4 * Read uncompressed and compressed metadata and file resources.
6 * Copyright (C) 2010 Carl Thijssen
7 * Copyright (C) 2012 Eric Biggers
9 * wimlib - Library for working with WIM files
11 * This library is free software; you can redistribute it and/or modify it under
12 * the terms of the GNU Lesser General Public License as published by the Free
13 * Software Foundation; either version 2.1 of the License, or (at your option) any
16 * This library is distributed in the hope that it will be useful, but WITHOUT ANY
17 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
18 * PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public License along
21 * with this library; if not, write to the Free Software Foundation, Inc., 59
22 * Temple Place, Suite 330, Boston, MA 02111-1307 USA
25 #include "wimlib_internal.h"
26 #include "lookup_table.h"
37 * Reads all or part of a compressed resource into an in-memory buffer.
39 * @fp: The FILE* for the WIM file.
40 * @resource_compressed_size: The compressed size of the resource.
41 * @resource_uncompressed_size: The uncompressed size of the resource.
42 * @resource_offset: The offset of the start of the resource from
43 * the start of the stream @fp.
44 * @resource_ctype: The compression type of the resource.
45 * @len: The number of bytes of uncompressed data to read from
47 * @offset: The offset of the bytes to read within the uncompressed
49 * @contents_len: An array into which the uncompressed data is written.
50 * It must be at least @len bytes long.
52 * Returns zero on success, nonzero on failure.
54 static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
55 u64 resource_uncompressed_size,
56 u64 resource_offset, int resource_ctype,
57 u64 len, u64 offset, u8 contents_ret[])
60 DEBUG2("comp size = %"PRIu64", "
61 "uncomp size = %"PRIu64", "
62 "res offset = %"PRIu64"\n",
63 resource_compressed_size,
64 resource_uncompressed_size,
66 DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"\n",
67 wimlib_get_compression_type_string(resource_ctype),
73 int (*decompress)(const void *, uint, void *, uint);
74 /* Set the appropriate decompress function. */
75 if (resource_ctype == WIM_COMPRESSION_TYPE_LZX)
76 decompress = lzx_decompress;
78 decompress = xpress_decompress;
80 /* The structure of a compressed resource consists of a table of chunk
81 * offsets followed by the chunks themselves. Each chunk consists of
82 * compressed data, and there is one chunk for each WIM_CHUNK_SIZE =
83 * 32768 bytes of the uncompressed file, with the last chunk having any
86 * The chunk offsets are measured relative to the end of the chunk
87 * table. The first chunk is omitted from the table in the WIM file
88 * because its offset is implicitly given by the fact that it directly
89 * follows the chunk table and therefore must have an offset of 0.
92 /* Calculate how many chunks the resource conists of in its entirety. */
93 u64 num_chunks = (resource_uncompressed_size + WIM_CHUNK_SIZE - 1) /
95 /* As mentioned, the first chunk has no entry in the chunk table. */
96 u64 num_chunk_entries = num_chunks - 1;
99 /* The index of the chunk that the read starts at. */
100 u64 start_chunk = offset / WIM_CHUNK_SIZE;
101 /* The byte offset at which the read starts, within the start chunk. */
102 u64 start_chunk_offset = offset % WIM_CHUNK_SIZE;
104 /* The index of the chunk that contains the last byte of the read. */
105 u64 end_chunk = (offset + len - 1) / WIM_CHUNK_SIZE;
106 /* The byte offset of the last byte of the read, within the end chunk */
107 u64 end_chunk_offset = (offset + len - 1) % WIM_CHUNK_SIZE;
109 /* Number of chunks that are actually needed to read the requested part
111 u64 num_needed_chunks = end_chunk - start_chunk + 1;
113 /* If the end chunk is not the last chunk, an extra chunk entry is
114 * needed because we need to know the offset of the chunk after the last
115 * chunk read to figure out the size of the last read chunk. */
116 if (end_chunk != num_chunks - 1)
119 /* Declare the chunk table. It will only contain offsets for the chunks
120 * that are actually needed for this read. */
121 u64 chunk_offsets[num_needed_chunks];
123 /* Set the implicit offset of the first chunk if it is included in the
126 * Note: M$'s documentation includes a picture that shows the first
127 * chunk starting right after the chunk entry table, labeled as offset
128 * 0x10. However, in the actual file format, the offset is measured
129 * from the end of the chunk entry table, so the first chunk has an
131 if (start_chunk == 0)
132 chunk_offsets[0] = 0;
134 /* According to M$'s documentation, if the uncompressed size of
135 * the file is greater than 4 GB, the chunk entries are 8-byte
136 * integers. Otherwise, they are 4-byte integers. */
137 u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
140 /* Size of the full chunk table in the WIM file. */
141 u64 chunk_table_size = chunk_entry_size * num_chunk_entries;
143 /* Read the needed chunk offsets from the table in the WIM file. */
145 /* Index, in the WIM file, of the first needed entry in the
147 u64 start_table_idx = (start_chunk == 0) ? 0 : start_chunk - 1;
149 /* Number of entries we need to actually read from the chunk
150 * table (excludes the implicit first chunk). */
151 u64 num_needed_chunk_entries = (start_chunk == 0) ?
