| @@ -1,255 +1,257 @@ | | | @@ -1,255 +1,257 @@ |
1 | /* $NetBSD: in_offload.c,v 1.4 2011/04/14 15:53:36 yamt Exp $ */ | | 1 | /* $NetBSD: in_offload.c,v 1.5 2011/04/25 22:11:31 yamt Exp $ */ |
2 | | | 2 | |
3 | /*- | | 3 | /*- |
4 | * Copyright (c)2005, 2006 YAMAMOTO Takashi, | | 4 | * Copyright (c)2005, 2006 YAMAMOTO Takashi, |
5 | * All rights reserved. | | 5 | * All rights reserved. |
6 | * | | 6 | * |
7 | * Redistribution and use in source and binary forms, with or without | | 7 | * Redistribution and use in source and binary forms, with or without |
8 | * modification, are permitted provided that the following conditions | | 8 | * modification, are permitted provided that the following conditions |
9 | * are met: | | 9 | * are met: |
10 | * 1. Redistributions of source code must retain the above copyright | | 10 | * 1. Redistributions of source code must retain the above copyright |
11 | * notice, this list of conditions and the following disclaimer. | | 11 | * notice, this list of conditions and the following disclaimer. |
12 | * 2. Redistributions in binary form must reproduce the above copyright | | 12 | * 2. Redistributions in binary form must reproduce the above copyright |
13 | * notice, this list of conditions and the following disclaimer in the | | 13 | * notice, this list of conditions and the following disclaimer in the |
14 | * documentation and/or other materials provided with the distribution. | | 14 | * documentation and/or other materials provided with the distribution. |
15 | * | | 15 | * |
16 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND | | 16 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND |
17 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | | 17 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
18 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | | 18 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
19 | * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE | | 19 | * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
20 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | | 20 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
21 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | | 21 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
22 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | | 22 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
23 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | | 23 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
24 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | | 24 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
25 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | | 25 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
26 | * SUCH DAMAGE. | | 26 | * SUCH DAMAGE. |
27 | */ | | 27 | */ |
28 | | | 28 | |
29 | #include <sys/cdefs.h> | | 29 | #include <sys/cdefs.h> |
30 | __KERNEL_RCSID(0, "$NetBSD: in_offload.c,v 1.4 2011/04/14 15:53:36 yamt Exp $"); | | 30 | __KERNEL_RCSID(0, "$NetBSD: in_offload.c,v 1.5 2011/04/25 22:11:31 yamt Exp $"); |
31 | | | 31 | |
32 | #include <sys/param.h> | | 32 | #include <sys/param.h> |
33 | #include <sys/mbuf.h> | | 33 | #include <sys/mbuf.h> |
