1 #ifndef _INET_ECN_H_ 2 #define _INET_ECN_H_ 3 4 #include <linux/ip.h> 5 #include <linux/skbuff.h> 6 7 #include <net/inet_sock.h> 8 #include <net/dsfield.h> 9 10 enum { 11 INET_ECN_NOT_ECT = 0, 12 INET_ECN_ECT_1 = 1, 13 INET_ECN_ECT_0 = 2, 14 INET_ECN_CE = 3, 15 INET_ECN_MASK = 3, 16 }; 17 18 extern int sysctl_tunnel_ecn_log; 19 20 static inline int INET_ECN_is_ce(__u8 dsfield) 21 { 22 return (dsfield & INET_ECN_MASK) == INET_ECN_CE; 23 } 24 25 static inline int INET_ECN_is_not_ect(__u8 dsfield) 26 { 27 return (dsfield & INET_ECN_MASK) == INET_ECN_NOT_ECT; 28 } 29 30 static inline int INET_ECN_is_capable(__u8 dsfield) 31 { 32 return dsfield & INET_ECN_ECT_0; 33 } 34 35 /* 36 * RFC 3168 9.1.1 37 * The full-functionality option for ECN encapsulation is to copy the 38 * ECN codepoint of the inside header to the outside header on 39 * encapsulation if the inside header is not-ECT or ECT, and to set the 40 * ECN codepoint of the outside header to ECT(0) if the ECN codepoint of 41 * the inside header is CE. 42 */ 43 static inline __u8 INET_ECN_encapsulate(__u8 outer, __u8 inner) 44 { 45 outer &= ~INET_ECN_MASK; 46 outer |= !INET_ECN_is_ce(inner) ? (inner & INET_ECN_MASK) : 47 INET_ECN_ECT_0; 48 return outer; 49 } 50 51 static inline void INET_ECN_xmit(struct sock *sk) 52 { 53 inet_sk(sk)->tos |= INET_ECN_ECT_0; 54 if (inet6_sk(sk) != NULL) 55 inet6_sk(sk)->tclass |= INET_ECN_ECT_0; 56 } 57 58 static inline void INET_ECN_dontxmit(struct sock *sk) 59 { 60 inet_sk(sk)->tos &= ~INET_ECN_MASK; 61 if (inet6_sk(sk) != NULL) 62 inet6_sk(sk)->tclass &= ~INET_ECN_MASK; 63 } 64 65 #define IP6_ECN_flow_init(label) do { \ 66 (label) &= ~htonl(INET_ECN_MASK << 20); \ 67 } while (0) 68 69 #define IP6_ECN_flow_xmit(sk, label) do { \ 70 if (INET_ECN_is_capable(inet6_sk(sk)->tclass)) \ 71 (label) |= htonl(INET_ECN_ECT_0 << 20); \ 72 } while (0) 73 74 static inline int IP_ECN_set_ce(struct iphdr *iph) 75 { 76 u32 check = (__force u32)iph->check; 77 u32 ecn = (iph->tos + 1) & INET_ECN_MASK; 78 79 /* 80 * After the last operation we have (in binary): 81 * INET_ECN_NOT_ECT => 01 82 * INET_ECN_ECT_1 => 10 83 * INET_ECN_ECT_0 => 11 84 * INET_ECN_CE => 00 85 */ 86 if (!(ecn & 2)) 87 return !ecn; 88 89 /* 90 * The following gives us: 91 * INET_ECN_ECT_1 => check += htons(0xFFFD) 92 * INET_ECN_ECT_0 => check += htons(0xFFFE) 93 */ 94 check += (__force u16)htons(0xFFFB) + (__force u16)htons(ecn); 95 96 iph->check = (__force __sum16)(check + (check>=0xFFFF)); 97 iph->tos |= INET_ECN_CE; 98 return 1; 99 } 100 101 static inline void IP_ECN_clear(struct iphdr *iph) 102 { 103 iph->tos &= ~INET_ECN_MASK; 104 } 105 106 static inline void ipv4_copy_dscp(unsigned int dscp, struct iphdr *inner) 107 { 108 dscp &= ~INET_ECN_MASK; 109 ipv4_change_dsfield(inner, INET_ECN_MASK, dscp); 110 } 111 112 struct ipv6hdr; 113 114 static inline int IP6_ECN_set_ce(struct ipv6hdr *iph) 115 { 116 if (INET_ECN_is_not_ect(ipv6_get_dsfield(iph))) 117 return 0; 118 *(__be32*)iph |= htonl(INET_ECN_CE << 20); 119 return 1; 120 } 121 