xref: /linux/lib/ts_bm.c (revision de2fe5e07d58424bc286fff3fd3c1b0bf933cd58)
1 /*
2  * lib/ts_bm.c		Boyer-Moore text search implementation
3  *
4  *		This program is free software; you can redistribute it and/or
5  *		modify it under the terms of the GNU General Public License
6  *		as published by the Free Software Foundation; either version
7  *		2 of the License, or (at your option) any later version.
8  *
9  * Authors:	Pablo Neira Ayuso <pablo@eurodev.net>
10  *
11  * ==========================================================================
12  *
13  *   Implements Boyer-Moore string matching algorithm:
14  *
15  *   [1] A Fast String Searching Algorithm, R.S. Boyer and Moore.
16  *       Communications of the Association for Computing Machinery,
17  *       20(10), 1977, pp. 762-772.
18  *       http://www.cs.utexas.edu/users/moore/publications/fstrpos.pdf
19  *
20  *   [2] Handbook of Exact String Matching Algorithms, Thierry Lecroq, 2004
21  *       http://www-igm.univ-mlv.fr/~lecroq/string/string.pdf
22  *
23  *   Note: Since Boyer-Moore (BM) performs searches for matchings from right
24  *   to left, it's still possible that a matching could be spread over
25  *   multiple blocks, in that case this algorithm won't find any coincidence.
26  *
27  *   If you're willing to ensure that such thing won't ever happen, use the
28  *   Knuth-Pratt-Morris (KMP) implementation instead. In conclusion, choose
29  *   the proper string search algorithm depending on your setting.
30  *
31  *   Say you're using the textsearch infrastructure for filtering, NIDS or
32  *   any similar security focused purpose, then go KMP. Otherwise, if you
33  *   really care about performance, say you're classifying packets to apply
34  *   Quality of Service (QoS) policies, and you don't mind about possible
35  *   matchings spread over multiple fragments, then go BM.
36  */
37 
38 #include <linux/config.h>
39 #include <linux/kernel.h>
40 #include <linux/module.h>
41 #include <linux/types.h>
42 #include <linux/string.h>
43 #include <linux/textsearch.h>
44 
45 /* Alphabet size, use ASCII */
46 #define ASIZE 256
47 
48 #if 0
49 #define DEBUGP printk
50 #else
51 #define DEBUGP(args, format...)
52 #endif
53 
54 struct ts_bm
55 {
56 	u8 *		pattern;
57 	unsigned int	patlen;
58 	unsigned int 	bad_shift[ASIZE];
59 	unsigned int	good_shift[0];
60 };
61 
62 static unsigned int bm_find(struct ts_config *conf, struct ts_state *state)
63 {
64 	struct ts_bm *bm = ts_config_priv(conf);
65 	unsigned int i, text_len, consumed = state->offset;
66 	const u8 *text;
67 	int shift = bm->patlen, bs;
68 
69 	for (;;) {
70 		text_len = conf->get_next_block(consumed, &text, conf, state);
71 
72 		if (unlikely(text_len == 0))
73 			break;
74 
75 		while (shift < text_len) {
76 			DEBUGP("Searching in position %d (%c)\n",
77 				shift, text[shift]);
78 			for (i = 0; i < bm->patlen; i++)
79 			     if (text[shift-i] != bm->pattern[bm->patlen-1-i])
80 				     goto next;
81 
82 			/* London calling... */
83 			DEBUGP("found!\n");
84 			return consumed += (shift-(bm->patlen-1));
85 
86 next:			bs = bm->bad_shift[text[shift-i]];
87 
88 			/* Now jumping to... */
89 			shift = max_t(int, shift-i+bs, shift+bm->good_shift[i]);
90 		}
91 		consumed += text_len;
92 	}
93 
94 	return UINT_MAX;
95 }
96 
97 static int subpattern(u8 *pattern, int i, int j, int g)
98 {
99 	int x = i+g-1, y = j+g-1, ret = 0;
100 
101 	while(pattern[x--] == pattern[y--]) {
102 		if (y < 0) {
103 			ret = 1;
104 			break;
105 		}
106 		if (--g == 0) {
107 			ret = pattern[i-1] != pattern[j-1];
108 			break;
109 		}
110 	}
111 
112 	return ret;
113 }
114 
115 static void compute_prefix_tbl(struct ts_bm *bm, const u8 *pattern,
116 			       unsigned int len)
117 {
118 	int i, j, g;
119 
120 	for (i = 0; i < ASIZE; i++)
121 		bm->bad_shift[i] = len;
122 	for (i = 0; i < len - 1; i++)
123 		bm->bad_shift[pattern[i]] = len - 1 - i;
124 
125 	/* Compute the good shift array, used to match reocurrences
126 	 * of a subpattern */
127 	bm->good_shift[0] = 1;
128 	for (i = 1; i < bm->patlen; i++)
129 		bm->good_shift[i] = bm->patlen;
130         for (i = bm->patlen-1, g = 1; i > 0; g++, i--) {
131 		for (j = i-1; j >= 1-g ; j--)
132 			if (subpattern(bm->pattern, i, j, g)) {
133 				bm->good_shift[g] = bm->patlen-j-g;
134 				break;
135 			}
136 	}
137 }
138 
139 static struct ts_config *bm_init(const void *pattern, unsigned int len,
140 				 gfp_t gfp_mask)
141 {
142 	struct ts_config *conf;
143 	struct ts_bm *bm;
144 	unsigned int prefix_tbl_len = len * sizeof(unsigned int);
145 	size_t priv_size = sizeof(*bm) + len + prefix_tbl_len;
146 
147 	conf = alloc_ts_config(priv_size, gfp_mask);
148 	if (IS_ERR(conf))
149 		return conf;
150 
151 	bm = ts_config_priv(conf);
152 	bm->patlen = len;
153 	bm->pattern = (u8 *) bm->good_shift + prefix_tbl_len;
154 	compute_prefix_tbl(bm, pattern, len);
155 	memcpy(bm->pattern, pattern, len);
156 
157 	return conf;
158 }
159 
160 static void *bm_get_pattern(struct ts_config *conf)
161 {
162 	struct ts_bm *bm = ts_config_priv(conf);
163 	return bm->pattern;
164 }
165 
166 static unsigned int bm_get_pattern_len(struct ts_config *conf)
167 {
168 	struct ts_bm *bm = ts_config_priv(conf);
169 	return bm->patlen;
170 }
171 
172 static struct ts_ops bm_ops = {
173 	.name		  = "bm",
174 	.find		  = bm_find,
175 	.init		  = bm_init,
176 	.get_pattern	  = bm_get_pattern,
177 	.get_pattern_len  = bm_get_pattern_len,
178 	.owner		  = THIS_MODULE,
179 	.list		  = LIST_HEAD_INIT(bm_ops.list)
180 };
181 
182 static int __init init_bm(void)
183 {
184 	return textsearch_register(&bm_ops);
185 }
186 
187 static void __exit exit_bm(void)
188 {
189 	textsearch_unregister(&bm_ops);
190 }
191 
192 MODULE_LICENSE("GPL");
193 
194 module_init(init_bm);
195 module_exit(exit_bm);
196