1 // SPDX-License-Identifier: GPL-2.0-or-later <<
2 /* 1 /*
3 * lib/textsearch.c Generic text search in 2 * lib/textsearch.c Generic text search interface
4 * 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 *
5 * Authors: Thomas Graf <tgraf@suug.ch> 9 * Authors: Thomas Graf <tgraf@suug.ch>
6 * Pablo Neira Ayuso <pablo@netfi 10 * Pablo Neira Ayuso <pablo@netfilter.org>
7 * 11 *
8 * =========================================== 12 * ==========================================================================
9 */ 13 */
10 14
11 /** 15 /**
12 * DOC: ts_intro 16 * DOC: ts_intro
13 * INTRODUCTION 17 * INTRODUCTION
14 * 18 *
15 * The textsearch infrastructure provides te 19 * The textsearch infrastructure provides text searching facilities for
16 * both linear and non-linear data. Individu 20 * both linear and non-linear data. Individual search algorithms are
17 * implemented in modules and chosen by the 21 * implemented in modules and chosen by the user.
18 * 22 *
19 * ARCHITECTURE 23 * ARCHITECTURE
20 * 24 *
21 * .. code-block:: none 25 * .. code-block:: none
22 * 26 *
23 * User 27 * User
24 * +----------------+ 28 * +----------------+
25 * | finish()|<--------------(6)--- 29 * | finish()|<--------------(6)-----------------+
26 * |get_next_block()|<--------------(5)--- 30 * |get_next_block()|<--------------(5)---------------+ |
27 * | | 31 * | | Algorithm | |
28 * | | + 32 * | | +------------------------------+
29 * | | | 33 * | | | init() find() destroy() |
30 * | | + 34 * | | +------------------------------+
31 * | | Core API 35 * | | Core API ^ ^ ^
32 * | | +-------------- 36 * | | +---------------+ (2) (4) (8)
33 * | (1)|----->| prepare() 37 * | (1)|----->| prepare() |---+ | |
34 * | (3)|----->| find()/next() 38 * | (3)|----->| find()/next() |-----------+ |
35 * | (7)|----->| destroy() 39 * | (7)|----->| destroy() |----------------------+
36 * +----------------+ +-------------- 40 * +----------------+ +---------------+
37 * 41 *
38 * (1) User configures a search by calling t 42 * (1) User configures a search by calling textsearch_prepare() specifying
39 * the search parameters such as the pat 43 * the search parameters such as the pattern and algorithm name.
40 * (2) Core requests the algorithm to alloca 44 * (2) Core requests the algorithm to allocate and initialize a search
41 * configuration according to the specif 45 * configuration according to the specified parameters.
42 * (3) User starts the search(es) by calling 46 * (3) User starts the search(es) by calling textsearch_find() or
43 * textsearch_next() to fetch subsequent 47 * textsearch_next() to fetch subsequent occurrences. A state variable
44 * is provided to the algorithm to store 48 * is provided to the algorithm to store persistent variables.
45 * (4) Core eventually resets the search off 49 * (4) Core eventually resets the search offset and forwards the find()
46 * request to the algorithm. 50 * request to the algorithm.
47 * (5) Algorithm calls get_next_block() prov 51 * (5) Algorithm calls get_next_block() provided by the user continuously
48 * to fetch the data to be searched in b 52 * to fetch the data to be searched in block by block.
49 * (6) Algorithm invokes finish() after the 53 * (6) Algorithm invokes finish() after the last call to get_next_block
50 * to clean up any leftovers from get_ne 54 * to clean up any leftovers from get_next_block. (Optional)
51 * (7) User destroys the configuration by ca 55 * (7) User destroys the configuration by calling textsearch_destroy().
