4 * Read uncompressed and compressed metadata and file resources.
8 * Copyright (C) 2010 Carl Thijssen
9 * Copyright (C) 2012 Eric Biggers
11 * This file is part of wimlib, a library for working with WIM files.
13 * wimlib is free software; you can redistribute it and/or modify it under the
14 * terms of the GNU Lesser General Public License as published by the Free
15 * Software Foundation; either version 2.1 of the License, or (at your option)
18 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20 * A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
23 * You should have received a copy of the GNU Lesser General Public License
24 * along with wimlib; if not, see http://www.gnu.org/licenses/.
27 #include "wimlib_internal.h"
28 #include "lookup_table.h"
41 * Reads all or part of a compressed resource into an in-memory buffer.
43 * @fp: The FILE* for the WIM file.
44 * @resource_compressed_size: The compressed size of the resource.
45 * @resource_uncompressed_size: The uncompressed size of the resource.
46 * @resource_offset: The offset of the start of the resource from
47 * the start of the stream @fp.
48 * @resource_ctype: The compression type of the resource.
49 * @len: The number of bytes of uncompressed data to read from
51 * @offset: The offset of the bytes to read within the uncompressed
53 * @contents_len: An array into which the uncompressed data is written.
54 * It must be at least @len bytes long.
56 * Returns zero on success, nonzero on failure.
58 static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
59 u64 resource_uncompressed_size,
60 u64 resource_offset, int resource_ctype,
61 u64 len, u64 offset, u8 contents_ret[])
64 DEBUG2("comp size = %"PRIu64", uncomp size = %"PRIu64", "
65 "res offset = %"PRIu64"",
66 resource_compressed_size,
67 resource_uncompressed_size,
69 DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"",
70 wimlib_get_compression_type_string(resource_ctype), len, offset);
75 int (*decompress)(const void *, uint, void *, uint);
76 /* Set the appropriate decompress function. */
77 if (resource_ctype == WIM_COMPRESSION_TYPE_LZX)
78 decompress = lzx_decompress;
80 decompress = xpress_decompress;
82 /* The structure of a compressed resource consists of a table of chunk
83 * offsets followed by the chunks themselves. Each chunk consists of
84 * compressed data, and there is one chunk for each WIM_CHUNK_SIZE =
85 * 32768 bytes of the uncompressed file, with the last chunk having any
88 * The chunk offsets are measured relative to the end of the chunk
89 * table. The first chunk is omitted from the table in the WIM file
90 * because its offset is implicitly given by the fact that it directly
91 * follows the chunk table and therefore must have an offset of 0.
94 /* Calculate how many chunks the resource conists of in its entirety. */
95 u64 num_chunks = (resource_uncompressed_size + WIM_CHUNK_SIZE - 1) /
97 /* As mentioned, the first chunk has no entry in the chunk table. */
98 u64 num_chunk_entries = num_chunks - 1;
101 /* The index of the chunk that the read starts at. */
102 u64 start_chunk = offset / WIM_CHUNK_SIZE;
103 /* The byte offset at which the read starts, within the start chunk. */
104 u64 start_chunk_offset = offset % WIM_CHUNK_SIZE;
106 /* The index of the chunk that contains the last byte of the read. */
107 u64 end_chunk = (offset + len - 1) / WIM_CHUNK_SIZE;
108 /* The byte offset of the last byte of the read, within the end chunk */
109 u64 end_chunk_offset = (offset + len - 1) % WIM_CHUNK_SIZE;
111 /* Number of chunks that are actually needed to read the requested part
113 u64 num_needed_chunks = end_chunk - start_chunk + 1;
115 /* If the end chunk is not the last chunk, an extra chunk entry is
116 * needed because we need to know the offset of the chunk after the last
117 * chunk read to figure out the size of the last read chunk. */
118 if (end_chunk != num_chunks - 1)
121 /* Declare the chunk table. It will only contain offsets for the chunks
122 * that are actually needed for this read. */
123 u64 chunk_offsets[num_needed_chunks];
125 /* Set the implicit offset of the first chunk if it is included in the
128 * Note: M$'s documentation includes a picture that shows the first
129 * chunk starting right after the chunk entry table, labeled as offset
130 * 0x10. However, in the actual file format, the offset is measured
131 * from the end of the chunk entry table, so the first chunk has an
133 if (start_chunk == 0)
134 chunk_offsets[0] = 0;
136 /* According to M$'s documentation, if the uncompressed size of
137 * the file is greater than 4 GB, the chunk entries are 8-byte
138 * integers. Otherwise, they are 4-byte integers. */
139 u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
142 /* Size of the full chunk table in the WIM file. */
143 u64 chunk_table_size = chunk_entry_size * num_chunk_entries;
145 /* Read the needed chunk offsets from the table in the WIM file. */
147 /* Index, in the WIM file, of the first needed entry in the
149 u64 start_table_idx = (start_chunk == 0) ? 0 : start_chunk - 1;
151 /* Number of entries we need to actually read from the chunk
152 * table (excludes the implicit first chunk). */
153 u64 num_needed_chunk_entries = (start_chunk == 0) ?
