]> git.ipfire.org Git - people/ms/u-boot.git/blame - lib/hashtable.c
cmd_elf: add an option for loading ELFs according to PHDRs
[people/ms/u-boot.git] / lib / hashtable.c
CommitLineData
a6826fbc
WD
1/*
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
ea882baf
WD
48#ifndef CONFIG_ENV_MAX_ENTRIES /* maximum number of entries */
49#define CONFIG_ENV_MAX_ENTRIES 512
50#endif
51
a6826fbc
WD
52#include "search.h"
53
54/*
55 * [Aho,Sethi,Ullman] Compilers: Principles, Techniques and Tools, 1986
56 * [Knuth] The Art of Computer Programming, part 3 (6.4)
57 */
58
59/*
60 * The non-reentrant version use a global space for storing the hash table.
61 */
62static struct hsearch_data htab;
63
64/*
65 * The reentrant version has no static variables to maintain the state.
66 * Instead the interface of all functions is extended to take an argument
67 * which describes the current status.
68 */
69typedef struct _ENTRY {
70 unsigned int used;
71 ENTRY entry;
72} _ENTRY;
73
74
75/*
76 * hcreate()
77 */
78
79/*
80 * For the used double hash method the table size has to be a prime. To
81 * correct the user given table size we need a prime test. This trivial
82 * algorithm is adequate because
83 * a) the code is (most probably) called a few times per program run and
84 * b) the number is small because the table must fit in the core
85 * */
86static int isprime(unsigned int number)
87{
88 /* no even number will be passed */
89 unsigned int div = 3;
90
91 while (div * div < number && number % div != 0)
92 div += 2;
93
94 return number % div != 0;
95}
96
97int hcreate(size_t nel)
98{
99 return hcreate_r(nel, &htab);
100}
101
102/*
103 * Before using the hash table we must allocate memory for it.
104 * Test for an existing table are done. We allocate one element
105 * more as the found prime number says. This is done for more effective
106 * indexing as explained in the comment for the hsearch function.
107 * The contents of the table is zeroed, especially the field used
108 * becomes zero.
109 */
110int hcreate_r(size_t nel, struct hsearch_data *htab)
111{
112 /* Test for correct arguments. */
113 if (htab == NULL) {
114 __set_errno(EINVAL);
115 return 0;
116 }
117
118 /* There is still another table active. Return with error. */
119 if (htab->table != NULL)
120 return 0;
121
122 /* Change nel to the first prime number not smaller as nel. */
123 nel |= 1; /* make odd */
124 while (!isprime(nel))
125 nel += 2;
126
127 htab->size = nel;
128 htab->filled = 0;
129
130 /* allocate memory and zero out */
131 htab->table = (_ENTRY *) calloc(htab->size + 1, sizeof(_ENTRY));
132 if (htab->table == NULL)
133 return 0;
134
135 /* everything went alright */
136 return 1;
137}
138
139
140/*
141 * hdestroy()
142 */
143void hdestroy(void)
144{
145 hdestroy_r(&htab);
146}
147
148/*
149 * After using the hash table it has to be destroyed. The used memory can
150 * be freed and the local static variable can be marked as not used.
151 */
152void hdestroy_r(struct hsearch_data *htab)
153{
154 int i;
155
156 /* Test for correct arguments. */
157 if (htab == NULL) {
158 __set_errno(EINVAL);
159 return;
160 }
161
162 /* free used memory */
163 for (i = 1; i <= htab->size; ++i) {
164 if (htab->table[i].used) {
165 ENTRY *ep = &htab->table[i].entry;
166
167 free(ep->key);
168 free(ep->data);
169 }
170 }
171 free(htab->table);
172
173 /* the sign for an existing table is an value != NULL in htable */
174 htab->table = NULL;
175}
176
177/*
178 * hsearch()
179 */
180
181/*
182 * This is the search function. It uses double hashing with open addressing.
183 * The argument item.key has to be a pointer to an zero terminated, most
184 * probably strings of chars. The function for generating a number of the
185 * strings is simple but fast. It can be replaced by a more complex function
186 * like ajw (see [Aho,Sethi,Ullman]) if the needs are shown.
