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"
40 * Reads all or part of a compressed resource into an in-memory buffer.
42 * @fp: The FILE* for the WIM file.
43 * @resource_compressed_size: The compressed size of the resource.
44 * @resource_uncompressed_size: The uncompressed size of the resource.
45 * @resource_offset: The offset of the start of the resource from
46 * the start of the stream @fp.
47 * @resource_ctype: The compression type of the resource.
48 * @len: The number of bytes of uncompressed data to read from
50 * @offset: The offset of the bytes to read within the uncompressed
52 * @contents_len: An array into which the uncompressed data is written.
53 * It must be at least @len bytes long.
55 * Returns zero on success, nonzero on failure.
57 static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
58 u64 resource_uncompressed_size,
59 u64 resource_offset, int resource_ctype,
60 u64 len, u64 offset, u8 contents_ret[])
63 DEBUG2("comp size = %"PRIu64", uncomp size = %"PRIu64", "
64 "res offset = %"PRIu64"",
65 resource_compressed_size,
66 resource_uncompressed_size,
68 DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"",
69 wimlib_get_compression_type_string(resource_ctype), len, offset);
74 int (*decompress)(const void *, uint, void *, uint);
75 /* Set the appropriate decompress function. */
76 if (resource_ctype == WIM_COMPRESSION_TYPE_LZX)
77 decompress = lzx_decompress;
79 decompress = xpress_decompress;
81 /* The structure of a compressed resource consists of a table of chunk
82 * offsets followed by the chunks themselves. Each chunk consists of
83 * compressed data, and there is one chunk for each WIM_CHUNK_SIZE =
84 * 32768 bytes of the uncompressed file, with the last chunk having any
87 * The chunk offsets are measured relative to the end of the chunk
88 * table. The first chunk is omitted from the table in the WIM file
89 * because its offset is implicitly given by the fact that it directly
90 * follows the chunk table and therefore must have an offset of 0.
93 /* Calculate how many chunks the resource conists of in its entirety. */
94 u64 num_chunks = (resource_uncompressed_size + WIM_CHUNK_SIZE - 1) /
96 /* As mentioned, the first chunk has no entry in the chunk table. */
97 u64 num_chunk_entries = num_chunks - 1;
100 /* The index of the chunk that the read starts at. */
101 u64 start_chunk = offset / WIM_CHUNK_SIZE;
102 /* The byte offset at which the read starts, within the start chunk. */
103 u64 start_chunk_offset = offset % WIM_CHUNK_SIZE;
105 /* The index of the chunk that contains the last byte of the read. */
106 u64 end_chunk = (offset + len - 1) / WIM_CHUNK_SIZE;
107 /* The byte offset of the last byte of the read, within the end chunk */
108 u64 end_chunk_offset = (offset + len - 1) % WIM_CHUNK_SIZE;
110 /* Number of chunks that are actually needed to read the requested part
112 u64 num_needed_chunks = end_chunk - start_chunk + 1;
114 /* If the end chunk is not the last chunk, an extra chunk entry is
115 * needed because we need to know the offset of the chunk after the last
116 * chunk read to figure out the size of the last read chunk. */
117 if (end_chunk != num_chunks - 1)
120 /* Declare the chunk table. It will only contain offsets for the chunks
121 * that are actually needed for this read. */
122 u64 chunk_offsets[num_needed_chunks];
124 /* Set the implicit offset of the first chunk if it is included in the
127 * Note: M$'s documentation includes a picture that shows the first
128 * chunk starting right after the chunk entry table, labeled as offset
129 * 0x10. However, in the actual file format, the offset is measured
130 * from the end of the chunk entry table, so the first chunk has an
132 if (start_chunk == 0)
133 chunk_offsets[0] = 0;
135 /* According to M$'s documentation, if the uncompressed size of
136 * the file is greater than 4 GB, the chunk entries are 8-byte
137 * integers. Otherwise, they are 4-byte integers. */
138 u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
141 /* Size of the full chunk table in the WIM file. */
142 u64 chunk_table_size = chunk_entry_size * num_chunk_entries;
144 /* Read the needed chunk offsets from the table in the WIM file. */
146 /* Index, in the WIM file, of the first needed entry in the
148 u64 start_table_idx = (start_chunk == 0) ? 0 : start_chunk - 1;
150 /* Number of entries we need to actually read from the chunk
151 * table (excludes the implicit first chunk). */
152 u64 num_needed_chunk_entries = (start_chunk == 0) ?