152 num_needed_chunks - 1 : num_needed_chunks;
154 /* Skip over unneeded chunk table entries. */
155 u64 file_offset_of_needed_chunk_entries = resource_offset +
156 start_table_idx * chunk_entry_size;
157 if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) {
158 ERROR("Failed to seek to byte %"PRIu64" "
159 "to read chunk table of compressed "
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("Failed to seek to byte %"PRIu64" "
197 "to read first chunk of compressed "
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")\n",
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, uncompressed_chunk_size = %u\n",
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\n", start_offset,
277 DEBUG2("partial_chunk_size = %u\n", 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("Uncompressed partial chunk "
288 "fseek() error: %m\n");
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 DEBUG2("out_p = %p\n");
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\n");
339 ERROR("Error reading compressed file resource: %m\n");
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 "
352 "(len = %"PRIu64")!\n", offset, len);
353 return WIMLIB_ERR_READ;
355 if (fread(contents_ret, 1, len, fp) != len) {
357 ERROR("Unexpected EOF in uncompressed file resource!\n");
359 ERROR("Failed to read %"PRIu64" bytes from "
360 "uncompressed resource at offset "
361 "%"PRIu64"\n", len, offset);
363 return WIMLIB_ERR_READ;
370 * Reads a WIM resource.
372 * @fp: The FILE* for the WIM file.
373 * @resource_size: The compressed size of the resource.
374 * @resource_original_size: The uncompressed size of the resource.
375 * @resource_offset: The offset of the resource in the stream @fp.
376 * @resource_ctype: The compression type of the resource.
377 * (WIM_COMPRESSION_TYPE_*)
378 * @len: How many bytes of the resource should be read.
379 * @offset: The offset within the resource at which the read
382 * To read the whole file resource, specify offset =
383 * 0 and len = resource_original_size, or call
384 * read_full_resource().
386 * @contents_ret: An array, that must have length at least @len,
387 * into which the uncompressed contents of
388 * the file resource starting at @offset and
389 * continuing for @len bytes will be written.
391 * @return: Zero on success, nonzero on failure. Failure may be due to
392 * being unable to read the data from the WIM file at the
393 * specified length and offset, or it may be due to the
394 * compressed data (if the data is compressed) being
397 int read_resource(FILE *fp, u64 resource_size, u64 resource_original_size,
398 u64 resource_offset, int resource_ctype, u64 len,
399 u64 offset, void *contents_ret)
401 if (resource_ctype == WIM_COMPRESSION_TYPE_NONE) {
402 if (resource_size != resource_original_size) {
403 ERROR("Resource with original size %"PRIu64" "
404 "bytes is marked as uncompressed, \n",
405 resource_original_size);
406 ERROR(" but its actual size is %"PRIu64" "
409 return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
411 return read_uncompressed_resource(fp,
412 resource_offset + offset,
415 return read_compressed_resource(fp, resource_size,
416 resource_original_size, resource_offset,
417 resource_ctype, len, offset, contents_ret);
423 * Extracts the first @size bytes file resource specified by @entry to the open
424 * file @fd. Returns nonzero on error.
427 * This function is somewhat redundant with uncompress_resource(). The
428 * main difference is that this function writes to a file descriptor using
429 * low-level calls to write() rather than to a FILE* with fwrite(); also this
430 * function allows only up to @size bytes to be extracted.
432 int extract_resource_to_fd(WIMStruct *w, const struct resource_entry *entry,
437 u8 buf[min(size, WIM_CHUNK_SIZE)];
445 num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
447 res_ctype = wim_resource_compression_type(w, entry);
449 for (i = 0; i < num_chunks; i++) {
450 if (i == num_chunks - 1) {
451 n = size % WIM_CHUNK_SIZE;
457 ret = read_resource(w->fp, entry->size, entry->original_size,
458 entry->offset, res_ctype, n, offset, buf);
462 if (full_write(fd, buf, n) != n)
463 return WIMLIB_ERR_WRITE;
470 * Copies the file resource specified by the lookup table entry @lte from the
471 * input WIM, pointed to by the fp field of the WIMStruct, to the output WIM,
472 * pointed to by the out_fp field of the WIMStruct.
