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/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#define DEBUGP(args, format...)
49
50struct ts_bm
51{
52	u8 *		pattern;
53	unsigned int	patlen;
54	unsigned int 	bad_shift[ASIZE];
55	unsigned int	good_shift[0];
56};
57
58static unsigned int bm_find(struct ts_config *conf, struct ts_state *state)
59{
60	struct ts_bm *bm = ts_config_priv(conf);
61	unsigned int i, text_len, consumed = state->offset;
62	const u8 *text;
63	int shift = bm->patlen - 1, bs;
64	const u8 icase = conf->flags & TS_IGNORECASE;
65
66	for (;;) {
67		text_len = conf->get_next_block(consumed, &text, conf, state);
68
69		if (unlikely(text_len == 0))
70			break;
71
72		while (shift < text_len) {
73			DEBUGP("Searching in position %d (%c)\n",
74				shift, text[shift]);
75			for (i = 0; i < bm->patlen; i++)
76				if ((icase ? toupper(text[shift-i])
77				    : text[shift-i])
78					!= bm->pattern[bm->patlen-1-i])
79				     goto next;
80
81			/* London calling... */
82			DEBUGP("found!\n");
83			return consumed += (shift-(bm->patlen-1));
84
85next:			bs = bm->bad_shift[text[shift-i]];
86
87			/* Now jumping to... */
88			shift = max_t(int, shift-i+bs, shift+bm->good_shift[i]);
89		}
90		consumed += text_len;
91	}
92
93	return UINT_MAX;
94}
95
96static int subpattern(u8 *pattern, int i, int j, int g)
97{
98	int x = i+g-1, y = j+g-1, ret = 0;
99
100	while(pattern[x--] == pattern[y--]) {
101		if (y < 0) {
102			ret = 1;
103			break;
104		}
105		if (--g == 0) {
106			ret = pattern[i-1] != pattern[j-1];
107			break;
108		}
109	}
110
111	return ret;
112}
113
114static void compute_prefix_tbl(struct ts_bm *bm, int flags)
115{
116	int i, j, g;
117
118	for (i = 0; i < ASIZE; i++)
119		bm->bad_shift[i] = bm->patlen;
120	for (i = 0; i < bm->patlen - 1; i++) {
121		bm->bad_shift[bm->pattern[i]] = bm->patlen - 1 - i;
122		if (flags & TS_IGNORECASE)
123			bm->bad_shift[tolower(bm->pattern[i])]
124			    = bm->patlen - 1 - i;
125	}
126
127	/* Compute the good shift array, used to match reocurrences
128	 * of a subpattern */
129	bm->good_shift[0] = 1;
130	for (i = 1; i < bm->patlen; i++)
131		bm->good_shift[i] = bm->patlen;
132        for (i = bm->patlen-1, g = 1; i > 0; g++, i--) {
133		for (j = i-1; j >= 1-g ; j--)
134			if (subpattern(bm->pattern, i, j, g)) {
135				bm->good_shift[g] = bm->patlen-j-g;
136				break;
137			}
138	}
139}
140
141static struct ts_config *bm_init(const void *pattern, unsigned int len,
142				 gfp_t gfp_mask, int flags)
143{
144	struct ts_config *conf;
145	struct ts_bm *bm;
146	int i;
147	unsigned int prefix_tbl_len = len * sizeof(unsigned int);
148	size_t priv_size = sizeof(*bm) + len + prefix_tbl_len;
149
150	conf = alloc_ts_config(priv_size, gfp_mask);
151	if (IS_ERR(conf))
152		return conf;
153
154	conf->flags = flags;
155	bm = ts_config_priv(conf);
156	bm->patlen = len;
157	bm->pattern = (u8 *) bm->good_shift + prefix_tbl_len;
158	if (flags & TS_IGNORECASE)
159		for (i = 0; i < len; i++)
160			bm->pattern[i] = toupper(((u8 *)pattern)[i]);
161	else
162		memcpy(bm->pattern, pattern, len);
163	compute_prefix_tbl(bm, flags);
164
165	return conf;
166}
167
168static void *bm_get_pattern(struct ts_config *conf)
169{
170	struct ts_bm *bm = ts_config_priv(conf);
171	return bm->pattern;
172}
173
174static unsigned int bm_get_pattern_len(struct ts_config *conf)
175{
176	struct ts_bm *bm = ts_config_priv(conf);
177	return bm->patlen;
178}
179
180static struct ts_ops bm_ops = {
181	.name		  = "bm",
182	.find		  = bm_find,
183	.init		  = bm_init,
184	.get_pattern	  = bm_get_pattern,
185	.get_pattern_len  = bm_get_pattern_len,
186	.owner		  = THIS_MODULE,
187	.list		  = LIST_HEAD_INIT(bm_ops.list)
188};
189
190static int __init init_bm(void)
191{
192	return textsearch_register(&bm_ops);
193}
194
195static void __exit exit_bm(void)
196{
197	textsearch_unregister(&bm_ops);
198}
199
200MODULE_LICENSE("GPL");
201
202module_init(init_bm);
203module_exit(exit_bm);
204