[deliverable/linux.git] / net / netfilter / ipvs / ip_vs_proto_udp.c

/*
 * ip_vs_proto_udp.c:	UDP load balancing support for IPVS
 *
 * Authors:     Wensong Zhang <wensong@linuxvirtualserver.org>
 *              Julian Anastasov <ja@ssi.bg>
 *
 *              This program is free software; you can redistribute it and/or
 *              modify it under the terms of the GNU General Public License
 *              as published by the Free Software Foundation; either version
 *              2 of the License, or (at your option) any later version.
 *
 * Changes:     Hans Schillstrom <hans.schillstrom@ericsson.com>
 *              Network name space (netns) aware.
 *
 */

#define KMSG_COMPONENT "IPVS"
#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt

#include <linux/in.h>
#include <linux/ip.h>
#include <linux/kernel.h>
#include <linux/netfilter.h>
#include <linux/netfilter_ipv4.h>
#include <linux/udp.h>

#include <net/ip_vs.h>
#include <net/ip.h>
#include <net/ip6_checksum.h>

static int
udp_conn_schedule(struct netns_ipvs *ipvs, int af, struct sk_buff *skb,
		  struct ip_vs_proto_data *pd,
		  int *verdict, struct ip_vs_conn **cpp,
		  struct ip_vs_iphdr *iph)
{
	struct ip_vs_service *svc;
	struct udphdr _udph, *uh;
	__be16 _ports[2], *ports = NULL;

	if (likely(!ip_vs_iph_icmp(iph))) {
		/* IPv6 fragments, only first fragment will hit this */
		uh = skb_header_pointer(skb, iph->len, sizeof(_udph), &_udph);
		if (uh)
			ports = &uh->source;
	} else {
		ports = skb_header_pointer(
			skb, iph->len, sizeof(_ports), &_ports);
	}

	if (!ports) {
		*verdict = NF_DROP;
		return 0;
	}

	rcu_read_lock();
	if (likely(!ip_vs_iph_inverse(iph)))
		svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
					 &iph->daddr, ports[1]);
	else
		svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
					 &iph->saddr, ports[0]);

	if (svc) {
		int ignored;

		if (ip_vs_todrop(ipvs)) {
			/*
			 * It seems that we are very loaded.
			 * We have to drop this packet :(
			 */
			rcu_read_unlock();
			*verdict = NF_DROP;
			return 0;
		}

		/*
		 * Let the virtual server select a real server for the
		 * incoming connection, and create a connection entry.
		 */
		*cpp = ip_vs_schedule(svc, skb, pd, &ignored, iph);
		if (!*cpp && ignored <= 0) {
			if (!ignored)
				*verdict = ip_vs_leave(svc, skb, pd, iph);
			else
				*verdict = NF_DROP;
			rcu_read_unlock();
			return 0;
		}
	}
	rcu_read_unlock();
	/* NF_ACCEPT */
	return 1;
}


static inline void
udp_fast_csum_update(int af, struct udphdr *uhdr,
		     const union nf_inet_addr *oldip,
		     const union nf_inet_addr *newip,
		     __be16 oldport, __be16 newport)
{
#ifdef CONFIG_IP_VS_IPV6
	if (af == AF_INET6)
		uhdr->check =
			csum_fold(ip_vs_check_diff16(oldip->ip6, newip->ip6,
					 ip_vs_check_diff2(oldport, newport,
						~csum_unfold(uhdr->check))));
	else
#endif
		uhdr->check =
			csum_fold(ip_vs_check_diff4(oldip->ip, newip->ip,
					 ip_vs_check_diff2(oldport, newport,
						~csum_unfold(uhdr->check))));
	if (!uhdr->check)
		uhdr->check = CSUM_MANGLED_0;
}

static inline void
udp_partial_csum_update(int af, struct udphdr *uhdr,
		     const union nf_inet_addr *oldip,
		     const union nf_inet_addr *newip,
		     __be16 oldlen, __be16 newlen)
{
#ifdef CONFIG_IP_VS_IPV6
	if (af == AF_INET6)
		uhdr->check =
			~csum_fold(ip_vs_check_diff16(oldip->ip6, newip->ip6,
					 ip_vs_check_diff2(oldlen, newlen,
						csum_unfold(uhdr->check))));
	else
#endif
	uhdr->check =
		~csum_fold(ip_vs_check_diff4(oldip->ip, newip->ip,
				ip_vs_check_diff2(oldlen, newlen,
						csum_unfold(uhdr->check))));
}


static int
udp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
		 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
{
	struct udphdr *udph;
	unsigned int udphoff = iph->len;
	int oldlen;
	int payload_csum = 0;

#ifdef CONFIG_IP_VS_IPV6
	if (cp->af == AF_INET6 && iph->fragoffs)
		return 1;
#endif
	oldlen = skb->len - udphoff;

	/* csum_check requires unshared skb */
	if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
		return 0;

	if (unlikely(cp->app != NULL)) {
		int ret;

