1 /*-
2 * Copyright (c) 1990 The Regents of the University of California.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 */
29 /*-
30 * Copyright (c) 2001 by Thomas Moestl <tmm@FreeBSD.org>.
31 * All rights reserved.
32 *
33 * Redistribution and use in source and binary forms, with or without
34 * modification, are permitted provided that the following conditions
35 * are met:
36 * 1. Redistributions of source code must retain the above copyright
37 * notice, this list of conditions and the following disclaimer.
38 * 2. Redistributions in binary form must reproduce the above copyright
39 * notice, this list of conditions and the following disclaimer in the
40 * documentation and/or other materials provided with the distribution.
41 *
42 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
43 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
44 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
45 * IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
46 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
47 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
48 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
49 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
50 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
51 * USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
52 *
53 * from tahoe: in_cksum.c 1.2 86/01/05
54 * from: @(#)in_cksum.c 1.3 (Berkeley) 1/19/91
55 * from: FreeBSD: src/sys/i386/i386/in_cksum.c,v 1.22 2000/11/25
56 *
57 * $FreeBSD: releng/6.0/sys/sparc64/sparc64/in_cksum.c 127977 2004-04-07 05:00:01Z imp $
58 */
59
60 #include <sys/param.h>
61 #include <sys/systm.h>
62 #include <sys/mbuf.h>
63
64 #include <netinet/in.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/ip.h>
67
68 #include <machine/in_cksum.h>
69
70 /*
71 * Checksum routine for Internet Protocol family headers.
72 *
73 * This routine is very heavily used in the network
74 * code and should be modified for each CPU to be as fast as possible.
75 *
76 * This implementation is a sparc64 version. Most code was taken over and
77 * adapted from the i386. Some optimizations were changed to achieve (hopefully)
78 * better performance.
79 * This uses 64 bit loads, but 32 bit additions due to the lack of a 64-bit
80 * add-with-carry operation.
81 */
82
83 /*
84 * REDUCE() is actually not used that frequently... maybe a C implementation
85 * would suffice.
86 */
87 #define REDUCE(sum, tmp) __asm __volatile( \
88 "sll %2, 16, %1\n" \
89 "addcc %2, %1, %0\n" \
90 "srl %0, 16, %0\n" \
91 "addc %0, 0, %0" : "=r" (sum), "=r" (tmp) : "" (sum))
92
93 /*
94 * Note that some of these macros depend on the flags being preserved between
95 * calls, so they should not be intermixed with other C statements.
96 */
97 #define LD64_ADD32(sum, tmp, addr, n, mod) __asm __volatile( \
98 "ldx [%3 + " #n "], %1\n" \
99 "add" #mod " %2, %1, %0\n" \
100 "srlx %1, 32, %1\n" \
101 "addccc %0, %1, %0" : "=r" (sum), "=r" (tmp) : "" (sum), "r" (addr))
102
103 #define LD32_ADD32(sum, tmp, addr, n, mod) __asm __volatile( \
104 "lduw [%3 + " #n "], %1\n" \
105 "add" #mod " %2, %1, %0\n" \
106 : "=r" (sum), "=r" (tmp) : "" (sum), "r" (addr))
107
108 #define MOP(sum) __asm __volatile( \
109 "addc %1, 0, %0" : "=r" (sum) : "" (sum))
110
111 u_short
112 in_cksum_skip(struct mbuf *m, int len, int skip)
113 {
114 u_short *w;
115 unsigned long tmp, sum = 0;
116 int mlen = 0;
117 int byte_swapped = 0;
118 u_short su = 0;
119
120 len -= skip;
121 for (; skip > 0 && m != NULL; m = m->m_next) {
122 if (m->m_len > skip) {
123 mlen = m->m_len - skip;
124 w = (u_short *)(mtod(m, u_char *) + skip);
125 goto skip_start;
126 } else
127 skip -= m->m_len;
128 }
129
130 for (; m != NULL && len > 0; m = m->m_next) {
131 if (m->m_len == 0)
132 continue;
133 w = mtod(m, u_short *);
134 if (mlen == -1) {
135 /*
136 * The first byte of this mbuf is the continuation
137 * of a word spanning between this mbuf and the
138 * last mbuf.
