]>
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 hsearch_r(ENTRY item
, ACTION action
, ENTRY
** retval
,
206 struct hsearch_data
*htab
)
210 unsigned int len
= strlen(item
.key
);
213 /* Compute an value for the given string. Perhaps use a better method. */
216 while (count
-- > 0) {
218 hval
+= item
.key
[count
];
222 * First hash function:
223 * simply take the modul but prevent zero.
229 /* The first index tried. */
232 if (htab
->table
[idx
].used
) {
234 * Further action might be required according to the
239 if (htab
->table
[idx
].used
== hval
240 && strcmp(item
.key
, htab
->table
[idx
].entry
.key
) == 0) {
241 /* Overwrite existing value? */
242 if ((action
== ENTER
) && (item
.data
!= NULL
)) {
243 free(htab
->table
[idx
].entry
.data
);
244 htab
->table
[idx
].entry
.data
=
246 if (!htab
->table
[idx
].entry
.data
) {
252 /* return found entry */
253 *retval
= &htab
->table
[idx
].entry
;
258 * Second hash function:
259 * as suggested in [Knuth]
261 hval2
= 1 + hval
% (htab
->size
- 2);
265 * Because SIZE is prime this guarantees to
266 * step through all available indices.
269 idx
= htab
->size
+ idx
- hval2
;
274 * If we visited all entries leave the loop
280 /* If entry is found use it. */
281 if ((htab
->table
[idx
].used
== hval
)
282 && strcmp(item
.key
, htab
->table
[idx
].entry
.key
) == 0) {
283 /* Overwrite existing value? */
284 if ((action
== ENTER
) && (item
.data
!= NULL
)) {
285 free(htab
->table
[idx
].entry
.data
);
286 htab
->table
[idx
].entry
.data
=
288 if (!htab
->table
[idx
].entry
.data
) {
294 /* return found entry */
295 *retval
= &htab
->table
[idx
].entry
;
299 while (htab
->table
[idx
].used
);
302 /* An empty bucket has been found. */
303 if (action
== ENTER
) {
305 * If table is full and another entry should be
306 * entered return with error.
308 if (htab
->filled
== htab
->size
) {
316 * create copies of item.key and item.data
318 htab
->table
[idx
].used
= hval
;
319 htab
->table
[idx
].entry
.key
= strdup(item
.key
);
320 htab
->table
[idx
].entry
.data
= strdup(item
.data
);
321 if (!htab
->table
[idx
].entry
.key
||
322 !htab
->table
[idx
].entry
.data
) {
330 /* return new entry */
331 *retval
= &htab
->table
[idx
].entry
;
346 * The standard implementation of hsearch(3) does not provide any way
347 * to delete any entries from the hash table. We extend the code to
351 int hdelete_r(const char *key
, struct hsearch_data
*htab
)
356 debug("hdelete: DELETE key \"%s\"\n", key
);
360 if ((idx
= hsearch_r(e
, FIND
, &ep
, htab
)) == 0) {
362 return 0; /* not found */
365 /* free used ENTRY */
366 debug("hdelete: DELETING key \"%s\"\n", key
);
370 htab
->table
[idx
].used
= 0;
382 * Export the data stored in the hash table in linearized form.
384 * Entries are exported as "name=value" strings, separated by an
385 * arbitrary (non-NUL, of course) separator character. This allows to
386 * use this function both when formatting the U-Boot environment for
387 * external storage (using '\0' as separator), but also when using it
388 * for the "printenv" command to print all variables, simply by using
389 * as '\n" as separator. This can also be used for new features like
390 * exporting the environment data as text file, including the option
391 * for later re-import.
393 * The entries in the result list will be sorted by ascending key
396 * If the separator character is different from NUL, then any
397 * separator characters and backslash characters in the values will
398 * be escaped by a preceeding backslash in output. This is needed for
399 * example to enable multi-line values, especially when the output
400 * shall later be parsed (for example, for re-import).
