Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * NETLINK Kernel-user communication protocol. |
| 3 | * |
| 4 | * Authors: Alan Cox <alan@redhat.com> |
| 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
| 6 | * |
| 7 | * This program is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU General Public License |
| 9 | * as published by the Free Software Foundation; either version |
| 10 | * 2 of the License, or (at your option) any later version. |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 11 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 12 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
| 13 | * added netlink_proto_exit |
| 14 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> |
| 15 | * use nlk_sk, as sk->protinfo is on a diet 8) |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 16 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
| 17 | * - inc module use count of module that owns |
| 18 | * the kernel socket in case userspace opens |
| 19 | * socket of same protocol |
| 20 | * - remove all module support, since netlink is |
| 21 | * mandatory if CONFIG_NET=y these days |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | */ |
| 23 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #include <linux/module.h> |
| 25 | |
Randy Dunlap | 4fc268d | 2006-01-11 12:17:47 -0800 | [diff] [blame] | 26 | #include <linux/capability.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 27 | #include <linux/kernel.h> |
| 28 | #include <linux/init.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 29 | #include <linux/signal.h> |
| 30 | #include <linux/sched.h> |
| 31 | #include <linux/errno.h> |
| 32 | #include <linux/string.h> |
| 33 | #include <linux/stat.h> |
| 34 | #include <linux/socket.h> |
| 35 | #include <linux/un.h> |
| 36 | #include <linux/fcntl.h> |
| 37 | #include <linux/termios.h> |
| 38 | #include <linux/sockios.h> |
| 39 | #include <linux/net.h> |
| 40 | #include <linux/fs.h> |
| 41 | #include <linux/slab.h> |
| 42 | #include <asm/uaccess.h> |
| 43 | #include <linux/skbuff.h> |
| 44 | #include <linux/netdevice.h> |
| 45 | #include <linux/rtnetlink.h> |
| 46 | #include <linux/proc_fs.h> |
| 47 | #include <linux/seq_file.h> |
| 48 | #include <linux/smp_lock.h> |
| 49 | #include <linux/notifier.h> |
| 50 | #include <linux/security.h> |
| 51 | #include <linux/jhash.h> |
| 52 | #include <linux/jiffies.h> |
| 53 | #include <linux/random.h> |
| 54 | #include <linux/bitops.h> |
| 55 | #include <linux/mm.h> |
| 56 | #include <linux/types.h> |
Andrew Morton | 54e0f52 | 2005-04-30 07:07:04 +0100 | [diff] [blame] | 57 | #include <linux/audit.h> |
Steve Grubb | e7c3497 | 2006-04-03 09:08:13 -0400 | [diff] [blame] | 58 | #include <linux/selinux.h> |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 59 | #include <linux/mutex.h> |
Andrew Morton | 54e0f52 | 2005-04-30 07:07:04 +0100 | [diff] [blame] | 60 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 61 | #include <net/sock.h> |
| 62 | #include <net/scm.h> |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 63 | #include <net/netlink.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 64 | |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 65 | #define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 66 | |
| 67 | struct netlink_sock { |
| 68 | /* struct sock has to be the first member of netlink_sock */ |
| 69 | struct sock sk; |
| 70 | u32 pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 71 | u32 dst_pid; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 72 | u32 dst_group; |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 73 | u32 flags; |
| 74 | u32 subscriptions; |
| 75 | u32 ngroups; |
| 76 | unsigned long *groups; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 77 | unsigned long state; |
| 78 | wait_queue_head_t wait; |
| 79 | struct netlink_callback *cb; |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 80 | struct mutex *cb_mutex; |
| 81 | struct mutex cb_def_mutex; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 82 | void (*data_ready)(struct sock *sk, int bytes); |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 83 | struct module *module; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 84 | }; |
| 85 | |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 86 | #define NETLINK_KERNEL_SOCKET 0x1 |
Patrick McHardy | 9a4595b | 2005-08-15 12:32:15 -0700 | [diff] [blame] | 87 | #define NETLINK_RECV_PKTINFO 0x2 |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 88 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 89 | static inline struct netlink_sock *nlk_sk(struct sock *sk) |
| 90 | { |
| 91 | return (struct netlink_sock *)sk; |
| 92 | } |
| 93 | |
| 94 | struct nl_pid_hash { |
| 95 | struct hlist_head *table; |
| 96 | unsigned long rehash_time; |
| 97 | |
| 98 | unsigned int mask; |
| 99 | unsigned int shift; |
| 100 | |
| 101 | unsigned int entries; |
| 102 | unsigned int max_shift; |
| 103 | |
| 104 | u32 rnd; |
| 105 | }; |
| 106 | |
| 107 | struct netlink_table { |
| 108 | struct nl_pid_hash hash; |
| 109 | struct hlist_head mc_list; |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 110 | unsigned long *listeners; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 111 | unsigned int nl_nonroot; |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 112 | unsigned int groups; |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 113 | struct mutex *cb_mutex; |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 114 | struct module *module; |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 115 | int registered; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 116 | }; |
| 117 | |
| 118 | static struct netlink_table *nl_table; |
| 119 | |
| 120 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); |
| 121 | |
| 122 | static int netlink_dump(struct sock *sk); |
| 123 | static void netlink_destroy_callback(struct netlink_callback *cb); |
| 124 | |
| 125 | static DEFINE_RWLOCK(nl_table_lock); |
| 126 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
| 127 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 128 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 129 | |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 130 | static u32 netlink_group_mask(u32 group) |
| 131 | { |
| 132 | return group ? 1 << (group - 1) : 0; |
| 133 | } |
| 134 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 135 | static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid) |
| 136 | { |
| 137 | return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask]; |
| 138 | } |
| 139 | |
| 140 | static void netlink_sock_destruct(struct sock *sk) |
| 141 | { |
| 142 | skb_queue_purge(&sk->sk_receive_queue); |
| 143 | |
| 144 | if (!sock_flag(sk, SOCK_DEAD)) { |
| 145 | printk("Freeing alive netlink socket %p\n", sk); |
| 146 | return; |
| 147 | } |
| 148 | BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc)); |
| 149 | BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc)); |
| 150 | BUG_TRAP(!nlk_sk(sk)->cb); |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 151 | BUG_TRAP(!nlk_sk(sk)->groups); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 152 | } |
| 153 | |
| 154 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP. |
| 155 | * Look, when several writers sleep and reader wakes them up, all but one |
| 156 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
| 157 | * this, _but_ remember, it adds useless work on UP machines. |
| 158 | */ |
| 159 | |
| 160 | static void netlink_table_grab(void) |
| 161 | { |
Arjan van de Ven | 6abd219 | 2006-07-03 00:24:07 -0700 | [diff] [blame] | 162 | write_lock_irq(&nl_table_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 163 | |
| 164 | if (atomic_read(&nl_table_users)) { |
| 165 | DECLARE_WAITQUEUE(wait, current); |
| 166 | |
| 167 | add_wait_queue_exclusive(&nl_table_wait, &wait); |
| 168 | for(;;) { |
| 169 | set_current_state(TASK_UNINTERRUPTIBLE); |
| 170 | if (atomic_read(&nl_table_users) == 0) |
| 171 | break; |
Arjan van de Ven | 6abd219 | 2006-07-03 00:24:07 -0700 | [diff] [blame] | 172 | write_unlock_irq(&nl_table_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 173 | schedule(); |
Arjan van de Ven | 6abd219 | 2006-07-03 00:24:07 -0700 | [diff] [blame] | 174 | write_lock_irq(&nl_table_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 175 | } |
| 176 | |
| 177 | __set_current_state(TASK_RUNNING); |
| 178 | remove_wait_queue(&nl_table_wait, &wait); |
| 179 | } |
| 180 | } |
| 181 | |
| 182 | static __inline__ void netlink_table_ungrab(void) |
| 183 | { |
Arjan van de Ven | 6abd219 | 2006-07-03 00:24:07 -0700 | [diff] [blame] | 184 | write_unlock_irq(&nl_table_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 185 | wake_up(&nl_table_wait); |
| 186 | } |
| 187 | |
| 188 | static __inline__ void |
| 189 | netlink_lock_table(void) |
| 190 | { |
| 191 | /* read_lock() synchronizes us to netlink_table_grab */ |
| 192 | |
| 193 | read_lock(&nl_table_lock); |
| 194 | atomic_inc(&nl_table_users); |
| 195 | read_unlock(&nl_table_lock); |
| 196 | } |
| 197 | |
| 198 | static __inline__ void |
| 199 | netlink_unlock_table(void) |
| 200 | { |
| 201 | if (atomic_dec_and_test(&nl_table_users)) |
| 202 | wake_up(&nl_table_wait); |
| 203 | } |
| 204 | |
| 205 | static __inline__ struct sock *netlink_lookup(int protocol, u32 pid) |
| 206 | { |
| 207 | struct nl_pid_hash *hash = &nl_table[protocol].hash; |
| 208 | struct hlist_head *head; |
| 209 | struct sock *sk; |
| 210 | struct hlist_node *node; |
| 211 | |
