Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Syncookies implementation for the Linux kernel |
| 3 | * |
| 4 | * Copyright (C) 1997 Andi Kleen |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 5 | * Based on ideas by D.J.Bernstein and Eric Schenk. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 11 | */ |
| 12 | |
| 13 | #include <linux/tcp.h> |
| 14 | #include <linux/slab.h> |
| 15 | #include <linux/random.h> |
| 16 | #include <linux/cryptohash.h> |
| 17 | #include <linux/kernel.h> |
Paul Gortmaker | bc3b2d7 | 2011-07-15 11:47:34 -0400 | [diff] [blame] | 18 | #include <linux/export.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 19 | #include <net/tcp.h> |
KOVACS Krisztian | 86b08d8 | 2008-10-01 07:44:42 -0700 | [diff] [blame] | 20 | #include <net/route.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 21 | |
| 22 | extern int sysctl_tcp_syncookies; |
| 23 | |
Florian Westphal | 253ff51 | 2014-08-26 12:55:53 +0200 | [diff] [blame] | 24 | static u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS] __read_mostly; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 25 | |
| 26 | #define COOKIEBITS 24 /* Upper bits store count */ |
| 27 | #define COOKIEMASK (((__u32)1 << COOKIEBITS) - 1) |
| 28 | |
Florian Westphal | 274e2da | 2014-11-03 17:35:01 +0100 | [diff] [blame] | 29 | /* TCP Timestamp: 6 lowest bits of timestamp sent in the cookie SYN-ACK |
| 30 | * stores TCP options: |
| 31 | * |
| 32 | * MSB LSB |
| 33 | * | 31 ... 6 | 5 | 4 | 3 2 1 0 | |
| 34 | * | Timestamp | ECN | SACK | WScale | |
| 35 | * |
| 36 | * When we receive a valid cookie-ACK, we look at the echoed tsval (if |
| 37 | * any) to figure out which TCP options we should use for the rebuilt |
| 38 | * connection. |
| 39 | * |
| 40 | * A WScale setting of '0xf' (which is an invalid scaling value) |
| 41 | * means that original syn did not include the TCP window scaling option. |
| 42 | */ |
| 43 | #define TS_OPT_WSCALE_MASK 0xf |
| 44 | #define TS_OPT_SACK BIT(4) |
| 45 | #define TS_OPT_ECN BIT(5) |
| 46 | /* There is no TS_OPT_TIMESTAMP: |
| 47 | * if ACK contains timestamp option, we already know it was |
| 48 | * requested/supported by the syn/synack exchange. |
| 49 | */ |
| 50 | #define TSBITS 6 |
| 51 | #define TSMASK (((__u32)1 << TSBITS) - 1) |
| 52 | |
Tejun Heo | 245b2e7 | 2009-06-24 15:13:48 +0900 | [diff] [blame] | 53 | static DEFINE_PER_CPU(__u32 [16 + 5 + SHA_WORKSPACE_WORDS], |
| 54 | ipv4_cookie_scratch); |
Eric Dumazet | 11baab7 | 2008-02-07 10:40:19 +0100 | [diff] [blame] | 55 | |
Al Viro | 714e85b | 2006-11-14 20:51:49 -0800 | [diff] [blame] | 56 | static u32 cookie_hash(__be32 saddr, __be32 daddr, __be16 sport, __be16 dport, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 57 | u32 count, int c) |
| 58 | { |
Hannes Frederic Sowa | b23a002 | 2013-10-19 21:48:56 +0200 | [diff] [blame] | 59 | __u32 *tmp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 60 | |
Hannes Frederic Sowa | b23a002 | 2013-10-19 21:48:56 +0200 | [diff] [blame] | 61 | net_get_random_once(syncookie_secret, sizeof(syncookie_secret)); |
| 62 | |
Christoph Lameter | 903ceff | 2014-08-17 12:30:35 -0500 | [diff] [blame] | 63 | tmp = this_cpu_ptr(ipv4_cookie_scratch); |
Florian Westphal | 2051f11 | 2008-03-23 22:21:28 -0700 | [diff] [blame] | 64 | memcpy(tmp + 4, syncookie_secret[c], sizeof(syncookie_secret[c])); |
