blob: 2cc79112ecc39fc2c945047b67e1c1ac4e67723f [file] [log] [blame]
Pablo Neira Ayuso8082e4e2005-08-25 16:12:22 -07001/*
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
54struct ts_bm
55{
56 u8 * pattern;
57 unsigned int patlen;
58 unsigned int bad_shift[ASIZE];
59 unsigned int good_shift[0];
60};
61
62static 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
86next: 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
97static void compute_prefix_tbl(struct ts_bm *bm, const u8 *pattern,
98 unsigned int len)
99{
100 int i, j, ended, l[ASIZE];
101
102 for (i = 0; i < ASIZE; i++)
103 bm->bad_shift[i] = len;
104 for (i = 0; i < len - 1; i++)
105 bm->bad_shift[pattern[i]] = len - 1 - i;
106
107 /* Compute the good shift array, used to match reocurrences
108 * of a subpattern */
109 for (i = 1; i < bm->patlen; i++) {
110 for (j = 0; j < bm->patlen && bm->pattern[bm->patlen - 1 - j]
111 == bm->pattern[bm->patlen - 1 - i - j]; j++);
112 l[i] = j;
113 }
114
115 bm->good_shift[0] = 1;
116 for (i = 1; i < bm->patlen; i++)
117 bm->good_shift[i] = bm->patlen;
118 for (i = bm->patlen - 1; i > 0; i--)
119 bm->good_shift[l[i]] = i;
120 ended = 0;
121 for (i = 0; i < bm->patlen; i++) {
122 if (l[i] == bm->patlen - 1 - i)
123 ended = i;
124 if (ended)
125 bm->good_shift[i] = ended;
126 }
127}
128
129static struct ts_config *bm_init(const void *pattern, unsigned int len,
130 int gfp_mask)
131{
132 struct ts_config *conf;
133 struct ts_bm *bm;
134 unsigned int prefix_tbl_len = len * sizeof(unsigned int);
135 size_t priv_size = sizeof(*bm) + len + prefix_tbl_len;
136
137 conf = alloc_ts_config(priv_size, gfp_mask);
138 if (IS_ERR(conf))
139 return conf;
140
141 bm = ts_config_priv(conf);
142 bm->patlen = len;
143 bm->pattern = (u8 *) bm->good_shift + prefix_tbl_len;
144 compute_prefix_tbl(bm, pattern, len);
145 memcpy(bm->pattern, pattern, len);
146
147 return conf;
148}
149
150static void *bm_get_pattern(struct ts_config *conf)
151{
152 struct ts_bm *bm = ts_config_priv(conf);
153 return bm->pattern;
154}
155
156static unsigned int bm_get_pattern_len(struct ts_config *conf)
157{
158 struct ts_bm *bm = ts_config_priv(conf);
159 return bm->patlen;
160}
161
162static struct ts_ops bm_ops = {
163 .name = "bm",
164 .find = bm_find,
165 .init = bm_init,
166 .get_pattern = bm_get_pattern,
167 .get_pattern_len = bm_get_pattern_len,
168 .owner = THIS_MODULE,
169 .list = LIST_HEAD_INIT(bm_ops.list)
170};
171
172static int __init init_bm(void)
173{
174 return textsearch_register(&bm_ops);
175}
176
177static void __exit exit_bm(void)
178{
179 textsearch_unregister(&bm_ops);
180}
181
182MODULE_LICENSE("GPL");
183
184module_init(init_bm);
185module_exit(exit_bm);