34 | | | 34 | |
35 | #include <net/if.h> | | 35 | #include <net/if.h> |
36 | | | 36 | |
37 | #include <netinet/in.h> | | 37 | #include <netinet/in.h> |
38 | #include <netinet/in_systm.h> | | 38 | #include <netinet/in_systm.h> |
39 | #include <netinet/ip.h> | | 39 | #include <netinet/ip.h> |
40 | #include <netinet/tcp.h> | | 40 | #include <netinet/tcp.h> |
41 | #include <netinet/in_offload.h> | | 41 | #include <netinet/in_offload.h> |
42 | | | 42 | |
43 | struct ip_tso_output_args { | | 43 | struct ip_tso_output_args { |
44 | struct ifnet *ifp; | | 44 | struct ifnet *ifp; |
45 | const struct sockaddr *sa; | | 45 | const struct sockaddr *sa; |
46 | struct rtentry *rt; | | 46 | struct rtentry *rt; |
47 | }; | | 47 | }; |
48 | | | 48 | |
49 | static int ip_tso_output_callback(void *, struct mbuf *); | | 49 | static int ip_tso_output_callback(void *, struct mbuf *); |
50 | | | 50 | |
51 | static int | | 51 | static int |
52 | ip_tso_output_callback(void *vp, struct mbuf *m) | | 52 | ip_tso_output_callback(void *vp, struct mbuf *m) |
53 | { | | 53 | { |
54 | struct ip_tso_output_args *args = vp; | | 54 | struct ip_tso_output_args *args = vp; |
55 | struct ifnet *ifp = args->ifp; | | 55 | struct ifnet *ifp = args->ifp; |
56 | int error; | | 56 | int error; |
57 | | | 57 | |
58 | KERNEL_LOCK(1, NULL); | | 58 | KERNEL_LOCK(1, NULL); |
59 | error = (*ifp->if_output)(ifp, m, args->sa, args->rt); | | 59 | error = (*ifp->if_output)(ifp, m, args->sa, args->rt); |
60 | KERNEL_UNLOCK_ONE(NULL); | | 60 | KERNEL_UNLOCK_ONE(NULL); |
61 | return error; | | 61 | return error; |
62 | } | | 62 | } |
63 | | | 63 | |
64 | int | | 64 | int |
65 | ip_tso_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, | | 65 | ip_tso_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, |
66 | struct rtentry *rt) | | 66 | struct rtentry *rt) |
67 | { | | 67 | { |
68 | struct ip_tso_output_args args; | | 68 | struct ip_tso_output_args args; |
69 | | | 69 | |
70 | args.ifp = ifp; | | 70 | args.ifp = ifp; |
71 | args.sa = sa; | | 71 | args.sa = sa; |
72 | args.rt = rt; | | 72 | args.rt = rt; |
73 | | | 73 | |
74 | return tcp4_segment(m, ip_tso_output_callback, &args); | | 74 | return tcp4_segment(m, ip_tso_output_callback, &args); |
75 | } | | 75 | } |
76 | | | 76 | |
77 | /* | | 77 | /* |
78 | * tcp4_segment: handle M_CSUM_TSOv4 by software. | | 78 | * tcp4_segment: handle M_CSUM_TSOv4 by software. |
79 | * | | 79 | * |
80 | * => always consume m. | | 80 | * => always consume m. |
81 | * => call output_func with output_arg for each segments. | | 81 | * => call output_func with output_arg for each segments. |
82 | */ | | 82 | */ |
83 | | | 83 | |
84 | int | | 84 | int |
85 | tcp4_segment(struct mbuf *m, int (*output_func)(void *, struct mbuf *), | | 85 | tcp4_segment(struct mbuf *m, int (*output_func)(void *, struct mbuf *), |
86 | void *output_arg) | | 86 | void *output_arg) |
87 | { | | 87 | { |
88 | int mss; | | 88 | int mss; |
89 | int iphlen; | | 89 | int iphlen; |
90 | int thlen; | | 90 | int thlen; |
91 | int hlen; | | 91 | int hlen; |
92 | int len; | | 92 | int len; |
93 | struct ip *iph; | | 93 | struct ip *iph; |