153 num_needed_chunks - 1 : num_needed_chunks;
155 /* Skip over unneeded chunk table entries. */
156 u64 file_offset_of_needed_chunk_entries = resource_offset +
157 start_table_idx * chunk_entry_size;
158 if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) {
159 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
160 "chunk table of compressed resource",
161 file_offset_of_needed_chunk_entries);
162 return WIMLIB_ERR_READ;
165 /* Number of bytes we need to read from the chunk table. */
166 size_t size = num_needed_chunk_entries * chunk_entry_size;
168 u8 chunk_tab_buf[size];
170 if (fread(chunk_tab_buf, 1, size, fp) != size)
173 /* Now fill in chunk_offsets from the entries we have read in
176 u64 *chunk_tab_p = chunk_offsets;
177 if (start_chunk == 0)
180 if (chunk_entry_size == 4) {
181 u32 *entries = (u32*)chunk_tab_buf;
182 while (num_needed_chunk_entries--)
183 *chunk_tab_p++ = to_le32(*entries++);
185 u64 *entries = (u64*)chunk_tab_buf;
186 while (num_needed_chunk_entries--)
187 *chunk_tab_p++ = to_le64(*entries++);
190 /* Done with the chunk table now. We must now seek to the first chunk
191 * that is needed for the read. */
193 u64 file_offset_of_first_needed_chunk = resource_offset +
194 chunk_table_size + chunk_offsets[0];
195 if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) {
196 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
197 "first chunk of compressed resource",
198 file_offset_of_first_needed_chunk);
199 return WIMLIB_ERR_READ;
202 /* Pointer to current position in the output buffer for uncompressed
204 u8 *out_p = (u8*)contents_ret;
206 /* Buffer for compressed data. While most compressed chunks will have a
207 * size much less than WIM_CHUNK_SIZE, WIM_CHUNK_SIZE - 1 is the maximum
208 * size in the worst-case. This assumption is valid only if chunks that
209 * happen to compress to more than the uncompressed size (i.e. a
210 * sequence of random bytes) are always stored uncompressed. But this seems
211 * to be the case in M$'s WIM files, even though it is undocumented. */
212 u8 compressed_buf[WIM_CHUNK_SIZE - 1];
215 /* Decompress all the chunks. */
216 for (u64 i = start_chunk; i <= end_chunk; i++) {
218 DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64").",
219 i, start_chunk, end_chunk);
221 /* Calculate the sizes of the compressed chunk and of the
222 * uncompressed chunk. */
223 uint compressed_chunk_size, uncompressed_chunk_size;
224 if (i != num_chunks - 1) {
225 /* All the chunks except the last one in the resource
226 * expand to WIM_CHUNK_SIZE uncompressed, and the amount
227 * of compressed data for the chunk is given by the
228 * difference of offsets in the chunk offset table. */
229 compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
230 chunk_offsets[i - start_chunk];
231 uncompressed_chunk_size = WIM_CHUNK_SIZE;
233 /* The last compressed chunk consists of the remaining
234 * bytes in the file resource, and the last uncompressed
235 * chunk has size equal to however many bytes are left-
236 * that is, the remainder of the uncompressed size when
237 * divided by WIM_CHUNK_SIZE.
239 * Note that the resource_compressed_size includes the
240 * chunk table, so the size of it must be subtracted. */
241 compressed_chunk_size = resource_compressed_size -
243 chunk_offsets[i - start_chunk];
245 uncompressed_chunk_size = resource_uncompressed_size %
248 /* If the remainder is 0, the last chunk actually
249 * uncompresses to a full WIM_CHUNK_SIZE bytes. */
250 if (uncompressed_chunk_size == 0)
251 uncompressed_chunk_size = WIM_CHUNK_SIZE;
254 DEBUG2("compressed_chunk_size = %u, "
255 "uncompressed_chunk_size = %u",
256 compressed_chunk_size, uncompressed_chunk_size);
259 /* Figure out how much of this chunk we actually need to read */
261 if (i == start_chunk)
262 start_offset = start_chunk_offset;
267 end_offset = end_chunk_offset;
269 end_offset = WIM_CHUNK_SIZE - 1;
271 u64 partial_chunk_size = end_offset + 1 - start_offset;
272 bool is_partial_chunk = (partial_chunk_size !=
273 uncompressed_chunk_size);
275 DEBUG2("start_offset = %u, end_offset = %u", start_offset,
277 DEBUG2("partial_chunk_size = %u", partial_chunk_size);
279 /* This is undocumented, but chunks can be uncompressed. This
280 * appears to always be the case when the compressed chunk size
281 * is equal to the uncompressed chunk size. */
282 if (compressed_chunk_size == uncompressed_chunk_size) {
283 /* Probably an uncompressed chunk */
285 if (start_offset != 0) {
286 if (fseeko(fp, start_offset, SEEK_CUR) != 0) {
287 ERROR_WITH_ERRNO("Uncompressed partial "
288 "chunk fseek() error");
289 return WIMLIB_ERR_READ;
292 if (fread(out_p, 1, partial_chunk_size, fp) !=
296 /* Compressed chunk */
299 /* Read the compressed data into compressed_buf. */
300 if (fread(compressed_buf, 1, compressed_chunk_size,
301 fp) != compressed_chunk_size)
304 /* For partial chunks we must buffer the uncompressed
305 * data because we don't need all of it. */
306 if (is_partial_chunk) {
307 u8 uncompressed_buf[uncompressed_chunk_size];
309 ret = decompress(compressed_buf,
310 compressed_chunk_size,
312 uncompressed_chunk_size);
314 return WIMLIB_ERR_DECOMPRESSION;
315 memcpy(out_p, uncompressed_buf + start_offset,
318 ret = decompress(compressed_buf,
319 compressed_chunk_size,
321 uncompressed_chunk_size);
323 return WIMLIB_ERR_DECOMPRESSION;
327 /* Advance the pointer into the uncompressed output data by the
328 * number of uncompressed bytes that were written. */
329 out_p += partial_chunk_size;
336 ERROR("Unexpected EOF in compressed file resource");
338 ERROR_WITH_ERRNO("Error reading compressed file resource");
339 return WIMLIB_ERR_READ;
343 * Reads uncompressed data from an open file stream.
