[net-next-2.6.git] / net / ipv4 / syncookies.c

/*
 *  Syncookies implementation for the Linux kernel
 *
 *  Copyright (C) 1997 Andi Kleen
 *  Based on ideas by D.J.Bernstein and Eric Schenk.
 *
 *	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.
 *
 *  $Id: syncookies.c,v 1.18 2002/02/01 22:01:04 davem Exp $
 *
 *  Missing: IPv6 support.
 */

#include <linux/tcp.h>
#include <linux/slab.h>
#include <linux/random.h>
#include <linux/cryptohash.h>
#include <linux/kernel.h>
#include <net/tcp.h>

extern int sysctl_tcp_syncookies;

static __u32 syncookie_secret[2][16-3+SHA_DIGEST_WORDS];

static __init int init_syncookies(void)
{
	get_random_bytes(syncookie_secret, sizeof(syncookie_secret));
	return 0;
}
module_init(init_syncookies);

#define COOKIEBITS 24	/* Upper bits store count */
#define COOKIEMASK (((__u32)1 << COOKIEBITS) - 1)

static DEFINE_PER_CPU(__u32, cookie_scratch)[16 + 5 + SHA_WORKSPACE_WORDS];

static u32 cookie_hash(__be32 saddr, __be32 daddr, __be16 sport, __be16 dport,
		       u32 count, int c)
{
	__u32 *tmp = __get_cpu_var(cookie_scratch);

	memcpy(tmp + 3, syncookie_secret[c], sizeof(syncookie_secret[c]));
	tmp[0] = (__force u32)saddr;
	tmp[1] = (__force u32)daddr;
	tmp[2] = ((__force u32)sport << 16) + (__force u32)dport;
	tmp[3] = count;
	sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);

	return tmp[17];
}

static __u32 secure_tcp_syn_cookie(__be32 saddr, __be32 daddr, __be16 sport,
				   __be16 dport, __u32 sseq, __u32 count,
				   __u32 data)
{
	/*
	 * Compute the secure sequence number.
	 * The output should be:
	 *   HASH(sec1,saddr,sport,daddr,dport,sec1) + sseq + (count * 2^24)
	 *      + (HASH(sec2,saddr,sport,daddr,dport,count,sec2) % 2^24).
	 * Where sseq is their sequence number and count increases every
	 * minute by 1.
	 * As an extra hack, we add a small "data" value that encodes the
	 * MSS into the second hash value.
	 */

	return (cookie_hash(saddr, daddr, sport, dport, 0, 0) +
		sseq + (count << COOKIEBITS) +
		((cookie_hash(saddr, daddr, sport, dport, count, 1) + data)
		 & COOKIEMASK));
}

/*
 * This retrieves the small "data" value from the syncookie.
 * If the syncookie is bad, the data returned will be out of
 * range.  This must be checked by the caller.
 *
 * The count value used to generate the cookie must be within
 * "maxdiff" if the current (passed-in) "count".  The return value
 * is (__u32)-1 if this test fails.
 */
static __u32 check_tcp_syn_cookie(__u32 cookie, __be32 saddr, __be32 daddr,
				  __be16 sport, __be16 dport, __u32 sseq,
				  __u32 count, __u32 maxdiff)
{
	__u32 diff;

	/* Strip away the layers from the cookie */
	cookie -= cookie_hash(saddr, daddr, sport, dport, 0, 0) + sseq;

	/* Cookie is now reduced to (count * 2^24) ^ (hash % 2^24) */
	diff = (count - (cookie >> COOKIEBITS)) & ((__u32) - 1 >> COOKIEBITS);
	if (diff >= maxdiff)
		return (__u32)-1;

	return (cookie -
		cookie_hash(saddr, daddr, sport, dport, count - diff, 1))
		& COOKIEMASK;	/* Leaving the data behind */
}

