1 /*	$OpenBSD: memmem.c,v 1.5 2020/04/16 12:39:28 claudio Exp $ */
2 
3 /*
4  * Copyright (c) 2005-2020 Rich Felker, et al.
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining
7  * a copy of this software and associated documentation files (the
8  * "Software"), to deal in the Software without restriction, including
9  * without limitation the rights to use, copy, modify, merge, publish,
10  * distribute, sublicense, and/or sell copies of the Software, and to
11  * permit persons to whom the Software is furnished to do so, subject to
12  * the following conditions:
13  *
14  * The above copyright notice and this permission notice shall be
15  * included in all copies or substantial portions of the Software.
16  *
17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
18  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
19  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
20  * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
21  * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
22  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
23  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
24  */
25 
26 /* OPENBSD ORIGINAL: lib/libc/string/memmem.c */
27 
28 #include "includes.h"
29 
30 #ifndef HAVE_MEMMEM
31 
32 #include <string.h>
33 #ifdef HAVE_STDINT_H
34 #include <stdint.h>
35 #endif
36 
37 static char *
twobyte_memmem(const unsigned char * h,size_t k,const unsigned char * n)38 twobyte_memmem(const unsigned char *h, size_t k, const unsigned char *n)
39 {
40 	uint16_t nw = n[0]<<8 | n[1], hw = h[0]<<8 | h[1];
41 	for (h+=2, k-=2; k; k--, hw = hw<<8 | *h++)
42 		if (hw == nw) return (char *)h-2;
43 	return hw == nw ? (char *)h-2 : 0;
44 }
45 
46 static char *
threebyte_memmem(const unsigned char * h,size_t k,const unsigned char * n)47 threebyte_memmem(const unsigned char *h, size_t k, const unsigned char *n)
48 {
49 	uint32_t nw = n[0]<<24 | n[1]<<16 | n[2]<<8;
50 	uint32_t hw = h[0]<<24 | h[1]<<16 | h[2]<<8;
51 	for (h+=3, k-=3; k; k--, hw = (hw|*h++)<<8)
52 		if (hw == nw) return (char *)h-3;
53 	return hw == nw ? (char *)h-3 : 0;
54 }
55 
56 static char *
fourbyte_memmem(const unsigned char * h,size_t k,const unsigned char * n)57 fourbyte_memmem(const unsigned char *h, size_t k, const unsigned char *n)
58 {
59 	uint32_t nw = n[0]<<24 | n[1]<<16 | n[2]<<8 | n[3];
60 	uint32_t hw = h[0]<<24 | h[1]<<16 | h[2]<<8 | h[3];
61 	for (h+=4, k-=4; k; k--, hw = hw<<8 | *h++)
62 		if (hw == nw) return (char *)h-4;
63 	return hw == nw ? (char *)h-4 : 0;
64 }
65 
66 #if 0
67 /* In -portable, defines.h ensures that these are already defined. */
68 #define MAX(a,b) ((a)>(b)?(a):(b))
69 #define MIN(a,b) ((a)<(b)?(a):(b))
70 #endif
71 
72 #define BITOP(a,b,op) \
73  ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
74 
75 /*
76  * Maxime Crochemore and Dominique Perrin, Two-way string-matching,
77  * Journal of the ACM, 38(3):651-675, July 1991.
78  */
79 static char *
twoway_memmem(const unsigned char * h,const unsigned char * z,const unsigned char * n,size_t l)80 twoway_memmem(const unsigned char *h, const unsigned char *z,
81     const unsigned char *n, size_t l)
82 {
83 	size_t i, ip, jp, k, p, ms, p0, mem, mem0;
84 	size_t byteset[32 / sizeof(size_t)] = { 0 };
85 	size_t shift[256];
86 
87 	/* Computing length of needle and fill shift table */
88 	for (i=0; i<l; i++)
89 		BITOP(byteset, n[i], |=), shift[n[i]] = i+1;
90 
91 	/* Compute maximal suffix */
92 	ip = -1; jp = 0; k = p = 1;
93 	while (jp+k<l) {
94 		if (n[ip+k] == n[jp+k]) {
95 			if (k == p) {
96 				jp += p;
97 				k = 1;
98 			} else k++;
99 		} else if (n[ip+k] > n[jp+k]) {
100 			jp += k;
101 			k = 1;
102 			p = jp - ip;
103 		} else {
104 			ip = jp++;
105 			k = p = 1;
106 		}
107 	}
108 	ms = ip;
109 	p0 = p;
110 
111 	/* And with the opposite comparison */
112 	ip = -1; jp = 0; k = p = 1;
113 	while (jp+k<l) {
114 		if (n[ip+k] == n[jp+k]) {
115 			if (k == p) {
116 				jp += p;
117 				k = 1;
118 			} else k++;
119 		} else if (n[ip+k] < n[jp+k]) {
120 			jp += k;
121 			k = 1;
122 			p = jp - ip;
123 		} else {
124 			ip = jp++;
125 			k = p = 1;
126 		}
127 	}
128 	if (ip+1 > ms+1) ms = ip;
129 	else p = p0;
130 
131 	/* Periodic needle? */
132 	if (memcmp(n, n+p, ms+1)) {
133 		mem0 = 0;
134 		p = MAX(ms, l-ms-1) + 1;
135 	} else mem0 = l-p;
136 	mem = 0;
137 
138 	/* Search loop */
139 	for (;;) {
140 		/* If remainder of haystack is shorter than needle, done */
141 		if (z-h < l) return 0;
142 
143 		/* Check last byte first; advance by shift on mismatch */
144 		if (BITOP(byteset, h[l-1], &)) {
145 			k = l-shift[h[l-1]];
146 			if (k) {
147 				if (k < mem) k = mem;
148 				h += k;
149 				mem = 0;
150 				continue;
151 			}
152 		} else {
153 			h += l;
154 			mem = 0;
155 			continue;
156 		}
157 
158 		/* Compare right half */
159 		for (k=MAX(ms+1,mem); k<l && n[k] == h[k]; k++);
160 		if (k < l) {
161 			h += k-ms;
162 			mem = 0;
163 			continue;
164 		}
165 		/* Compare left half */
166 		for (k=ms+1; k>mem && n[k-1] == h[k-1]; k--);
167 		if (k <= mem) return (char *)h;
168 		h += p;
169 		mem = mem0;
170 	}
171 }
172 
173 void *
memmem(const void * h0,size_t k,const void * n0,size_t l)174 memmem(const void *h0, size_t k, const void *n0, size_t l)
175 {
176 	const unsigned char *h = h0, *n = n0;
177 
178 	/* Return immediately on empty needle */
179 	if (!l) return (void *)h;
180 
181 	/* Return immediately when needle is longer than haystack */
182 	if (k<l) return 0;
183 
184 	/* Use faster algorithms for short needles */
185 	h = memchr(h0, *n, k);
186 	if (!h || l==1) return (void *)h;
187 	k -= h - (const unsigned char *)h0;
188 	if (k<l) return 0;
189 	if (l==2) return twobyte_memmem(h, k, n);
190 	if (l==3) return threebyte_memmem(h, k, n);
191 	if (l==4) return fourbyte_memmem(h, k, n);
192 
193 	return twoway_memmem(h, h+k, n, l);
194 }
195 DEF_WEAK(memmem);
196 #endif /* HAVE_MEMMEM */
197