |
3289 | goto out_mount; | |
3290 | } | |
77d76ea3 AK |
3291 | |
3292 | /* The real protocol initialization is performed in later initcalls. | |
1da177e4 LT |
3293 | */ |
3294 | ||
3295 | #ifdef CONFIG_NETFILTER | |
6d11cfdb PNA |
3296 | err = netfilter_init(); |
3297 | if (err) | |
3298 | goto out; | |
1da177e4 | 3299 | #endif |
cbeb321a | 3300 | |
408eccce | 3301 | ptp_classifier_init(); |
c1f19b51 | 3302 | |
b3e19d92 NP |
3303 | out: |
3304 | return err; | |
3305 | ||
3306 | out_mount: | |
3307 | unregister_filesystem(&sock_fs_type); | |
b3e19d92 | 3308 | goto out; |
1da177e4 LT |
3309 | } |
3310 | ||
77d76ea3 AK |
3311 | core_initcall(sock_init); /* early initcall */ |
3312 | ||
1da177e4 LT |
3313 | #ifdef CONFIG_PROC_FS |
3314 | void socket_seq_show(struct seq_file *seq) | |
3315 | { | |
648845ab TZ |
3316 | seq_printf(seq, "sockets: used %d\n", |
3317 | sock_inuse_get(seq->private)); | |
1da177e4 | 3318 | } |
89bddce5 | 3319 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 3320 | |
29c49648 AB |
3321 | /* Handle the fact that while struct ifreq has the same *layout* on |
3322 | * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data, | |
3323 | * which are handled elsewhere, it still has different *size* due to | |
3324 | * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit, | |
3325 | * resulting in struct ifreq being 32 and 40 bytes respectively). | |
3326 | * As a result, if the struct happens to be at the end of a page and | |
3327 | * the next page isn't readable/writable, we get a fault. To prevent | |
3328 | * that, copy back and forth to the full size. | |
3329 | */ | |
3330 | int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg) | |
7a229387 | 3331 | { |
29c49648 AB |
3332 | if (in_compat_syscall()) { |
3333 | struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr; | |
7a229387 | 3334 | |
29c49648 AB |
3335 | memset(ifr, 0, sizeof(*ifr)); |
3336 | if (copy_from_user(ifr32, arg, sizeof(*ifr32))) | |
3337 | return -EFAULT; | |
7a229387 | 3338 | |
29c49648 AB |
3339 | if (ifrdata) |
3340 | *ifrdata = compat_ptr(ifr32->ifr_data); | |
7a229387 | 3341 | |
29c49648 AB |
3342 | return 0; |
3343 | } | |
7a229387 | 3344 | |
29c49648 | 3345 | if (copy_from_user(ifr, arg, sizeof(*ifr))) |
7a229387 AB |
3346 | return -EFAULT; |
3347 | ||
29c49648 AB |
3348 | if (ifrdata) |
3349 | *ifrdata = ifr->ifr_data; | |
3350 | ||
7a229387 AB |
3351 | return 0; |
3352 | } | |
29c49648 | 3353 | EXPORT_SYMBOL(get_user_ifreq); |
7a229387 | 3354 | |
29c49648 | 3355 | int put_user_ifreq(struct ifreq *ifr, void __user *arg) |
7a229387 | 3356 | { |
29c49648 | 3357 | size_t size = sizeof(*ifr); |
7a229387 | 3358 | |
29c49648 AB |
3359 | if (in_compat_syscall()) |
3360 | size = sizeof(struct compat_ifreq); | |
7a229387 | 3361 | |
29c49648 | 3362 | if (copy_to_user(arg, ifr, size)) |
7a229387 AB |
3363 | return -EFAULT; |
3364 | ||
3a7da39d | 3365 | return 0; |
7a229387 | 3366 | } |
29c49648 | 3367 | EXPORT_SYMBOL(put_user_ifreq); |
7a229387 | 3368 | |
89bbfc95 | 3369 | #ifdef CONFIG_COMPAT |
7a50a240 AB |
3370 | static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) |
3371 | { | |
7a50a240 | 3372 | compat_uptr_t uptr32; |
44c02a2c AV |
3373 | struct ifreq ifr; |
3374 | void __user *saved; | |
3375 | int err; | |