| 212 | read_lock(&nl_table_lock); |
| 213 | head = nl_pid_hashfn(hash, pid); |
| 214 | sk_for_each(sk, node, head) { |
| 215 | if (nlk_sk(sk)->pid == pid) { |
| 216 | sock_hold(sk); |
| 217 | goto found; |
| 218 | } |
| 219 | } |
| 220 | sk = NULL; |
| 221 | found: |
| 222 | read_unlock(&nl_table_lock); |
| 223 | return sk; |
| 224 | } |
| 225 | |
| 226 | static inline struct hlist_head *nl_pid_hash_alloc(size_t size) |
| 227 | { |
| 228 | if (size <= PAGE_SIZE) |
| 229 | return kmalloc(size, GFP_ATOMIC); |
| 230 | else |
| 231 | return (struct hlist_head *) |
| 232 | __get_free_pages(GFP_ATOMIC, get_order(size)); |
| 233 | } |
| 234 | |
| 235 | static inline void nl_pid_hash_free(struct hlist_head *table, size_t size) |
| 236 | { |
| 237 | if (size <= PAGE_SIZE) |
| 238 | kfree(table); |
| 239 | else |
| 240 | free_pages((unsigned long)table, get_order(size)); |
| 241 | } |
| 242 | |
| 243 | static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow) |
| 244 | { |
| 245 | unsigned int omask, mask, shift; |
| 246 | size_t osize, size; |
| 247 | struct hlist_head *otable, *table; |
| 248 | int i; |
| 249 | |
| 250 | omask = mask = hash->mask; |
| 251 | osize = size = (mask + 1) * sizeof(*table); |
| 252 | shift = hash->shift; |
| 253 | |
| 254 | if (grow) { |
| 255 | if (++shift > hash->max_shift) |
| 256 | return 0; |
| 257 | mask = mask * 2 + 1; |
| 258 | size *= 2; |
| 259 | } |
| 260 | |
| 261 | table = nl_pid_hash_alloc(size); |
| 262 | if (!table) |
| 263 | return 0; |
| 264 | |
| 265 | memset(table, 0, size); |
| 266 | otable = hash->table; |
| 267 | hash->table = table; |
| 268 | hash->mask = mask; |
| 269 | hash->shift = shift; |
| 270 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); |
| 271 | |
| 272 | for (i = 0; i <= omask; i++) { |
| 273 | struct sock *sk; |
| 274 | struct hlist_node *node, *tmp; |
| 275 | |
| 276 | sk_for_each_safe(sk, node, tmp, &otable[i]) |
| 277 | __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid)); |
| 278 | } |
| 279 | |
| 280 | nl_pid_hash_free(otable, osize); |
| 281 | hash->rehash_time = jiffies + 10 * 60 * HZ; |
| 282 | return 1; |
| 283 | } |
| 284 | |
| 285 | static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len) |
| 286 | { |
| 287 | int avg = hash->entries >> hash->shift; |
| 288 | |
| 289 | if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1)) |
| 290 | return 1; |
| 291 | |
| 292 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { |
| 293 | nl_pid_hash_rehash(hash, 0); |
| 294 | return 1; |
| 295 | } |
| 296 | |
| 297 | return 0; |
| 298 | } |
| 299 | |
Eric Dumazet | 90ddc4f | 2005-12-22 12:49:22 -0800 | [diff] [blame] | 300 | static const struct proto_ops netlink_ops; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 301 | |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 302 | static void |
| 303 | netlink_update_listeners(struct sock *sk) |
| 304 | { |
| 305 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
| 306 | struct hlist_node *node; |
| 307 | unsigned long mask; |
| 308 | unsigned int i; |
| 309 | |
| 310 | for (i = 0; i < NLGRPSZ(tbl->groups)/sizeof(unsigned long); i++) { |
| 311 | mask = 0; |
| 312 | sk_for_each_bound(sk, node, &tbl->mc_list) |
| 313 | mask |= nlk_sk(sk)->groups[i]; |
| 314 | tbl->listeners[i] = mask; |
| 315 | } |
| 316 | /* this function is only called with the netlink table "grabbed", which |
| 317 | * makes sure updates are visible before bind or setsockopt return. */ |
| 318 | } |
| 319 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 320 | static int netlink_insert(struct sock *sk, u32 pid) |
| 321 | { |
| 322 | struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; |
| 323 | struct hlist_head *head; |
| 324 | int err = -EADDRINUSE; |
| 325 | struct sock *osk; |
| 326 | struct hlist_node *node; |
| 327 | int len; |
| 328 | |
| 329 | netlink_table_grab(); |
| 330 | head = nl_pid_hashfn(hash, pid); |
| 331 | len = 0; |
| 332 | sk_for_each(osk, node, head) { |
| 333 | if (nlk_sk(osk)->pid == pid) |
| 334 | break; |
| 335 | len++; |
| 336 | } |
| 337 | if (node) |
| 338 | goto err; |
| 339 | |
| 340 | err = -EBUSY; |
| 341 | if (nlk_sk(sk)->pid) |
| 342 | goto err; |
| 343 | |
| 344 | err = -ENOMEM; |
| 345 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) |
| 346 | goto err; |
| 347 | |
| 348 | if (len && nl_pid_hash_dilute(hash, len)) |
| 349 | head = nl_pid_hashfn(hash, pid); |
| 350 | hash->entries++; |
| 351 | nlk_sk(sk)->pid = pid; |
| 352 | sk_add_node(sk, head); |
| 353 | err = 0; |
| 354 | |
| 355 | err: |
| 356 | netlink_table_ungrab(); |
| 357 | return err; |
| 358 | } |
| 359 | |
| 360 | static void netlink_remove(struct sock *sk) |
| 361 | { |
| 362 | netlink_table_grab(); |
David S. Miller | d470e3b | 2005-06-26 15:31:51 -0700 | [diff] [blame] | 363 | if (sk_del_node_init(sk)) |
| 364 | nl_table[sk->sk_protocol].hash.entries--; |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 365 | if (nlk_sk(sk)->subscriptions) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 366 | __sk_del_bind_node(sk); |
| 367 | netlink_table_ungrab(); |
| 368 | } |
| 369 | |
| 370 | static struct proto netlink_proto = { |
| 371 | .name = "NETLINK", |
| 372 | .owner = THIS_MODULE, |
| 373 | .obj_size = sizeof(struct netlink_sock), |
| 374 | }; |
| 375 | |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 376 | static int __netlink_create(struct socket *sock, struct mutex *cb_mutex, |
| 377 | int protocol) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 378 | { |
| 379 | struct sock *sk; |
| 380 | struct netlink_sock *nlk; |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 381 | |
| 382 | sock->ops = &netlink_ops; |
| 383 | |
| 384 | sk = sk_alloc(PF_NETLINK, GFP_KERNEL, &netlink_proto, 1); |
| 385 | if (!sk) |
| 386 | return -ENOMEM; |
| 387 | |
| 388 | sock_init_data(sock, sk); |
| 389 | |
| 390 | nlk = nlk_sk(sk); |
Patrick McHardy | ffa4d72 | 2007-04-25 14:01:17 -0700 | [diff] [blame] | 391 | if (cb_mutex) |
| 392 | nlk->cb_mutex = cb_mutex; |
| 393 | else { |
| 394 | nlk->cb_mutex = &nlk->cb_def_mutex; |
| 395 | mutex_init(nlk->cb_mutex); |
| 396 | } |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 397 | init_waitqueue_head(&nlk->wait); |
| 398 | |
| 399 | sk->sk_destruct = netlink_sock_destruct; |
| 400 | sk->sk_protocol = protocol; |
| 401 | return 0; |
| 402 | } |
| 403 | |
| 404 | static int netlink_create(struct socket *sock, int protocol) |
| 405 | { |
| 406 | struct module *module = NULL; |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 407 | struct mutex *cb_mutex; |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 408 | struct netlink_sock *nlk; |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 409 | int err = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 410 | |
| 411 | sock->state = SS_UNCONNECTED; |
| 412 | |
| 413 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) |
| 414 | return -ESOCKTNOSUPPORT; |
| 415 | |
| 416 | if (protocol<0 || protocol >= MAX_LINKS) |
| 417 | return -EPROTONOSUPPORT; |
| 418 | |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 419 | netlink_lock_table(); |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 420 | #ifdef CONFIG_KMOD |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 421 | if (!nl_table[protocol].registered) { |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 422 | netlink_unlock_table(); |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 423 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 424 | netlink_lock_table(); |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 425 | } |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 426 | #endif |
| 427 | if (nl_table[protocol].registered && |
| 428 | try_module_get(nl_table[protocol].module)) |
| 429 | module = nl_table[protocol].module; |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 430 | cb_mutex = nl_table[protocol].cb_mutex; |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 431 | netlink_unlock_table(); |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 432 | |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 433 | if ((err = __netlink_create(sock, cb_mutex, protocol)) < 0) |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 434 | goto out_module; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 435 | |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 436 | nlk = nlk_sk(sock->sk); |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 437 | nlk->module = module; |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 438 | out: |
| 439 | return err; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 440 | |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 441 | out_module: |
| 442 | module_put(module); |
| 443 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 444 | } |
| 445 | |
| 446 | static int netlink_release(struct socket *sock) |
| 447 | { |
| 448 | struct sock *sk = sock->sk; |
| 449 | struct netlink_sock *nlk; |
| 450 | |
| 451 | if (!sk) |
| 452 | return 0; |
| 453 | |
| 454 | netlink_remove(sk); |
Denis Lunev | ac57b3a | 2007-04-18 17:05:58 -0700 | [diff] [blame] | 455 | sock_orphan(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 456 | nlk = nlk_sk(sk); |
| 457 | |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 458 | mutex_lock(nlk->cb_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 459 | if (nlk->cb) { |
Thomas Graf | a8f74b2 | 2005-11-10 02:25:52 +0100 | [diff] [blame] | 460 | if (nlk->cb->done) |
| 461 | nlk->cb->done(nlk->cb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 462 | netlink_destroy_callback(nlk->cb); |
