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