187 *
188 * We use an trick to speed up the lookup. The table is created by hcreate
189 * with one more element available. This enables us to use the index zero
190 * special. This index will never be used because we store the first hash
191 * index in the field used where zero means not used. Every other value
192 * means used. The used field can be used as a first fast comparison for
193 * equality of the stored and the parameter value. This helps to prevent
194 * unnecessary expensive calls of strcmp.
195 *
196 * This implementation differs from the standard library version of
197 * this function in a number of ways:
198 *
199 * - While the standard version does not make any assumptions about
200 * the type of the stored data objects at all, this implementation
201 * works with NUL terminated strings only.
202 * - Instead of storing just pointers to the original objects, we
203 * create local copies so the caller does not need to care about the
204 * data any more.
205 * - The standard implementation does not provide a way to update an
206 * existing entry. This version will create a new entry or update an
207 * existing one when both "action == ENTER" and "item.data != NULL".
208 * - Instead of returning 1 on success, we return the index into the
209 * internal hash table, which is also guaranteed to be positive.
210 * This allows us direct access to the found hash table slot for
211 * example for functions like hdelete().
212 */
213
214ENTRY *hsearch(ENTRY item, ACTION action)
215{
216 ENTRY *result;
217
218 (void) hsearch_r(item, action, &result, &htab);
219
220 return result;
221}
222
223int hsearch_r(ENTRY item, ACTION action, ENTRY ** retval,
224 struct hsearch_data *htab)
225{
226 unsigned int hval;
227 unsigned int count;
228 unsigned int len = strlen(item.key);
229 unsigned int idx;
230
231 /* Compute an value for the given string. Perhaps use a better method. */
232 hval = len;
233 count = len;
234 while (count-- > 0) {
235 hval <<= 4;
236 hval += item.key[count];
237 }
238
239 /*
240 * First hash function:
241 * simply take the modul but prevent zero.
242 */
243 hval %= htab->size;
244 if (hval == 0)
245 ++hval;
246
247 /* The first index tried. */
248 idx = hval;
249
250 if (htab->table[idx].used) {
251 /*
252 * Further action might be required according to the
253 * action value.
254 */
255 unsigned hval2;
256
257 if (htab->table[idx].used == hval
258 && strcmp(item.key, htab->table[idx].entry.key) == 0) {
259 /* Overwrite existing value? */
260 if ((action == ENTER) && (item.data != NULL)) {
261 free(htab->table[idx].entry.data);
262 htab->table[idx].entry.data =
263 strdup(item.data);
264 if (!htab->table[idx].entry.data) {
265 __set_errno(ENOMEM);
266 *retval = NULL;
267 return 0;
268 }
269 }
270 /* return found entry */
271 *retval = &htab->table[idx].entry;
272 return idx;
273 }
274
275 /*
276 * Second hash function:
277 * as suggested in [Knuth]
278 */
279 hval2 = 1 + hval % (htab->size - 2);
280
281 do {
282 /*
283 * Because SIZE is prime this guarantees to
284 * step through all available indices.
285 */
286 if (idx <= hval2)
287 idx = htab->size + idx - hval2;
288 else
289 idx -= hval2;
290
291 /*
292 * If we visited all entries leave the loop
293 * unsuccessfully.
294 */
295 if (idx == hval)
296 break;
297
298 /* If entry is found use it. */
299 if ((htab->table[idx].used == hval)
300 && strcmp(item.key, htab->table[idx].entry.key) == 0) {
301 /* Overwrite existing value? */
302 if ((action == ENTER) && (item.data != NULL)) {
303 free(htab->table[idx].entry.data);
304 htab->table[idx].entry.data =
305 strdup(item.data);
306 if (!htab->table[idx].entry.data) {
307 __set_errno(ENOMEM);
308 *retval = NULL;
309 return 0;
310 }
311 }
312 /* return found entry */
313 *retval = &htab->table[idx].entry;
314 return idx;
315 }
316 }
317 while (htab->table[idx].used);
318 }
319
320 /* An empty bucket has been found. */
321 if (action == ENTER) {
322 /*
323 * If table is full and another entry should be
324 * entered return with error.