52 * (8) Core notifies the algorithm to destro 56 * (8) Core notifies the algorithm to destroy algorithm specific
53 * allocations. (Optional) 57 * allocations. (Optional)
54 * 58 *
55 * USAGE 59 * USAGE
56 * 60 *
57 * Before a search can be performed, a confi 61 * Before a search can be performed, a configuration must be created
58 * by calling textsearch_prepare() specifyin 62 * by calling textsearch_prepare() specifying the searching algorithm,
59 * the pattern to look for and flags. As a f 63 * the pattern to look for and flags. As a flag, you can set TS_IGNORECASE
60 * to perform case insensitive matching. But 64 * to perform case insensitive matching. But it might slow down
61 * performance of algorithm, so you should u 65 * performance of algorithm, so you should use it at own your risk.
62 * The returned configuration may then be us 66 * The returned configuration may then be used for an arbitrary
63 * amount of times and even in parallel as l 67 * amount of times and even in parallel as long as a separate struct
64 * ts_state variable is provided to every in 68 * ts_state variable is provided to every instance.
65 * 69 *
66 * The actual search is performed by either 70 * The actual search is performed by either calling
67 * textsearch_find_continuous() for linear d 71 * textsearch_find_continuous() for linear data or by providing
68 * an own get_next_block() implementation an 72 * an own get_next_block() implementation and
69 * calling textsearch_find(). Both functions 73 * calling textsearch_find(). Both functions return
70 * the position of the first occurrence of t 74 * the position of the first occurrence of the pattern or UINT_MAX if
71 * no match was found. Subsequent occurrence 75 * no match was found. Subsequent occurrences can be found by calling
72 * textsearch_next() regardless of the linea 76 * textsearch_next() regardless of the linearity of the data.
73 * 77 *
74 * Once you're done using a configuration it 78 * Once you're done using a configuration it must be given back via
75 * textsearch_destroy. 79 * textsearch_destroy.
76 * 80 *
77 * EXAMPLE:: 81 * EXAMPLE::
78 * 82 *
79 * int pos; 83 * int pos;
80 * struct ts_config *conf; 84 * struct ts_config *conf;
81 * struct ts_state state; 85 * struct ts_state state;
82 * const char *pattern = "chicken"; 86 * const char *pattern = "chicken";
83 * const char *example = "We dance the funky 87 * const char *example = "We dance the funky chicken";
84 * 88 *
85 * conf = textsearch_prepare("kmp", pattern, 89 * conf = textsearch_prepare("kmp", pattern, strlen(pattern),
86 * GFP_KERNEL, TS_ 90 * GFP_KERNEL, TS_AUTOLOAD);
87 * if (IS_ERR(conf)) { 91 * if (IS_ERR(conf)) {
88 * err = PTR_ERR(conf); 92 * err = PTR_ERR(conf);
89 * goto errout; 93 * goto errout;
90 * } 94 * }
91 * 95 *
92 * pos = textsearch_find_continuous(conf, &s !! 96 * pos = textsearch_find_continuous(conf, \&state, example, strlen(example));
93 * if (pos != UINT_MAX) 97 * if (pos != UINT_MAX)
94 * panic("Oh my god, dancing chickens at !! 98 * panic("Oh my god, dancing chickens at \%d\n", pos);
95 * 99 *
96 * textsearch_destroy(conf); 100 * textsearch_destroy(conf);
97 */ 101 */
98 /* =========================================== 102 /* ========================================================================== */
99 103
100 #include <linux/module.h> 104 #include <linux/module.h>
101 #include <linux/types.h> 105 #include <linux/types.h>
102 #include <linux/string.h> 106 #include <linux/string.h>
103 #include <linux/init.h> 107 #include <linux/init.h>
104 #include <linux/rculist.h> 108 #include <linux/rculist.h>
105 #include <linux/rcupdate.h> 109 #include <linux/rcupdate.h>
106 #include <linux/err.h> 110 #include <linux/err.h>
