1 /*
2 * Copyright © 2009, 2010 Codethink Limited
3 * Copyright © 2011 Collabora Ltd.
4 *
5 * SPDX-License-Identifier: LGPL-2.1-or-later
6 *
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
19 *
20 * Author: Ryan Lortie <desrt@desrt.ca>
21 * Stef Walter <stefw@collabora.co.uk>
22 */
23
24 #include "config.h"
25
26 #include "gbytes.h"
27
28 #include <glib/garray.h>
29 #include <glib/gstrfuncs.h>
30 #include <glib/gatomic.h>
31 #include <glib/gslice.h>
32 #include <glib/gtestutils.h>
33 #include <glib/gmem.h>
34 #include <glib/gmessages.h>
35 #include <glib/grefcount.h>
36
37 #include <string.h>
38
39 /**
40 * GBytes:
41 *
42 * A simple refcounted data type representing an immutable sequence of zero or
43 * more bytes from an unspecified origin.
44 *
45 * The purpose of a #GBytes is to keep the memory region that it holds
46 * alive for as long as anyone holds a reference to the bytes. When
47 * the last reference count is dropped, the memory is released. Multiple
48 * unrelated callers can use byte data in the #GBytes without coordinating
49 * their activities, resting assured that the byte data will not change or
50 * move while they hold a reference.
51 *
52 * A #GBytes can come from many different origins that may have
53 * different procedures for freeing the memory region. Examples are
54 * memory from g_malloc(), from memory slices, from a #GMappedFile or
55 * memory from other allocators.
56 *
57 * #GBytes work well as keys in #GHashTable. Use g_bytes_equal() and
58 * g_bytes_hash() as parameters to g_hash_table_new() or g_hash_table_new_full().
59 * #GBytes can also be used as keys in a #GTree by passing the g_bytes_compare()
60 * function to g_tree_new().
61 *
62 * The data pointed to by this bytes must not be modified. For a mutable
63 * array of bytes see #GByteArray. Use g_bytes_unref_to_array() to create a
64 * mutable array for a #GBytes sequence. To create an immutable #GBytes from
65 * a mutable #GByteArray, use the g_byte_array_free_to_bytes() function.
66 *
67 * Since: 2.32
68 **/
69
70 /* Keep in sync with glib/tests/bytes.c */
71 struct _GBytes
72 {
73 gconstpointer data; /* may be NULL iff (size == 0) */
74 gsize size; /* may be 0 */
75 gatomicrefcount ref_count;
76 GDestroyNotify free_func;
77 gpointer user_data;
78 };
79
80 /**
81 * g_bytes_new:
82 * @data: (transfer none) (array length=size) (element-type guint8) (nullable):
83 * the data to be used for the bytes
84 * @size: the size of @data
85 *
86 * Creates a new #GBytes from @data.
87 *
88 * @data is copied. If @size is 0, @data may be %NULL.
89 *
90 * Returns: (transfer full): a new #GBytes
91 *
92 * Since: 2.32
93 */
94 GBytes *
95 g_bytes_new (gconstpointer data,
96 gsize size)
97 {
98 g_return_val_if_fail (data != NULL || size == 0, NULL);
99
100 return g_bytes_new_take (g_memdup2 (data, size), size);
101 }
102
103 /**
104 * g_bytes_new_take:
105 * @data: (transfer full) (array length=size) (element-type guint8) (nullable):
106 * the data to be used for the bytes
107 * @size: the size of @data
108 *
109 * Creates a new #GBytes from @data.
110 *
111 * After this call, @data belongs to the #GBytes and may no longer be
112 * modified by the caller. The memory of @data has to be dynamically
113 * allocated and will eventually be freed with g_free().
114 *
115 * For creating #GBytes with memory from other allocators, see
116 * g_bytes_new_with_free_func().
117 *
118 * @data may be %NULL if @size is 0.