154 num_needed_chunks - 1 : num_needed_chunks;
156 /* Skip over unneeded chunk table entries. */
157 u64 file_offset_of_needed_chunk_entries = resource_offset +
158 start_table_idx * chunk_entry_size;
159 if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) {
160 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
161 "chunk table of compressed resource",
162 file_offset_of_needed_chunk_entries);
163 return WIMLIB_ERR_READ;
166 /* Number of bytes we need to read from the chunk table. */
167 size_t size = num_needed_chunk_entries * chunk_entry_size;
169 u8 chunk_tab_buf[size];
171 if (fread(chunk_tab_buf, 1, size, fp) != size)
174 /* Now fill in chunk_offsets from the entries we have read in
177 u64 *chunk_tab_p = chunk_offsets;
178 if (start_chunk == 0)
181 if (chunk_entry_size == 4) {
182 u32 *entries = (u32*)chunk_tab_buf;
183 while (num_needed_chunk_entries--)
184 *chunk_tab_p++ = to_le32(*entries++);
186 u64 *entries = (u64*)chunk_tab_buf;
187 while (num_needed_chunk_entries--)
188 *chunk_tab_p++ = to_le64(*entries++);
191 /* Done with the chunk table now. We must now seek to the first chunk
192 * that is needed for the read. */
194 u64 file_offset_of_first_needed_chunk = resource_offset +
195 chunk_table_size + chunk_offsets[0];
196 if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) {
197 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
198 "first chunk of compressed resource",
199 file_offset_of_first_needed_chunk);
200 return WIMLIB_ERR_READ;
203 /* Pointer to current position in the output buffer for uncompressed
205 u8 *out_p = (u8*)contents_ret;
207 /* Buffer for compressed data. While most compressed chunks will have a
208 * size much less than WIM_CHUNK_SIZE, WIM_CHUNK_SIZE - 1 is the maximum
209 * size in the worst-case. This assumption is valid only if chunks that
210 * happen to compress to more than the uncompressed size (i.e. a
211 * sequence of random bytes) are always stored uncompressed. But this seems
212 * to be the case in M$'s WIM files, even though it is undocumented. */
213 u8 compressed_buf[WIM_CHUNK_SIZE - 1];
216 /* Decompress all the chunks. */
217 for (u64 i = start_chunk; i <= end_chunk; i++) {
219 DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64").",
220 i, start_chunk, end_chunk);
222 /* Calculate the sizes of the compressed chunk and of the
223 * uncompressed chunk. */
224 uint compressed_chunk_size, uncompressed_chunk_size;
225 if (i != num_chunks - 1) {
226 /* All the chunks except the last one in the resource
227 * expand to WIM_CHUNK_SIZE uncompressed, and the amount
228 * of compressed data for the chunk is given by the
229 * difference of offsets in the chunk offset table. */
230 compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
231 chunk_offsets[i - start_chunk];
232 uncompressed_chunk_size = WIM_CHUNK_SIZE;
234 /* The last compressed chunk consists of the remaining
235 * bytes in the file resource, and the last uncompressed
236 * chunk has size equal to however many bytes are left-
237 * that is, the remainder of the uncompressed size when
238 * divided by WIM_CHUNK_SIZE.
