blob: 9f069c07692b4a0460b3edf35049597944063237 [file] [log] [blame]
Wolfgang Denka6826fb2010-06-20 13:17:12 +02001/*
2 * This implementation is based on code from uClibc-0.9.30.3 but was
3 * modified and extended for use within U-Boot.
4 *
5 * Copyright (C) 2010 Wolfgang Denk <wd@denx.de>
6 *
7 * Original license header:
8 *
9 * Copyright (C) 1993, 1995, 1996, 1997, 2002 Free Software Foundation, Inc.
10 * This file is part of the GNU C Library.
11 * Contributed by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1993.
12 *
13 * The GNU C Library is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU Lesser General Public
15 * License as published by the Free Software Foundation; either
16 * version 2.1 of the License, or (at your option) any later version.
17 *
18 * The GNU C Library is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 * Lesser General Public License for more details.
22 *
23 * You should have received a copy of the GNU Lesser General Public
24 * License along with the GNU C Library; if not, write to the Free
25 * Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
26 * 02111-1307 USA.
27 */
28
29#include <errno.h>
30#include <malloc.h>
31
32#ifdef USE_HOSTCC /* HOST build */
33# include <string.h>
34# include <assert.h>
35
36# ifndef debug
37# ifdef DEBUG
38# define debug(fmt,args...) printf(fmt ,##args)
39# else
40# define debug(fmt,args...)
41# endif
42# endif
43#else /* U-Boot build */
44# include <common.h>
45# include <linux/string.h>
46#endif
47
Andreas Bießmannfc5fc762010-10-01 22:51:02 +020048#ifndef CONFIG_ENV_MIN_ENTRIES /* minimum number of entries */
49#define CONFIG_ENV_MIN_ENTRIES 64
50#endif
Wolfgang Denkea882ba2010-06-20 23:33:59 +020051#ifndef CONFIG_ENV_MAX_ENTRIES /* maximum number of entries */
52#define CONFIG_ENV_MAX_ENTRIES 512
53#endif
54
Wolfgang Denka6826fb2010-06-20 13:17:12 +020055#include "search.h"
56
57/*
58 * [Aho,Sethi,Ullman] Compilers: Principles, Techniques and Tools, 1986
Wolfgang Denk071bc922010-10-27 22:48:30 +020059 * [Knuth] The Art of Computer Programming, part 3 (6.4)
Wolfgang Denka6826fb2010-06-20 13:17:12 +020060 */
61
62/*
Wolfgang Denka6826fb2010-06-20 13:17:12 +020063 * The reentrant version has no static variables to maintain the state.
64 * Instead the interface of all functions is extended to take an argument
65 * which describes the current status.
66 */
67typedef struct _ENTRY {
68 unsigned int used;
69 ENTRY entry;
70} _ENTRY;
71
72
73/*
74 * hcreate()
75 */
76
77/*
78 * For the used double hash method the table size has to be a prime. To
79 * correct the user given table size we need a prime test. This trivial
80 * algorithm is adequate because
81 * a) the code is (most probably) called a few times per program run and
82 * b) the number is small because the table must fit in the core
83 * */
84static int isprime(unsigned int number)
85{
86 /* no even number will be passed */
87 unsigned int div = 3;
88
89 while (div * div < number && number % div != 0)
90 div += 2;
91
92 return number % div != 0;
93}
94
Wolfgang Denka6826fb2010-06-20 13:17:12 +020095/*
96 * Before using the hash table we must allocate memory for it.
97 * Test for an existing table are done. We allocate one element
98 * more as the found prime number says. This is done for more effective
99 * indexing as explained in the comment for the hsearch function.
100 * The contents of the table is zeroed, especially the field used
101 * becomes zero.
