1 /* SPDX-License-Identifier: GPL-2.0 */ 1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _INET_ECN_H_ 2 #ifndef _INET_ECN_H_
3 #define _INET_ECN_H_ 3 #define _INET_ECN_H_
4 4
5 #include <linux/ip.h> 5 #include <linux/ip.h>
6 #include <linux/skbuff.h> 6 #include <linux/skbuff.h>
7 #include <linux/if_vlan.h> <<
8 7
9 #include <net/inet_sock.h> 8 #include <net/inet_sock.h>
10 #include <net/dsfield.h> 9 #include <net/dsfield.h>
11 #include <net/checksum.h> <<
12 10
13 enum { 11 enum {
14 INET_ECN_NOT_ECT = 0, 12 INET_ECN_NOT_ECT = 0,
15 INET_ECN_ECT_1 = 1, 13 INET_ECN_ECT_1 = 1,
16 INET_ECN_ECT_0 = 2, 14 INET_ECN_ECT_0 = 2,
17 INET_ECN_CE = 3, 15 INET_ECN_CE = 3,
18 INET_ECN_MASK = 3, 16 INET_ECN_MASK = 3,
19 }; 17 };
20 18
21 extern int sysctl_tunnel_ecn_log; 19 extern int sysctl_tunnel_ecn_log;
22 20
23 static inline int INET_ECN_is_ce(__u8 dsfield) 21 static inline int INET_ECN_is_ce(__u8 dsfield)
24 { 22 {
25 return (dsfield & INET_ECN_MASK) == IN 23 return (dsfield & INET_ECN_MASK) == INET_ECN_CE;
26 } 24 }
27 25
28 static inline int INET_ECN_is_not_ect(__u8 dsf 26 static inline int INET_ECN_is_not_ect(__u8 dsfield)
29 { 27 {
30 return (dsfield & INET_ECN_MASK) == IN 28 return (dsfield & INET_ECN_MASK) == INET_ECN_NOT_ECT;
31 } 29 }
32 30
33 static inline int INET_ECN_is_capable(__u8 dsf 31 static inline int INET_ECN_is_capable(__u8 dsfield)
34 { 32 {
35 return dsfield & INET_ECN_ECT_0; 33 return dsfield & INET_ECN_ECT_0;
36 } 34 }
37 35
38 /* 36 /*
39 * RFC 3168 9.1.1 37 * RFC 3168 9.1.1
40 * The full-functionality option for ECN enca 38 * The full-functionality option for ECN encapsulation is to copy the
41 * ECN codepoint of the inside header to the 39 * ECN codepoint of the inside header to the outside header on
42 * encapsulation if the inside header is not- 40 * encapsulation if the inside header is not-ECT or ECT, and to set the
43 * ECN codepoint of the outside header to ECT 41 * ECN codepoint of the outside header to ECT(0) if the ECN codepoint of
44 * the inside header is CE. 42 * the inside header is CE.