Al Viro | 714e85b | 2006-11-14 20:51:49 -0800 | [diff] [blame] | 65 | tmp[0] = (__force u32)saddr; |
| 66 | tmp[1] = (__force u32)daddr; |
| 67 | tmp[2] = ((__force u32)sport << 16) + (__force u32)dport; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 68 | tmp[3] = count; |
| 69 | sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5); |
| 70 | |
| 71 | return tmp[17]; |
| 72 | } |
| 73 | |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 74 | |
| 75 | /* |
| 76 | * when syncookies are in effect and tcp timestamps are enabled we encode |
Florian Westphal | 734f614 | 2010-06-21 11:48:44 +0000 | [diff] [blame] | 77 | * tcp options in the lower bits of the timestamp value that will be |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 78 | * sent in the syn-ack. |
| 79 | * Since subsequent timestamps use the normal tcp_time_stamp value, we |
| 80 | * must make sure that the resulting initial timestamp is <= tcp_time_stamp. |
| 81 | */ |
| 82 | __u32 cookie_init_timestamp(struct request_sock *req) |
| 83 | { |
| 84 | struct inet_request_sock *ireq; |
| 85 | u32 ts, ts_now = tcp_time_stamp; |
| 86 | u32 options = 0; |
| 87 | |
| 88 | ireq = inet_rsk(req); |
Florian Westphal | 734f614 | 2010-06-21 11:48:44 +0000 | [diff] [blame] | 89 | |
Florian Westphal | 274e2da | 2014-11-03 17:35:01 +0100 | [diff] [blame] | 90 | options = ireq->wscale_ok ? ireq->snd_wscale : TS_OPT_WSCALE_MASK; |
| 91 | if (ireq->sack_ok) |
| 92 | options |= TS_OPT_SACK; |
| 93 | if (ireq->ecn_ok) |
| 94 | options |= TS_OPT_ECN; |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 95 | |
| 96 | ts = ts_now & ~TSMASK; |
| 97 | ts |= options; |
| 98 | if (ts > ts_now) { |
| 99 | ts >>= TSBITS; |
| 100 | ts--; |
| 101 | ts <<= TSBITS; |
| 102 | ts |= options; |
| 103 | } |
| 104 | return ts; |
| 105 | } |
| 106 | |
| 107 | |
Al Viro | 714e85b | 2006-11-14 20:51:49 -0800 | [diff] [blame] | 108 | static __u32 secure_tcp_syn_cookie(__be32 saddr, __be32 daddr, __be16 sport, |
Florian Westphal | 8c27bd7 | 2013-09-20 22:32:55 +0200 | [diff] [blame] | 109 | __be16 dport, __u32 sseq, __u32 data) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 110 | { |
| 111 | /* |
| 112 | * Compute the secure sequence number. |
| 113 | * The output should be: |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 114 | * HASH(sec1,saddr,sport,daddr,dport,sec1) + sseq + (count * 2^24) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 115 | * + (HASH(sec2,saddr,sport,daddr,dport,count,sec2) % 2^24). |
| 116 | * Where sseq is their sequence number and count increases every |
| 117 | * minute by 1. |
| 118 | * As an extra hack, we add a small "data" value that encodes the |
| 119 | * MSS into the second hash value. |
| 120 | */ |
Florian Westphal | 8c27bd7 | 2013-09-20 22:32:55 +0200 | [diff] [blame] | 121 | u32 count = tcp_cookie_time(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 122 | return (cookie_hash(saddr, daddr, sport, dport, 0, 0) + |
| 123 | sseq + (count << COOKIEBITS) + |
| 124 | ((cookie_hash(saddr, daddr, sport, dport, count, 1) + data) |
| 125 | & COOKIEMASK)); |
| 126 | } |
| 127 | |
| 128 | /* |
| 129 | * This retrieves the small "data" value from the syncookie. |
| 130 | * If the syncookie is bad, the data returned will be out of |
| 131 | * range. This must be checked by the caller. |
| 132 | * |