240 * Note that the resource_compressed_size includes the
241 * chunk table, so the size of it must be subtracted. */
242 compressed_chunk_size = resource_compressed_size -
244 chunk_offsets[i - start_chunk];
246 uncompressed_chunk_size = resource_uncompressed_size %
249 /* If the remainder is 0, the last chunk actually
250 * uncompresses to a full WIM_CHUNK_SIZE bytes. */
251 if (uncompressed_chunk_size == 0)
252 uncompressed_chunk_size = WIM_CHUNK_SIZE;
255 DEBUG2("compressed_chunk_size = %u, "
256 "uncompressed_chunk_size = %u",
257 compressed_chunk_size, uncompressed_chunk_size);
260 /* Figure out how much of this chunk we actually need to read */
262 if (i == start_chunk)
263 start_offset = start_chunk_offset;
268 end_offset = end_chunk_offset;
270 end_offset = WIM_CHUNK_SIZE - 1;
272 u64 partial_chunk_size = end_offset + 1 - start_offset;
273 bool is_partial_chunk = (partial_chunk_size !=
274 uncompressed_chunk_size);
276 DEBUG2("start_offset = %u, end_offset = %u", start_offset,
278 DEBUG2("partial_chunk_size = %u", partial_chunk_size);
280 /* This is undocumented, but chunks can be uncompressed. This
281 * appears to always be the case when the compressed chunk size
282 * is equal to the uncompressed chunk size. */
283 if (compressed_chunk_size == uncompressed_chunk_size) {
284 /* Probably an uncompressed chunk */
286 if (start_offset != 0) {
287 if (fseeko(fp, start_offset, SEEK_CUR) != 0) {
288 ERROR_WITH_ERRNO("Uncompressed partial "
289 "chunk fseek() error");
290 return WIMLIB_ERR_READ;
293 if (fread(out_p, 1, partial_chunk_size, fp) !=
297 /* Compressed chunk */
300 /* Read the compressed data into compressed_buf. */
301 if (fread(compressed_buf, 1, compressed_chunk_size,
302 fp) != compressed_chunk_size)
305 /* For partial chunks we must buffer the uncompressed
306 * data because we don't need all of it. */
307 if (is_partial_chunk) {
308 u8 uncompressed_buf[uncompressed_chunk_size];
310 ret = decompress(compressed_buf,
311 compressed_chunk_size,
313 uncompressed_chunk_size);
315 return WIMLIB_ERR_DECOMPRESSION;
316 memcpy(out_p, uncompressed_buf + start_offset,
319 ret = decompress(compressed_buf,
320 compressed_chunk_size,
322 uncompressed_chunk_size);
324 return WIMLIB_ERR_DECOMPRESSION;
328 /* Advance the pointer into the uncompressed output data by the
329 * number of uncompressed bytes that were written. */
330 out_p += partial_chunk_size;
337 ERROR("Unexpected EOF in compressed file resource");
339 ERROR_WITH_ERRNO("Error reading compressed file resource");
340 return WIMLIB_ERR_READ;
344 * Reads uncompressed data from an open file stream.
346 int read_uncompressed_resource(FILE *fp, u64 offset, u64 len,
349 if (fseeko(fp, offset, SEEK_SET) != 0) {
350 ERROR("Failed to seek to byte %"PRIu64" of input file "
351 "to read uncompressed resource (len = %"PRIu64")",
353 return WIMLIB_ERR_READ;
355 if (fread(contents_ret, 1, len, fp) != len) {
357 ERROR("Unexpected EOF in uncompressed file resource");
359 ERROR("Failed to read %"PRIu64" bytes from "
360 "uncompressed resource at offset %"PRIu64,
363 return WIMLIB_ERR_READ;
371 /* Reads the contents of a struct resource_entry, as represented in the on-disk
372 * format, from the memory pointed to by @p, and fills in the fields of @entry.
373 * A pointer to the byte after the memory read at @p is returned. */
374 const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry)
379 p = get_u56(p, &size);
380 p = get_u8(p, &flags);
382 entry->flags = flags;
383 p = get_u64(p, &entry->offset);
384 p = get_u64(p, &entry->original_size);
388 /* Copies the struct resource_entry @entry to the memory pointed to by @p in the
389 * on-disk format. A pointer to the byte after the memory written at @p is
391 u8 *put_resource_entry(u8 *p, const struct resource_entry *entry)
393 p = put_u56(p, entry->size);
394 p = put_u8(p, entry->flags);
395 p = put_u64(p, entry->offset);
396 p = put_u64(p, entry->original_size);
401 * Reads some data from the resource corresponding to a WIM lookup table entry.
403 * @lte: The WIM lookup table entry for the resource.
404 * @buf: Buffer into which to write the data.
405 * @size: Number of bytes to read.
406 * @offset: Offset at which to start reading the resource.
407 * @raw: If %true, compressed data is read literally rather than being
408 * decompressed first.
410 * Returns zero on success, nonzero on failure.
412 int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[],
413 size_t size, u64 offset, bool raw)
415 /* We shouldn't be allowing read over-runs in any part of the library.
