net/ipv4/ipmr_base.c

   1 /* Linux multicast routing support
   2  * Common logic shared by IPv4 [ipmr] and IPv6 [ip6mr] implementation
   3  */
   4
   5 #include <linux/rhashtable.h>
   6 #include <linux/mroute_base.h>
   7
   8 /* Sets everything common except 'dev', since that is done under locking */
   9 void vif_device_init(struct vif_device *v,
  10                      struct net_device *dev,
  11                      unsigned long rate_limit,
  12                      unsigned char threshold,
  13                      unsigned short flags,
  14                      unsigned short get_iflink_mask)
  15 {
  16         v->dev = NULL;
  17         v->bytes_in = 0;
  18         v->bytes_out = 0;
  19         v->pkt_in = 0;
  20         v->pkt_out = 0;
  21         v->rate_limit = rate_limit;
  22         v->flags = flags;
  23         v->threshold = threshold;
  24         if (v->flags & get_iflink_mask)
  25                 v->link = dev_get_iflink(dev);
  26         else
  27                 v->link = dev->ifindex;
  28 }
  29 EXPORT_SYMBOL(vif_device_init);
  30
  31 struct mr_table *
  32 mr_table_alloc(struct net *net, u32 id,
  33                struct mr_table_ops *ops,
  34                void (*expire_func)(struct timer_list *t),
  35                void (*table_set)(struct mr_table *mrt,
  36                                  struct net *net))
  37 {
  38         struct mr_table *mrt;
  39         int err;
  40
  41         mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
  42         if (!mrt)
  43                 return ERR_PTR(-ENOMEM);
  44         mrt->id = id;
  45         write_pnet(&mrt->net, net);
  46
  47         mrt->ops = *ops;
  48         err = rhltable_init(&mrt->mfc_hash, mrt->ops.rht_params);
  49         if (err) {
  50                 kfree(mrt);
  51                 return ERR_PTR(err);
  52         }
  53         INIT_LIST_HEAD(&mrt->mfc_cache_list);
  54         INIT_LIST_HEAD(&mrt->mfc_unres_queue);
  55
  56         timer_setup(&mrt->ipmr_expire_timer, expire_func, 0);
  57
  58         mrt->mroute_reg_vif_num = -1;
  59         table_set(mrt, net);
  60         return mrt;
  61 }
  62 EXPORT_SYMBOL(mr_table_alloc);
  63
  64 void *mr_mfc_find_parent(struct mr_table *mrt, void *hasharg, int parent)
  65 {
  66         struct rhlist_head *tmp, *list;
  67         struct mr_mfc *c;
  68
  69         list = rhltable_lookup(&mrt->mfc_hash, hasharg, *mrt->ops.rht_params);
  70         rhl_for_each_entry_rcu(c, tmp, list, mnode)
  71                 if (parent == -1 || parent == c->mfc_parent)
  72                         return c;
  73
  74         return NULL;
  75 }
  76 EXPORT_SYMBOL(mr_mfc_find_parent);
  77
  78 void *mr_mfc_find_any_parent(struct mr_table *mrt, int vifi)
  79 {
  80         struct rhlist_head *tmp, *list;
  81         struct mr_mfc *c;
  82
  83         list = rhltable_lookup(&mrt->mfc_hash, mrt->ops.cmparg_any,
  84                                *mrt->ops.rht_params);
  85         rhl_for_each_entry_rcu(c, tmp, list, mnode)
  86                 if (c->mfc_un.res.ttls[vifi] < 255)
  87                         return c;
  88
  89         return NULL;
  90 }
  91 EXPORT_SYMBOL(mr_mfc_find_any_parent);
  92
  93 void *mr_mfc_find_any(struct mr_table *mrt, int vifi, void *hasharg)
  94 {
  95         struct rhlist_head *tmp, *list;
  96         struct mr_mfc *c, *proxy;
  97
  98         list = rhltable_lookup(&mrt->mfc_hash, hasharg, *mrt->ops.rht_params);
  99         rhl_for_each_entry_rcu(c, tmp, list, mnode) {
 100                 if (c->mfc_un.res.ttls[vifi] < 255)
 101                         return c;
 102
 103                 /* It's ok if the vifi is part of the static tree */
 104                 proxy = mr_mfc_find_any_parent(mrt, c->mfc_parent);
 105                 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
 106                         return c;
 107         }
 108
 109         return mr_mfc_find_any_parent(mrt, vifi);
 110 }
 111 EXPORT_SYMBOL(mr_mfc_find_any);
 112
 113 #ifdef CONFIG_PROC_FS
 114 void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter, loff_t pos)
 115 {
 116         struct mr_table *mrt = iter->mrt;
 117
 118         for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
 119                 if (!VIF_EXISTS(mrt, iter->ct))
 120                         continue;
 121                 if (pos-- == 0)
 122                         return &mrt->vif_table[iter->ct];
 123         }
 124         return NULL;
 125 }
 126 EXPORT_SYMBOL(mr_vif_seq_idx);
 127
 128 void *mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 129 {