325 */
326 if (htab->filled == htab->size) {
327 __set_errno(ENOMEM);
328 *retval = NULL;
329 return 0;
330 }
331
332 /*
333 * Create new entry;
334 * create copies of item.key and item.data
335 */
336 htab->table[idx].used = hval;
337 htab->table[idx].entry.key = strdup(item.key);
338 htab->table[idx].entry.data = strdup(item.data);
339 if (!htab->table[idx].entry.key ||
340 !htab->table[idx].entry.data) {
341 __set_errno(ENOMEM);
342 *retval = NULL;
343 return 0;
344 }
345
346 ++htab->filled;
347
348 /* return new entry */
349 *retval = &htab->table[idx].entry;
350 return 1;
351 }
352
353 __set_errno(ESRCH);
354 *retval = NULL;
355 return 0;
356}
357
358
359/*
360 * hdelete()
361 */
362
363/*
364 * The standard implementation of hsearch(3) does not provide any way
365 * to delete any entries from the hash table. We extend the code to
366 * do that.
367 */
368
369int hdelete(const char *key)
370{
371 return hdelete_r(key, &htab);
372}
373
374int hdelete_r(const char *key, struct hsearch_data *htab)
375{
376 ENTRY e, *ep;
377 int idx;
378
379 debug("hdelete: DELETE key \"%s\"\n", key);
380
381 e.key = (char *)key;
382
383 if ((idx = hsearch_r(e, FIND, &ep, htab)) == 0) {
384 __set_errno(ESRCH);
385 return 0; /* not found */
386 }
387
388 /* free used ENTRY */
389 debug("hdelete: DELETING key \"%s\"\n", key);
390
391 free(ep->key);
392 free(ep->data);
393 htab->table[idx].used = 0;
394
395 --htab->filled;
396
397 return 1;
398}
399
400/*
401 * hexport()
402 */
403
404/*
405 * Export the data stored in the hash table in linearized form.
406 *
407 * Entries are exported as "name=value" strings, separated by an
408 * arbitrary (non-NUL, of course) separator character. This allows to
409 * use this function both when formatting the U-Boot environment for
410 * external storage (using '\0' as separator), but also when using it
411 * for the "printenv" command to print all variables, simply by using
412 * as '\n" as separator. This can also be used for new features like
413 * exporting the environment data as text file, including the option
414 * for later re-import.
415 *
416 * The entries in the result list will be sorted by ascending key
417 * values.
418 *
419 * If the separator character is different from NUL, then any
420 * separator characters and backslash characters in the values will
421 * be escaped by a preceeding backslash in output. This is needed for
422 * example to enable multi-line values, especially when the output
423 * shall later be parsed (for example, for re-import).
424 *
425 * There are several options how the result buffer is handled:
426 *
427 * *resp size
428 * -----------
429 * NULL 0 A string of sufficient length will be allocated.
430 * NULL >0 A string of the size given will be
431 * allocated. An error will be returned if the size is
432 * not sufficient. Any unused bytes in the string will
433 * be '\0'-padded.
434 * !NULL 0 The user-supplied buffer will be used. No length
435 * checking will be performed, i. e. it is assumed that
436 * the buffer size will always be big enough. DANGEROUS.
437 * !NULL >0 The user-supplied buffer will be used. An error will
438 * be returned if the size is not sufficient. Any unused
439 * bytes in the string will be '\0'-padded.