/*
 * This table has to be sorted and terminated with (__u16)-1.
 * XXX generate a better table.
 * Unresolved Issues: HIPPI with a 64k MSS is not well supported.
 */
static __u16 const msstab[] = {
	64 - 1,
	256 - 1,
	512 - 1,
	536 - 1,
	1024 - 1,
	1440 - 1,
	1460 - 1,
	4312 - 1,
	(__u16)-1
};
/* The number doesn't include the -1 terminator */
#define NUM_MSS (ARRAY_SIZE(msstab) - 1)

/*
 * Generate a syncookie.  mssp points to the mss, which is returned
 * rounded down to the value encoded in the cookie.
 */
__u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, __u16 *mssp)
{
	struct tcp_sock *tp = tcp_sk(sk);
	const struct iphdr *iph = ip_hdr(skb);
	const struct tcphdr *th = tcp_hdr(skb);
	int mssind;
	const __u16 mss = *mssp;

	tp->last_synq_overflow = jiffies;

	/* XXX sort msstab[] by probability?  Binary search? */
	for (mssind = 0; mss > msstab[mssind + 1]; mssind++)
		;
	*mssp = msstab[mssind] + 1;

	NET_INC_STATS_BH(LINUX_MIB_SYNCOOKIESSENT);

	return secure_tcp_syn_cookie(iph->saddr, iph->daddr,
				     th->source, th->dest, ntohl(th->seq),
				     jiffies / (HZ * 60), mssind);
}

/*
 * This (misnamed) value is the age of syncookie which is permitted.
 * Its ideal value should be dependent on TCP_TIMEOUT_INIT and
 * sysctl_tcp_retries1. It's a rather complicated formula (exponential
 * backoff) to compute at runtime so it's currently hardcoded here.
 */
#define COUNTER_TRIES 4
/*
 * Check if a ack sequence number is a valid syncookie.
 * Return the decoded mss if it is, or 0 if not.
 */
static inline int cookie_check(struct sk_buff *skb, __u32 cookie)
{
	const struct iphdr *iph = ip_hdr(skb);
	const struct tcphdr *th = tcp_hdr(skb);
	__u32 seq = ntohl(th->seq) - 1;
	__u32 mssind = check_tcp_syn_cookie(cookie, iph->saddr, iph->daddr,
					    th->source, th->dest, seq,
					    jiffies / (HZ * 60),
					    COUNTER_TRIES);

	return mssind < NUM_MSS ? msstab[mssind] + 1 : 0;
}

static inline struct sock *get_cookie_sock(struct sock *sk, struct sk_buff *skb,
					   struct request_sock *req,
					   struct dst_entry *dst)
{
	struct inet_connection_sock *icsk = inet_csk(sk);
	struct sock *child;

	child = icsk->icsk_af_ops->syn_recv_sock(sk, skb, req, dst);
	if (child)
		inet_csk_reqsk_queue_add(sk, req, child);
	else
		reqsk_free(req);

	return child;
}

struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
			     struct ip_options *opt)
{
	struct inet_request_sock *ireq;
	struct tcp_request_sock *treq;
	struct tcp_sock *tp = tcp_sk(sk);
	const struct tcphdr *th = tcp_hdr(skb);
	__u32 cookie = ntohl(th->ack_seq) - 1;
	struct sock *ret = sk;
	struct request_sock *req;
	int mss;
	struct rtable *rt;
	__u8 rcv_wscale;

	if (!sysctl_tcp_syncookies || !th->ack)
		goto out;

	if (time_after(jiffies, tp->last_synq_overflow + TCP_TIMEOUT_INIT) ||
	    (mss = cookie_check(skb, cookie)) == 0) {
		NET_INC_STATS_BH(LINUX_MIB_SYNCOOKIESFAILED);
		goto out;
	}

	NET_INC_STATS_BH(LINUX_MIB_SYNCOOKIESRECV);

	ret = NULL;
	req = reqsk_alloc(&tcp_request_sock_ops); /* for safety */
	if (!req)
		goto out;

	if (security_inet_conn_request(sk, skb, req)) {
		reqsk_free(req);
		goto out;
	}
	ireq = inet_rsk(req);
	treq = tcp_rsk(req);
	treq->rcv_isn		= ntohl(th->seq) - 1;
	treq->snt_isn		= cookie;
	req->mss		= mss;
	ireq->rmt_port		= th->source;
	ireq->loc_addr		= ip_hdr(skb)->daddr;
	ireq->rmt_addr		= ip_hdr(skb)->saddr;
	ireq->opt		= NULL;