 130         struct mr_vif_iter *iter = seq->private;
 131         struct net *net = seq_file_net(seq);
 132         struct mr_table *mrt = iter->mrt;
 133
 134         ++*pos;
 135         if (v == SEQ_START_TOKEN)
 136                 return mr_vif_seq_idx(net, iter, 0);
 137
 138         while (++iter->ct < mrt->maxvif) {
 139                 if (!VIF_EXISTS(mrt, iter->ct))
 140                         continue;
 141                 return &mrt->vif_table[iter->ct];
 142         }
 143         return NULL;
 144 }
 145 EXPORT_SYMBOL(mr_vif_seq_next);
 146
 147 void *mr_mfc_seq_idx(struct net *net,
 148                      struct mr_mfc_iter *it, loff_t pos)
 149 {
 150         struct mr_table *mrt = it->mrt;
 151         struct mr_mfc *mfc;
 152
 153         rcu_read_lock();
 154         it->cache = &mrt->mfc_cache_list;
 155         list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
 156                 if (pos-- == 0)
 157                         return mfc;
 158         rcu_read_unlock();
 159
 160         spin_lock_bh(it->lock);
 161         it->cache = &mrt->mfc_unres_queue;
 162         list_for_each_entry(mfc, it->cache, list)
 163                 if (pos-- == 0)
 164                         return mfc;
 165         spin_unlock_bh(it->lock);
 166
 167         it->cache = NULL;
 168         return NULL;
 169 }
 170 EXPORT_SYMBOL(mr_mfc_seq_idx);
 171
 172 void *mr_mfc_seq_next(struct seq_file *seq, void *v,
 173                       loff_t *pos)
 174 {
 175         struct mr_mfc_iter *it = seq->private;
 176         struct net *net = seq_file_net(seq);
 177         struct mr_table *mrt = it->mrt;
 178         struct mr_mfc *c = v;
 179
 180         ++*pos;
 181
 182         if (v == SEQ_START_TOKEN)
 183                 return mr_mfc_seq_idx(net, seq->private, 0);
 184
 185         if (c->list.next != it->cache)
 186                 return list_entry(c->list.next, struct mr_mfc, list);
 187
 188         if (it->cache == &mrt->mfc_unres_queue)
 189                 goto end_of_list;
 190
 191         /* exhausted cache_array, show unresolved */
 192         rcu_read_unlock();
 193         it->cache = &mrt->mfc_unres_queue;
 194
 195         spin_lock_bh(it->lock);
 196         if (!list_empty(it->cache))
 197                 return list_first_entry(it->cache, struct mr_mfc, list);
 198
 199 end_of_list:
 200         spin_unlock_bh(it->lock);
 201         it->cache = NULL;
 202
 203         return NULL;
 204 }
 205 EXPORT_SYMBOL(mr_mfc_seq_next);
 206 #endif
 207
 208 int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
 209                    struct mr_mfc *c, struct rtmsg *rtm)
 210 {
 211         struct rta_mfc_stats mfcs;
 212         struct nlattr *mp_attr;
 213         struct rtnexthop *nhp;
 214         unsigned long lastuse;
 215         int ct;
 216
 217         /* If cache is unresolved, don't try to parse IIF and OIF */
 218         if (c->mfc_parent >= MAXVIFS) {
 219                 rtm->rtm_flags |= RTNH_F_UNRESOLVED;
 220                 return -ENOENT;
 221         }
 222
 223         if (VIF_EXISTS(mrt, c->mfc_parent) &&
 224             nla_put_u32(skb, RTA_IIF,
 225                         mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
 226                 return -EMSGSIZE;
 227
 228         if (c->mfc_flags & MFC_OFFLOAD)
 229                 rtm->rtm_flags |= RTNH_F_OFFLOAD;
 230
 231         mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
 232         if (!mp_attr)
 233                 return -EMSGSIZE;
 234
 235         for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
 236                 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
 237                         struct vif_device *vif;
 238
 239                         nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
 240                         if (!nhp) {
 241                                 nla_nest_cancel(skb, mp_attr);
 242                                 return -EMSGSIZE;
 243                         }
 244
 245                         nhp->rtnh_flags = 0;
 246                         nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
 247                         vif = &mrt->vif_table[ct];
 248                         nhp->rtnh_ifindex = vif->dev->ifindex;
 249                         nhp->rtnh_len = sizeof(*nhp);
 250                 }
 251         }
 252
 253         nla_nest_end(skb, mp_attr);
 254
 255         lastuse = READ_ONCE(c->mfc_un.res.lastuse);
 256         lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
 257