102 */
Mike Frysinger2eb15732010-12-08 06:26:04 -0500103
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200104int hcreate_r(size_t nel, struct hsearch_data *htab)
105{
106 /* Test for correct arguments. */
107 if (htab == NULL) {
108 __set_errno(EINVAL);
109 return 0;
110 }
111
112 /* There is still another table active. Return with error. */
113 if (htab->table != NULL)
114 return 0;
115
116 /* Change nel to the first prime number not smaller as nel. */
117 nel |= 1; /* make odd */
118 while (!isprime(nel))
119 nel += 2;
120
121 htab->size = nel;
122 htab->filled = 0;
123
124 /* allocate memory and zero out */
125 htab->table = (_ENTRY *) calloc(htab->size + 1, sizeof(_ENTRY));
126 if (htab->table == NULL)
127 return 0;
128
129 /* everything went alright */
130 return 1;
131}
132
133
134/*
135 * hdestroy()
136 */
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200137
138/*
139 * After using the hash table it has to be destroyed. The used memory can
140 * be freed and the local static variable can be marked as not used.
141 */
Mike Frysinger2eb15732010-12-08 06:26:04 -0500142
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200143void hdestroy_r(struct hsearch_data *htab)
144{
145 int i;
146
147 /* Test for correct arguments. */
148 if (htab == NULL) {
149 __set_errno(EINVAL);
150 return;
151 }
152
153 /* free used memory */
154 for (i = 1; i <= htab->size; ++i) {
155 if (htab->table[i].used) {
156 ENTRY *ep = &htab->table[i].entry;
157
158 free(ep->key);
159 free(ep->data);
160 }
161 }
162 free(htab->table);
163
164 /* the sign for an existing table is an value != NULL in htable */
165 htab->table = NULL;
166}
167
168/*
169 * hsearch()
170 */
171
172/*
173 * This is the search function. It uses double hashing with open addressing.
174 * The argument item.key has to be a pointer to an zero terminated, most
175 * probably strings of chars. The function for generating a number of the
176 * strings is simple but fast. It can be replaced by a more complex function
177 * like ajw (see [Aho,Sethi,Ullman]) if the needs are shown.
178 *
179 * We use an trick to speed up the lookup. The table is created by hcreate
180 * with one more element available. This enables us to use the index zero
181 * special. This index will never be used because we store the first hash
182 * index in the field used where zero means not used. Every other value
183 * means used. The used field can be used as a first fast comparison for
184 * equality of the stored and the parameter value. This helps to prevent
185 * unnecessary expensive calls of strcmp.
186 *
187 * This implementation differs from the standard library version of
188 * this function in a number of ways:
189 *
190 * - While the standard version does not make any assumptions about
191 * the type of the stored data objects at all, this implementation
192 * works with NUL terminated strings only.
193 * - Instead of storing just pointers to the original objects, we
194 * create local copies so the caller does not need to care about the
195 * data any more.
196 * - The standard implementation does not provide a way to update an
197 * existing entry. This version will create a new entry or update an
198 * existing one when both "action == ENTER" and "item.data != NULL".
199 * - Instead of returning 1 on success, we return the index into the
200 * internal hash table, which is also guaranteed to be positive.
201 * This allows us direct access to the found hash table slot for
202 * example for functions like hdelete().
203 */
204
Mike Frysinger560d4242010-12-17 16:51:59 -0500205int hmatch_r(const char *match, int last_idx, ENTRY ** retval,
206 struct hsearch_data *htab)
207{
208 unsigned int idx;
209 size_t key_len = strlen(match);
210
211 for (idx = last_idx + 1; idx < htab->size; ++idx) {
212 if (!htab->table[idx].used)
213 continue;
214 if (!strncmp(match, htab->table[idx].entry.key, key_len)) {
215 *retval = &htab->table[idx].entry;
216 return idx;
217 }
218 }
219
220 __set_errno(ESRCH);
221 *retval = NULL;
222 return 0;
223}
224
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200225int hsearch_r(ENTRY item, ACTION action, ENTRY ** retval,
226 struct hsearch_data *htab)
227{
228 unsigned int hval;
229 unsigned int count;
230 unsigned int len = strlen(item.key);
231 unsigned int idx;
232
233 /* Compute an value for the given string. Perhaps use a better method. */
234 hval = len;
235 count = len;
236 while (count-- > 0) {
237 hval <<= 4;
238 hval += item.key[count];
239 }
240
241 /*
242 * First hash function:
243 * simply take the modul but prevent zero.