	/* We throwed the options of the initial SYN away, so we hope
	 * the ACK carries the same options again (see RFC1122 4.2.3.8)
	 */
	if (opt && opt->optlen) {
		int opt_size = sizeof(struct ip_options) + opt->optlen;

		ireq->opt = kmalloc(opt_size, GFP_ATOMIC);
		if (ireq->opt != NULL && ip_options_echo(ireq->opt, skb)) {
			kfree(ireq->opt);
			ireq->opt = NULL;
		}
	}

	ireq->snd_wscale = ireq->rcv_wscale = ireq->tstamp_ok = 0;
	ireq->wscale_ok	 = ireq->sack_ok = 0;
	req->expires	= 0UL;
	req->retrans	= 0;

	/*
	 * We need to lookup the route here to get at the correct
	 * window size. We should better make sure that the window size
	 * hasn't changed since we received the original syn, but I see
	 * no easy way to do this.
	 */
	{
		struct flowi fl = { .nl_u = { .ip4_u =
					      { .daddr = ((opt && opt->srr) ?
							  opt->faddr :
							  ireq->rmt_addr),
						.saddr = ireq->loc_addr,
						.tos = RT_CONN_FLAGS(sk) } },
				    .proto = IPPROTO_TCP,
				    .uli_u = { .ports =
					       { .sport = th->dest,
						 .dport = th->source } } };
		security_req_classify_flow(req, &fl);
		if (ip_route_output_key(&init_net, &rt, &fl)) {
			reqsk_free(req);
			goto out;
		}
	}

	/* Try to redo what tcp_v4_send_synack did. */
	req->window_clamp = dst_metric(&rt->u.dst, RTAX_WINDOW);
	tcp_select_initial_window(tcp_full_space(sk), req->mss,
				  &req->rcv_wnd, &req->window_clamp,
				  0, &rcv_wscale);
	/* BTW win scale with syncookies is 0 by definition */
	ireq->rcv_wscale  = rcv_wscale;

