1 /* SPDX-License-Identifier: GPL-2.0 */ <<
2 #ifndef _INET_ECN_H_ 1 #ifndef _INET_ECN_H_
3 #define _INET_ECN_H_ 2 #define _INET_ECN_H_
4 3
5 #include <linux/ip.h> 4 #include <linux/ip.h>
6 #include <linux/skbuff.h> 5 #include <linux/skbuff.h>
7 #include <linux/if_vlan.h> <<
8 6
9 #include <net/inet_sock.h> 7 #include <net/inet_sock.h>
10 #include <net/dsfield.h> 8 #include <net/dsfield.h>
11 #include <net/checksum.h> <<
12 9
13 enum { 10 enum {
14 INET_ECN_NOT_ECT = 0, 11 INET_ECN_NOT_ECT = 0,
15 INET_ECN_ECT_1 = 1, 12 INET_ECN_ECT_1 = 1,
16 INET_ECN_ECT_0 = 2, 13 INET_ECN_ECT_0 = 2,
17 INET_ECN_CE = 3, 14 INET_ECN_CE = 3,
18 INET_ECN_MASK = 3, 15 INET_ECN_MASK = 3,
19 }; 16 };
20 17
21 extern int sysctl_tunnel_ecn_log; 18 extern int sysctl_tunnel_ecn_log;
22 19
23 static inline int INET_ECN_is_ce(__u8 dsfield) 20 static inline int INET_ECN_is_ce(__u8 dsfield)
24 { 21 {
25 return (dsfield & INET_ECN_MASK) == IN 22 return (dsfield & INET_ECN_MASK) == INET_ECN_CE;
26 } 23 }
27 24
28 static inline int INET_ECN_is_not_ect(__u8 dsf 25 static inline int INET_ECN_is_not_ect(__u8 dsfield)
29 { 26 {
30 return (dsfield & INET_ECN_MASK) == IN 27 return (dsfield & INET_ECN_MASK) == INET_ECN_NOT_ECT;
31 } 28 }
32 29
33 static inline int INET_ECN_is_capable(__u8 dsf 30 static inline int INET_ECN_is_capable(__u8 dsfield)
34 { 31 {
35 return dsfield & INET_ECN_ECT_0; 32 return dsfield & INET_ECN_ECT_0;
36 } 33 }
37 34
38 /* 35 /*
39 * RFC 3168 9.1.1 36 * RFC 3168 9.1.1
40 * The full-functionality option for ECN enca 37 * The full-functionality option for ECN encapsulation is to copy the
41 * ECN codepoint of the inside header to the 38 * ECN codepoint of the inside header to the outside header on
42 * encapsulation if the inside header is not- 39 * encapsulation if the inside header is not-ECT or ECT, and to set the
43 * ECN codepoint of the outside header to ECT 40 * ECN codepoint of the outside header to ECT(0) if the ECN codepoint of
44 * the inside header is CE. 41 * the inside header is CE.
45 */ 42 */
46 static inline __u8 INET_ECN_encapsulate(__u8 o 43 static inline __u8 INET_ECN_encapsulate(__u8 outer, __u8 inner)
47 { 44 {
48 outer &= ~INET_ECN_MASK; 45 outer &= ~INET_ECN_MASK;
49 outer |= !INET_ECN_is_ce(inner) ? (inn 46 outer |= !INET_ECN_is_ce(inner) ? (inner & INET_ECN_MASK) :
50 INET 47 INET_ECN_ECT_0;
51 return outer; 48 return outer;
52 } 49 }
53 50
54 static inline void INET_ECN_xmit(struct sock * 51 static inline void INET_ECN_xmit(struct sock *sk)
55 { 52 {
56 inet_sk(sk)->tos |= INET_ECN_ECT_0; 53 inet_sk(sk)->tos |= INET_ECN_ECT_0;
57 if (inet6_sk(sk) != NULL) 54 if (inet6_sk(sk) != NULL)
58 inet6_sk(sk)->tclass |= INET_E 55 inet6_sk(sk)->tclass |= INET_ECN_ECT_0;
59 } 56 }
60 57
61 static inline void INET_ECN_dontxmit(struct so 58 static inline void INET_ECN_dontxmit(struct sock *sk)
62 { 59 {
63 inet_sk(sk)->tos &= ~INET_ECN_MASK; 60 inet_sk(sk)->tos &= ~INET_ECN_MASK;
