4 * Code for reading streams and resources, including compressed WIM resources.
8 * Copyright (C) 2012, 2013 Eric Biggers
10 * This file is part of wimlib, a library for working with WIM files.
12 * wimlib is free software; you can redistribute it and/or modify it under the
13 * terms of the GNU General Public License as published by the Free Software
14 * Foundation; either version 3 of the License, or (at your option) any later
17 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
18 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
19 * A PARTICULAR PURPOSE. See the GNU General Public License for more details.
21 * You should have received a copy of the GNU General Public License along with
22 * wimlib; if not, see http://www.gnu.org/licenses/.
30 #include "wimlib/endianness.h"
31 #include "wimlib/error.h"
32 #include "wimlib/file_io.h"
33 #include "wimlib/lookup_table.h"
34 #include "wimlib/resource.h"
35 #include "wimlib/sha1.h"
38 /* for read_win32_file_prefix(), read_win32_encrypted_file_prefix() */
39 # include "wimlib/win32.h"
43 /* for read_ntfs_file_prefix() */
44 # include "wimlib/ntfs_3g.h"
56 * Compressed WIM resources
58 * A compressed resource in a WIM consists of a number of compressed chunks,
59 * each of which decompresses to a fixed chunk size (given in the WIM header;
60 * usually 32768) except possibly the last, which always decompresses to any
61 * remaining bytes. In addition, immediately before the chunks, a table (the
62 * "chunk table") provides the offset, in bytes relative to the end of the chunk
63 * table, of the start of each compressed chunk, except for the first chunk
64 * which is omitted as it always has an offset of 0. Therefore, a compressed
65 * resource with N chunks will have a chunk table with N - 1 entries.
67 * Additional information:
69 * - Entries in the chunk table are 4 bytes each, except if the uncompressed
70 * size of the resource is greater than 4 GiB, in which case the entries in
71 * the chunk table are 8 bytes each. In either case, the entries are unsigned
72 * little-endian integers.
74 * - The chunk table is included in the compressed size of the resource provided
75 * in the corresponding entry in the WIM's stream lookup table.
77 * - The compressed size of a chunk is never greater than the uncompressed size.
78 * From the compressor's point of view, chunks that would have compressed to a
79 * size greater than or equal to their original size are in fact stored
80 * uncompressed. From the decompresser's point of view, chunks with
81 * compressed size equal to their uncompressed size are in fact uncompressed.
83 * Furthermore, wimlib supports its own "pipable" WIM format, and for this the
84 * structure of compressed resources was modified to allow piped reading and
85 * writing. To make sequential writing possible, the chunk table is placed
86 * after the chunks rather than before the chunks, and to make sequential
87 * reading possible, each chunk is prefixed with a 4-byte header giving its
88 * compressed size as a 32-bit, unsigned, little-endian integer. Otherwise the
89 * details are the same.
99 * read_compressed_wim_resource() -
101 * Read data from a compressed WIM resource.
104 * Specification of the compressed WIM resource to read from.
106 * Nonoverlapping, nonempty ranges of the uncompressed resource data to
107 * read, sorted by increasing offset.
109 * Number of ranges in @ranges; must be at least 1.
111 * Callback function to feed the data being read. Each call provides the
112 * next chunk of the requested data, uncompressed. Each chunk will be of
113 * nonzero size and will not cross range boundaries, but otherwise will be
114 * of unspecified size.
116 * Parameter to pass to @cb_ctx.
118 * Possible return values:
120 * WIMLIB_ERR_SUCCESS (0)
121 * WIMLIB_ERR_READ (errno set)
122 * WIMLIB_ERR_UNEXPECTED_END_OF_FILE (errno set to 0)
123 * WIMLIB_ERR_NOMEM (errno set to ENOMEM)
124 * WIMLIB_ERR_DECOMPRESSION (errno set to EINVAL)
126 * or other error code returned by the @cb function.
129 read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
130 const struct data_range * const ranges,
131 const size_t num_ranges,
132 const consume_data_callback_t cb,
138 u64 *chunk_offsets = NULL;
141 bool chunk_offsets_malloced = false;
142 bool ubuf_malloced = false;
143 bool cbuf_malloced = false;
144 struct wimlib_decompressor *decompressor = NULL;
147 wimlib_assert(rspec != NULL);
148 wimlib_assert(resource_is_compressed(rspec));
149 wimlib_assert(cb != NULL);
150 wimlib_assert(num_ranges != 0);
151 for (size_t i = 0; i < num_ranges; i++) {
152 DEBUG("Range %zu/%zu: %"PRIu64"@+%"PRIu64" / %"PRIu64,
153 i + 1, num_ranges, ranges[i].size, ranges[i].offset,
154 rspec->uncompressed_size);
155 wimlib_assert(ranges[i].size != 0);
156 wimlib_assert(ranges[i].offset + ranges[i].size >= ranges[i].size);
157 wimlib_assert(ranges[i].offset + ranges[i].size <= rspec->uncompressed_size);
159 for (size_t i = 0; i < num_ranges - 1; i++)
