[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / netfilter / xt_cluster.c

/*
 * (C) 2008-2009 Pablo Neira Ayuso <pablo@netfilter.org>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */
#include <linux/module.h>
#include <linux/skbuff.h>
#include <linux/jhash.h>
#include <linux/ip.h>
#include <net/ipv6.h>

#include <linux/netfilter/x_tables.h>
#include <net/netfilter/nf_conntrack.h>
#include <linux/netfilter/xt_cluster.h>

static inline u_int32_t nf_ct_orig_ipv4_src(const struct nf_conn *ct)
{
	return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip;
}

static inline const void *nf_ct_orig_ipv6_src(const struct nf_conn *ct)
{
	return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip6;
}

static inline u_int32_t
xt_cluster_hash_ipv4(u_int32_t ip, const struct xt_cluster_match_info *info)
{
	return jhash_1word(ip, info->hash_seed);
}

static inline u_int32_t
xt_cluster_hash_ipv6(const void *ip, const struct xt_cluster_match_info *info)
{
	return jhash2(ip, NF_CT_TUPLE_L3SIZE / sizeof(__u32), info->hash_seed);
}

static inline u_int32_t
xt_cluster_hash(const struct nf_conn *ct,
		const struct xt_cluster_match_info *info)
{
	u_int32_t hash = 0;

	switch(nf_ct_l3num(ct)) {
	case AF_INET:
		hash = xt_cluster_hash_ipv4(nf_ct_orig_ipv4_src(ct), info);
		break;
	case AF_INET6:
		hash = xt_cluster_hash_ipv6(nf_ct_orig_ipv6_src(ct), info);
		break;
	default:
		WARN_ON(1);
		break;
	}
	return (((u64)hash * info->total_nodes) >> 32);
}

static inline bool
xt_cluster_is_multicast_addr(const struct sk_buff *skb, u_int8_t family)
{
	bool is_multicast = false;

	switch(family) {
	case NFPROTO_IPV4:
		is_multicast = ipv4_is_multicast(ip_hdr(skb)->daddr);
		break;
	case NFPROTO_IPV6:
		is_multicast = ipv6_addr_type(&ipv6_hdr(skb)->daddr) &
						IPV6_ADDR_MULTICAST;
		break;
	default:
		WARN_ON(1);
		break;
	}
	return is_multicast;
}

static bool
xt_cluster_mt(const struct sk_buff *skb, const struct xt_match_param *par)
{
	struct sk_buff *pskb = (struct sk_buff *)skb;
	const struct xt_cluster_match_info *info = par->matchinfo;
	const struct nf_conn *ct;
	enum ip_conntrack_info ctinfo;
	unsigned long hash;

	/* This match assumes that all nodes see the same packets. This can be
	 * achieved if the switch that connects the cluster nodes support some
	 * sort of 'port mirroring'. However, if your switch does not support
	 * this, your cluster nodes can reply ARP request using a multicast MAC
	 * address. Thus, your switch will flood the same packets to the
	 * cluster nodes with the same multicast MAC address. Using a multicast
	 * link address is a RFC 1812 (section 3.3.2) violation, but this works
	 * fine in practise.
	 *
	 * Unfortunately, if you use the multicast MAC address, the link layer
	 * sets skbuff's pkt_type to PACKET_MULTICAST, which is not accepted
	 * by TCP and others for packets coming to this node. For that reason,
	 * this match mangles skbuff's pkt_type if it detects a packet
	 * addressed to a unicast address but using PACKET_MULTICAST. Yes, I
	 * know, matches should not alter packets, but we are doing this here
	 * because we would need to add a PKTTYPE target for this sole purpose.
	 */
	if (!xt_cluster_is_multicast_addr(skb, par->family) &&
	    skb->pkt_type == PACKET_MULTICAST) {
	    	pskb->pkt_type = PACKET_HOST;
	}

	ct = nf_ct_get(skb, &ctinfo);
	if (ct == NULL)
		return false;

