]>
git.ipfire.org Git - people/ms/u-boot.git/blob - lib/hashtable.c
2 * This implementation is based on code from uClibc-0.9.30.3 but was
3 * modified and extended for use within U-Boot.
5 * Copyright (C) 2010 Wolfgang Denk <wd@denx.de>
7 * Original license header:
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.
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.
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.
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
32 #ifdef USE_HOSTCC /* HOST build */
39 # define debug(fmt,args...) printf(fmt ,##args)
41 # define debug(fmt,args...)
44 #else /* U-Boot build */
46 # include <linux/string.h>
47 # include <linux/ctype.h>
50 #ifndef CONFIG_ENV_MIN_ENTRIES /* minimum number of entries */
51 #define CONFIG_ENV_MIN_ENTRIES 64
53 #ifndef CONFIG_ENV_MAX_ENTRIES /* maximum number of entries */
54 #define CONFIG_ENV_MAX_ENTRIES 512
60 * [Aho,Sethi,Ullman] Compilers: Principles, Techniques and Tools, 1986
61 * [Knuth] The Art of Computer Programming, part 3 (6.4)
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.
69 typedef struct _ENTRY
{
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
86 static int isprime(unsigned int number
)
88 /* no even number will be passed */
91 while (div
* div
< number
&& number
% div
!= 0)
94 return number
% div
!= 0;
98 * Before using the hash table we must allocate memory for it.
99 * Test for an existing table are done. We allocate one element
100 * more as the found prime number says. This is done for more effective
101 * indexing as explained in the comment for the hsearch function.
102 * The contents of the table is zeroed, especially the field used
106 int hcreate_r(size_t nel
, struct hsearch_data
*htab
)
108 /* Test for correct arguments. */
114 /* There is still another table active. Return with error. */
115 if (htab
->table
!= NULL
)
118 /* Change nel to the first prime number not smaller as nel. */
119 nel
|= 1; /* make odd */
120 while (!isprime(nel
))
126 /* allocate memory and zero out */
127 htab
->table
= (_ENTRY
*) calloc(htab
->size
+ 1, sizeof(_ENTRY
));
128 if (htab
->table
== NULL
)
131 /* everything went alright */
141 * After using the hash table it has to be destroyed. The used memory can
142 * be freed and the local static variable can be marked as not used.
145 void hdestroy_r(struct hsearch_data
*htab
)
149 /* Test for correct arguments. */
155 /* free used memory */
156 for (i
= 1; i
<= htab
->size
; ++i
) {
157 if (htab
->table
[i
].used
> 0) {
158 ENTRY
*ep
= &htab
->table
[i
].entry
;
160 free((void *)ep
->key
);
166 /* the sign for an existing table is an value != NULL in htable */
175 * This is the search function. It uses double hashing with open addressing.
176 * The argument item.key has to be a pointer to an zero terminated, most
177 * probably strings of chars. The function for generating a number of the
178 * strings is simple but fast. It can be replaced by a more complex function
179 * like ajw (see [Aho,Sethi,Ullman]) if the needs are shown.
181 * We use an trick to speed up the lookup. The table is created by hcreate
182 * with one more element available. This enables us to use the index zero
183 * special. This index will never be used because we store the first hash
184 * index in the field used where zero means not used. Every other value
185 * means used. The used field can be used as a first fast comparison for
186 * equality of the stored and the parameter value. This helps to prevent
187 * unnecessary expensive calls of strcmp.
189 * This implementation differs from the standard library version of
190 * this function in a number of ways:
192 * - While the standard version does not make any assumptions about
193 * the type of the stored data objects at all, this implementation
194 * works with NUL terminated strings only.
195 * - Instead of storing just pointers to the original objects, we
196 * create local copies so the caller does not need to care about the
198 * - The standard implementation does not provide a way to update an
199 * existing entry. This version will create a new entry or update an
200 * existing one when both "action == ENTER" and "item.data != NULL".
201 * - Instead of returning 1 on success, we return the index into the
202 * internal hash table, which is also guaranteed to be positive.
203 * This allows us direct access to the found hash table slot for
204 * example for functions like hdelete().