		/* Some checks before mangling */
		if (pp->csum_check && !pp->csum_check(cp->af, skb, pp))
			return 0;

		/*
		 *	Call application helper if needed
		 */
		if (!(ret = ip_vs_app_pkt_out(cp, skb)))
			return 0;
		/* ret=2: csum update is needed after payload mangling */
		if (ret == 1)
			oldlen = skb->len - udphoff;
		else
			payload_csum = 1;
	}

	udph = (void *)skb_network_header(skb) + udphoff;
	udph->source = cp->vport;

	/*
	 *	Adjust UDP checksums
	 */
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
		udp_partial_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
					htons(oldlen),
					htons(skb->len - udphoff));
	} else if (!payload_csum && (udph->check != 0)) {
		/* Only port and addr are changed, do fast csum update */
		udp_fast_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
				     cp->dport, cp->vport);
		if (skb->ip_summed == CHECKSUM_COMPLETE)
			skb->ip_summed = (cp->app && pp->csum_check) ?
					 CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
	} else {
		/* full checksum calculation */
		udph->check = 0;
		skb->csum = skb_checksum(skb, udphoff, skb->len - udphoff, 0);
#ifdef CONFIG_IP_VS_IPV6
		if (cp->af == AF_INET6)
			udph->check = csum_ipv6_magic(&cp->vaddr.in6,
						      &cp->caddr.in6,
						      skb->len - udphoff,
						      cp->protocol, skb->csum);
		else
#endif
			udph->check = csum_tcpudp_magic(cp->vaddr.ip,
							cp->caddr.ip,
							skb->len - udphoff,
							cp->protocol,
							skb->csum);
		if (udph->check == 0)
			udph->check = CSUM_MANGLED_0;
		skb->ip_summed = CHECKSUM_UNNECESSARY;
		IP_VS_DBG(11, "O-pkt: %s O-csum=%d (+%zd)\n",
			  pp->name, udph->check,
			  (char*)&(udph->check) - (char*)udph);
	}
	return 1;
}


static int
udp_dnat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
		 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph)
{
	struct udphdr *udph;
	unsigned int udphoff = iph->len;
	int oldlen;
	int payload_csum = 0;

#ifdef CONFIG_IP_VS_IPV6
	if (cp->af == AF_INET6 && iph->fragoffs)
		return 1;
#endif
	oldlen = skb->len - udphoff;

	/* csum_check requires unshared skb */
	if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
		return 0;

	if (unlikely(cp->app != NULL)) {
		int ret;

		/* Some checks before mangling */
		if (pp->csum_check && !pp->csum_check(cp->af, skb, pp))
			return 0;

		/*
		 *	Attempt ip_vs_app call.
		 *	It will fix ip_vs_conn
		 */
		if (!(ret = ip_vs_app_pkt_in(cp, skb)))
			return 0;
		/* ret=2: csum update is needed after payload mangling */
		if (ret == 1)
			oldlen = skb->len - udphoff;
		else
			payload_csum = 1;
	}

	udph = (void *)skb_network_header(skb) + udphoff;
	udph->dest = cp->dport;

	/*
	 *	Adjust UDP checksums
	 */
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
		udp_partial_csum_update(cp->af, udph, &cp->vaddr, &cp->daddr,
					htons(oldlen),
					htons(skb->len - udphoff));
	} else if (!payload_csum && (udph->check != 0)) {
		/* Only port and addr are changed, do fast csum update */
		udp_fast_csum_update(cp->af, udph, &cp->vaddr, &cp->daddr,
				     cp->vport, cp->dport);
		if (skb->ip_summed == CHECKSUM_COMPLETE)
			skb->ip_summed = (cp->app && pp->csum_check) ?
					 CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
	} else {
		/* full checksum calculation */
		udph->check = 0;
		skb->csum = skb_checksum(skb, udphoff, skb->len - udphoff, 0);
#ifdef CONFIG_IP_VS_IPV6
		if (cp->af == AF_INET6)
			udph->check = csum_ipv6_magic(&cp->caddr.in6,
						      &cp->daddr.in6,
						      skb->len - udphoff,
						      cp->protocol, skb->csum);
		else
#endif
			udph->check = csum_tcpudp_magic(cp->caddr.ip,
							cp->daddr.ip,
							skb->len - udphoff,
							cp->protocol,
							skb->csum);
		if (udph->check == 0)
			udph->check = CSUM_MANGLED_0;
		skb->ip_summed = CHECKSUM_UNNECESSARY;
	}
	return 1;
}


static int
udp_csum_check(int af, struct sk_buff *skb, struct ip_vs_protocol *pp)
{
	struct udphdr _udph, *uh;
	unsigned int udphoff;

