4 * Read uncompressed and compressed metadata and file resources from a WIM file.
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/lzms.h"
35 #include "wimlib/resource.h"
36 #include "wimlib/sha1.h"
39 /* for read_win32_file_prefix(), read_win32_encrypted_file_prefix() */
40 # include "wimlib/win32.h"
44 /* for read_ntfs_file_prefix() */
45 # include "wimlib/ntfs_3g.h"
57 * Compressed WIM resources
59 * A compressed resource in a WIM consists of a number of compressed chunks,
60 * each of which decompresses to a fixed chunk size (given in the WIM header;
61 * usually 32768) except possibly the last, which always decompresses to any
62 * remaining bytes. In addition, immediately before the chunks, a table (the
63 * "chunk table") provides the offset, in bytes relative to the end of the chunk
64 * table, of the start of each compressed chunk, except for the first chunk
65 * which is omitted as it always has an offset of 0. Therefore, a compressed
66 * resource with N chunks will have a chunk table with N - 1 entries.
68 * Additional information:
70 * - Entries in the chunk table are 4 bytes each, except if the uncompressed
71 * size of the resource is greater than 4 GiB, in which case the entries in
72 * the chunk table are 8 bytes each. In either case, the entries are unsigned
73 * little-endian integers.
75 * - The chunk table is included in the compressed size of the resource provided
76 * in the corresponding entry in the WIM's stream lookup table.
78 * - The compressed size of a chunk is never greater than the uncompressed size.
79 * From the compressor's point of view, chunks that would have compressed to a
80 * size greater than or equal to their original size are in fact stored
81 * uncompressed. From the decompresser's point of view, chunks with
82 * compressed size equal to their uncompressed size are in fact uncompressed.
84 * Furthermore, wimlib supports its own "pipable" WIM format, and for this the
85 * structure of compressed resources was modified to allow piped reading and
86 * writing. To make sequential writing possible, the chunk table is placed
87 * after the chunks rather than before the chunks, and to make sequential
88 * reading possible, each chunk is prefixed with a 4-byte header giving its
89 * compressed size as a 32-bit, unsigned, little-endian integer. Otherwise the
90 * details are the same.
94 /* Decompress the specified chunk that uses the specified compression type
95 * @ctype, part of a WIM with default chunk size @wim_chunk_size. For LZX the
96 * separate @wim_chunk_size is needed because it determines the window size used
97 * for LZX compression. */
99 decompress(const void *cchunk, unsigned clen, void *uchunk, unsigned ulen,
100 int ctype, u32 wim_chunk_size)
103 case WIMLIB_COMPRESSION_TYPE_LZX:
104 return wimlib_lzx_decompress2(cchunk, clen,
105 uchunk, ulen, wim_chunk_size);
106 case WIMLIB_COMPRESSION_TYPE_XPRESS:
107 return wimlib_xpress_decompress(cchunk, clen,
109 case WIMLIB_COMPRESSION_TYPE_LZMS:
110 return lzms_decompress(cchunk, clen,
111 uchunk, ulen, wim_chunk_size);
118 /* Read data from a compressed WIM resource. Assumes parameters were already
119 * verified by read_partial_wim_resource(). */
121 read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
122 const u64 size, const consume_data_callback_t cb,
123 const u32 cb_chunk_size, void * const ctx_or_buf,
124 const int flags, const u64 offset)
129 const u32 orig_chunk_size = rspec->cchunk_size;
130 const u32 orig_chunk_order = bsr32(orig_chunk_size);
132 wimlib_assert(is_power_of_2(orig_chunk_size));
134 /* Handle the trivial case. */
138 if (rspec->ctype == WIMLIB_COMPRESSION_TYPE_LZMS) {
141 unsigned clen = rspec->size_in_wim;
142 unsigned ulen = rspec->uncompressed_size;
144 fprintf(stderr, "clen=%u, ulen=%u, offset=%lu\n", clen, ulen,
145 rspec->offset_in_wim);
147 u8 *cbuf = MALLOC(clen);
148 u8 *ubuf = MALLOC(ulen);
150 ret = full_pread(&rspec->wim->in_fd,
151 cbuf, clen, rspec->offset_in_wim);
153 ERROR_WITH_ERRNO("Can't read compressed data");
157 ret = lzms_decompress(cbuf, clen, ubuf, ulen,
160 ERROR("LZMS decompression error.");
162 ret = WIMLIB_ERR_DECOMPRESSION;
167 for (u64 i = offset; i < offset + size; i += chunk_size) {
168 chunk_size = min(offset + size - i, cb_chunk_size);
169 ret = cb(&ubuf[i], chunk_size, ctx_or_buf);
174 memcpy(ctx_or_buf, &ubuf[offset], size);
183 u64 *chunk_offsets = NULL;
186 void *compressed_buf = NULL;
187 bool chunk_offsets_malloced = false;
188 bool out_buf_malloced = false;
189 bool tmp_buf_malloced = false;
190 bool compressed_buf_malloced = false;
192 /* Get the file descriptor for the WIM. */
193 struct filedes * const in_fd = &rspec->wim->in_fd;
195 /* Determine if we're reading a pipable resource from a pipe or not. */
196 const bool is_pipe_read = !filedes_is_seekable(in_fd);
198 /* Calculate the number of chunks the resource is divided into. */
199 const u64 num_chunks = (rspec->uncompressed_size + orig_chunk_size - 1) >> orig_chunk_order;
201 /* Calculate the 0-based index of the chunk at which the read starts.