Florian Westphal | 8c27bd7 | 2013-09-20 22:32:55 +0200 | [diff] [blame] | 133 | * The count value used to generate the cookie must be less than |
| 134 | * MAX_SYNCOOKIE_AGE minutes in the past. |
| 135 | * The return value (__u32)-1 if this test fails. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 136 | */ |
Al Viro | 714e85b | 2006-11-14 20:51:49 -0800 | [diff] [blame] | 137 | static __u32 check_tcp_syn_cookie(__u32 cookie, __be32 saddr, __be32 daddr, |
Florian Westphal | 8c27bd7 | 2013-09-20 22:32:55 +0200 | [diff] [blame] | 138 | __be16 sport, __be16 dport, __u32 sseq) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 139 | { |
Florian Westphal | 8c27bd7 | 2013-09-20 22:32:55 +0200 | [diff] [blame] | 140 | u32 diff, count = tcp_cookie_time(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 141 | |
| 142 | /* Strip away the layers from the cookie */ |
| 143 | cookie -= cookie_hash(saddr, daddr, sport, dport, 0, 0) + sseq; |
| 144 | |
| 145 | /* Cookie is now reduced to (count * 2^24) ^ (hash % 2^24) */ |
Weilong Chen | 0c9a67d | 2013-12-23 14:37:27 +0800 | [diff] [blame] | 146 | diff = (count - (cookie >> COOKIEBITS)) & ((__u32) -1 >> COOKIEBITS); |
Florian Westphal | 8c27bd7 | 2013-09-20 22:32:55 +0200 | [diff] [blame] | 147 | if (diff >= MAX_SYNCOOKIE_AGE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 148 | return (__u32)-1; |
| 149 | |
| 150 | return (cookie - |
| 151 | cookie_hash(saddr, daddr, sport, dport, count - diff, 1)) |
| 152 | & COOKIEMASK; /* Leaving the data behind */ |
| 153 | } |
| 154 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 155 | /* |
Florian Westphal | 0862935 | 2013-09-20 22:32:56 +0200 | [diff] [blame] | 156 | * MSS Values are chosen based on the 2011 paper |
| 157 | * 'An Analysis of TCP Maximum Segement Sizes' by S. Alcock and R. Nelson. |
| 158 | * Values .. |
| 159 | * .. lower than 536 are rare (< 0.2%) |
| 160 | * .. between 537 and 1299 account for less than < 1.5% of observed values |
| 161 | * .. in the 1300-1349 range account for about 15 to 20% of observed mss values |
| 162 | * .. exceeding 1460 are very rare (< 0.04%) |
Florian Westphal | 5918e2f | 2010-06-03 00:43:57 +0000 | [diff] [blame] | 163 | * |
Florian Westphal | 0862935 | 2013-09-20 22:32:56 +0200 | [diff] [blame] | 164 | * 1460 is the single most frequently announced mss value (30 to 46% depending |
| 165 | * on monitor location). Table must be sorted. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 166 | */ |
| 167 | static __u16 const msstab[] = { |
Florian Westphal | 5918e2f | 2010-06-03 00:43:57 +0000 | [diff] [blame] | 168 | 536, |
Florian Westphal | 0862935 | 2013-09-20 22:32:56 +0200 | [diff] [blame] | 169 | 1300, |
| 170 | 1440, /* 1440, 1452: PPPoE */ |
Florian Westphal | 5918e2f | 2010-06-03 00:43:57 +0000 | [diff] [blame] | 171 | 1460, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 172 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 173 | |
| 174 | /* |
| 175 | * Generate a syncookie. mssp points to the mss, which is returned |
| 176 | * rounded down to the value encoded in the cookie. |
| 177 | */ |
Patrick McHardy | 0198230 | 2013-08-27 08:50:13 +0200 | [diff] [blame] | 178 | u32 __cookie_v4_init_sequence(const struct iphdr *iph, const struct tcphdr *th, |