#ifdef CONFIG_IP_VS_IPV6
	if (af == AF_INET6)
		udphoff = sizeof(struct ipv6hdr);
	else
#endif
		udphoff = ip_hdrlen(skb);

	uh = skb_header_pointer(skb, udphoff, sizeof(_udph), &_udph);
	if (uh == NULL)
		return 0;

	if (uh->check != 0) {
		switch (skb->ip_summed) {
		case CHECKSUM_NONE:
			skb->csum = skb_checksum(skb, udphoff,
						 skb->len - udphoff, 0);
		case CHECKSUM_COMPLETE:
#ifdef CONFIG_IP_VS_IPV6
			if (af == AF_INET6) {
				if (csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
						    &ipv6_hdr(skb)->daddr,
						    skb->len - udphoff,
						    ipv6_hdr(skb)->nexthdr,
						    skb->csum)) {
					IP_VS_DBG_RL_PKT(0, af, pp, skb, 0,
							 "Failed checksum for");
					return 0;
				}
			} else
#endif
				if (csum_tcpudp_magic(ip_hdr(skb)->saddr,
						      ip_hdr(skb)->daddr,
						      skb->len - udphoff,
						      ip_hdr(skb)->protocol,
						      skb->csum)) {
					IP_VS_DBG_RL_PKT(0, af, pp, skb, 0,
							 "Failed checksum for");
					return 0;
				}
			break;
		default:
			/* No need to checksum. */
			break;
		}
	}
	return 1;
}

static inline __u16 udp_app_hashkey(__be16 port)
{
	return (((__force u16)port >> UDP_APP_TAB_BITS) ^ (__force u16)port)
		& UDP_APP_TAB_MASK;
}


static int udp_register_app(struct netns_ipvs *ipvs, struct ip_vs_app *inc)
{
	struct ip_vs_app *i;
	__u16 hash;
	__be16 port = inc->port;
	int ret = 0;
	struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);

	hash = udp_app_hashkey(port);

	list_for_each_entry(i, &ipvs->udp_apps[hash], p_list) {
		if (i->port == port) {
			ret = -EEXIST;
			goto out;
		}
	}
	list_add_rcu(&inc->p_list, &ipvs->udp_apps[hash]);
	atomic_inc(&pd->appcnt);

  out:
	return ret;
}


static void
udp_unregister_app(struct netns_ipvs *ipvs, struct ip_vs_app *inc)
{
	struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);

	atomic_dec(&pd->appcnt);
	list_del_rcu(&inc->p_list);
}


static int udp_app_conn_bind(struct ip_vs_conn *cp)
{
	struct netns_ipvs *ipvs = cp->ipvs;
	int hash;
	struct ip_vs_app *inc;
	int result = 0;

	/* Default binding: bind app only for NAT */
	if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)
		return 0;

	/* Lookup application incarnations and bind the right one */
	hash = udp_app_hashkey(cp->vport);

	rcu_read_lock();
	list_for_each_entry_rcu(inc, &ipvs->udp_apps[hash], p_list) {
		if (inc->port == cp->vport) {
			if (unlikely(!ip_vs_app_inc_get(inc)))
				break;
			rcu_read_unlock();

			IP_VS_DBG_BUF(9, "%s(): Binding conn %s:%u->"
				      "%s:%u to app %s on port %u\n",
				      __func__,
				      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
				      ntohs(cp->cport),
				      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
				      ntohs(cp->vport),
				      inc->name, ntohs(inc->port));

			cp->app = inc;
			if (inc->init_conn)
				result = inc->init_conn(inc, cp);
			goto out;
		}
	}
	rcu_read_unlock();

  out:
	return result;
}


static const int udp_timeouts[IP_VS_UDP_S_LAST+1] = {
	[IP_VS_UDP_S_NORMAL]		=	5*60*HZ,
	[IP_VS_UDP_S_LAST]		=	2*HZ,
};

static const char *const udp_state_name_table[IP_VS_UDP_S_LAST+1] = {
	[IP_VS_UDP_S_NORMAL]		=	"UDP",
	[IP_VS_UDP_S_LAST]		=	"BUG!",
};

static const char * udp_state_name(int state)
{
	if (state >= IP_VS_UDP_S_LAST)
		return "ERR!";
	return udp_state_name_table[state] ? udp_state_name_table[state] : "?";
}

static void
udp_state_transition(struct ip_vs_conn *cp, int direction,
		     const struct sk_buff *skb,
		     struct ip_vs_proto_data *pd)
{
	if (unlikely(!pd)) {
		pr_err("UDP no ns data\n");
		return;
	}

	cp->timeout = pd->timeout_table[IP_VS_UDP_S_NORMAL];
}

static int __udp_init(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
{
	ip_vs_init_hash_table(ipvs->udp_apps, UDP_APP_TAB_SIZE);
	pd->timeout_table = ip_vs_create_timeout_table((int *)udp_timeouts,
							sizeof(udp_timeouts));
	if (!pd->timeout_table)
		return -ENOMEM;
	return 0;
}

static void __udp_exit(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
{
	kfree(pd->timeout_table);
}