160 wimlib_assert(ranges[i].offset + ranges[i].size <= ranges[i + 1].offset);
162 /* Get the offsets of the first and last bytes of the read. */
163 const u64 first_offset = ranges[0].offset;
164 const u64 last_offset = ranges[num_ranges - 1].offset + ranges[num_ranges - 1].size - 1;
166 /* Get the file descriptor for the WIM. */
167 struct filedes * const in_fd = &rspec->wim->in_fd;
169 /* Determine if we're reading a pipable resource from a pipe or not. */
170 const bool is_pipe_read = !filedes_is_seekable(in_fd);
172 /* Determine if the chunk table is in an altenate format. */
173 const bool alt_chunk_table = (rspec->flags & WIM_RESHDR_FLAG_PACKED_STREAMS)
176 /* Get the maximum size of uncompressed chunks in this resource, which
177 * we require be a power of 2. */
178 u64 cur_read_offset = rspec->offset_in_wim;
179 int ctype = rspec->compression_type;
180 u32 chunk_size = rspec->chunk_size;
181 if (alt_chunk_table) {
182 /* Alternate chunk table format. Its header specifies the chunk
183 * size and compression format. Note: it could be read here;
184 * however, the relevant data was already loaded into @rspec by
185 * read_wim_lookup_table(). */
186 cur_read_offset += sizeof(struct alt_chunk_table_header_disk);
189 if (!is_power_of_2(chunk_size)) {
190 ERROR("Invalid compressed resource: "
191 "expected power-of-2 chunk size (got %"PRIu32")",
193 ret = WIMLIB_ERR_INVALID_CHUNK_SIZE;
194 goto out_free_memory;
197 /* Get valid decompressor. */
198 if (ctype == rspec->wim->decompressor_ctype &&
199 chunk_size == rspec->wim->decompressor_max_block_size)
201 /* Cached decompressor. */
202 decompressor = rspec->wim->decompressor;
203 rspec->wim->decompressor_ctype = WIMLIB_COMPRESSION_TYPE_NONE;
204 rspec->wim->decompressor = NULL;
206 ret = wimlib_create_decompressor(ctype, chunk_size, NULL,
209 goto out_free_memory;
212 const u32 chunk_order = bsr32(chunk_size);
214 /* Calculate the total number of chunks the resource is divided into. */
215 const u64 num_chunks = (rspec->uncompressed_size + chunk_size - 1) >> chunk_order;
217 /* Calculate the 0-based indices of the first and last chunks containing
218 * data that needs to be passed to the callback. */
219 const u64 first_needed_chunk = first_offset >> chunk_order;
220 const u64 last_needed_chunk = last_offset >> chunk_order;
222 /* Calculate the 0-based index of the first chunk that actually needs to
223 * be read. This is normally first_needed_chunk, but for pipe reads we
224 * must always start from the 0th chunk. */
225 const u64 read_start_chunk = (is_pipe_read ? 0 : first_needed_chunk);
227 /* Calculate the number of chunk offsets that are needed for the chunks
229 const u64 num_needed_chunk_offsets =
230 last_needed_chunk - read_start_chunk + 1 +
231 (last_needed_chunk < num_chunks - 1);
233 /* Calculate the number of entries in the chunk table. Normally, it's
234 * one less than the number of chunks, since the first chunk has no
235 * entry. But in the alternate chunk table format, the chunk entries
236 * contain chunk sizes, not offsets, and there is one per chunk. */
237 const u64 num_chunk_entries = (alt_chunk_table ? num_chunks : num_chunks - 1);
239 /* Set the size of each chunk table entry based on the resource's
240 * uncompressed size. */
241 const u64 chunk_entry_size = get_chunk_entry_size(rspec->uncompressed_size,
244 /* Calculate the size of the chunk table in bytes. */
245 const u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
247 /* Calculate the size of the chunk table in bytes, including the header
248 * in the case of the alternate chunk table format. */
249 const u64 chunk_table_full_size =
250 (alt_chunk_table) ? chunk_table_size + sizeof(struct alt_chunk_table_header_disk)
254 /* Read the needed chunk table entries into memory and use them
255 * to initialize the chunk_offsets array. */
257 u64 first_chunk_entry_to_read;
258 u64 last_chunk_entry_to_read;
260 if (alt_chunk_table) {
261 /* The alternate chunk table contains chunk sizes, not
262 * offsets, so we always must read all preceding entries
263 * in order to determine offsets. */
264 first_chunk_entry_to_read = 0;
265 last_chunk_entry_to_read = last_needed_chunk;
267 /* Here we must account for the fact that the first
268 * chunk has no explicit chunk table entry. */
270 if (read_start_chunk == 0)
271 first_chunk_entry_to_read = 0;
273 first_chunk_entry_to_read = read_start_chunk - 1;
275 if (last_needed_chunk == 0)
276 last_chunk_entry_to_read = 0;
278 last_chunk_entry_to_read = last_needed_chunk - 1;
280 if (last_needed_chunk < num_chunks - 1)
281 last_chunk_entry_to_read++;
284 const u64 num_chunk_entries_to_read =
285 last_chunk_entry_to_read - first_chunk_entry_to_read + 1;
287 const u64 chunk_offsets_alloc_size =
288 max(num_chunk_entries_to_read,
289 num_needed_chunk_offsets) * sizeof(chunk_offsets[0]);
291 if ((size_t)chunk_offsets_alloc_size != chunk_offsets_alloc_size)
294 if (chunk_offsets_alloc_size <= STACK_MAX) {
295 chunk_offsets = alloca(chunk_offsets_alloc_size);