64 if (inet6_sk(sk) != NULL) 61 if (inet6_sk(sk) != NULL)
65 inet6_sk(sk)->tclass &= ~INET_ 62 inet6_sk(sk)->tclass &= ~INET_ECN_MASK;
66 } 63 }
67 64
68 #define IP6_ECN_flow_init(label) do { 65 #define IP6_ECN_flow_init(label) do { \
69 (label) &= ~htonl(INET_ECN_MASK << 20); 66 (label) &= ~htonl(INET_ECN_MASK << 20); \
70 } while (0) 67 } while (0)
71 68
72 #define IP6_ECN_flow_xmit(sk, label) do { 69 #define IP6_ECN_flow_xmit(sk, label) do { \
73 if (INET_ECN_is_capable(inet6_sk(sk)-> 70 if (INET_ECN_is_capable(inet6_sk(sk)->tclass)) \
74 (label) |= htonl(INET_ECN_ECT_ 71 (label) |= htonl(INET_ECN_ECT_0 << 20); \
75 } while (0) 72 } while (0)
76 73
77 static inline int IP_ECN_set_ce(struct iphdr * 74 static inline int IP_ECN_set_ce(struct iphdr *iph)
78 { 75 {
>> 76 u32 check = (__force u32)iph->check;
79 u32 ecn = (iph->tos + 1) & INET_ECN_MA 77 u32 ecn = (iph->tos + 1) & INET_ECN_MASK;
80 __be16 check_add; <<
81 78
82 /* 79 /*
83 * After the last operation we have (i 80 * After the last operation we have (in binary):
84 * INET_ECN_NOT_ECT => 01 81 * INET_ECN_NOT_ECT => 01
85 * INET_ECN_ECT_1 => 10 82 * INET_ECN_ECT_1 => 10
86 * INET_ECN_ECT_0 => 11 83 * INET_ECN_ECT_0 => 11
87 * INET_ECN_CE => 00 84 * INET_ECN_CE => 00
88 */ 85 */
89 if (!(ecn & 2)) 86 if (!(ecn & 2))
90 return !ecn; 87 return !ecn;
91 88
92 /* 89 /*
93 * The following gives us: 90 * The following gives us:
94 * INET_ECN_ECT_1 => check += htons(0x 91 * INET_ECN_ECT_1 => check += htons(0xFFFD)
95 * INET_ECN_ECT_0 => check += htons(0x 92 * INET_ECN_ECT_0 => check += htons(0xFFFE)
96 */ 93 */
97 check_add = (__force __be16)((__force !! 94 check += (__force u16)htons(0xFFFB) + (__force u16)htons(ecn);
98 (__force <<
99 95
100 iph->check = csum16_add(iph->check, ch !! 96 iph->check = (__force __sum16)(check + (check>=0xFFFF));
101 iph->tos |= INET_ECN_CE; 97 iph->tos |= INET_ECN_CE;
102 return 1; 98 return 1;
103 } 99 }
104 100
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 * 101 static inline void IP_ECN_clear(struct iphdr *iph)
116 { 102 {
117 iph->tos &= ~INET_ECN_MASK; 103 iph->tos &= ~INET_ECN_MASK;
118 } 104 }
119 105
120 static inline void ipv4_copy_dscp(unsigned int 106 static inline void ipv4_copy_dscp(unsigned int dscp, struct iphdr *inner)
121 { 107 {
122 dscp &= ~INET_ECN_MASK; 108 dscp &= ~INET_ECN_MASK;
123 ipv4_change_dsfield(inner, INET_ECN_MA 109 ipv4_change_dsfield(inner, INET_ECN_MASK, dscp);
124 } 110 }
125 111
126 struct ipv6hdr; 112 struct ipv6hdr;
127 113
128 /* Note: 114 /* Note:
129 * IP_ECN_set_ce() has to tweak IPV4 checksum 115 * IP_ECN_set_ce() has to tweak IPV4 checksum when setting CE,
130 * meaning both changes have no effect on skb- 116 * meaning both changes have no effect on skb->csum if/when CHECKSUM_COMPLETE
131 * In IPv6 case, no checksum compensates the c 117 * In IPv6 case, no checksum compensates the change in IPv6 header,
132 * so we have to update skb->csum. 118 * so we have to update skb->csum.