struct ip_vs_protocol ip_vs_protocol_udp = {
	.name =			"UDP",
	.protocol =		IPPROTO_UDP,
	.num_states =		IP_VS_UDP_S_LAST,
	.dont_defrag =		0,
	.init =			NULL,
	.exit =			NULL,
	.init_netns =		__udp_init,
	.exit_netns =		__udp_exit,
	.conn_schedule =	udp_conn_schedule,
	.conn_in_get =		ip_vs_conn_in_get_proto,
	.conn_out_get =		ip_vs_conn_out_get_proto,
	.snat_handler =		udp_snat_handler,
	.dnat_handler =		udp_dnat_handler,
	.csum_check =		udp_csum_check,
	.state_transition =	udp_state_transition,
	.state_name =		udp_state_name,
	.register_app =		udp_register_app,
	.unregister_app =	udp_unregister_app,
	.app_conn_bind =	udp_app_conn_bind,
	.debug_packet =		ip_vs_tcpudp_debug_packet,
	.timeout_change =	NULL,
};
Commit	Line	Data
1da177e4 LT	1	/*
	2	* ip_vs_proto_udp.c: UDP load balancing support for IPVS
	3	*
1da177e4 LT	4	* Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
	5	* Julian Anastasov <ja@ssi.bg>
	6	*
	7	* This program is free software; you can redistribute it and/or
	8	* modify it under the terms of the GNU General Public License
	9	* as published by the Free Software Foundation; either version
	10	* 2 of the License, or (at your option) any later version.
	11	*
78b16bde HS	12	* Changes: Hans Schillstrom <hans.schillstrom@ericsson.com>
78b16bde HS	13	* Network name space (netns) aware.
1da177e4 LT	14	*
	15	*/
	16
9aada7ac HE	17	#define KMSG_COMPONENT "IPVS"
	18	#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
	19
14c85021 ACM	20	#include <linux/in.h>
14c85021 ACM	21	#include <linux/ip.h>
1da177e4	22	#include <linux/kernel.h>
af1e1cf0	23	#include <linux/netfilter.h>
1da177e4	24	#include <linux/netfilter_ipv4.h>
14c85021	25	#include <linux/udp.h>
1da177e4 LT	26
1da177e4 LT	27	#include <net/ip_vs.h>
c9bdd4b5	28	#include <net/ip.h>
63f2c046	29	#include <net/ip6_checksum.h>
1da177e4	30
1da177e4	31	static int
d8f44c33 EB	32	udp_conn_schedule(struct netns_ipvs ipvs, int af, struct sk_buff skb,
d8f44c33 EB	33	struct ip_vs_proto_data *pd,
d4383f04 JDB	34	int verdict, struct ip_vs_conn *cpp,
d4383f04 JDB	35	struct ip_vs_iphdr *iph)
1da177e4 LT	36	{
	37	struct ip_vs_service *svc;
	38	struct udphdr _udph, *uh;
2b0f39ef	39	__be16 _ports[2], *ports = NULL;
1da177e4	40
2b0f39ef AG	41	if (likely(!ip_vs_iph_icmp(iph))) {
	42	/* IPv6 fragments, only first fragment will hit this */
	43	uh = skb_header_pointer(skb, iph->len, sizeof(_udph), &_udph);
	44	if (uh)
	45	ports = &uh->source;
	46	} else {
	47	ports = skb_header_pointer(
	48	skb, iph->len, sizeof(_ports), &_ports);
6044eeff AG	49	}
6044eeff AG	50
2b0f39ef	51	if (!ports) {
1da177e4 LT	52	*verdict = NF_DROP;
	53	return 0;
	54	}
2b0f39ef	55
ceec4c38	56	rcu_read_lock();
2b0f39ef	57	if (likely(!ip_vs_iph_inverse(iph)))
0a4fd6ce	58	svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
2b0f39ef AG	59	&iph->daddr, ports[1]);
2b0f39ef AG	60	else
0a4fd6ce	61	svc = ip_vs_service_find(ipvs, af, skb->mark, iph->protocol,
2b0f39ef AG	62	&iph->saddr, ports[0]);
2b0f39ef AG	63
3c2e0505	64	if (svc) {
190ecd27 JA	65	int ignored;
190ecd27 JA	66
d8f44c33	67	if (ip_vs_todrop(ipvs)) {
1da177e4 LT	68	/*
	69	* It seems that we are very loaded.
	70	* We have to drop this packet :(
	71	*/
ceec4c38	72	rcu_read_unlock();
1da177e4 LT	73	*verdict = NF_DROP;
	74	return 0;
	75	}
	76
	77	/*
	78	* Let the virtual server select a real server for the
	79	* incoming connection, and create a connection entry.
	80	*/