203 const u64 start_chunk = offset >> orig_chunk_order;
205 /* For pipe reads, we always must start from the 0th chunk. */
206 const u64 actual_start_chunk = (is_pipe_read ? 0 : start_chunk);
208 /* Calculate the offset, within the start chunk, of the first byte of
210 const u32 start_offset_in_chunk = offset & (orig_chunk_size - 1);
212 /* Calculate the index of the chunk that contains the last byte of the
214 const u64 end_chunk = (offset + size - 1) >> orig_chunk_order;
216 /* Calculate the offset, within the end chunk, of the last byte of the
218 const u32 end_offset_in_chunk = (offset + size - 1) & (orig_chunk_size - 1);
220 /* Calculate the number of entries in the chunk table; it's one less
221 * than the number of chunks, since the first chunk has no entry. */
222 const u64 num_chunk_entries = num_chunks - 1;
224 /* Set the size of each chunk table entry based on the resource's
225 * uncompressed size. */
226 const u64 chunk_entry_size = (rspec->uncompressed_size > (1ULL << 32)) ? 8 : 4;
228 /* Calculate the size, in bytes, of the full chunk table. */
229 const u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
231 /* Current offset to read from. */
232 u64 cur_read_offset = rspec->offset_in_wim;
234 /* Read the chunk table into memory. */
236 /* Calculate the number of chunk entries are actually needed to
237 * read the requested part of the resource. Include an entry
238 * for the first chunk even though that doesn't exist in the
239 * on-disk table, but take into account that if the last chunk
240 * required for the read is not the last chunk of the resource,
241 * an extra chunk entry is needed so that the compressed size of
242 * the last chunk of the read can be determined. */
243 const u64 num_alloc_chunk_entries = end_chunk - start_chunk +
244 1 + (end_chunk != num_chunks - 1);
246 /* Allocate a buffer to hold a subset of the chunk table. It
247 * will only contain offsets for the chunks that are actually
248 * needed for this read. For speed, allocate the buffer on the
249 * stack unless it's too large. */
250 if ((size_t)(num_alloc_chunk_entries * sizeof(u64)) !=
251 (num_alloc_chunk_entries * sizeof(u64)))
254 if (num_alloc_chunk_entries <= STACK_MAX / sizeof(u64)) {
255 chunk_offsets = alloca(num_alloc_chunk_entries * sizeof(u64));
257 chunk_offsets = MALLOC(num_alloc_chunk_entries * sizeof(u64));
258 if (chunk_offsets == NULL)
260 chunk_offsets_malloced = true;
263 /* Set the implicit offset of the first chunk if it's included
264 * in the needed chunks. */
265 if (start_chunk == 0)
266 chunk_offsets[0] = 0;
268 /* Calculate the index of the first needed entry in the chunk
270 const u64 start_table_idx = (start_chunk == 0) ?
273 /* Calculate the number of entries that need to be read from the
275 const u64 num_needed_chunk_entries = (start_chunk == 0) ?