418 wimlib_assert(offset + size <= lte->resource_entry.size);
420 wimlib_assert(offset + size <= lte->resource_entry.original_size);
425 switch (lte->resource_location) {
426 case RESOURCE_IN_WIM:
427 /* The resource is in a WIM file, and its WIMStruct is given by
428 * the lte->wim member. The resource may be either compressed
429 * or uncompressed. */
430 wimlib_assert(lte->wim);
431 wimlib_assert(lte->wim->fp);
432 ctype = wim_resource_compression_type(lte);
434 wimlib_assert(ctype != WIM_COMPRESSION_TYPE_NONE ||
435 (lte->resource_entry.original_size ==
436 lte->resource_entry.size));
438 if (raw || ctype == WIM_COMPRESSION_TYPE_NONE)
439 return read_uncompressed_resource(lte->wim->fp,
440 lte->resource_entry.offset + offset,
443 return read_compressed_resource(lte->wim->fp,
444 lte->resource_entry.size,
445 lte->resource_entry.original_size,
446 lte->resource_entry.offset,
447 ctype, size, offset, buf);
449 case RESOURCE_IN_STAGING_FILE:
450 case RESOURCE_IN_FILE_ON_DISK:
451 /* The resource is in some file on the external filesystem and
452 * needs to be read uncompressed */
453 wimlib_assert(lte->file_on_disk);
454 wimlib_assert(<e->file_on_disk == <e->staging_file_name);
455 /* Use existing file pointer if available; otherwise open one
457 if (lte->file_on_disk_fp) {
458 fp = lte->file_on_disk_fp;
460 fp = fopen(lte->file_on_disk, "rb");
462 ERROR_WITH_ERRNO("Failed to open the file "
463 "`%s'", lte->file_on_disk);
466 ret = read_uncompressed_resource(fp, offset, size, buf);
467 if (fp != lte->file_on_disk_fp)
471 case RESOURCE_IN_ATTACHED_BUFFER:
472 /* The resource is directly attached uncompressed in an
473 * in-memory buffer. */
474 wimlib_assert(lte->attached_buffer);
475 memcpy(buf, lte->attached_buffer + offset, size);
484 * Reads all the data from the resource corresponding to a WIM lookup table
487 * @lte: The WIM lookup table entry for the resource.
488 * @buf: Buffer into which to write the data. It must be at least
489 * wim_resource_size(lte) bytes long.
491 * Returns 0 on success; nonzero on failure.
493 int read_full_wim_resource(const struct lookup_table_entry *lte, u8 buf[])
495 return read_wim_resource(lte, buf, wim_resource_size(lte), 0, false);
498 /* Chunk table that's located at the beginning of each compressed resource in
499 * the WIM. (This is not the on-disk format; the on-disk format just has an
500 * array of offsets.) */
504 u64 original_resource_size;
505 u64 bytes_per_chunk_entry;
513 * Allocates and initializes a chunk table, and reserves space for it in the
517 begin_wim_resource_chunk_tab(const struct lookup_table_entry *lte,
520 struct chunk_table **chunk_tab_ret)
522 u64 size = wim_resource_size(lte);
523 u64 num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
524 struct chunk_table *chunk_tab = MALLOC(sizeof(struct chunk_table) +
525 num_chunks * sizeof(u64));
528 wimlib_assert(size != 0);
531 ERROR("Failed to allocate chunk table for %"PRIu64" byte "
533 ret = WIMLIB_ERR_NOMEM;
536 chunk_tab->file_offset = file_offset;
537 chunk_tab->num_chunks = num_chunks;
538 chunk_tab->original_resource_size = size;
539 chunk_tab->bytes_per_chunk_entry = (size >= (1ULL << 32)) ? 8 : 4;
540 chunk_tab->table_disk_size = chunk_tab->bytes_per_chunk_entry *
542 chunk_tab->cur_offset = 0;
543 chunk_tab->cur_offset_p = chunk_tab->offsets;
545 if (fwrite(chunk_tab, 1, chunk_tab->table_disk_size, out_fp) !=
546 chunk_tab->table_disk_size) {
547 ERROR_WITH_ERRNO("Failed to write chunk table in compressed "
549 ret = WIMLIB_ERR_WRITE;
553 *chunk_tab_ret = chunk_tab;
559 * Compresses a chunk of a WIM resource.
561 * @chunk: Uncompressed data of the chunk.
562 * @chunk_size: Size of the uncompressed chunk in bytes.
563 * @compressed_chunk: Pointer to output buffer of size at least
564 * (@chunk_size - 1) bytes.
565 * @compressed_chunk_len_ret: Pointer to an unsigned int into which the size
566 * of the compressed chunk will be
568 * @ctype: Type of compression to use. Must be WIM_COMPRESSION_TYPE_LZX
569 * or WIM_COMPRESSION_TYPE_XPRESS.
571 * Returns zero if compressed succeeded, and nonzero if the chunk could not be
572 * compressed to any smaller than @chunk_size. This function cannot fail for
575 static int compress_chunk(const u8 chunk[], unsigned chunk_size,
576 u8 compressed_chunk[],
577 unsigned *compressed_chunk_len_ret,
580 unsigned compressed_chunk_sz;
581 int (*compress)(const void *, unsigned, void *, unsigned *);
583 case WIM_COMPRESSION_TYPE_LZX:
584 compress = lzx_compress;
586 case WIM_COMPRESSION_TYPE_XPRESS:
587 compress = xpress_compress;
593 return (*compress)(chunk, chunk_size, compressed_chunk,
594 compressed_chunk_len_ret);
598 * Writes a chunk of a WIM resource to an output file.
600 * @chunk: Uncompressed data of the chunk.
601 * @chunk_size: Size of the chunk (<= WIM_CHUNK_SIZE)
602 * @out_fp: FILE * to write tho chunk to.
