[NETFILTER]: Add new iptables "connbytes" match

This patch ads a new "connbytes" match that utilizes the CONFIG_NF_CT_ACCT
per-connection byte and packet counters.  Using it you can do things like
packet classification on average packet size within a connection.

Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

4 files changed, 202 insertions, 1 deletions

diff --git a/include/linux/netfilter_ipv4/ipt_connbytes.h b/include/linux/netfilter_ipv4/ipt_connbytes.h
new file mode 100644
index 000000000000..abaa65afd4e9
--- /dev/null
+++ b/include/linux/netfilter_ipv4/ipt_connbytes.h
@@ -0,0 +1,25 @@
+#ifndef _IPT_CONNBYTES_H
+#define _IPT_CONNBYTES_H
+
+enum ipt_connbytes_what {
+        IPT_CONNBYTES_WHAT_PKTS,
+        IPT_CONNBYTES_WHAT_BYTES,
+        IPT_CONNBYTES_WHAT_AVGPKT,
+};
+
+enum ipt_connbytes_direction {
+        IPT_CONNBYTES_DIR_ORIGINAL,
+        IPT_CONNBYTES_DIR_REPLY,
+        IPT_CONNBYTES_DIR_BOTH,
+};
+
+struct ipt_connbytes_info
+{
+        struct {
+                aligned_u64 from;       /* count to be matched */
+                aligned_u64 to;         /* count to be matched */
+        } count;
+        u_int8_t what;          /* ipt_connbytes_what */
+        u_int8_t direction;     /* ipt_connbytes_direction */
+};
+#endif
diff --git a/net/ipv4/netfilter/Kconfig b/net/ipv4/netfilter/Kconfig
index 9f5e1d769b5f..3f7e6e49cbdd 100644
--- a/net/ipv4/netfilter/Kconfig
+++ b/net/ipv4/netfilter/Kconfig
@@ -386,6 +386,16 @@ config IP_NF_MATCH_CONNMARK
           <file:Documentation/modules.txt>.  The module will be called
           ipt_connmark.o.  If unsure, say `N'.
 
+config IP_NF_MATCH_CONNBYTES
+        tristate  'Connection byte/packet counter match support'
+        depends on IP_NF_CT_ACCT && IP_NF_IPTABLES
+        help
+          This option adds a `connbytes' match, which allows you to match the
+          number of bytes and/or packets for each direction within a connection.
+
+          If you want to compile it as a module, say M here and read
+          <file:Documentation/modules.txt>.  If unsure, say `N'.
+
 config IP_NF_MATCH_HASHLIMIT
         tristate  'hashlimit match support'
         depends on IP_NF_IPTABLES
@@ -723,6 +733,5 @@ config IP_NF_CONNTRACK_NETLINK
         help
           This option enables support for a netlink-based userspace interface
 