297 chunk_offsets = MALLOC(chunk_offsets_alloc_size);
298 if (chunk_offsets == NULL)
300 chunk_offsets_malloced = true;
303 const size_t chunk_table_size_to_read =
304 num_chunk_entries_to_read * chunk_entry_size;
306 const u64 file_offset_of_needed_chunk_entries =
308 + (first_chunk_entry_to_read * chunk_entry_size)
309 + (rspec->is_pipable ? (rspec->size_in_wim - chunk_table_size) : 0);
311 void * const chunk_table_data =
313 chunk_offsets_alloc_size -
314 chunk_table_size_to_read;
316 ret = full_pread(in_fd, chunk_table_data, chunk_table_size_to_read,
317 file_offset_of_needed_chunk_entries);
321 /* Now fill in chunk_offsets from the entries we have read in
322 * chunk_tab_data. We break aliasing rules here to avoid having
323 * to allocate yet another array. */
324 typedef le64 __attribute__((may_alias)) aliased_le64_t;
325 typedef le32 __attribute__((may_alias)) aliased_le32_t;
326 u64 * chunk_offsets_p = chunk_offsets;
328 if (alt_chunk_table) {
330 aliased_le32_t *raw_entries = chunk_table_data;
332 for (size_t i = 0; i < num_chunk_entries_to_read; i++) {
333 u32 entry = le32_to_cpu(raw_entries[i]);
334 if (i >= read_start_chunk)
335 *chunk_offsets_p++ = cur_offset;
338 if (last_needed_chunk < num_chunks - 1)
339 *chunk_offsets_p = cur_offset;
341 if (read_start_chunk == 0)
342 *chunk_offsets_p++ = 0;
344 if (chunk_entry_size == 4) {
345 aliased_le32_t *raw_entries = chunk_table_data;
346 for (size_t i = 0; i < num_chunk_entries_to_read; i++)
347 *chunk_offsets_p++ = le32_to_cpu(raw_entries[i]);
349 aliased_le64_t *raw_entries = chunk_table_data;
350 for (size_t i = 0; i < num_chunk_entries_to_read; i++)
351 *chunk_offsets_p++ = le64_to_cpu(raw_entries[i]);
355 /* Set offset to beginning of first chunk to read. */
356 cur_read_offset += chunk_offsets[0];
357 if (rspec->is_pipable)
358 cur_read_offset += read_start_chunk * sizeof(struct pwm_chunk_hdr);
360 cur_read_offset += chunk_table_size;
363 /* Allocate buffer for holding the uncompressed data of each chunk. */
364 if (chunk_size <= STACK_MAX) {
365 ubuf = alloca(chunk_size);
367 ubuf = MALLOC(chunk_size);
370 ubuf_malloced = true;
373 /* Allocate a temporary buffer for reading compressed chunks, each of
374 * which can be at most @chunk_size - 1 bytes. This excludes compressed
375 * chunks that are a full @chunk_size bytes, which are actually stored
377 if (chunk_size - 1 <= STACK_MAX) {
378 cbuf = alloca(chunk_size - 1);
380 cbuf = MALLOC(chunk_size - 1);
383 cbuf_malloced = true;
386 /* Set current data range. */
387 const struct data_range *cur_range = ranges;
388 const struct data_range * const end_range = &ranges[num_ranges];
389 u64 cur_range_pos = cur_range->offset;
390 u64 cur_range_end = cur_range->offset + cur_range->size;
392 /* Read and process each needed chunk. */
393 for (u64 i = read_start_chunk; i <= last_needed_chunk; i++) {
395 /* Calculate uncompressed size of next chunk. */
397 if ((i == num_chunks - 1) && (rspec->uncompressed_size & (chunk_size - 1)))
398 chunk_usize = (rspec->uncompressed_size & (chunk_size - 1));
400 chunk_usize = chunk_size;
402 /* Calculate compressed size of next chunk. */
405 struct pwm_chunk_hdr chunk_hdr;
407 ret = full_pread(in_fd, &chunk_hdr,
408 sizeof(chunk_hdr), cur_read_offset);
411 chunk_csize = le32_to_cpu(chunk_hdr.compressed_size);
413 if (i == num_chunks - 1) {
414 chunk_csize = rspec->size_in_wim -
415 chunk_table_full_size -
416 chunk_offsets[i - read_start_chunk];
417 if (rspec->is_pipable)
418 chunk_csize -= num_chunks * sizeof(struct pwm_chunk_hdr);
420 chunk_csize = chunk_offsets[i + 1 - read_start_chunk] -
421 chunk_offsets[i - read_start_chunk];
424 if (chunk_csize == 0 || chunk_csize > chunk_usize) {
425 ERROR("Invalid chunk size in compressed resource!");
427 ret = WIMLIB_ERR_DECOMPRESSION;
428 goto out_free_memory;
430 if (rspec->is_pipable)
431 cur_read_offset += sizeof(struct pwm_chunk_hdr);
433 /* Offsets in the uncompressed resource at which this chunk
434 * starts and ends. */
435 const u64 chunk_start_offset = i << chunk_order;
436 const u64 chunk_end_offset = chunk_start_offset + chunk_usize;
438 if (chunk_end_offset <= cur_range_pos) {
440 /* The next range does not require data in this chunk,
442 cur_read_offset += chunk_csize;
446 ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
452 /* Read the chunk and feed data to the callback
456 if (chunk_csize == chunk_usize)
461 ret = full_pread(in_fd,
468 if (read_buf == cbuf) {
469 DEBUG("Decompressing chunk %"PRIu64" "
470 "(csize=%"PRIu32" usize=%"PRIu32")",
471 i, chunk_csize, chunk_usize);
472 ret = wimlib_decompress(cbuf,
478 ERROR("Failed to decompress data!");
479 ret = WIMLIB_ERR_DECOMPRESSION;
481 goto out_free_memory;
484 cur_read_offset += chunk_csize;
486 /* At least one range requires data in this chunk. */
488 size_t start, end, size;
490 /* Calculate how many bytes of data should be
491 * sent to the callback function, taking into
492 * account that data sent to the callback
493 * function must not overlap range boundaries.
