1 /* SPDX-License-Identifier: GPL-2.0 */
6 #include <linux/skbuff.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 int IP_ECN_set_ect1(struct iphdr *iph)
104 u32 check = (__force u32)iph->check;
106 if ((iph->tos & INET_ECN_MASK) != INET_ECN_ECT_0)
109 check += (__force u16)htons(0x100);
111 iph->check = (__force __sum16)(check + (check>=0xFFFF));
112 iph->tos ^= INET_ECN_MASK;
116 static inline void IP_ECN_clear(struct iphdr *iph)
118 iph->tos &= ~INET_ECN_MASK;
121 static inline void ipv4_copy_dscp(unsigned int dscp, struct iphdr *inner)
123 dscp &= ~INET_ECN_MASK;
124 ipv4_change_dsfield(inner, INET_ECN_MASK, dscp);
130 * IP_ECN_set_ce() has to tweak IPV4 checksum when setting CE,
131 * meaning both changes have no effect on skb->csum if/when CHECKSUM_COMPLETE
132 * In IPv6 case, no checksum compensates the change in IPv6 header,
133 * so we have to update skb->csum.
135 static inline int IP6_ECN_set_ce(struct sk_buff *skb, struct ipv6hdr *iph)
139 if (INET_ECN_is_not_ect(ipv6_get_dsfield(iph)))
142 from = *(__be32 *)iph;
143 to = from | htonl(INET_ECN_CE << 20);
145 if (skb->ip_summed == CHECKSUM_COMPLETE)
146 skb->csum = csum_add(csum_sub(skb->csum, (__force __wsum)from),
151 static inline int IP6_ECN_set_ect1(struct sk_buff *skb, struct ipv6hdr *iph)
155 if ((ipv6_get_dsfield(iph) & INET_ECN_MASK) != INET_ECN_ECT_0)
158 from = *(__be32 *)iph;
159 to = from ^ htonl(INET_ECN_MASK << 20);
161 if (skb->ip_summed == CHECKSUM_COMPLETE)
162 skb->csum = csum_add(csum_sub(skb->csum, (__force __wsum)from),
167 static inline void ipv6_copy_dscp(unsigned int dscp, struct ipv6hdr *inner)
169 dscp &= ~INET_ECN_MASK;
170 ipv6_change_dsfield(inner, INET_ECN_MASK, dscp);
173 static inline int INET_ECN_set_ce(struct sk_buff *skb)
175 switch (skb->protocol) {
176 case cpu_to_be16(ETH_P_IP):
177 if (skb_network_header(skb) + sizeof(struct iphdr) <=
178 skb_tail_pointer(skb))
179 return IP_ECN_set_ce(ip_hdr(skb));
182 case cpu_to_be16(ETH_P_IPV6):
183 if (skb_network_header(skb) + sizeof(struct ipv6hdr) <=
184 skb_tail_pointer(skb))
185 return IP6_ECN_set_ce(skb, ipv6_hdr(skb));
192 static inline int INET_ECN_set_ect1(struct sk_buff *skb)
194 switch (skb->protocol) {
195 case cpu_to_be16(ETH_P_IP):
196 if (skb_network_header(skb) + sizeof(struct iphdr) <=
197 skb_tail_pointer(skb))
198 return IP_ECN_set_ect1(ip_hdr(skb));
201 case cpu_to_be16(ETH_P_IPV6):
202 if (skb_network_header(skb) + sizeof(struct ipv6hdr) <=
203 skb_tail_pointer(skb))
204 return IP6_ECN_set_ect1(skb, ipv6_hdr(skb));
213 * To decapsulate the inner header at the tunnel egress, a compliant
214 * tunnel egress MUST set the outgoing ECN field to the codepoint at the
215 * intersection of the appropriate arriving inner header (row) and outer
216 * header (column) in Figure 4
218 * +---------+------------------------------------------------+
219 * |Arriving | Arriving Outer Header |
220 * | Inner +---------+------------+------------+------------+
221 * | Header | Not-ECT | ECT(0) | ECT(1) | CE |
222 * +---------+---------+------------+------------+------------+
223 * | Not-ECT | Not-ECT |Not-ECT(!!!)|Not-ECT(!!!)| <drop>(!!!)|
224 * | ECT(0) | ECT(0) | ECT(0) | ECT(1) | CE |
225 * | ECT(1) | ECT(1) | ECT(1) (!) | ECT(1) | CE |
226 * | CE | CE | CE | CE(!!!)| CE |
227 * +---------+---------+------------+------------+------------+
229 * Figure 4: New IP in IP Decapsulation Behaviour
231 * returns 0 on success
232 * 1 if something is broken and should be logged (!!! above)
233 * 2 if packet should be dropped
235 static inline int __INET_ECN_decapsulate(__u8 outer, __u8 inner, bool *set_ce)
237 if (INET_ECN_is_not_ect(inner)) {
238 switch (outer & INET_ECN_MASK) {
239 case INET_ECN_NOT_ECT:
249 *set_ce = INET_ECN_is_ce(outer);
253 static inline int INET_ECN_decapsulate(struct sk_buff *skb,
254 __u8 outer, __u8 inner)
259 rc = __INET_ECN_decapsulate(outer, inner, &set_ce);
262 INET_ECN_set_ce(skb);
263 else if ((outer & INET_ECN_MASK) == INET_ECN_ECT_1)
264 INET_ECN_set_ect1(skb);
270 static inline int IP_ECN_decapsulate(const struct iphdr *oiph,
275 if (skb->protocol == htons(ETH_P_IP))
276 inner = ip_hdr(skb)->tos;
277 else if (skb->protocol == htons(ETH_P_IPV6))
278 inner = ipv6_get_dsfield(ipv6_hdr(skb));
282 return INET_ECN_decapsulate(skb, oiph->tos, inner);
285 static inline int IP6_ECN_decapsulate(const struct ipv6hdr *oipv6h,
290 if (skb->protocol == htons(ETH_P_IP))
291 inner = ip_hdr(skb)->tos;
292 else if (skb->protocol == htons(ETH_P_IPV6))
293 inner = ipv6_get_dsfield(ipv6_hdr(skb));
297 return INET_ECN_decapsulate(skb, ipv6_get_dsfield(oipv6h), inner);