345 int read_uncompressed_resource(FILE *fp, u64 offset, u64 len,
348 if (fseeko(fp, offset, SEEK_SET) != 0) {
349 ERROR("Failed to seek to byte %"PRIu64" of input file "
350 "to read uncompressed resource (len = %"PRIu64")",
352 return WIMLIB_ERR_READ;
354 if (fread(contents_ret, 1, len, fp) != len) {
356 ERROR("Unexpected EOF in uncompressed file resource");
358 ERROR("Failed to read %"PRIu64" bytes from "
359 "uncompressed resource at offset %"PRIu64,
362 return WIMLIB_ERR_READ;
369 * Reads a WIM resource.
371 * @fp: The FILE* for the WIM file.
372 * @resource_size: The compressed size of the resource.
373 * @resource_original_size: The uncompressed size of the resource.
374 * @resource_offset: The offset of the resource in the stream @fp.
375 * @resource_ctype: The compression type of the resource.
376 * (WIM_COMPRESSION_TYPE_*)
377 * @len: How many bytes of the resource should be read.
378 * @offset: The offset within the resource at which the read
381 * To read the whole file resource, specify offset =
382 * 0 and len = resource_original_size, or call
383 * read_full_resource().
385 * @contents_ret: An array, that must have length at least @len,
386 * into which the uncompressed contents of
387 * the file resource starting at @offset and
388 * continuing for @len bytes will be written.
390 * @return: Zero on success, nonzero on failure. Failure may be due to
391 * being unable to read the data from the WIM file at the
392 * specified length and offset, or it may be due to the
393 * compressed data (if the data is compressed) being
396 int read_resource(FILE *fp, u64 resource_size, u64 resource_original_size,
397 u64 resource_offset, int resource_ctype, u64 len,
398 u64 offset, void *contents_ret)
400 if (resource_ctype == WIM_COMPRESSION_TYPE_NONE) {
401 if (resource_size != resource_original_size) {
402 ERROR("Resource with original size %"PRIu64" bytes is "
403 "marked as uncompressed, but its actual size is "
405 resource_original_size, resource_size);
406 return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
408 return read_uncompressed_resource(fp,
409 resource_offset + offset,
412 return read_compressed_resource(fp, resource_size,
413 resource_original_size, resource_offset,
414 resource_ctype, len, offset, contents_ret);
420 * Extracts the first @size bytes file resource specified by @entry to the open
421 * file @fd. Returns nonzero on error.
424 * This function is somewhat redundant with uncompress_resource(). The
425 * main difference is that this function writes to a file descriptor using
426 * low-level calls to write() rather than to a FILE* with fwrite(); also this
427 * function allows only up to @size bytes to be extracted.
429 int extract_resource_to_fd(WIMStruct *w, const struct resource_entry *entry,
434 u8 buf[min(size, WIM_CHUNK_SIZE)];
442 num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
444 res_ctype = wim_resource_compression_type(w, entry);
446 for (i = 0; i < num_chunks; i++) {
447 if (i == num_chunks - 1) {
448 n = size % WIM_CHUNK_SIZE;
454 ret = read_resource(w->fp, entry->size, entry->original_size,
455 entry->offset, res_ctype, n, offset, buf);
459 if (full_write(fd, buf, n) != n)
460 return WIMLIB_ERR_WRITE;
467 * Copies the file resource specified by the lookup table entry @lte from the
468 * input WIM, pointed to by the fp field of the WIMStruct, to the output WIM,
469 * pointed to by the out_fp field of the WIMStruct.
