net/netfilter/ipvs/ip_vs_nfct.c

   1 /*
   2  * ip_vs_nfct.c:        Netfilter connection tracking support for IPVS
   3  *
   4  * Portions Copyright (C) 2001-2002
   5  * Antefacto Ltd, 181 Parnell St, Dublin 1, Ireland.
   6  *
   7  * Portions Copyright (C) 2003-2010
   8  * Julian Anastasov
   9  *
  10  *
  11  * This code is free software; you can redistribute it and/or modify
  12  * it under the terms of the GNU General Public License as published by
  13  * the Free Software Foundation; either version 2 of the License, or
  14  * (at your option) any later version.
  15  *
  16  * This program is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  19  * GNU General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU General Public License
  22  * along with this program; if not, see <http://www.gnu.org/licenses/>.
  23  *
  24  *
  25  * Authors:
  26  * Ben North <ben@redfrontdoor.org>
  27  * Julian Anastasov <ja@ssi.bg>         Reorganize and sync with latest kernels
  28  * Hannes Eder <heder@google.com>       Extend NFCT support for FTP, ipvs match
  29  *
  30  *
  31  * Current status:
  32  *
  33  * - provide conntrack confirmation for new and related connections, by
  34  * this way we can see their proper conntrack state in all hooks
  35  * - support for all forwarding methods, not only NAT
  36  * - FTP support (NAT), ability to support other NAT apps with expectations
  37  * - to correctly create expectations for related NAT connections the proper
  38  * NF conntrack support must be already installed, eg. ip_vs_ftp requires
  39  * nf_conntrack_ftp ... iptables_nat for the same ports (but no iptables
  40  * NAT rules are needed)
  41  * - alter reply for NAT when forwarding packet in original direction:
  42  * conntrack from client in NEW or RELATED (Passive FTP DATA) state or
  43  * when RELATED conntrack is created from real server (Active FTP DATA)
  44  * - if iptables_nat is not loaded the Passive FTP will not work (the
  45  * PASV response can not be NAT-ed) but Active FTP should work
  46  *
  47  */
  48
  49 #define KMSG_COMPONENT "IPVS"
  50 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  51
  52 #include <linux/module.h>
  53 #include <linux/types.h>
  54 #include <linux/kernel.h>
  55 #include <linux/errno.h>
  56 #include <linux/compiler.h>
  57 #include <linux/vmalloc.h>
  58 #include <linux/skbuff.h>
  59 #include <net/ip.h>
  60 #include <linux/netfilter.h>
  61 #include <linux/netfilter_ipv4.h>
  62 #include <net/ip_vs.h>
  63 #include <net/netfilter/nf_conntrack_core.h>
  64 #include <net/netfilter/nf_conntrack_expect.h>
  65 #include <net/netfilter/nf_conntrack_seqadj.h>
  66 #include <net/netfilter/nf_conntrack_helper.h>
  67 #include <net/netfilter/nf_conntrack_zones.h>
  68
  69
  70 #define FMT_TUPLE       "%pI4:%u->%pI4:%u/%u"
  71 #define ARG_TUPLE(T)    &(T)->src.u3.ip, ntohs((T)->src.u.all), \
  72                         &(T)->dst.u3.ip, ntohs((T)->dst.u.all), \
  73                         (T)->dst.protonum
  74
  75 #define FMT_CONN        "%pI4:%u->%pI4:%u->%pI4:%u/%u:%u"
  76 #define ARG_CONN(C)     &((C)->caddr.ip), ntohs((C)->cport), \
  77                         &((C)->vaddr.ip), ntohs((C)->vport), \
  78                         &((C)->daddr.ip), ntohs((C)->dport), \
  79                         (C)->protocol, (C)->state
  80
  81 void
  82 ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp, int outin)
  83 {
  84         enum ip_conntrack_info ctinfo;
  85         struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
  86         struct nf_conntrack_tuple new_tuple;
  87
  88         if (ct == NULL || nf_ct_is_confirmed(ct) || nf_ct_is_untracked(ct) ||
  89             nf_ct_is_dying(ct))
  90                 return;
  91
  92         /* Never alter conntrack for non-NAT conns */
  93         if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)
  94                 return;
  95
  96         /* Alter reply only in original direction */
  97         if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL)
  98                 return;
  99
 100         /* Applications may adjust TCP seqs */
 101         if (cp->app && nf_ct_protonum(ct) == IPPROTO_TCP &&
 102             !nfct_seqadj(ct) && !nfct_seqadj_ext_add(ct))
 103                 return;
 104
 105         /*
 106          * The connection is not yet in the hashtable, so we update it.
 107          * CIP->VIP will remain the same, so leave the tuple in
 108          * IP_CT_DIR_ORIGINAL untouched.  When the reply comes back from the
 109          * real-server we will see RIP->DIP.