| 463 | nlk->cb = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 464 | } |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 465 | mutex_unlock(nlk->cb_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 466 | |
| 467 | /* OK. Socket is unlinked, and, therefore, |
| 468 | no new packets will arrive */ |
| 469 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 470 | sock->sk = NULL; |
| 471 | wake_up_interruptible_all(&nlk->wait); |
| 472 | |
| 473 | skb_queue_purge(&sk->sk_write_queue); |
| 474 | |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 475 | if (nlk->pid && !nlk->subscriptions) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 476 | struct netlink_notify n = { |
| 477 | .protocol = sk->sk_protocol, |
| 478 | .pid = nlk->pid, |
| 479 | }; |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 480 | atomic_notifier_call_chain(&netlink_chain, |
| 481 | NETLINK_URELEASE, &n); |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 482 | } |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 483 | |
Mariusz Kozlowski | 5e7c001 | 2007-01-02 15:24:30 -0800 | [diff] [blame] | 484 | module_put(nlk->module); |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 485 | |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 486 | netlink_table_grab(); |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 487 | if (nlk->flags & NETLINK_KERNEL_SOCKET) { |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 488 | kfree(nl_table[sk->sk_protocol].listeners); |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 489 | nl_table[sk->sk_protocol].module = NULL; |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 490 | nl_table[sk->sk_protocol].registered = 0; |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 491 | } else if (nlk->subscriptions) |
| 492 | netlink_update_listeners(sk); |
| 493 | netlink_table_ungrab(); |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 494 | |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 495 | kfree(nlk->groups); |
| 496 | nlk->groups = NULL; |
| 497 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 498 | sock_put(sk); |
| 499 | return 0; |
| 500 | } |
| 501 | |
| 502 | static int netlink_autobind(struct socket *sock) |
| 503 | { |
| 504 | struct sock *sk = sock->sk; |
| 505 | struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; |
| 506 | struct hlist_head *head; |
| 507 | struct sock *osk; |
| 508 | struct hlist_node *node; |
Herbert Xu | c27bd49 | 2005-11-22 14:41:50 -0800 | [diff] [blame] | 509 | s32 pid = current->tgid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 510 | int err; |
| 511 | static s32 rover = -4097; |
| 512 | |
| 513 | retry: |
| 514 | cond_resched(); |
| 515 | netlink_table_grab(); |
| 516 | head = nl_pid_hashfn(hash, pid); |
| 517 | sk_for_each(osk, node, head) { |
| 518 | if (nlk_sk(osk)->pid == pid) { |
| 519 | /* Bind collision, search negative pid values. */ |
| 520 | pid = rover--; |
| 521 | if (rover > -4097) |
| 522 | rover = -4097; |
| 523 | netlink_table_ungrab(); |
| 524 | goto retry; |
| 525 | } |
| 526 | } |
| 527 | netlink_table_ungrab(); |
| 528 | |
| 529 | err = netlink_insert(sk, pid); |
| 530 | if (err == -EADDRINUSE) |
| 531 | goto retry; |
David S. Miller | d470e3b | 2005-06-26 15:31:51 -0700 | [diff] [blame] | 532 | |
| 533 | /* If 2 threads race to autobind, that is fine. */ |
| 534 | if (err == -EBUSY) |
| 535 | err = 0; |
| 536 | |
| 537 | return err; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 538 | } |
| 539 | |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 540 | static inline int netlink_capable(struct socket *sock, unsigned int flag) |
| 541 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 542 | return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) || |
| 543 | capable(CAP_NET_ADMIN); |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 544 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 545 | |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 546 | static void |
| 547 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) |
| 548 | { |
| 549 | struct netlink_sock *nlk = nlk_sk(sk); |
| 550 | |
| 551 | if (nlk->subscriptions && !subscriptions) |
| 552 | __sk_del_bind_node(sk); |
| 553 | else if (!nlk->subscriptions && subscriptions) |
| 554 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); |
| 555 | nlk->subscriptions = subscriptions; |
| 556 | } |
| 557 | |
Patrick McHardy | 513c250 | 2005-09-06 15:43:59 -0700 | [diff] [blame] | 558 | static int netlink_alloc_groups(struct sock *sk) |
| 559 | { |
| 560 | struct netlink_sock *nlk = nlk_sk(sk); |
| 561 | unsigned int groups; |
| 562 | int err = 0; |
| 563 | |
| 564 | netlink_lock_table(); |
| 565 | groups = nl_table[sk->sk_protocol].groups; |
| 566 | if (!nl_table[sk->sk_protocol].registered) |
| 567 | err = -ENOENT; |
| 568 | netlink_unlock_table(); |
| 569 | |
| 570 | if (err) |
| 571 | return err; |
| 572 | |
Panagiotis Issaris | 0da974f | 2006-07-21 14:51:30 -0700 | [diff] [blame] | 573 | nlk->groups = kzalloc(NLGRPSZ(groups), GFP_KERNEL); |
Patrick McHardy | 513c250 | 2005-09-06 15:43:59 -0700 | [diff] [blame] | 574 | if (nlk->groups == NULL) |
| 575 | return -ENOMEM; |
Patrick McHardy | 513c250 | 2005-09-06 15:43:59 -0700 | [diff] [blame] | 576 | nlk->ngroups = groups; |
| 577 | return 0; |
| 578 | } |
| 579 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 580 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len) |
| 581 | { |
| 582 | struct sock *sk = sock->sk; |
| 583 | struct netlink_sock *nlk = nlk_sk(sk); |
| 584 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
| 585 | int err; |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 586 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 587 | if (nladdr->nl_family != AF_NETLINK) |
| 588 | return -EINVAL; |
| 589 | |
| 590 | /* Only superuser is allowed to listen multicasts */ |
Patrick McHardy | 513c250 | 2005-09-06 15:43:59 -0700 | [diff] [blame] | 591 | if (nladdr->nl_groups) { |
| 592 | if (!netlink_capable(sock, NL_NONROOT_RECV)) |
| 593 | return -EPERM; |
| 594 | if (nlk->groups == NULL) { |
| 595 | err = netlink_alloc_groups(sk); |
| 596 | if (err) |
| 597 | return err; |
| 598 | } |
| 599 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 600 | |
| 601 | if (nlk->pid) { |
| 602 | if (nladdr->nl_pid != nlk->pid) |
| 603 | return -EINVAL; |
| 604 | } else { |
| 605 | err = nladdr->nl_pid ? |
| 606 | netlink_insert(sk, nladdr->nl_pid) : |
| 607 | netlink_autobind(sock); |
| 608 | if (err) |
| 609 | return err; |
| 610 | } |
| 611 | |
Patrick McHardy | 513c250 | 2005-09-06 15:43:59 -0700 | [diff] [blame] | 612 | if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 613 | return 0; |
| 614 | |
| 615 | netlink_table_grab(); |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 616 | netlink_update_subscriptions(sk, nlk->subscriptions + |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 617 | hweight32(nladdr->nl_groups) - |
| 618 | hweight32(nlk->groups[0])); |
| 619 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 620 | netlink_update_listeners(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 621 | netlink_table_ungrab(); |
| 622 | |
| 623 | return 0; |
| 624 | } |
| 625 | |
| 626 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, |
| 627 | int alen, int flags) |
| 628 | { |
| 629 | int err = 0; |
| 630 | struct sock *sk = sock->sk; |
| 631 | struct netlink_sock *nlk = nlk_sk(sk); |
| 632 | struct sockaddr_nl *nladdr=(struct sockaddr_nl*)addr; |
| 633 | |
| 634 | if (addr->sa_family == AF_UNSPEC) { |
| 635 | sk->sk_state = NETLINK_UNCONNECTED; |
| 636 | nlk->dst_pid = 0; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 637 | nlk->dst_group = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 638 | return 0; |
| 639 | } |
| 640 | if (addr->sa_family != AF_NETLINK) |
| 641 | return -EINVAL; |
| 642 | |
| 643 | /* Only superuser is allowed to send multicasts */ |
| 644 | if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND)) |
| 645 | return -EPERM; |
| 646 | |
| 647 | if (!nlk->pid) |
| 648 | err = netlink_autobind(sock); |
| 649 | |
| 650 | if (err == 0) { |
| 651 | sk->sk_state = NETLINK_CONNECTED; |
| 652 | nlk->dst_pid = nladdr->nl_pid; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 653 | nlk->dst_group = ffs(nladdr->nl_groups); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 654 | } |
| 655 | |
| 656 | return err; |
| 657 | } |
| 658 | |
| 659 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, int *addr_len, int peer) |
| 660 | { |
| 661 | struct sock *sk = sock->sk; |
| 662 | struct netlink_sock *nlk = nlk_sk(sk); |
| 663 | struct sockaddr_nl *nladdr=(struct sockaddr_nl *)addr; |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 664 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 665 | nladdr->nl_family = AF_NETLINK; |
| 666 | nladdr->nl_pad = 0; |
| 667 | *addr_len = sizeof(*nladdr); |
| 668 | |
| 669 | if (peer) { |
| 670 | nladdr->nl_pid = nlk->dst_pid; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 671 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 672 | } else { |
| 673 | nladdr->nl_pid = nlk->pid; |
Patrick McHardy | 513c250 | 2005-09-06 15:43:59 -0700 | [diff] [blame] | 674 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 675 | } |
| 676 | return 0; |
| 677 | } |
| 678 | |
| 679 | static void netlink_overrun(struct sock *sk) |
| 680 | { |
| 681 | if (!test_and_set_bit(0, &nlk_sk(sk)->state)) { |
| 682 | sk->sk_err = ENOBUFS; |
| 683 | sk->sk_error_report(sk); |
| 684 | } |