107 #include <linux/textsearch.h> 111 #include <linux/textsearch.h>
108 #include <linux/slab.h> 112 #include <linux/slab.h>
109 113
110 static LIST_HEAD(ts_ops); 114 static LIST_HEAD(ts_ops);
111 static DEFINE_SPINLOCK(ts_mod_lock); 115 static DEFINE_SPINLOCK(ts_mod_lock);
112 116
113 static inline struct ts_ops *lookup_ts_algo(co 117 static inline struct ts_ops *lookup_ts_algo(const char *name)
114 { 118 {
115 struct ts_ops *o; 119 struct ts_ops *o;
116 120
117 rcu_read_lock(); 121 rcu_read_lock();
118 list_for_each_entry_rcu(o, &ts_ops, li 122 list_for_each_entry_rcu(o, &ts_ops, list) {
119 if (!strcmp(name, o->name)) { 123 if (!strcmp(name, o->name)) {
120 if (!try_module_get(o- 124 if (!try_module_get(o->owner))
121 o = NULL; 125 o = NULL;
122 rcu_read_unlock(); 126 rcu_read_unlock();
123 return o; 127 return o;
124 } 128 }
125 } 129 }
126 rcu_read_unlock(); 130 rcu_read_unlock();
127 131
128 return NULL; 132 return NULL;
129 } 133 }
130 134
131 /** 135 /**
132 * textsearch_register - register a textsearch 136 * textsearch_register - register a textsearch module
133 * @ops: operations lookup table 137 * @ops: operations lookup table
134 * 138 *
135 * This function must be called by textsearch 139 * This function must be called by textsearch modules to announce
136 * their presence. The specified &@ops must ha 140 * their presence. The specified &@ops must have %name set to a
137 * unique identifier and the callbacks find(), 141 * unique identifier and the callbacks find(), init(), get_pattern(),
138 * and get_pattern_len() must be implemented. 142 * and get_pattern_len() must be implemented.
139 * 143 *
140 * Returns 0 or -EEXISTS if another module has 144 * Returns 0 or -EEXISTS if another module has already registered
141 * with same name. 145 * with same name.
142 */ 146 */
143 int textsearch_register(struct ts_ops *ops) 147 int textsearch_register(struct ts_ops *ops)
144 { 148 {
145 int err = -EEXIST; 149 int err = -EEXIST;
146 struct ts_ops *o; 150 struct ts_ops *o;
147 151
148 if (ops->name == NULL || ops->find == 152 if (ops->name == NULL || ops->find == NULL || ops->init == NULL ||
149 ops->get_pattern == NULL || ops->g 153 ops->get_pattern == NULL || ops->get_pattern_len == NULL)
150 return -EINVAL; 154 return -EINVAL;
151 155
152 spin_lock(&ts_mod_lock); 156 spin_lock(&ts_mod_lock);
153 list_for_each_entry(o, &ts_ops, list) 157 list_for_each_entry(o, &ts_ops, list) {
154 if (!strcmp(ops->name, o->name 158 if (!strcmp(ops->name, o->name))
155 goto errout; 159 goto errout;
156 } 160 }
157 161
158 list_add_tail_rcu(&ops->list, &ts_ops) 162 list_add_tail_rcu(&ops->list, &ts_ops);
159 err = 0; 163 err = 0;
160 errout: 164 errout:
161 spin_unlock(&ts_mod_lock); 165 spin_unlock(&ts_mod_lock);
162 return err; 166 return err;
163 } 167 }
164 EXPORT_SYMBOL(textsearch_register); 168 EXPORT_SYMBOL(textsearch_register);
165 169
166 /** 170 /**
167 * textsearch_unregister - unregister a textse 171 * textsearch_unregister - unregister a textsearch module
168 * @ops: operations lookup table 172 * @ops: operations lookup table
169 * 173 *
170 * This function must be called by textsearch 174 * This function must be called by textsearch modules to announce
171 * their disappearance for examples when the m 175 * their disappearance for examples when the module gets unloaded.
172 * The &ops parameter must be the same as the 176 * The &ops parameter must be the same as the one during the
173 * registration. 177 * registration.
174 * 178 *
175 * Returns 0 on success or -ENOENT if no match 179 * Returns 0 on success or -ENOENT if no matching textsearch
176 * registration was found. 180 * registration was found.