276 num_alloc_chunk_entries - 1 : num_alloc_chunk_entries;
278 /* Calculate the number of bytes of data that need to be read
279 * from the chunk table. */
280 const size_t chunk_table_needed_size =
281 num_needed_chunk_entries * chunk_entry_size;
283 /* Calculate the byte offset, in the WIM file, of the first
284 * chunk table entry to read. Take into account that if the WIM
285 * file is in the special "pipable" format, then the chunk table
286 * is at the end of the resource, not the beginning. */
287 const u64 file_offset_of_needed_chunk_entries =
289 + (start_table_idx * chunk_entry_size)
290 + (rspec->is_pipable ? (rspec->size_in_wim - chunk_table_size) : 0);
292 /* Read the needed chunk table entries into the end of the
293 * chunk_offsets buffer. */
294 void * const chunk_tab_data = (u8*)&chunk_offsets[num_alloc_chunk_entries] -
295 chunk_table_needed_size;
296 ret = full_pread(in_fd, chunk_tab_data, chunk_table_needed_size,
297 file_offset_of_needed_chunk_entries);
301 /* Now fill in chunk_offsets from the entries we have read in
302 * chunk_tab_data. Careful: chunk_offsets aliases
303 * chunk_tab_data, which breaks C's aliasing rules when we read
304 * 32-bit integers and store 64-bit integers. But since the
305 * operations are safe as long as the compiler doesn't mess with
306 * their order, we use the gcc may_alias extension to tell the
307 * compiler that loads from the 32-bit integers may alias stores
308 * to the 64-bit integers. */
310 typedef le64 __attribute__((may_alias)) aliased_le64_t;
311 typedef le32 __attribute__((may_alias)) aliased_le32_t;
312 u64 * const chunk_offsets_p = chunk_offsets + (start_chunk == 0);
315 if (chunk_entry_size == 4) {
316 aliased_le32_t *raw_entries = (aliased_le32_t*)chunk_tab_data;
317 for (i = 0; i < num_needed_chunk_entries; i++)
318 chunk_offsets_p[i] = le32_to_cpu(raw_entries[i]);
320 aliased_le64_t *raw_entries = (aliased_le64_t*)chunk_tab_data;
321 for (i = 0; i < num_needed_chunk_entries; i++)
322 chunk_offsets_p[i] = le64_to_cpu(raw_entries[i]);
326 /* Set offset to beginning of first chunk to read. */
327 cur_read_offset += chunk_offsets[0];
328 if (rspec->is_pipable)
329 cur_read_offset += start_chunk * sizeof(struct pwm_chunk_hdr);
331 cur_read_offset += chunk_table_size;
334 /* If using a callback function, allocate a temporary buffer that will
335 * hold data being passed to it. If writing directly to a buffer
336 * instead, arrange to write data directly into it. */
338 u8 *out_buf_end, *out_p;
340 out_buf_size = max(cb_chunk_size, orig_chunk_size);
341 if (out_buf_size <= STACK_MAX) {
342 out_buf = alloca(out_buf_size);
344 out_buf = MALLOC(out_buf_size);
347 out_buf_malloced = true;
351 out_buf = ctx_or_buf;
353 out_buf_end = out_buf + out_buf_size;
356 /* Unless the raw compressed data was requested, allocate a temporary
357 * buffer for reading compressed chunks, each of which can be at most
358 * @orig_chunk_size - 1 bytes. This excludes compressed chunks that are
359 * a full @orig_chunk_size bytes, which are actually stored
361 if (!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS)) {
362 if (orig_chunk_size - 1 <= STACK_MAX) {
363 compressed_buf = alloca(orig_chunk_size - 1);
365 compressed_buf = MALLOC(orig_chunk_size - 1);
366 if (compressed_buf == NULL)
368 compressed_buf_malloced = true;
372 /* Allocate yet another temporary buffer, this one for decompressing
373 * chunks for which only part of the data is needed. */
374 if (start_offset_in_chunk != 0 ||
375 (end_offset_in_chunk != orig_chunk_size - 1 &&
376 offset + size != rspec->uncompressed_size))
378 if (orig_chunk_size <= STACK_MAX) {
379 tmp_buf = alloca(orig_chunk_size);
381 tmp_buf = MALLOC(orig_chunk_size);
384 tmp_buf_malloced = true;
388 /* Read, and possibly decompress, each needed chunk, either writing the
389 * data directly into the @ctx_or_buf buffer or passing it to the @cb
390 * callback function. */
391 for (u64 i = actual_start_chunk; i <= end_chunk; i++) {
393 /* Calculate uncompressed size of next chunk. */
395 if ((i == num_chunks - 1) && (rspec->uncompressed_size & (orig_chunk_size - 1)))