94 | struct tcphdr *th; | | 94 | struct tcphdr *th; |
95 | uint16_t ipid; | | 95 | uint16_t ipid; |
96 | uint32_t tcpseq; | | 96 | uint32_t tcpseq; |
97 | struct mbuf *hdr = NULL; | | 97 | struct mbuf *hdr = NULL; |
98 | struct mbuf *t; | | 98 | struct mbuf *t; |
99 | int error = 0; | | 99 | int error = 0; |
100 | | | 100 | |
101 | KASSERT((m->m_flags & M_PKTHDR) != 0); | | 101 | KASSERT((m->m_flags & M_PKTHDR) != 0); |
102 | KASSERT((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0); | | 102 | KASSERT((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0); |
103 | | | 103 | |
104 | m->m_pkthdr.csum_flags = 0; | | 104 | m->m_pkthdr.csum_flags = 0; |
105 | | | 105 | |
106 | len = m->m_pkthdr.len; | | 106 | len = m->m_pkthdr.len; |
107 | KASSERT(len >= sizeof(*iph) + sizeof(*th)); | | 107 | KASSERT(len >= sizeof(*iph) + sizeof(*th)); |
108 | | | 108 | |
109 | if (m->m_len < sizeof(*iph)) { | | 109 | if (m->m_len < sizeof(*iph)) { |
110 | m = m_pullup(m, sizeof(*iph)); | | 110 | m = m_pullup(m, sizeof(*iph)); |
111 | if (m == NULL) { | | 111 | if (m == NULL) { |
112 | error = ENOMEM; | | 112 | error = ENOMEM; |
113 | goto quit; | | 113 | goto quit; |
114 | } | | 114 | } |
115 | } | | 115 | } |
116 | iph = mtod(m, struct ip *); | | 116 | iph = mtod(m, struct ip *); |
117 | iphlen = iph->ip_hl * 4; | | 117 | iphlen = iph->ip_hl * 4; |
118 | KASSERT(iph->ip_v == IPVERSION); | | 118 | KASSERT(iph->ip_v == IPVERSION); |
119 | KASSERT(iphlen >= sizeof(*iph)); | | 119 | KASSERT(iphlen >= sizeof(*iph)); |
120 | KASSERT(iph->ip_p == IPPROTO_TCP); | | 120 | KASSERT(iph->ip_p == IPPROTO_TCP); |
121 | ipid = ntohs(iph->ip_id); | | 121 | ipid = ntohs(iph->ip_id); |
122 | | | 122 | |
123 | hlen = iphlen + sizeof(*th); | | 123 | hlen = iphlen + sizeof(*th); |
124 | if (m->m_len < hlen) { | | 124 | if (m->m_len < hlen) { |
125 | m = m_pullup(m, hlen); | | 125 | m = m_pullup(m, hlen); |
126 | if (m == NULL) { | | 126 | if (m == NULL) { |
127 | error = ENOMEM; | | 127 | error = ENOMEM; |
128 | goto quit; | | 128 | goto quit; |
129 | } | | 129 | } |
130 | } | | 130 | } |
131 | th = (void *)(mtod(m, char *) + iphlen); | | 131 | th = (void *)(mtod(m, char *) + iphlen); |
132 | tcpseq = ntohl(th->th_seq); | | 132 | tcpseq = ntohl(th->th_seq); |
133 | thlen = th->th_off * 4; | | 133 | thlen = th->th_off * 4; |
134 | hlen = iphlen + thlen; | | 134 | hlen = iphlen + thlen; |
135 | | | 135 | |
136 | mss = m->m_pkthdr.segsz; | | 136 | mss = m->m_pkthdr.segsz; |
137 | KASSERT(mss != 0); | | 137 | KASSERT(mss != 0); |
138 | KASSERT(len > hlen); | | 138 | KASSERT(len > hlen); |
139 | | | 139 | |
140 | t = m_split(m, hlen, M_NOWAIT); | | 140 | t = m_split(m, hlen, M_NOWAIT); |
141 | if (t == NULL) { | | 141 | if (t == NULL) { |
142 | error = ENOMEM; | | 142 | error = ENOMEM; |
143 | goto quit; | | 143 | goto quit; |
144 | } | | 144 | } |
145 | hdr = m; | | 145 | hdr = m; |
146 | m = t; | | 146 | m = t; |
147 | len -= hlen; | | 147 | len -= hlen; |
148 | KASSERT(len % mss == 0); | | 148 | KASSERT(len % mss == 0); |
149 | while (len > 0) { | | 149 | while (len > 0) { |
150 | struct mbuf *n; | | 150 | struct mbuf *n; |
151 | | | 151 | |