| 179 | u16 *mssp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 180 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 181 | int mssind; |
| 182 | const __u16 mss = *mssp; |
| 183 | |
Florian Westphal | 5918e2f | 2010-06-03 00:43:57 +0000 | [diff] [blame] | 184 | for (mssind = ARRAY_SIZE(msstab) - 1; mssind ; mssind--) |
| 185 | if (mss >= msstab[mssind]) |
| 186 | break; |
| 187 | *mssp = msstab[mssind]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 188 | |
Arnaldo Carvalho de Melo | aa8223c | 2007-04-10 21:04:22 -0700 | [diff] [blame] | 189 | return secure_tcp_syn_cookie(iph->saddr, iph->daddr, |
| 190 | th->source, th->dest, ntohl(th->seq), |
Florian Westphal | 8c27bd7 | 2013-09-20 22:32:55 +0200 | [diff] [blame] | 191 | mssind); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 192 | } |
Patrick McHardy | 0198230 | 2013-08-27 08:50:13 +0200 | [diff] [blame] | 193 | EXPORT_SYMBOL_GPL(__cookie_v4_init_sequence); |
| 194 | |
Eric Dumazet | 3f684b4 | 2015-09-29 07:42:49 -0700 | [diff] [blame] | 195 | __u32 cookie_v4_init_sequence(const struct sk_buff *skb, __u16 *mssp) |
Patrick McHardy | 0198230 | 2013-08-27 08:50:13 +0200 | [diff] [blame] | 196 | { |
| 197 | const struct iphdr *iph = ip_hdr(skb); |
| 198 | const struct tcphdr *th = tcp_hdr(skb); |
| 199 | |
Patrick McHardy | 0198230 | 2013-08-27 08:50:13 +0200 | [diff] [blame] | 200 | return __cookie_v4_init_sequence(iph, th, mssp); |
| 201 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 202 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 203 | /* |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 204 | * Check if a ack sequence number is a valid syncookie. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 205 | * Return the decoded mss if it is, or 0 if not. |
| 206 | */ |
Patrick McHardy | 0198230 | 2013-08-27 08:50:13 +0200 | [diff] [blame] | 207 | int __cookie_v4_check(const struct iphdr *iph, const struct tcphdr *th, |
| 208 | u32 cookie) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 209 | { |
Arnaldo Carvalho de Melo | aa8223c | 2007-04-10 21:04:22 -0700 | [diff] [blame] | 210 | __u32 seq = ntohl(th->seq) - 1; |
| 211 | __u32 mssind = check_tcp_syn_cookie(cookie, iph->saddr, iph->daddr, |
Florian Westphal | 8c27bd7 | 2013-09-20 22:32:55 +0200 | [diff] [blame] | 212 | th->source, th->dest, seq); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 213 | |
Florian Westphal | 5918e2f | 2010-06-03 00:43:57 +0000 | [diff] [blame] | 214 | return mssind < ARRAY_SIZE(msstab) ? msstab[mssind] : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 215 | } |
Patrick McHardy | 0198230 | 2013-08-27 08:50:13 +0200 | [diff] [blame] | 216 | EXPORT_SYMBOL_GPL(__cookie_v4_check); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 217 | |
Eric Dumazet | b80c0e7 | 2015-06-04 18:30:43 -0700 | [diff] [blame] | 218 | struct sock *tcp_get_cookie_sock(struct sock *sk, struct sk_buff *skb, |
| 219 | struct request_sock *req, |
| 220 | struct dst_entry *dst) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 221 | { |
Arnaldo Carvalho de Melo | 8292a17 | 2005-12-13 23:15:52 -0800 | [diff] [blame] | 222 | struct inet_connection_sock *icsk = inet_csk(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 223 | struct sock *child; |
Eric Dumazet | 5e0724d | 2015-10-22 08:20:46 -0700 | [diff] [blame] | 224 | bool own_req; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 225 | |