471 * The output_resource_entry, out_refcnt, and part_number fields of @lte are
474 * Metadata resources are not copied (they are handled elsewhere for joining and
477 int copy_resource(struct lookup_table_entry *lte, void *w)
479 if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
480 !((WIMStruct*)w)->write_metadata) {
484 FILE *in_fp = ((WIMStruct*)w)->fp;
485 FILE *out_fp = ((WIMStruct*)w)->out_fp;
487 u64 size = lte->resource_entry.size;
488 u64 offset = lte->resource_entry.offset;
489 off_t new_offset = ftello(out_fp);
491 if (new_offset == -1)
492 return WIMLIB_ERR_WRITE;
494 ret = copy_between_files(in_fp, offset, out_fp, size);
498 memcpy(<e->output_resource_entry, <e->resource_entry,
499 sizeof(struct resource_entry));
501 lte->output_resource_entry.offset = new_offset;
502 lte->out_refcnt = lte->refcnt;
503 lte->part_number = ((WIMStruct*)w)->hdr.part_number;
507 /* Reads the contents of a struct resource_entry, as represented in the on-disk
508 * format, from the memory pointed to by @p, and fills in the fields of @entry.
509 * A pointer to the byte after the memory read at @p is returned. */
510 const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry)
515 p = get_u56(p, &size);
516 p = get_u8(p, &flags);
518 entry->flags = flags;
519 p = get_u64(p, &entry->offset);
520 p = get_u64(p, &entry->original_size);
524 /* Copies the struct resource_entry @entry to the memory pointed to by @p in the
525 * on-disk format. A pointer to the byte after the memory written at @p is
527 u8 *put_resource_entry(u8 *p, const struct resource_entry *entry)
529 p = put_u56(p, entry->size);
530 p = put_u8(p, entry->flags);
531 p = put_u64(p, entry->offset);
532 p = put_u64(p, entry->original_size);
536 /* Given the compression type for the WIM file as a whole as the flags field of
537 * a resource entry, returns the compression type for that resource entry. */
538 int resource_compression_type(int wim_ctype, int reshdr_flags)
540 if (wim_ctype != WIM_COMPRESSION_TYPE_NONE &&
541 (reshdr_flags & WIM_RESHDR_FLAG_COMPRESSED))
544 return WIM_COMPRESSION_TYPE_NONE;
550 * Copies bytes between two file streams.
552 * Copies @len bytes from @in_fp to @out_fp, at the current position in @out_fp,
553 * and at an offset of @in_offset in @in_fp.
555 int copy_between_files(FILE *in_fp, off_t in_offset, FILE *out_fp, size_t len)
560 if (fseeko(in_fp, in_offset, SEEK_SET) != 0) {
561 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" of "
562 "input file", in_offset);
563 return WIMLIB_ERR_READ;
565 /* To reduce memory usage and improve speed, read and write BUFFER_SIZE
566 * bytes at a time. */
568 n = min(len, BUFFER_SIZE);
569 if (fread(buf, 1, n, in_fp) != n) {
571 ERROR("Unexpected EOF when copying data "
574 ERROR_WITH_ERRNO("Error copying data between "
577 return WIMLIB_ERR_READ;
580 if (fwrite(buf, 1, n, out_fp) != n) {
581 ERROR_WITH_ERRNO("Error copying data between files");
582 return WIMLIB_ERR_WRITE;
591 * Uncompresses a WIM file resource and writes it uncompressed to a file stream.
593 * @in_fp: The file stream that contains the file resource.
594 * @size: The size of the resource in the input file.
595 * @original_size: The original (uncompressed) size of the resource.
596 * @offset: The offset of the start of the resource in @in.
597 * @input_ctype: The compression type of the resource in @in.
598 * @out_fp: The file stream to write the file resource to.
600 static int uncompress_resource(FILE *in_fp, u64 size, u64 original_size,
601 off_t offset, int input_ctype, FILE *out_fp)
604 u8 buf[WIM_CHUNK_SIZE];
605 /* Determine how many compressed chunks the file is divided into. */
608 u64 uncompressed_offset;
609 u64 uncompressed_chunk_size;
611 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
613 for (i = 0; i < num_chunks; i++) {
615 uncompressed_offset = i * WIM_CHUNK_SIZE;
616 uncompressed_chunk_size = min(WIM_CHUNK_SIZE, original_size -
617 uncompressed_offset);
619 ret = read_resource(in_fp, size, original_size, offset,
620 input_ctype, uncompressed_chunk_size,
621 uncompressed_offset, buf);
625 if (fwrite(buf, 1, uncompressed_chunk_size, out_fp) !=
626 uncompressed_chunk_size)
628 ERROR_WITH_ERRNO("Failed to write file resource");
629 return WIMLIB_ERR_WRITE;
636 * Transfers a file resource between two files, writing it compressed. The file
637 * resource in the input file may be either compressed or uncompressed.