7a50a240 | 3376 | |
29c49648 | 3377 | if (get_user_ifreq(&ifr, NULL, uifr32)) |
7a50a240 AB |
3378 | return -EFAULT; |
3379 | ||
3380 | if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) | |
3381 | return -EFAULT; | |
3382 | ||
44c02a2c AV |
3383 | saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; |
3384 | ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); | |
7a229387 | 3385 | |
a554bf96 | 3386 | err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL); |
44c02a2c AV |
3387 | if (!err) { |
3388 | ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; | |
29c49648 | 3389 | if (put_user_ifreq(&ifr, uifr32)) |
44c02a2c | 3390 | err = -EFAULT; |
ccbd6a5a | 3391 | } |
44c02a2c | 3392 | return err; |
7a229387 AB |
3393 | } |
3394 | ||
590d4693 BH |
3395 | /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ |
3396 | static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, | |
6b96018b | 3397 | struct compat_ifreq __user *u_ifreq32) |
7a229387 | 3398 | { |
44c02a2c | 3399 | struct ifreq ifreq; |
a554bf96 | 3400 | void __user *data; |
7a229387 | 3401 | |
d0efb162 PC |
3402 | if (!is_socket_ioctl_cmd(cmd)) |
3403 | return -ENOTTY; | |
a554bf96 | 3404 | if (get_user_ifreq(&ifreq, &data, u_ifreq32)) |
7a229387 | 3405 | return -EFAULT; |
a554bf96 | 3406 | ifreq.ifr_data = data; |
7a229387 | 3407 | |
a554bf96 | 3408 | return dev_ioctl(net, cmd, &ifreq, data, NULL); |
a2116ed2 AB |
3409 | } |
3410 | ||
6b96018b AB |
3411 | static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, |
3412 | unsigned int cmd, unsigned long arg) | |
3413 | { | |
3414 | void __user *argp = compat_ptr(arg); | |
3415 | struct sock *sk = sock->sk; | |
3416 | struct net *net = sock_net(sk); | |
1ded5e5a | 3417 | const struct proto_ops *ops; |
7a229387 | 3418 | |
6b96018b | 3419 | if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) |
88fc023f | 3420 | return sock_ioctl(file, cmd, (unsigned long)argp); |
6b96018b AB |
3421 | |
3422 | switch (cmd) { | |
7a50a240 AB |
3423 | case SIOCWANDEV: |
3424 | return compat_siocwandev(net, argp); | |
0768e170 AB |
3425 | case SIOCGSTAMP_OLD: |
3426 | case SIOCGSTAMPNS_OLD: | |
1ded5e5a ED |
3427 | ops = READ_ONCE(sock->ops); |
3428 | if (!ops->gettstamp) | |
c7cbdbf2 | 3429 | return -ENOIOCTLCMD; |
1ded5e5a ED |
3430 | return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD, |
3431 | !COMPAT_USE_64BIT_TIME); | |
c7cbdbf2 | 3432 | |
dd98d289 | 3433 | case SIOCETHTOOL: |
590d4693 BH |
3434 | case SIOCBONDSLAVEINFOQUERY: |
3435 | case SIOCBONDINFOQUERY: | |
a2116ed2 | 3436 | case SIOCSHWTSTAMP: |
fd468c74 | 3437 | case SIOCGHWTSTAMP: |
590d4693 | 3438 | return compat_ifr_data_ioctl(net, cmd, argp); |
6b96018b AB |
3439 | |
3440 | case FIOSETOWN: | |
3441 | case SIOCSPGRP: | |
3442 | case FIOGETOWN: | |
3443 | case SIOCGPGRP: | |
3444 | case SIOCBRADDBR: | |
3445 | case SIOCBRDELBR: | |
3446 | case SIOCGIFVLAN: | |
3447 | case SIOCSIFVLAN: | |
c62cce2c | 3448 | case SIOCGSKNS: |
0768e170 AB |
3449 | case SIOCGSTAMP_NEW: |
3450 | case SIOCGSTAMPNS_NEW: | |
876f0bf9 | 3451 | case SIOCGIFCONF: |
fd3a4590 RP |
3452 | case SIOCSIFBR: |
3453 | case SIOCGIFBR: | |
6b96018b AB |
3454 | return sock_ioctl(file, cmd, arg); |
3455 | ||
3456 | case SIOCGIFFLAGS: | |
3457 | case SIOCSIFFLAGS: | |