| 685 | } |
| 686 | |
| 687 | static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid) |
| 688 | { |
| 689 | int protocol = ssk->sk_protocol; |
| 690 | struct sock *sock; |
| 691 | struct netlink_sock *nlk; |
| 692 | |
| 693 | sock = netlink_lookup(protocol, pid); |
| 694 | if (!sock) |
| 695 | return ERR_PTR(-ECONNREFUSED); |
| 696 | |
| 697 | /* Don't bother queuing skb if kernel socket has no input function */ |
| 698 | nlk = nlk_sk(sock); |
| 699 | if ((nlk->pid == 0 && !nlk->data_ready) || |
| 700 | (sock->sk_state == NETLINK_CONNECTED && |
| 701 | nlk->dst_pid != nlk_sk(ssk)->pid)) { |
| 702 | sock_put(sock); |
| 703 | return ERR_PTR(-ECONNREFUSED); |
| 704 | } |
| 705 | return sock; |
| 706 | } |
| 707 | |
| 708 | struct sock *netlink_getsockbyfilp(struct file *filp) |
| 709 | { |
Josef Sipek | 6db5fc5 | 2006-12-08 02:37:25 -0800 | [diff] [blame] | 710 | struct inode *inode = filp->f_path.dentry->d_inode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 711 | struct sock *sock; |
| 712 | |
| 713 | if (!S_ISSOCK(inode->i_mode)) |
| 714 | return ERR_PTR(-ENOTSOCK); |
| 715 | |
| 716 | sock = SOCKET_I(inode)->sk; |
| 717 | if (sock->sk_family != AF_NETLINK) |
| 718 | return ERR_PTR(-EINVAL); |
| 719 | |
| 720 | sock_hold(sock); |
| 721 | return sock; |
| 722 | } |
| 723 | |
| 724 | /* |
| 725 | * Attach a skb to a netlink socket. |
| 726 | * The caller must hold a reference to the destination socket. On error, the |
| 727 | * reference is dropped. The skb is not send to the destination, just all |
| 728 | * all error checks are performed and memory in the queue is reserved. |
| 729 | * Return values: |
| 730 | * < 0: error. skb freed, reference to sock dropped. |
| 731 | * 0: continue |
| 732 | * 1: repeat lookup - reference dropped while waiting for socket memory. |
| 733 | */ |
Alexey Kuznetsov | a70ea99 | 2006-02-09 16:40:11 -0800 | [diff] [blame] | 734 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock, |
| 735 | long timeo, struct sock *ssk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 736 | { |
| 737 | struct netlink_sock *nlk; |
| 738 | |
| 739 | nlk = nlk_sk(sk); |
| 740 | |
| 741 | if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
| 742 | test_bit(0, &nlk->state)) { |
| 743 | DECLARE_WAITQUEUE(wait, current); |
| 744 | if (!timeo) { |
Alexey Kuznetsov | a70ea99 | 2006-02-09 16:40:11 -0800 | [diff] [blame] | 745 | if (!ssk || nlk_sk(ssk)->pid == 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 746 | netlink_overrun(sk); |
| 747 | sock_put(sk); |
| 748 | kfree_skb(skb); |
| 749 | return -EAGAIN; |
| 750 | } |
| 751 | |
| 752 | __set_current_state(TASK_INTERRUPTIBLE); |
| 753 | add_wait_queue(&nlk->wait, &wait); |
| 754 | |
| 755 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
| 756 | test_bit(0, &nlk->state)) && |
| 757 | !sock_flag(sk, SOCK_DEAD)) |
| 758 | timeo = schedule_timeout(timeo); |
| 759 | |
| 760 | __set_current_state(TASK_RUNNING); |
| 761 | remove_wait_queue(&nlk->wait, &wait); |
| 762 | sock_put(sk); |
| 763 | |
| 764 | if (signal_pending(current)) { |
| 765 | kfree_skb(skb); |
| 766 | return sock_intr_errno(timeo); |
| 767 | } |
| 768 | return 1; |
| 769 | } |
| 770 | skb_set_owner_r(skb, sk); |
| 771 | return 0; |
| 772 | } |
| 773 | |
| 774 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb, int protocol) |
| 775 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 776 | int len = skb->len; |
| 777 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 778 | skb_queue_tail(&sk->sk_receive_queue, skb); |
| 779 | sk->sk_data_ready(sk, len); |
| 780 | sock_put(sk); |
| 781 | return len; |
| 782 | } |
| 783 | |
| 784 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) |
| 785 | { |
| 786 | kfree_skb(skb); |
| 787 | sock_put(sk); |
| 788 | } |
| 789 | |
Victor Fusco | 37da647 | 2005-07-18 13:35:43 -0700 | [diff] [blame] | 790 | static inline struct sk_buff *netlink_trim(struct sk_buff *skb, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 791 | gfp_t allocation) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 792 | { |
| 793 | int delta; |
| 794 | |
| 795 | skb_orphan(skb); |
| 796 | |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 797 | delta = skb->end - skb->tail; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 798 | if (delta * 2 < skb->truesize) |
| 799 | return skb; |
| 800 | |
| 801 | if (skb_shared(skb)) { |
| 802 | struct sk_buff *nskb = skb_clone(skb, allocation); |
| 803 | if (!nskb) |
| 804 | return skb; |
| 805 | kfree_skb(skb); |
| 806 | skb = nskb; |
| 807 | } |
| 808 | |
| 809 | if (!pskb_expand_head(skb, 0, -delta, allocation)) |
| 810 | skb->truesize -= delta; |
| 811 | |
| 812 | return skb; |
| 813 | } |
| 814 | |
| 815 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, u32 pid, int nonblock) |
| 816 | { |
| 817 | struct sock *sk; |
| 818 | int err; |
| 819 | long timeo; |
| 820 | |
| 821 | skb = netlink_trim(skb, gfp_any()); |
| 822 | |
| 823 | timeo = sock_sndtimeo(ssk, nonblock); |
| 824 | retry: |
| 825 | sk = netlink_getsockbypid(ssk, pid); |
| 826 | if (IS_ERR(sk)) { |
| 827 | kfree_skb(skb); |
| 828 | return PTR_ERR(sk); |
| 829 | } |
Alexey Kuznetsov | a70ea99 | 2006-02-09 16:40:11 -0800 | [diff] [blame] | 830 | err = netlink_attachskb(sk, skb, nonblock, timeo, ssk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 831 | if (err == 1) |
| 832 | goto retry; |
| 833 | if (err) |
| 834 | return err; |
| 835 | |
| 836 | return netlink_sendskb(sk, skb, ssk->sk_protocol); |
| 837 | } |
| 838 | |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 839 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
| 840 | { |
| 841 | int res = 0; |
| 842 | |
| 843 | BUG_ON(!(nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET)); |
| 844 | if (group - 1 < nl_table[sk->sk_protocol].groups) |
| 845 | res = test_bit(group - 1, nl_table[sk->sk_protocol].listeners); |
| 846 | return res; |
| 847 | } |
| 848 | EXPORT_SYMBOL_GPL(netlink_has_listeners); |
| 849 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 850 | static __inline__ int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
| 851 | { |
| 852 | struct netlink_sock *nlk = nlk_sk(sk); |
| 853 | |
| 854 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && |
| 855 | !test_bit(0, &nlk->state)) { |
| 856 | skb_set_owner_r(skb, sk); |
| 857 | skb_queue_tail(&sk->sk_receive_queue, skb); |
| 858 | sk->sk_data_ready(sk, skb->len); |
| 859 | return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf; |
| 860 | } |
| 861 | return -1; |
| 862 | } |
| 863 | |
| 864 | struct netlink_broadcast_data { |
| 865 | struct sock *exclude_sk; |
| 866 | u32 pid; |
| 867 | u32 group; |
| 868 | int failure; |
| 869 | int congested; |
| 870 | int delivered; |
Al Viro | 7d877f3 | 2005-10-21 03:20:43 -0400 | [diff] [blame] | 871 | gfp_t allocation; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 872 | struct sk_buff *skb, *skb2; |
| 873 | }; |
| 874 | |
| 875 | static inline int do_one_broadcast(struct sock *sk, |
| 876 | struct netlink_broadcast_data *p) |
| 877 | { |
| 878 | struct netlink_sock *nlk = nlk_sk(sk); |
| 879 | int val; |
| 880 | |
| 881 | if (p->exclude_sk == sk) |
| 882 | goto out; |
| 883 | |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 884 | if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups || |
| 885 | !test_bit(p->group - 1, nlk->groups)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 886 | goto out; |
| 887 | |
| 888 | if (p->failure) { |
| 889 | netlink_overrun(sk); |
| 890 | goto out; |
| 891 | } |
| 892 | |
| 893 | sock_hold(sk); |
| 894 | if (p->skb2 == NULL) { |
Tommy S. Christensen | 68acc02 | 2005-05-19 13:06:35 -0700 | [diff] [blame] | 895 | if (skb_shared(p->skb)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 896 | p->skb2 = skb_clone(p->skb, p->allocation); |
| 897 | } else { |
Tommy S. Christensen | 68acc02 | 2005-05-19 13:06:35 -0700 | [diff] [blame] | 898 | p->skb2 = skb_get(p->skb); |
| 899 | /* |
| 900 | * skb ownership may have been set when |
| 901 | * delivered to a previous socket. |
| 902 | */ |
| 903 | skb_orphan(p->skb2); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 904 | } |
| 905 | } |
| 906 | if (p->skb2 == NULL) { |
| 907 | netlink_overrun(sk); |
| 908 | /* Clone failed. Notify ALL listeners. */ |
| 909 | p->failure = 1; |
| 910 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
| 911 | netlink_overrun(sk); |
| 912 | } else { |
| 913 | p->congested |= val; |
| 914 | p->delivered = 1; |
| 915 | p->skb2 = NULL; |
| 916 | } |
| 917 | sock_put(sk); |
| 918 | |
| 919 | out: |
| 920 | return 0; |
| 921 | } |
| 922 | |
| 923 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 924 | u32 group, gfp_t allocation) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 925 | { |
| 926 | struct netlink_broadcast_data info; |
| 927 | struct hlist_node *node; |
| 928 | struct sock *sk; |
| 929 | |
| 930 | skb = netlink_trim(skb, allocation); |
| 931 | |
| 932 | info.exclude_sk = ssk; |
| 933 | info.pid = pid; |
| 934 | info.group = group; |
| 935 | info.failure = 0; |
| 936 | info.congested = 0; |
| 937 | info.delivered = 0; |
| 938 | info.allocation = allocation; |
| 939 | info.skb = skb; |
| 940 | info.skb2 = NULL; |
| 941 | |
| 942 | /* While we sleep in clone, do not allow to change socket list */ |
| 943 | |
| 944 | netlink_lock_table(); |
| 945 | |
| 946 | sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) |
| 947 | do_one_broadcast(sk, &info); |
| 948 | |
Tommy S. Christensen | aa1c6a6 | 2005-05-19 13:07:32 -0700 | [diff] [blame] | 949 | kfree_skb(skb); |