638 * Alternatively, the input resource may be in-memory, but it must be
641 * @in_fp: The file stream that contains the file resource. Ignored
642 * if uncompressed_resource != NULL.
643 * @uncompressed_resource: If this pointer is not NULL, it points to an
644 * array of @original_size bytes that are
645 * the uncompressed input resource.
646 * @size: The size of the resource in the input file.
647 * @original_size: The original (uncompressed) size of the resource.
648 * @offset: The offset of the start of the resource in @in. Ignored
649 * if uncompressed_resource != NULL.
650 * @input_ctype: The compression type of the resource in @in. Ignored if
651 * uncompressed_resource != NULL.
652 * @out_fp: The file stream to write the file resource to.
653 * @output_type: The compression type to use when writing the resource to
655 * @new_size_ret: A location into which the new compressed size of the file
656 * resource in returned.
658 static int recompress_resource(FILE *in_fp, const u8 *uncompressed_resource,
659 u64 size, u64 original_size,
660 off_t offset, int input_ctype, FILE *out_fp,
661 int output_ctype, u64 *new_size_ret)
664 int (*compress)(const void *, uint, void *, uint *);
665 if (output_ctype == WIM_COMPRESSION_TYPE_LZX)
666 compress = lzx_compress;
668 compress = xpress_compress;
670 u8 uncompressed_buf[WIM_CHUNK_SIZE];
671 u8 compressed_buf[WIM_CHUNK_SIZE - 1];
673 /* Determine how many compressed chunks the file needs to be divided
675 u64 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
677 u64 num_chunk_entries = num_chunks - 1;
679 /* Size of the chunk entries--- 8 bytes for files over 4GB, otherwise 4
681 uint chunk_entry_size = (original_size >= (u64)1 << 32) ? 8 : 4;
683 /* Array in which to construct the chunk offset table. */
684 u64 chunk_offsets[num_chunk_entries];
686 /* Offset of the start of the chunk table in the output file. */
687 off_t chunk_tab_offset = ftello(out_fp);
689 if (chunk_tab_offset == -1) {
690 ERROR_WITH_ERRNO("Failed to get offset of output file");
691 return WIMLIB_ERR_WRITE;
694 /* Total size of the chunk table (as written to the file) */
695 u64 chunk_tab_size = chunk_entry_size * num_chunk_entries;
697 /* Reserve space for the chunk table. */
698 if (fwrite(chunk_offsets, 1, chunk_tab_size, out_fp) !=
701 ERROR_WITH_ERRNO("Failed to write chunk offset table");
702 return WIMLIB_ERR_WRITE;
705 /* Read each chunk of the file, compress it, write it to the output
706 * file, and update th chunk offset table. */
707 u64 cur_chunk_offset = 0;
708 for (u64 i = 0; i < num_chunks; i++) {
710 u64 uncompressed_offset = i * WIM_CHUNK_SIZE;
711 u64 uncompressed_chunk_size = min(WIM_CHUNK_SIZE,
712 original_size - uncompressed_offset);
714 const u8 *uncompressed_p;
715 if (uncompressed_resource != NULL) {
716 uncompressed_p = uncompressed_resource +
720 /* Read chunk i of the file into uncompressed_buf. */
721 ret = read_resource(in_fp, size, original_size, offset,
723 uncompressed_chunk_size,
728 uncompressed_p = uncompressed_buf;
732 chunk_offsets[i - 1] = cur_chunk_offset;
736 ret = compress(uncompressed_p, uncompressed_chunk_size,
737 compressed_buf, &compressed_len);
739 /* if compress() returned nonzero, the compressed chunk would
740 * have been at least as large as the uncompressed chunk. In
741 * this situation, the WIM format requires that the uncompressed
742 * chunk be written instead. */
743 const u8 *buf_to_write;
746 buf_to_write = compressed_buf;
747 len_to_write = compressed_len;
749 buf_to_write = uncompressed_p;
750 len_to_write = uncompressed_chunk_size;
753 if (fwrite(buf_to_write, 1, len_to_write, out_fp) !=
756 ERROR_WITH_ERRNO("Failed to write compressed "
758 return WIMLIB_ERR_WRITE;
760 cur_chunk_offset += len_to_write;
763 /* The chunk offset after the last chunk, plus the size of the chunk
764 * table, gives the total compressed size of the resource. */
765 *new_size_ret = cur_chunk_offset + chunk_tab_size;
767 /* Now that all entries of the chunk table are determined, rewind the
768 * stream to where the chunk table was, and write it back out. */
770 if (fseeko(out_fp, chunk_tab_offset, SEEK_SET) != 0) {
771 ERROR_WITH_ERRNO("Failed to seek to beginning of chunk table");
772 return WIMLIB_ERR_READ;
775 if (chunk_entry_size == 8) {