709566d7 AB |
3458 | case SIOCGIFMAP: |
3459 | case SIOCSIFMAP: | |
6b96018b AB |
3460 | case SIOCGIFMETRIC: |
3461 | case SIOCSIFMETRIC: | |
3462 | case SIOCGIFMTU: | |
3463 | case SIOCSIFMTU: | |
3464 | case SIOCGIFMEM: | |
3465 | case SIOCSIFMEM: | |
3466 | case SIOCGIFHWADDR: | |
3467 | case SIOCSIFHWADDR: | |
3468 | case SIOCADDMULTI: | |
3469 | case SIOCDELMULTI: | |
3470 | case SIOCGIFINDEX: | |
6b96018b AB |
3471 | case SIOCGIFADDR: |
3472 | case SIOCSIFADDR: | |
3473 | case SIOCSIFHWBROADCAST: | |
6b96018b | 3474 | case SIOCDIFADDR: |
6b96018b AB |
3475 | case SIOCGIFBRDADDR: |
3476 | case SIOCSIFBRDADDR: | |
3477 | case SIOCGIFDSTADDR: | |
3478 | case SIOCSIFDSTADDR: | |
3479 | case SIOCGIFNETMASK: | |
3480 | case SIOCSIFNETMASK: | |
3481 | case SIOCSIFPFLAGS: | |
3482 | case SIOCGIFPFLAGS: | |
3483 | case SIOCGIFTXQLEN: | |
3484 | case SIOCSIFTXQLEN: | |
3485 | case SIOCBRADDIF: | |
3486 | case SIOCBRDELIF: | |
c6c9fee3 | 3487 | case SIOCGIFNAME: |
9177efd3 AB |
3488 | case SIOCSIFNAME: |
3489 | case SIOCGMIIPHY: | |
3490 | case SIOCGMIIREG: | |
3491 | case SIOCSMIIREG: | |
f92d4fc9 AV |
3492 | case SIOCBONDENSLAVE: |
3493 | case SIOCBONDRELEASE: | |
3494 | case SIOCBONDSETHWADDR: | |
3495 | case SIOCBONDCHANGEACTIVE: | |
6b96018b AB |
3496 | case SIOCSARP: |
3497 | case SIOCGARP: | |
3498 | case SIOCDARP: | |
c7dc504e | 3499 | case SIOCOUTQ: |
9d7bf41f | 3500 | case SIOCOUTQNSD: |
6b96018b | 3501 | case SIOCATMARK: |
63ff03ab | 3502 | return sock_do_ioctl(net, sock, cmd, arg); |
9177efd3 AB |
3503 | } |
3504 | ||
6b96018b AB |
3505 | return -ENOIOCTLCMD; |
3506 | } | |
7a229387 | 3507 | |
95c96174 | 3508 | static long compat_sock_ioctl(struct file *file, unsigned int cmd, |
89bddce5 | 3509 | unsigned long arg) |
89bbfc95 SP |
3510 | { |
3511 | struct socket *sock = file->private_data; | |
1ded5e5a | 3512 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
89bbfc95 | 3513 | int ret = -ENOIOCTLCMD; |
87de87d5 DM |
3514 | struct sock *sk; |
3515 | struct net *net; | |
3516 | ||
3517 | sk = sock->sk; | |
3518 | net = sock_net(sk); | |
89bbfc95 | 3519 | |
1ded5e5a ED |
3520 | if (ops->compat_ioctl) |
3521 | ret = ops->compat_ioctl(sock, cmd, arg); | |
89bbfc95 | 3522 | |
87de87d5 DM |
3523 | if (ret == -ENOIOCTLCMD && |
3524 | (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) | |
3525 | ret = compat_wext_handle_ioctl(net, cmd, arg); | |
3526 | ||
6b96018b AB |
3527 | if (ret == -ENOIOCTLCMD) |
3528 | ret = compat_sock_ioctl_trans(file, sock, cmd, arg); | |
3529 | ||
89bbfc95 SP |
3530 | return ret; |
3531 | } | |
3532 | #endif | |
3533 | ||
8a3c245c PT |
3534 | /** |
3535 | * kernel_bind - bind an address to a socket (kernel space) | |
3536 | * @sock: socket | |
3537 | * @addr: address | |
3538 | * @addrlen: length of address | |
3539 | * | |
3540 | * Returns 0 or an error. | |
3541 | */ | |
3542 | ||
ac5a488e SS |
3543 | int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) |
3544 | { | |
c889a99a JR |
3545 | struct sockaddr_storage address; |
3546 | ||
3547 | memcpy(&address, addr, addrlen); | |
3548 | ||
3549 | return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address, | |
3550 | addrlen); | |
ac5a488e | 3551 | } |
c6d409cf | 3552 | EXPORT_SYMBOL(kernel_bind); |
ac5a488e | 3553 | |
