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