Eric Dumazet | 5e0724d | 2015-10-22 08:20:46 -0700 | [diff] [blame] | 226 | child = icsk->icsk_af_ops->syn_recv_sock(sk, skb, req, dst, |
| 227 | NULL, &own_req); |
Eric Dumazet | 0470c8c | 2015-03-17 18:32:31 -0700 | [diff] [blame] | 228 | if (child) { |
| 229 | atomic_set(&req->rsk_refcnt, 1); |
Eric Dumazet | 6bcfd7f | 2015-10-08 11:16:48 -0700 | [diff] [blame] | 230 | sock_rps_save_rxhash(child, skb); |
Arnaldo Carvalho de Melo | 463c84b | 2005-08-09 20:10:42 -0700 | [diff] [blame] | 231 | inet_csk_reqsk_queue_add(sk, req, child); |
Eric Dumazet | 0470c8c | 2015-03-17 18:32:31 -0700 | [diff] [blame] | 232 | } else { |
Arnaldo Carvalho de Melo | 60236fd | 2005-06-18 22:47:21 -0700 | [diff] [blame] | 233 | reqsk_free(req); |
Eric Dumazet | 0470c8c | 2015-03-17 18:32:31 -0700 | [diff] [blame] | 234 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 235 | return child; |
| 236 | } |
Eric Dumazet | b80c0e7 | 2015-06-04 18:30:43 -0700 | [diff] [blame] | 237 | EXPORT_SYMBOL(tcp_get_cookie_sock); |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 238 | |
| 239 | /* |
| 240 | * when syncookies are in effect and tcp timestamps are enabled we stored |
| 241 | * additional tcp options in the timestamp. |
| 242 | * This extracts these options from the timestamp echo. |
| 243 | * |
Florian Westphal | f167338 | 2014-11-03 17:35:02 +0100 | [diff] [blame] | 244 | * return false if we decode a tcp option that is disabled |
| 245 | * on the host. |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 246 | */ |
Florian Westphal | f167338 | 2014-11-03 17:35:02 +0100 | [diff] [blame] | 247 | bool cookie_timestamp_decode(struct tcp_options_received *tcp_opt) |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 248 | { |
Florian Westphal | 734f614 | 2010-06-21 11:48:44 +0000 | [diff] [blame] | 249 | /* echoed timestamp, lowest bits contain options */ |
Florian Westphal | 274e2da | 2014-11-03 17:35:01 +0100 | [diff] [blame] | 250 | u32 options = tcp_opt->rcv_tsecr; |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 251 | |
Florian Westphal | 8c76368 | 2010-06-16 14:42:15 -0700 | [diff] [blame] | 252 | if (!tcp_opt->saw_tstamp) { |
| 253 | tcp_clear_options(tcp_opt); |
| 254 | return true; |
| 255 | } |
| 256 | |
| 257 | if (!sysctl_tcp_timestamps) |
| 258 | return false; |
| 259 | |
Florian Westphal | 274e2da | 2014-11-03 17:35:01 +0100 | [diff] [blame] | 260 | tcp_opt->sack_ok = (options & TS_OPT_SACK) ? TCP_SACK_SEEN : 0; |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 261 | |
Florian Westphal | 8c76368 | 2010-06-16 14:42:15 -0700 | [diff] [blame] | 262 | if (tcp_opt->sack_ok && !sysctl_tcp_sack) |
| 263 | return false; |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 264 | |
Florian Westphal | 274e2da | 2014-11-03 17:35:01 +0100 | [diff] [blame] | 265 | if ((options & TS_OPT_WSCALE_MASK) == TS_OPT_WSCALE_MASK) |
Florian Westphal | 734f614 | 2010-06-21 11:48:44 +0000 | [diff] [blame] | 266 | return true; /* no window scaling */ |
| 267 | |
| 268 | tcp_opt->wscale_ok = 1; |
Florian Westphal | 274e2da | 2014-11-03 17:35:01 +0100 | [diff] [blame] | 269 | tcp_opt->snd_wscale = options & TS_OPT_WSCALE_MASK; |
| 270 | |
Florian Westphal | 734f614 | 2010-06-21 11:48:44 +0000 | [diff] [blame] | 271 | return sysctl_tcp_window_scaling != 0; |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 272 | } |
