1 /* SPDX-License-Identifier: GPL-2.0 */
6 #include <linux/skbuff.h>
7 #include <linux/if_vlan.h>
9 #include <net/inet_sock.h>
10 #include <net/dsfield.h>
20 extern int sysctl_tunnel_ecn_log;
22 static inline int INET_ECN_is_ce(__u8 dsfield)
24 return (dsfield & INET_ECN_MASK) == INET_ECN_CE;
27 static inline int INET_ECN_is_not_ect(__u8 dsfield)
29 return (dsfield & INET_ECN_MASK) == INET_ECN_NOT_ECT;
32 static inline int INET_ECN_is_capable(__u8 dsfield)
34 return dsfield & INET_ECN_ECT_0;
39 * The full-functionality option for ECN encapsulation is to copy the
40 * ECN codepoint of the inside header to the outside header on
41 * encapsulation if the inside header is not-ECT or ECT, and to set the
42 * ECN codepoint of the outside header to ECT(0) if the ECN codepoint of
43 * the inside header is CE.
45 static inline __u8 INET_ECN_encapsulate(__u8 outer, __u8 inner)
47 outer &= ~INET_ECN_MASK;
48 outer |= !INET_ECN_is_ce(inner) ? (inner & INET_ECN_MASK) :
53 static inline void INET_ECN_xmit(struct sock *sk)
55 inet_sk(sk)->tos |= INET_ECN_ECT_0;
56 if (inet6_sk(sk) != NULL)
57 inet6_sk(sk)->tclass |= INET_ECN_ECT_0;
60 static inline void INET_ECN_dontxmit(struct sock *sk)
62 inet_sk(sk)->tos &= ~INET_ECN_MASK;
63 if (inet6_sk(sk) != NULL)
64 inet6_sk(sk)->tclass &= ~INET_ECN_MASK;
67 #define IP6_ECN_flow_init(label) do { \
68 (label) &= ~htonl(INET_ECN_MASK << 20); \
71 #define IP6_ECN_flow_xmit(sk, label) do { \
72 if (INET_ECN_is_capable(inet6_sk(sk)->tclass)) \
73 (label) |= htonl(INET_ECN_ECT_0 << 20); \
76 static inline int IP_ECN_set_ce(struct iphdr *iph)
78 u32 check = (__force u32)iph->check;
79 u32 ecn = (iph->tos + 1) & INET_ECN_MASK;
82 * After the last operation we have (in binary):
83 * INET_ECN_NOT_ECT => 01
84 * INET_ECN_ECT_1 => 10
85 * INET_ECN_ECT_0 => 11
92 * The following gives us:
93 * INET_ECN_ECT_1 => check += htons(0xFFFD)
94 * INET_ECN_ECT_0 => check += htons(0xFFFE)
96 check += (__force u16)htons(0xFFFB) + (__force u16)htons(ecn);
98 iph->check = (__force __sum16)(check + (check>=0xFFFF));
99 iph->tos |= INET_ECN_CE;
103 static inline int IP_ECN_set_ect1(struct iphdr *iph)
105 u32 check = (__force u32)iph->check;
107 if ((iph->tos & INET_ECN_MASK) != INET_ECN_ECT_0)
110 check += (__force u16)htons(0x1);
112 iph->check = (__force __sum16)(check + (check>=0xFFFF));
113 iph->tos ^= INET_ECN_MASK;
117 static inline void IP_ECN_clear(struct iphdr *iph)
119 iph->tos &= ~INET_ECN_MASK;
122 static inline void ipv4_copy_dscp(unsigned int dscp, struct iphdr *inner)
124 dscp &= ~INET_ECN_MASK;
125 ipv4_change_dsfield(inner, INET_ECN_MASK, dscp);
131 * IP_ECN_set_ce() has to tweak IPV4 checksum when setting CE,
132 * meaning both changes have no effect on skb->csum if/when CHECKSUM_COMPLETE
133 * In IPv6 case, no checksum compensates the change in IPv6 header,
134 * so we have to update skb->csum.