 110          */
 111         new_tuple = ct->tuplehash[IP_CT_DIR_REPLY].tuple;
 112         /*
 113          * This will also take care of UDP and other protocols.
 114          */
 115         if (outin) {
 116                 new_tuple.src.u3 = cp->daddr;
 117                 if (new_tuple.dst.protonum != IPPROTO_ICMP &&
 118                     new_tuple.dst.protonum != IPPROTO_ICMPV6)
 119                         new_tuple.src.u.tcp.port = cp->dport;
 120         } else {
 121                 new_tuple.dst.u3 = cp->vaddr;
 122                 if (new_tuple.dst.protonum != IPPROTO_ICMP &&
 123                     new_tuple.dst.protonum != IPPROTO_ICMPV6)
 124                         new_tuple.dst.u.tcp.port = cp->vport;
 125         }
 126         IP_VS_DBG(7, "%s: Updating conntrack ct=%p, status=0x%lX, "
 127                   "ctinfo=%d, old reply=" FMT_TUPLE
 128                   ", new reply=" FMT_TUPLE ", cp=" FMT_CONN "\n",
 129                   __func__, ct, ct->status, ctinfo,
 130                   ARG_TUPLE(&ct->tuplehash[IP_CT_DIR_REPLY].tuple),
 131                   ARG_TUPLE(&new_tuple), ARG_CONN(cp));
 132         nf_conntrack_alter_reply(ct, &new_tuple);
 133 }
 134
 135 int ip_vs_confirm_conntrack(struct sk_buff *skb)
 136 {
 137         return nf_conntrack_confirm(skb);
 138 }
 139
 140 /*
 141  * Called from init_conntrack() as expectfn handler.
 142  */
 143 static void ip_vs_nfct_expect_callback(struct nf_conn *ct,
 144         struct nf_conntrack_expect *exp)
 145 {
 146         struct nf_conntrack_tuple *orig, new_reply;
 147         struct ip_vs_conn *cp;
 148         struct ip_vs_conn_param p;
 149         struct net *net = nf_ct_net(ct);
 150
 151         if (exp->tuple.src.l3num != PF_INET)
 152                 return;
 153
 154         /*
 155          * We assume that no NF locks are held before this callback.
 156          * ip_vs_conn_out_get and ip_vs_conn_in_get should match their
 157          * expectations even if they use wildcard values, now we provide the
 158          * actual values from the newly created original conntrack direction.
 159          * The conntrack is confirmed when packet reaches IPVS hooks.
 160          */
 161
 162         /* RS->CLIENT */
 163         orig = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
 164         ip_vs_conn_fill_param(net, exp->tuple.src.l3num, orig->dst.protonum,
 165                               &orig->src.u3, orig->src.u.tcp.port,
 166                               &orig->dst.u3, orig->dst.u.tcp.port, &p);
 167         cp = ip_vs_conn_out_get(&p);
 168         if (cp) {
 169                 /* Change reply CLIENT->RS to CLIENT->VS */
 170                 new_reply = ct->tuplehash[IP_CT_DIR_REPLY].tuple;
 171                 IP_VS_DBG(7, "%s: ct=%p, status=0x%lX, tuples=" FMT_TUPLE ", "
 172                           FMT_TUPLE ", found inout cp=" FMT_CONN "\n",
 173                           __func__, ct, ct->status,
 174                           ARG_TUPLE(orig), ARG_TUPLE(&new_reply),
 175                           ARG_CONN(cp));
 176                 new_reply.dst.u3 = cp->vaddr;
 177                 new_reply.dst.u.tcp.port = cp->vport;
 178                 IP_VS_DBG(7, "%s: ct=%p, new tuples=" FMT_TUPLE ", " FMT_TUPLE
 179                           ", inout cp=" FMT_CONN "\n",
 180                           __func__, ct,
 181                           ARG_TUPLE(orig), ARG_TUPLE(&new_reply),
 182                           ARG_CONN(cp));
 183                 goto alter;
 184         }
 185
 186         /* CLIENT->VS */
 187         cp = ip_vs_conn_in_get(&p);
 188         if (cp) {
 189                 /* Change reply VS->CLIENT to RS->CLIENT */
 190                 new_reply = ct->tuplehash[IP_CT_DIR_REPLY].tuple;
 191                 IP_VS_DBG(7, "%s: ct=%p, status=0x%lX, tuples=" FMT_TUPLE ", "
 192                           FMT_TUPLE ", found outin cp=" FMT_CONN "\n",
 193                           __func__, ct, ct->status,
 194                           ARG_TUPLE(orig), ARG_TUPLE(&new_reply),
 195                           ARG_CONN(cp));
 196                 new_reply.src.u3 = cp->daddr;
 197                 new_reply.src.u.tcp.port = cp->dport;
 198                 IP_VS_DBG(7, "%s: ct=%p, new tuples=" FMT_TUPLE ", "
 199                           FMT_TUPLE ", outin cp=" FMT_CONN "\n",
 200                           __func__, ct,
 201                           ARG_TUPLE(orig), ARG_TUPLE(&new_reply),
 202                           ARG_CONN(cp));
 203                 goto alter;
 204         }
 205
 206         IP_VS_DBG(7, "%s: ct=%p, status=0x%lX, tuple=" FMT_TUPLE
 207                   " - unknown expect\n",
 208                   __func__, ct, ct->status, ARG_TUPLE(orig));
 209         return;
 210
 211 alter:
 212         /* Never alter conntrack for non-NAT conns */
 213         if (IP_VS_FWD_METHOD(cp) == IP_VS_CONN_F_MASQ)
 214                 nf_conntrack_alter_reply(ct, &new_reply);
 215         ip_vs_conn_put(cp);
 216         return;
 217 }
 218
 219 /*
 220  * Create NF conntrack expectation with wildcard (optional) source port.