396 chunk_usize = (rspec->uncompressed_size & (orig_chunk_size - 1));
398 chunk_usize = orig_chunk_size;
400 /* Calculate compressed size of next chunk. */
403 struct pwm_chunk_hdr chunk_hdr;
405 ret = full_pread(in_fd, &chunk_hdr,
406 sizeof(chunk_hdr), cur_read_offset);
409 chunk_csize = le32_to_cpu(chunk_hdr.compressed_size);
411 if (i == num_chunks - 1) {
412 chunk_csize = rspec->size_in_wim -
414 chunk_offsets[i - start_chunk];
415 if (rspec->is_pipable)
416 chunk_csize -= num_chunks * sizeof(struct pwm_chunk_hdr);
418 chunk_csize = chunk_offsets[i + 1 - start_chunk] -
419 chunk_offsets[i - start_chunk];
422 if (chunk_csize == 0 || chunk_csize > chunk_usize) {
423 ERROR("Invalid chunk size in compressed resource!");
425 ret = WIMLIB_ERR_DECOMPRESSION;
426 goto out_free_memory;
428 if (rspec->is_pipable)
429 cur_read_offset += sizeof(struct pwm_chunk_hdr);
431 if (i >= start_chunk) {
432 /* Calculate how much of this chunk needs to be read. */
433 u32 chunk_needed_size;
434 u32 start_offset = 0;
435 u32 end_offset = orig_chunk_size - 1;
437 if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
438 chunk_needed_size = chunk_csize;
440 if (i == start_chunk)
441 start_offset = start_offset_in_chunk;
444 end_offset = end_offset_in_chunk;
446 chunk_needed_size = end_offset + 1 - start_offset;
449 if (chunk_csize == chunk_usize ||
450 (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS))
452 /* Read the raw chunk data. */
454 ret = full_pread(in_fd,
457 cur_read_offset + start_offset);
461 /* Read and decompress the chunk. */
465 ret = full_pread(in_fd,
472 if (chunk_needed_size == chunk_usize)
477 ret = decompress(compressed_buf,
484 ERROR("Failed to decompress data!");
485 ret = WIMLIB_ERR_DECOMPRESSION;
487 goto out_free_memory;
489 if (chunk_needed_size != chunk_usize)
490 memcpy(out_p, tmp_buf + start_offset,
494 out_p += chunk_needed_size;
497 /* Feed the data to the callback function. */
499 if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
500 ret = cb(out_buf, out_p - out_buf, ctx_or_buf);
502 goto out_free_memory;
504 } else if (i == end_chunk || out_p == out_buf_end) {
508 for (p = out_buf; p != out_p; p += bytes_sent) {
509 bytes_sent = min(cb_chunk_size, out_p - p);
510 ret = cb(p, bytes_sent, ctx_or_buf);
512 goto out_free_memory;
517 cur_read_offset += chunk_csize;
521 /* Skip data only. */
522 cur_read_offset += chunk_csize;
523 ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
530 && size == rspec->uncompressed_size
534 /* Skip chunk table at end of pipable resource. */
536 cur_read_offset += chunk_table_size;
537 ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
544 if (chunk_offsets_malloced)
546 if (out_buf_malloced)
548 if (compressed_buf_malloced)
549 FREE(compressed_buf);
550 if (tmp_buf_malloced)
556 ERROR("Not enough memory available to read size=%"PRIu64" bytes "
557 "from compressed resource!", size);
559 ret = WIMLIB_ERR_NOMEM;
560 goto out_free_memory;
563 ERROR_WITH_ERRNO("Error reading compressed file resource!");
564 goto out_free_memory;
567 /* Read raw data from a file descriptor at the specified offset. */
569 read_raw_file_data(struct filedes *in_fd, u64 size, consume_data_callback_t cb,
570 u32 cb_chunk_size, void *ctx_or_buf, u64 offset)
574 bool tmp_buf_malloced = false;
577 /* Send data to callback function in chunks. */
578 if (cb_chunk_size <= STACK_MAX) {
579 tmp_buf = alloca(cb_chunk_size);
581 tmp_buf = MALLOC(cb_chunk_size);
582 if (tmp_buf == NULL) {
583 ret = WIMLIB_ERR_NOMEM;
586 tmp_buf_malloced = true;
590 size_t bytes_to_read = min(cb_chunk_size, size);
591 ret = full_pread(in_fd, tmp_buf, bytes_to_read,
595 ret = cb(tmp_buf, bytes_to_read, ctx_or_buf);
598 size -= bytes_to_read;
599 offset += bytes_to_read;
602 /* Read data directly into buffer. */
603 ret = full_pread(in_fd, ctx_or_buf, size, offset);
611 ERROR_WITH_ERRNO("Read error");
613 if (tmp_buf_malloced)
619 * read_partial_wim_resource()-
621 * Read a range of data from an uncompressed or compressed resource in a WIM
622 * file. Data is written into a buffer or fed into a callback function, as
623 * documented in read_stream_prefix().