139 *
140 * The high order byte of su is already saved when
141 * scanning previous mbuf. sum was REDUCEd when we
142 * found mlen == -1
143 */
144 sum += su | *(u_char *)w;
145 w = (u_short *)((u_char *)w + 1);
146 mlen = m->m_len - 1;
147 len--;
148 } else
149 mlen = m->m_len;
150 skip_start:
151 if (len < mlen)
152 mlen = len;
153 len -= mlen;
154 /*
155 * Force to a 8-byte boundary first so that we can use
156 * LD64_ADD32.
157 */
158 if (((u_long)w & 7) != 0) {
159 REDUCE(sum, tmp);
160 if (((u_long)w & 1) != 0 && mlen >= 1) {
161 sum <<= 8;
162 su = *(u_char *)w << 8;
163 w = (u_short *)((u_char *)w + 1);
164 mlen--;
165 byte_swapped = 1;
166 }
167 if (((u_long)w & 2) != 0 && mlen >= 2) {
168 sum += *w++;
169 mlen -= 2;
170 }
171 if (((u_long)w & 4) != 0 && mlen >= 4) {
172 LD32_ADD32(sum, tmp, w, 0, cc);
173 MOP(sum);
174 w += 2;
175 mlen -= 4;
176 }
177 }
178 /*
179 * Do as much of the checksum as possible 64 bits at at time.
180 * In fact, this loop is unrolled to make overhead from
181 * branches &c small.
182 */
183 for (; mlen >= 64; mlen -= 64) {
184 LD64_ADD32(sum, tmp, w, 0, cc);
185 LD64_ADD32(sum, tmp, w, 8, ccc);
186 LD64_ADD32(sum, tmp, w, 16, ccc);
187 LD64_ADD32(sum, tmp, w, 24, ccc);
188 LD64_ADD32(sum, tmp, w, 32, ccc);
189 LD64_ADD32(sum, tmp, w, 40, ccc);
190 LD64_ADD32(sum, tmp, w, 48, ccc);
191 LD64_ADD32(sum, tmp, w, 56, ccc);
192 MOP(sum);
193 w += 32;
194 }
195 if (mlen >= 32) {
196 LD64_ADD32(sum, tmp, w, 0, cc);
197 LD64_ADD32(sum, tmp, w, 8, ccc);
198 LD64_ADD32(sum, tmp, w, 16, ccc);
199 LD64_ADD32(sum, tmp, w, 24, ccc);
200 MOP(sum);
201 w += 16;
202 mlen -= 32;
203 }
204 if (mlen >= 16) {
205 LD64_ADD32(sum, tmp, w, 0, cc);
206 LD64_ADD32(sum, tmp, w, 8, ccc);
207 MOP(sum);
208 w += 8;
209 mlen -= 16;
210 }
211 if (mlen >= 8) {
212 LD64_ADD32(sum, tmp, w, 0, cc);
213 MOP(sum);
214 w += 4;
215 mlen -= 8;
216 }
217 REDUCE(sum, tmp);
218 while ((mlen -= 2) >= 0)
219 sum += *w++;
220 if (byte_swapped) {
221 sum <<= 8;
222 byte_swapped = 0;
223 if (mlen == -1) {
224 su |= *(u_char *)w;
225 sum += su;
226 mlen = 0;
227 } else
228 mlen = -1;
229 } else if (mlen == -1) {
230 /*
231 * This mbuf has odd number of bytes.
232 * There could be a word split betwen
233 * this mbuf and the next mbuf.
234 * Save the last byte (to prepend to next mbuf).
235 */
236 su = *(u_char *)w << 8;
237 }
238 }
239
240 if (len)
241 printf("%s: out of data by %d\n", __func__, len);
242 if (mlen == -1) {
243 /* The last mbuf has odd # of bytes. Follow the
244 standard (the odd byte is shifted left by 8 bits) */
245 sum += su & 0xff00;
246 }
247 REDUCE(sum, tmp);
248 return (~sum & 0xffff);
249 }
Cache object: 47fccae24321abb6fed6c4b927741059
|