xref: /netbsd/sys/arch/powerpc/powerpc/in_cksum.c (revision c4a72b64)
1 /*	$NetBSD: in_cksum.c,v 1.6 2002/08/05 02:56:58 enami Exp $	*/
2 
3 /*
4  * Copyright 2001 Wasabi Systems, Inc.
5  * All rights reserved.
6  *
7  * Written by Simon Burge and Eduardo Horvath for Wasabi Systems, Inc.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *      This product includes software developed for the NetBSD Project by
20  *      Wasabi Systems, Inc.
21  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
22  *    or promote products derived from this software without specific prior
23  *    written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
27  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
29  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <sys/param.h>
39 #include <sys/mbuf.h>
40 #include <sys/systm.h>
41 #include <netinet/in_systm.h>
42 #include <netinet/in.h>
43 #include <netinet/ip.h>
44 #include <netinet/ip_var.h>
45 
46 /*
47  * Checksum routine for Internet Protocol family headers.
48  *
49  * This routine is very heavily used in the network
50  * code and should be modified for each CPU to be as fast as possible.
51  *
52  * PowerPC version.
53  */
54 
55 #define	REDUCE1		sum = (sum & 0xffff) + (sum >> 16)
56 /* Two REDUCE1s is faster than REDUCE1; if (sum > 65535) sum -= 65536; */
57 #define	REDUCE		{ REDUCE1; REDUCE1; }
58 
59 static __inline__ int
60 in_cksum_internal(struct mbuf *m, int off, int len, u_int sum)
61 {
62 	uint8_t *w;
63 	int mlen = 0;
64 	int byte_swapped = 0;
65 	int n;
66 
67 	union {
68 		uint8_t  c[2];
69 		uint16_t s;
70 	} s_util;
71 
72 	for (;m && len; m = m->m_next) {
73 		if (m->m_len == 0)
74 			continue;
75 		w = mtod(m, uint8_t *) + off;
76 
77 		/*
78 		 * 'off' can only be non-zero on the first pass of this
79 		 * loop when mlen != -1, so we don't need to worry about
80 		 * 'off' in the if clause below.
81 		 */
82 		if (mlen == -1) {
83 			/*
84 			 * The first byte of this mbuf is the continuation
85 			 * of a word spanning between this mbuf and the
86 			 * last mbuf.
87 			 *
88 			 * s_util.c[0] is already saved when scanning previous
89 			 * mbuf.
90 			 */
91 			s_util.c[1] = *w++;
92 			sum += s_util.s;
93 			mlen = m->m_len - 1;
94 			len--;
95 		} else {
96 			mlen = m->m_len - off;
97 			off = 0;
98 		}
99 		if (len < mlen)
100 			mlen = len;
101 		len -= mlen;
102 
103 		/*
104 		 * Force to a word boundary.
105 		 */
106 		if ((3 & (long) w) && (mlen > 0)) {
107 			if ((1 & (long) w)) {
108 				REDUCE;
109 				sum <<= 8;
110 				s_util.c[0] = *w++;
111 				mlen--;
112 				byte_swapped = 1;
113 			}
114 			if ((2 & (long) w) && (mlen > 1)) {
115 				/*
116 				 * Since the `sum' may contain full 32 bit
117 				 * value, we can't simply add any value.