625 * By default, this function provides the uncompressed data of the resource, and
626 * @size and @offset and interpreted relative to the uncompressed contents of
627 * the resource. This behavior can be modified by either of the following
630 * WIMLIB_READ_RESOURCE_FLAG_RAW_FULL:
631 * Read @size bytes at @offset of the raw contents of the compressed
632 * resource. In the case of pipable resources, this excludes the stream
633 * header. Exclusive with WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS.
635 * WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS:
636 * Read the raw compressed chunks of the compressed resource. @size must
637 * be the full uncompressed size, @offset must be 0, and @cb_chunk_size
638 * must be the resource chunk size.
641 * WIMLIB_ERR_SUCCESS (0)
642 * WIMLIB_ERR_READ (errno set)
643 * WIMLIB_ERR_UNEXPECTED_END_OF_FILE (errno set to 0)
644 * WIMLIB_ERR_NOMEM (errno set to ENOMEM)
645 * WIMLIB_ERR_DECOMPRESSION (errno set to EINVAL)
647 * or other error code returned by the @cb function.
650 read_partial_wim_resource(const struct wim_lookup_table_entry *lte,
651 u64 size, consume_data_callback_t cb,
652 u32 cb_chunk_size, void *ctx_or_buf,
653 int flags, u64 offset)
655 const struct wim_resource_spec *rspec;
656 struct filedes *in_fd;
658 /* Verify parameters. */
659 wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
661 in_fd = &rspec->wim->in_fd;
663 wimlib_assert(is_power_of_2(cb_chunk_size));
664 if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
665 /* Raw chunks mode is subject to the restrictions noted. */
666 wimlib_assert(!lte_is_partial(lte));
667 wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL));
668 wimlib_assert(cb_chunk_size == rspec->cchunk_size);
669 wimlib_assert(size == rspec->uncompressed_size);
670 wimlib_assert(offset == 0);
671 } else if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) {
672 /* Raw full mode: read must not overrun end of store size. */
673 wimlib_assert(!lte_is_partial(lte));
674 wimlib_assert(offset + size >= size &&
675 offset + size <= rspec->size_in_wim);
677 /* Normal mode: read must not overrun end of original size. */
678 wimlib_assert(offset + size >= size &&
679 lte->offset_in_res + offset + size <= rspec->uncompressed_size);
682 DEBUG("Reading WIM resource: %"PRIu64" @ +%"PRIu64"[+%"PRIu64"] "
683 "from %"PRIu64"(%"PRIu64") @ +%"PRIu64" "
684 "(readflags 0x%08x, resflags 0x%02x%s)",
685 size, offset, lte->offset_in_res,
687 rspec->uncompressed_size,
688 rspec->offset_in_wim,
690 (rspec->is_pipable ? ", pipable" : ""));
692 if ((flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) ||
693 rspec->ctype == WIMLIB_COMPRESSION_TYPE_NONE)
695 return read_raw_file_data(in_fd,
700 offset + rspec->offset_in_wim);
702 return read_compressed_wim_resource(rspec, size, cb,
704 ctx_or_buf, flags, offset + lte->offset_in_res);
709 read_partial_wim_stream_into_buf(const struct wim_lookup_table_entry *lte,
710 size_t size, u64 offset, void *buf)
712 return read_partial_wim_resource(lte, size, NULL, 0, buf, 0, offset);
716 read_wim_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
717 consume_data_callback_t cb, u32 cb_chunk_size,
718 void *ctx_or_buf, int flags)
720 return read_partial_wim_resource(lte, size, cb, cb_chunk_size,
721 ctx_or_buf, flags, 0);
725 /* This function handles reading stream data that is located in an external
726 * file, such as a file that has been added to the WIM image through execution
727 * of a wimlib_add_command.
729 * This assumes the file can be accessed using the standard POSIX open(),
730 * read(), and close(). On Windows this will not necessarily be the case (since
731 * the file may need FILE_FLAG_BACKUP_SEMANTICS to be opened, or the file may be
732 * encrypted), so Windows uses its own code for its equivalent case.