45 */ 43 */
46 static inline __u8 INET_ECN_encapsulate(__u8 o 44 static inline __u8 INET_ECN_encapsulate(__u8 outer, __u8 inner)
47 { 45 {
48 outer &= ~INET_ECN_MASK; 46 outer &= ~INET_ECN_MASK;
49 outer |= !INET_ECN_is_ce(inner) ? (inn 47 outer |= !INET_ECN_is_ce(inner) ? (inner & INET_ECN_MASK) :
50 INET 48 INET_ECN_ECT_0;
51 return outer; 49 return outer;
52 } 50 }
53 51
54 static inline void INET_ECN_xmit(struct sock * 52 static inline void INET_ECN_xmit(struct sock *sk)
55 { 53 {
56 inet_sk(sk)->tos |= INET_ECN_ECT_0; 54 inet_sk(sk)->tos |= INET_ECN_ECT_0;
57 if (inet6_sk(sk) != NULL) 55 if (inet6_sk(sk) != NULL)
58 inet6_sk(sk)->tclass |= INET_E 56 inet6_sk(sk)->tclass |= INET_ECN_ECT_0;
59 } 57 }
60 58
61 static inline void INET_ECN_dontxmit(struct so 59 static inline void INET_ECN_dontxmit(struct sock *sk)
62 { 60 {
63 inet_sk(sk)->tos &= ~INET_ECN_MASK; 61 inet_sk(sk)->tos &= ~INET_ECN_MASK;
64 if (inet6_sk(sk) != NULL) 62 if (inet6_sk(sk) != NULL)
65 inet6_sk(sk)->tclass &= ~INET_ 63 inet6_sk(sk)->tclass &= ~INET_ECN_MASK;
66 } 64 }
67 65
68 #define IP6_ECN_flow_init(label) do { 66 #define IP6_ECN_flow_init(label) do { \
69 (label) &= ~htonl(INET_ECN_MASK << 20); 67 (label) &= ~htonl(INET_ECN_MASK << 20); \
70 } while (0) 68 } while (0)
71 69
72 #define IP6_ECN_flow_xmit(sk, label) do { 70 #define IP6_ECN_flow_xmit(sk, label) do { \
73 if (INET_ECN_is_capable(inet6_sk(sk)-> 71 if (INET_ECN_is_capable(inet6_sk(sk)->tclass)) \
74 (label) |= htonl(INET_ECN_ECT_ 72 (label) |= htonl(INET_ECN_ECT_0 << 20); \
75 } while (0) 73 } while (0)
76 74
77 static inline int IP_ECN_set_ce(struct iphdr * 75 static inline int IP_ECN_set_ce(struct iphdr *iph)
78 { 76 {
>> 77 u32 check = (__force u32)iph->check;
79 u32 ecn = (iph->tos + 1) & INET_ECN_MA 78 u32 ecn = (iph->tos + 1) & INET_ECN_MASK;
80 __be16 check_add; <<
81 79
82 /* 80 /*
83 * After the last operation we have (i 81 * After the last operation we have (in binary):
84 * INET_ECN_NOT_ECT => 01 82 * INET_ECN_NOT_ECT => 01
85 * INET_ECN_ECT_1 => 10 83 * INET_ECN_ECT_1 => 10
86 * INET_ECN_ECT_0 => 11 84 * INET_ECN_ECT_0 => 11
87 * INET_ECN_CE => 00 85 * INET_ECN_CE => 00
88 */ 86 */
89 if (!(ecn & 2)) 87 if (!(ecn & 2))
90 return !ecn; 88 return !ecn;
91 89
92 /* 90 /*
93 * The following gives us: 91 * The following gives us:
94 * INET_ECN_ECT_1 => check += htons(0x 92 * INET_ECN_ECT_1 => check += htons(0xFFFD)
95 * INET_ECN_ECT_0 => check += htons(0x 93 * INET_ECN_ECT_0 => check += htons(0xFFFE)
96 */ 94 */
97 check_add = (__force __be16)((__force !! 95 check += (__force u16)htons(0xFFFB) + (__force u16)htons(ecn);
98 (__force <<
99 96
100 iph->check = csum16_add(iph->check, ch !! 97 iph->check = (__force __sum16)(check + (check>=0xFFFF));
101 iph->tos |= INET_ECN_CE; 98 iph->tos |= INET_ECN_CE;
102 return 1; 99 return 1;
103 } 100 }
104 101
105 static inline int IP_ECN_set_ect1(struct iphdr <<
106 { <<
107 if ((iph->tos & INET_ECN_MASK) != INET <<
108 return 0; <<
109 <<
110 iph->check = csum16_add(iph->check, ht <<
111 iph->tos ^= INET_ECN_MASK; <<
112 return 1; <<
113 } <<
114 <<
115 static inline void IP_ECN_clear(struct iphdr * 102 static inline void IP_ECN_clear(struct iphdr *iph)
116 { 103 {
117 iph->tos &= ~INET_ECN_MASK; 104 iph->tos &= ~INET_ECN_MASK;
118 } 105 }
119 106
120 static inline void ipv4_copy_dscp(unsigned int 107 static inline void ipv4_copy_dscp(unsigned int dscp, struct iphdr *inner)
121 { 108 {
122 dscp &= ~INET_ECN_MASK; 109 dscp &= ~INET_ECN_MASK;
123 ipv4_change_dsfield(inner, INET_ECN_MA 110 ipv4_change_dsfield(inner, INET_ECN_MASK, dscp);
124 } 111 }
125 112
126 struct ipv6hdr; 113 struct ipv6hdr;
127 114
128 /* Note: 115 /* Note:
129 * IP_ECN_set_ce() has to tweak IPV4 checksum 116 * IP_ECN_set_ce() has to tweak IPV4 checksum when setting CE,
130 * meaning both changes have no effect on skb- 117 * meaning both changes have no effect on skb->csum if/when CHECKSUM_COMPLETE
131 * In IPv6 case, no checksum compensates the c 118 * In IPv6 case, no checksum compensates the change in IPv6 header,
132 * so we have to update skb->csum. 119 * so we have to update skb->csum.
