xref: /kernel/linux/linux-5.10/lib/ts_bm.c (revision 8c2ecf20) 
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * lib/ts_bm.c		Boyer-Moore text search implementation
4 *
5 * Authors:	Pablo Neira Ayuso <pablo@eurodev.net>
6 *
7 * ==========================================================================
8 *
9 *   Implements Boyer-Moore string matching algorithm:
10 *
11 *   [1] A Fast String Searching Algorithm, R.S. Boyer and Moore.
12 *       Communications of the Association for Computing Machinery,
13 *       20(10), 1977, pp. 762-772.
14 *       https://www.cs.utexas.edu/users/moore/publications/fstrpos.pdf
15 *
16 *   [2] Handbook of Exact String Matching Algorithms, Thierry Lecroq, 2004
17 *       http://www-igm.univ-mlv.fr/~lecroq/string/string.pdf
18 *
19 *   Note: Since Boyer-Moore (BM) performs searches for matchings from right
20 *   to left, it's still possible that a matching could be spread over
21 *   multiple blocks, in that case this algorithm won't find any coincidence.
22 *
23 *   If you're willing to ensure that such thing won't ever happen, use the
24 *   Knuth-Pratt-Morris (KMP) implementation instead. In conclusion, choose
25 *   the proper string search algorithm depending on your setting.
26 *
27 *   Say you're using the textsearch infrastructure for filtering, NIDS or
28 *   any similar security focused purpose, then go KMP. Otherwise, if you
29 *   really care about performance, say you're classifying packets to apply
30 *   Quality of Service (QoS) policies, and you don't mind about possible
31 *   matchings spread over multiple fragments, then go BM.
32 */
33
34#include <linux/kernel.h>
35#include <linux/module.h>
36#include <linux/types.h>
37#include <linux/string.h>
38#include <linux/ctype.h>
39#include <linux/textsearch.h>
40
41/* Alphabet size, use ASCII */
42#define ASIZE 256
43
44#if 0
45#define DEBUGP printk
46#else
47#define DEBUGP(args, format...)
48#endif
49
50struct ts_bm
51{
52	u8 *		pattern;
53	unsigned int	patlen;
54	unsigned int 	bad_shift[ASIZE];
55	unsigned int	good_shift[];
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 bs;
64	const u8 icase = conf->flags & TS_IGNORECASE;
65
66	for (;;) {
67		int shift = bm->patlen - 1;
68
69		text_len = conf->get_next_block(consumed, &text, conf, state);
70
71		if (unlikely(text_len == 0))
72			break;
73
74		while (shift < text_len) {
75			DEBUGP("Searching in position %d (%c)\n",
76				shift, text[shift]);
77			for (i = 0; i < bm->patlen; i++)
78				if ((icase ? toupper(text[shift-i])
79				    : text[shift-i])
80					!= bm->pattern[bm->patlen-1-i])
81				     goto next;
82
83			/* London calling... */
84			DEBUGP("found!\n");
85			return consumed += (shift-(bm->patlen-1));
86
87next:			bs = bm->bad_shift[text[shift-i]];
88
89			/* Now jumping to... */
90			shift = max_t(int, shift-i+bs, shift+bm->good_shift[i]);
91		}
92		consumed += text_len;
93	}
94
95	return UINT_MAX;
96}
97
98static int subpattern(u8 *pattern, int i, int j, int g)
99{
100	int x = i+g-1, y = j+g-1, ret = 0;
101
102	while(pattern[x--] == pattern[y--]) {
103		if (y < 0) {
104			ret = 1;
105			break;
106		}
107		if (--g == 0) {
108			ret = pattern[i-1] != pattern[j-1];
109			break;
110		}
111	}
112
113	return ret;
114}
115
116static void compute_prefix_tbl(struct ts_bm *bm, int flags)
117{
118	int i, j, g;
119
120	for (i = 0; i < ASIZE; i++)
121		bm->bad_shift[i] = bm->patlen;
122	for (i = 0; i < bm->patlen - 1; i++) {
123		bm->bad_shift[bm->pattern[i]] = bm->patlen - 1 - i;
124		if (flags & TS_IGNORECASE)
125			bm->bad_shift[tolower(bm->pattern[i])]
126			    = bm->patlen - 1 - i;
127	}
128
129	/* Compute the good shift array, used to match reocurrences
130	 * of a subpattern */
131	bm->good_shift[0] = 1;
132	for (i = 1; i < bm->patlen; i++)
133		bm->good_shift[i] = bm->patlen;
134        for (i = bm->patlen-1, g = 1; i > 0; g++, i--) {
135		for (j = i-1; j >= 1-g ; j--)
136			if (subpattern(bm->pattern, i, j, g)) {
137				bm->good_shift[g] = bm->patlen-j-g;
138				break;
139			}
140	}
141}
142
143static struct ts_config *bm_init(const void *pattern, unsigned int len,
144				 gfp_t gfp_mask, int flags)
145{
146	struct ts_config *conf;
147	struct ts_bm *bm;
148	int i;
149	unsigned int prefix_tbl_len = len * sizeof(unsigned int);
150	size_t priv_size = sizeof(*bm) + len + prefix_tbl_len;
151
152	conf = alloc_ts_config(priv_size, gfp_mask);
153	if (IS_ERR(conf))
154		return conf;
155
156	conf->flags = flags;
157	bm = ts_config_priv(conf);
158	bm->patlen = len;
159	bm->pattern = (u8 *) bm->good_shift + prefix_tbl_len;
160	if (flags & TS_IGNORECASE)
161		for (i = 0; i < len; i++)
162			bm->pattern[i] = toupper(((u8 *)pattern)[i]);
163	else
164		memcpy(bm->pattern, pattern, len);
165	compute_prefix_tbl(bm, flags);
166
167	return conf;
168}
169
170static void *bm_get_pattern(struct ts_config *conf)
171{
172	struct ts_bm *bm = ts_config_priv(conf);
173	return bm->pattern;
174}
175
176static unsigned int bm_get_pattern_len(struct ts_config *conf)
177{
178	struct ts_bm *bm = ts_config_priv(conf);
179	return bm->patlen;
180}
181
182static struct ts_ops bm_ops = {
183	.name		  = "bm",
184	.find		  = bm_find,
185	.init		  = bm_init,
186	.get_pattern	  = bm_get_pattern,
187	.get_pattern_len  = bm_get_pattern_len,
188	.owner		  = THIS_MODULE,
189	.list		  = LIST_HEAD_INIT(bm_ops.list)
190};
191
192static int __init init_bm(void)
193{
194	return textsearch_register(&bm_ops);
195}
196
197static void __exit exit_bm(void)
198{
199	textsearch_unregister(&bm_ops);
200}
201
202MODULE_LICENSE("GPL");
203
204module_init(init_bm);
205module_exit(exit_bm);
206