net/netfilter/nfnetlink.c

   1 /* Netfilter messages via netlink socket. Allows for user space
   2  * protocol helpers and general trouble making from userspace.
   3  *
   4  * (C) 2001 by Jay Schulist <jschlst@samba.org>,
   5  * (C) 2002-2005 by Harald Welte <laforge@gnumonks.org>
   6  * (C) 2005 by Pablo Neira Ayuso <pablo@eurodev.net>
   7  *
   8  * Initial netfilter messages via netlink development funded and
   9  * generally made possible by Network Robots, Inc. (www.networkrobots.com)
  10  *
  11  * Further development of this code funded by Astaro AG (http://www.astaro.com)
  12  *
  13  * This software may be used and distributed according to the terms
  14  * of the GNU General Public License, incorporated herein by reference.
  15  */
  16
  17 #include <linux/config.h>
  18 #include <linux/module.h>
  19 #include <linux/types.h>
  20 #include <linux/socket.h>
  21 #include <linux/kernel.h>
  22 #include <linux/major.h>
  23 #include <linux/sched.h>
  24 #include <linux/timer.h>
  25 #include <linux/string.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/fcntl.h>
  29 #include <linux/skbuff.h>
  30 #include <asm/uaccess.h>
  31 #include <asm/system.h>
  32 #include <net/sock.h>
  33 #include <linux/init.h>
  34 #include <linux/spinlock.h>
  35
  36 #include <linux/netfilter.h>
  37 #include <linux/netlink.h>
  38 #include <linux/netfilter/nfnetlink.h>
  39
  40 MODULE_LICENSE("GPL");
  41 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
  42 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_NETFILTER);
  43
  44 static char __initdata nfversion[] = "0.30";
  45
  46 #if 0
  47 #define DEBUGP(format, args...) \
  48                 printk(KERN_DEBUG "%s(%d):%s(): " format, __FILE__, \
  49                         __LINE__, __FUNCTION__, ## args)
  50 #else
  51 #define DEBUGP(format, args...)
  52 #endif
  53
  54 static struct sock *nfnl = NULL;
  55 static struct nfnetlink_subsystem *subsys_table[NFNL_SUBSYS_COUNT];
  56 DECLARE_MUTEX(nfnl_sem);
  57
  58 void nfnl_lock(void)
  59 {
  60         nfnl_shlock();
  61 }
  62
  63 void nfnl_unlock(void)
  64 {
  65         nfnl_shunlock();
  66 }
  67
  68 int nfnetlink_subsys_register(struct nfnetlink_subsystem *n)
  69 {
  70         DEBUGP("registering subsystem ID %u\n", n->subsys_id);
  71
  72         nfnl_lock();
  73         if (subsys_table[n->subsys_id]) {
  74                 nfnl_unlock();
  75                 return -EBUSY;
  76         }
  77         subsys_table[n->subsys_id] = n;
  78         nfnl_unlock();
  79
  80         return 0;
  81 }
  82
  83 int nfnetlink_subsys_unregister(struct nfnetlink_subsystem *n)
  84 {
  85         DEBUGP("unregistering subsystem ID %u\n", n->subsys_id);
  86
  87         nfnl_lock();
  88         subsys_table[n->subsys_id] = NULL;
  89         nfnl_unlock();
  90
  91         return 0;
  92 }
  93
  94 static inline struct nfnetlink_subsystem *nfnetlink_get_subsys(u_int16_t type)
  95 {
  96         u_int8_t subsys_id = NFNL_SUBSYS_ID(type);
  97
  98         if (subsys_id >= NFNL_SUBSYS_COUNT
  99             || subsys_table[subsys_id] == NULL)
 100                 return NULL;
 101
 102         return subsys_table[subsys_id];
 103 }
 104
 105 static inline struct nfnl_callback *
 106 nfnetlink_find_client(u_int16_t type, struct nfnetlink_subsystem *ss)
 107 {
 108         u_int8_t cb_id = NFNL_MSG_TYPE(type);
 109
 110         if (cb_id >= ss->cb_count) {
 111                 DEBUGP("msgtype %u >= %u, returning\n", type, ss->cb_count);
 112                 return NULL;
 113         }
 114
 115         return &ss->cb[cb_id];
 116 }
 117
 118 void __nfa_fill(struct sk_buff *skb, int attrtype, int attrlen,
 119                 const void *data)
 120 {
 121         struct nfattr *nfa;
 122         int size = NFA_LENGTH(attrlen);
 123
 124         nfa = (struct nfattr *)skb_put(skb, NFA_ALIGN(size));
 125         nfa->nfa_type = attrtype;
 126         nfa->nfa_len  = size;
 127         memcpy(NFA_DATA(nfa), data, attrlen);
 128         memset(NFA_DATA(nfa) + attrlen, 0, NFA_ALIGN(size) - size);
 129 }
 130
 131 void nfattr_parse(struct nfattr *tb[], int maxattr, struct nfattr *nfa, int len)
 132 {
 133         memset(tb, 0, sizeof(struct nfattr *) * maxattr);
 134
 135         while (NFA_OK(nfa, len)) {
 136                 unsigned flavor = NFA_TYPE(nfa);
 137                 if (flavor && flavor <= maxattr)
 138                         tb[flavor-1] = nfa;