603 * @out_ctype: Compression type to use when writing the chunk (ignored if no
604 * chunk table provided)
605 * @chunk_tab: Pointer to chunk table being created. It is updated with the
606 * offset of the chunk we write.
608 * Returns 0 on success; nonzero on failure.
610 static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size,
611 FILE *out_fp, int out_ctype,
612 struct chunk_table *chunk_tab)
615 unsigned out_chunk_size;
617 wimlib_assert(chunk_size <= WIM_CHUNK_SIZE);
621 out_chunk_size = chunk_size;
623 u8 *compressed_chunk = alloca(chunk_size);
625 unsigned compressed_chunk_len;
627 ret = compress_chunk(chunk, chunk_size, compressed_chunk,
628 &out_chunk_size, out_ctype);
630 out_chunk = compressed_chunk;
633 out_chunk_size = chunk_size;
635 *chunk_tab->cur_offset_p++ = chunk_tab->cur_offset;
636 chunk_tab->cur_offset += out_chunk_size;
639 if (fwrite(out_chunk, 1, out_chunk_size, out_fp) != out_chunk_size) {
640 ERROR_WITH_ERRNO("Failed to write WIM resource chunk");
641 return WIMLIB_ERR_WRITE;
647 * Finishes a WIM chunk tale and writes it to the output file at the correct
650 * The final size of the full compressed resource is returned in the
651 * @compressed_size_p.
654 finish_wim_resource_chunk_tab(struct chunk_table *chunk_tab,
655 FILE *out_fp, u64 *compressed_size_p)
657 size_t bytes_written;
658 if (fseeko(out_fp, chunk_tab->file_offset, SEEK_SET) != 0) {
659 ERROR_WITH_ERRNO("Failed to seek to byte "PRIu64" of output "
660 "WIM file", chunk_tab->file_offset);
661 return WIMLIB_ERR_WRITE;
664 if (chunk_tab->bytes_per_chunk_entry == 8) {
665 array_to_le64(chunk_tab->offsets, chunk_tab->num_chunks);
667 for (u64 i = 0; i < chunk_tab->num_chunks; i++)
668 ((u32*)chunk_tab->offsets)[i] =
669 to_le32(chunk_tab->offsets[i]);
671 bytes_written = fwrite((u8*)chunk_tab->offsets +
672 chunk_tab->bytes_per_chunk_entry,
673 1, chunk_tab->table_disk_size, out_fp);
674 if (bytes_written != chunk_tab->table_disk_size) {
675 ERROR_WITH_ERRNO("Failed to write chunk table in compressed "
677 return WIMLIB_ERR_WRITE;
679 if (fseeko(out_fp, 0, SEEK_END) != 0) {
680 ERROR_WITH_ERRNO("Failed to seek to end of output WIM file");
681 return WIMLIB_ERR_WRITE;
683 *compressed_size_p = chunk_tab->cur_offset + chunk_tab->table_disk_size;
688 * Writes a WIM resource to a FILE * opened for writing. The resource may be
689 * written uncompressed or compressed depending on the @out_ctype parameter.
691 * If by chance the resource compresses to more than the original size (this may
692 * happen with random data or files than are pre-compressed), the resource is
693 * instead written uncompressed (and this is reflected in the @out_res_entry by
694 * removing the WIM_RESHDR_FLAG_COMPRESSED flag).
696 * @lte: The lookup table entry for the WIM resource.
697 * @out_fp: The FILE * to write the resource to.
698 * @out_ctype: The compression type of the resource to write. Note: if this is
699 * the same as the compression type of the WIM resource we
700 * need to read, we simply copy the data (i.e. we do not
701 * uncompress it, then compress it again).
702 * @out_res_entry: If non-NULL, a resource entry that is filled in with the
703 * offset, original size, compressed size, and compression flag
704 * of the output resource.
706 * Returns 0 on success; nonzero on failure.