152 | n = m_dup(hdr, 0, hlen, M_NOWAIT); | | 152 | n = m_dup(hdr, 0, hlen, M_NOWAIT); |
153 | if (n == NULL) { | | 153 | if (n == NULL) { |
154 | error = ENOMEM; | | 154 | error = ENOMEM; |
155 | goto quit; | | 155 | goto quit; |
156 | } | | 156 | } |
157 | KASSERT(n->m_len == hlen); /* XXX */ | | 157 | KASSERT(n->m_len == hlen); /* XXX */ |
158 | | | 158 | |
159 | t = m_split(m, mss, M_NOWAIT); | | 159 | t = m_split(m, mss, M_NOWAIT); |
160 | if (t == NULL) { | | 160 | if (t == NULL) { |
161 | m_freem(n); | | 161 | m_freem(n); |
162 | error = ENOMEM; | | 162 | error = ENOMEM; |
163 | goto quit; | | 163 | goto quit; |
164 | } | | 164 | } |
165 | m_cat(n, m); | | 165 | m_cat(n, m); |
166 | m = t; | | 166 | m = t; |
167 | | | 167 | |
168 | KASSERT(n->m_len >= hlen); /* XXX */ | | 168 | KASSERT(n->m_len >= hlen); /* XXX */ |
169 | | | 169 | |
170 | n->m_pkthdr.len = hlen + mss; | | 170 | n->m_pkthdr.len = hlen + mss; |
171 | iph = mtod(n, struct ip *); | | 171 | iph = mtod(n, struct ip *); |
172 | KASSERT(iph->ip_v == IPVERSION); | | 172 | KASSERT(iph->ip_v == IPVERSION); |
173 | iph->ip_len = htons(n->m_pkthdr.len); | | 173 | iph->ip_len = htons(n->m_pkthdr.len); |
174 | iph->ip_id = htons(ipid); | | 174 | iph->ip_id = htons(ipid); |
175 | th = (void *)(mtod(n, char *) + iphlen); | | 175 | th = (void *)(mtod(n, char *) + iphlen); |
176 | th->th_seq = htonl(tcpseq); | | 176 | th->th_seq = htonl(tcpseq); |
177 | iph->ip_sum = 0; | | 177 | iph->ip_sum = 0; |
178 | iph->ip_sum = in_cksum(n, iphlen); | | 178 | iph->ip_sum = in_cksum(n, iphlen); |
179 | th->th_sum = 0; | | 179 | th->th_sum = 0; |
180 | th->th_sum = in4_cksum(n, IPPROTO_TCP, iphlen, thlen + mss); | | 180 | th->th_sum = in4_cksum(n, IPPROTO_TCP, iphlen, thlen + mss); |
181 | | | 181 | |
182 | error = (*output_func)(output_arg, n); | | 182 | error = (*output_func)(output_arg, n); |
183 | if (error) { | | 183 | if (error) { |
184 | goto quit; | | 184 | goto quit; |
185 | } | | 185 | } |
186 | | | 186 | |
187 | tcpseq += mss; | | 187 | tcpseq += mss; |
188 | ipid++; | | 188 | ipid++; |
189 | len -= mss; | | 189 | len -= mss; |
190 | } | | 190 | } |
191 | | | 191 | |
192 | quit: | | 192 | quit: |
193 | if (hdr != NULL) { | | 193 | if (hdr != NULL) { |
194 | m_freem(hdr); | | 194 | m_freem(hdr); |
195 | } | | 195 | } |
196 | if (m != NULL) { | | 196 | if (m != NULL) { |
197 | m_freem(m); | | 197 | m_freem(m); |
198 | } | | 198 | } |
199 | | | 199 | |
200 | return error; | | 200 | return error; |
201 | } | | 201 | } |
202 | | | 202 | |
203 | void | | 203 | void |
204 | ip_undefer_csum(struct mbuf *m, size_t hdrlen, int csum_flags) | | 204 | ip_undefer_csum(struct mbuf *m, size_t hdrlen, int csum_flags) |
205 | { | | 205 | { |
206 | KASSERT(m->m_flags & M_PKTHDR); | | 206 | const size_t iphdrlen = M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data); |
207 | KASSERT((m->m_pkthdr.csum_flags & csum_flags) == csum_flags); | | | |
208 | uint16_t csum; | | 207 | uint16_t csum; |
209 | uint16_t ip_len; | | 208 | uint16_t ip_len; |
210 | uint16_t *csump; | | 209 | uint16_t *csump; |
211 | size_t iphdrlen; | | 210 | |
| | | 211 | KASSERT(m->m_flags & M_PKTHDR); |