402 * There are several options how the result buffer is handled:
406 * NULL 0 A string of sufficient length will be allocated.
407 * NULL >0 A string of the size given will be
408 * allocated. An error will be returned if the size is
409 * not sufficient. Any unused bytes in the string will
411 * !NULL 0 The user-supplied buffer will be used. No length
412 * checking will be performed, i. e. it is assumed that
413 * the buffer size will always be big enough. DANGEROUS.
414 * !NULL >0 The user-supplied buffer will be used. An error will
415 * be returned if the size is not sufficient. Any unused
416 * bytes in the string will be '\0'-padded.
419 static int cmpkey(const void *p1
, const void *p2
)
421 ENTRY
*e1
= *(ENTRY
**) p1
;
422 ENTRY
*e2
= *(ENTRY
**) p2
;
424 return (strcmp(e1
->key
, e2
->key
));
427 ssize_t
hexport_r(struct hsearch_data
*htab
, const char sep
,
428 char **resp
, size_t size
)
430 ENTRY
*list
[htab
->size
];
435 /* Test for correct arguments. */
436 if ((resp
== NULL
) || (htab
== NULL
)) {
441 debug("EXPORT table = %p, htab.size = %d, htab.filled = %d, size = %d\n",
442 htab
, htab
->size
, htab
->filled
, size
);
445 * search used entries,
446 * save addresses and compute total length
448 for (i
= 1, n
= 0, totlen
= 0; i
<= htab
->size
; ++i
) {
450 if (htab
->table
[i
].used
) {
451 ENTRY
*ep
= &htab
->table
[i
].entry
;
455 totlen
+= strlen(ep
->key
) + 2;
458 totlen
+= strlen(ep
->data
);
459 } else { /* check if escapes are needed */
464 /* add room for needed escape chars */
465 if ((*s
== sep
) || (*s
== '\\'))
470 totlen
+= 2; /* for '=' and 'sep' char */
475 /* Pass 1a: print unsorted list */
476 printf("Unsorted: n=%d\n", n
);
477 for (i
= 0; i
< n
; ++i
) {
478 printf("\t%3d: %p ==> %-10s => %s\n",
479 i
, list
[i
], list
[i
]->key
, list
[i
]->data
);
483 /* Sort list by keys */
484 qsort(list
, n
, sizeof(ENTRY
*), cmpkey
);
486 /* Check if the user supplied buffer size is sufficient */
488 if (size
< totlen
+ 1) { /* provided buffer too small */
489 debug("### buffer too small: %d, but need %d\n",
498 /* Check if the user provided a buffer */
502 memset(res
, '\0', size
);
504 /* no, allocate and clear one */
505 *resp
= res
= calloc(1, size
);
513 * export sorted list of result data
515 for (i
= 0, p
= res
; i
< n
; ++i
) {
526 if ((*s
== sep
) || (*s
== '\\'))
527 *p
++ = '\\'; /* escape */
532 *p
= '\0'; /* terminate result */
543 * Import linearized data into hash table.
545 * This is the inverse function to hexport(): it takes a linear list
546 * of "name=value" pairs and creates hash table entries from it.
548 * Entries without "value", i. e. consisting of only "name" or
549 * "name=", will cause this entry to be deleted from the hash table.
551 * The "flag" argument can be used to control the behaviour: when the
552 * H_NOCLEAR bit is set, then an existing hash table will kept, i. e.
553 * new data will be added to an existing hash table; otherwise, old
554 * data will be discarded and a new hash table will be created.
556 * The separator character for the "name=value" pairs can be selected,
557 * so we both support importing from externally stored environment
558 * data (separated by NUL characters) and from plain text files
559 * (entries separated by newline characters).
561 * To allow for nicely formatted text input, leading white space
562 * (sequences of SPACE and TAB chars) is ignored, and entries starting
563 * (after removal of any leading white space) with a '#' character are
564 * considered comments and ignored.