	ret = get_cookie_sock(sk, skb, req, &rt->u.dst);
out:	return ret;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* Syncookies implementation for the Linux kernel
	3	*
	4	* Copyright (C) 1997 Andi Kleen
e905a9ed	5	* Based on ideas by D.J.Bernstein and Eric Schenk.
1da177e4 LT	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.
e905a9ed	11	*
1da177e4 LT	12	* $Id: syncookies.c,v 1.18 2002/02/01 22:01:04 davem Exp $
1da177e4 LT	13	*
e905a9ed	14	* Missing: IPv6 support.
1da177e4 LT	15	*/
	16
	17	#include <linux/tcp.h>
	18	#include <linux/slab.h>
	19	#include <linux/random.h>
	20	#include <linux/cryptohash.h>
	21	#include <linux/kernel.h>
	22	#include <net/tcp.h>
	23
	24	extern int sysctl_tcp_syncookies;
	25
	26	static __u32 syncookie_secret[2][16-3+SHA_DIGEST_WORDS];
	27
	28	static __init int init_syncookies(void)
	29	{
	30	get_random_bytes(syncookie_secret, sizeof(syncookie_secret));
	31	return 0;
	32	}
	33	module_init(init_syncookies);
	34
	35	#define COOKIEBITS 24 /* Upper bits store count */
	36	#define COOKIEMASK (((__u32)1 << COOKIEBITS) - 1)
	37
11baab7a ED	38	static DEFINE_PER_CPU(__u32, cookie_scratch)[16 + 5 + SHA_WORKSPACE_WORDS];
11baab7a ED	39
714e85be	40	static u32 cookie_hash(__be32 saddr, __be32 daddr, __be16 sport, __be16 dport,
1da177e4 LT	41	u32 count, int c)
1da177e4 LT	42	{
11baab7a	43	__u32 *tmp = __get_cpu_var(cookie_scratch);
1da177e4 LT	44
1da177e4 LT	45	memcpy(tmp + 3, syncookie_secret[c], sizeof(syncookie_secret[c]));
714e85be AV	46	tmp[0] = (__force u32)saddr;
	47	tmp[1] = (__force u32)daddr;
	48	tmp[2] = ((__force u32)sport << 16) + (__force u32)dport;
1da177e4 LT	49	tmp[3] = count;
	50	sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
	51
	52	return tmp[17];
	53	}
	54
714e85be AV	55	static __u32 secure_tcp_syn_cookie(__be32 saddr, __be32 daddr, __be16 sport,
714e85be AV	56	__be16 dport, __u32 sseq, __u32 count,
1da177e4 LT	57	__u32 data)
	58	{
	59	/*
	60	* Compute the secure sequence number.
	61	* The output should be:
e905a9ed	62	* HASH(sec1,saddr,sport,daddr,dport,sec1) + sseq + (count * 2^24)
1da177e4 LT	63	* + (HASH(sec2,saddr,sport,daddr,dport,count,sec2) % 2^24).
	64	* Where sseq is their sequence number and count increases every
	65	* minute by 1.
	66	* As an extra hack, we add a small "data" value that encodes the
	67	* MSS into the second hash value.
	68	*/
	69
	70	return (cookie_hash(saddr, daddr, sport, dport, 0, 0) +
	71	sseq + (count << COOKIEBITS) +
	72	((cookie_hash(saddr, daddr, sport, dport, count, 1) + data)
	73	& COOKIEMASK));
	74	}
	75
	76	/*
	77	* This retrieves the small "data" value from the syncookie.
	78	* If the syncookie is bad, the data returned will be out of
	79	* range. This must be checked by the caller.
	80	*
	81	* The count value used to generate the cookie must be within
	82	* "maxdiff" if the current (passed-in) "count". The return value
	83	* is (__u32)-1 if this test fails.
	84	*/
714e85be AV	85	static __u32 check_tcp_syn_cookie(__u32 cookie, __be32 saddr, __be32 daddr,
714e85be AV	86	__be16 sport, __be16 dport, __u32 sseq,
1da177e4 LT	87	__u32 count, __u32 maxdiff)
	88	{
	89	__u32 diff;
	90
	91	/* Strip away the layers from the cookie */
	92	cookie -= cookie_hash(saddr, daddr, sport, dport, 0, 0) + sseq;
	93
	94	/* Cookie is now reduced to (count * 2^24) ^ (hash % 2^24) */