| 950 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 951 | netlink_unlock_table(); |
| 952 | |
| 953 | if (info.skb2) |
| 954 | kfree_skb(info.skb2); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 955 | |
| 956 | if (info.delivered) { |
| 957 | if (info.congested && (allocation & __GFP_WAIT)) |
| 958 | yield(); |
| 959 | return 0; |
| 960 | } |
| 961 | if (info.failure) |
| 962 | return -ENOBUFS; |
| 963 | return -ESRCH; |
| 964 | } |
| 965 | |
| 966 | struct netlink_set_err_data { |
| 967 | struct sock *exclude_sk; |
| 968 | u32 pid; |
| 969 | u32 group; |
| 970 | int code; |
| 971 | }; |
| 972 | |
| 973 | static inline int do_one_set_err(struct sock *sk, |
| 974 | struct netlink_set_err_data *p) |
| 975 | { |
| 976 | struct netlink_sock *nlk = nlk_sk(sk); |
| 977 | |
| 978 | if (sk == p->exclude_sk) |
| 979 | goto out; |
| 980 | |
Patrick McHardy | f7fa9b1 | 2005-08-15 12:29:13 -0700 | [diff] [blame] | 981 | if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups || |
| 982 | !test_bit(p->group - 1, nlk->groups)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 983 | goto out; |
| 984 | |
| 985 | sk->sk_err = p->code; |
| 986 | sk->sk_error_report(sk); |
| 987 | out: |
| 988 | return 0; |
| 989 | } |
| 990 | |
| 991 | void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code) |
| 992 | { |
| 993 | struct netlink_set_err_data info; |
| 994 | struct hlist_node *node; |
| 995 | struct sock *sk; |
| 996 | |
| 997 | info.exclude_sk = ssk; |
| 998 | info.pid = pid; |
| 999 | info.group = group; |
| 1000 | info.code = code; |
| 1001 | |
| 1002 | read_lock(&nl_table_lock); |
| 1003 | |
| 1004 | sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) |
| 1005 | do_one_set_err(sk, &info); |
| 1006 | |
| 1007 | read_unlock(&nl_table_lock); |
| 1008 | } |
| 1009 | |
Patrick McHardy | 9a4595b | 2005-08-15 12:32:15 -0700 | [diff] [blame] | 1010 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1011 | char __user *optval, int optlen) |
Patrick McHardy | 9a4595b | 2005-08-15 12:32:15 -0700 | [diff] [blame] | 1012 | { |
| 1013 | struct sock *sk = sock->sk; |
| 1014 | struct netlink_sock *nlk = nlk_sk(sk); |
| 1015 | int val = 0, err; |
| 1016 | |
| 1017 | if (level != SOL_NETLINK) |
| 1018 | return -ENOPROTOOPT; |
| 1019 | |
| 1020 | if (optlen >= sizeof(int) && |
| 1021 | get_user(val, (int __user *)optval)) |
| 1022 | return -EFAULT; |
| 1023 | |
| 1024 | switch (optname) { |
| 1025 | case NETLINK_PKTINFO: |
| 1026 | if (val) |
| 1027 | nlk->flags |= NETLINK_RECV_PKTINFO; |
| 1028 | else |
| 1029 | nlk->flags &= ~NETLINK_RECV_PKTINFO; |
| 1030 | err = 0; |
| 1031 | break; |
| 1032 | case NETLINK_ADD_MEMBERSHIP: |
| 1033 | case NETLINK_DROP_MEMBERSHIP: { |
| 1034 | unsigned int subscriptions; |
| 1035 | int old, new = optname == NETLINK_ADD_MEMBERSHIP ? 1 : 0; |
| 1036 | |
| 1037 | if (!netlink_capable(sock, NL_NONROOT_RECV)) |
| 1038 | return -EPERM; |
Patrick McHardy | 513c250 | 2005-09-06 15:43:59 -0700 | [diff] [blame] | 1039 | if (nlk->groups == NULL) { |
| 1040 | err = netlink_alloc_groups(sk); |
| 1041 | if (err) |
| 1042 | return err; |
| 1043 | } |
Patrick McHardy | 9a4595b | 2005-08-15 12:32:15 -0700 | [diff] [blame] | 1044 | if (!val || val - 1 >= nlk->ngroups) |
| 1045 | return -EINVAL; |
| 1046 | netlink_table_grab(); |
| 1047 | old = test_bit(val - 1, nlk->groups); |
| 1048 | subscriptions = nlk->subscriptions - old + new; |
| 1049 | if (new) |
| 1050 | __set_bit(val - 1, nlk->groups); |
| 1051 | else |
| 1052 | __clear_bit(val - 1, nlk->groups); |
| 1053 | netlink_update_subscriptions(sk, subscriptions); |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 1054 | netlink_update_listeners(sk); |
Patrick McHardy | 9a4595b | 2005-08-15 12:32:15 -0700 | [diff] [blame] | 1055 | netlink_table_ungrab(); |
| 1056 | err = 0; |
| 1057 | break; |
| 1058 | } |
| 1059 | default: |
| 1060 | err = -ENOPROTOOPT; |
| 1061 | } |
| 1062 | return err; |
| 1063 | } |
| 1064 | |
| 1065 | static int netlink_getsockopt(struct socket *sock, int level, int optname, |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1066 | char __user *optval, int __user *optlen) |
Patrick McHardy | 9a4595b | 2005-08-15 12:32:15 -0700 | [diff] [blame] | 1067 | { |
| 1068 | struct sock *sk = sock->sk; |
| 1069 | struct netlink_sock *nlk = nlk_sk(sk); |
| 1070 | int len, val, err; |
| 1071 | |
| 1072 | if (level != SOL_NETLINK) |
| 1073 | return -ENOPROTOOPT; |
| 1074 | |
| 1075 | if (get_user(len, optlen)) |
| 1076 | return -EFAULT; |
| 1077 | if (len < 0) |
| 1078 | return -EINVAL; |
| 1079 | |
| 1080 | switch (optname) { |
| 1081 | case NETLINK_PKTINFO: |
| 1082 | if (len < sizeof(int)) |
| 1083 | return -EINVAL; |
| 1084 | len = sizeof(int); |
| 1085 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; |
Heiko Carstens | a27b58f | 2006-10-30 15:06:12 -0800 | [diff] [blame] | 1086 | if (put_user(len, optlen) || |
| 1087 | put_user(val, optval)) |
| 1088 | return -EFAULT; |
Patrick McHardy | 9a4595b | 2005-08-15 12:32:15 -0700 | [diff] [blame] | 1089 | err = 0; |
| 1090 | break; |
| 1091 | default: |
| 1092 | err = -ENOPROTOOPT; |
| 1093 | } |
| 1094 | return err; |
| 1095 | } |
| 1096 | |
| 1097 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) |
| 1098 | { |
| 1099 | struct nl_pktinfo info; |
| 1100 | |
| 1101 | info.group = NETLINK_CB(skb).dst_group; |
| 1102 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); |
| 1103 | } |
| 1104 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1105 | static inline void netlink_rcv_wake(struct sock *sk) |
| 1106 | { |
| 1107 | struct netlink_sock *nlk = nlk_sk(sk); |
| 1108 | |
David S. Miller | b03efcf | 2005-07-08 14:57:23 -0700 | [diff] [blame] | 1109 | if (skb_queue_empty(&sk->sk_receive_queue)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1110 | clear_bit(0, &nlk->state); |
| 1111 | if (!test_bit(0, &nlk->state)) |
| 1112 | wake_up_interruptible(&nlk->wait); |
| 1113 | } |
| 1114 | |
| 1115 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
| 1116 | struct msghdr *msg, size_t len) |
| 1117 | { |
| 1118 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); |
| 1119 | struct sock *sk = sock->sk; |
| 1120 | struct netlink_sock *nlk = nlk_sk(sk); |
| 1121 | struct sockaddr_nl *addr=msg->msg_name; |
| 1122 | u32 dst_pid; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 1123 | u32 dst_group; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1124 | struct sk_buff *skb; |
| 1125 | int err; |
| 1126 | struct scm_cookie scm; |
| 1127 | |
| 1128 | if (msg->msg_flags&MSG_OOB) |
| 1129 | return -EOPNOTSUPP; |
| 1130 | |
| 1131 | if (NULL == siocb->scm) |
| 1132 | siocb->scm = &scm; |
| 1133 | err = scm_send(sock, msg, siocb->scm); |
| 1134 | if (err < 0) |
| 1135 | return err; |
| 1136 | |
| 1137 | if (msg->msg_namelen) { |
| 1138 | if (addr->nl_family != AF_NETLINK) |
| 1139 | return -EINVAL; |
| 1140 | dst_pid = addr->nl_pid; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 1141 | dst_group = ffs(addr->nl_groups); |
| 1142 | if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1143 | return -EPERM; |
| 1144 | } else { |
| 1145 | dst_pid = nlk->dst_pid; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 1146 | dst_group = nlk->dst_group; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1147 | } |
| 1148 | |
| 1149 | if (!nlk->pid) { |
| 1150 | err = netlink_autobind(sock); |
| 1151 | if (err) |
| 1152 | goto out; |
| 1153 | } |
| 1154 | |
| 1155 | err = -EMSGSIZE; |
| 1156 | if (len > sk->sk_sndbuf - 32) |
| 1157 | goto out; |
| 1158 | err = -ENOBUFS; |
Thomas Graf | 339bf98 | 2006-11-10 14:10:15 -0800 | [diff] [blame] | 1159 | skb = alloc_skb(len, GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1160 | if (skb==NULL) |
| 1161 | goto out; |
| 1162 | |
| 1163 | NETLINK_CB(skb).pid = nlk->pid; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 1164 | NETLINK_CB(skb).dst_group = dst_group; |
Serge Hallyn | c94c257 | 2005-04-29 16:27:17 +0100 | [diff] [blame] | 1165 | NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context); |
Steve Grubb | e7c3497 | 2006-04-03 09:08:13 -0400 | [diff] [blame] | 1166 | selinux_get_task_sid(current, &(NETLINK_CB(skb).sid)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1167 | memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred)); |
| 1168 | |
| 1169 | /* What can I do? Netlink is asynchronous, so that |
| 1170 | we will have to save current capabilities to |
| 1171 | check them, when this message will be delivered |
| 1172 | to corresponding kernel module. --ANK (980802) |
| 1173 | */ |
| 1174 | |
| 1175 | err = -EFAULT; |
| 1176 | if (memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len)) { |
| 1177 | kfree_skb(skb); |
| 1178 | goto out; |
| 1179 | } |
| 1180 | |
| 1181 | err = security_netlink_send(sk, skb); |
| 1182 | if (err) { |
| 1183 | kfree_skb(skb); |
| 1184 | goto out; |
| 1185 | } |
| 1186 | |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 1187 | if (dst_group) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1188 | atomic_inc(&skb->users); |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 1189 | netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1190 | } |
| 1191 | err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT); |
| 1192 | |
| 1193 | out: |
| 1194 | return err; |
| 1195 | } |
| 1196 | |
| 1197 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, |
| 1198 | struct msghdr *msg, size_t len, |
| 1199 | int flags) |
| 1200 | { |
| 1201 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); |
| 1202 | struct scm_cookie scm; |