 139                 nfa = NFA_NEXT(nfa, len);
 140         }
 141 }
 142
 143 /**
 144  * nfnetlink_check_attributes - check and parse nfnetlink attributes
 145  *
 146  * subsys: nfnl subsystem for which this message is to be parsed
 147  * nlmsghdr: netlink message to be checked/parsed
 148  * cda: array of pointers, needs to be at least subsys->attr_count big
 149  *
 150  */
 151 static int
 152 nfnetlink_check_attributes(struct nfnetlink_subsystem *subsys,
 153                            struct nlmsghdr *nlh, struct nfattr *cda[])
 154 {
 155         int min_len;
 156         u_int16_t attr_count;
 157         u_int8_t cb_id = NFNL_MSG_TYPE(nlh->nlmsg_type);
 158
 159         if (unlikely(cb_id >= subsys->cb_count)) {
 160                 DEBUGP("msgtype %u >= %u, returning\n",
 161                         cb_id, subsys->cb_count);
 162                 return -EINVAL;
 163         }
 164
 165         min_len = NLMSG_ALIGN(sizeof(struct nfgenmsg));
 166         if (unlikely(nlh->nlmsg_len < min_len))
 167                 return -EINVAL;
 168
 169         attr_count = subsys->cb[cb_id].attr_count;
 170         memset(cda, 0, sizeof(struct nfattr *) * attr_count);
 171
 172         /* check attribute lengths. */
 173         if (likely(nlh->nlmsg_len > min_len)) {
 174                 struct nfattr *attr = NFM_NFA(NLMSG_DATA(nlh));
 175                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
 176
 177                 while (NFA_OK(attr, attrlen)) {
 178                         unsigned flavor = NFA_TYPE(attr);
 179                         if (flavor) {
 180                                 if (flavor > attr_count)
 181                                         return -EINVAL;
 182                                 cda[flavor - 1] = attr;
 183                         }
 184                         attr = NFA_NEXT(attr, attrlen);
 185                 }
 186         }
 187
 188         /* implicit: if nlmsg_len == min_len, we return 0, and an empty
 189          * (zeroed) cda[] array. The message is valid, but empty. */
 190
 191         return 0;
 192 }
 193
 194 int nfnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
 195 {
 196         gfp_t allocation = in_interrupt() ? GFP_ATOMIC : GFP_KERNEL;
 197         int err = 0;
 198
 199         NETLINK_CB(skb).dst_group = group;
 200         if (echo)
 201                 atomic_inc(&skb->users);
 202         netlink_broadcast(nfnl, skb, pid, group, allocation);
 203         if (echo)
 204                 err = netlink_unicast(nfnl, skb, pid, MSG_DONTWAIT);
 205
 206         return err;
 207 }
 208
 209 int nfnetlink_unicast(struct sk_buff *skb, u_int32_t pid, int flags)
 210 {
 211         return netlink_unicast(nfnl, skb, pid, flags);
 212 }
 213
 214 /* Process one complete nfnetlink message. */
 215 static inline int nfnetlink_rcv_msg(struct sk_buff *skb,
 216                                     struct nlmsghdr *nlh, int *errp)
 217 {
 218         struct nfnl_callback *nc;
 219         struct nfnetlink_subsystem *ss;
 220         int type, err = 0;
 221
 222         DEBUGP("entered; subsys=%u, msgtype=%u\n",
 223                  NFNL_SUBSYS_ID(nlh->nlmsg_type),
 224                  NFNL_MSG_TYPE(nlh->nlmsg_type));
 225
 226         /* Only requests are handled by kernel now. */
 227         if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) {
 228                 DEBUGP("received non-request message\n");
 229                 return 0;
 230         }
 231
 232         /* All the messages must at least contain nfgenmsg */
 233         if (nlh->nlmsg_len <
 234                         NLMSG_LENGTH(NLMSG_ALIGN(sizeof(struct nfgenmsg)))) {
 235                 DEBUGP("received message was too short\n");
 236                 return 0;
 237         }
 238
 239         type = nlh->nlmsg_type;
 240         ss = nfnetlink_get_subsys(type);
 241         if (!ss) {
 242 #ifdef CONFIG_KMOD
 243                 if (cap_raised(NETLINK_CB(skb).eff_cap, CAP_NET_ADMIN)) {
 244                         /* don't call nfnl_shunlock, since it would reenter
 245                          * with further packet processing */
 246                         up(&nfnl_sem);
 247                         request_module("nfnetlink-subsys-%d",
 248                                         NFNL_SUBSYS_ID(type));
 249                         nfnl_shlock();
 250                         ss = nfnetlink_get_subsys(type);
 251                 }