495 start = cur_range_pos - chunk_start_offset;
496 end = min(cur_range_end, chunk_end_offset) - chunk_start_offset;
499 ret = (*cb)(&ubuf[start], size, cb_ctx);
502 goto out_free_memory;
504 cur_range_pos += size;
505 if (cur_range_pos == cur_range_end) {
506 /* Advance to next range. */
507 if (++cur_range == end_range) {
508 cur_range_pos = ~0ULL;
510 cur_range_pos = cur_range->offset;
511 cur_range_end = cur_range->offset + cur_range->size;
514 } while (cur_range_pos < chunk_end_offset);
519 last_offset == rspec->uncompressed_size - 1 &&
523 /* If reading a pipable resource from a pipe and the full data
524 * was requested, skip the chunk table at the end so that the
525 * file descriptor is fully clear of the resource after this
527 cur_read_offset += chunk_table_size;
528 ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
537 wimlib_free_decompressor(rspec->wim->decompressor);
538 rspec->wim->decompressor = decompressor;
539 rspec->wim->decompressor_ctype = ctype;
540 rspec->wim->decompressor_max_block_size = chunk_size;
542 if (chunk_offsets_malloced)
552 ERROR("Not enough memory available to read size=%"PRIu64" bytes "
553 "from compressed WIM resource!", last_offset - first_offset + 1);
555 ret = WIMLIB_ERR_NOMEM;
556 goto out_free_memory;
559 ERROR_WITH_ERRNO("Error reading compressed WIM resource!");
560 goto out_free_memory;
563 /* Read raw data from a file descriptor at the specified offset, feeding the
564 * data it in chunks into the specified callback function. */
566 read_raw_file_data(struct filedes *in_fd, u64 offset, u64 size,
567 consume_data_callback_t cb, void *cb_ctx)
570 size_t bytes_to_read;
574 bytes_to_read = min(sizeof(buf), size);
575 ret = full_pread(in_fd, buf, bytes_to_read, offset);
577 ERROR_WITH_ERRNO("Read error");
580 ret = cb(buf, bytes_to_read, cb_ctx);
583 size -= bytes_to_read;
584 offset += bytes_to_read;
589 /* A consume_data_callback_t implementation that simply concatenates all chunks
592 bufferer_cb(const void *chunk, size_t size, void *_ctx)
596 *buf_p = mempcpy(*buf_p, chunk, size);
601 * read_partial_wim_resource()-
603 * Read a range of data from an uncompressed or compressed resource in a WIM
607 * Specification of the WIM resource to read from.
609 * Offset within the uncompressed resource at which to start reading.
611 * Number of bytes to read.
613 * Callback function to feed the data being read. Each call provides the
614 * next chunk of the requested data, uncompressed. Each chunk will be of
615 * nonzero size and will not cross range boundaries, but otherwise will be
616 * of unspecified size.
618 * Parameter to pass to @cb_ctx.
621 * WIMLIB_ERR_SUCCESS (0)
622 * WIMLIB_ERR_READ (errno set)
623 * WIMLIB_ERR_UNEXPECTED_END_OF_FILE (errno set to 0)
624 * WIMLIB_ERR_NOMEM (errno set to ENOMEM)
625 * WIMLIB_ERR_DECOMPRESSION (errno set to EINVAL)
627 * or other error code returned by the @cb function.
630 read_partial_wim_resource(const struct wim_resource_spec *rspec,
631 u64 offset, u64 size,
632 consume_data_callback_t cb, void *cb_ctx)
635 wimlib_assert(offset + size >= offset);
636 wimlib_assert(offset + size <= rspec->uncompressed_size);
638 DEBUG("Reading %"PRIu64" @ %"PRIu64" from WIM resource "
639 "%"PRIu64" => %"PRIu64" @ %"PRIu64,
640 size, offset, rspec->uncompressed_size,
641 rspec->size_in_wim, rspec->offset_in_wim);
647 if (resource_is_compressed(rspec)) {
648 struct data_range range = {
652 return read_compressed_wim_resource(rspec, &range, 1,
655 return read_raw_file_data(&rspec->wim->in_fd,
656 rspec->offset_in_wim + offset,
663 /* Read the specified range of uncompressed data from the specified stream,
664 * which must be located into a WIM file, into the specified buffer. */
666 read_partial_wim_stream_into_buf(const struct wim_lookup_table_entry *lte,
667 size_t size, u64 offset, void *_buf)
671 wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
673 return read_partial_wim_resource(lte->rspec,
674 lte->offset_in_res + offset,
680 /* A consume_data_callback_t implementation that simply ignores the data
683 skip_chunk_cb(const void *chunk, size_t size, void *_ctx)
688 /* Skip over the data of the specified stream, which must correspond to a full
691 skip_wim_stream(struct wim_lookup_table_entry *lte)
693 wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
694 wimlib_assert(!(lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS));
695 DEBUG("Skipping stream (size=%"PRIu64")", lte->size);
696 return read_partial_wim_resource(lte->rspec,
698 lte->rspec->uncompressed_size,
704 read_wim_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
705 consume_data_callback_t cb, void *cb_ctx)
707 return read_partial_wim_resource(lte->rspec, lte->offset_in_res, size,
712 /* This function handles reading stream data that is located in an external
713 * file, such as a file that has been added to the WIM image through execution
714 * of a wimlib_add_command.
716 * This assumes the file can be accessed using the standard POSIX open(),
717 * read(), and close(). On Windows this will not necessarily be the case (since
718 * the file may need FILE_FLAG_BACKUP_SEMANTICS to be opened, or the file may be
719 * encrypted), so Windows uses its own code for its equivalent case. */
721 read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte, u64 size,
722 consume_data_callback_t cb, void *cb_ctx)
728 wimlib_assert(size <= lte->size);
730 DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", size, lte->file_on_disk);
732 raw_fd = open(lte->file_on_disk, O_BINARY | O_RDONLY);
734 ERROR_WITH_ERRNO("Can't open \"%"TS"\"", lte->file_on_disk);
735 return WIMLIB_ERR_OPEN;
737 filedes_init(&fd, raw_fd);
738 ret = read_raw_file_data(&fd, 0, size, cb, cb_ctx);
742 #endif /* !__WIN32__ */
744 /* This function handles the trivial case of reading stream data that is, in
745 * fact, already located in an in-memory buffer. */
747 read_buffer_prefix(const struct wim_lookup_table_entry *lte,
748 u64 size, consume_data_callback_t cb, void *cb_ctx)
750 wimlib_assert(size <= lte->size);
751 return (*cb)(lte->attached_buffer, size, cb_ctx);
754 typedef int (*read_stream_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
756 consume_data_callback_t cb,
760 * read_stream_prefix()-
762 * Reads the first @size bytes from a generic "stream", which may be located in
763 * any one of several locations, such as in a WIM file (compressed or
764 * uncompressed), in an external file, or directly in an in-memory buffer.