244 */
245 hval %= htab->size;
246 if (hval == 0)
247 ++hval;
248
249 /* The first index tried. */
250 idx = hval;
251
252 if (htab->table[idx].used) {
253 /*
Wolfgang Denk071bc922010-10-27 22:48:30 +0200254 * Further action might be required according to the
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200255 * action value.
256 */
257 unsigned hval2;
258
259 if (htab->table[idx].used == hval
260 && strcmp(item.key, htab->table[idx].entry.key) == 0) {
261 /* Overwrite existing value? */
262 if ((action == ENTER) && (item.data != NULL)) {
263 free(htab->table[idx].entry.data);
264 htab->table[idx].entry.data =
265 strdup(item.data);
266 if (!htab->table[idx].entry.data) {
267 __set_errno(ENOMEM);
268 *retval = NULL;
269 return 0;
270 }
271 }
272 /* return found entry */
273 *retval = &htab->table[idx].entry;
274 return idx;
275 }
276
277 /*
278 * Second hash function:
279 * as suggested in [Knuth]
280 */
281 hval2 = 1 + hval % (htab->size - 2);
282
283 do {
284 /*
Wolfgang Denk071bc922010-10-27 22:48:30 +0200285 * Because SIZE is prime this guarantees to
286 * step through all available indices.
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200287 */
288 if (idx <= hval2)
289 idx = htab->size + idx - hval2;
290 else
291 idx -= hval2;
292
293 /*
294 * If we visited all entries leave the loop
295 * unsuccessfully.
296 */
297 if (idx == hval)
298 break;
299
300 /* If entry is found use it. */
301 if ((htab->table[idx].used == hval)
302 && strcmp(item.key, htab->table[idx].entry.key) == 0) {
303 /* Overwrite existing value? */
304 if ((action == ENTER) && (item.data != NULL)) {
305 free(htab->table[idx].entry.data);
306 htab->table[idx].entry.data =
307 strdup(item.data);
308 if (!htab->table[idx].entry.data) {
309 __set_errno(ENOMEM);
310 *retval = NULL;
311 return 0;
312 }
313 }
314 /* return found entry */
315 *retval = &htab->table[idx].entry;
316 return idx;
317 }
318 }
319 while (htab->table[idx].used);
320 }
321
322 /* An empty bucket has been found. */
323 if (action == ENTER) {
324 /*
Wolfgang Denk071bc922010-10-27 22:48:30 +0200325 * If table is full and another entry should be
326 * entered return with error.
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200327 */
328 if (htab->filled == htab->size) {
329 __set_errno(ENOMEM);
330 *retval = NULL;
331 return 0;
332 }
333
334 /*
335 * Create new entry;
336 * create copies of item.key and item.data
337 */
338 htab->table[idx].used = hval;
339 htab->table[idx].entry.key = strdup(item.key);
340 htab->table[idx].entry.data = strdup(item.data);
341 if (!htab->table[idx].entry.key ||
342 !htab->table[idx].entry.data) {
343 __set_errno(ENOMEM);
344 *retval = NULL;
345 return 0;
346 }
347
348 ++htab->filled;
349
350 /* return new entry */
351 *retval = &htab->table[idx].entry;
352 return 1;
353 }
354
355 __set_errno(ESRCH);
356 *retval = NULL;
357 return 0;
358}
359
360
361/*
362 * hdelete()
363 */
364
365/*
366 * The standard implementation of hsearch(3) does not provide any way
367 * to delete any entries from the hash table. We extend the code to
368 * do that.