708 static int write_wim_resource(struct lookup_table_entry *lte,
709 FILE *out_fp, int out_ctype,
710 struct resource_entry *out_res_entry)
714 u64 old_compressed_size;
715 u64 new_compressed_size;
718 struct chunk_table *chunk_tab = NULL;
722 /* Original size of the resource */
723 original_size = wim_resource_size(lte);
725 /* Compressed size of the resource (as it exists now) */
726 old_compressed_size = wim_resource_compressed_size(lte);
728 /* Current offset in output file */
729 file_offset = ftello(out_fp);
730 if (file_offset == -1) {
731 ERROR_WITH_ERRNO("Failed to get offset in output "
733 return WIMLIB_ERR_WRITE;
736 /* Are the compression types the same? If so, do a raw copy (copy
737 * without decompressing and recompressing the data). */
738 raw = (wim_resource_compression_type(lte) == out_ctype
739 && out_ctype != WIM_COMPRESSION_TYPE_NONE);
741 bytes_remaining = old_compressed_size;
743 bytes_remaining = original_size;
745 /* Empty resource; nothing needs to be done, so just return success. */
746 if (bytes_remaining == 0)
749 /* Buffer for reading chunks for the resource */
750 char buf[min(WIM_CHUNK_SIZE, bytes_remaining)];
752 /* If we are writing a compressed resource and not doing a raw copy, we
753 * need to initialize the chunk table */
754 if (out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw) {
755 ret = begin_wim_resource_chunk_tab(lte, out_fp, file_offset,
761 /* If the WIM resource is in an external file, open a FILE * to it so we
762 * don't have to open a temporary one in read_wim_resource() for each
764 if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK
765 && !lte->file_on_disk_fp)
767 wimlib_assert(lte->file_on_disk);
768 lte->file_on_disk_fp = fopen(lte->file_on_disk, "rb");
769 if (!lte->file_on_disk_fp) {
770 ERROR_WITH_ERRNO("Failed to open the file `%s' for "
771 "reading", lte->file_on_disk);
772 ret = WIMLIB_ERR_OPEN;
777 /* If we aren't doing a raw copy, we will compute the SHA1 message
778 * digest of the resource as we read it, and verify it's the same as the
779 * hash given in the lookup table entry once we've finished reading the
785 /* While there are still bytes remaining in the WIM resource, read a
786 * chunk of the resource, update SHA1, then write that chunk using the
787 * desired compression type. */
789 u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE);
790 ret = read_wim_resource(lte, buf, to_read, offset, raw);
794 sha1_update(&ctx, buf, to_read);
795 ret = write_wim_resource_chunk(buf, to_read, out_fp,
796 out_ctype, chunk_tab);
799 bytes_remaining -= to_read;
801 } while (bytes_remaining);
803 /* If writing a compressed resource and not doing a raw copy, write the
804 * chunk table, and finish_wim_resource_chunk_tab() will provide the
805 * compressed size of the resource we wrote. Otherwise, the compressed
806 * size of the written resource is the same as the compressed size of
807 * the existing resource. */
808 if (out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw) {
809 ret = finish_wim_resource_chunk_tab(chunk_tab, out_fp,
810 &new_compressed_size);
814 new_compressed_size = old_compressed_size;
817 /* Verify SHA1 message digest of the resource, unless we are doing a raw
818 * write (in which case we never even saw the uncompressed data). Or,
819 * if the hash we had before is all 0's, just re-set it to be the new
822 u8 md[SHA1_HASH_SIZE];
823 sha1_final(md, &ctx);
824 if (is_zero_hash(lte->hash)) {
825 copy_hash(lte->hash, md);
826 } else if (!hashes_equal(md, lte->hash)) {
827 ERROR("WIM resource has incorrect hash!");
828 if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK) {
829 ERROR("We were reading it from `%s'; maybe it changed "
830 "while we were reading it.",
833 ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
838 if (new_compressed_size > original_size) {
839 /* Oops! We compressed the resource to larger than the original
840 * size. Write the resource uncompressed instead. */
841 if (fseeko(out_fp, file_offset, SEEK_SET) != 0) {
842 ERROR_WITH_ERRNO("Failed to seek to byte "PRIu64" "
843 "of output WIM file", file_offset);
844 ret = WIMLIB_ERR_WRITE;
847 ret = write_wim_resource(lte, out_fp, WIM_COMPRESSION_TYPE_NONE,
851 if (fflush(out_fp) != 0) {
852 ERROR_WITH_ERRNO("Failed to flush output WIM file");
853 ret = WIMLIB_ERR_WRITE;
856 if (ftruncate(fileno(out_fp), file_offset + out_res_entry->size) != 0) {
857 ERROR_WITH_ERRNO("Failed to truncate output WIM file");
858 ret = WIMLIB_ERR_WRITE;
862 wimlib_assert(new_compressed_size <= original_size);
864 out_res_entry->size = new_compressed_size;
865 out_res_entry->original_size = original_size;
866 out_res_entry->offset = file_offset;
867 out_res_entry->flags = lte->resource_entry.flags
868 & ~WIM_RESHDR_FLAG_COMPRESSED;
869 if (out_ctype != WIM_COMPRESSION_TYPE_NONE)
870 out_res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
873 if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK
874 && lte->file_on_disk_fp) {
875 fclose(lte->file_on_disk_fp);
876 lte->file_on_disk_fp = NULL;
883 /* Like write_wim_resource(), but the resource is specified by a buffer of
884 * uncompressed data rather a lookup table entry; also writes the SHA1 hash of
885 * the buffer to @hash. */
886 static int write_wim_resource_from_buffer(const u8 *buf, u64 buf_size,
887 FILE *out_fp, int out_ctype,
888 struct resource_entry *out_res_entry,
889 u8 hash[SHA1_HASH_SIZE])
891 /* Set up a temporary lookup table entry that we provide to
892 * write_wim_resource(). */
893 struct lookup_table_entry lte;
895 lte.resource_entry.flags = 0;
896 lte.resource_entry.original_size = buf_size;
897 lte.resource_entry.size = buf_size;
898 lte.resource_entry.offset = 0;
899 lte.resource_location = RESOURCE_IN_ATTACHED_BUFFER;
900 lte.attached_buffer = (u8*)buf;
903 ret = write_wim_resource(<e, out_fp, out_ctype, out_res_entry);
906 copy_hash(hash, lte.hash);
911 * Extracts the first @size bytes of the WIM resource specified by @lte to the
912 * open file descriptor @fd.