118 				 */
119 				__asm __volatile(
120 				    "lhz 7,0(%1);"	/* load current data
121 							   half word */
122 				    "addc %0,%0,7;"	/* add to sum */
123 				    "addze %0,%0;"	/* add carry bit */
124 				    : "+r"(sum)
125 				    : "b"(w)
126 				    : "7");		/* clobber r7 */
127 				w += 2;
128 				mlen -= 2;
129 			}
130 		}
131 
132 		if (mlen >= 64) {
133 			n = mlen >> 6;
134 			__asm __volatile(
135 			    "addic 0,0,0;"		/* clear carry */
136 			    "mtctr %1;"			/* load loop count */
137 			    "1:"
138 			    "lwz 7,4(%2);"		/* load current data
139 							   word */
140 			    "lwz 8,8(%2);"
141 			    "lwz 9,12(%2);"
142 			    "lwz 10,16(%2);"
143 			    "adde %0,%0,7;"		/* add to sum */
144 			    "adde %0,%0,8;"
145 			    "adde %0,%0,9;"
146 			    "adde %0,%0,10;"
147 			    "lwz 7,20(%2);"
148 			    "lwz 8,24(%2);"
149 			    "lwz 9,28(%2);"
150 			    "lwz 10,32(%2);"
151 			    "adde %0,%0,7;"
152 			    "adde %0,%0,8;"
153 			    "adde %0,%0,9;"
154 			    "adde %0,%0,10;"
155 			    "lwz 7,36(%2);"
156 			    "lwz 8,40(%2);"
157 			    "lwz 9,44(%2);"
158 			    "lwz 10,48(%2);"
159 			    "adde %0,%0,7;"
160 			    "adde %0,%0,8;"
161 			    "adde %0,%0,9;"
162 			    "adde %0,%0,10;"
163 			    "lwz 7,52(%2);"
164 			    "lwz 8,56(%2);"
165 			    "lwz 9,60(%2);"
166 			    "lwzu 10,64(%2);"
167 			    "adde %0,%0,7;"
168 			    "adde %0,%0,8;"
169 			    "adde %0,%0,9;"
170 			    "adde %0,%0,10;"
171 			    "bdnz 1b;"			/* loop */
172 			    "addze %0,%0;"		/* add carry bit */
173 			    : "+r"(sum)
174 			    : "r"(n), "b"(w - 4)
175 			    : "7", "8", "9", "10");	/* clobber r7, r8, r9,
176 							   r10 */
177 			w += n * 64;
178 			mlen -= n * 64;
179 		}
180 
181 		if (mlen >= 8) {
182 			n = mlen >> 3;
183 			__asm __volatile(
184 			    "addic 0,0,0;"		/* clear carry */
185 			    "mtctr %1;"			/* load loop count */
186 			    "1:"
187 			    "lwz 7,4(%2);"		/* load current data
188 							   word */
189 			    "lwzu 8,8(%2);"
190 			    "adde %0,%0,7;"		/* add to sum */
191 			    "adde %0,%0,8;"
192 			    "bdnz 1b;"			/* loop */
193 			    "addze %0,%0;"		/* add carry bit */
194 			    : "+r"(sum)
195 			    : "r"(n), "b"(w - 4)
196 			    : "7", "8");		/* clobber r7, r8 */
197 			w += n * 8;
198 			mlen -= n * 8;
199 		}
200 
201 		if (mlen == 0 && byte_swapped == 0)
202 			continue;
203 		REDUCE;
204 
205 		while ((mlen -= 2) >= 0) {
206 			sum += *(uint16_t *)w;
207 			w += 2;
208 		}
209 
210 		if (byte_swapped) {
211 			REDUCE;
212 			sum <<= 8;
213 			byte_swapped = 0;
214 			if (mlen == -1) {
215 				s_util.c[1] = *w;
216 				sum += s_util.s;
217 				mlen = 0;
218 			} else
219 				mlen = -1;
220 		} else if (mlen == -1)
221 			s_util.c[0] = *w;
222 	}
223 	if (len)
224 		printf("cksum: out of data\n");
225 	if (mlen == -1) {
226 		/* The last mbuf has odd # of bytes. Follow the
227 		   standard (the odd byte may be shifted left by 8 bits
228 		   or not as determined by endian-ness of the machine) */
229 		s_util.c[1] = 0;
230 		sum += s_util.s;
231 	}
232 	REDUCE;
233 	return (~sum & 0xffff);
234 }
235 
236 int
237 in_cksum(struct mbuf *m, int len)
238 {
239 
240 	return (in_cksum_internal(m, 0, len, 0));
241 }
242 
243 int
244 in4_cksum(struct mbuf *m, uint8_t nxt, int off, int len)
245 {
246 	uint16_t *w;
247 	u_int sum = 0;
248 	union {
249 		struct ipovly ipov;
250 		u_int16_t w[10];
251 	} u;
252 
253 	if (nxt != 0) {
254 		/* pseudo header */
255 		memset(&u.ipov, 0, sizeof(u.ipov));
256 		u.ipov.ih_len = htons(len);
257 		u.ipov.ih_pr = nxt;
258 		u.ipov.ih_src = mtod(m, struct ip *)->ip_src;
259 		u.ipov.ih_dst = mtod(m, struct ip *)->ip_dst;
260 		w = u.w;
261 		/* assumes sizeof(ipov) == 20 */
262 		sum += w[0]; sum += w[1]; sum += w[2]; sum += w[3]; sum += w[4];
263 		sum += w[5]; sum += w[6]; sum += w[7]; sum += w[8]; sum += w[9];
264 	}
265 
266 	/* skip unnecessary part */
267 	while (m && off > 0) {
268 		if (m->m_len > off)
269 			break;
270 		off -= m->m_len;
271 		m = m->m_next;
272 	}
273 
274 	return (in_cksum_internal(m, off, len, sum));
275 }
276