119 *
120 * Returns: (transfer full): a new #GBytes
121 *
122 * Since: 2.32
123 */
124 GBytes *
125 g_bytes_new_take (gpointer data,
126 gsize size)
127 {
128 return g_bytes_new_with_free_func (data, size, g_free, data);
129 }
130
131
132 /**
133 * g_bytes_new_static: (skip)
134 * @data: (transfer full) (array length=size) (element-type guint8) (nullable):
135 * the data to be used for the bytes
136 * @size: the size of @data
137 *
138 * Creates a new #GBytes from static data.
139 *
140 * @data must be static (ie: never modified or freed). It may be %NULL if @size
141 * is 0.
142 *
143 * Returns: (transfer full): a new #GBytes
144 *
145 * Since: 2.32
146 */
147 GBytes *
148 g_bytes_new_static (gconstpointer data,
149 gsize size)
150 {
151 return g_bytes_new_with_free_func (data, size, NULL, NULL);
152 }
153
154 /**
155 * g_bytes_new_with_free_func: (skip)
156 * @data: (array length=size) (element-type guint8) (nullable):
157 * the data to be used for the bytes
158 * @size: the size of @data
159 * @free_func: the function to call to release the data
160 * @user_data: data to pass to @free_func
161 *
162 * Creates a #GBytes from @data.
163 *
164 * When the last reference is dropped, @free_func will be called with the
165 * @user_data argument.
166 *
167 * @data must not be modified after this call is made until @free_func has
168 * been called to indicate that the bytes is no longer in use.
169 *
170 * @data may be %NULL if @size is 0.
171 *
172 * Returns: (transfer full): a new #GBytes
173 *
174 * Since: 2.32
175 */
176 GBytes *
177 g_bytes_new_with_free_func (gconstpointer data,
178 gsize size,
179 GDestroyNotify free_func,
180 gpointer user_data)
181 {
182 GBytes *bytes;
183
184 g_return_val_if_fail (data != NULL || size == 0, NULL);
185
186 bytes = g_slice_new (GBytes);
187 bytes->data = data;
188 bytes->size = size;
189 bytes->free_func = free_func;
190 bytes->user_data = user_data;
191 g_atomic_ref_count_init (&bytes->ref_count);
192
193 return (GBytes *)bytes;
194 }
195
196 /**
197 * g_bytes_new_from_bytes:
198 * @bytes: a #GBytes
199 * @offset: offset which subsection starts at
200 * @length: length of subsection
201 *
202 * Creates a #GBytes which is a subsection of another #GBytes. The @offset +
203 * @length may not be longer than the size of @bytes.
204 *
205 * A reference to @bytes will be held by the newly created #GBytes until
206 * the byte data is no longer needed.
207 *
208 * Since 2.56, if @offset is 0 and @length matches the size of @bytes, then
209 * @bytes will be returned with the reference count incremented by 1. If @bytes
210 * is a slice of another #GBytes, then the resulting #GBytes will reference
211 * the same #GBytes instead of @bytes. This allows consumers to simplify the
212 * usage of #GBytes when asynchronously writing to streams.
213 *
214 * Returns: (transfer full): a new #GBytes
215 *
216 * Since: 2.32
217 */
218 GBytes *
219 g_bytes_new_from_bytes (GBytes *bytes,
220 gsize offset,
221 gsize length)
222 {
223 gchar *base;
224
225 /* Note that length may be 0. */
226 g_return_val_if_fail (bytes != NULL, NULL);
227 g_return_val_if_fail (offset <= bytes->size, NULL);
228 g_return_val_if_fail (offset + length <= bytes->size, NULL);
229
230 /* Avoid an extra GBytes if all bytes were requested */
231 if (offset == 0 && length == bytes->size)
232 return g_bytes_ref (bytes);
233
234 base = (gchar *)bytes->data + offset;
235
236 /* Avoid referencing intermediate GBytes. In practice, this should
237 * only loop once.