776 array_to_le64(chunk_offsets, num_chunk_entries);
778 for (u64 i = 0; i < num_chunk_entries; i++)
779 ((u32*)chunk_offsets)[i] = to_le32(chunk_offsets[i]);
781 if (fwrite(chunk_offsets, 1, chunk_tab_size, out_fp) != chunk_tab_size)
783 ERROR_WITH_ERRNO("Failed to write chunk table");
784 return WIMLIB_ERR_WRITE;
787 if (fseeko(out_fp, 0, SEEK_END) != 0) {
788 ERROR_WITH_ERRNO("Failed to seek to end of output file");
789 return WIMLIB_ERR_WRITE;
795 int write_resource_from_memory(const u8 resource[], int out_ctype,
796 u64 resource_original_size, FILE *out_fp,
797 u64 *resource_size_ret)
799 if (out_ctype == WIM_COMPRESSION_TYPE_NONE) {
800 if (fwrite(resource, 1, resource_original_size, out_fp) !=
801 resource_original_size)
803 ERROR_WITH_ERRNO("Failed to write resource of length "
804 "%"PRIu64, resource_original_size);
805 return WIMLIB_ERR_WRITE;
807 *resource_size_ret = resource_original_size;
810 return recompress_resource(NULL, resource,
811 resource_original_size,
812 resource_original_size, 0, 0, out_fp,
813 out_ctype, resource_size_ret);
819 * Transfers a file resource from a FILE* opened for reading to a FILE* opened
820 * for writing, possibly changing the compression type.
822 * @in_fp: The FILE* that contains the file resource.
823 * @size: The (compressed) size of the file resource.
824 * @original_size: The uncompressed size of the file resource.
825 * @offset: The offset of the file resource in the input file.
826 * @input_ctype: The compression type of the file resource in the input
828 * @out_fp: The FILE* for the output file. The file resource is
829 * written at the current position of @out.
830 * @output_ctype: The compression type to which the file resource will be
832 * @output_res_entry: A pointer to a resource entry that, upon successful
833 * return of this function, will have the size,
834 * original size, offset, and flags fields filled
835 * in for the file resource written to the output
838 static int transfer_file_resource(FILE *in_fp, u64 size, u64 original_size,
839 off_t offset, int input_ctype, FILE *out_fp,
841 struct resource_entry *output_res_entry)
845 /* Handle zero-length files */
846 if (original_size == 0) {
847 memset(output_res_entry, 0, sizeof(*output_res_entry));
851 /* Get current offset in the output file. */
852 off_t out_offset = ftello(out_fp);
853 if (out_offset == -1) {
854 ERROR_WITH_ERRNO("Failed to get output position");
855 return WIMLIB_ERR_WRITE;
857 output_res_entry->offset = (u64)out_offset;
859 if (output_ctype == input_ctype) {
860 /* The same compression types; simply copy the resource. */
862 ret = copy_between_files(in_fp, offset, out_fp, size);
865 output_res_entry->size = size;
867 /* Different compression types. */
869 if (output_ctype == WIM_COMPRESSION_TYPE_NONE) {
870 /* Uncompress a compressed file resource */
871 ret = uncompress_resource(in_fp, size,
872 original_size, offset,
873 input_ctype, out_fp);
876 output_res_entry->size = original_size;
879 /* Compress an uncompressed file resource, or compress a
880 * compressed file resource using a different
881 * compression type */
882 ret = recompress_resource(in_fp, NULL, size,
884 offset, input_ctype, out_fp,
885 output_ctype, &new_size);
888 output_res_entry->size = new_size;
893 output_res_entry->original_size = original_size;
894 if (output_ctype == WIM_COMPRESSION_TYPE_NONE)
895 output_res_entry->flags = 0;
897 output_res_entry->flags = WIM_RESHDR_FLAG_COMPRESSED;
902 * Reads the metadata metadata resource from the WIM file. The metadata
903 * resource consists of the security data, followed by the directory entry for
904 * the root directory, followed by all the other directory entries in the
905 * filesystem. The subdir_offset field of each directory entry gives the start
906 * of its child entries from the beginning of the metadata resource. An
907 * end-of-directory is signaled by a directory entry of length '0', really of
908 * length 8, because that's how long the 'length' field is.
910 * @fp: The FILE* for the input WIM file.
911 * @wim_ctype: The compression type of the WIM file.
912 * @imd: Pointer to the image metadata structure. Its
913 * `lookup_table_entry' member specifies the lookup table entry for
914 * the metadata resource. The rest of the image metadata entry
915 * will be filled in by this function.