d4383f04	81	*cpp = ip_vs_schedule(svc, skb, pd, &ignored, iph);
a5959d53 HS	82	if (!*cpp && ignored <= 0) {
a5959d53 HS	83	if (!ignored)
d4383f04	84	*verdict = ip_vs_leave(svc, skb, pd, iph);
ceec4c38	85	else
a5959d53	86	*verdict = NF_DROP;
ceec4c38	87	rcu_read_unlock();
1da177e4 LT	88	return 0;
1da177e4 LT	89	}
1da177e4	90	}
ceec4c38	91	rcu_read_unlock();
a5959d53	92	/* NF_ACCEPT */
1da177e4 LT	93	return 1;
	94	}
	95
	96
	97	static inline void
0bbdd42b JV	98	udp_fast_csum_update(int af, struct udphdr *uhdr,
	99	const union nf_inet_addr *oldip,
	100	const union nf_inet_addr *newip,
014d730d	101	__be16 oldport, __be16 newport)
1da177e4	102	{
0bbdd42b JV	103	#ifdef CONFIG_IP_VS_IPV6
	104	if (af == AF_INET6)
	105	uhdr->check =
	106	csum_fold(ip_vs_check_diff16(oldip->ip6, newip->ip6,
	107	ip_vs_check_diff2(oldport, newport,
	108	~csum_unfold(uhdr->check))));
	109	else
	110	#endif
	111	uhdr->check =
	112	csum_fold(ip_vs_check_diff4(oldip->ip, newip->ip,
	113	ip_vs_check_diff2(oldport, newport,
	114	~csum_unfold(uhdr->check))));
1da177e4	115	if (!uhdr->check)
f6ab0288	116	uhdr->check = CSUM_MANGLED_0;
1da177e4 LT	117	}
1da177e4 LT	118
503e81f6 SH	119	static inline void
	120	udp_partial_csum_update(int af, struct udphdr *uhdr,
	121	const union nf_inet_addr *oldip,
	122	const union nf_inet_addr *newip,
	123	__be16 oldlen, __be16 newlen)
	124	{
	125	#ifdef CONFIG_IP_VS_IPV6
	126	if (af == AF_INET6)
	127	uhdr->check =
5bc9068e	128	~csum_fold(ip_vs_check_diff16(oldip->ip6, newip->ip6,
503e81f6	129	ip_vs_check_diff2(oldlen, newlen,
5bc9068e	130	csum_unfold(uhdr->check))));
503e81f6 SH	131	else
	132	#endif
	133	uhdr->check =
5bc9068e	134	~csum_fold(ip_vs_check_diff4(oldip->ip, newip->ip,
503e81f6	135	ip_vs_check_diff2(oldlen, newlen,
5bc9068e	136	csum_unfold(uhdr->check))));
503e81f6 SH	137	}
	138
	139
1da177e4	140	static int
d4383f04 JDB	141	udp_snat_handler(struct sk_buff skb, struct ip_vs_protocol pp,
d4383f04 JDB	142	struct ip_vs_conn cp, struct ip_vs_iphdr iph)
1da177e4 LT	143	{
1da177e4 LT	144	struct udphdr *udph;
d4383f04	145	unsigned int udphoff = iph->len;
503e81f6	146	int oldlen;
8b27b10f	147	int payload_csum = 0;
0bbdd42b JV	148
0bbdd42b JV	149	#ifdef CONFIG_IP_VS_IPV6
d4383f04	150	if (cp->af == AF_INET6 && iph->fragoffs)
63dca2c0	151	return 1;
0bbdd42b	152	#endif
503e81f6	153	oldlen = skb->len - udphoff;
1da177e4 LT	154
1da177e4 LT	155	/* csum_check requires unshared skb */
3db05fea	156	if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
1da177e4 LT	157	return 0;
	158
	159	if (unlikely(cp->app != NULL)) {
8b27b10f JA	160	int ret;
8b27b10f JA	161
1da177e4	162	/* Some checks before mangling */
0bbdd42b	163	if (pp->csum_check && !pp->csum_check(cp->af, skb, pp))
1da177e4 LT	164	return 0;
	165
	166	/*
	167	* Call application helper if needed
	168	*/
8b27b10f	169	if (!(ret = ip_vs_app_pkt_out(cp, skb)))
1da177e4	170	return 0;
8b27b10f JA	171	/* ret=2: csum update is needed after payload mangling */
	172	if (ret == 1)
	173	oldlen = skb->len - udphoff;
	174	else
	175	payload_csum = 1;
1da177e4 LT	176	}
1da177e4 LT	177
0bbdd42b	178	udph = (void *)skb_network_header(skb) + udphoff;
1da177e4 LT	179	udph->source = cp->vport;
	180
	181	/*
	182	* Adjust UDP checksums
	183	*/
503e81f6 SH	184	if (skb->ip_summed == CHECKSUM_PARTIAL) {
503e81f6 SH	185	udp_partial_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
ca62059b HH	186	htons(oldlen),
ca62059b HH	187	htons(skb->len - udphoff));