Florian Westphal | f167338 | 2014-11-03 17:35:02 +0100 | [diff] [blame] | 273 | EXPORT_SYMBOL(cookie_timestamp_decode); |
| 274 | |
| 275 | bool cookie_ecn_ok(const struct tcp_options_received *tcp_opt, |
Florian Westphal | f7b3bec | 2014-11-03 17:35:03 +0100 | [diff] [blame] | 276 | const struct net *net, const struct dst_entry *dst) |
Florian Westphal | f167338 | 2014-11-03 17:35:02 +0100 | [diff] [blame] | 277 | { |
| 278 | bool ecn_ok = tcp_opt->rcv_tsecr & TS_OPT_ECN; |
| 279 | |
| 280 | if (!ecn_ok) |
| 281 | return false; |
| 282 | |
| 283 | if (net->ipv4.sysctl_tcp_ecn) |
| 284 | return true; |
| 285 | |
Florian Westphal | f7b3bec | 2014-11-03 17:35:03 +0100 | [diff] [blame] | 286 | return dst_feature(dst, RTAX_FEATURE_ECN); |
Florian Westphal | f167338 | 2014-11-03 17:35:02 +0100 | [diff] [blame] | 287 | } |
| 288 | EXPORT_SYMBOL(cookie_ecn_ok); |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 289 | |
Eric Dumazet | 079096f | 2015-10-02 11:43:32 -0700 | [diff] [blame] | 290 | /* On input, sk is a listener. |
| 291 | * Output is listener if incoming packet would not create a child |
| 292 | * NULL if memory could not be allocated. |
| 293 | */ |
Cong Wang | 461b74c | 2014-10-15 14:33:22 -0700 | [diff] [blame] | 294 | struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 295 | { |
Cong Wang | 461b74c | 2014-10-15 14:33:22 -0700 | [diff] [blame] | 296 | struct ip_options *opt = &TCP_SKB_CB(skb)->header.h4.opt; |
William Allen Simpson | 4957faade | 2009-12-02 18:25:27 +0000 | [diff] [blame] | 297 | struct tcp_options_received tcp_opt; |
Arnaldo Carvalho de Melo | 2e6599c | 2005-06-18 22:46:52 -0700 | [diff] [blame] | 298 | struct inet_request_sock *ireq; |
| 299 | struct tcp_request_sock *treq; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 300 | struct tcp_sock *tp = tcp_sk(sk); |
Arnaldo Carvalho de Melo | aa8223c | 2007-04-10 21:04:22 -0700 | [diff] [blame] | 301 | const struct tcphdr *th = tcp_hdr(skb); |
| 302 | __u32 cookie = ntohl(th->ack_seq) - 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 303 | struct sock *ret = sk; |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 304 | struct request_sock *req; |
| 305 | int mss; |
| 306 | struct rtable *rt; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 307 | __u8 rcv_wscale; |
Eric Dumazet | dfd25ff | 2012-03-10 09:20:21 +0000 | [diff] [blame] | 308 | struct flowi4 fl4; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 309 | |
Florian Westphal | af9b473 | 2010-06-03 00:43:44 +0000 | [diff] [blame] | 310 | if (!sysctl_tcp_syncookies || !th->ack || th->rst) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 311 | goto out; |
| 312 | |
Florian Westphal | 646697b | 2014-10-30 02:55:38 +0100 | [diff] [blame] | 313 | if (tcp_synq_no_recent_overflow(sk)) |
| 314 | goto out; |
| 315 | |
| 316 | mss = __cookie_v4_check(ip_hdr(skb), th, cookie); |
| 317 | if (mss == 0) { |
Pavel Emelyanov | de0744a | 2008-07-16 20:31:16 -0700 | [diff] [blame] | 318 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_SYNCOOKIESFAILED); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 319 | goto out; |
| 320 | } |
| 321 | |
Pavel Emelyanov | de0744a | 2008-07-16 20:31:16 -0700 | [diff] [blame] | 322 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_SYNCOOKIESRECV); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 323 | |