369 */
370
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200371int hdelete_r(const char *key, struct hsearch_data *htab)
372{
373 ENTRY e, *ep;
374 int idx;
375
376 debug("hdelete: DELETE key \"%s\"\n", key);
377
378 e.key = (char *)key;
379
380 if ((idx = hsearch_r(e, FIND, &ep, htab)) == 0) {
381 __set_errno(ESRCH);
382 return 0; /* not found */
383 }
384
385 /* free used ENTRY */
386 debug("hdelete: DELETING key \"%s\"\n", key);
387
388 free(ep->key);
389 free(ep->data);
390 htab->table[idx].used = 0;
391
392 --htab->filled;
393
394 return 1;
395}
396
397/*
398 * hexport()
399 */
400
401/*
402 * Export the data stored in the hash table in linearized form.
403 *
404 * Entries are exported as "name=value" strings, separated by an
405 * arbitrary (non-NUL, of course) separator character. This allows to
406 * use this function both when formatting the U-Boot environment for
407 * external storage (using '\0' as separator), but also when using it
408 * for the "printenv" command to print all variables, simply by using
409 * as '\n" as separator. This can also be used for new features like
410 * exporting the environment data as text file, including the option
411 * for later re-import.
412 *
413 * The entries in the result list will be sorted by ascending key
414 * values.
415 *
416 * If the separator character is different from NUL, then any
417 * separator characters and backslash characters in the values will
418 * be escaped by a preceeding backslash in output. This is needed for
419 * example to enable multi-line values, especially when the output
420 * shall later be parsed (for example, for re-import).
421 *
422 * There are several options how the result buffer is handled:
423 *
424 * *resp size
425 * -----------
426 * NULL 0 A string of sufficient length will be allocated.
427 * NULL >0 A string of the size given will be
428 * allocated. An error will be returned if the size is
429 * not sufficient. Any unused bytes in the string will
430 * be '\0'-padded.
431 * !NULL 0 The user-supplied buffer will be used. No length
432 * checking will be performed, i. e. it is assumed that
433 * the buffer size will always be big enough. DANGEROUS.
434 * !NULL >0 The user-supplied buffer will be used. An error will
435 * be returned if the size is not sufficient. Any unused
436 * bytes in the string will be '\0'-padded.
437 */
438
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200439static int cmpkey(const void *p1, const void *p2)
440{
441 ENTRY *e1 = *(ENTRY **) p1;
442 ENTRY *e2 = *(ENTRY **) p2;
443
444 return (strcmp(e1->key, e2->key));
445}
446
447ssize_t hexport_r(struct hsearch_data *htab, const char sep,
448 char **resp, size_t size)
449{
450 ENTRY *list[htab->size];
451 char *res, *p;
452 size_t totlen;
453 int i, n;
454
455 /* Test for correct arguments. */
456 if ((resp == NULL) || (htab == NULL)) {
457 __set_errno(EINVAL);
458 return (-1);
459 }
460
461 debug("EXPORT table = %p, htab.size = %d, htab.filled = %d, size = %d\n",
462 htab, htab->size, htab->filled, size);
463 /*
464 * Pass 1:
465 * search used entries,
466 * save addresses and compute total length
467 */
468 for (i = 1, n = 0, totlen = 0; i <= htab->size; ++i) {
469
470 if (htab->table[i].used) {
471 ENTRY *ep = &htab->table[i].entry;
472
473 list[n++] = ep;
474
475 totlen += strlen(ep->key) + 2;
476
477 if (sep == '\0') {
478 totlen += strlen(ep->data);
479 } else { /* check if escapes are needed */
480 char *s = ep->data;
481
482 while (*s) {
483 ++totlen;
484 /* add room for needed escape chars */
485 if ((*s == sep) || (*s == '\\'))
486 ++totlen;
487 ++s;
488 }
489 }
490 totlen += 2; /* for '=' and 'sep' char */
491 }
492 }
493
494#ifdef DEBUG
495 /* Pass 1a: print unsorted list */
496 printf("Unsorted: n=%d\n", n);
497 for (i = 0; i < n; ++i) {
498 printf("\t%3d: %p ==> %-10s => %s\n",
499 i, list[i], list[i]->key, list[i]->data);
500 }
501#endif
502
503 /* Sort list by keys */
504 qsort(list, n, sizeof(ENTRY *), cmpkey);
505
506 /* Check if the user supplied buffer size is sufficient */
507 if (size) {
508 if (size < totlen + 1) { /* provided buffer too small */
509 debug("### buffer too small: %d, but need %d\n",
510 size, totlen + 1);