566 * [NOTE: this means that a variable name cannot start with a '#'
569 * When using a non-NUL separator character, backslash is used as
570 * escape character in the value part, allowing for example for
573 * In theory, arbitrary separator characters can be used, but only
574 * '\0' and '\n' have really been tested.
577 int himport_r(struct hsearch_data
*htab
,
578 const char *env
, size_t size
, const char sep
, int flag
)
580 char *data
, *sp
, *dp
, *name
, *value
;
582 /* Test for correct arguments. */
588 /* we allocate new space to make sure we can write to the array */
589 if ((data
= malloc(size
)) == NULL
) {
590 debug("himport_r: can't malloc %d bytes\n", size
);
594 memcpy(data
, env
, size
);
597 if ((flag
& H_NOCLEAR
) == 0) {
598 /* Destroy old hash table if one exists */
599 debug("Destroy Hash Table: %p table = %p\n", htab
,
606 * Create new hash table (if needed). The computation of the hash
607 * table size is based on heuristics: in a sample of some 70+
608 * existing systems we found an average size of 39+ bytes per entry
609 * in the environment (for the whole key=value pair). Assuming a
610 * size of 8 per entry (= safety factor of ~5) should provide enough
611 * safety margin for any existing environment definitions and still
612 * allow for more than enough dynamic additions. Note that the
613 * "size" argument is supposed to give the maximum enviroment size
614 * (CONFIG_ENV_SIZE). This heuristics will result in
615 * unreasonably large numbers (and thus memory footprint) for
616 * big flash environments (>8,000 entries for 64 KB
617 * envrionment size), so we clip it to a reasonable value.
618 * On the other hand we need to add some more entries for free
619 * space when importing very small buffers. Both boundaries can
620 * be overwritten in the board config file if needed.
624 int nent
= CONFIG_ENV_MIN_ENTRIES
+ size
/ 8;
626 if (nent
> CONFIG_ENV_MAX_ENTRIES
)
627 nent
= CONFIG_ENV_MAX_ENTRIES
;
629 debug("Create Hash Table: N=%d\n", nent
);
631 if (hcreate_r(nent
, htab
) == 0) {
637 /* Parse environment; allow for '\0' and 'sep' as separators */
641 /* skip leading white space */
642 while ((*dp
== ' ') || (*dp
== '\t'))
645 /* skip comment lines */
647 while (*dp
&& (*dp
!= sep
))
654 for (name
= dp
; *dp
!= '=' && *dp
&& *dp
!= sep
; ++dp
)
657 /* deal with "name" and "name=" entries (delete var) */
658 if (*dp
== '\0' || *(dp
+ 1) == '\0' ||
659 *dp
== sep
|| *(dp
+ 1) == sep
) {
662 *dp
++ = '\0'; /* terminate name */
664 debug("DELETE CANDIDATE: \"%s\"\n", name
);
666 if (hdelete_r(name
, htab
) == 0)
667 debug("DELETE ERROR ##############################\n");
671 *dp
++ = '\0'; /* terminate name */
673 /* parse value; deal with escapes */
674 for (value
= sp
= dp
; *dp
&& (*dp
!= sep
); ++dp
) {
675 if ((*dp
== '\\') && *(dp
+ 1))
679 *sp
++ = '\0'; /* terminate value */
682 /* enter into hash table */
686 hsearch_r(e
, ENTER
, &rv
, htab
);
688 printf("himport_r: can't insert \"%s=%s\" into hash table\n",
693 debug("INSERT: table %p, filled %d/%d rv %p ==> name=\"%s\" value=\"%s\"\n",
694 htab
, htab
->filled
, htab
->size
,
696 } while ((dp
< data
+ size
) && *dp
); /* size check needed for text */
697 /* without '\0' termination */
698 debug("INSERT: free(data = %p)\n", data
);
701 debug("INSERT: done\n");
702 return 1; /* everything OK */