177 */ 181 */
178 int textsearch_unregister(struct ts_ops *ops) 182 int textsearch_unregister(struct ts_ops *ops)
179 { 183 {
180 int err = 0; 184 int err = 0;
181 struct ts_ops *o; 185 struct ts_ops *o;
182 186
183 spin_lock(&ts_mod_lock); 187 spin_lock(&ts_mod_lock);
184 list_for_each_entry(o, &ts_ops, list) 188 list_for_each_entry(o, &ts_ops, list) {
185 if (o == ops) { 189 if (o == ops) {
186 list_del_rcu(&o->list) 190 list_del_rcu(&o->list);
187 goto out; 191 goto out;
188 } 192 }
189 } 193 }
190 194
191 err = -ENOENT; 195 err = -ENOENT;
192 out: 196 out:
193 spin_unlock(&ts_mod_lock); 197 spin_unlock(&ts_mod_lock);
194 return err; 198 return err;
195 } 199 }
196 EXPORT_SYMBOL(textsearch_unregister); 200 EXPORT_SYMBOL(textsearch_unregister);
197 201
198 struct ts_linear_state 202 struct ts_linear_state
199 { 203 {
200 unsigned int len; 204 unsigned int len;
201 const void *data; 205 const void *data;
202 }; 206 };
203 207
204 static unsigned int get_linear_data(unsigned i 208 static unsigned int get_linear_data(unsigned int consumed, const u8 **dst,
205 struct ts_ 209 struct ts_config *conf,
206 struct ts_ 210 struct ts_state *state)
207 { 211 {
208 struct ts_linear_state *st = (struct t 212 struct ts_linear_state *st = (struct ts_linear_state *) state->cb;
209 213
210 if (likely(consumed < st->len)) { 214 if (likely(consumed < st->len)) {
211 *dst = st->data + consumed; 215 *dst = st->data + consumed;
212 return st->len - consumed; 216 return st->len - consumed;
213 } 217 }
214 218
215 return 0; 219 return 0;
216 } 220 }
217 221
218 /** 222 /**
219 * textsearch_find_continuous - search a patte 223 * textsearch_find_continuous - search a pattern in continuous/linear data
220 * @conf: search configuration 224 * @conf: search configuration
221 * @state: search state 225 * @state: search state
222 * @data: data to search in 226 * @data: data to search in
223 * @len: length of data 227 * @len: length of data
224 * 228 *
225 * A simplified version of textsearch_find() f 229 * A simplified version of textsearch_find() for continuous/linear data.
226 * Call textsearch_next() to retrieve subseque 230 * Call textsearch_next() to retrieve subsequent matches.
227 * 231 *
228 * Returns the position of first occurrence of 232 * Returns the position of first occurrence of the pattern or
229 * %UINT_MAX if no occurrence was found. 233 * %UINT_MAX if no occurrence was found.
230 */ 234 */
231 unsigned int textsearch_find_continuous(struct 235 unsigned int textsearch_find_continuous(struct ts_config *conf,
232 struct 236 struct ts_state *state,
233 const 237 const void *data, unsigned int len)
234 { 238 {
235 struct ts_linear_state *st = (struct t 239 struct ts_linear_state *st = (struct ts_linear_state *) state->cb;
236 240
237 conf->get_next_block = get_linear_data 241 conf->get_next_block = get_linear_data;
238 st->data = data; 242 st->data = data;
239 st->len = len; 243 st->len = len;
240 244
241 return textsearch_find(conf, state); 245 return textsearch_find(conf, state);
242 } 246 }
243 EXPORT_SYMBOL(textsearch_find_continuous); 247 EXPORT_SYMBOL(textsearch_find_continuous);
244 248
245 /** 249 /**
246 * textsearch_prepare - Prepare a search 250 * textsearch_prepare - Prepare a search
247 * @algo: name of search algorithm 251 * @algo: name of search algorithm
248 * @pattern: pattern data 252 * @pattern: pattern data
249 * @len: length of pattern 253 * @len: length of pattern
250 * @gfp_mask: allocation mask 254 * @gfp_mask: allocation mask
251 * @flags: search flags 255 * @flags: search flags
252 * 256 *
253 * Looks up the search algorithm module and cr 257 * Looks up the search algorithm module and creates a new textsearch
254 * configuration for the specified pattern. 258 * configuration for the specified pattern.