 221  * Then the default callback function will alter the reply and will confirm
 222  * the conntrack entry when the first packet comes.
 223  * Use port 0 to expect connection from any port.
 224  */
 225 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
 226                                struct ip_vs_conn *cp, u_int8_t proto,
 227                                const __be16 port, int from_rs)
 228 {
 229         struct nf_conntrack_expect *exp;
 230
 231         if (ct == NULL || nf_ct_is_untracked(ct))
 232                 return;
 233
 234         exp = nf_ct_expect_alloc(ct);
 235         if (!exp)
 236                 return;
 237
 238         nf_ct_expect_init(exp, NF_CT_EXPECT_CLASS_DEFAULT, nf_ct_l3num(ct),
 239                         from_rs ? &cp->daddr : &cp->caddr,
 240                         from_rs ? &cp->caddr : &cp->vaddr,
 241                         proto, port ? &port : NULL,
 242                         from_rs ? &cp->cport : &cp->vport);
 243
 244         exp->expectfn = ip_vs_nfct_expect_callback;
 245
 246         IP_VS_DBG(7, "%s: ct=%p, expect tuple=" FMT_TUPLE "\n",
 247                 __func__, ct, ARG_TUPLE(&exp->tuple));
 248         nf_ct_expect_related(exp);
 249         nf_ct_expect_put(exp);
 250 }
 251 EXPORT_SYMBOL(ip_vs_nfct_expect_related);
 252
 253 /*
 254  * Our connection was terminated, try to drop the conntrack immediately
 255  */
 256 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
 257 {
 258         struct nf_conntrack_tuple_hash *h;
 259         struct nf_conn *ct;
 260         struct nf_conntrack_tuple tuple;
 261
 262         if (!cp->cport)
 263                 return;
 264
 265         tuple = (struct nf_conntrack_tuple) {
 266                 .dst = { .protonum = cp->protocol, .dir = IP_CT_DIR_ORIGINAL } };
 267         tuple.src.u3 = cp->caddr;
 268         tuple.src.u.all = cp->cport;
 269         tuple.src.l3num = cp->af;
 270         tuple.dst.u3 = cp->vaddr;
 271         tuple.dst.u.all = cp->vport;
 272
 273         IP_VS_DBG(7, "%s: dropping conntrack with tuple=" FMT_TUPLE
 274                 " for conn " FMT_CONN "\n",
 275                 __func__, ARG_TUPLE(&tuple), ARG_CONN(cp));
 276
 277         h = nf_conntrack_find_get(cp->ipvs->net, &nf_ct_zone_dflt, &tuple);
 278         if (h) {
 279                 ct = nf_ct_tuplehash_to_ctrack(h);
 280                 /* Show what happens instead of calling nf_ct_kill() */
 281                 if (del_timer(&ct->timeout)) {
 282                         IP_VS_DBG(7, "%s: ct=%p, deleted conntrack timer for tuple="
 283                                 FMT_TUPLE "\n",
 284                                 __func__, ct, ARG_TUPLE(&tuple));
 285                         if (ct->timeout.function)
 286                                 ct->timeout.function(ct->timeout.data);
 287                 } else {
 288                         IP_VS_DBG(7, "%s: ct=%p, no conntrack timer for tuple="
 289                                 FMT_TUPLE "\n",
 290                                 __func__, ct, ARG_TUPLE(&tuple));
 291                 }
 292                 nf_ct_put(ct);
 293         } else {
 294                 IP_VS_DBG(7, "%s: no conntrack for tuple=" FMT_TUPLE "\n",
 295                         __func__, ARG_TUPLE(&tuple));
 296         }
 297 }
 298