133 */ 119 */
134 static inline int IP6_ECN_set_ce(struct sk_buf 120 static inline int IP6_ECN_set_ce(struct sk_buff *skb, struct ipv6hdr *iph)
135 { 121 {
136 __be32 from, to; 122 __be32 from, to;
137 123
138 if (INET_ECN_is_not_ect(ipv6_get_dsfie 124 if (INET_ECN_is_not_ect(ipv6_get_dsfield(iph)))
139 return 0; 125 return 0;
140 126
141 from = *(__be32 *)iph; 127 from = *(__be32 *)iph;
142 to = from | htonl(INET_ECN_CE << 20); 128 to = from | htonl(INET_ECN_CE << 20);
143 *(__be32 *)iph = to; 129 *(__be32 *)iph = to;
144 if (skb->ip_summed == CHECKSUM_COMPLET 130 if (skb->ip_summed == CHECKSUM_COMPLETE)
145 skb->csum = csum_add(csum_sub( 131 skb->csum = csum_add(csum_sub(skb->csum, (__force __wsum)from),
146 (__force 132 (__force __wsum)to);
147 return 1; 133 return 1;
148 } 134 }
149 135
150 static inline int IP6_ECN_set_ect1(struct sk_b !! 136 static inline void IP6_ECN_clear(struct ipv6hdr *iph)
151 { 137 {
152 __be32 from, to; !! 138 *(__be32*)iph &= ~htonl(INET_ECN_MASK << 20);
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 } 139 }
165 140
166 static inline void ipv6_copy_dscp(unsigned int 141 static inline void ipv6_copy_dscp(unsigned int dscp, struct ipv6hdr *inner)
167 { 142 {
168 dscp &= ~INET_ECN_MASK; 143 dscp &= ~INET_ECN_MASK;
169 ipv6_change_dsfield(inner, INET_ECN_MA 144 ipv6_change_dsfield(inner, INET_ECN_MASK, dscp);
170 } 145 }
171 146
172 static inline int INET_ECN_set_ce(struct sk_bu 147 static inline int INET_ECN_set_ce(struct sk_buff *skb)
173 { 148 {
174 switch (skb_protocol(skb, true)) { !! 149 switch (skb->protocol) {
175 case cpu_to_be16(ETH_P_IP): 150 case cpu_to_be16(ETH_P_IP):
176 if (skb_network_header(skb) + 151 if (skb_network_header(skb) + sizeof(struct iphdr) <=
177 skb_tail_pointer(skb)) 152 skb_tail_pointer(skb))
178 return IP_ECN_set_ce(i 153 return IP_ECN_set_ce(ip_hdr(skb));
179 break; 154 break;
180 155
181 case cpu_to_be16(ETH_P_IPV6): 156 case cpu_to_be16(ETH_P_IPV6):
182 if (skb_network_header(skb) + 157 if (skb_network_header(skb) + sizeof(struct ipv6hdr) <=
183 skb_tail_pointer(skb)) 158 skb_tail_pointer(skb))
184 return IP6_ECN_set_ce( 159 return IP6_ECN_set_ce(skb, ipv6_hdr(skb));
185 break; 160 break;
186 } 161 }
187 162
188 return 0; 163 return 0;
189 } 164 }
190 165
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 /* 166 /*
228 * RFC 6040 4.2 167 * RFC 6040 4.2
229 * To decapsulate the inner header at the tun 168 * To decapsulate the inner header at the tunnel egress, a compliant
230 * tunnel egress MUST set the outgoing ECN fi 169 * tunnel egress MUST set the outgoing ECN field to the codepoint at the
231 * intersection of the appropriate arriving i 170 * intersection of the appropriate arriving inner header (row) and outer
232 * header (column) in Figure 4 171 * header (column) in Figure 4
233 * 172 *
234 * +---------+--------------------------- 173 * +---------+------------------------------------------------+
235 * |Arriving | Arriving Outer 174 * |Arriving | Arriving Outer Header |
236 * | Inner +---------+------------+---- 175 * | Inner +---------+------------+------------+------------+
237 * | Header | Not-ECT | ECT(0) | ECT 176 * | Header | Not-ECT | ECT(0) | ECT(1) | CE |
238 * +---------+---------+------------+---- 177 * +---------+---------+------------+------------+------------+