440 */
441
442ssize_t hexport(const char sep, char **resp, size_t size)
443{
444 return hexport_r(&htab, sep, resp, size);
445}
446
447static int cmpkey(const void *p1, const void *p2)
448{
449 ENTRY *e1 = *(ENTRY **) p1;
450 ENTRY *e2 = *(ENTRY **) p2;
451
452 return (strcmp(e1->key, e2->key));
453}
454
455ssize_t hexport_r(struct hsearch_data *htab, const char sep,
456 char **resp, size_t size)
457{
458 ENTRY *list[htab->size];
459 char *res, *p;
460 size_t totlen;
461 int i, n;
462
463 /* Test for correct arguments. */
464 if ((resp == NULL) || (htab == NULL)) {
465 __set_errno(EINVAL);
466 return (-1);
467 }
468
469 debug("EXPORT table = %p, htab.size = %d, htab.filled = %d, size = %d\n",
470 htab, htab->size, htab->filled, size);
471 /*
472 * Pass 1:
473 * search used entries,
474 * save addresses and compute total length
475 */
476 for (i = 1, n = 0, totlen = 0; i <= htab->size; ++i) {
477
478 if (htab->table[i].used) {
479 ENTRY *ep = &htab->table[i].entry;
480
481 list[n++] = ep;
482
483 totlen += strlen(ep->key) + 2;
484
485 if (sep == '\0') {
486 totlen += strlen(ep->data);
487 } else { /* check if escapes are needed */
488 char *s = ep->data;
489
490 while (*s) {
491 ++totlen;
492 /* add room for needed escape chars */
493 if ((*s == sep) || (*s == '\\'))
494 ++totlen;
495 ++s;
496 }
497 }
498 totlen += 2; /* for '=' and 'sep' char */
499 }
500 }
501
502#ifdef DEBUG
503 /* Pass 1a: print unsorted list */
504 printf("Unsorted: n=%d\n", n);
505 for (i = 0; i < n; ++i) {
506 printf("\t%3d: %p ==> %-10s => %s\n",
507 i, list[i], list[i]->key, list[i]->data);
508 }
509#endif
510
511 /* Sort list by keys */
512 qsort(list, n, sizeof(ENTRY *), cmpkey);
513
514 /* Check if the user supplied buffer size is sufficient */
515 if (size) {
516 if (size < totlen + 1) { /* provided buffer too small */
517 debug("### buffer too small: %d, but need %d\n",
518 size, totlen + 1);
519 __set_errno(ENOMEM);
520 return (-1);
521 }
522 } else {
523 size = totlen + 1;
524 }
525
526 /* Check if the user provided a buffer */
527 if (*resp) {
528 /* yes; clear it */
529 res = *resp;
530 memset(res, '\0', size);
531 } else {
532 /* no, allocate and clear one */
533 *resp = res = calloc(1, size);
534 if (res == NULL) {
535 __set_errno(ENOMEM);
536 return (-1);
537 }
538 }
539 /*
540 * Pass 2:
541 * export sorted list of result data
542 */
543 for (i = 0, p = res; i < n; ++i) {
544 char *s;
545
546 s = list[i]->key;
547 while (*s)
548 *p++ = *s++;
549 *p++ = '=';
550
551 s = list[i]->data;
552
553 while (*s) {
554 if ((*s == sep) || (*s == '\\'))
555 *p++ = '\\'; /* escape */
556 *p++ = *s++;
557 }
558 *p++ = sep;
559 }
560 *p = '\0'; /* terminate result */
561
562 return size;
563}
564
565
566/*
567 * himport()
568 */
569
570/*
571 * Import linearized data into hash table.
572 *
573 * This is the inverse function to hexport(): it takes a linear list
574 * of "name=value" pairs and creates hash table entries from it.
575 *
576 * Entries without "value", i. e. consisting of only "name" or
577 * "name=", will cause this entry to be deleted from the hash table.
578 *
579 * The "flag" argument can be used to control the behaviour: when the
580 * H_NOCLEAR bit is set, then an existing hash table will kept, i. e.
581 * new data will be added to an existing hash table; otherwise, old
582 * data will be discarded and a new hash table will be created.
583 *
584 * The separator character for the "name=value" pairs can be selected,
585 * so we both support importing from externally stored environment
586 * data (separated by NUL characters) and from plain text files
587 * (entries separated by newline characters).
588 *
589 * To allow for nicely formatted text input, leading white space
590 * (sequences of SPACE and TAB chars) is ignored, and entries starting
591 * (after removal of any leading white space) with a '#' character are
592 * considered comments and ignored.
593 *
594 * [NOTE: this means that a variable name cannot start with a '#'
595 * character.]
596 *
597 * When using a non-NUL separator character, backslash is used as
598 * escape character in the value part, allowing for example for
599 * multi-line values.
600 *
601 * In theory, arbitrary separator characters can be used, but only
602 * '\0' and '\n' have really been tested.