	if (ct == &nf_conntrack_untracked)
		return false;

	if (ct->master)
		hash = xt_cluster_hash(ct->master, info);
	else
		hash = xt_cluster_hash(ct, info);

	return !!((1 << hash) & info->node_mask) ^
	       !!(info->flags & XT_CLUSTER_F_INV);
}

static bool xt_cluster_mt_checkentry(const struct xt_mtchk_param *par)
{
	struct xt_cluster_match_info *info = par->matchinfo;

	if (info->node_mask >= (1 << info->total_nodes)) {
		printk(KERN_ERR "xt_cluster: this node mask cannot be "
				"higher than the total number of nodes\n");
		return false;
	}
	return true;
}

static struct xt_match xt_cluster_match __read_mostly = {
	.name		= "cluster",
	.family		= NFPROTO_UNSPEC,
	.match		= xt_cluster_mt,
	.checkentry	= xt_cluster_mt_checkentry,
	.matchsize	= sizeof(struct xt_cluster_match_info),
	.me		= THIS_MODULE,
};

static int __init xt_cluster_mt_init(void)
{
	return xt_register_match(&xt_cluster_match);
}

static void __exit xt_cluster_mt_fini(void)
{
	xt_unregister_match(&xt_cluster_match);
}

MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("Xtables: hash-based cluster match");
MODULE_ALIAS("ipt_cluster");
MODULE_ALIAS("ip6t_cluster");
module_init(xt_cluster_mt_init);
module_exit(xt_cluster_mt_fini);
Commit	Line	Data
0269ea49 PNA	1	/*
	2	* (C) 2008-2009 Pablo Neira Ayuso <pablo@netfilter.org>
	3	*
	4	* This program is free software; you can redistribute it and/or modify
	5	* it under the terms of the GNU General Public License version 2 as
	6	* published by the Free Software Foundation.
	7	*/
	8	#include <linux/module.h>
	9	#include <linux/skbuff.h>
	10	#include <linux/jhash.h>
	11	#include <linux/ip.h>
	12	#include <net/ipv6.h>
	13
	14	#include <linux/netfilter/x_tables.h>
	15	#include <net/netfilter/nf_conntrack.h>
	16	#include <linux/netfilter/xt_cluster.h>
	17
	18	static inline u_int32_t nf_ct_orig_ipv4_src(const struct nf_conn *ct)
	19	{
	20	return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip;
	21	}
	22
	23	static inline const void nf_ct_orig_ipv6_src(const struct nf_conn ct)
	24	{
	25	return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip6;
	26	}
	27
	28	static inline u_int32_t
	29	xt_cluster_hash_ipv4(u_int32_t ip, const struct xt_cluster_match_info *info)
	30	{
	31	return jhash_1word(ip, info->hash_seed);
	32	}
	33
	34	static inline u_int32_t
	35	xt_cluster_hash_ipv6(const void ip, const struct xt_cluster_match_info info)
	36	{
	37	return jhash2(ip, NF_CT_TUPLE_L3SIZE / sizeof(__u32), info->hash_seed);
	38	}
	39
	40	static inline u_int32_t
	41	xt_cluster_hash(const struct nf_conn *ct,
	42	const struct xt_cluster_match_info *info)
	43	{
	44	u_int32_t hash = 0;
	45
	46	switch(nf_ct_l3num(ct)) {
	47	case AF_INET:
	48	hash = xt_cluster_hash_ipv4(nf_ct_orig_ipv4_src(ct), info);
	49	break;
	50	case AF_INET6:
	51	hash = xt_cluster_hash_ipv6(nf_ct_orig_ipv6_src(ct), info);
	52	break;
	53	default:
	54	WARN_ON(1);
	55	break;
	56	}
	57	return (((u64)hash * info->total_nodes) >> 32);
	58	}
	59
	60	static inline bool
	61	xt_cluster_is_multicast_addr(const struct sk_buff *skb, u_int8_t family)
	62	{
	63	bool is_multicast = false;
	64
65	switch(family) {
66	case NFPROTO_IPV4:
67	is_multicast = ipv4_is_multicast(ip_hdr(skb)->daddr);