208 * hstrstr_r - return index to entry whose key and/or data contains match
210 int hstrstr_r(const char *match
, int last_idx
, ENTRY
** retval
,
211 struct hsearch_data
*htab
)
215 for (idx
= last_idx
+ 1; idx
< htab
->size
; ++idx
) {
216 if (htab
->table
[idx
].used
<= 0)
218 if (strstr(htab
->table
[idx
].entry
.key
, match
) ||
219 strstr(htab
->table
[idx
].entry
.data
, match
)) {
220 *retval
= &htab
->table
[idx
].entry
;
230 int hmatch_r(const char *match
, int last_idx
, ENTRY
** retval
,
231 struct hsearch_data
*htab
)
234 size_t key_len
= strlen(match
);
236 for (idx
= last_idx
+ 1; idx
< htab
->size
; ++idx
) {
237 if (htab
->table
[idx
].used
<= 0)
239 if (!strncmp(match
, htab
->table
[idx
].entry
.key
, key_len
)) {
240 *retval
= &htab
->table
[idx
].entry
;
250 int hsearch_r(ENTRY item
, ACTION action
, ENTRY
** retval
,
251 struct hsearch_data
*htab
)
255 unsigned int len
= strlen(item
.key
);
257 unsigned int first_deleted
= 0;
259 /* Compute an value for the given string. Perhaps use a better method. */
262 while (count
-- > 0) {
264 hval
+= item
.key
[count
];
268 * First hash function:
269 * simply take the modul but prevent zero.
275 /* The first index tried. */
278 if (htab
->table
[idx
].used
) {
280 * Further action might be required according to the
285 if (htab
->table
[idx
].used
== -1
289 if (htab
->table
[idx
].used
== hval
290 && strcmp(item
.key
, htab
->table
[idx
].entry
.key
) == 0) {
291 /* Overwrite existing value? */
292 if ((action
== ENTER
) && (item
.data
!= NULL
)) {
293 free(htab
->table
[idx
].entry
.data
);
294 htab
->table
[idx
].entry
.data
=
296 if (!htab
->table
[idx
].entry
.data
) {
302 /* return found entry */
303 *retval
= &htab
->table
[idx
].entry
;
308 * Second hash function:
309 * as suggested in [Knuth]
311 hval2
= 1 + hval
% (htab
->size
- 2);
315 * Because SIZE is prime this guarantees to
316 * step through all available indices.
319 idx
= htab
->size
+ idx
- hval2
;
324 * If we visited all entries leave the loop
330 /* If entry is found use it. */
331 if ((htab
->table
[idx
].used
== hval
)
332 && strcmp(item
.key
, htab
->table
[idx
].entry
.key
) == 0) {
333 /* Overwrite existing value? */
334 if ((action
== ENTER
) && (item
.data
!= NULL
)) {
335 free(htab
->table
[idx
].entry
.data
);
336 htab
->table
[idx
].entry
.data
=
338 if (!htab
->table
[idx
].entry
.data
) {
344 /* return found entry */
345 *retval
= &htab
->table
[idx
].entry
;
349 while (htab
->table
[idx
].used
);
352 /* An empty bucket has been found. */
353 if (action
== ENTER
) {
355 * If table is full and another entry should be
356 * entered return with error.
358 if (htab
->filled
== htab
->size
) {
366 * create copies of item.key and item.data
371 htab
->table
[idx
].used
= hval
;
372 htab
->table
[idx
].entry
.key
= strdup(item
.key
);
373 htab
->table
[idx
].entry
.data
= strdup(item
.data
);
374 if (!htab
->table
[idx
].entry
.key
||
375 !htab
->table
[idx
].entry
.data
) {
383 /* return new entry */
384 *retval
= &htab
->table
[idx
].entry
;
399 * The standard implementation of hsearch(3) does not provide any way
400 * to delete any entries from the hash table. We extend the code to
404 int hdelete_r(const char *key
, struct hsearch_data
*htab
)
409 debug("hdelete: DELETE key \"%s\"\n", key
);
413 if ((idx
= hsearch_r(e
, FIND
, &ep
, htab
)) == 0) {
415 return 0; /* not found */
418 /* free used ENTRY */
419 debug("hdelete: DELETING key \"%s\"\n", key
);
421 free((void *)ep
->key
);
423 htab
->table
[idx
].used
= -1;
435 * Export the data stored in the hash table in linearized form.