 258         mfcs.mfcs_packets = c->mfc_un.res.pkt;
 259         mfcs.mfcs_bytes = c->mfc_un.res.bytes;
 260         mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
 261         if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) ||
 262             nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse),
 263                               RTA_PAD))
 264                 return -EMSGSIZE;
 265
 266         rtm->rtm_type = RTN_MULTICAST;
 267         return 1;
 268 }
 269 EXPORT_SYMBOL(mr_fill_mroute);
 270
 271 int mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
 272                      struct mr_table *(*iter)(struct net *net,
 273                                               struct mr_table *mrt),
 274                      int (*fill)(struct mr_table *mrt,
 275                                  struct sk_buff *skb,
 276                                  u32 portid, u32 seq, struct mr_mfc *c,
 277                                  int cmd, int flags),
 278                      spinlock_t *lock)
 279 {
 280         unsigned int t = 0, e = 0, s_t = cb->args[0], s_e = cb->args[1];
 281         struct net *net = sock_net(skb->sk);
 282         struct mr_table *mrt;
 283         struct mr_mfc *mfc;
 284
 285         rcu_read_lock();
 286         for (mrt = iter(net, NULL); mrt; mrt = iter(net, mrt)) {
 287                 if (t < s_t)
 288                         goto next_table;
 289                 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) {
 290                         if (e < s_e)
 291                                 goto next_entry;
 292                         if (fill(mrt, skb, NETLINK_CB(cb->skb).portid,
 293                                  cb->nlh->nlmsg_seq, mfc,
 294                                  RTM_NEWROUTE, NLM_F_MULTI) < 0)
 295                                 goto done;
 296 next_entry:
 297                         e++;
 298                 }
 299
 300                 spin_lock_bh(lock);
 301                 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
 302                         if (e < s_e)
 303                                 goto next_entry2;
 304                         if (fill(mrt, skb, NETLINK_CB(cb->skb).portid,
 305                                  cb->nlh->nlmsg_seq, mfc,
 306                                  RTM_NEWROUTE, NLM_F_MULTI) < 0) {
 307                                 spin_unlock_bh(lock);
 308                                 goto done;
 309                         }
 310 next_entry2:
 311                         e++;
 312                 }
 313                 spin_unlock_bh(lock);
 314                 e = 0;
 315                 s_e = 0;
 316 next_table:
 317                 t++;
 318         }
 319 done:
 320         rcu_read_unlock();
 321
 322         cb->args[1] = e;
 323         cb->args[0] = t;
 324
 325         return skb->len;
 326 }
 327 EXPORT_SYMBOL(mr_rtm_dumproute);
 328
 329 int mr_dump(struct net *net, struct notifier_block *nb, unsigned short family,
 330             int (*rules_dump)(struct net *net,
 331                               struct notifier_block *nb),
 332             struct mr_table *(*mr_iter)(struct net *net,
 333                                         struct mr_table *mrt),
 334             rwlock_t *mrt_lock)
 335 {
 336         struct mr_table *mrt;
 337         int err;
 338
 339         err = rules_dump(net, nb);
 340         if (err)
 341                 return err;
 342
 343         for (mrt = mr_iter(net, NULL); mrt; mrt = mr_iter(net, mrt)) {
 344                 struct vif_device *v = &mrt->vif_table[0];
 345                 struct mr_mfc *mfc;
 346                 int vifi;
 347
 348                 /* Notifiy on table VIF entries */
 349                 read_lock(mrt_lock);
 350                 for (vifi = 0; vifi < mrt->maxvif; vifi++, v++) {
 351                         if (!v->dev)
 352                                 continue;
 353
 354                         mr_call_vif_notifier(nb, net, family,
 355                                              FIB_EVENT_VIF_ADD,
 356                                              v, vifi, mrt->id);
 357                 }
 358                 read_unlock(mrt_lock);
 359
 360                 /* Notify on table MFC entries */
 361                 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
 362                         mr_call_mfc_notifier(nb, net, family,
 363                                              FIB_EVENT_ENTRY_ADD,
 364                                              mfc, mrt->id);
 365         }
 366
 367         return 0;
 368 }
 369 EXPORT_SYMBOL(mr_dump);