766 * This function feeds the data to a callback function @cb in chunks of
769 * Returns 0 on success; nonzero on error. A nonzero value will be returned if
770 * the stream data cannot be successfully read (for a number of different
771 * reasons, depending on the stream location), or if @cb returned nonzero in
772 * which case that error code will be returned.
775 read_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
776 consume_data_callback_t cb, void *cb_ctx)
778 static const read_stream_prefix_handler_t handlers[] = {
779 [RESOURCE_IN_WIM] = read_wim_stream_prefix,
781 [RESOURCE_IN_FILE_ON_DISK] = read_win32_file_prefix,
783 [RESOURCE_IN_FILE_ON_DISK] = read_file_on_disk_prefix,
785 [RESOURCE_IN_ATTACHED_BUFFER] = read_buffer_prefix,
787 [RESOURCE_IN_STAGING_FILE] = read_file_on_disk_prefix,
790 [RESOURCE_IN_NTFS_VOLUME] = read_ntfs_file_prefix,
793 [RESOURCE_WIN32_ENCRYPTED] = read_win32_encrypted_file_prefix,
796 wimlib_assert(lte->resource_location < ARRAY_LEN(handlers)
797 && handlers[lte->resource_location] != NULL);
798 return handlers[lte->resource_location](lte, size, cb, cb_ctx);
801 /* Read the full uncompressed data of the specified stream into the specified
802 * buffer, which must have space for at least lte->size bytes. */
804 read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *_buf)
807 return read_stream_prefix(lte, lte->size, bufferer_cb, &buf);
810 /* Retrieve the full uncompressed data of the specified stream. A buffer large
811 * enough hold the data is allocated and returned in @buf_ret. */
813 read_full_stream_into_alloc_buf(const struct wim_lookup_table_entry *lte,
819 if ((size_t)lte->size != lte->size) {
820 ERROR("Can't read %"PRIu64" byte stream into "
821 "memory", lte->size);
822 return WIMLIB_ERR_NOMEM;
825 buf = MALLOC(lte->size);
827 return WIMLIB_ERR_NOMEM;
829 ret = read_full_stream_into_buf(lte, buf);
839 /* Retrieve the full uncompressed data of the specified WIM resource. A buffer
840 * large enough hold the data is allocated and returned in @buf_ret. */
842 wim_resource_spec_to_data(struct wim_resource_spec *rspec, void **buf_ret)
845 struct wim_lookup_table_entry *lte;
847 lte = new_lookup_table_entry();
849 return WIMLIB_ERR_NOMEM;
851 lte_bind_wim_resource_spec(lte, rspec);
852 lte->flags = rspec->flags;
853 lte->size = rspec->uncompressed_size;
854 lte->offset_in_res = 0;
856 ret = read_full_stream_into_alloc_buf(lte, buf_ret);
858 lte_unbind_wim_resource_spec(lte);
859 free_lookup_table_entry(lte);
863 /* Retrieve the full uncompressed data of a WIM resource specified as a raw
864 * `wim_reshdr' and the corresponding WIM file. A large enough hold the data is
865 * allocated and returned in @buf_ret. */
867 wim_reshdr_to_data(const struct wim_reshdr *reshdr, WIMStruct *wim, void **buf_ret)
869 DEBUG("offset_in_wim=%"PRIu64", size_in_wim=%"PRIu64", "
870 "uncompressed_size=%"PRIu64,
871 reshdr->offset_in_wim, reshdr->size_in_wim,
872 reshdr->uncompressed_size);
874 struct wim_resource_spec rspec;
875 wim_res_hdr_to_spec(reshdr, wim, &rspec);
876 return wim_resource_spec_to_data(&rspec, buf_ret);
879 struct streamifier_context {
880 struct read_stream_list_callbacks cbs;
881 struct wim_lookup_table_entry *cur_stream;
882 struct wim_lookup_table_entry *next_stream;
883 u64 cur_stream_offset;
884 struct wim_lookup_table_entry *final_stream;
885 size_t list_head_offset;
888 static struct wim_lookup_table_entry *
889 next_stream(struct wim_lookup_table_entry *lte, size_t list_head_offset)
891 struct list_head *cur;
893 cur = (struct list_head*)((u8*)lte + list_head_offset);
895 return (struct wim_lookup_table_entry*)((u8*)cur->next - list_head_offset);
898 /* A consume_data_callback_t implementation that translates raw resource data
899 * into streams, calling the begin_stream, consume_chunk, and end_stream
900 * callback functions as appropriate. */
902 streamifier_cb(const void *chunk, size_t size, void *_ctx)
904 struct streamifier_context *ctx = _ctx;
907 DEBUG("%zu bytes passed to streamifier", size);
909 wimlib_assert(ctx->cur_stream != NULL);
910 wimlib_assert(size <= ctx->cur_stream->size - ctx->cur_stream_offset);
912 if (ctx->cur_stream_offset == 0) {
913 /* Starting a new stream. */
914 DEBUG("Begin new stream (size=%"PRIu64").",
915 ctx->cur_stream->size);
916 ret = (*ctx->cbs.begin_stream)(ctx->cur_stream, true,
917 ctx->cbs.begin_stream_ctx);
922 /* Consume the chunk. */
923 ret = (*ctx->cbs.consume_chunk)(chunk, size,
924 ctx->cbs.consume_chunk_ctx);
927 ctx->cur_stream_offset += size;
929 if (ctx->cur_stream_offset == ctx->cur_stream->size) {