917 * @return: Zero on success, nonzero on failure.
919 int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd)
925 const struct resource_entry *res_entry;
926 struct dentry *dentry;
927 struct wim_security_data *sd;
928 struct link_group_table *lgt;
930 res_entry = &imd->metadata_lte->resource_entry;
932 DEBUG("Reading metadata resource: length = %"PRIu64", "
933 "offset = %"PRIu64"",
934 res_entry->original_size, res_entry->offset);
936 if (res_entry->original_size < 8) {
937 ERROR("Expected at least 8 bytes for the metadata resource");
938 return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
941 /* Allocate memory for the uncompressed metadata resource. */
942 buf = MALLOC(res_entry->original_size);
945 ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
946 "metadata resource", res_entry->original_size);
947 return WIMLIB_ERR_NOMEM;
950 /* Determine the compression type of the metadata resource. */
951 ctype = resource_compression_type(wim_ctype, res_entry->flags);
953 /* Read the metadata resource into memory. (It may be compressed.) */
954 ret = read_full_resource(fp, res_entry->size,
955 res_entry->original_size, res_entry->offset,
960 DEBUG("Finished reading metadata resource into memory.");
962 /* The root directory entry starts after security data, on an 8-byte
965 * The security data starts with a 4-byte integer giving its total
968 /* Read the security data into a wim_security_data structure. */
969 ret = read_security_data(buf, res_entry->original_size, &sd);
973 dentry = MALLOC(sizeof(struct dentry));
975 ERROR("Failed to allocate %zu bytes for root dentry",
976 sizeof(struct dentry));
977 ret = WIMLIB_ERR_NOMEM;
978 goto out_free_security_data;
981 get_u32(buf, &dentry_offset);
982 if (dentry_offset == 0)
984 dentry_offset = (dentry_offset + 7) & ~7;
986 ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry);
987 /* This is the root dentry, so set its pointers correctly. */
988 dentry->parent = dentry;
989 dentry->next = dentry;
990 dentry->prev = dentry;
992 goto out_free_dentry_tree;
994 DEBUG("Reading dentry tree");
995 /* Now read the entire directory entry tree. */
996 ret = read_dentry_tree(buf, res_entry->original_size, dentry);
998 goto out_free_dentry_tree;
1000 DEBUG("Calculating dentry full paths");
1001 /* Calculate the full paths in the dentry tree. */
1002 ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
1004 goto out_free_dentry_tree;
1006 DEBUG("Building link group table");
1007 /* Build hash table that maps hard link group IDs to dentry sets */
1008 lgt = new_link_group_table(9001);
1010 goto out_free_dentry_tree;
1011 ret = for_dentry_in_tree(dentry, link_group_table_insert, lgt);
1015 DEBUG("Freeing duplicate ADS entries in link group table");
1016 ret = link_groups_free_duplicate_data(lgt);
1019 DEBUG("Done reading image metadata");
1022 imd->security_data = sd;
1023 imd->root_dentry = dentry;
1026 free_link_group_table(lgt);
1027 out_free_dentry_tree:
1028 free_dentry_tree(dentry, NULL, false);
1029 out_free_security_data:
1030 free_security_data(sd);
1036 /* Write the metadata resource for the current image. */
1037 int write_metadata_resource(WIMStruct *w)
1044 struct dentry *root;
1045 struct lookup_table_entry *lte;
1046 struct resource_entry *res_entry;
1047 off_t metadata_offset;
1048 u64 metadata_original_size;
1049 u64 metadata_compressed_size;
1051 u8 hash[WIM_HASH_SIZE];
1053 DEBUG("Writing metadata resource for image %d", w->current_image);
1056 root = wim_root_dentry(w);
1057 metadata_ctype = wimlib_get_compression_type(w);
1058 metadata_offset = ftello(out);
1059 if (metadata_offset == -1)
1060 return WIMLIB_ERR_WRITE;
1062 struct wim_security_data *sd = wim_security_data(w);
1064 subdir_offset = sd->total_length + root->length + 8;
1066 subdir_offset = 8 + root->length + 8;
1067 calculate_subdir_offsets(root, &subdir_offset);
1068 metadata_original_size = subdir_offset;
1069 buf = MALLOC(metadata_original_size);
1071 ERROR("Failed to allocate %"PRIu64" bytes for "
1072 "metadata resource", metadata_original_size);
1073 return WIMLIB_ERR_NOMEM;
1076 p = write_security_data(sd, buf);
1078 DEBUG("Writing dentry tree.");
1079 p = write_dentry_tree(root, p);