8a3c245c PT |
3554 | /** |
3555 | * kernel_listen - move socket to listening state (kernel space) | |
3556 | * @sock: socket | |
3557 | * @backlog: pending connections queue size | |
3558 | * | |
3559 | * Returns 0 or an error. | |
3560 | */ | |
3561 | ||
ac5a488e SS |
3562 | int kernel_listen(struct socket *sock, int backlog) |
3563 | { | |
1ded5e5a | 3564 | return READ_ONCE(sock->ops)->listen(sock, backlog); |
ac5a488e | 3565 | } |
c6d409cf | 3566 | EXPORT_SYMBOL(kernel_listen); |
ac5a488e | 3567 | |
8a3c245c PT |
3568 | /** |
3569 | * kernel_accept - accept a connection (kernel space) | |
3570 | * @sock: listening socket | |
3571 | * @newsock: new connected socket | |
3572 | * @flags: flags | |
3573 | * | |
3574 | * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0. | |
3575 | * If it fails, @newsock is guaranteed to be %NULL. | |
3576 | * Returns 0 or an error. | |
3577 | */ | |
3578 | ||
ac5a488e SS |
3579 | int kernel_accept(struct socket *sock, struct socket **newsock, int flags) |
3580 | { | |
3581 | struct sock *sk = sock->sk; | |
1ded5e5a | 3582 | const struct proto_ops *ops = READ_ONCE(sock->ops); |
ac5a488e SS |
3583 | int err; |
3584 | ||
3585 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
3586 | newsock); | |
3587 | if (err < 0) | |
3588 | goto done; | |
3589 | ||
1ded5e5a | 3590 | err = ops->accept(sock, *newsock, flags, true); |
ac5a488e SS |
3591 | if (err < 0) { |
3592 | sock_release(*newsock); | |
fa8705b0 | 3593 | *newsock = NULL; |
ac5a488e SS |
3594 | goto done; |
3595 | } | |
3596 | ||
1ded5e5a ED |
3597 | (*newsock)->ops = ops; |
3598 | __module_get(ops->owner); | |
ac5a488e SS |
3599 | |
3600 | done: | |
3601 | return err; | |
3602 | } | |
c6d409cf | 3603 | EXPORT_SYMBOL(kernel_accept); |
ac5a488e | 3604 | |
8a3c245c PT |
3605 | /** |
3606 | * kernel_connect - connect a socket (kernel space) | |
3607 | * @sock: socket | |
3608 | * @addr: address | |
3609 | * @addrlen: address length | |
3610 | * @flags: flags (O_NONBLOCK, ...) | |
3611 | * | |
f1dcffcc | 3612 | * For datagram sockets, @addr is the address to which datagrams are sent |
8a3c245c PT |
3613 | * by default, and the only address from which datagrams are received. |
3614 | * For stream sockets, attempts to connect to @addr. | |
3615 | * Returns 0 or an error code. | |
3616 | */ | |
3617 | ||
ac5a488e | 3618 | int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, |
4768fbcb | 3619 | int flags) |
ac5a488e | 3620 | { |
0bdf3993 JR |
3621 | struct sockaddr_storage address; |
3622 | ||
3623 | memcpy(&address, addr, addrlen); | |
3624 | ||
3625 | return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address, | |
3626 | addrlen, flags); | |
ac5a488e | 3627 | } |
c6d409cf | 3628 | EXPORT_SYMBOL(kernel_connect); |
ac5a488e | 3629 | |
8a3c245c PT |
3630 | /** |
3631 | * kernel_getsockname - get the address which the socket is bound (kernel space) | |
3632 | * @sock: socket | |
3633 | * @addr: address holder | |
3634 | * | |
3635 | * Fills the @addr pointer with the address which the socket is bound. | |
0fc95dec | 3636 | * Returns the length of the address in bytes or an error code. |
8a3c245c PT |
3637 | */ |
3638 | ||
9b2c45d4 | 3639 | int kernel_getsockname(struct socket *sock, struct sockaddr *addr) |