603 */
604
605int himport(const char *env, size_t size, const char sep, int flag)
606{
607 return himport_r(&htab, env, size, sep, flag);
608}
609
610int himport_r(struct hsearch_data *htab,
611 const char *env, size_t size, const char sep, int flag)
612{
613 char *data, *sp, *dp, *name, *value;
614
615 /* Test for correct arguments. */
616 if (htab == NULL) {
617 __set_errno(EINVAL);
618 return 0;
619 }
620
621 /* we allocate new space to make sure we can write to the array */
622 if ((data = malloc(size)) == NULL) {
623 debug("himport_r: can't malloc %d bytes\n", size);
624 __set_errno(ENOMEM);
625 return 0;
626 }
627 memcpy(data, env, size);
628 dp = data;
629
630 if ((flag & H_NOCLEAR) == 0) {
631 /* Destroy old hash table if one exists */
632 debug("Destroy Hash Table: %p table = %p\n", htab,
633 htab->table);
634 if (htab->table)
635 hdestroy_r(htab);
636 }
637
638 /*
639 * Create new hash table (if needed). The computation of the hash
640 * table size is based on heuristics: in a sample of some 70+
641 * existing systems we found an average size of 39+ bytes per entry
642 * in the environment (for the whole key=value pair). Assuming a
ea882baf
WD
643 * size of 8 per entry (= safety factor of ~5) should provide enough
644 * safety margin for any existing environment definitions and still
a6826fbc
WD
645 * allow for more than enough dynamic additions. Note that the
646 * "size" argument is supposed to give the maximum enviroment size
ea882baf
WD
647 * (CONFIG_ENV_SIZE). This heuristics will result in
648 * unreasonably large numbers (and thus memory footprint) for
649 * big flash environments (>8,000 entries for 64 KB
650 * envrionment size), so we clip it to a reasonable value
651 * (which can be overwritten in the board config file if
652 * needed).
a6826fbc
WD
653 */
654
655 if (!htab->table) {
ea882baf
WD
656 int nent = size / 8;
657
658 if (nent > CONFIG_ENV_MAX_ENTRIES)
659 nent = CONFIG_ENV_MAX_ENTRIES;
a6826fbc
WD
660
661 debug("Create Hash Table: N=%d\n", nent);
662
663 if (hcreate_r(nent, htab) == 0) {
664 free(data);
665 return 0;
666 }
667 }
668
669 /* Parse environment; allow for '\0' and 'sep' as separators */
670 do {
671 ENTRY e, *rv;
672
673 /* skip leading white space */
674 while ((*dp == ' ') || (*dp == '\t'))
675 ++dp;
676
677 /* skip comment lines */
678 if (*dp == '#') {
679 while (*dp && (*dp != sep))
680 ++dp;
681 ++dp;
682 continue;
683 }
684
685 /* parse name */
686 for (name = dp; *dp != '=' && *dp && *dp != sep; ++dp)
687 ;
688
689 /* deal with "name" and "name=" entries (delete var) */
690 if (*dp == '\0' || *(dp + 1) == '\0' ||
691 *dp == sep || *(dp + 1) == sep) {
692 if (*dp == '=')
693 *dp++ = '\0';
694 *dp++ = '\0'; /* terminate name */
695
696 debug("DELETE CANDIDATE: \"%s\"\n", name);
697
698 if (hdelete_r(name, htab) == 0)
699 debug("DELETE ERROR ##############################\n");
700
701 continue;
702 }
703 *dp++ = '\0'; /* terminate name */
704
705 /* parse value; deal with escapes */
706 for (value = sp = dp; *dp && (*dp != sep); ++dp) {
707 if ((*dp == '\\') && *(dp + 1))
708 ++dp;
709 *sp++ = *dp;
710 }
711 *sp++ = '\0'; /* terminate value */
712 ++dp;
713
714 /* enter into hash table */
715 e.key = name;
716 e.data = value;
717
718 hsearch_r(e, ENTER, &rv, htab);
719 if (rv == NULL) {
ea882baf
WD
720 printf("himport_r: can't insert \"%s=%s\" into hash table\n",
721 name, value);
a6826fbc
WD
722 return 0;
723 }
724
ea882baf
WD
725 debug("INSERT: table %p, filled %d/%d rv %p ==> name=\"%s\" value=\"%s\"\n",
726 htab, htab->filled, htab->size,
727 rv, name, value);
a6826fbc
WD
728 } while ((dp < data + size) && *dp); /* size check needed for text */
729 /* without '\0' termination */
ea882baf 730 debug("INSERT: free(data = %p)\n", data);
a6826fbc
WD
731 free(data);
732
ea882baf 733 debug("INSERT: done\n");
a6826fbc
WD
734 return 1; /* everything OK */
735}