1081 /* Like file resources, the lookup table entry for a metadata resource
1082 * uses for the hash code a SHA1 message digest of its uncompressed
1084 sha1_buffer(buf, metadata_original_size, hash);
1086 ret = write_resource_from_memory(buf,
1088 metadata_original_size,
1090 &metadata_compressed_size);
1095 DEBUG("Updating metadata lookup table entry (size %zu)",
1096 metadata_original_size);
1098 /* Update the lookup table entry, including the hash and output resource
1099 * entry fields, for this image's metadata resource. */
1100 lte = wim_metadata_lookup_table_entry(w);
1101 res_entry = <e->output_resource_entry;
1103 if (memcmp(hash, lte->hash, WIM_HASH_SIZE) != 0) {
1104 lookup_table_unlink(w->lookup_table, lte);
1105 memcpy(lte->hash, hash, WIM_HASH_SIZE);
1106 lookup_table_insert(w->lookup_table, lte);
1108 res_entry->original_size = metadata_original_size;
1109 res_entry->offset = metadata_offset;
1110 res_entry->size = metadata_compressed_size;
1111 res_entry->flags = WIM_RESHDR_FLAG_METADATA;
1112 if (metadata_ctype != WIM_COMPRESSION_TYPE_NONE)
1113 res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
1117 static int write_file_resource(WIMStruct *w, const u8 hash[])
1119 /* Get the lookup entry for the file resource. */
1120 struct lookup_table_entry *lte;
1122 lte = __lookup_resource(w->lookup_table, hash);
1126 /* No need to write file resources twice. */
1127 if (++lte->out_refcnt != 1)
1130 /* do not write empty resources */
1131 if (lte->resource_entry.original_size == 0)
1134 int out_wim_ctype = wimlib_get_compression_type(w);
1135 struct resource_entry *output_res_entry = <e->output_resource_entry;
1138 FILE *out_fp = w->out_fp;
1141 /* Figure out if we can read the resource from the WIM file, or
1142 * if we have to read it from the filesystem outside, or if it's a
1143 * symbolic link with the data already in memory pointed to by a field
1144 * of the lookup table entry. */
1145 if (lte->is_symlink) {
1146 off_t offset = ftello(w->out_fp);
1150 ERROR_WITH_ERRNO("Could not get position in output "
1152 return WIMLIB_ERR_WRITE;
1155 wimlib_assert(lte->symlink_buf);
1157 len = lte->resource_entry.original_size;
1159 ret = recompress_resource(NULL, lte->symlink_buf, len, len, 0,
1160 0, out_fp, out_wim_ctype, &new_size);
1161 output_res_entry->size = new_size;
1162 output_res_entry->original_size = len;
1163 output_res_entry->offset = offset;
1164 output_res_entry->flags = (out_wim_ctype == WIM_COMPRESSION_TYPE_NONE)
1165 ? 0 : WIM_RESHDR_FLAG_COMPRESSED;
1166 } else if (lte->file_on_disk) {
1168 /* Read from disk (uncompressed) */
1170 len = lte->resource_entry.original_size;
1172 in_fp = fopen(lte->file_on_disk, "rb");
1174 ERROR_WITH_ERRNO("Failed to open the file `%s'",
1176 return WIMLIB_ERR_OPEN;
1179 ret = transfer_file_resource(in_fp, len, len, 0,
1180 WIM_COMPRESSION_TYPE_NONE, out_fp,
1181 out_wim_ctype, output_res_entry);
1186 /* Read from input WIM (possibly compressed) */
1188 /* It may be a different WIM file, in the case of
1189 * exporting images from one WIM file to another */
1190 if (lte->other_wim_fp) {
1191 /* Different WIM file. */
1192 in_fp = lte->other_wim_fp;
1193 in_wim_ctype = lte->other_wim_ctype;
1195 /* Same WIM file. */
1197 in_wim_ctype = out_wim_ctype;
1199 int input_res_ctype = resource_compression_type(
1201 lte->resource_entry.flags);
1203 ret = transfer_file_resource(in_fp,
1204 lte->resource_entry.size,
1205 lte->resource_entry.original_size,
1206 lte->resource_entry.offset,
1216 * Writes a dentry's resources to the output file.
1218 * @dentry: The dentry for the file resource.
1219 * @wim_p: A pointer to the WIMStruct. The fields of interest to this
1220 * function are the input and output file streams and the lookup
1221 * table, and the alternate data streams.
1223 * @return zero on success, nonzero on failure.
1225 int write_dentry_resources(struct dentry *dentry, void *wim_p)
1227 WIMStruct *w = wim_p;
1230 /* Directories don't need file resources. */
1231 if (dentry_is_directory(dentry))
1234 ret = write_file_resource(w, dentry->hash);
1237 for (u16 i = 0; i < dentry->num_ads; i++) {
1238 ret = write_file_resource(w, dentry->ads_entries[i].hash);