ac5a488e | 3640 | { |
1ded5e5a | 3641 | return READ_ONCE(sock->ops)->getname(sock, addr, 0); |
ac5a488e | 3642 | } |
c6d409cf | 3643 | EXPORT_SYMBOL(kernel_getsockname); |
ac5a488e | 3644 | |
8a3c245c | 3645 | /** |
645f0897 | 3646 | * kernel_getpeername - get the address which the socket is connected (kernel space) |
8a3c245c PT |
3647 | * @sock: socket |
3648 | * @addr: address holder | |
3649 | * | |
3650 | * Fills the @addr pointer with the address which the socket is connected. | |
0fc95dec | 3651 | * Returns the length of the address in bytes or an error code. |
8a3c245c PT |
3652 | */ |
3653 | ||
9b2c45d4 | 3654 | int kernel_getpeername(struct socket *sock, struct sockaddr *addr) |
ac5a488e | 3655 | { |
1ded5e5a | 3656 | return READ_ONCE(sock->ops)->getname(sock, addr, 1); |
ac5a488e | 3657 | } |
c6d409cf | 3658 | EXPORT_SYMBOL(kernel_getpeername); |
ac5a488e | 3659 | |
8a3c245c | 3660 | /** |
645f0897 | 3661 | * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space) |
8a3c245c PT |
3662 | * @sock: socket |
3663 | * @how: connection part | |
3664 | * | |
3665 | * Returns 0 or an error. | |
3666 | */ | |
3667 | ||
91cf45f0 TM |
3668 | int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) |
3669 | { | |
1ded5e5a | 3670 | return READ_ONCE(sock->ops)->shutdown(sock, how); |
91cf45f0 | 3671 | } |
91cf45f0 | 3672 | EXPORT_SYMBOL(kernel_sock_shutdown); |
113c3075 | 3673 | |
8a3c245c PT |
3674 | /** |
3675 | * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket | |
3676 | * @sk: socket | |
3677 | * | |
3678 | * This routine returns the IP overhead imposed by a socket i.e. | |
3679 | * the length of the underlying IP header, depending on whether | |
3680 | * this is an IPv4 or IPv6 socket and the length from IP options turned | |
3681 | * on at the socket. Assumes that the caller has a lock on the socket. | |
113c3075 | 3682 | */ |
8a3c245c | 3683 | |
113c3075 P |
3684 | u32 kernel_sock_ip_overhead(struct sock *sk) |
3685 | { | |
3686 | struct inet_sock *inet; | |
3687 | struct ip_options_rcu *opt; | |
3688 | u32 overhead = 0; | |
113c3075 P |
3689 | #if IS_ENABLED(CONFIG_IPV6) |
3690 | struct ipv6_pinfo *np; | |
3691 | struct ipv6_txoptions *optv6 = NULL; | |
3692 | #endif /* IS_ENABLED(CONFIG_IPV6) */ | |
3693 | ||
3694 | if (!sk) | |
3695 | return overhead; | |
3696 | ||
113c3075 P |
3697 | switch (sk->sk_family) { |
3698 | case AF_INET: | |
3699 | inet = inet_sk(sk); | |
3700 | overhead += sizeof(struct iphdr); | |
3701 | opt = rcu_dereference_protected(inet->inet_opt, | |
614d79c0 | 3702 | sock_owned_by_user(sk)); |
113c3075 P |
3703 | if (opt) |
3704 | overhead += opt->opt.optlen; | |
3705 | return overhead; | |
3706 | #if IS_ENABLED(CONFIG_IPV6) | |
3707 | case AF_INET6: | |
3708 | np = inet6_sk(sk); | |
3709 | overhead += sizeof(struct ipv6hdr); | |
3710 | if (np) | |
3711 | optv6 = rcu_dereference_protected(np->opt, | |
614d79c0 | 3712 | sock_owned_by_user(sk)); |
113c3075 P |
3713 | if (optv6) |
3714 | overhead += (optv6->opt_flen + optv6->opt_nflen); | |
3715 | return overhead; | |
3716 | #endif /* IS_ENABLED(CONFIG_IPV6) */ | |
3717 | default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ | |
3718 | return overhead; | |
3719 | } | |
3720 | } | |
3721 | EXPORT_SYMBOL(kernel_sock_ip_overhead); |