122 static inline void IP6_ECN_clear(struct ipv6hdr *iph) 123 { 124 *(__be32*)iph &= ~htonl(INET_ECN_MASK << 20); 125 } 126 127 static inline void ipv6_copy_dscp(unsigned int dscp, struct ipv6hdr *inner) 128 { 129 dscp &= ~INET_ECN_MASK; 130 ipv6_change_dsfield(inner, INET_ECN_MASK, dscp); 131 } 132 133 static inline int INET_ECN_set_ce(struct sk_buff *skb) 134 { 135 switch (skb->protocol) { 136 case cpu_to_be16(ETH_P_IP): 137 if (skb_network_header(skb) + sizeof(struct iphdr) <= 138 skb_tail_pointer(skb)) 139 return IP_ECN_set_ce(ip_hdr(skb)); 140 break; 141 142 case cpu_to_be16(ETH_P_IPV6): 143 if (skb_network_header(skb) + sizeof(struct ipv6hdr) <= 144 skb_tail_pointer(skb)) 145 return IP6_ECN_set_ce(ipv6_hdr(skb)); 146 break; 147 } 148 149 return 0; 150 } 151 152 /* 153 * RFC 6080 4.2 154 * To decapsulate the inner header at the tunnel egress, a compliant 155 * tunnel egress MUST set the outgoing ECN field to the codepoint at the 156 * intersection of the appropriate arriving inner header (row) and outer 157 * header (column) in Figure 4 158 * 159 * +---------+------------------------------------------------+ 160 * |Arriving | Arriving Outer Header | 161 * | Inner +---------+------------+------------+------------+ 162 * | Header | Not-ECT | ECT(0) | ECT(1) | CE | 163 * +---------+---------+------------+------------+------------+ 164 * | Not-ECT | Not-ECT |Not-ECT(!!!)|Not-ECT(!!!)| <drop>(!!!)| 165 * | ECT(0) | ECT(0) | ECT(0) | ECT(1) | CE | 166 * | ECT(1) | ECT(1) | ECT(1) (!) | ECT(1) | CE | 167 * | CE | CE | CE | CE(!!!)| CE | 168 * +---------+---------+------------+------------+------------+ 169 * 170 * Figure 4: New IP in IP Decapsulation Behaviour 171 * 172 * returns 0 on success 173 * 1 if something is broken and should be logged (!!! above) 174 * 2 if packet should be dropped 175 */ 176 static inline int INET_ECN_decapsulate(struct sk_buff *skb, 177 __u8 outer, __u8 inner) 178 { 179 if (INET_ECN_is_not_ect(inner)) { 180 switch (outer & INET_ECN_MASK) { 181 case INET_ECN_NOT_ECT: 182 return 0; 183 case INET_ECN_ECT_0: 184 case INET_ECN_ECT_1: 185 return 1; 186 case INET_ECN_CE: 187 return 2; 188 } 189 } 190 191 if (INET_ECN_is_ce(outer)) 192 INET_ECN_set_ce(skb); 193 194 return 0; 195 } 196 197 static inline int IP_ECN_decapsulate(const struct iphdr *oiph, 198 struct sk_buff *skb) 199 { 200 __u8 inner; 201 202 if (skb->protocol == htons(ETH_P_IP)) 203 inner = ip_hdr(skb)->tos; 204 else if (skb->protocol == htons(ETH_P_IPV6)) 205 inner = ipv6_get_dsfield(ipv6_hdr(skb)); 206 else 207 return 0; 208 209 return INET_ECN_decapsulate(skb, oiph->tos, inner); 210 } 211 212 static inline int IP6_ECN_decapsulate(const struct ipv6hdr *oipv6h, 213 struct sk_buff *skb) 214 { 215 __u8 inner; 216 217 if (skb->protocol == htons(ETH_P_IP)) 218 inner = ip_hdr(skb)->tos; 219 else if (skb->protocol == htons(ETH_P_IPV6)) 220 inner = ipv6_get_dsfield(ipv6_hdr(skb)); 221 else 222 return 0; 223 224 return INET_ECN_decapsulate(skb, ipv6_get_dsfield(oipv6h), inner); 225 } 226 #endif 227