	95	diff = (count - (cookie >> COOKIEBITS)) & ((__u32) - 1 >> COOKIEBITS);
	96	if (diff >= maxdiff)
	97	return (__u32)-1;
	98
	99	return (cookie -
	100	cookie_hash(saddr, daddr, sport, dport, count - diff, 1))
	101	& COOKIEMASK; /* Leaving the data behind */
	102	}
	103
e905a9ed	104	/*
1da177e4 LT	105	* This table has to be sorted and terminated with (__u16)-1.
	106	* XXX generate a better table.
	107	* Unresolved Issues: HIPPI with a 64k MSS is not well supported.
	108	*/
	109	static __u16 const msstab[] = {
	110	64 - 1,
e905a9ed	111	256 - 1,
1da177e4 LT	112	512 - 1,
1da177e4 LT	113	536 - 1,
e905a9ed	114	1024 - 1,
1da177e4 LT	115	1440 - 1,
	116	1460 - 1,
	117	4312 - 1,
	118	(__u16)-1
	119	};
	120	/* The number doesn't include the -1 terminator */
	121	#define NUM_MSS (ARRAY_SIZE(msstab) - 1)
	122
	123	/*
	124	* Generate a syncookie. mssp points to the mss, which is returned
	125	* rounded down to the value encoded in the cookie.
	126	*/
	127	__u32 cookie_v4_init_sequence(struct sock sk, struct sk_buff skb, __u16 *mssp)
	128	{
	129	struct tcp_sock *tp = tcp_sk(sk);
aa8223c7 ACM	130	const struct iphdr *iph = ip_hdr(skb);
aa8223c7 ACM	131	const struct tcphdr *th = tcp_hdr(skb);
1da177e4 LT	132	int mssind;
	133	const __u16 mss = *mssp;
	134
1da177e4 LT	135	tp->last_synq_overflow = jiffies;
	136
	137	/* XXX sort msstab[] by probability? Binary search? */
	138	for (mssind = 0; mss > msstab[mssind + 1]; mssind++)
	139	;
	140	*mssp = msstab[mssind] + 1;
	141
	142	NET_INC_STATS_BH(LINUX_MIB_SYNCOOKIESSENT);
	143
aa8223c7 ACM	144	return secure_tcp_syn_cookie(iph->saddr, iph->daddr,
aa8223c7 ACM	145	th->source, th->dest, ntohl(th->seq),
1da177e4 LT	146	jiffies / (HZ * 60), mssind);
	147	}
	148
e905a9ed	149	/*
1da177e4 LT	150	* This (misnamed) value is the age of syncookie which is permitted.
	151	* Its ideal value should be dependent on TCP_TIMEOUT_INIT and
	152	* sysctl_tcp_retries1. It's a rather complicated formula (exponential
	153	* backoff) to compute at runtime so it's currently hardcoded here.
	154	*/
	155	#define COUNTER_TRIES 4
e905a9ed YH	156	/*
e905a9ed YH	157	* Check if a ack sequence number is a valid syncookie.
1da177e4 LT	158	* Return the decoded mss if it is, or 0 if not.
	159	*/
	160	static inline int cookie_check(struct sk_buff *skb, __u32 cookie)
	161	{
aa8223c7 ACM	162	const struct iphdr *iph = ip_hdr(skb);
	163	const struct tcphdr *th = tcp_hdr(skb);
	164	__u32 seq = ntohl(th->seq) - 1;
	165	__u32 mssind = check_tcp_syn_cookie(cookie, iph->saddr, iph->daddr,
	166	th->source, th->dest, seq,
	167	jiffies / (HZ * 60),
	168	COUNTER_TRIES);
1da177e4 LT	169
	170	return mssind < NUM_MSS ? msstab[mssind] + 1 : 0;
	171	}
	172
1da177e4	173	static inline struct sock get_cookie_sock(struct sock sk, struct sk_buff *skb,
60236fdd	174	struct request_sock *req,
1da177e4 LT	175	struct dst_entry *dst)
1da177e4 LT	176	{
8292a17a	177	struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4 LT	178	struct sock *child;
1da177e4 LT	179
8292a17a	180	child = icsk->icsk_af_ops->syn_recv_sock(sk, skb, req, dst);
1da177e4	181	if (child)
463c84b9	182	inet_csk_reqsk_queue_add(sk, req, child);
1da177e4	183	else
60236fdd	184	reqsk_free(req);
1da177e4 LT	185
	186	return child;
	187	}
	188
	189	struct sock cookie_v4_check(struct sock sk, struct sk_buff *skb,
	190	struct ip_options *opt)
	191	{
2e6599cb ACM	192	struct inet_request_sock *ireq;