930 /* Finished reading all the data for a stream. */
932 ctx->cur_stream_offset = 0;
934 DEBUG("End stream (size=%"PRIu64").", ctx->cur_stream->size);
935 ret = (*ctx->cbs.end_stream)(ctx->cur_stream, 0,
936 ctx->cbs.end_stream_ctx);
940 /* Advance to next stream. */
941 ctx->cur_stream = ctx->next_stream;
942 if (ctx->cur_stream != NULL) {
943 if (ctx->cur_stream != ctx->final_stream)
944 ctx->next_stream = next_stream(ctx->cur_stream,
945 ctx->list_head_offset);
947 ctx->next_stream = NULL;
953 struct hasher_context {
956 struct read_stream_list_callbacks cbs;
959 /* Callback for starting to read a stream while calculating its SHA1 message
962 hasher_begin_stream(struct wim_lookup_table_entry *lte, bool is_partial_res,
965 struct hasher_context *ctx = _ctx;
967 sha1_init(&ctx->sha_ctx);
969 if (ctx->cbs.begin_stream == NULL)
972 return (*ctx->cbs.begin_stream)(lte, is_partial_res,
973 ctx->cbs.begin_stream_ctx);
976 /* A consume_data_callback_t implementation that continues calculating the SHA1
977 * message digest of the stream being read, then optionally passes the data on
978 * to another consume_data_callback_t implementation. This allows checking the
979 * SHA1 message digest of a stream being extracted, for example. */
981 hasher_consume_chunk(const void *chunk, size_t size, void *_ctx)
983 struct hasher_context *ctx = _ctx;
985 sha1_update(&ctx->sha_ctx, chunk, size);
986 if (ctx->cbs.consume_chunk == NULL)
989 return (*ctx->cbs.consume_chunk)(chunk, size, ctx->cbs.consume_chunk_ctx);
992 /* Callback for finishing reading a stream while calculating its SHA1 message
995 hasher_end_stream(struct wim_lookup_table_entry *lte, int status, void *_ctx)
997 struct hasher_context *ctx = _ctx;
998 u8 hash[SHA1_HASH_SIZE];
1002 /* Error occurred; the full stream may not have been read. */
1007 /* Retrieve the final SHA1 message digest. */
1008 sha1_final(hash, &ctx->sha_ctx);
1010 if (lte->unhashed) {
1011 if (ctx->flags & COMPUTE_MISSING_STREAM_HASHES) {
1012 /* No SHA1 message digest was previously present for the
1013 * stream. Set it to the one just calculated. */
1014 DEBUG("Set SHA1 message digest for stream "
1015 "(size=%"PRIu64").", lte->size);
1016 copy_hash(lte->hash, hash);
1019 if (ctx->flags & VERIFY_STREAM_HASHES) {
1020 /* The stream already had a SHA1 message digest present. Verify
1021 * that it is the same as the calculated value. */
1022 if (!hashes_equal(hash, lte->hash)) {
1023 if (wimlib_print_errors) {
1024 ERROR("Invalid SHA1 message digest "
1025 "on the following WIM stream:");
1026 print_lookup_table_entry(lte, stderr);
1028 ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
1032 DEBUG("SHA1 message digest okay for "
1033 "stream (size=%"PRIu64").", lte->size);
1038 if (ctx->cbs.end_stream == NULL)
1041 return (*ctx->cbs.end_stream)(lte, ret, ctx->cbs.end_stream_ctx);
1045 read_full_stream_with_cbs(struct wim_lookup_table_entry *lte,
1046 const struct read_stream_list_callbacks *cbs)
1050 ret = (*cbs->begin_stream)(lte, false, cbs->begin_stream_ctx);
1054 ret = read_stream_prefix(lte, lte->size, cbs->consume_chunk,
1055 cbs->consume_chunk_ctx);
1057 return (*cbs->end_stream)(lte, ret, cbs->end_stream_ctx);
1060 /* Read the full data of the specified stream, passing the data into the
1061 * specified callbacks (all of which are optional) and either checking or
1062 * computing the SHA1 message digest of the stream. */
1064 read_full_stream_with_sha1(struct wim_lookup_table_entry *lte,
1065 const struct read_stream_list_callbacks *cbs)
1067 struct hasher_context hasher_ctx = {
1068 .flags = VERIFY_STREAM_HASHES | COMPUTE_MISSING_STREAM_HASHES,
1071 struct read_stream_list_callbacks hasher_cbs = {
1072 .begin_stream = hasher_begin_stream,
1073 .begin_stream_ctx = &hasher_ctx,
1074 .consume_chunk = hasher_consume_chunk,
1075 .consume_chunk_ctx = &hasher_ctx,
1076 .end_stream = hasher_end_stream,
1077 .end_stream_ctx = &hasher_ctx,
1080 return read_full_stream_with_cbs(lte, &hasher_cbs);
1084 read_packed_streams(struct wim_lookup_table_entry *first_stream,
1085 struct wim_lookup_table_entry *last_stream,
1087 size_t list_head_offset,
1088 const struct read_stream_list_callbacks *sink_cbs)
1090 struct data_range *ranges;
1091 bool ranges_malloced;
1092 struct wim_lookup_table_entry *cur_stream;
1095 u64 ranges_alloc_size;
1097 DEBUG("Reading %"PRIu64" streams combined in same WIM resource",
1100 /* Setup data ranges array (one range per stream to read); this way
1101 * read_compressed_wim_resource() does not need to be aware of streams.