8b27b10f	188	} else if (!payload_csum && (udph->check != 0)) {
1da177e4	189	/* Only port and addr are changed, do fast csum update */
0bbdd42b	190	udp_fast_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
1da177e4	191	cp->dport, cp->vport);
3db05fea	192	if (skb->ip_summed == CHECKSUM_COMPLETE)
8b27b10f JA	193	skb->ip_summed = (cp->app && pp->csum_check) ?
8b27b10f JA	194	CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
1da177e4 LT	195	} else {
	196	/* full checksum calculation */
	197	udph->check = 0;
3db05fea	198	skb->csum = skb_checksum(skb, udphoff, skb->len - udphoff, 0);
0bbdd42b JV	199	#ifdef CONFIG_IP_VS_IPV6
	200	if (cp->af == AF_INET6)
	201	udph->check = csum_ipv6_magic(&cp->vaddr.in6,
	202	&cp->caddr.in6,
	203	skb->len - udphoff,
	204	cp->protocol, skb->csum);
	205	else
	206	#endif
	207	udph->check = csum_tcpudp_magic(cp->vaddr.ip,
	208	cp->caddr.ip,
	209	skb->len - udphoff,
	210	cp->protocol,
	211	skb->csum);
1da177e4	212	if (udph->check == 0)
f6ab0288	213	udph->check = CSUM_MANGLED_0;
8b27b10f	214	skb->ip_summed = CHECKSUM_UNNECESSARY;
1da177e4 LT	215	IP_VS_DBG(11, "O-pkt: %s O-csum=%d (+%zd)\n",
	216	pp->name, udph->check,
	217	(char)&(udph->check) - (char)udph);
	218	}
	219	return 1;
	220	}
	221
	222
	223	static int
d4383f04 JDB	224	udp_dnat_handler(struct sk_buff skb, struct ip_vs_protocol pp,
d4383f04 JDB	225	struct ip_vs_conn cp, struct ip_vs_iphdr iph)
1da177e4 LT	226	{
1da177e4 LT	227	struct udphdr *udph;
d4383f04	228	unsigned int udphoff = iph->len;
503e81f6	229	int oldlen;
8b27b10f	230	int payload_csum = 0;
0bbdd42b JV	231
0bbdd42b JV	232	#ifdef CONFIG_IP_VS_IPV6
d4383f04	233	if (cp->af == AF_INET6 && iph->fragoffs)
63dca2c0	234	return 1;
0bbdd42b	235	#endif
503e81f6	236	oldlen = skb->len - udphoff;
1da177e4 LT	237
1da177e4 LT	238	/* csum_check requires unshared skb */
3db05fea	239	if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
1da177e4 LT	240	return 0;
	241
	242	if (unlikely(cp->app != NULL)) {
8b27b10f JA	243	int ret;
8b27b10f JA	244
1da177e4	245	/* Some checks before mangling */
0bbdd42b	246	if (pp->csum_check && !pp->csum_check(cp->af, skb, pp))
1da177e4 LT	247	return 0;
	248
	249	/*
	250	* Attempt ip_vs_app call.
	251	* It will fix ip_vs_conn
	252	*/
8b27b10f	253	if (!(ret = ip_vs_app_pkt_in(cp, skb)))
1da177e4	254	return 0;
8b27b10f JA	255	/* ret=2: csum update is needed after payload mangling */
	256	if (ret == 1)
	257	oldlen = skb->len - udphoff;
	258	else
	259	payload_csum = 1;
1da177e4 LT	260	}
1da177e4 LT	261
0bbdd42b	262	udph = (void *)skb_network_header(skb) + udphoff;
1da177e4 LT	263	udph->dest = cp->dport;
	264
	265	/*
	266	* Adjust UDP checksums
	267	*/
503e81f6	268	if (skb->ip_summed == CHECKSUM_PARTIAL) {
5bc9068e	269	udp_partial_csum_update(cp->af, udph, &cp->vaddr, &cp->daddr,
ca62059b HH	270	htons(oldlen),
ca62059b HH	271	htons(skb->len - udphoff));
8b27b10f	272	} else if (!payload_csum && (udph->check != 0)) {
1da177e4	273	/* Only port and addr are changed, do fast csum update */
0bbdd42b	274	udp_fast_csum_update(cp->af, udph, &cp->vaddr, &cp->daddr,
1da177e4	275	cp->vport, cp->dport);
3db05fea	276	if (skb->ip_summed == CHECKSUM_COMPLETE)
8b27b10f JA	277	skb->ip_summed = (cp->app && pp->csum_check) ?
8b27b10f JA	278	CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
1da177e4 LT	279	} else {
	280	/* full checksum calculation */
	281	udph->check = 0;
3db05fea	282	skb->csum = skb_checksum(skb, udphoff, skb->len - udphoff, 0);