735 read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte, u64 size,
736 consume_data_callback_t cb, u32 cb_chunk_size,
737 void *ctx_or_buf, int _ignored_flags)
743 wimlib_assert(size <= lte->size);
744 DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", size, lte->file_on_disk);
746 raw_fd = open(lte->file_on_disk, O_BINARY | O_RDONLY);
748 ERROR_WITH_ERRNO("Can't open \"%"TS"\"", lte->file_on_disk);
749 return WIMLIB_ERR_OPEN;
751 filedes_init(&fd, raw_fd);
752 ret = read_raw_file_data(&fd, size, cb, cb_chunk_size, ctx_or_buf, 0);
756 #endif /* !__WIN32__ */
758 /* This function handles the trivial case of reading stream data that is, in
759 * fact, already located in an in-memory buffer. */
761 read_buffer_prefix(const struct wim_lookup_table_entry *lte,
762 u64 size, consume_data_callback_t cb,
763 u32 cb_chunk_size, void *ctx_or_buf, int _ignored_flags)
765 wimlib_assert(size <= lte->size);
768 /* Feed the data into the callback function in
769 * appropriately-sized chunks. */
773 for (u64 offset = 0; offset < size; offset += chunk_size) {
774 chunk_size = min(cb_chunk_size, size - offset);
775 ret = cb((const u8*)lte->attached_buffer + offset,
776 chunk_size, ctx_or_buf);
781 /* Copy the data directly into the specified buffer. */
782 memcpy(ctx_or_buf, lte->attached_buffer, size);
787 typedef int (*read_stream_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
788 u64 size, consume_data_callback_t cb,
789 u32 cb_chunk_size, void *ctx_or_buf,
793 * read_stream_prefix()-
795 * Reads the first @size bytes from a generic "stream", which may be located in
796 * any one of several locations, such as in a WIM file (compressed or
797 * uncompressed), in an external file, or directly in an in-memory buffer.
799 * This function feeds the data either to a callback function (@cb != NULL,
800 * passing it @ctx_or_buf), or write it directly into a buffer (@cb == NULL,
801 * @ctx_or_buf specifies the buffer, which must have room for at least @size
804 * When (@cb != NULL), @cb_chunk_size specifies the maximum size of data chunks
805 * to feed the callback function. @cb_chunk_size must be positive, and if the
806 * stream is in a WIM file, must be a power of 2. All chunks, except possibly
807 * the last one, will be this size. If (@cb == NULL), @cb_chunk_size is
810 * If the stream is located in a WIM file, @flags can be set as documented in
811 * read_partial_wim_resource(). Otherwise @flags are ignored.
813 * Returns 0 on success; nonzero on error. A nonzero value will be returned if
814 * the stream data cannot be successfully read (for a number of different
815 * reasons, depending on the stream location), or if a callback function was
816 * specified and it returned nonzero.
819 read_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
820 consume_data_callback_t cb, u32 cb_chunk_size,
821 void *ctx_or_buf, int flags)
823 /* This function merely verifies several preconditions, then passes
824 * control to an appropriate function for understanding each possible
825 * stream location. */
826 static const read_stream_prefix_handler_t handlers[] = {
827 [RESOURCE_IN_WIM] = read_wim_stream_prefix,
829 [RESOURCE_IN_FILE_ON_DISK] = read_win32_file_prefix,
831 [RESOURCE_IN_FILE_ON_DISK] = read_file_on_disk_prefix,
833 [RESOURCE_IN_ATTACHED_BUFFER] = read_buffer_prefix,
835 [RESOURCE_IN_STAGING_FILE] = read_file_on_disk_prefix,
838 [RESOURCE_IN_NTFS_VOLUME] = read_ntfs_file_prefix,
841 [RESOURCE_WIN32_ENCRYPTED] = read_win32_encrypted_file_prefix,
844 wimlib_assert(lte->resource_location < ARRAY_LEN(handlers)
845 && handlers[lte->resource_location] != NULL);