133 */ 120 */
134 static inline int IP6_ECN_set_ce(struct sk_buf 121 static inline int IP6_ECN_set_ce(struct sk_buff *skb, struct ipv6hdr *iph)
135 { 122 {
136 __be32 from, to; 123 __be32 from, to;
137 124
138 if (INET_ECN_is_not_ect(ipv6_get_dsfie 125 if (INET_ECN_is_not_ect(ipv6_get_dsfield(iph)))
139 return 0; 126 return 0;
140 127
141 from = *(__be32 *)iph; 128 from = *(__be32 *)iph;
142 to = from | htonl(INET_ECN_CE << 20); 129 to = from | htonl(INET_ECN_CE << 20);
143 *(__be32 *)iph = to; 130 *(__be32 *)iph = to;
144 if (skb->ip_summed == CHECKSUM_COMPLET 131 if (skb->ip_summed == CHECKSUM_COMPLETE)
145 skb->csum = csum_add(csum_sub( 132 skb->csum = csum_add(csum_sub(skb->csum, (__force __wsum)from),
146 (__force 133 (__force __wsum)to);
147 return 1; 134 return 1;
148 } 135 }
149 136
150 static inline int IP6_ECN_set_ect1(struct sk_b <<
151 { <<
152 __be32 from, to; <<
153 <<
154 if ((ipv6_get_dsfield(iph) & INET_ECN_ <<
155 return 0; <<
156 <<
157 from = *(__be32 *)iph; <<
158 to = from ^ htonl(INET_ECN_MASK << 20) <<
159 *(__be32 *)iph = to; <<
160 if (skb->ip_summed == CHECKSUM_COMPLET <<
161 skb->csum = csum_add(csum_sub( <<
162 (__force <<
163 return 1; <<
164 } <<
165 <<
166 static inline void ipv6_copy_dscp(unsigned int 137 static inline void ipv6_copy_dscp(unsigned int dscp, struct ipv6hdr *inner)
167 { 138 {
168 dscp &= ~INET_ECN_MASK; 139 dscp &= ~INET_ECN_MASK;
169 ipv6_change_dsfield(inner, INET_ECN_MA 140 ipv6_change_dsfield(inner, INET_ECN_MASK, dscp);
170 } 141 }
171 142
172 static inline int INET_ECN_set_ce(struct sk_bu 143 static inline int INET_ECN_set_ce(struct sk_buff *skb)
173 { 144 {
174 switch (skb_protocol(skb, true)) { !! 145 switch (skb->protocol) {
175 case cpu_to_be16(ETH_P_IP): 146 case cpu_to_be16(ETH_P_IP):
176 if (skb_network_header(skb) + 147 if (skb_network_header(skb) + sizeof(struct iphdr) <=
177 skb_tail_pointer(skb)) 148 skb_tail_pointer(skb))
178 return IP_ECN_set_ce(i 149 return IP_ECN_set_ce(ip_hdr(skb));
179 break; 150 break;
180 151
181 case cpu_to_be16(ETH_P_IPV6): 152 case cpu_to_be16(ETH_P_IPV6):
182 if (skb_network_header(skb) + 153 if (skb_network_header(skb) + sizeof(struct ipv6hdr) <=
183 skb_tail_pointer(skb)) 154 skb_tail_pointer(skb))
184 return IP6_ECN_set_ce( 155 return IP6_ECN_set_ce(skb, ipv6_hdr(skb));
185 break; 156 break;
186 } 157 }
187 158
188 return 0; 159 return 0;
189 } 160 }
190 161
191 static inline int skb_get_dsfield(struct sk_bu <<
192 { <<
193 switch (skb_protocol(skb, true)) { <<
194 case cpu_to_be16(ETH_P_IP): <<
195 if (!pskb_network_may_pull(skb <<
196 break; <<
197 return ipv4_get_dsfield(ip_hdr <<