0bbdd42b JV	283	#ifdef CONFIG_IP_VS_IPV6
	284	if (cp->af == AF_INET6)
	285	udph->check = csum_ipv6_magic(&cp->caddr.in6,
	286	&cp->daddr.in6,
	287	skb->len - udphoff,
	288	cp->protocol, skb->csum);
	289	else
	290	#endif
	291	udph->check = csum_tcpudp_magic(cp->caddr.ip,
	292	cp->daddr.ip,
	293	skb->len - udphoff,
	294	cp->protocol,
	295	skb->csum);
1da177e4	296	if (udph->check == 0)
f6ab0288	297	udph->check = CSUM_MANGLED_0;
3db05fea	298	skb->ip_summed = CHECKSUM_UNNECESSARY;
1da177e4 LT	299	}
	300	return 1;
	301	}
	302
	303
	304	static int
51ef348b	305	udp_csum_check(int af, struct sk_buff skb, struct ip_vs_protocol pp)
1da177e4 LT	306	{
1da177e4 LT	307	struct udphdr _udph, *uh;
51ef348b JV	308	unsigned int udphoff;
	309
	310	#ifdef CONFIG_IP_VS_IPV6
	311	if (af == AF_INET6)
	312	udphoff = sizeof(struct ipv6hdr);
	313	else
	314	#endif
	315	udphoff = ip_hdrlen(skb);
1da177e4 LT	316
	317	uh = skb_header_pointer(skb, udphoff, sizeof(_udph), &_udph);
	318	if (uh == NULL)
	319	return 0;
	320
	321	if (uh->check != 0) {
	322	switch (skb->ip_summed) {
	323	case CHECKSUM_NONE:
	324	skb->csum = skb_checksum(skb, udphoff,
	325	skb->len - udphoff, 0);
84fa7933	326	case CHECKSUM_COMPLETE:
51ef348b JV	327	#ifdef CONFIG_IP_VS_IPV6
	328	if (af == AF_INET6) {
	329	if (csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
	330	&ipv6_hdr(skb)->daddr,
	331	skb->len - udphoff,
	332	ipv6_hdr(skb)->nexthdr,
	333	skb->csum)) {
0d79641a	334	IP_VS_DBG_RL_PKT(0, af, pp, skb, 0,
51ef348b JV	335	"Failed checksum for");
	336	return 0;
	337	}
	338	} else
	339	#endif
	340	if (csum_tcpudp_magic(ip_hdr(skb)->saddr,
	341	ip_hdr(skb)->daddr,
	342	skb->len - udphoff,
	343	ip_hdr(skb)->protocol,
	344	skb->csum)) {
0d79641a	345	IP_VS_DBG_RL_PKT(0, af, pp, skb, 0,
51ef348b JV	346	"Failed checksum for");
	347	return 0;
	348	}
1da177e4 LT	349	break;
1da177e4 LT	350	default:
84fa7933	351	/* No need to checksum. */
1da177e4 LT	352	break;
	353	}
	354	}
	355	return 1;
	356	}
	357
75e7ce66	358	static inline __u16 udp_app_hashkey(__be16 port)
1da177e4	359	{
75e7ce66 AV	360	return (((__force u16)port >> UDP_APP_TAB_BITS) ^ (__force u16)port)
75e7ce66 AV	361	& UDP_APP_TAB_MASK;
1da177e4 LT	362	}
	363
	364
19648918	365	static int udp_register_app(struct netns_ipvs ipvs, struct ip_vs_app inc)
1da177e4 LT	366	{
1da177e4 LT	367	struct ip_vs_app *i;
75e7ce66 AV	368	__u16 hash;
75e7ce66 AV	369	__be16 port = inc->port;
1da177e4	370	int ret = 0;
18d6ade6	371	struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
1da177e4 LT	372
	373	hash = udp_app_hashkey(port);
	374
78b16bde	375	list_for_each_entry(i, &ipvs->udp_apps[hash], p_list) {
1da177e4 LT	376	if (i->port == port) {
	377	ret = -EEXIST;
	378	goto out;
	379	}
	380	}
363c97d7	381	list_add_rcu(&inc->p_list, &ipvs->udp_apps[hash]);
9bbac6a9	382	atomic_inc(&pd->appcnt);
1da177e4 LT	383
1da177e4 LT	384	out:
1da177e4 LT	385	return ret;
	386	}
	387
	388
	389	static void
19648918	390	udp_unregister_app(struct netns_ipvs ipvs, struct ip_vs_app inc)
1da177e4	391	{
19648918	392	struct ip_vs_proto_data *pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
78b16bde	393
9bbac6a9	394	atomic_dec(&pd->appcnt);
363c97d7	395	list_del_rcu(&inc->p_list);
1da177e4 LT	396	}
	397
	398
	399	static int udp_app_conn_bind(struct ip_vs_conn *cp)
	400	{
58dbc6f2	401	struct netns_ipvs *ipvs = cp->ipvs;
1da177e4 LT	402	int hash;
	403	struct ip_vs_app *inc;
	404	int result = 0;
	405