| 1203 | struct sock *sk = sock->sk; |
| 1204 | struct netlink_sock *nlk = nlk_sk(sk); |
| 1205 | int noblock = flags&MSG_DONTWAIT; |
| 1206 | size_t copied; |
| 1207 | struct sk_buff *skb; |
| 1208 | int err; |
| 1209 | |
| 1210 | if (flags&MSG_OOB) |
| 1211 | return -EOPNOTSUPP; |
| 1212 | |
| 1213 | copied = 0; |
| 1214 | |
| 1215 | skb = skb_recv_datagram(sk,flags,noblock,&err); |
| 1216 | if (skb==NULL) |
| 1217 | goto out; |
| 1218 | |
| 1219 | msg->msg_namelen = 0; |
| 1220 | |
| 1221 | copied = skb->len; |
| 1222 | if (len < copied) { |
| 1223 | msg->msg_flags |= MSG_TRUNC; |
| 1224 | copied = len; |
| 1225 | } |
| 1226 | |
Arnaldo Carvalho de Melo | badff6d | 2007-03-13 13:06:52 -0300 | [diff] [blame] | 1227 | skb_reset_transport_header(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1228 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); |
| 1229 | |
| 1230 | if (msg->msg_name) { |
| 1231 | struct sockaddr_nl *addr = (struct sockaddr_nl*)msg->msg_name; |
| 1232 | addr->nl_family = AF_NETLINK; |
| 1233 | addr->nl_pad = 0; |
| 1234 | addr->nl_pid = NETLINK_CB(skb).pid; |
Patrick McHardy | d629b83 | 2005-08-14 19:27:50 -0700 | [diff] [blame] | 1235 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1236 | msg->msg_namelen = sizeof(*addr); |
| 1237 | } |
| 1238 | |
Patrick McHardy | cc9a06c | 2006-03-12 20:34:27 -0800 | [diff] [blame] | 1239 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
| 1240 | netlink_cmsg_recv_pktinfo(msg, skb); |
| 1241 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1242 | if (NULL == siocb->scm) { |
| 1243 | memset(&scm, 0, sizeof(scm)); |
| 1244 | siocb->scm = &scm; |
| 1245 | } |
| 1246 | siocb->scm->creds = *NETLINK_CREDS(skb); |
| 1247 | skb_free_datagram(sk, skb); |
| 1248 | |
| 1249 | if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) |
| 1250 | netlink_dump(sk); |
| 1251 | |
| 1252 | scm_recv(sock, msg, siocb->scm, flags); |
| 1253 | |
David S. Miller | b558ff7 | 2007-03-06 17:02:35 -0800 | [diff] [blame] | 1254 | if (flags & MSG_TRUNC) |
| 1255 | copied = skb->len; |
| 1256 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1257 | out: |
| 1258 | netlink_rcv_wake(sk); |
| 1259 | return err ? : copied; |
| 1260 | } |
| 1261 | |
| 1262 | static void netlink_data_ready(struct sock *sk, int len) |
| 1263 | { |
| 1264 | struct netlink_sock *nlk = nlk_sk(sk); |
| 1265 | |
| 1266 | if (nlk->data_ready) |
| 1267 | nlk->data_ready(sk, len); |
| 1268 | netlink_rcv_wake(sk); |
| 1269 | } |
| 1270 | |
| 1271 | /* |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1272 | * We export these functions to other modules. They provide a |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1273 | * complete set of kernel non-blocking support for message |
| 1274 | * queueing. |
| 1275 | */ |
| 1276 | |
| 1277 | struct sock * |
Patrick McHardy | 0662860 | 2005-08-15 12:33:26 -0700 | [diff] [blame] | 1278 | netlink_kernel_create(int unit, unsigned int groups, |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1279 | void (*input)(struct sock *sk, int len), |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1280 | struct mutex *cb_mutex, struct module *module) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1281 | { |
| 1282 | struct socket *sock; |
| 1283 | struct sock *sk; |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 1284 | struct netlink_sock *nlk; |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 1285 | unsigned long *listeners = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1286 | |
Akinobu Mita | fab2caf | 2006-08-29 02:15:24 -0700 | [diff] [blame] | 1287 | BUG_ON(!nl_table); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1288 | |
| 1289 | if (unit<0 || unit>=MAX_LINKS) |
| 1290 | return NULL; |
| 1291 | |
| 1292 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) |
| 1293 | return NULL; |
| 1294 | |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1295 | if (__netlink_create(sock, cb_mutex, unit) < 0) |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 1296 | goto out_sock_release; |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 1297 | |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 1298 | if (groups < 32) |
| 1299 | groups = 32; |
| 1300 | |
| 1301 | listeners = kzalloc(NLGRPSZ(groups), GFP_KERNEL); |
| 1302 | if (!listeners) |
| 1303 | goto out_sock_release; |
| 1304 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1305 | sk = sock->sk; |
| 1306 | sk->sk_data_ready = netlink_data_ready; |
| 1307 | if (input) |
| 1308 | nlk_sk(sk)->data_ready = input; |
| 1309 | |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 1310 | if (netlink_insert(sk, 0)) |
| 1311 | goto out_sock_release; |
| 1312 | |
| 1313 | nlk = nlk_sk(sk); |
| 1314 | nlk->flags |= NETLINK_KERNEL_SOCKET; |
| 1315 | |
| 1316 | netlink_table_grab(); |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 1317 | nl_table[unit].groups = groups; |
| 1318 | nl_table[unit].listeners = listeners; |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1319 | nl_table[unit].cb_mutex = cb_mutex; |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 1320 | nl_table[unit].module = module; |
Patrick McHardy | ab33a17 | 2005-08-14 19:31:36 -0700 | [diff] [blame] | 1321 | nl_table[unit].registered = 1; |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 1322 | netlink_table_ungrab(); |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 1323 | |
| 1324 | return sk; |
| 1325 | |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 1326 | out_sock_release: |
Patrick McHardy | 4277a08 | 2006-03-20 18:52:01 -0800 | [diff] [blame] | 1327 | kfree(listeners); |
Harald Welte | 4fdb3bb | 2005-08-09 19:40:55 -0700 | [diff] [blame] | 1328 | sock_release(sock); |
Patrick McHardy | 77247bb | 2005-08-14 19:27:13 -0700 | [diff] [blame] | 1329 | return NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1330 | } |
| 1331 | |
| 1332 | void netlink_set_nonroot(int protocol, unsigned int flags) |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1333 | { |
| 1334 | if ((unsigned int)protocol < MAX_LINKS) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1335 | nl_table[protocol].nl_nonroot = flags; |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1336 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1337 | |
| 1338 | static void netlink_destroy_callback(struct netlink_callback *cb) |
| 1339 | { |
| 1340 | if (cb->skb) |
| 1341 | kfree_skb(cb->skb); |
| 1342 | kfree(cb); |
| 1343 | } |
| 1344 | |
| 1345 | /* |
| 1346 | * It looks a bit ugly. |
| 1347 | * It would be better to create kernel thread. |
| 1348 | */ |
| 1349 | |
| 1350 | static int netlink_dump(struct sock *sk) |
| 1351 | { |
| 1352 | struct netlink_sock *nlk = nlk_sk(sk); |
| 1353 | struct netlink_callback *cb; |
| 1354 | struct sk_buff *skb; |
| 1355 | struct nlmsghdr *nlh; |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1356 | int len, err = -ENOBUFS; |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1357 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1358 | skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL); |
| 1359 | if (!skb) |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1360 | goto errout; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1361 | |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1362 | mutex_lock(nlk->cb_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1363 | |
| 1364 | cb = nlk->cb; |
| 1365 | if (cb == NULL) { |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1366 | err = -EINVAL; |
| 1367 | goto errout_skb; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1368 | } |
| 1369 | |
| 1370 | len = cb->dump(skb, cb); |
| 1371 | |
| 1372 | if (len > 0) { |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1373 | mutex_unlock(nlk->cb_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1374 | skb_queue_tail(&sk->sk_receive_queue, skb); |
| 1375 | sk->sk_data_ready(sk, len); |
| 1376 | return 0; |
| 1377 | } |
| 1378 | |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1379 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
| 1380 | if (!nlh) |
| 1381 | goto errout_skb; |
| 1382 | |
| 1383 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
| 1384 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1385 | skb_queue_tail(&sk->sk_receive_queue, skb); |
| 1386 | sk->sk_data_ready(sk, skb->len); |
| 1387 | |
Thomas Graf | a8f74b2 | 2005-11-10 02:25:52 +0100 | [diff] [blame] | 1388 | if (cb->done) |
| 1389 | cb->done(cb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1390 | nlk->cb = NULL; |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1391 | mutex_unlock(nlk->cb_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1392 | |
| 1393 | netlink_destroy_callback(cb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1394 | return 0; |
Thomas Graf | 1797754 | 2005-06-18 22:53:48 -0700 | [diff] [blame] | 1395 | |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1396 | errout_skb: |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1397 | mutex_unlock(nlk->cb_mutex); |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1398 | kfree_skb(skb); |
| 1399 | errout: |
| 1400 | return err; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1401 | } |
| 1402 | |
| 1403 | int netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
| 1404 | struct nlmsghdr *nlh, |
| 1405 | int (*dump)(struct sk_buff *skb, struct netlink_callback*), |
| 1406 | int (*done)(struct netlink_callback*)) |
| 1407 | { |
| 1408 | struct netlink_callback *cb; |
| 1409 | struct sock *sk; |