2e6599cb ACM	193	struct tcp_request_sock *treq;
1da177e4	194	struct tcp_sock *tp = tcp_sk(sk);
aa8223c7 ACM	195	const struct tcphdr *th = tcp_hdr(skb);
aa8223c7 ACM	196	__u32 cookie = ntohl(th->ack_seq) - 1;
1da177e4	197	struct sock *ret = sk;
e905a9ed YH	198	struct request_sock *req;
	199	int mss;
	200	struct rtable *rt;
1da177e4 LT	201	__u8 rcv_wscale;
1da177e4 LT	202
aa8223c7	203	if (!sysctl_tcp_syncookies \|\| !th->ack)
1da177e4 LT	204	goto out;
1da177e4 LT	205
e905a9ed	206	if (time_after(jiffies, tp->last_synq_overflow + TCP_TIMEOUT_INIT) \|\|
1da177e4	207	(mss = cookie_check(skb, cookie)) == 0) {
e905a9ed	208	NET_INC_STATS_BH(LINUX_MIB_SYNCOOKIESFAILED);
1da177e4 LT	209	goto out;
	210	}
	211
	212	NET_INC_STATS_BH(LINUX_MIB_SYNCOOKIESRECV);
	213
1da177e4	214	ret = NULL;
60236fdd	215	req = reqsk_alloc(&tcp_request_sock_ops); /* for safety */
1da177e4 LT	216	if (!req)
	217	goto out;
	218
4237c75c VY	219	if (security_inet_conn_request(sk, skb, req)) {
	220	reqsk_free(req);
	221	goto out;
	222	}
2e6599cb ACM	223	ireq = inet_rsk(req);
2e6599cb ACM	224	treq = tcp_rsk(req);
aa8223c7	225	treq->rcv_isn = ntohl(th->seq) - 1;
e905a9ed	226	treq->snt_isn = cookie;
1da177e4	227	req->mss = mss;
aa8223c7	228	ireq->rmt_port = th->source;
eddc9ec5 ACM	229	ireq->loc_addr = ip_hdr(skb)->daddr;
eddc9ec5 ACM	230	ireq->rmt_addr = ip_hdr(skb)->saddr;
2e6599cb	231	ireq->opt = NULL;
1da177e4 LT	232
	233	/* We throwed the options of the initial SYN away, so we hope
	234	* the ACK carries the same options again (see RFC1122 4.2.3.8)
	235	*/
	236	if (opt && opt->optlen) {
	237	int opt_size = sizeof(struct ip_options) + opt->optlen;
	238
2e6599cb ACM	239	ireq->opt = kmalloc(opt_size, GFP_ATOMIC);
	240	if (ireq->opt != NULL && ip_options_echo(ireq->opt, skb)) {
	241	kfree(ireq->opt);
	242	ireq->opt = NULL;
1da177e4 LT	243	}
	244	}
	245
2e6599cb	246	ireq->snd_wscale = ireq->rcv_wscale = ireq->tstamp_ok = 0;
e905a9ed YH	247	ireq->wscale_ok = ireq->sack_ok = 0;
	248	req->expires = 0UL;
	249	req->retrans = 0;
	250
1da177e4 LT	251	/*
	252	* We need to lookup the route here to get at the correct
	253	* window size. We should better make sure that the window size
	254	* hasn't changed since we received the original syn, but I see
e905a9ed	255	* no easy way to do this.
1da177e4 LT	256	*/
	257	{
	258	struct flowi fl = { .nl_u = { .ip4_u =
	259	{ .daddr = ((opt && opt->srr) ?
	260	opt->faddr :
2e6599cb ACM	261	ireq->rmt_addr),
2e6599cb ACM	262	.saddr = ireq->loc_addr,
1da177e4 LT	263	.tos = RT_CONN_FLAGS(sk) } },
	264	.proto = IPPROTO_TCP,
	265	.uli_u = { .ports =
aa8223c7 ACM	266	{ .sport = th->dest,
aa8223c7 ACM	267	.dport = th->source } } };
4237c75c	268	security_req_classify_flow(req, &fl);
f206351a	269	if (ip_route_output_key(&init_net, &rt, &fl)) {
60236fdd	270	reqsk_free(req);
e905a9ed	271	goto out;
1da177e4 LT	272	}
	273	}
	274
	275	/* Try to redo what tcp_v4_send_synack did. */
	276	req->window_clamp = dst_metric(&rt->u.dst, RTAX_WINDOW);
	277	tcp_select_initial_window(tcp_full_space(sk), req->mss,
e905a9ed	278	&req->rcv_wnd, &req->window_clamp,
1da177e4 LT	279	0, &rcv_wscale);
1da177e4 LT	280	/* BTW win scale with syncookies is 0 by definition */
e905a9ed	281	ireq->rcv_wscale = rcv_wscale;
1da177e4 LT	282
	283	ret = get_cookie_sock(sk, skb, req, &rt->u.dst);
	284	out: return ret;
	285	}