	406	/* Default binding: bind app only for NAT */
	407	if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)
	408	return 0;
	409
	410	/* Lookup application incarnations and bind the right one */
	411	hash = udp_app_hashkey(cp->vport);
	412
363c97d7 JA	413	rcu_read_lock();
363c97d7 JA	414	list_for_each_entry_rcu(inc, &ipvs->udp_apps[hash], p_list) {
1da177e4 LT	415	if (inc->port == cp->vport) {
	416	if (unlikely(!ip_vs_app_inc_get(inc)))
	417	break;
363c97d7	418	rcu_read_unlock();
1da177e4	419
1e3e238e	420	IP_VS_DBG_BUF(9, "%s(): Binding conn %s:%u->"
cfc78c5a JV	421	"%s:%u to app %s on port %u\n",
	422	__func__,
	423	IP_VS_DBG_ADDR(cp->af, &cp->caddr),
	424	ntohs(cp->cport),
	425	IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
	426	ntohs(cp->vport),
	427	inc->name, ntohs(inc->port));
	428
1da177e4 LT	429	cp->app = inc;
	430	if (inc->init_conn)
	431	result = inc->init_conn(inc, cp);
	432	goto out;
	433	}
	434	}
363c97d7	435	rcu_read_unlock();
1da177e4 LT	436
	437	out:
	438	return result;
	439	}
	440
	441
78b16bde	442	static const int udp_timeouts[IP_VS_UDP_S_LAST+1] = {
1da177e4 LT	443	[IP_VS_UDP_S_NORMAL] = 560HZ,
	444	[IP_VS_UDP_S_LAST] = 2*HZ,
	445	};
	446
36cbd3dc	447	static const char *const udp_state_name_table[IP_VS_UDP_S_LAST+1] = {
1da177e4 LT	448	[IP_VS_UDP_S_NORMAL] = "UDP",
	449	[IP_VS_UDP_S_LAST] = "BUG!",
	450	};
	451
1da177e4 LT	452	static const char * udp_state_name(int state)
	453	{
	454	if (state >= IP_VS_UDP_S_LAST)
	455	return "ERR!";
	456	return udp_state_name_table[state] ? udp_state_name_table[state] : "?";
	457	}
	458
4a516f11	459	static void
1da177e4 LT	460	udp_state_transition(struct ip_vs_conn *cp, int direction,
1da177e4 LT	461	const struct sk_buff *skb,
9330419d	462	struct ip_vs_proto_data *pd)
1da177e4	463	{
78b16bde HS	464	if (unlikely(!pd)) {
78b16bde HS	465	pr_err("UDP no ns data\n");
4a516f11	466	return;
78b16bde HS	467	}
	468
	469	cp->timeout = pd->timeout_table[IP_VS_UDP_S_NORMAL];
1da177e4 LT	470	}
1da177e4 LT	471
1281a9c2	472	static int __udp_init(struct netns_ipvs ipvs, struct ip_vs_proto_data pd)
1da177e4	473	{
78b16bde	474	ip_vs_init_hash_table(ipvs->udp_apps, UDP_APP_TAB_SIZE);
78b16bde HS	475	pd->timeout_table = ip_vs_create_timeout_table((int *)udp_timeouts,
78b16bde HS	476	sizeof(udp_timeouts));
582b8e3e HS	477	if (!pd->timeout_table)
	478	return -ENOMEM;
	479	return 0;
1da177e4 LT	480	}
1da177e4 LT	481
1281a9c2	482	static void __udp_exit(struct netns_ipvs ipvs, struct ip_vs_proto_data pd)
1da177e4	483	{
78b16bde	484	kfree(pd->timeout_table);
1da177e4 LT	485	}
	486
	487
	488	struct ip_vs_protocol ip_vs_protocol_udp = {
	489	.name = "UDP",
	490	.protocol = IPPROTO_UDP,
2ad17def	491	.num_states = IP_VS_UDP_S_LAST,
1da177e4	492	.dont_defrag = 0,
78b16bde HS	493	.init = NULL,
	494	.exit = NULL,
	495	.init_netns = __udp_init,
	496	.exit_netns = __udp_exit,
1da177e4	497	.conn_schedule = udp_conn_schedule,
5c0d2374 SH	498	.conn_in_get = ip_vs_conn_in_get_proto,
5c0d2374 SH	499	.conn_out_get = ip_vs_conn_out_get_proto,
1da177e4 LT	500	.snat_handler = udp_snat_handler,
	501	.dnat_handler = udp_dnat_handler,
	502	.csum_check = udp_csum_check,
	503	.state_transition = udp_state_transition,
	504	.state_name = udp_state_name,
	505	.register_app = udp_register_app,
	506	.unregister_app = udp_unregister_app,
	507	.app_conn_bind = udp_app_conn_bind,
	508	.debug_packet = ip_vs_tcpudp_debug_packet,
	509	.timeout_change = NULL,
1da177e4	510	};