]>
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 */
38 # define debug(fmt,args...) printf(fmt ,##args)
40 # define debug(fmt,args...)
43 #else /* U-Boot build */
45 # include <linux/string.h>
48 #ifndef CONFIG_ENV_MIN_ENTRIES /* minimum number of entries */
49 #define CONFIG_ENV_MIN_ENTRIES 64
51 #ifndef CONFIG_ENV_MAX_ENTRIES /* maximum number of entries */
52 #define CONFIG_ENV_MAX_ENTRIES 512
58 * [Aho,Sethi,Ullman] Compilers: Principles, Techniques and Tools, 1986
59 * [Knuth] The Art of Computer Programming, part 3 (6.4)
63 * 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.
67 typedef struct _ENTRY
{
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
84 static int isprime(unsigned int number
)
86 /* no even number will be passed */
89 while (div
* div
< number
&& number
% div
!= 0)
92 return number
% div
!= 0;
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
104 int hcreate_r(size_t nel
, struct hsearch_data
*htab
)
106 /* Test for correct arguments. */
112 /* There is still another table active. Return with error. */
113 if (htab
->table
!= NULL
)
116 /* Change nel to the first prime number not smaller as nel. */
117 nel
|= 1; /* make odd */
118 while (!isprime(nel
))
124 /* allocate memory and zero out */
125 htab
->table
= (_ENTRY
*) calloc(htab
->size
+ 1, sizeof(_ENTRY
));
126 if (htab
->table
== NULL
)
129 /* everything went alright */
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.
143 void hdestroy_r(struct hsearch_data
*htab
)
147 /* Test for correct arguments. */
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
;
164 /* the sign for an existing table is an value != NULL in htable */
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.
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.
187 * This implementation differs from the standard library version of
188 * this function in a number of ways:
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
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().
205 int hmatch_r(const char *match
, int last_idx
, ENTRY
** retval
,
206 struct hsearch_data
*htab
)
209 size_t key_len
= strlen(match
);
211 for (idx
= last_idx
+ 1; idx
< htab
->size
; ++idx
) {
212 if (!htab
->table
[idx
].used
)
214 if (!strncmp(match
, htab
->table
[idx
].entry
.key
, key_len
)) {
215 *retval
= &htab
->table
[idx
].entry
;
225 int hsearch_r(ENTRY item
, ACTION action
, ENTRY
** retval
,
226 struct hsearch_data
*htab
)
230 unsigned int len
= strlen(item
.key
);
233 /* Compute an value for the given string. Perhaps use a better method. */
236 while (count
-- > 0) {
238 hval
+= item
.key
[count
];
242 * First hash function:
243 * simply take the modul but prevent zero.
249 /* The first index tried. */
252 if (htab
->table
[idx
].used
) {
254 * Further action might be required according to the
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
=
266 if (!htab
->table
[idx
].entry
.data
) {
272 /* return found entry */
273 *retval
= &htab
->table
[idx
].entry
;
278 * Second hash function:
279 * as suggested in [Knuth]
281 hval2
= 1 + hval
% (htab
->size
- 2);
285 * Because SIZE is prime this guarantees to
286 * step through all available indices.
289 idx
= htab
->size
+ idx
- hval2
;
294 * If we visited all entries leave the loop
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
=
308 if (!htab
->table
[idx
].entry
.data
) {
314 /* return found entry */
315 *retval
= &htab
->table
[idx
].entry
;
319 while (htab
->table
[idx
].used
);
322 /* An empty bucket has been found. */
323 if (action
== ENTER
) {
325 * If table is full and another entry should be
326 * entered return with error.
328 if (htab
->filled
== htab
->size
) {
336 * create copies of item.key and item.data
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
) {
350 /* return new entry */
351 *retval
= &htab
->table
[idx
].entry
;
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
371 int hdelete_r(const char *key
, struct hsearch_data
*htab
)
376 debug("hdelete: DELETE key \"%s\"\n", key
);
380 if ((idx
= hsearch_r(e
, FIND
, &ep
, htab
)) == 0) {
382 return 0; /* not found */
385 /* free used ENTRY */
386 debug("hdelete: DELETING key \"%s\"\n", key
);
390 htab
->table
[idx
].used
= 0;
402 * Export the data stored in the hash table in linearized form.
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.
413 * The entries in the result list will be sorted by ascending key
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).
422 * There are several options how the result buffer is handled:
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
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.