238 */
239 while (bytes->free_func == (gpointer)g_bytes_unref)
240 bytes = bytes->user_data;
241
242 g_return_val_if_fail (bytes != NULL, NULL);
243 g_return_val_if_fail (base >= (gchar *)bytes->data, NULL);
244 g_return_val_if_fail (base <= (gchar *)bytes->data + bytes->size, NULL);
245 g_return_val_if_fail (base + length <= (gchar *)bytes->data + bytes->size, NULL);
246
247 return g_bytes_new_with_free_func (base, length,
248 (GDestroyNotify)g_bytes_unref, g_bytes_ref (bytes));
249 }
250
251 /**
252 * g_bytes_get_data:
253 * @bytes: a #GBytes
254 * @size: (out) (optional): location to return size of byte data
255 *
256 * Get the byte data in the #GBytes. This data should not be modified.
257 *
258 * This function will always return the same pointer for a given #GBytes.
259 *
260 * %NULL may be returned if @size is 0. This is not guaranteed, as the #GBytes
261 * may represent an empty string with @data non-%NULL and @size as 0. %NULL will
262 * not be returned if @size is non-zero.
263 *
264 * Returns: (transfer none) (array length=size) (element-type guint8) (nullable):
265 * a pointer to the byte data, or %NULL
266 *
267 * Since: 2.32
268 */
269 gconstpointer
270 g_bytes_get_data (GBytes *bytes,
271 gsize *size)
272 {
273 g_return_val_if_fail (bytes != NULL, NULL);
274 if (size)
275 *size = bytes->size;
276 return bytes->data;
277 }
278
279 /**
280 * g_bytes_get_size:
281 * @bytes: a #GBytes
282 *
283 * Get the size of the byte data in the #GBytes.
284 *
285 * This function will always return the same value for a given #GBytes.
286 *
287 * Returns: the size
288 *
289 * Since: 2.32
290 */
291 gsize
292 g_bytes_get_size (GBytes *bytes)
293 {
294 g_return_val_if_fail (bytes != NULL, 0);
295 return bytes->size;
296 }
297
298
299 /**
300 * g_bytes_ref:
301 * @bytes: a #GBytes
302 *
303 * Increase the reference count on @bytes.
304 *
305 * Returns: the #GBytes
306 *
307 * Since: 2.32
308 */
309 GBytes *
310 g_bytes_ref (GBytes *bytes)
311 {
312 g_return_val_if_fail (bytes != NULL, NULL);
313
314 g_atomic_ref_count_inc (&bytes->ref_count);
315
316 return bytes;
317 }
318
319 /**
320 * g_bytes_unref:
321 * @bytes: (nullable): a #GBytes
322 *
323 * Releases a reference on @bytes. This may result in the bytes being
324 * freed. If @bytes is %NULL, it will return immediately.
325 *
326 * Since: 2.32
327 */
328 void
329 g_bytes_unref (GBytes *bytes)
330 {
331 if (bytes == NULL)
332 return;
333
334 if (g_atomic_ref_count_dec (&bytes->ref_count))
335 {
336 if (bytes->free_func != NULL)
337 bytes->free_func (bytes->user_data);
338 g_slice_free (GBytes, bytes);
339 }
340 }
341
342 /**
343 * g_bytes_equal:
344 * @bytes1: (type GLib.Bytes): a pointer to a #GBytes
345 * @bytes2: (type GLib.Bytes): a pointer to a #GBytes to compare with @bytes1
346 *
347 * Compares the two #GBytes values being pointed to and returns
348 * %TRUE if they are equal.
349 *
350 * This function can be passed to g_hash_table_new() as the @key_equal_func
351 * parameter, when using non-%NULL #GBytes pointers as keys in a #GHashTable.
352 *
353 * Returns: %TRUE if the two keys match.
354 *
355 * Since: 2.32
356 */
357 gboolean
358 g_bytes_equal (gconstpointer bytes1,
359 gconstpointer bytes2)
360 {
361 const GBytes *b1 = bytes1;
362 const GBytes *b2 = bytes2;
363
364 g_return_val_if_fail (bytes1 != NULL, FALSE);
365 g_return_val_if_fail (bytes2 != NULL, FALSE);
366
367 return b1->size == b2->size &&
368 (b1->size == 0 || memcmp (b1->data, b2->data, b1->size) == 0);
369 }
370
371 /**
372 * g_bytes_hash:
373 * @bytes: (type GLib.Bytes): a pointer to a #GBytes key
374 *
375 * Creates an integer hash code for the byte data in the #GBytes.