136 static inline int IP6_ECN_set_ce(struct sk_buff *skb, struct ipv6hdr *iph)
140 if (INET_ECN_is_not_ect(ipv6_get_dsfield(iph)))
143 from = *(__be32 *)iph;
144 to = from | htonl(INET_ECN_CE << 20);
146 if (skb->ip_summed == CHECKSUM_COMPLETE)
147 skb->csum = csum_add(csum_sub(skb->csum, (__force __wsum)from),
152 static inline int IP6_ECN_set_ect1(struct sk_buff *skb, struct ipv6hdr *iph)
156 if ((ipv6_get_dsfield(iph) & INET_ECN_MASK) != INET_ECN_ECT_0)
159 from = *(__be32 *)iph;
160 to = from ^ htonl(INET_ECN_MASK << 20);
162 if (skb->ip_summed == CHECKSUM_COMPLETE)
163 skb->csum = csum_add(csum_sub(skb->csum, (__force __wsum)from),
168 static inline void ipv6_copy_dscp(unsigned int dscp, struct ipv6hdr *inner)
170 dscp &= ~INET_ECN_MASK;
171 ipv6_change_dsfield(inner, INET_ECN_MASK, dscp);
174 static inline int INET_ECN_set_ce(struct sk_buff *skb)
176 switch (skb_protocol(skb, true)) {
177 case cpu_to_be16(ETH_P_IP):
178 if (skb_network_header(skb) + sizeof(struct iphdr) <=
179 skb_tail_pointer(skb))
180 return IP_ECN_set_ce(ip_hdr(skb));
183 case cpu_to_be16(ETH_P_IPV6):
184 if (skb_network_header(skb) + sizeof(struct ipv6hdr) <=
185 skb_tail_pointer(skb))
186 return IP6_ECN_set_ce(skb, ipv6_hdr(skb));
193 static inline int INET_ECN_set_ect1(struct sk_buff *skb)
195 switch (skb_protocol(skb, true)) {
196 case cpu_to_be16(ETH_P_IP):
197 if (skb_network_header(skb) + sizeof(struct iphdr) <=
198 skb_tail_pointer(skb))
199 return IP_ECN_set_ect1(ip_hdr(skb));
202 case cpu_to_be16(ETH_P_IPV6):
203 if (skb_network_header(skb) + sizeof(struct ipv6hdr) <=
204 skb_tail_pointer(skb))
205 return IP6_ECN_set_ect1(skb, ipv6_hdr(skb));
214 * To decapsulate the inner header at the tunnel egress, a compliant
215 * tunnel egress MUST set the outgoing ECN field to the codepoint at the
216 * intersection of the appropriate arriving inner header (row) and outer
217 * header (column) in Figure 4
219 * +---------+------------------------------------------------+
220 * |Arriving | Arriving Outer Header |
221 * | Inner +---------+------------+------------+------------+
222 * | Header | Not-ECT | ECT(0) | ECT(1) | CE |
223 * +---------+---------+------------+------------+------------+
224 * | Not-ECT | Not-ECT |Not-ECT(!!!)|Not-ECT(!!!)| <drop>(!!!)|
225 * | ECT(0) | ECT(0) | ECT(0) | ECT(1) | CE |
226 * | ECT(1) | ECT(1) | ECT(1) (!) | ECT(1) | CE |
227 * | CE | CE | CE | CE(!!!)| CE |
228 * +---------+---------+------------+------------+------------+
230 * Figure 4: New IP in IP Decapsulation Behaviour
232 * returns 0 on success
233 * 1 if something is broken and should be logged (!!! above)
234 * 2 if packet should be dropped
236 static inline int __INET_ECN_decapsulate(__u8 outer, __u8 inner, bool *set_ce)
238 if (INET_ECN_is_not_ect(inner)) {
239 switch (outer & INET_ECN_MASK) {
240 case INET_ECN_NOT_ECT:
250 *set_ce = INET_ECN_is_ce(outer);
254 static inline int INET_ECN_decapsulate(struct sk_buff *skb,
255 __u8 outer, __u8 inner)
260 rc = __INET_ECN_decapsulate(outer, inner, &set_ce);
263 INET_ECN_set_ce(skb);
264 else if ((outer & INET_ECN_MASK) == INET_ECN_ECT_1)
265 INET_ECN_set_ect1(skb);
271 static inline int IP_ECN_decapsulate(const struct iphdr *oiph,
276 switch (skb_protocol(skb, true)) {
277 case htons(ETH_P_IP):
278 inner = ip_hdr(skb)->tos;
280 case htons(ETH_P_IPV6):
281 inner = ipv6_get_dsfield(ipv6_hdr(skb));
287 return INET_ECN_decapsulate(skb, oiph->tos, inner);
290 static inline int IP6_ECN_decapsulate(const struct ipv6hdr *oipv6h,
295 switch (skb_protocol(skb, true)) {
296 case htons(ETH_P_IP):
297 inner = ip_hdr(skb)->tos;
299 case htons(ETH_P_IPV6):
300 inner = ipv6_get_dsfield(ipv6_hdr(skb));
306 return INET_ECN_decapsulate(skb, ipv6_get_dsfield(oipv6h), inner);