846 wimlib_assert(cb == NULL || cb_chunk_size > 0);
847 return handlers[lte->resource_location](lte, size, cb, cb_chunk_size,
851 /* Read the full uncompressed data of the specified stream into the specified
852 * buffer, which must have space for at least lte->size bytes. */
854 read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *buf)
856 return read_stream_prefix(lte, lte->size, NULL, 0, buf, 0);
859 /* Read the full uncompressed data of the specified stream. A buffer sufficient
860 * to hold the data is allocated and returned in @buf_ret. */
862 read_full_stream_into_alloc_buf(const struct wim_lookup_table_entry *lte,
868 if ((size_t)lte->size != lte->size) {
869 ERROR("Can't read %"PRIu64" byte stream into "
870 "memory", lte->size);
871 return WIMLIB_ERR_NOMEM;
874 buf = MALLOC(lte->size);
876 return WIMLIB_ERR_NOMEM;
878 ret = read_full_stream_into_buf(lte, buf);
888 /* Retrieve the full uncompressed data of the specified WIM resource. */
890 wim_resource_spec_to_data(struct wim_resource_spec *rspec, void **buf_ret)
893 struct wim_lookup_table_entry *lte;
895 lte = new_lookup_table_entry();
897 return WIMLIB_ERR_NOMEM;
900 lte_bind_wim_resource_spec(lte, rspec);
901 lte->flags = rspec->flags;
902 lte->size = rspec->uncompressed_size;
903 lte->offset_in_res = 0;
905 ret = read_full_stream_into_alloc_buf(lte, buf_ret);
907 lte_unbind_wim_resource_spec(lte);
908 free_lookup_table_entry(lte);
912 /* Retrieve the full uncompressed data of the specified WIM resource. */
914 wim_reshdr_to_data(const struct wim_reshdr *reshdr, WIMStruct *wim, void **buf_ret)
916 DEBUG("offset_in_wim=%"PRIu64", size_in_wim=%"PRIu64", "
917 "uncompressed_size=%"PRIu64,
918 reshdr->offset_in_wim, reshdr->size_in_wim, reshdr->uncompressed_size);
920 struct wim_resource_spec rspec;
921 wim_res_hdr_to_spec(reshdr, wim, &rspec);
922 return wim_resource_spec_to_data(&rspec, buf_ret);
927 consume_data_callback_t extract_chunk;
928 void *extract_chunk_arg;
932 extract_chunk_sha1_wrapper(const void *chunk, size_t chunk_size, void *_ctx)
934 struct extract_ctx *ctx = _ctx;
936 sha1_update(&ctx->sha_ctx, chunk, chunk_size);
937 return ctx->extract_chunk(chunk, chunk_size, ctx->extract_chunk_arg);
940 /* Extracts the first @size bytes of a stream to somewhere. In the process, the
941 * SHA1 message digest of the uncompressed stream is checked if the full stream
942 * is being extracted.
944 * @extract_chunk is a function that will be called to extract each chunk of the
947 extract_stream(const struct wim_lookup_table_entry *lte, u64 size,
948 consume_data_callback_t extract_chunk, void *extract_chunk_arg)
951 if (size == lte->size) {
953 struct extract_ctx ctx;
954 ctx.extract_chunk = extract_chunk;
955 ctx.extract_chunk_arg = extract_chunk_arg;
956 sha1_init(&ctx.sha_ctx);
957 ret = read_stream_prefix(lte, size,
958 extract_chunk_sha1_wrapper,
959 lte_cchunk_size(lte),
962 u8 hash[SHA1_HASH_SIZE];
963 sha1_final(hash, &ctx.sha_ctx);
964 if (!hashes_equal(hash, lte->hash)) {
965 if (wimlib_print_errors) {
966 ERROR("Invalid SHA1 message digest "
967 "on the following WIM stream:");
968 print_lookup_table_entry(lte, stderr);
969 if (lte->resource_location == RESOURCE_IN_WIM)
970 ERROR("The WIM file appears to be corrupt!");
972 ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
977 ret = read_stream_prefix(lte, size, extract_chunk,
978 lte_cchunk_size(lte),
979 extract_chunk_arg, 0);
985 extract_wim_chunk_to_fd(const void *buf, size_t len, void *_fd_p)
987 struct filedes *fd = _fd_p;
988 int ret = full_write(fd, buf, len);
990 ERROR_WITH_ERRNO("Error writing to file descriptor");