376 *
377 * This function can be passed to g_hash_table_new() as the @key_hash_func
378 * parameter, when using non-%NULL #GBytes pointers as keys in a #GHashTable.
379 *
380 * Returns: a hash value corresponding to the key.
381 *
382 * Since: 2.32
383 */
384 guint
385 g_bytes_hash (gconstpointer bytes)
386 {
387 const GBytes *a = bytes;
388 const signed char *p, *e;
389 guint32 h = 5381;
390
391 g_return_val_if_fail (bytes != NULL, 0);
392
393 for (p = (signed char *)a->data, e = (signed char *)a->data + a->size; p != e; p++)
394 h = (h << 5) + h + *p;
395
396 return h;
397 }
398
399 /**
400 * g_bytes_compare:
401 * @bytes1: (type GLib.Bytes): a pointer to a #GBytes
402 * @bytes2: (type GLib.Bytes): a pointer to a #GBytes to compare with @bytes1
403 *
404 * Compares the two #GBytes values.
405 *
406 * This function can be used to sort GBytes instances in lexicographical order.
407 *
408 * If @bytes1 and @bytes2 have different length but the shorter one is a
409 * prefix of the longer one then the shorter one is considered to be less than
410 * the longer one. Otherwise the first byte where both differ is used for
411 * comparison. If @bytes1 has a smaller value at that position it is
412 * considered less, otherwise greater than @bytes2.
413 *
414 * Returns: a negative value if @bytes1 is less than @bytes2, a positive value
415 * if @bytes1 is greater than @bytes2, and zero if @bytes1 is equal to
416 * @bytes2
417 *
418 *
419 * Since: 2.32
420 */
421 gint
422 g_bytes_compare (gconstpointer bytes1,
423 gconstpointer bytes2)
424 {
425 const GBytes *b1 = bytes1;
426 const GBytes *b2 = bytes2;
427 gint ret;
428
429 g_return_val_if_fail (bytes1 != NULL, 0);
430 g_return_val_if_fail (bytes2 != NULL, 0);
431
432 ret = memcmp (b1->data, b2->data, MIN (b1->size, b2->size));
433 if (ret == 0 && b1->size != b2->size)
434 ret = b1->size < b2->size ? -1 : 1;
435 return ret;
436 }
437
438 static gpointer
439 try_steal_and_unref (GBytes *bytes,
440 GDestroyNotify free_func,
441 gsize *size)
442 {
443 gpointer result;
444
445 if (bytes->free_func != free_func || bytes->data == NULL ||
446 bytes->user_data != bytes->data)
447 return NULL;
448
449 /* Are we the only reference? */
450 if (g_atomic_ref_count_compare (&bytes->ref_count, 1))
451 {
452 *size = bytes->size;
453 result = (gpointer)bytes->data;
454 g_slice_free (GBytes, bytes);
455 return result;
456 }
457
458 return NULL;
459 }
460
461
462 /**
463 * g_bytes_unref_to_data:
464 * @bytes: (transfer full): a #GBytes
465 * @size: (out): location to place the length of the returned data
466 *
467 * Unreferences the bytes, and returns a pointer the same byte data
468 * contents.
469 *
470 * As an optimization, the byte data is returned without copying if this was
471 * the last reference to bytes and bytes was created with g_bytes_new(),
472 * g_bytes_new_take() or g_byte_array_free_to_bytes(). In all other cases the
473 * data is copied.
474 *
475 * Returns: (transfer full) (array length=size) (element-type guint8)
476 * (not nullable): a pointer to the same byte data, which should be
477 * freed with g_free()
478 *
479 * Since: 2.32
480 */
481 gpointer
482 g_bytes_unref_to_data (GBytes *bytes,
483 gsize *size)
484 {
485 gpointer result;
486
487 g_return_val_if_fail (bytes != NULL, NULL);
488 g_return_val_if_fail (size != NULL, NULL);
489
490 /*
491 * Optimal path: if this is was the last reference, then we can return
492 * the data from this GBytes without copying.