-
 endmenu
 
diff --git a/net/ipv4/netfilter/Makefile b/net/ipv4/netfilter/Makefile
index 58aa7c616e1f..7c8ae858aa43 100644
--- a/net/ipv4/netfilter/Makefile
+++ b/net/ipv4/netfilter/Makefile
@@ -59,6 +59,7 @@ obj-$(CONFIG_IP_NF_MATCH_TTL) += ipt_ttl.o
 obj-$(CONFIG_IP_NF_MATCH_STATE) += ipt_state.o
 obj-$(CONFIG_IP_NF_MATCH_CONNMARK) += ipt_connmark.o
 obj-$(CONFIG_IP_NF_MATCH_CONNTRACK) += ipt_conntrack.o
+obj-$(CONFIG_IP_NF_MATCH_CONNBYTES) += ipt_connbytes.o
 obj-$(CONFIG_IP_NF_MATCH_TCPMSS) += ipt_tcpmss.o
 obj-$(CONFIG_IP_NF_MATCH_REALM) += ipt_realm.o
 obj-$(CONFIG_IP_NF_MATCH_ADDRTYPE) += ipt_addrtype.o
diff --git a/net/ipv4/netfilter/ipt_connbytes.c b/net/ipv4/netfilter/ipt_connbytes.c
new file mode 100644
index 000000000000..0dfb52c0e808
--- /dev/null
+++ b/net/ipv4/netfilter/ipt_connbytes.c
@@ -0,0 +1,166 @@
+/* Kernel module to match connection tracking byte counter.
+ * GPL (C) 2002 Martin Devera (devik@cdi.cz).
+ *
+ * 2004-07-20 Harald Welte <laforge@netfilter.org>
+ *      - reimplemented to use per-connection accounting counters
+ *      - add functionality to match number of packets
+ *      - add functionality to match average packet size
+ *      - add support to match directions seperately
+ *
+ */
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/netfilter_ipv4/ip_conntrack.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ipt_connbytes.h>
+
+#include <asm/div64.h>
+#include <asm/bitops.h>
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
+MODULE_DESCRIPTION("iptables match for matching number of pkts/bytes per connection");
+
+/* 64bit divisor, dividend and result. dynamic precision */
+static u_int64_t div64_64(u_int64_t divisor, u_int64_t dividend)
+{
+        u_int64_t result = divisor;
+
+        if (dividend > 0xffffffff) {
+                int first_bit = find_first_bit((unsigned long *) &dividend, sizeof(dividend));
+                /* calculate number of bits to shift. shift exactly enough
+                 * bits to make dividend fit in 32bits. */
+                int num_shift = (64 - 32 - first_bit);
+                /* first bit has to be < 32, since dividend was > 0xffffffff */
+                result = result >> num_shift;
+                dividend = dividend >> num_shift;
+        }
+
+        do_div(divisor, dividend);
+
+        return divisor;
+}
+
+static int
+match(const struct sk_buff *skb,
+      const struct net_device *in,
+      const struct net_device *out,
+      const void *matchinfo,
+      int offset,
+      int *hotdrop)
+{
+        const struct ipt_connbytes_info *sinfo = matchinfo;
+        enum ip_conntrack_info ctinfo;
+        struct ip_conntrack *ct;
+        u_int64_t what = 0;     /* initialize to make gcc happy */
+
+        if (!(ct = ip_conntrack_get((struct sk_buff *)skb, &ctinfo)))
+                return 0; /* no match */
+
+        switch (sinfo->what) {
+        case IPT_CONNBYTES_WHAT_PKTS:
+                switch (sinfo->direction) {
+                case IPT_CONNBYTES_DIR_ORIGINAL:
+                        what = ct->counters[IP_CT_DIR_ORIGINAL].packets;
+                        break;
+                case IPT_CONNBYTES_DIR_REPLY:
+                        what = ct->counters[IP_CT_DIR_REPLY].packets;
+                        break;
+                case IPT_CONNBYTES_DIR_BOTH:
+                        what = ct->counters[IP_CT_DIR_ORIGINAL].packets;
+                        what += ct->counters[IP_CT_DIR_REPLY].packets;
+                        break;
+                }
+                break;
+        case IPT_CONNBYTES_WHAT_BYTES:
+                switch (sinfo->direction) {
+                case IPT_CONNBYTES_DIR_ORIGINAL:
+                        what = ct->counters[IP_CT_DIR_ORIGINAL].bytes;
+                        break;
+                case IPT_CONNBYTES_DIR_REPLY:
+                        what = ct->counters[IP_CT_DIR_REPLY].bytes;
+                        break;
+                case IPT_CONNBYTES_DIR_BOTH:
+                        what = ct->counters[IP_CT_DIR_ORIGINAL].bytes;
+                        what += ct->counters[IP_CT_DIR_REPLY].bytes;
+                        break;
+                }
+                break;
+        case IPT_CONNBYTES_WHAT_AVGPKT:
+                switch (sinfo->direction) {
+                case IPT_CONNBYTES_DIR_ORIGINAL:
+                        what = div64_64(ct->counters[IP_CT_DIR_ORIGINAL].bytes,
+                                        ct->counters[IP_CT_DIR_ORIGINAL].packets);
+                        break;
+                case IPT_CONNBYTES_DIR_REPLY:
+                        what = div64_64(ct->counters[IP_CT_DIR_REPLY].bytes,
+                                        ct->counters[IP_CT_DIR_REPLY].packets);
+                        break;
+                case IPT_CONNBYTES_DIR_BOTH:
+                        {
+                                u_int64_t bytes;
+                                u_int64_t pkts;
+                                bytes = ct->counters[IP_CT_DIR_ORIGINAL].bytes +
+                                        ct->counters[IP_CT_DIR_REPLY].bytes;
+                                pkts = ct->counters[IP_CT_DIR_ORIGINAL].packets+
+                                        ct->counters[IP_CT_DIR_REPLY].packets;
+
+                                /* FIXME_THEORETICAL: what to do if sum
+                                 * overflows ? */
+
+                                what = div64_64(bytes, pkts);
+                        }
+                        break;
+                }
+                break;
+        }
+
+        if (sinfo->count.to)
+                return (what <= sinfo->count.to && what >= sinfo->count.from);
+        else
+                return (what >= sinfo->count.from);
+}
+
+static int check(const char *tablename,
+                 const struct ipt_ip *ip,
+                 void *matchinfo,
+                 unsigned int matchsize,
+                 unsigned int hook_mask)
+{
+        const struct ipt_connbytes_info *sinfo = matchinfo;
+
+        if (matchsize != IPT_ALIGN(sizeof(struct ipt_connbytes_info)))
+                return 0;
+
+        if (sinfo->what != IPT_CONNBYTES_WHAT_PKTS &&
+            sinfo->what != IPT_CONNBYTES_WHAT_BYTES &&
+            sinfo->what != IPT_CONNBYTES_WHAT_AVGPKT)
+                return 0;
+
+        if (sinfo->direction != IPT_CONNBYTES_DIR_ORIGINAL &&
+            sinfo->direction != IPT_CONNBYTES_DIR_REPLY &&
+            sinfo->direction != IPT_CONNBYTES_DIR_BOTH)
+                return 0;
+
+        return 1;
+}
+
+static struct ipt_match state_match = {
+        .name           = "connbytes",
+        .match          = &match,
+        .checkentry     = &check,
+        .me             = THIS_MODULE
+};
+
+static int __init init(void)
+{
+        return ipt_register_match(&state_match);
+}
+
+static void __exit fini(void)
+{
+        ipt_unregister_match(&state_match);
+}
+
+module_init(init);
+module_exit(fini);