914 * Returns 0 on success; nonzero on failure.
916 int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd,
919 u64 bytes_remaining = size;
920 char buf[min(WIM_CHUNK_SIZE, bytes_remaining)];
923 u8 hash[SHA1_HASH_SIZE];
928 while (bytes_remaining) {
929 u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE);
930 ret = read_wim_resource(lte, buf, to_read, offset, false);
933 sha1_update(&ctx, buf, to_read);
934 if (full_write(fd, buf, to_read) < 0) {
935 ERROR_WITH_ERRNO("Error extracting WIM resource");
936 return WIMLIB_ERR_WRITE;
938 bytes_remaining -= to_read;
941 sha1_final(hash, &ctx);
942 if (!hashes_equal(hash, lte->hash)) {
943 ERROR("Invalid checksum on a WIM resource "
944 "(detected when extracting to external file)");
945 ERROR("The following WIM resource is invalid:");
946 print_lookup_table_entry(lte);
947 return WIMLIB_ERR_INVALID_RESOURCE_HASH;
953 * Extracts the WIM resource specified by @lte to the open file descriptor @fd.
955 * Returns 0 on success; nonzero on failure.
957 int extract_full_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd)
959 return extract_wim_resource_to_fd(lte, fd, wim_resource_size(lte));
963 * Copies the file resource specified by the lookup table entry @lte from the
964 * input WIM to the output WIM that has its FILE * given by
965 * ((WIMStruct*)wim)->out_fp.
967 * The output_resource_entry, out_refcnt, and part_number fields of @lte are
970 * Metadata resources are not copied (they are handled elsewhere for joining and
973 int copy_resource(struct lookup_table_entry *lte, void *wim)
978 if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
982 ret = write_wim_resource(lte, w->out_fp,
983 wim_resource_compression_type(lte),
984 <e->output_resource_entry);
987 lte->out_refcnt = lte->refcnt;
988 lte->part_number = w->hdr.part_number;
993 * Writes a dentry's resources, including the main file resource as well as all
994 * alternate data streams, to the output file.
996 * @dentry: The dentry for the file.
997 * @wim_p: A pointer to the WIMStruct containing @dentry.
999 * @return zero on success, nonzero on failure.
1001 int write_dentry_resources(struct dentry *dentry, void *wim_p)
1003 WIMStruct *w = wim_p;
1005 struct lookup_table_entry *lte;
1006 int ctype = wimlib_get_compression_type(w);
1008 if (w->write_flags & WIMLIB_WRITE_FLAG_VERBOSE) {
1009 wimlib_assert(dentry->full_path_utf8);
1010 printf("Writing streams for `%s'\n", dentry->full_path_utf8);
1013 for (unsigned i = 0; i <= dentry->num_ads; i++) {
1014 lte = dentry_stream_lte(dentry, i, w->lookup_table);
1015 if (lte && ++lte->out_refcnt == 1) {
1016 ret = write_wim_resource(lte, w->out_fp, ctype,
1017 <e->output_resource_entry);
1026 * Reads the metadata metadata resource from the WIM file. The metadata
1027 * resource consists of the security data, followed by the directory entry for
1028 * the root directory, followed by all the other directory entries in the
1029 * filesystem. The subdir_offset field of each directory entry gives the start
1030 * of its child entries from the beginning of the metadata resource. An
1031 * end-of-directory is signaled by a directory entry of length '0', really of
1032 * length 8, because that's how long the 'length' field is.
1034 * @fp: The FILE* for the input WIM file.
1035 * @wim_ctype: The compression type of the WIM file.
1036 * @imd: Pointer to the image metadata structure. Its `metadata_lte'
1037 * member specifies the lookup table entry for the metadata
1038 * resource. The rest of the image metadata entry will be filled
1039 * in by this function.
1041 * @return: Zero on success, nonzero on failure.