| | | 212 | KASSERT((m->m_pkthdr.csum_flags & csum_flags) == csum_flags); |
212 | | | 213 | |
213 | if (__predict_true(hdrlen + sizeof(struct ip) <= m->m_len)) { | | 214 | if (__predict_true(hdrlen + sizeof(struct ip) <= m->m_len)) { |
214 | struct ip *ip = (struct ip *)(mtod(m, uint8_t *) + hdrlen); | | 215 | struct ip *ip = (struct ip *)(mtod(m, uint8_t *) + hdrlen); |
| | | 216 | |
215 | ip_len = ip->ip_len; | | 217 | ip_len = ip->ip_len; |
216 | iphdrlen = ip->ip_hl << 2; | | | |
217 | csump = &ip->ip_sum; | | 218 | csump = &ip->ip_sum; |
218 | } else { | | 219 | } else { |
219 | uint8_t ip_vhl; | | 220 | const size_t ip_len_offset = |
220 | const size_t ip_len_offset = hdrlen + offsetof(struct ip, ip_len); | | 221 | hdrlen + offsetof(struct ip, ip_len); |
221 | m_copydata(m, hdrlen, sizeof(ip_vhl), &ip_vhl); | | 222 | |
222 | m_copydata(m, ip_len_offset, sizeof(ip_len), &ip_len); | | 223 | m_copydata(m, ip_len_offset, sizeof(ip_len), &ip_len); |
223 | iphdrlen = (ip_vhl & 0x0f) << 2; | | | |
224 | csump = NULL; | | 224 | csump = NULL; |
225 | } | | 225 | } |
| | | 226 | ip_len = ntohs(ip_len); |
226 | | | 227 | |
227 | if (csum_flags & M_CSUM_IPv4) { | | 228 | if (csum_flags & M_CSUM_IPv4) { |
228 | const size_t offset = hdrlen + offsetof(struct ip, ip_sum); | | | |
229 | csum = in4_cksum(m, 0, hdrlen, iphdrlen); | | 229 | csum = in4_cksum(m, 0, hdrlen, iphdrlen); |
230 | if (csump != NULL) { | | 230 | if (csump != NULL) { |
231 | *csump = csum; | | 231 | *csump = csum; |
232 | } else { | | 232 | } else { |
| | | 233 | const size_t offset = hdrlen + |
| | | 234 | offsetof(struct ip, ip_sum); |
| | | 235 | |
233 | m_copyback(m, offset, sizeof(uint16_t), &csum); | | 236 | m_copyback(m, offset, sizeof(uint16_t), &csum); |
234 | } | | 237 | } |
235 | } | | 238 | } |
236 | | | 239 | |
237 | if (csum_flags & (M_CSUM_UDPv4|M_CSUM_TCPv4)) { | | 240 | if (csum_flags & (M_CSUM_UDPv4|M_CSUM_TCPv4)) { |
238 | size_t l4offset = hdrlen | | 241 | size_t l4offset = hdrlen + iphdrlen; |
239 | + M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data); | | | |
240 | | | 242 | |
241 | csum = in4_cksum(m, 0, hdrlen + l4offset, ip_len - l4offset); | | 243 | csum = in4_cksum(m, 0, l4offset, ip_len - l4offset - hdrlen); |
242 | if (csum == 0 && (csum_flags & M_CSUM_UDPv4) != 0) | | 244 | if (csum == 0 && (csum_flags & M_CSUM_UDPv4) != 0) |
243 | csum = 0xffff; | | 245 | csum = 0xffff; |
244 | | | 246 | |
245 | l4offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); | | 247 | l4offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); |
246 | | | 248 | |
247 | if (__predict_true(l4offset + sizeof(uint16_t) <= m->m_len)) { | | 249 | if (__predict_true(l4offset + sizeof(uint16_t) <= m->m_len)) { |
248 | *(uint16_t *)(mtod(m, char *) + l4offset) = csum; | | 250 | *(uint16_t *)(mtod(m, char *) + l4offset) = csum; |
249 | } else { | | 251 | } else { |
250 | m_copyback(m, l4offset, sizeof(csum), (void *) &csum); | | 252 | m_copyback(m, l4offset, sizeof(csum), (void *) &csum); |
251 | } | | 253 | } |
252 | } | | 254 | } |
253 | | | 255 | |
254 | m->m_pkthdr.csum_flags ^= csum_flags; | | 256 | m->m_pkthdr.csum_flags ^= csum_flags; |
255 | } | | 257 | } |