511 __set_errno(ENOMEM);
512 return (-1);
513 }
514 } else {
515 size = totlen + 1;
516 }
517
518 /* Check if the user provided a buffer */
519 if (*resp) {
520 /* yes; clear it */
521 res = *resp;
522 memset(res, '\0', size);
523 } else {
524 /* no, allocate and clear one */
525 *resp = res = calloc(1, size);
526 if (res == NULL) {
527 __set_errno(ENOMEM);
528 return (-1);
529 }
530 }
531 /*
532 * Pass 2:
533 * export sorted list of result data
534 */
535 for (i = 0, p = res; i < n; ++i) {
536 char *s;
537
538 s = list[i]->key;
539 while (*s)
540 *p++ = *s++;
541 *p++ = '=';
542
543 s = list[i]->data;
544
545 while (*s) {
546 if ((*s == sep) || (*s == '\\'))
547 *p++ = '\\'; /* escape */
548 *p++ = *s++;
549 }
550 *p++ = sep;
551 }
552 *p = '\0'; /* terminate result */
553
554 return size;
555}
556
557
558/*
559 * himport()
560 */
561
562/*
563 * Import linearized data into hash table.
564 *
565 * This is the inverse function to hexport(): it takes a linear list
566 * of "name=value" pairs and creates hash table entries from it.
567 *
568 * Entries without "value", i. e. consisting of only "name" or
569 * "name=", will cause this entry to be deleted from the hash table.
570 *
571 * The "flag" argument can be used to control the behaviour: when the
572 * H_NOCLEAR bit is set, then an existing hash table will kept, i. e.
573 * new data will be added to an existing hash table; otherwise, old
574 * data will be discarded and a new hash table will be created.
575 *
576 * The separator character for the "name=value" pairs can be selected,
577 * so we both support importing from externally stored environment
578 * data (separated by NUL characters) and from plain text files
579 * (entries separated by newline characters).
580 *
581 * To allow for nicely formatted text input, leading white space
582 * (sequences of SPACE and TAB chars) is ignored, and entries starting
583 * (after removal of any leading white space) with a '#' character are
584 * considered comments and ignored.
585 *
586 * [NOTE: this means that a variable name cannot start with a '#'
587 * character.]
588 *
589 * When using a non-NUL separator character, backslash is used as
590 * escape character in the value part, allowing for example for
591 * multi-line values.
592 *
593 * In theory, arbitrary separator characters can be used, but only
594 * '\0' and '\n' have really been tested.
595 */
596
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200597int himport_r(struct hsearch_data *htab,
598 const char *env, size_t size, const char sep, int flag)
599{
600 char *data, *sp, *dp, *name, *value;
601
602 /* Test for correct arguments. */
603 if (htab == NULL) {
604 __set_errno(EINVAL);
605 return 0;
606 }
607
608 /* we allocate new space to make sure we can write to the array */
609 if ((data = malloc(size)) == NULL) {
610 debug("himport_r: can't malloc %d bytes\n", size);
611 __set_errno(ENOMEM);
612 return 0;
613 }
614 memcpy(data, env, size);
615 dp = data;
616
617 if ((flag & H_NOCLEAR) == 0) {
618 /* Destroy old hash table if one exists */
619 debug("Destroy Hash Table: %p table = %p\n", htab,
620 htab->table);
621 if (htab->table)
622 hdestroy_r(htab);
623 }
624
625 /*
626 * Create new hash table (if needed). The computation of the hash
627 * table size is based on heuristics: in a sample of some 70+
628 * existing systems we found an average size of 39+ bytes per entry
629 * in the environment (for the whole key=value pair). Assuming a
Wolfgang Denkea882ba2010-06-20 23:33:59 +0200630 * size of 8 per entry (= safety factor of ~5) should provide enough
631 * safety margin for any existing environment definitions and still
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200632 * allow for more than enough dynamic additions. Note that the
633 * "size" argument is supposed to give the maximum enviroment size
Wolfgang Denkea882ba2010-06-20 23:33:59 +0200634 * (CONFIG_ENV_SIZE). This heuristics will result in
635 * unreasonably large numbers (and thus memory footprint) for
636 * big flash environments (>8,000 entries for 64 KB
Andreas Bießmannfc5fc762010-10-01 22:51:02 +0200637 * envrionment size), so we clip it to a reasonable value.