 252                 if (!ss)
 253 #endif
 254                         goto err_inval;
 255         }
 256
 257         nc = nfnetlink_find_client(type, ss);
 258         if (!nc) {
 259                 DEBUGP("unable to find client for type %d\n", type);
 260                 goto err_inval;
 261         }
 262
 263         if (nc->cap_required &&
 264             !cap_raised(NETLINK_CB(skb).eff_cap, nc->cap_required)) {
 265                 DEBUGP("permission denied for type %d\n", type);
 266                 *errp = -EPERM;
 267                 return -1;
 268         }
 269
 270         {
 271                 u_int16_t attr_count =
 272                         ss->cb[NFNL_MSG_TYPE(nlh->nlmsg_type)].attr_count;
 273                 struct nfattr *cda[attr_count];
 274
 275                 memset(cda, 0, sizeof(struct nfattr *) * attr_count);
 276
 277                 err = nfnetlink_check_attributes(ss, nlh, cda);
 278                 if (err < 0)
 279                         goto err_inval;
 280
 281                 DEBUGP("calling handler\n");
 282                 err = nc->call(nfnl, skb, nlh, cda, errp);
 283                 *errp = err;
 284                 return err;
 285         }
 286
 287 err_inval:
 288         DEBUGP("returning -EINVAL\n");
 289         *errp = -EINVAL;
 290         return -1;
 291 }
 292
 293 /* Process one packet of messages. */
 294 static inline int nfnetlink_rcv_skb(struct sk_buff *skb)
 295 {
 296         int err;
 297         struct nlmsghdr *nlh;
 298
 299         while (skb->len >= NLMSG_SPACE(0)) {
 300                 u32 rlen;
 301
 302                 nlh = (struct nlmsghdr *)skb->data;
 303                 if (nlh->nlmsg_len < sizeof(struct nlmsghdr)
 304                     || skb->len < nlh->nlmsg_len)
 305                         return 0;
 306                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 307                 if (rlen > skb->len)
 308                         rlen = skb->len;
 309                 if (nfnetlink_rcv_msg(skb, nlh, &err)) {
 310                         if (!err)
 311                                 return -1;
 312                         netlink_ack(skb, nlh, err);
 313                 } else
 314                         if (nlh->nlmsg_flags & NLM_F_ACK)
 315                                 netlink_ack(skb, nlh, 0);
 316                 skb_pull(skb, rlen);
 317         }
 318
 319         return 0;
 320 }
 321
 322 static void nfnetlink_rcv(struct sock *sk, int len)
 323 {
 324         do {
 325                 struct sk_buff *skb;
 326
 327                 if (nfnl_shlock_nowait())
 328                         return;
 329
 330                 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
 331                         if (nfnetlink_rcv_skb(skb)) {
 332                                 if (skb->len)
 333                                         skb_queue_head(&sk->sk_receive_queue,
 334                                                        skb);
 335                                 else
 336                                         kfree_skb(skb);
 337                                 break;
 338                         }
 339                         kfree_skb(skb);
 340                 }
 341
 342                 /* don't call nfnl_shunlock, since it would reenter
 343                  * with further packet processing */
 344                 up(&nfnl_sem);
 345         } while(nfnl && nfnl->sk_receive_queue.qlen);
 346 }
 347
 348 static void __exit nfnetlink_exit(void)
 349 {
 350         printk("Removing netfilter NETLINK layer.\n");
 351         sock_release(nfnl->sk_socket);
 352         return;
 353 }
 354
 355 static int __init nfnetlink_init(void)
 356 {
 357         printk("Netfilter messages via NETLINK v%s.\n", nfversion);
 358
 359         nfnl = netlink_kernel_create(NETLINK_NETFILTER, NFNLGRP_MAX,
 360                                      nfnetlink_rcv, THIS_MODULE);
 361         if (!nfnl) {
 362                 printk(KERN_ERR "cannot initialize nfnetlink!\n");
 363                 return -1;
 364         }
 365
 366         return 0;
 367 }
 368
 369 module_init(nfnetlink_init);
 370 module_exit(nfnetlink_exit);
 371
 372 EXPORT_SYMBOL_GPL(nfnetlink_subsys_register);
 373 EXPORT_SYMBOL_GPL(nfnetlink_subsys_unregister);
 374 EXPORT_SYMBOL_GPL(nfnetlink_send);
 375 EXPORT_SYMBOL_GPL(nfnetlink_unicast);
 376 EXPORT_SYMBOL_GPL(nfattr_parse);
 377 EXPORT_SYMBOL_GPL(__nfa_fill);