5 #include <linux/skbuff.h>
6 #include <linux/if_vlan.h>
8 #include <net/inet_sock.h>
9 #include <net/dsfield.h>
19 extern int sysctl_tunnel_ecn_log;
21 static inline int INET_ECN_is_ce(__u8 dsfield)
23 return (dsfield & INET_ECN_MASK) == INET_ECN_CE;
26 static inline int INET_ECN_is_not_ect(__u8 dsfield)
28 return (dsfield & INET_ECN_MASK) == INET_ECN_NOT_ECT;
31 static inline int INET_ECN_is_capable(__u8 dsfield)
33 return dsfield & INET_ECN_ECT_0;
38 * The full-functionality option for ECN encapsulation is to copy the
39 * ECN codepoint of the inside header to the outside header on
40 * encapsulation if the inside header is not-ECT or ECT, and to set the
41 * ECN codepoint of the outside header to ECT(0) if the ECN codepoint of
42 * the inside header is CE.
44 static inline __u8 INET_ECN_encapsulate(__u8 outer, __u8 inner)
46 outer &= ~INET_ECN_MASK;
47 outer |= !INET_ECN_is_ce(inner) ? (inner & INET_ECN_MASK) :
52 static inline void INET_ECN_xmit(struct sock *sk)
54 inet_sk(sk)->tos |= INET_ECN_ECT_0;
55 if (inet6_sk(sk) != NULL)
56 inet6_sk(sk)->tclass |= INET_ECN_ECT_0;
59 static inline void INET_ECN_dontxmit(struct sock *sk)
61 inet_sk(sk)->tos &= ~INET_ECN_MASK;
62 if (inet6_sk(sk) != NULL)
63 inet6_sk(sk)->tclass &= ~INET_ECN_MASK;
66 #define IP6_ECN_flow_init(label) do { \
67 (label) &= ~htonl(INET_ECN_MASK << 20); \
70 #define IP6_ECN_flow_xmit(sk, label) do { \
71 if (INET_ECN_is_capable(inet6_sk(sk)->tclass)) \
72 (label) |= htonl(INET_ECN_ECT_0 << 20); \
75 static inline int IP_ECN_set_ce(struct iphdr *iph)
77 u32 check = (__force u32)iph->check;
78 u32 ecn = (iph->tos + 1) & INET_ECN_MASK;
81 * After the last operation we have (in binary):
82 * INET_ECN_NOT_ECT => 01
83 * INET_ECN_ECT_1 => 10
84 * INET_ECN_ECT_0 => 11
91 * The following gives us:
92 * INET_ECN_ECT_1 => check += htons(0xFFFD)
93 * INET_ECN_ECT_0 => check += htons(0xFFFE)
95 check += (__force u16)htons(0xFFFB) + (__force u16)htons(ecn);
97 iph->check = (__force __sum16)(check + (check>=0xFFFF));
98 iph->tos |= INET_ECN_CE;
102 static inline void IP_ECN_clear(struct iphdr *iph)
104 iph->tos &= ~INET_ECN_MASK;
107 static inline void ipv4_copy_dscp(unsigned int dscp, struct iphdr *inner)
109 dscp &= ~INET_ECN_MASK;
110 ipv4_change_dsfield(inner, INET_ECN_MASK, dscp);
116 * IP_ECN_set_ce() has to tweak IPV4 checksum when setting CE,
117 * meaning both changes have no effect on skb->csum if/when CHECKSUM_COMPLETE
118 * In IPv6 case, no checksum compensates the change in IPv6 header,
119 * so we have to update skb->csum.
121 static inline int IP6_ECN_set_ce(struct sk_buff *skb, struct ipv6hdr *iph)
125 if (INET_ECN_is_not_ect(ipv6_get_dsfield(iph)))
128 from = *(__be32 *)iph;
129 to = from | htonl(INET_ECN_CE << 20);
131 if (skb->ip_summed == CHECKSUM_COMPLETE)
132 skb->csum = csum_add(csum_sub(skb->csum, (__force __wsum)from),
137 static inline void IP6_ECN_clear(struct ipv6hdr *iph)
139 *(__be32*)iph &= ~htonl(INET_ECN_MASK << 20);
142 static inline void ipv6_copy_dscp(unsigned int dscp, struct ipv6hdr *inner)
144 dscp &= ~INET_ECN_MASK;
145 ipv6_change_dsfield(inner, INET_ECN_MASK, dscp);
148 static inline int INET_ECN_set_ce(struct sk_buff *skb)
150 switch (skb->protocol) {
151 case cpu_to_be16(ETH_P_IP):
152 if (skb_network_header(skb) + sizeof(struct iphdr) <=
153 skb_tail_pointer(skb))
154 return IP_ECN_set_ce(ip_hdr(skb));
157 case cpu_to_be16(ETH_P_IPV6):
158 if (skb_network_header(skb) + sizeof(struct ipv6hdr) <=
159 skb_tail_pointer(skb))
160 return IP6_ECN_set_ce(skb, ipv6_hdr(skb));
169 * To decapsulate the inner header at the tunnel egress, a compliant
170 * tunnel egress MUST set the outgoing ECN field to the codepoint at the
171 * intersection of the appropriate arriving inner header (row) and outer
172 * header (column) in Figure 4
174 * +---------+------------------------------------------------+
175 * |Arriving | Arriving Outer Header |
176 * | Inner +---------+------------+------------+------------+
177 * | Header | Not-ECT | ECT(0) | ECT(1) | CE |
178 * +---------+---------+------------+------------+------------+
179 * | Not-ECT | Not-ECT |Not-ECT(!!!)|Not-ECT(!!!)| <drop>(!!!)|
180 * | ECT(0) | ECT(0) | ECT(0) | ECT(1) | CE |
181 * | ECT(1) | ECT(1) | ECT(1) (!) | ECT(1) | CE |
182 * | CE | CE | CE | CE(!!!)| CE |
183 * +---------+---------+------------+------------+------------+
185 * Figure 4: New IP in IP Decapsulation Behaviour
187 * returns 0 on success
188 * 1 if something is broken and should be logged (!!! above)
189 * 2 if packet should be dropped
191 static inline int INET_ECN_decapsulate(struct sk_buff *skb,
192 __u8 outer, __u8 inner)
194 if (INET_ECN_is_not_ect(inner)) {
195 switch (outer & INET_ECN_MASK) {
196 case INET_ECN_NOT_ECT:
206 if (INET_ECN_is_ce(outer))
207 INET_ECN_set_ce(skb);
212 static inline int IP_ECN_decapsulate(const struct iphdr *oiph,
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));
224 return INET_ECN_decapsulate(skb, oiph->tos, inner);
227 static inline int IP6_ECN_decapsulate(const struct ipv6hdr *oipv6h,
232 if (skb->protocol == htons(ETH_P_IP))
233 inner = ip_hdr(skb)->tos;
234 else if (skb->protocol == htons(ETH_P_IPV6))
235 inner = ipv6_get_dsfield(ipv6_hdr(skb));
239 return INET_ECN_decapsulate(skb, ipv6_get_dsfield(oipv6h), inner);