437 * Entries are exported as "name=value" strings, separated by an
438 * arbitrary (non-NUL, of course) separator character. This allows to
439 * use this function both when formatting the U-Boot environment for
440 * external storage (using '\0' as separator), but also when using it
441 * for the "printenv" command to print all variables, simply by using
442 * as '\n" as separator. This can also be used for new features like
443 * exporting the environment data as text file, including the option
444 * for later re-import.
446 * The entries in the result list will be sorted by ascending key
449 * If the separator character is different from NUL, then any
450 * separator characters and backslash characters in the values will
451 * be escaped by a preceeding backslash in output. This is needed for
452 * example to enable multi-line values, especially when the output
453 * shall later be parsed (for example, for re-import).
455 * There are several options how the result buffer is handled:
459 * NULL 0 A string of sufficient length will be allocated.
460 * NULL >0 A string of the size given will be
461 * allocated. An error will be returned if the size is
462 * not sufficient. Any unused bytes in the string will
464 * !NULL 0 The user-supplied buffer will be used. No length
465 * checking will be performed, i. e. it is assumed that
466 * the buffer size will always be big enough. DANGEROUS.
467 * !NULL >0 The user-supplied buffer will be used. An error will
468 * be returned if the size is not sufficient. Any unused
469 * bytes in the string will be '\0'-padded.
472 static int cmpkey(const void *p1
, const void *p2
)
474 ENTRY
*e1
= *(ENTRY
**) p1
;
475 ENTRY
*e2
= *(ENTRY
**) p2
;
477 return (strcmp(e1
->key
, e2
->key
));
480 ssize_t
hexport_r(struct hsearch_data
*htab
, const char sep
,
481 char **resp
, size_t size
,
482 int argc
, char * const argv
[])
484 ENTRY
*list
[htab
->size
];
489 /* Test for correct arguments. */
490 if ((resp
== NULL
) || (htab
== NULL
)) {
495 debug("EXPORT table = %p, htab.size = %d, htab.filled = %d, size = %d\n",
496 htab
, htab
->size
, htab
->filled
, size
);
499 * search used entries,
500 * save addresses and compute total length
502 for (i
= 1, n
= 0, totlen
= 0; i
<= htab
->size
; ++i
) {
504 if (htab
->table
[i
].used
> 0) {
505 ENTRY
*ep
= &htab
->table
[i
].entry
;
508 for (arg
= 0; arg
< argc
; ++arg
) {
509 if (strcmp(argv
[arg
], ep
->key
) == 0) {
514 if ((argc
> 0) && (found
== 0))
519 totlen
+= strlen(ep
->key
) + 2;
522 totlen
+= strlen(ep
->data
);
523 } else { /* check if escapes are needed */
528 /* add room for needed escape chars */
529 if ((*s
== sep
) || (*s
== '\\'))
534 totlen
+= 2; /* for '=' and 'sep' char */
539 /* Pass 1a: print unsorted list */
540 printf("Unsorted: n=%d\n", n
);
541 for (i
= 0; i
< n
; ++i
) {
542 printf("\t%3d: %p ==> %-10s => %s\n",
543 i
, list
[i
], list
[i
]->key
, list
[i
]->data
);
547 /* Sort list by keys */
548 qsort(list
, n
, sizeof(ENTRY
*), cmpkey
);
550 /* Check if the user supplied buffer size is sufficient */
552 if (size
< totlen
+ 1) { /* provided buffer too small */
553 printf("Env export buffer too small: %d, but need %d\n",
562 /* Check if the user provided a buffer */
566 memset(res
, '\0', size
);
568 /* no, allocate and clear one */
569 *resp
= res
= calloc(1, size
);
577 * export sorted list of result data
579 for (i
= 0, p
= res
; i
< n
; ++i
) {
590 if ((*s
== sep
) || (*s
== '\\'))
591 *p
++ = '\\'; /* escape */
596 *p
= '\0'; /* terminate result */
607 * Import linearized data into hash table.
609 * This is the inverse function to hexport(): it takes a linear list
610 * of "name=value" pairs and creates hash table entries from it.
612 * Entries without "value", i. e. consisting of only "name" or
613 * "name=", will cause this entry to be deleted from the hash table.
615 * The "flag" argument can be used to control the behaviour: when the
616 * H_NOCLEAR bit is set, then an existing hash table will kept, i. e.