1104 ranges_alloc_size = stream_count * sizeof(ranges[0]);
1106 if (unlikely((size_t)ranges_alloc_size != ranges_alloc_size)) {
1107 ERROR("Too many streams in one resource!");
1108 return WIMLIB_ERR_NOMEM;
1110 if (likely(ranges_alloc_size <= STACK_MAX)) {
1111 ranges = alloca(ranges_alloc_size);
1112 ranges_malloced = false;
1114 ranges = MALLOC(ranges_alloc_size);
1115 if (ranges == NULL) {
1116 ERROR("Too many streams in one resource!");
1117 return WIMLIB_ERR_NOMEM;
1119 ranges_malloced = true;
1122 for (i = 0, cur_stream = first_stream;
1124 i++, cur_stream = next_stream(cur_stream, list_head_offset))
1126 ranges[i].offset = cur_stream->offset_in_res;
1127 ranges[i].size = cur_stream->size;
1130 struct streamifier_context streamifier_ctx = {
1132 .cur_stream = first_stream,
1133 .next_stream = next_stream(first_stream, list_head_offset),
1134 .cur_stream_offset = 0,
1135 .final_stream = last_stream,
1136 .list_head_offset = list_head_offset,
1139 ret = read_compressed_wim_resource(first_stream->rspec,
1145 if (ranges_malloced)
1149 if (streamifier_ctx.cur_stream_offset != 0) {
1150 ret = (*streamifier_ctx.cbs.end_stream)
1151 (streamifier_ctx.cur_stream,
1153 streamifier_ctx.cbs.end_stream_ctx);
1160 * Read a list of streams, each of which may be in any supported location (e.g.
1161 * in a WIM or in an external file). Unlike read_stream_prefix() or the
1162 * functions which call it, this function optimizes the case where multiple
1163 * streams are packed into a single compressed WIM resource and reads them all
1164 * consecutively, only decompressing the data one time.
1167 * List of streams (represented as `struct wim_lookup_table_entry's) to
1170 * Offset of the `struct list_head' within each `struct
1171 * wim_lookup_table_entry' that makes up the @stream_list.
1173 * Callback functions to accept the stream data.
1175 * Bitwise OR of zero or more of the following flags:
1177 * VERIFY_STREAM_HASHES:
1178 * For all streams being read that have already had SHA1 message
1179 * digests computed, calculate the SHA1 message digest of the read
1180 * data and compare it with the previously computed value. If they
1181 * do not match, return WIMLIB_ERR_INVALID_RESOURCE_HASH.
1183 * COMPUTE_MISSING_STREAM_HASHES
1184 * For all streams being read that have not yet had their SHA1
1185 * message digests computed, calculate and save their SHA1 message
1188 * STREAM_LIST_ALREADY_SORTED
1189 * @stream_list is already sorted in sequential order for reading.
1191 * The callback functions are allowed to delete the current stream from the list
1194 * Returns 0 on success; a nonzero error code on failure. Failure can occur due
1195 * to an error reading the data or due to an error status being returned by any
1196 * of the callback functions.
1199 read_stream_list(struct list_head *stream_list,
1200 size_t list_head_offset,
1201 const struct read_stream_list_callbacks *cbs,
1205 struct list_head *cur, *next;
1206 struct wim_lookup_table_entry *lte;
1207 struct hasher_context *hasher_ctx;
1208 struct read_stream_list_callbacks *sink_cbs;
1210 if (!(flags & STREAM_LIST_ALREADY_SORTED)) {
1211 ret = sort_stream_list_by_sequential_order(stream_list, list_head_offset);
1216 if (flags & (VERIFY_STREAM_HASHES | COMPUTE_MISSING_STREAM_HASHES)) {
1217 hasher_ctx = alloca(sizeof(*hasher_ctx));
1218 *hasher_ctx = (struct hasher_context) {
1222 sink_cbs = alloca(sizeof(*sink_cbs));
1223 *sink_cbs = (struct read_stream_list_callbacks) {
1224 .begin_stream = hasher_begin_stream,
1225 .begin_stream_ctx = hasher_ctx,
1226 .consume_chunk = hasher_consume_chunk,
1227 .consume_chunk_ctx = hasher_ctx,
1228 .end_stream = hasher_end_stream,
1229 .end_stream_ctx = hasher_ctx,
1232 sink_cbs = (struct read_stream_list_callbacks*)cbs;
1235 for (cur = stream_list->next, next = cur->next;
1237 cur = next, next = cur->next)
1239 lte = (struct wim_lookup_table_entry*)((u8*)cur - list_head_offset);
1241 if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
1242 lte->size != lte->rspec->uncompressed_size)
1245 struct wim_lookup_table_entry *lte_next, *lte_last;
1246 struct list_head *next2;
1249 /* The next stream is a proper sub-sequence of a WIM
1250 * resource. See if there are other streams in the same
1251 * resource that need to be read. Since
1252 * sort_stream_list_by_sequential_order() sorted the
1253 * streams by offset in the WIM, this can be determined
1254 * by simply scanning forward in the list. */
1259 next2 != stream_list
1260 && (lte_next = (struct wim_lookup_table_entry*)
1261 ((u8*)next2 - list_head_offset),
1262 lte_next->resource_location == RESOURCE_IN_WIM
1263 && lte_next->rspec == lte->rspec);
1264 next2 = next2->next)
1266 lte_last = lte_next;
1269 if (stream_count > 1) {
1270 /* Reading multiple streams combined into a
1271 * single WIM resource. They are in the stream
1272 * list, sorted by offset; @lte specifies the
1273 * first stream in the resource that needs to be
1274 * read and @lte_last specifies the last stream
1275 * in the resource that needs to be read. */
1277 ret = read_packed_streams(lte, lte_last,
1287 ret = read_full_stream_with_cbs(lte, sink_cbs);
1288 if (ret && ret != BEGIN_STREAM_STATUS_SKIP_STREAM)
1294 /* Extract the first @size bytes of the specified stream.