439 static int cmpkey(const void *p1
, const void *p2
)
441 ENTRY
*e1
= *(ENTRY
**) p1
;
442 ENTRY
*e2
= *(ENTRY
**) p2
;
444 return (strcmp(e1
->key
, e2
->key
));
447 ssize_t
hexport_r(struct hsearch_data
*htab
, const char sep
,
448 char **resp
, size_t size
)
450 ENTRY
*list
[htab
->size
];
455 /* Test for correct arguments. */
456 if ((resp
== NULL
) || (htab
== NULL
)) {
461 debug("EXPORT table = %p, htab.size = %d, htab.filled = %d, size = %d\n",
462 htab
, htab
->size
, htab
->filled
, size
);
465 * search used entries,
466 * save addresses and compute total length
468 for (i
= 1, n
= 0, totlen
= 0; i
<= htab
->size
; ++i
) {
470 if (htab
->table
[i
].used
) {
471 ENTRY
*ep
= &htab
->table
[i
].entry
;
475 totlen
+= strlen(ep
->key
) + 2;
478 totlen
+= strlen(ep
->data
);
479 } else { /* check if escapes are needed */
484 /* add room for needed escape chars */
485 if ((*s
== sep
) || (*s
== '\\'))
490 totlen
+= 2; /* for '=' and 'sep' char */
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
);
503 /* Sort list by keys */
504 qsort(list
, n
, sizeof(ENTRY
*), cmpkey
);
506 /* Check if the user supplied buffer size is sufficient */
508 if (size
< totlen
+ 1) { /* provided buffer too small */
509 debug("### buffer too small: %d, but need %d\n",
518 /* Check if the user provided a buffer */
522 memset(res
, '\0', size
);
524 /* no, allocate and clear one */
525 *resp
= res
= calloc(1, size
);
533 * export sorted list of result data
535 for (i
= 0, p
= res
; i
< n
; ++i
) {
546 if ((*s
== sep
) || (*s
== '\\'))
547 *p
++ = '\\'; /* escape */
552 *p
= '\0'; /* terminate result */
563 * Import linearized data into hash table.
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.
568 * Entries without "value", i. e. consisting of only "name" or
569 * "name=", will cause this entry to be deleted from the hash table.
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.
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).
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.
586 * [NOTE: this means that a variable name cannot start with a '#'
589 * When using a non-NUL separator character, backslash is used as
590 * escape character in the value part, allowing for example for
593 * In theory, arbitrary separator characters can be used, but only
594 * '\0' and '\n' have really been tested.
597 int himport_r(struct hsearch_data
*htab
,
598 const char *env
, size_t size
, const char sep
, int flag
)
600 char *data
, *sp
, *dp
, *name
, *value
;
602 /* Test for correct arguments. */
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
);
614 memcpy(data
, env
, size
);
617 if ((flag
& H_NOCLEAR
) == 0) {
618 /* Destroy old hash table if one exists */
619 debug("Destroy Hash Table: %p table = %p\n", htab
,
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
630 * size of 8 per entry (= safety factor of ~5) should provide enough
631 * safety margin for any existing environment definitions and still
632 * allow for more than enough dynamic additions. Note that the
633 * "size" argument is supposed to give the maximum enviroment size
634 * (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
637 * 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.
644 int nent
= CONFIG_ENV_MIN_ENTRIES
+ size
/ 8;
646 if (nent
> CONFIG_ENV_MAX_ENTRIES
)
647 nent
= CONFIG_ENV_MAX_ENTRIES
;
649 debug("Create Hash Table: N=%d\n", nent
);
651 if (hcreate_r(nent
, htab
) == 0) {
657 /* Parse environment; allow for '\0' and 'sep' as separators */
661 /* skip leading white space */
662 while ((*dp
== ' ') || (*dp
== '\t'))
665 /* skip comment lines */
667 while (*dp
&& (*dp
!= sep
))
674 for (name
= dp
; *dp
!= '=' && *dp
&& *dp
!= sep
; ++dp
)
677 /* deal with "name" and "name=" entries (delete var) */
678 if (*dp
== '\0' || *(dp
+ 1) == '\0' ||
679 *dp
== sep
|| *(dp
+ 1) == sep
) {
682 *dp
++ = '\0'; /* terminate name */
684 debug("DELETE CANDIDATE: \"%s\"\n", name
);
686 if (hdelete_r(name
, htab
) == 0)
687 debug("DELETE ERROR ##############################\n");
691 *dp
++ = '\0'; /* terminate name */
693 /* parse value; deal with escapes */
694 for (value
= sp
= dp
; *dp
&& (*dp
!= sep
); ++dp
) {
695 if ((*dp
== '\\') && *(dp
+ 1))
699 *sp
++ = '\0'; /* terminate value */
702 /* enter into hash table */
706 hsearch_r(e
, ENTER
, &rv
, htab
);
708 printf("himport_r: can't insert \"%s=%s\" into hash table\n",
713 debug("INSERT: table %p, filled %d/%d rv %p ==> name=\"%s\" value=\"%s\"\n",
714 htab
, htab
->filled
, htab
->size
,
716 } while ((dp
< data
+ size
) && *dp
); /* size check needed for text */
717 /* without '\0' termination */
718 debug("INSERT: free(data = %p)\n", data
);
721 debug("INSERT: done\n");
722 return 1; /* everything OK */