638 * On the other hand we need to add some more entries for free
639 * space when importing very small buffers. Both boundaries can
640 * be overwritten in the board config file if needed.
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200641 */
642
643 if (!htab->table) {
Andreas Bießmannfc5fc762010-10-01 22:51:02 +0200644 int nent = CONFIG_ENV_MIN_ENTRIES + size / 8;
Wolfgang Denkea882ba2010-06-20 23:33:59 +0200645
646 if (nent > CONFIG_ENV_MAX_ENTRIES)
647 nent = CONFIG_ENV_MAX_ENTRIES;
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200648
649 debug("Create Hash Table: N=%d\n", nent);
650
651 if (hcreate_r(nent, htab) == 0) {
652 free(data);
653 return 0;
654 }
655 }
656
657 /* Parse environment; allow for '\0' and 'sep' as separators */
658 do {
659 ENTRY e, *rv;
660
661 /* skip leading white space */
662 while ((*dp == ' ') || (*dp == '\t'))
663 ++dp;
664
665 /* skip comment lines */
666 if (*dp == '#') {
667 while (*dp && (*dp != sep))
668 ++dp;
669 ++dp;
670 continue;
671 }
672
673 /* parse name */
674 for (name = dp; *dp != '=' && *dp && *dp != sep; ++dp)
675 ;
676
677 /* deal with "name" and "name=" entries (delete var) */
678 if (*dp == '\0' || *(dp + 1) == '\0' ||
679 *dp == sep || *(dp + 1) == sep) {
680 if (*dp == '=')
681 *dp++ = '\0';
682 *dp++ = '\0'; /* terminate name */
683
684 debug("DELETE CANDIDATE: \"%s\"\n", name);
685
686 if (hdelete_r(name, htab) == 0)
687 debug("DELETE ERROR ##############################\n");
688
689 continue;
690 }
691 *dp++ = '\0'; /* terminate name */
692
693 /* parse value; deal with escapes */
694 for (value = sp = dp; *dp && (*dp != sep); ++dp) {
695 if ((*dp == '\\') && *(dp + 1))
696 ++dp;
697 *sp++ = *dp;
698 }
699 *sp++ = '\0'; /* terminate value */
700 ++dp;
701
702 /* enter into hash table */
703 e.key = name;
704 e.data = value;
705
706 hsearch_r(e, ENTER, &rv, htab);
707 if (rv == NULL) {
Wolfgang Denkea882ba2010-06-20 23:33:59 +0200708 printf("himport_r: can't insert \"%s=%s\" into hash table\n",
709 name, value);
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200710 return 0;
711 }
712
Wolfgang Denkea882ba2010-06-20 23:33:59 +0200713 debug("INSERT: table %p, filled %d/%d rv %p ==> name=\"%s\" value=\"%s\"\n",
714 htab, htab->filled, htab->size,
715 rv, name, value);
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200716 } while ((dp < data + size) && *dp); /* size check needed for text */
717 /* without '\0' termination */
Wolfgang Denkea882ba2010-06-20 23:33:59 +0200718 debug("INSERT: free(data = %p)\n", data);
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200719 free(data);
720
Wolfgang Denkea882ba2010-06-20 23:33:59 +0200721 debug("INSERT: done\n");
Wolfgang Denka6826fb2010-06-20 13:17:12 +0200722 return 1; /* everything OK */
723}