493 */
494
495 result = try_steal_and_unref (bytes, g_free, size);
496 if (result == NULL)
497 {
498 /*
499 * Copy: Non g_malloc (or compatible) allocator, or static memory,
500 * so we have to copy, and then unref.
501 */
502 result = g_memdup2 (bytes->data, bytes->size);
503 *size = bytes->size;
504 g_bytes_unref (bytes);
505 }
506
507 return result;
508 }
509
510 /**
511 * g_bytes_unref_to_array:
512 * @bytes: (transfer full): a #GBytes
513 *
514 * Unreferences the bytes, and returns a new mutable #GByteArray containing
515 * the same byte data.
516 *
517 * As an optimization, the byte data is transferred to the array without copying
518 * if this was the last reference to bytes and bytes was created with
519 * g_bytes_new(), g_bytes_new_take() or g_byte_array_free_to_bytes(). In all
520 * other cases the data is copied.
521 *
522 * Do not use it if @bytes contains more than %G_MAXUINT
523 * bytes. #GByteArray stores the length of its data in #guint, which
524 * may be shorter than #gsize, that @bytes is using.
525 *
526 * Returns: (transfer full): a new mutable #GByteArray containing the same byte data
527 *
528 * Since: 2.32
529 */
530 GByteArray *
531 g_bytes_unref_to_array (GBytes *bytes)
532 {
533 gpointer data;
534 gsize size;
535
536 g_return_val_if_fail (bytes != NULL, NULL);
537
538 data = g_bytes_unref_to_data (bytes, &size);
539 return g_byte_array_new_take (data, size);
540 }
541
542 /**
543 * g_bytes_get_region:
544 * @bytes: a #GBytes
545 * @element_size: a non-zero element size
546 * @offset: an offset to the start of the region within the @bytes
547 * @n_elements: the number of elements in the region
548 *
549 * Gets a pointer to a region in @bytes.
550 *
551 * The region starts at @offset many bytes from the start of the data
552 * and contains @n_elements many elements of @element_size size.
553 *
554 * @n_elements may be zero, but @element_size must always be non-zero.
555 * Ideally, @element_size is a static constant (eg: sizeof a struct).
556 *
557 * This function does careful bounds checking (including checking for
558 * arithmetic overflows) and returns a non-%NULL pointer if the
559 * specified region lies entirely within the @bytes. If the region is
560 * in some way out of range, or if an overflow has occurred, then %NULL
561 * is returned.
562 *
563 * Note: it is possible to have a valid zero-size region. In this case,
564 * the returned pointer will be equal to the base pointer of the data of
565 * @bytes, plus @offset. This will be non-%NULL except for the case
566 * where @bytes itself was a zero-sized region. Since it is unlikely
567 * that you will be using this function to check for a zero-sized region
568 * in a zero-sized @bytes, %NULL effectively always means "error".
569 *
570 * Returns: (nullable): the requested region, or %NULL in case of an error
571 *
572 * Since: 2.70
573 */
574 gconstpointer
575 g_bytes_get_region (GBytes *bytes,
576 gsize element_size,
577 gsize offset,
578 gsize n_elements)
579 {
580 gsize total_size;
581 gsize end_offset;
582
583 g_return_val_if_fail (element_size > 0, NULL);
584
585 /* No other assertion checks here. If something is wrong then we will
586 * simply crash (via NULL dereference or divide-by-zero).
587 */
588
589 if (!g_size_checked_mul (&total_size, element_size, n_elements))
590 return NULL;
591
592 if (!g_size_checked_add (&end_offset, offset, total_size))
593 return NULL;
594
595 /* We now have:
596 *
597 * 0 <= offset <= end_offset
598 *
599 * So we need only check that end_offset is within the range of the
600 * size of @bytes and we're good to go.
601 */
602
603 if (end_offset > bytes->size)
604 return NULL;
605
606 /* We now have:
607 *
608 * 0 <= offset <= end_offset <= bytes->size
609 */
610
611 return ((guchar *) bytes->data) + offset;
612 }