198 <<
199 case cpu_to_be16(ETH_P_IPV6): <<
200 if (!pskb_network_may_pull(skb <<
201 break; <<
202 return ipv6_get_dsfield(ipv6_h <<
203 } <<
204 <<
205 return -1; <<
206 } <<
207 <<
208 static inline int INET_ECN_set_ect1(struct sk_ <<
209 { <<
210 switch (skb_protocol(skb, true)) { <<
211 case cpu_to_be16(ETH_P_IP): <<
212 if (skb_network_header(skb) + <<
213 skb_tail_pointer(skb)) <<
214 return IP_ECN_set_ect1 <<
215 break; <<
216 <<
217 case cpu_to_be16(ETH_P_IPV6): <<
218 if (skb_network_header(skb) + <<
219 skb_tail_pointer(skb)) <<
220 return IP6_ECN_set_ect <<
221 break; <<
222 } <<
223 <<
224 return 0; <<
225 } <<
226 <<
227 /* 162 /*
228 * RFC 6040 4.2 163 * RFC 6040 4.2
229 * To decapsulate the inner header at the tun 164 * To decapsulate the inner header at the tunnel egress, a compliant
230 * tunnel egress MUST set the outgoing ECN fi 165 * tunnel egress MUST set the outgoing ECN field to the codepoint at the
231 * intersection of the appropriate arriving i 166 * intersection of the appropriate arriving inner header (row) and outer
232 * header (column) in Figure 4 167 * header (column) in Figure 4
233 * 168 *
234 * +---------+--------------------------- 169 * +---------+------------------------------------------------+
235 * |Arriving | Arriving Outer 170 * |Arriving | Arriving Outer Header |
236 * | Inner +---------+------------+---- 171 * | Inner +---------+------------+------------+------------+
237 * | Header | Not-ECT | ECT(0) | ECT 172 * | Header | Not-ECT | ECT(0) | ECT(1) | CE |
238 * +---------+---------+------------+---- 173 * +---------+---------+------------+------------+------------+
239 * | Not-ECT | Not-ECT |Not-ECT(!!!)|Not- 174 * | Not-ECT | Not-ECT |Not-ECT(!!!)|Not-ECT(!!!)| <drop>(!!!)|
240 * | ECT(0) | ECT(0) | ECT(0) | ECT 175 * | ECT(0) | ECT(0) | ECT(0) | ECT(1) | CE |
241 * | ECT(1) | ECT(1) | ECT(1) (!) | ECT 176 * | ECT(1) | ECT(1) | ECT(1) (!) | ECT(1) | CE |
242 * | CE | CE | CE | 177 * | CE | CE | CE | CE(!!!)| CE |
243 * +---------+---------+------------+---- 178 * +---------+---------+------------+------------+------------+
244 * 179 *
245 * Figure 4: New IP in IP Decapsul 180 * Figure 4: New IP in IP Decapsulation Behaviour
246 * 181 *
247 * returns 0 on success 182 * returns 0 on success
248 * 1 if something is broken and shoul 183 * 1 if something is broken and should be logged (!!! above)
249 * 2 if packet should be dropped 184 * 2 if packet should be dropped
250 */ 185 */
251 static inline int __INET_ECN_decapsulate(__u8 !! 186 static inline int INET_ECN_decapsulate(struct sk_buff *skb,