255 * 259 *
256 * Note: The format of the pattern may not be 260 * Note: The format of the pattern may not be compatible between
257 * the various search algorithms. 261 * the various search algorithms.
258 * 262 *
259 * Returns a new textsearch configuration acco 263 * Returns a new textsearch configuration according to the specified
260 * parameters or a ERR_PTR(). If a zero length 264 * parameters or a ERR_PTR(). If a zero length pattern is passed, this
261 * function returns EINVAL. 265 * function returns EINVAL.
262 */ 266 */
263 struct ts_config *textsearch_prepare(const cha 267 struct ts_config *textsearch_prepare(const char *algo, const void *pattern,
264 unsigned 268 unsigned int len, gfp_t gfp_mask, int flags)
265 { 269 {
266 int err = -ENOENT; 270 int err = -ENOENT;
267 struct ts_config *conf; 271 struct ts_config *conf;
268 struct ts_ops *ops; 272 struct ts_ops *ops;
269 273
270 if (len == 0) 274 if (len == 0)
271 return ERR_PTR(-EINVAL); 275 return ERR_PTR(-EINVAL);
272 276
273 ops = lookup_ts_algo(algo); 277 ops = lookup_ts_algo(algo);
274 #ifdef CONFIG_MODULES 278 #ifdef CONFIG_MODULES
275 /* 279 /*
276 * Why not always autoload you may ask 280 * Why not always autoload you may ask. Some users are
277 * in a situation where requesting a m 281 * in a situation where requesting a module may deadlock,
278 * especially when the module is locat 282 * especially when the module is located on a NFS mount.
279 */ 283 */
280 if (ops == NULL && flags & TS_AUTOLOAD 284 if (ops == NULL && flags & TS_AUTOLOAD) {
281 request_module("ts_%s", algo); 285 request_module("ts_%s", algo);
282 ops = lookup_ts_algo(algo); 286 ops = lookup_ts_algo(algo);
283 } 287 }
284 #endif 288 #endif
285 289
286 if (ops == NULL) 290 if (ops == NULL)
287 goto errout; 291 goto errout;
288 292
289 conf = ops->init(pattern, len, gfp_mas 293 conf = ops->init(pattern, len, gfp_mask, flags);
290 if (IS_ERR(conf)) { 294 if (IS_ERR(conf)) {
291 err = PTR_ERR(conf); 295 err = PTR_ERR(conf);
292 goto errout; 296 goto errout;
293 } 297 }
294 298
295 conf->ops = ops; 299 conf->ops = ops;
296 return conf; 300 return conf;
297 301
298 errout: 302 errout:
299 if (ops) 303 if (ops)
300 module_put(ops->owner); 304 module_put(ops->owner);
301 305
302 return ERR_PTR(err); 306 return ERR_PTR(err);
303 } 307 }
304 EXPORT_SYMBOL(textsearch_prepare); 308 EXPORT_SYMBOL(textsearch_prepare);
305 309
306 /** 310 /**
307 * textsearch_destroy - destroy a search confi 311 * textsearch_destroy - destroy a search configuration
308 * @conf: search configuration 312 * @conf: search configuration
309 * 313 *
310 * Releases all references of the configuratio 314 * Releases all references of the configuration and frees
311 * up the memory. 315 * up the memory.
312 */ 316 */
313 void textsearch_destroy(struct ts_config *conf 317 void textsearch_destroy(struct ts_config *conf)
314 { 318 {
315 if (conf->ops) { 319 if (conf->ops) {
316 if (conf->ops->destroy) 320 if (conf->ops->destroy)
317 conf->ops->destroy(con 321 conf->ops->destroy(conf);
318 module_put(conf->ops->owner); 322 module_put(conf->ops->owner);
319 } 323 }
320 324
321 kfree(conf); 325 kfree(conf);
322 } 326 }
323 EXPORT_SYMBOL(textsearch_destroy); 327 EXPORT_SYMBOL(textsearch_destroy);
324 328
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.