68	break;
69	case NFPROTO_IPV6:
70	is_multicast = ipv6_addr_type(&ipv6_hdr(skb)->daddr) &
71	IPV6_ADDR_MULTICAST;
72	break;
73	default:
74	WARN_ON(1);
75	break;
76	}
77	return is_multicast;
78	}
79
80	static bool
81	xt_cluster_mt(const struct sk_buff skb, const struct xt_match_param par)
82	{
83	struct sk_buff pskb = (struct sk_buff )skb;
84	const struct xt_cluster_match_info *info = par->matchinfo;
85	const struct nf_conn *ct;
86	enum ip_conntrack_info ctinfo;
87	unsigned long hash;
88
89	/* This match assumes that all nodes see the same packets. This can be
90	* achieved if the switch that connects the cluster nodes support some
91	* sort of 'port mirroring'. However, if your switch does not support
92	* this, your cluster nodes can reply ARP request using a multicast MAC
93	* address. Thus, your switch will flood the same packets to the
94	* cluster nodes with the same multicast MAC address. Using a multicast
95	* link address is a RFC 1812 (section 3.3.2) violation, but this works
96	* fine in practise.
97	*
98	* Unfortunately, if you use the multicast MAC address, the link layer
99	* sets skbuff's pkt_type to PACKET_MULTICAST, which is not accepted
100	* by TCP and others for packets coming to this node. For that reason,
101	* this match mangles skbuff's pkt_type if it detects a packet
102	* addressed to a unicast address but using PACKET_MULTICAST. Yes, I
103	* know, matches should not alter packets, but we are doing this here
104	* because we would need to add a PKTTYPE target for this sole purpose.
105	*/
106	if (!xt_cluster_is_multicast_addr(skb, par->family) &&
107	skb->pkt_type == PACKET_MULTICAST) {
108	pskb->pkt_type = PACKET_HOST;
109	}
110
111	ct = nf_ct_get(skb, &ctinfo);
112	if (ct == NULL)
113	return false;
114
115	if (ct == &nf_conntrack_untracked)
116	return false;
117
118	if (ct->master)
119	hash = xt_cluster_hash(ct->master, info);
120	else
121	hash = xt_cluster_hash(ct, info);
122
123	return !!((1 << hash) & info->node_mask) ^
124	!!(info->flags & XT_CLUSTER_F_INV);
125	}
126
127	static bool xt_cluster_mt_checkentry(const struct xt_mtchk_param *par)
128	{
129	struct xt_cluster_match_info *info = par->matchinfo;
130
131	if (info->node_mask >= (1 << info->total_nodes)) {
132	printk(KERN_ERR "xt_cluster: this node mask cannot be "
133	"higher than the total number of nodes\n");
134	return false;
135	}
136	return true;
137	}
138
139	static struct xt_match xt_cluster_match __read_mostly = {
140	.name = "cluster",
141	.family = NFPROTO_UNSPEC,
142	.match = xt_cluster_mt,
143	.checkentry = xt_cluster_mt_checkentry,
144	.matchsize = sizeof(struct xt_cluster_match_info),
145	.me = THIS_MODULE,
146	};
147
148	static int __init xt_cluster_mt_init(void)
149	{
150	return xt_register_match(&xt_cluster_match);
151	}
152
153	static void __exit xt_cluster_mt_fini(void)
154	{
155	xt_unregister_match(&xt_cluster_match);
156	}
157
158	MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
159	MODULE_LICENSE("GPL");
160	MODULE_DESCRIPTION("Xtables: hash-based cluster match");
161	MODULE_ALIAS("ipt_cluster");
162	MODULE_ALIAS("ip6t_cluster");
163	module_init(xt_cluster_mt_init);
164	module_exit(xt_cluster_mt_fini);