>> 187 __u8 outer, __u8 inner)
252 { 188 {
253 if (INET_ECN_is_not_ect(inner)) { 189 if (INET_ECN_is_not_ect(inner)) {
254 switch (outer & INET_ECN_MASK) 190 switch (outer & INET_ECN_MASK) {
255 case INET_ECN_NOT_ECT: 191 case INET_ECN_NOT_ECT:
256 return 0; 192 return 0;
257 case INET_ECN_ECT_0: 193 case INET_ECN_ECT_0:
258 case INET_ECN_ECT_1: 194 case INET_ECN_ECT_1:
259 return 1; 195 return 1;
260 case INET_ECN_CE: 196 case INET_ECN_CE:
261 return 2; 197 return 2;
262 } 198 }
263 } 199 }
264 200
265 *set_ce = INET_ECN_is_ce(outer); !! 201 if (INET_ECN_is_ce(outer))
266 return 0; !! 202 INET_ECN_set_ce(skb);
267 } <<
268 <<
269 static inline int INET_ECN_decapsulate(struct <<
270 __u8 ou <<
271 { <<
272 bool set_ce = false; <<
273 int rc; <<
274 <<
275 rc = __INET_ECN_decapsulate(outer, inn <<
276 if (!rc) { <<
277 if (set_ce) <<
278 INET_ECN_set_ce(skb); <<
279 else if ((outer & INET_ECN_MAS <<
280 INET_ECN_set_ect1(skb) <<
281 } <<
282 203
283 return rc; !! 204 return 0;
284 } 205 }
285 206
286 static inline int IP_ECN_decapsulate(const str 207 static inline int IP_ECN_decapsulate(const struct iphdr *oiph,
287 struct sk 208 struct sk_buff *skb)
288 { 209 {
289 __u8 inner; 210 __u8 inner;
290 211
291 switch (skb_protocol(skb, true)) { !! 212 if (skb->protocol == htons(ETH_P_IP))
292 case htons(ETH_P_IP): <<
293 inner = ip_hdr(skb)->tos; 213 inner = ip_hdr(skb)->tos;
294 break; !! 214 else if (skb->protocol == htons(ETH_P_IPV6))
295 case htons(ETH_P_IPV6): <<
296 inner = ipv6_get_dsfield(ipv6_ 215 inner = ipv6_get_dsfield(ipv6_hdr(skb));
297 break; !! 216 else
298 default: <<
299 return 0; 217 return 0;
300 } <<
301 218
302 return INET_ECN_decapsulate(skb, oiph- 219 return INET_ECN_decapsulate(skb, oiph->tos, inner);
303 } 220 }
304 221
305 static inline int IP6_ECN_decapsulate(const st 222 static inline int IP6_ECN_decapsulate(const struct ipv6hdr *oipv6h,
306 struct s 223 struct sk_buff *skb)
307 { 224 {
308 __u8 inner; 225 __u8 inner;
309 226
310 switch (skb_protocol(skb, true)) { !! 227 if (skb->protocol == htons(ETH_P_IP))
311 case htons(ETH_P_IP): <<
312 inner = ip_hdr(skb)->tos; 228 inner = ip_hdr(skb)->tos;
313 break; !! 229 else if (skb->protocol == htons(ETH_P_IPV6))
314 case htons(ETH_P_IPV6): <<
315 inner = ipv6_get_dsfield(ipv6_ 230 inner = ipv6_get_dsfield(ipv6_hdr(skb));
316 break; !! 231 else
317 default: <<
318 return 0; 232 return 0;
319 } <<
320 233
321 return INET_ECN_decapsulate(skb, ipv6_ 234 return INET_ECN_decapsulate(skb, ipv6_get_dsfield(oipv6h), inner);
322 } 235 }
323 #endif 236 #endif
324 237
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.