David S. Miller | bb5b7c1 | 2009-12-15 20:56:42 -0800 | [diff] [blame] | 324 | /* check for timestamp cookie support */ |
| 325 | memset(&tcp_opt, 0, sizeof(tcp_opt)); |
Christoph Paasch | 1a2c618 | 2013-03-17 08:23:34 +0000 | [diff] [blame] | 326 | tcp_parse_options(skb, &tcp_opt, 0, NULL); |
David S. Miller | bb5b7c1 | 2009-12-15 20:56:42 -0800 | [diff] [blame] | 327 | |
Florian Westphal | f167338 | 2014-11-03 17:35:02 +0100 | [diff] [blame] | 328 | if (!cookie_timestamp_decode(&tcp_opt)) |
Florian Westphal | 8c76368 | 2010-06-16 14:42:15 -0700 | [diff] [blame] | 329 | goto out; |
David S. Miller | bb5b7c1 | 2009-12-15 20:56:42 -0800 | [diff] [blame] | 330 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 331 | ret = NULL; |
Eric Dumazet | a1a5344 | 2015-10-04 21:08:11 -0700 | [diff] [blame] | 332 | req = inet_reqsk_alloc(&tcp_request_sock_ops, sk, false); /* for safety */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 333 | if (!req) |
| 334 | goto out; |
| 335 | |
Arnaldo Carvalho de Melo | 2e6599c | 2005-06-18 22:46:52 -0700 | [diff] [blame] | 336 | ireq = inet_rsk(req); |
| 337 | treq = tcp_rsk(req); |
Arnaldo Carvalho de Melo | aa8223c | 2007-04-10 21:04:22 -0700 | [diff] [blame] | 338 | treq->rcv_isn = ntohl(th->seq) - 1; |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 339 | treq->snt_isn = cookie; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 340 | req->mss = mss; |
Eric Dumazet | b44084c | 2013-10-10 00:04:37 -0700 | [diff] [blame] | 341 | ireq->ir_num = ntohs(th->dest); |
| 342 | ireq->ir_rmt_port = th->source; |
Eric Dumazet | 08d2cc3b | 2015-03-18 14:05:38 -0700 | [diff] [blame] | 343 | sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr); |
| 344 | sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr); |
Lorenzo Colitti | 84f39b0 | 2014-05-13 10:17:35 -0700 | [diff] [blame] | 345 | ireq->ir_mark = inet_request_mark(sk, skb); |
David S. Miller | bb5b7c1 | 2009-12-15 20:56:42 -0800 | [diff] [blame] | 346 | ireq->snd_wscale = tcp_opt.snd_wscale; |
David S. Miller | bb5b7c1 | 2009-12-15 20:56:42 -0800 | [diff] [blame] | 347 | ireq->sack_ok = tcp_opt.sack_ok; |
| 348 | ireq->wscale_ok = tcp_opt.wscale_ok; |
| 349 | ireq->tstamp_ok = tcp_opt.saw_tstamp; |
| 350 | req->ts_recent = tcp_opt.saw_tstamp ? tcp_opt.rcv_tsval : 0; |
Yuchung Cheng | 0f1c28a | 2015-09-18 11:36:14 -0700 | [diff] [blame] | 351 | treq->snt_synack.v64 = 0; |
Eric Dumazet | 9439ce0 | 2015-03-17 18:32:29 -0700 | [diff] [blame] | 352 | treq->tfo_listener = false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 353 | |
Eric Dumazet | 16f8616 | 2015-03-13 15:51:10 -0700 | [diff] [blame] | 354 | ireq->ir_iif = sk->sk_bound_dev_if; |
| 355 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 356 | /* We throwed the options of the initial SYN away, so we hope |
| 357 | * the ACK carries the same options again (see RFC1122 4.2.3.8) |
| 358 | */ |
Cong Wang | e25f866 | 2014-10-15 14:33:21 -0700 | [diff] [blame] | 359 | ireq->opt = tcp_v4_save_options(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 360 | |
Paul Moore | 284904a | 2009-03-27 17:10:28 -0400 | [diff] [blame] | 361 | if (security_inet_conn_request(sk, skb, req)) { |
Eric Dumazet | 0470c8c | 2015-03-17 18:32:31 -0700 | [diff] [blame] | 362 | reqsk_free(req); |
