[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 u32 cookie_hash(__be32 saddr, __be32 daddr, __be16 sport, __be16 dport,
		       u32 count, int c)
{
	__u32 tmp[16 + 5 + SHA_WORKSPACE_WORDS];

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