239 * | Not-ECT | Not-ECT |Not-ECT(!!!)|Not- 178 * | Not-ECT | Not-ECT |Not-ECT(!!!)|Not-ECT(!!!)| <drop>(!!!)|
240 * | ECT(0) | ECT(0) | ECT(0) | ECT 179 * | ECT(0) | ECT(0) | ECT(0) | ECT(1) | CE |
241 * | ECT(1) | ECT(1) | ECT(1) (!) | ECT 180 * | ECT(1) | ECT(1) | ECT(1) (!) | ECT(1) | CE |
242 * | CE | CE | CE | 181 * | CE | CE | CE | CE(!!!)| CE |
243 * +---------+---------+------------+---- 182 * +---------+---------+------------+------------+------------+
244 * 183 *
245 * Figure 4: New IP in IP Decapsul 184 * Figure 4: New IP in IP Decapsulation Behaviour
246 * 185 *
247 * returns 0 on success 186 * returns 0 on success
248 * 1 if something is broken and shoul 187 * 1 if something is broken and should be logged (!!! above)
249 * 2 if packet should be dropped 188 * 2 if packet should be dropped
250 */ 189 */
251 static inline int __INET_ECN_decapsulate(__u8 !! 190 static inline int INET_ECN_decapsulate(struct sk_buff *skb,
>> 191 __u8 outer, __u8 inner)
252 { 192 {
253 if (INET_ECN_is_not_ect(inner)) { 193 if (INET_ECN_is_not_ect(inner)) {
254 switch (outer & INET_ECN_MASK) 194 switch (outer & INET_ECN_MASK) {
255 case INET_ECN_NOT_ECT: 195 case INET_ECN_NOT_ECT:
256 return 0; 196 return 0;
257 case INET_ECN_ECT_0: 197 case INET_ECN_ECT_0:
258 case INET_ECN_ECT_1: 198 case INET_ECN_ECT_1:
259 return 1; 199 return 1;
260 case INET_ECN_CE: 200 case INET_ECN_CE:
261 return 2; 201 return 2;
262 } 202 }
263 } 203 }
264 204
265 *set_ce = INET_ECN_is_ce(outer); !! 205 if (INET_ECN_is_ce(outer))
266 return 0; !! 206 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 } <<
282 207
283 return rc; !! 208 return 0;
284 } 209 }
285 210
286 static inline int IP_ECN_decapsulate(const str 211 static inline int IP_ECN_decapsulate(const struct iphdr *oiph,
287 struct sk 212 struct sk_buff *skb)
288 { 213 {
289 __u8 inner; 214 __u8 inner;
290 215
291 switch (skb_protocol(skb, true)) { !! 216 if (skb->protocol == htons(ETH_P_IP))
292 case htons(ETH_P_IP): <<
293 inner = ip_hdr(skb)->tos; 217 inner = ip_hdr(skb)->tos;
294 break; !! 218 else if (skb->protocol == htons(ETH_P_IPV6))
295 case htons(ETH_P_IPV6): <<
296 inner = ipv6_get_dsfield(ipv6_ 219 inner = ipv6_get_dsfield(ipv6_hdr(skb));
297 break; !! 220 else
298 default: <<
299 return 0; 221 return 0;
300 } <<
301 222
302 return INET_ECN_decapsulate(skb, oiph- 223 return INET_ECN_decapsulate(skb, oiph->tos, inner);
303 } 224 }
304 225
305 static inline int IP6_ECN_decapsulate(const st 226 static inline int IP6_ECN_decapsulate(const struct ipv6hdr *oipv6h,
306 struct s 227 struct sk_buff *skb)
307 { 228 {
308 __u8 inner; 229 __u8 inner;
309 230
310 switch (skb_protocol(skb, true)) { !! 231 if (skb->protocol == htons(ETH_P_IP))
311 case htons(ETH_P_IP): <<
312 inner = ip_hdr(skb)->tos; 232 inner = ip_hdr(skb)->tos;
313 break; !! 233 else if (skb->protocol == htons(ETH_P_IPV6))
314 case htons(ETH_P_IPV6): <<
315 inner = ipv6_get_dsfield(ipv6_ 234 inner = ipv6_get_dsfield(ipv6_hdr(skb));
316 break; !! 235 else
317 default: <<
318 return 0; 236 return 0;
319 } <<
320 237
321 return INET_ECN_decapsulate(skb, ipv6_ 238 return INET_ECN_decapsulate(skb, ipv6_get_dsfield(oipv6h), inner);
322 } 239 }
323 #endif 240 #endif
324 241
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.