617 * new data will be added to an existing hash table; otherwise, old
618 * data will be discarded and a new hash table will be created.
620 * The separator character for the "name=value" pairs can be selected,
621 * so we both support importing from externally stored environment
622 * data (separated by NUL characters) and from plain text files
623 * (entries separated by newline characters).
625 * To allow for nicely formatted text input, leading white space
626 * (sequences of SPACE and TAB chars) is ignored, and entries starting
627 * (after removal of any leading white space) with a '#' character are
628 * considered comments and ignored.
630 * [NOTE: this means that a variable name cannot start with a '#'
633 * When using a non-NUL separator character, backslash is used as
634 * escape character in the value part, allowing for example for
637 * In theory, arbitrary separator characters can be used, but only
638 * '\0' and '\n' have really been tested.
641 int himport_r(struct hsearch_data
*htab
,
642 const char *env
, size_t size
, const char sep
, int flag
)
644 char *data
, *sp
, *dp
, *name
, *value
;
646 /* Test for correct arguments. */
652 /* we allocate new space to make sure we can write to the array */
653 if ((data
= malloc(size
)) == NULL
) {
654 debug("himport_r: can't malloc %d bytes\n", size
);
658 memcpy(data
, env
, size
);
661 if ((flag
& H_NOCLEAR
) == 0) {
662 /* Destroy old hash table if one exists */
663 debug("Destroy Hash Table: %p table = %p\n", htab
,
670 * Create new hash table (if needed). The computation of the hash
671 * table size is based on heuristics: in a sample of some 70+
672 * existing systems we found an average size of 39+ bytes per entry
673 * in the environment (for the whole key=value pair). Assuming a
674 * size of 8 per entry (= safety factor of ~5) should provide enough
675 * safety margin for any existing environment definitions and still
676 * allow for more than enough dynamic additions. Note that the
677 * "size" argument is supposed to give the maximum enviroment size
678 * (CONFIG_ENV_SIZE). This heuristics will result in
679 * unreasonably large numbers (and thus memory footprint) for
680 * big flash environments (>8,000 entries for 64 KB
681 * envrionment size), so we clip it to a reasonable value.
682 * On the other hand we need to add some more entries for free
683 * space when importing very small buffers. Both boundaries can
684 * be overwritten in the board config file if needed.
688 int nent
= CONFIG_ENV_MIN_ENTRIES
+ size
/ 8;
690 if (nent
> CONFIG_ENV_MAX_ENTRIES
)
691 nent
= CONFIG_ENV_MAX_ENTRIES
;
693 debug("Create Hash Table: N=%d\n", nent
);
695 if (hcreate_r(nent
, htab
) == 0) {
701 /* Parse environment; allow for '\0' and 'sep' as separators */
705 /* skip leading white space */
709 /* skip comment lines */
711 while (*dp
&& (*dp
!= sep
))
718 for (name
= dp
; *dp
!= '=' && *dp
&& *dp
!= sep
; ++dp
)
721 /* deal with "name" and "name=" entries (delete var) */
722 if (*dp
== '\0' || *(dp
+ 1) == '\0' ||
723 *dp
== sep
|| *(dp
+ 1) == sep
) {
726 *dp
++ = '\0'; /* terminate name */
728 debug("DELETE CANDIDATE: \"%s\"\n", name
);
730 if (hdelete_r(name
, htab
) == 0)
731 debug("DELETE ERROR ##############################\n");
735 *dp
++ = '\0'; /* terminate name */
737 /* parse value; deal with escapes */
738 for (value
= sp
= dp
; *dp
&& (*dp
!= sep
); ++dp
) {
739 if ((*dp
== '\\') && *(dp
+ 1))
743 *sp
++ = '\0'; /* terminate value */
746 /* enter into hash table */
750 hsearch_r(e
, ENTER
, &rv
, htab
);
752 printf("himport_r: can't insert \"%s=%s\" into hash table\n",
757 debug("INSERT: table %p, filled %d/%d rv %p ==> name=\"%s\" value=\"%s\"\n",
758 htab
, htab
->filled
, htab
->size
,
760 } while ((dp
< data
+ size
) && *dp
); /* size check needed for text */
761 /* without '\0' termination */
762 debug("INSERT: free(data = %p)\n", data
);
765 debug("INSERT: done\n");
766 return 1; /* everything OK */