Linux Kernel  3.7.1
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
ts_bm.c
Go to the documentation of this file.
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 <[email protected]>
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/kernel.h>
39 #include <linux/module.h>
40 #include <linux/types.h>
41 #include <linux/string.h>
42 #include <linux/ctype.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 {
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 - 1, bs;
68  const u8 icase = conf->flags & TS_IGNORECASE;
69 
70  for (;;) {
71  text_len = conf->get_next_block(consumed, &text, conf, state);
72 
73  if (unlikely(text_len == 0))
74  break;
75 
76  while (shift < text_len) {
77  DEBUGP("Searching in position %d (%c)\n",
78  shift, text[shift]);
79  for (i = 0; i < bm->patlen; i++)
80  if ((icase ? toupper(text[shift-i])
81  : text[shift-i])
82  != bm->pattern[bm->patlen-1-i])
83  goto next;
84 
85  /* London calling... */
86  DEBUGP("found!\n");
87  return consumed += (shift-(bm->patlen-1));
88 
89 next: bs = bm->bad_shift[text[shift-i]];
90 
91  /* Now jumping to... */
92  shift = max_t(int, shift-i+bs, shift+bm->good_shift[i]);
93  }
94  consumed += text_len;
95  }
96 
97  return UINT_MAX;
98 }
99 
100 static int subpattern(u8 *pattern, int i, int j, int g)
101 {
102  int x = i+g-1, y = j+g-1, ret = 0;
103 
104  while(pattern[x--] == pattern[y--]) {
105  if (y < 0) {
106  ret = 1;
107  break;
108  }
109  if (--g == 0) {
110  ret = pattern[i-1] != pattern[j-1];
111  break;
112  }
113  }
114 
115  return ret;
116 }
117 
118 static void compute_prefix_tbl(struct ts_bm *bm, int flags)
119 {
120  int i, j, g;
121 
122  for (i = 0; i < ASIZE; i++)
123  bm->bad_shift[i] = bm->patlen;
124  for (i = 0; i < bm->patlen - 1; i++) {
125  bm->bad_shift[bm->pattern[i]] = bm->patlen - 1 - i;
126  if (flags & TS_IGNORECASE)
127  bm->bad_shift[tolower(bm->pattern[i])]
128  = bm->patlen - 1 - i;
129  }
130 
131  /* Compute the good shift array, used to match reocurrences
132  * of a subpattern */
133  bm->good_shift[0] = 1;
134  for (i = 1; i < bm->patlen; i++)
135  bm->good_shift[i] = bm->patlen;
136  for (i = bm->patlen-1, g = 1; i > 0; g++, i--) {
137  for (j = i-1; j >= 1-g ; j--)
138  if (subpattern(bm->pattern, i, j, g)) {
139  bm->good_shift[g] = bm->patlen-j-g;
140  break;
141  }
142  }
143 }
144 
145 static struct ts_config *bm_init(const void *pattern, unsigned int len,
146  gfp_t gfp_mask, int flags)
147 {
148  struct ts_config *conf;
149  struct ts_bm *bm;
150  int i;
151  unsigned int prefix_tbl_len = len * sizeof(unsigned int);
152  size_t priv_size = sizeof(*bm) + len + prefix_tbl_len;
153 
154  conf = alloc_ts_config(priv_size, gfp_mask);
155  if (IS_ERR(conf))
156  return conf;
157 
158  conf->flags = flags;
159  bm = ts_config_priv(conf);
160  bm->patlen = len;
161  bm->pattern = (u8 *) bm->good_shift + prefix_tbl_len;
162  if (flags & TS_IGNORECASE)
163  for (i = 0; i < len; i++)
164  bm->pattern[i] = toupper(((u8 *)pattern)[i]);
165  else
166  memcpy(bm->pattern, pattern, len);
167  compute_prefix_tbl(bm, flags);
168 
169  return conf;
170 }
171 
172 static void *bm_get_pattern(struct ts_config *conf)
173 {
174  struct ts_bm *bm = ts_config_priv(conf);
175  return bm->pattern;
176 }
177 
178 static unsigned int bm_get_pattern_len(struct ts_config *conf)
179 {
180  struct ts_bm *bm = ts_config_priv(conf);
181  return bm->patlen;
182 }
183 
184 static struct ts_ops bm_ops = {
185  .name = "bm",
186  .find = bm_find,
187  .init = bm_init,
188  .get_pattern = bm_get_pattern,
189  .get_pattern_len = bm_get_pattern_len,
190  .owner = THIS_MODULE,
191  .list = LIST_HEAD_INIT(bm_ops.list)
192 };
193 
194 static int __init init_bm(void)
195 {
196  return textsearch_register(&bm_ops);
197 }
198 
199 static void __exit exit_bm(void)
200 {
201  textsearch_unregister(&bm_ops);
202 }
203 
204 MODULE_LICENSE("GPL");
205 
206 module_init(init_bm);
207 module_exit(exit_bm);