| 1410 | struct netlink_sock *nlk; |
| 1411 | |
Panagiotis Issaris | 0da974f | 2006-07-21 14:51:30 -0700 | [diff] [blame] | 1412 | cb = kzalloc(sizeof(*cb), GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1413 | if (cb == NULL) |
| 1414 | return -ENOBUFS; |
| 1415 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1416 | cb->dump = dump; |
| 1417 | cb->done = done; |
| 1418 | cb->nlh = nlh; |
| 1419 | atomic_inc(&skb->users); |
| 1420 | cb->skb = skb; |
| 1421 | |
| 1422 | sk = netlink_lookup(ssk->sk_protocol, NETLINK_CB(skb).pid); |
| 1423 | if (sk == NULL) { |
| 1424 | netlink_destroy_callback(cb); |
| 1425 | return -ECONNREFUSED; |
| 1426 | } |
| 1427 | nlk = nlk_sk(sk); |
Denis Lunev | ac57b3a | 2007-04-18 17:05:58 -0700 | [diff] [blame] | 1428 | /* A dump or destruction is in progress... */ |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1429 | mutex_lock(nlk->cb_mutex); |
Denis Lunev | ac57b3a | 2007-04-18 17:05:58 -0700 | [diff] [blame] | 1430 | if (nlk->cb || sock_flag(sk, SOCK_DEAD)) { |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1431 | mutex_unlock(nlk->cb_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1432 | netlink_destroy_callback(cb); |
| 1433 | sock_put(sk); |
| 1434 | return -EBUSY; |
| 1435 | } |
| 1436 | nlk->cb = cb; |
Patrick McHardy | af65bdf | 2007-04-20 14:14:21 -0700 | [diff] [blame] | 1437 | mutex_unlock(nlk->cb_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1438 | |
| 1439 | netlink_dump(sk); |
| 1440 | sock_put(sk); |
Thomas Graf | c702e80 | 2007-03-22 23:30:55 -0700 | [diff] [blame] | 1441 | |
| 1442 | /* We successfully started a dump, by returning -EINTR we |
| 1443 | * signal the queue mangement to interrupt processing of |
| 1444 | * any netlink messages so userspace gets a chance to read |
| 1445 | * the results. */ |
| 1446 | return -EINTR; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1447 | } |
| 1448 | |
| 1449 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) |
| 1450 | { |
| 1451 | struct sk_buff *skb; |
| 1452 | struct nlmsghdr *rep; |
| 1453 | struct nlmsgerr *errmsg; |
Thomas Graf | 339bf98 | 2006-11-10 14:10:15 -0800 | [diff] [blame] | 1454 | size_t payload = sizeof(*errmsg); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1455 | |
Thomas Graf | 339bf98 | 2006-11-10 14:10:15 -0800 | [diff] [blame] | 1456 | /* error messages get the original request appened */ |
| 1457 | if (err) |
| 1458 | payload += nlmsg_len(nlh); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1459 | |
Thomas Graf | 339bf98 | 2006-11-10 14:10:15 -0800 | [diff] [blame] | 1460 | skb = nlmsg_new(payload, GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1461 | if (!skb) { |
| 1462 | struct sock *sk; |
| 1463 | |
| 1464 | sk = netlink_lookup(in_skb->sk->sk_protocol, |
| 1465 | NETLINK_CB(in_skb).pid); |
| 1466 | if (sk) { |
| 1467 | sk->sk_err = ENOBUFS; |
| 1468 | sk->sk_error_report(sk); |
| 1469 | sock_put(sk); |
| 1470 | } |
| 1471 | return; |
| 1472 | } |
| 1473 | |
| 1474 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, |
Thomas Graf | 1797754 | 2005-06-18 22:53:48 -0700 | [diff] [blame] | 1475 | NLMSG_ERROR, sizeof(struct nlmsgerr), 0); |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1476 | errmsg = nlmsg_data(rep); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1477 | errmsg->error = err; |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1478 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1479 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT); |
| 1480 | } |
| 1481 | |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1482 | static int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
Thomas Graf | 1d00a4e | 2007-03-22 23:30:12 -0700 | [diff] [blame] | 1483 | struct nlmsghdr *)) |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1484 | { |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1485 | struct nlmsghdr *nlh; |
| 1486 | int err; |
| 1487 | |
| 1488 | while (skb->len >= nlmsg_total_size(0)) { |
Arnaldo Carvalho de Melo | b529ccf | 2007-04-25 19:08:35 -0700 | [diff] [blame] | 1489 | nlh = nlmsg_hdr(skb); |
Thomas Graf | d35b685 | 2007-03-22 23:28:46 -0700 | [diff] [blame] | 1490 | err = 0; |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1491 | |
Martin Murray | ad8e4b7 | 2006-01-10 13:02:29 -0800 | [diff] [blame] | 1492 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1493 | return 0; |
| 1494 | |
Thomas Graf | d35b685 | 2007-03-22 23:28:46 -0700 | [diff] [blame] | 1495 | /* Only requests are handled by the kernel */ |
| 1496 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) |
| 1497 | goto skip; |
| 1498 | |
Thomas Graf | 45e7ae7 | 2007-03-22 23:29:10 -0700 | [diff] [blame] | 1499 | /* Skip control messages */ |
| 1500 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) |
| 1501 | goto skip; |
| 1502 | |
Thomas Graf | 1d00a4e | 2007-03-22 23:30:12 -0700 | [diff] [blame] | 1503 | err = cb(skb, nlh); |
| 1504 | if (err == -EINTR) { |
| 1505 | /* Not an error, but we interrupt processing */ |
| 1506 | netlink_queue_skip(nlh, skb); |
| 1507 | return err; |
Thomas Graf | d35b685 | 2007-03-22 23:28:46 -0700 | [diff] [blame] | 1508 | } |
| 1509 | skip: |
| 1510 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1511 | netlink_ack(skb, nlh, err); |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1512 | |
Thomas Graf | bf8b79e | 2006-08-04 23:03:29 -0700 | [diff] [blame] | 1513 | netlink_queue_skip(nlh, skb); |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1514 | } |
| 1515 | |
| 1516 | return 0; |
| 1517 | } |
| 1518 | |
| 1519 | /** |
| 1520 | * nelink_run_queue - Process netlink receive queue. |
| 1521 | * @sk: Netlink socket containing the queue |
| 1522 | * @qlen: Place to store queue length upon entry |
| 1523 | * @cb: Callback function invoked for each netlink message found |
| 1524 | * |
| 1525 | * Processes as much as there was in the queue upon entry and invokes |
| 1526 | * a callback function for each netlink message found. The callback |
| 1527 | * function may refuse a message by returning a negative error code |
| 1528 | * but setting the error pointer to 0 in which case this function |
| 1529 | * returns with a qlen != 0. |
| 1530 | * |
| 1531 | * qlen must be initialized to 0 before the initial entry, afterwards |
| 1532 | * the function may be called repeatedly until qlen reaches 0. |
Thomas Graf | 1d00a4e | 2007-03-22 23:30:12 -0700 | [diff] [blame] | 1533 | * |
| 1534 | * The callback function may return -EINTR to signal that processing |
| 1535 | * of netlink messages shall be interrupted. In this case the message |
| 1536 | * currently being processed will NOT be requeued onto the receive |
| 1537 | * queue. |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1538 | */ |
| 1539 | void netlink_run_queue(struct sock *sk, unsigned int *qlen, |
Thomas Graf | 1d00a4e | 2007-03-22 23:30:12 -0700 | [diff] [blame] | 1540 | int (*cb)(struct sk_buff *, struct nlmsghdr *)) |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1541 | { |
| 1542 | struct sk_buff *skb; |
| 1543 | |
| 1544 | if (!*qlen || *qlen > skb_queue_len(&sk->sk_receive_queue)) |
| 1545 | *qlen = skb_queue_len(&sk->sk_receive_queue); |
| 1546 | |
| 1547 | for (; *qlen; (*qlen)--) { |
| 1548 | skb = skb_dequeue(&sk->sk_receive_queue); |
| 1549 | if (netlink_rcv_skb(skb, cb)) { |
| 1550 | if (skb->len) |
| 1551 | skb_queue_head(&sk->sk_receive_queue, skb); |
| 1552 | else { |
| 1553 | kfree_skb(skb); |
| 1554 | (*qlen)--; |
| 1555 | } |
| 1556 | break; |
| 1557 | } |
| 1558 | |
| 1559 | kfree_skb(skb); |
| 1560 | } |
| 1561 | } |
| 1562 | |
| 1563 | /** |
| 1564 | * netlink_queue_skip - Skip netlink message while processing queue. |
| 1565 | * @nlh: Netlink message to be skipped |
| 1566 | * @skb: Socket buffer containing the netlink messages. |
| 1567 | * |
| 1568 | * Pulls the given netlink message off the socket buffer so the next |
| 1569 | * call to netlink_queue_run() will not reconsider the message. |
| 1570 | */ |
| 1571 | void netlink_queue_skip(struct nlmsghdr *nlh, struct sk_buff *skb) |
| 1572 | { |
| 1573 | int msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
| 1574 | |
| 1575 | if (msglen > skb->len) |
| 1576 | msglen = skb->len; |
| 1577 | |
| 1578 | skb_pull(skb, msglen); |
| 1579 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1580 | |
Thomas Graf | d387f6a | 2006-08-15 00:31:06 -0700 | [diff] [blame] | 1581 | /** |
| 1582 | * nlmsg_notify - send a notification netlink message |
| 1583 | * @sk: netlink socket to use |
| 1584 | * @skb: notification message |
| 1585 | * @pid: destination netlink pid for reports or 0 |
| 1586 | * @group: destination multicast group or 0 |
| 1587 | * @report: 1 to report back, 0 to disable |
| 1588 | * @flags: allocation flags |
| 1589 | */ |
| 1590 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 pid, |
| 1591 | unsigned int group, int report, gfp_t flags) |
| 1592 | { |
| 1593 | int err = 0; |
| 1594 | |
| 1595 | if (group) { |
| 1596 | int exclude_pid = 0; |
| 1597 | |
| 1598 | if (report) { |
| 1599 | atomic_inc(&skb->users); |
| 1600 | exclude_pid = pid; |
| 1601 | } |
| 1602 | |
| 1603 | /* errors reported via destination sk->sk_err */ |
| 1604 | nlmsg_multicast(sk, skb, exclude_pid, group, flags); |
| 1605 | } |
| 1606 | |
| 1607 | if (report) |
| 1608 | err = nlmsg_unicast(sk, skb, pid); |
| 1609 | |
| 1610 | return err; |
| 1611 | } |
| 1612 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1613 | #ifdef CONFIG_PROC_FS |
| 1614 | struct nl_seq_iter { |
| 1615 | int link; |