1296 * If @size specifies the full uncompressed size of the stream, then the SHA1
1297 * message digest of the uncompressed stream is checked while being extracted.
1299 * The uncompressed data of the resource is passed in chunks of unspecified size
1300 * to the @extract_chunk function, passing it @extract_chunk_arg. */
1302 extract_stream(struct wim_lookup_table_entry *lte, u64 size,
1303 consume_data_callback_t extract_chunk, void *extract_chunk_arg)
1305 wimlib_assert(size <= lte->size);
1306 if (size == lte->size) {
1308 struct read_stream_list_callbacks cbs = {
1309 .consume_chunk = extract_chunk,
1310 .consume_chunk_ctx = extract_chunk_arg,
1312 return read_full_stream_with_sha1(lte, &cbs);
1314 /* Don't do SHA1. */
1315 return read_stream_prefix(lte, size, extract_chunk,
1320 /* A consume_data_callback_t implementation that writes the chunk of data to a
1321 * file descriptor. */
1323 extract_chunk_to_fd(const void *chunk, size_t size, void *_fd_p)
1325 struct filedes *fd = _fd_p;
1327 int ret = full_write(fd, chunk, size);
1329 ERROR_WITH_ERRNO("Error writing to file descriptor");
1335 /* Extract the first @size bytes of the specified stream to the specified file
1338 extract_stream_to_fd(struct wim_lookup_table_entry *lte,
1339 struct filedes *fd, u64 size)
1341 return extract_stream(lte, size, extract_chunk_to_fd, fd);
1344 /* Extract the full uncompressed contents of the specified stream to the
1345 * specified file descriptor. */
1347 extract_full_stream_to_fd(struct wim_lookup_table_entry *lte,
1350 return extract_stream_to_fd(lte, fd, lte->size);
1353 /* Calculate the SHA1 message digest of a stream and store it in @lte->hash. */
1355 sha1_stream(struct wim_lookup_table_entry *lte)
1357 wimlib_assert(lte->unhashed);
1358 struct read_stream_list_callbacks cbs = {
1360 return read_full_stream_with_sha1(lte, &cbs);
1363 /* Convert a short WIM resource header to a stand-alone WIM resource
1366 * Note: for packed resources some fields still need to be overridden.
1369 wim_res_hdr_to_spec(const struct wim_reshdr *reshdr, WIMStruct *wim,
1370 struct wim_resource_spec *rspec)
1373 rspec->offset_in_wim = reshdr->offset_in_wim;
1374 rspec->size_in_wim = reshdr->size_in_wim;
1375 rspec->uncompressed_size = reshdr->uncompressed_size;
1376 INIT_LIST_HEAD(&rspec->stream_list);
1377 rspec->flags = reshdr->flags;
1378 rspec->is_pipable = wim_is_pipable(wim);
1379 if (rspec->flags & WIM_RESHDR_FLAG_COMPRESSED) {
1380 rspec->compression_type = wim->compression_type;
1381 rspec->chunk_size = wim->chunk_size;
1383 rspec->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
1384 rspec->chunk_size = 0;
1388 /* Convert a stand-alone resource specification to a WIM resource header. */
1390 wim_res_spec_to_hdr(const struct wim_resource_spec *rspec,
1391 struct wim_reshdr *reshdr)
1393 reshdr->offset_in_wim = rspec->offset_in_wim;
1394 reshdr->size_in_wim = rspec->size_in_wim;
1395 reshdr->flags = rspec->flags;
1396 reshdr->uncompressed_size = rspec->uncompressed_size;
1399 /* Translates a WIM resource header from the on-disk format into an in-memory
1402 get_wim_reshdr(const struct wim_reshdr_disk *disk_reshdr,
1403 struct wim_reshdr *reshdr)
1405 reshdr->offset_in_wim = le64_to_cpu(disk_reshdr->offset_in_wim);
1406 reshdr->size_in_wim = (((u64)disk_reshdr->size_in_wim[0] << 0) |
1407 ((u64)disk_reshdr->size_in_wim[1] << 8) |
1408 ((u64)disk_reshdr->size_in_wim[2] << 16) |
1409 ((u64)disk_reshdr->size_in_wim[3] << 24) |
1410 ((u64)disk_reshdr->size_in_wim[4] << 32) |
1411 ((u64)disk_reshdr->size_in_wim[5] << 40) |
1412 ((u64)disk_reshdr->size_in_wim[6] << 48));
1413 reshdr->uncompressed_size = le64_to_cpu(disk_reshdr->uncompressed_size);
1414 reshdr->flags = disk_reshdr->flags;
1417 /* Translates a WIM resource header from an in-memory format into the on-disk
1420 put_wim_reshdr(const struct wim_reshdr *reshdr,
1421 struct wim_reshdr_disk *disk_reshdr)
1423 disk_reshdr->size_in_wim[0] = reshdr->size_in_wim >> 0;
1424 disk_reshdr->size_in_wim[1] = reshdr->size_in_wim >> 8;
1425 disk_reshdr->size_in_wim[2] = reshdr->size_in_wim >> 16;
1426 disk_reshdr->size_in_wim[3] = reshdr->size_in_wim >> 24;
1427 disk_reshdr->size_in_wim[4] = reshdr->size_in_wim >> 32;
1428 disk_reshdr->size_in_wim[5] = reshdr->size_in_wim >> 40;
1429 disk_reshdr->size_in_wim[6] = reshdr->size_in_wim >> 48;
1430 disk_reshdr->flags = reshdr->flags;
1431 disk_reshdr->offset_in_wim = cpu_to_le64(reshdr->offset_in_wim);
1432 disk_reshdr->uncompressed_size = cpu_to_le64(reshdr->uncompressed_size);