994 /* Extract the first @size bytes of the specified stream to the specified file
995 * descriptor. If @size is the full size of the stream, its SHA1 message digest
996 * is also checked. */
998 extract_stream_to_fd(const struct wim_lookup_table_entry *lte,
999 struct filedes *fd, u64 size)
1001 return extract_stream(lte, size, extract_wim_chunk_to_fd, fd);
1006 sha1_chunk(const void *buf, size_t len, void *ctx)
1008 sha1_update(ctx, buf, len);
1012 /* Calculate the SHA1 message digest of a stream, storing it in @lte->hash. */
1014 sha1_stream(struct wim_lookup_table_entry *lte)
1019 sha1_init(&sha_ctx);
1020 ret = read_stream_prefix(lte, lte->size,
1021 sha1_chunk, lte_cchunk_size(lte),
1024 sha1_final(lte->hash, &sha_ctx);
1029 /* Convert a WIM resource header to a stand-alone resource specification. */
1031 wim_res_hdr_to_spec(const struct wim_reshdr *reshdr, WIMStruct *wim,
1032 struct wim_resource_spec *spec)
1035 spec->offset_in_wim = reshdr->offset_in_wim;
1036 spec->size_in_wim = reshdr->size_in_wim;
1037 spec->uncompressed_size = reshdr->uncompressed_size;
1038 INIT_LIST_HEAD(&spec->lte_list);
1039 spec->flags = reshdr->flags;
1040 spec->is_pipable = wim_is_pipable(wim);
1041 if (spec->flags & (WIM_RESHDR_FLAG_COMPRESSED | WIM_RESHDR_FLAG_CONCAT)) {
1042 spec->ctype = wim->compression_type;
1043 spec->cchunk_size = wim->chunk_size;
1045 spec->ctype = WIMLIB_COMPRESSION_TYPE_NONE;
1046 spec->cchunk_size = 0;
1050 /* Convert a stand-alone resource specification to a WIM resource header. */
1052 wim_res_spec_to_hdr(const struct wim_resource_spec *rspec,
1053 struct wim_reshdr *reshdr)
1055 reshdr->offset_in_wim = rspec->offset_in_wim;
1056 reshdr->size_in_wim = rspec->size_in_wim;
1057 reshdr->flags = rspec->flags;
1058 reshdr->uncompressed_size = rspec->uncompressed_size;
1061 /* Translates a WIM resource header from the on-disk format into an in-memory
1064 get_wim_reshdr(const struct wim_reshdr_disk *disk_reshdr,
1065 struct wim_reshdr *reshdr)
1067 reshdr->offset_in_wim = le64_to_cpu(disk_reshdr->offset_in_wim);
1068 reshdr->size_in_wim = (((u64)disk_reshdr->size_in_wim[0] << 0) |
1069 ((u64)disk_reshdr->size_in_wim[1] << 8) |
1070 ((u64)disk_reshdr->size_in_wim[2] << 16) |
1071 ((u64)disk_reshdr->size_in_wim[3] << 24) |
1072 ((u64)disk_reshdr->size_in_wim[4] << 32) |
1073 ((u64)disk_reshdr->size_in_wim[5] << 40) |
1074 ((u64)disk_reshdr->size_in_wim[6] << 48));
1075 reshdr->uncompressed_size = le64_to_cpu(disk_reshdr->uncompressed_size);
1076 reshdr->flags = disk_reshdr->flags;
1078 /* Truncate numbers to 62 bits to avoid possible overflows. */
1079 if (reshdr->offset_in_wim & 0xc000000000000000ULL)
1080 return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
1082 if (reshdr->uncompressed_size & 0xc000000000000000ULL)
1083 return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
1088 /* Translates a WIM resource header from an in-memory format into the on-disk
1091 put_wim_reshdr(const struct wim_reshdr *reshdr,
1092 struct wim_reshdr_disk *disk_reshdr)
1094 disk_reshdr->size_in_wim[0] = reshdr->size_in_wim >> 0;
1095 disk_reshdr->size_in_wim[1] = reshdr->size_in_wim >> 8;
1096 disk_reshdr->size_in_wim[2] = reshdr->size_in_wim >> 16;
1097 disk_reshdr->size_in_wim[3] = reshdr->size_in_wim >> 24;
1098 disk_reshdr->size_in_wim[4] = reshdr->size_in_wim >> 32;
1099 disk_reshdr->size_in_wim[5] = reshdr->size_in_wim >> 40;
1100 disk_reshdr->size_in_wim[6] = reshdr->size_in_wim >> 48;
1101 disk_reshdr->flags = reshdr->flags;
1102 disk_reshdr->offset_in_wim = cpu_to_le64(reshdr->offset_in_wim);
1103 disk_reshdr->uncompressed_size = cpu_to_le64(reshdr->uncompressed_size);