1043 int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd)
1049 struct dentry *dentry;
1050 struct wim_security_data *sd;
1051 struct link_group_table *lgt;
1052 const struct lookup_table_entry *metadata_lte;
1053 const struct resource_entry *res_entry;
1055 metadata_lte = imd->metadata_lte;
1056 res_entry = &metadata_lte->resource_entry;
1058 DEBUG("Reading metadata resource: length = %"PRIu64", "
1059 "offset = %"PRIu64"",
1060 res_entry->original_size, res_entry->offset);
1062 if (res_entry->original_size < 8) {
1063 ERROR("Expected at least 8 bytes for the metadata resource");
1064 return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
1067 /* Allocate memory for the uncompressed metadata resource. */
1068 buf = MALLOC(res_entry->original_size);
1071 ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
1072 "metadata resource", res_entry->original_size);
1073 return WIMLIB_ERR_NOMEM;
1076 /* Determine the compression type of the metadata resource. */
1078 /* Read the metadata resource into memory. (It may be compressed.) */
1079 ret = read_full_wim_resource(metadata_lte, buf);
1083 DEBUG("Finished reading metadata resource into memory.");
1085 /* The root directory entry starts after security data, on an 8-byte
1088 * The security data starts with a 4-byte integer giving its total
1091 /* Read the security data into a wim_security_data structure. */
1092 ret = read_security_data(buf, res_entry->original_size, &sd);
1096 dentry = MALLOC(sizeof(struct dentry));
1098 ERROR("Failed to allocate %zu bytes for root dentry",
1099 sizeof(struct dentry));
1100 ret = WIMLIB_ERR_NOMEM;
1101 goto out_free_security_data;
1104 get_u32(buf, &dentry_offset);
1105 if (dentry_offset == 0)
1107 dentry_offset = (dentry_offset + 7) & ~7;
1109 ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry);
1110 /* This is the root dentry, so set its pointers correctly. */
1111 dentry->parent = dentry;
1112 dentry->next = dentry;
1113 dentry->prev = dentry;
1115 goto out_free_dentry_tree;
1117 DEBUG("Reading dentry tree");
1118 /* Now read the entire directory entry tree. */
1119 ret = read_dentry_tree(buf, res_entry->original_size, dentry);
1121 goto out_free_dentry_tree;
1123 DEBUG("Calculating dentry full paths");
1124 /* Calculate the full paths in the dentry tree. */
1125 ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
1127 goto out_free_dentry_tree;
1129 DEBUG("Building link group table");
1130 /* Build hash table that maps hard link group IDs to dentry sets */
1131 lgt = new_link_group_table(9001);
1133 goto out_free_dentry_tree;
1134 ret = for_dentry_in_tree(dentry, link_group_table_insert, lgt);
1138 DEBUG("Freeing duplicate ADS entries in link group table");
1139 ret = link_groups_free_duplicate_data(lgt);
1142 DEBUG("Done reading image metadata");
1145 imd->security_data = sd;
1146 imd->root_dentry = dentry;
1149 free_link_group_table(lgt);
1150 out_free_dentry_tree:
1151 free_dentry_tree(dentry, NULL);
1152 out_free_security_data:
1153 free_security_data(sd);
1159 /* Write the metadata resource for the current WIM image. */
1160 int write_metadata_resource(WIMStruct *w)
1166 struct dentry *root;
1167 struct lookup_table_entry *lte, *duplicate_lte;
1168 u64 metadata_original_size;
1171 * We append 20 random bytes to the metadata resource so that we don't
1172 * have identical metadata resources if we happen to append exactly the
1173 * same image twice without any changes in timestamps. If this were to
1174 * happen, it would cause confusion about the number and order of images
1177 const unsigned random_tail_len = 20;
1179 DEBUG("Writing metadata resource for image %d", w->current_image);
1181 root = wim_root_dentry(w);
1183 const struct wim_security_data *sd = wim_security_data(w);
1185 subdir_offset = sd->total_length + root->length + 8;
1186 calculate_subdir_offsets(root, &subdir_offset);
1187 metadata_original_size = subdir_offset + random_tail_len;
1188 buf = MALLOC(metadata_original_size);
1190 ERROR("Failed to allocate %"PRIu64" bytes for "
1191 "metadata resource", metadata_original_size);
1192 return WIMLIB_ERR_NOMEM;
1195 p = write_security_data(sd, buf);
1197 DEBUG("Writing dentry tree.");
1198 p = write_dentry_tree(root, p);
1199 randomize_byte_array(p, random_tail_len);
1200 wimlib_assert(p - buf + random_tail_len == metadata_original_size);
1202 lte = wim_metadata_lookup_table_entry(w);
1204 ret = write_wim_resource_from_buffer(buf, metadata_original_size,
1206 wimlib_get_compression_type(w),
1207 <e->output_resource_entry,
1209 lookup_table_unlink(w->lookup_table, lte);
1210 lookup_table_insert(w->lookup_table, lte);
1211 wimlib_assert(lte->out_refcnt == 0);
1212 lte->out_refcnt = 1;
1213 lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;