Paul Moore | 284904a | 2009-03-27 17:10:28 -0400 | [diff] [blame] | 363 | goto out; |
| 364 | } |
| 365 | |
Eric Dumazet | e6c022a | 2012-10-27 23:16:46 +0000 | [diff] [blame] | 366 | req->num_retrans = 0; |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 367 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 368 | /* |
| 369 | * We need to lookup the route here to get at the correct |
| 370 | * window size. We should better make sure that the window size |
| 371 | * hasn't changed since we received the original syn, but I see |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 372 | * no easy way to do this. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 373 | */ |
Lorenzo Colitti | 84f39b0 | 2014-05-13 10:17:35 -0700 | [diff] [blame] | 374 | flowi4_init_output(&fl4, sk->sk_bound_dev_if, ireq->ir_mark, |
Dmitry Popov | d66954a | 2013-04-11 08:55:07 +0000 | [diff] [blame] | 375 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, IPPROTO_TCP, |
Eric Dumazet | dfd25ff | 2012-03-10 09:20:21 +0000 | [diff] [blame] | 376 | inet_sk_flowi_flags(sk), |
Cong Wang | 461b74c | 2014-10-15 14:33:22 -0700 | [diff] [blame] | 377 | opt->srr ? opt->faddr : ireq->ir_rmt_addr, |
Eric Dumazet | 634fb979 | 2013-10-09 15:21:29 -0700 | [diff] [blame] | 378 | ireq->ir_loc_addr, th->source, th->dest); |
Eric Dumazet | dfd25ff | 2012-03-10 09:20:21 +0000 | [diff] [blame] | 379 | security_req_classify_flow(req, flowi4_to_flowi(&fl4)); |
| 380 | rt = ip_route_output_key(sock_net(sk), &fl4); |
| 381 | if (IS_ERR(rt)) { |
Eric Dumazet | 0470c8c | 2015-03-17 18:32:31 -0700 | [diff] [blame] | 382 | reqsk_free(req); |
Eric Dumazet | dfd25ff | 2012-03-10 09:20:21 +0000 | [diff] [blame] | 383 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 384 | } |
| 385 | |
| 386 | /* Try to redo what tcp_v4_send_synack did. */ |
Eric Dumazet | ed53d0a | 2015-10-08 19:33:23 -0700 | [diff] [blame] | 387 | req->rsk_window_clamp = tp->window_clamp ? :dst_metric(&rt->dst, RTAX_WINDOW); |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 388 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 389 | tcp_select_initial_window(tcp_full_space(sk), req->mss, |
Eric Dumazet | ed53d0a | 2015-10-08 19:33:23 -0700 | [diff] [blame] | 390 | &req->rsk_rcv_wnd, &req->rsk_window_clamp, |
laurent chavey | 31d1292 | 2009-12-15 11:15:28 +0000 | [diff] [blame] | 391 | ireq->wscale_ok, &rcv_wscale, |
Changli Gao | d8d1f30 | 2010-06-10 23:31:35 -0700 | [diff] [blame] | 392 | dst_metric(&rt->dst, RTAX_INITRWND)); |
Florian Westphal | 4dfc281 | 2008-04-10 03:12:40 -0700 | [diff] [blame] | 393 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 394 | ireq->rcv_wscale = rcv_wscale; |
Florian Westphal | f7b3bec | 2014-11-03 17:35:03 +0100 | [diff] [blame] | 395 | ireq->ecn_ok = cookie_ecn_ok(&tcp_opt, sock_net(sk), &rt->dst); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 396 | |
Eric Dumazet | b80c0e7 | 2015-06-04 18:30:43 -0700 | [diff] [blame] | 397 | ret = tcp_get_cookie_sock(sk, skb, req, &rt->dst); |
Eric Dumazet | dfd25ff | 2012-03-10 09:20:21 +0000 | [diff] [blame] | 398 | /* ip_queue_xmit() depends on our flow being setup |
| 399 | * Normal sockets get it right from inet_csk_route_child_sock() |
| 400 | */ |
| 401 | if (ret) |
| 402 | inet_sk(ret)->cork.fl.u.ip4 = fl4; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 403 | out: return ret; |
| 404 | } |