| 1616 | int hash_idx; |
| 1617 | }; |
| 1618 | |
| 1619 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) |
| 1620 | { |
| 1621 | struct nl_seq_iter *iter = seq->private; |
| 1622 | int i, j; |
| 1623 | struct sock *s; |
| 1624 | struct hlist_node *node; |
| 1625 | loff_t off = 0; |
| 1626 | |
| 1627 | for (i=0; i<MAX_LINKS; i++) { |
| 1628 | struct nl_pid_hash *hash = &nl_table[i].hash; |
| 1629 | |
| 1630 | for (j = 0; j <= hash->mask; j++) { |
| 1631 | sk_for_each(s, node, &hash->table[j]) { |
| 1632 | if (off == pos) { |
| 1633 | iter->link = i; |
| 1634 | iter->hash_idx = j; |
| 1635 | return s; |
| 1636 | } |
| 1637 | ++off; |
| 1638 | } |
| 1639 | } |
| 1640 | } |
| 1641 | return NULL; |
| 1642 | } |
| 1643 | |
| 1644 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) |
| 1645 | { |
| 1646 | read_lock(&nl_table_lock); |
| 1647 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
| 1648 | } |
| 1649 | |
| 1650 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) |
| 1651 | { |
| 1652 | struct sock *s; |
| 1653 | struct nl_seq_iter *iter; |
| 1654 | int i, j; |
| 1655 | |
| 1656 | ++*pos; |
| 1657 | |
| 1658 | if (v == SEQ_START_TOKEN) |
| 1659 | return netlink_seq_socket_idx(seq, 0); |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1660 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1661 | s = sk_next(v); |
| 1662 | if (s) |
| 1663 | return s; |
| 1664 | |
| 1665 | iter = seq->private; |
| 1666 | i = iter->link; |
| 1667 | j = iter->hash_idx + 1; |
| 1668 | |
| 1669 | do { |
| 1670 | struct nl_pid_hash *hash = &nl_table[i].hash; |
| 1671 | |
| 1672 | for (; j <= hash->mask; j++) { |
| 1673 | s = sk_head(&hash->table[j]); |
| 1674 | if (s) { |
| 1675 | iter->link = i; |
| 1676 | iter->hash_idx = j; |
| 1677 | return s; |
| 1678 | } |
| 1679 | } |
| 1680 | |
| 1681 | j = 0; |
| 1682 | } while (++i < MAX_LINKS); |
| 1683 | |
| 1684 | return NULL; |
| 1685 | } |
| 1686 | |
| 1687 | static void netlink_seq_stop(struct seq_file *seq, void *v) |
| 1688 | { |
| 1689 | read_unlock(&nl_table_lock); |
| 1690 | } |
| 1691 | |
| 1692 | |
| 1693 | static int netlink_seq_show(struct seq_file *seq, void *v) |
| 1694 | { |
| 1695 | if (v == SEQ_START_TOKEN) |
| 1696 | seq_puts(seq, |
| 1697 | "sk Eth Pid Groups " |
| 1698 | "Rmem Wmem Dump Locks\n"); |
| 1699 | else { |
| 1700 | struct sock *s = v; |
| 1701 | struct netlink_sock *nlk = nlk_sk(s); |
| 1702 | |
| 1703 | seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n", |
| 1704 | s, |
| 1705 | s->sk_protocol, |
| 1706 | nlk->pid, |
Patrick McHardy | 513c250 | 2005-09-06 15:43:59 -0700 | [diff] [blame] | 1707 | nlk->groups ? (u32)nlk->groups[0] : 0, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1708 | atomic_read(&s->sk_rmem_alloc), |
| 1709 | atomic_read(&s->sk_wmem_alloc), |
| 1710 | nlk->cb, |
| 1711 | atomic_read(&s->sk_refcnt) |
| 1712 | ); |
| 1713 | |
| 1714 | } |
| 1715 | return 0; |
| 1716 | } |
| 1717 | |
| 1718 | static struct seq_operations netlink_seq_ops = { |
| 1719 | .start = netlink_seq_start, |
| 1720 | .next = netlink_seq_next, |
| 1721 | .stop = netlink_seq_stop, |
| 1722 | .show = netlink_seq_show, |
| 1723 | }; |
| 1724 | |
| 1725 | |
| 1726 | static int netlink_seq_open(struct inode *inode, struct file *file) |
| 1727 | { |
| 1728 | struct seq_file *seq; |
| 1729 | struct nl_seq_iter *iter; |
| 1730 | int err; |
| 1731 | |
Panagiotis Issaris | 0da974f | 2006-07-21 14:51:30 -0700 | [diff] [blame] | 1732 | iter = kzalloc(sizeof(*iter), GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1733 | if (!iter) |
| 1734 | return -ENOMEM; |
| 1735 | |
| 1736 | err = seq_open(file, &netlink_seq_ops); |
| 1737 | if (err) { |
| 1738 | kfree(iter); |
| 1739 | return err; |
| 1740 | } |
| 1741 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1742 | seq = file->private_data; |
| 1743 | seq->private = iter; |
| 1744 | return 0; |
| 1745 | } |
| 1746 | |
Arjan van de Ven | da7071d | 2007-02-12 00:55:36 -0800 | [diff] [blame] | 1747 | static const struct file_operations netlink_seq_fops = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1748 | .owner = THIS_MODULE, |
| 1749 | .open = netlink_seq_open, |
| 1750 | .read = seq_read, |
| 1751 | .llseek = seq_lseek, |
| 1752 | .release = seq_release_private, |
| 1753 | }; |
| 1754 | |
| 1755 | #endif |
| 1756 | |
| 1757 | int netlink_register_notifier(struct notifier_block *nb) |
| 1758 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 1759 | return atomic_notifier_chain_register(&netlink_chain, nb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1760 | } |
| 1761 | |
| 1762 | int netlink_unregister_notifier(struct notifier_block *nb) |
| 1763 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 1764 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1765 | } |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1766 | |
Eric Dumazet | 90ddc4f | 2005-12-22 12:49:22 -0800 | [diff] [blame] | 1767 | static const struct proto_ops netlink_ops = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1768 | .family = PF_NETLINK, |
| 1769 | .owner = THIS_MODULE, |
| 1770 | .release = netlink_release, |
| 1771 | .bind = netlink_bind, |
| 1772 | .connect = netlink_connect, |
| 1773 | .socketpair = sock_no_socketpair, |
| 1774 | .accept = sock_no_accept, |
| 1775 | .getname = netlink_getname, |
| 1776 | .poll = datagram_poll, |
| 1777 | .ioctl = sock_no_ioctl, |
| 1778 | .listen = sock_no_listen, |
| 1779 | .shutdown = sock_no_shutdown, |
Patrick McHardy | 9a4595b | 2005-08-15 12:32:15 -0700 | [diff] [blame] | 1780 | .setsockopt = netlink_setsockopt, |
| 1781 | .getsockopt = netlink_getsockopt, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1782 | .sendmsg = netlink_sendmsg, |
| 1783 | .recvmsg = netlink_recvmsg, |
| 1784 | .mmap = sock_no_mmap, |
| 1785 | .sendpage = sock_no_sendpage, |
| 1786 | }; |
| 1787 | |
| 1788 | static struct net_proto_family netlink_family_ops = { |
| 1789 | .family = PF_NETLINK, |
| 1790 | .create = netlink_create, |
| 1791 | .owner = THIS_MODULE, /* for consistency 8) */ |
| 1792 | }; |
| 1793 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1794 | static int __init netlink_proto_init(void) |
| 1795 | { |
| 1796 | struct sk_buff *dummy_skb; |
| 1797 | int i; |
| 1798 | unsigned long max; |
| 1799 | unsigned int order; |
| 1800 | int err = proto_register(&netlink_proto, 0); |
| 1801 | |
| 1802 | if (err != 0) |
| 1803 | goto out; |
| 1804 | |
YOSHIFUJI Hideaki | ef047f5 | 2006-09-01 00:29:06 -0700 | [diff] [blame] | 1805 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1806 | |
Panagiotis Issaris | 0da974f | 2006-07-21 14:51:30 -0700 | [diff] [blame] | 1807 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
Akinobu Mita | fab2caf | 2006-08-29 02:15:24 -0700 | [diff] [blame] | 1808 | if (!nl_table) |
| 1809 | goto panic; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1810 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1811 | if (num_physpages >= (128 * 1024)) |
| 1812 | max = num_physpages >> (21 - PAGE_SHIFT); |
| 1813 | else |
| 1814 | max = num_physpages >> (23 - PAGE_SHIFT); |
| 1815 | |
| 1816 | order = get_bitmask_order(max) - 1 + PAGE_SHIFT; |
| 1817 | max = (1UL << order) / sizeof(struct hlist_head); |
| 1818 | order = get_bitmask_order(max > UINT_MAX ? UINT_MAX : max) - 1; |
| 1819 | |
| 1820 | for (i = 0; i < MAX_LINKS; i++) { |
| 1821 | struct nl_pid_hash *hash = &nl_table[i].hash; |
| 1822 | |
| 1823 | hash->table = nl_pid_hash_alloc(1 * sizeof(*hash->table)); |
| 1824 | if (!hash->table) { |
| 1825 | while (i-- > 0) |
| 1826 | nl_pid_hash_free(nl_table[i].hash.table, |
| 1827 | 1 * sizeof(*hash->table)); |
| 1828 | kfree(nl_table); |
Akinobu Mita | fab2caf | 2006-08-29 02:15:24 -0700 | [diff] [blame] | 1829 | goto panic; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1830 | } |
| 1831 | memset(hash->table, 0, 1 * sizeof(*hash->table)); |
| 1832 | hash->max_shift = order; |
| 1833 | hash->shift = 0; |
| 1834 | hash->mask = 0; |
| 1835 | hash->rehash_time = jiffies; |
| 1836 | } |
| 1837 | |
| 1838 | sock_register(&netlink_family_ops); |
| 1839 | #ifdef CONFIG_PROC_FS |
| 1840 | proc_net_fops_create("netlink", 0, &netlink_seq_fops); |
| 1841 | #endif |
YOSHIFUJI Hideaki | 746fac4 | 2007-02-09 23:25:07 +0900 | [diff] [blame] | 1842 | /* The netlink device handler may be needed early. */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1843 | rtnetlink_init(); |
| 1844 | out: |
| 1845 | return err; |
Akinobu Mita | fab2caf | 2006-08-29 02:15:24 -0700 | [diff] [blame] | 1846 | panic: |
| 1847 | panic("netlink_init: Cannot allocate nl_table\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1848 | } |
| 1849 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1850 | core_initcall(netlink_proto_init); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1851 | |
| 1852 | EXPORT_SYMBOL(netlink_ack); |
Thomas Graf | 82ace47 | 2005-11-10 02:25:53 +0100 | [diff] [blame] | 1853 | EXPORT_SYMBOL(netlink_run_queue); |
| 1854 | EXPORT_SYMBOL(netlink_queue_skip); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1855 | EXPORT_SYMBOL(netlink_broadcast); |
| 1856 | EXPORT_SYMBOL(netlink_dump_start); |
| 1857 | EXPORT_SYMBOL(netlink_kernel_create); |
| 1858 | EXPORT_SYMBOL(netlink_register_notifier); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1859 | EXPORT_SYMBOL(netlink_set_nonroot); |
| 1860 | EXPORT_SYMBOL(netlink_unicast); |
| 1861 | EXPORT_SYMBOL(netlink_unregister_notifier); |
Thomas Graf | d387f6a | 2006-08-15 00:31:06 -0700 | [diff] [blame] | 1862 | EXPORT_SYMBOL(nlmsg_notify); |