474 * The output_resource_entry, out_refcnt, and part_number fields of @lte are
477 * Metadata resources are not copied (they are handled elsewhere for joining and
480 int copy_resource(struct lookup_table_entry *lte, void *w)
482 if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
483 !((WIMStruct*)w)->write_metadata) {
487 FILE *in_fp = ((WIMStruct*)w)->fp;
488 FILE *out_fp = ((WIMStruct*)w)->out_fp;
490 u64 size = lte->resource_entry.size;
491 u64 offset = lte->resource_entry.offset;
492 off_t new_offset = ftello(out_fp);
494 if (new_offset == -1)
495 return WIMLIB_ERR_WRITE;
497 ret = copy_between_files(in_fp, offset, out_fp, size);
501 memcpy(<e->output_resource_entry, <e->resource_entry,
502 sizeof(struct resource_entry));
504 lte->output_resource_entry.offset = new_offset;
505 lte->out_refcnt = lte->refcnt;
506 lte->part_number = ((WIMStruct*)w)->hdr.part_number;
510 /* Reads the contents of a struct resource_entry, as represented in the on-disk
511 * format, from the memory pointed to by @p, and fills in the fields of @entry.
512 * A pointer to the byte after the memory read at @p is returned. */
513 const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry)
518 p = get_u56(p, &size);
519 p = get_u8(p, &flags);
521 entry->flags = flags;
522 p = get_u64(p, &entry->offset);
523 p = get_u64(p, &entry->original_size);
527 /* Copies the struct resource_entry @entry to the memory pointed to by @p in the
528 * on-disk format. A pointer to the byte after the memory written at @p is
530 u8 *put_resource_entry(u8 *p, const struct resource_entry *entry)
532 p = put_u56(p, entry->size);
533 p = put_u8(p, entry->flags);
534 p = put_u64(p, entry->offset);
535 p = put_u64(p, entry->original_size);
539 /* Given the compression type for the WIM file as a whole as the flags field of
540 * a resource entry, returns the compression type for that resource entry. */
541 int resource_compression_type(int wim_ctype, int reshdr_flags)
543 if (wim_ctype == WIM_COMPRESSION_TYPE_NONE) {
544 return WIM_COMPRESSION_TYPE_NONE;
546 if (reshdr_flags & WIM_RESHDR_FLAG_COMPRESSED)
549 return WIM_COMPRESSION_TYPE_NONE;
556 * Copies bytes between two file streams.
558 * Copies @len bytes from @in to @out, at the current position in @out, and at
559 * an offset of @in_offset in @in.
561 int copy_between_files(FILE *in, off_t in_offset, FILE *out, size_t len)
566 if (fseeko(in, in_offset, SEEK_SET) != 0) {
567 ERROR("Failed to seek to byte %"PRIu64" of input file: %m\n",
569 return WIMLIB_ERR_READ;
571 /* To reduce memory usage and improve speed, read and write BUFFER_SIZE
572 * bytes at a time. */
574 n = min(len, BUFFER_SIZE);
575 if (fread(buf, 1, n, in) != n) {
577 ERROR("Unexpected EOF when copying data "
580 ERROR("Error copying data between files: %m\n");
582 return WIMLIB_ERR_READ;
585 if (fwrite(buf, 1, n, out) != n) {
586 ERROR("Error copying data between files: %m\n");
587 return WIMLIB_ERR_WRITE;
596 * Uncompresses a WIM file resource and writes it uncompressed to a file stream.
598 * @in: The file stream that contains the file resource.
599 * @size: The size of the resource in the input file.
600 * @original_size: The original (uncompressed) size of the resource.
601 * @offset: The offset of the start of the resource in @in.
602 * @input_ctype: The compression type of the resource in @in.
603 * @out: The file stream to write the file resource to.
605 static int uncompress_resource(FILE *in, u64 size, u64 original_size,
606 off_t offset, int input_ctype, FILE *out)
609 u8 buf[WIM_CHUNK_SIZE];
610 /* Determine how many compressed chunks the file is divided into. */
613 u64 uncompressed_offset;
614 u64 uncompressed_chunk_size;
616 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
618 for (i = 0; i < num_chunks; i++) {
620 uncompressed_offset = i * WIM_CHUNK_SIZE;
621 uncompressed_chunk_size = min(WIM_CHUNK_SIZE,
622 original_size - uncompressed_offset);
624 ret = read_resource(in, size, original_size, offset, input_ctype,
625 uncompressed_chunk_size,
626 uncompressed_offset, buf);
630 if (fwrite(buf, 1, uncompressed_chunk_size, out) !=
631 uncompressed_chunk_size) {
632 ERROR("Failed to write file resource: %m\n");
633 return WIMLIB_ERR_WRITE;
640 * Transfers a file resource between two files, writing it compressed. The file
641 * resource in the input file may be either compressed or uncompressed.
642 * Alternatively, the input resource may be in-memory, but it must be
645 * @in: The file stream that contains the file resource. Ignored
646 * if uncompressed_resource != NULL.
647 * @uncompressed_resource: If this pointer is not NULL, it points to an
648 * array of @original_size bytes that are
649 * the uncompressed input resource.
650 * @size: The size of the resource in the input file.
651 * @original_size: The original (uncompressed) size of the resource.
652 * @offset: The offset of the start of the resource in @in. Ignored
653 * if uncompressed_resource != NULL.
654 * @input_ctype: The compression type of the resource in @in. Ignored if
655 * uncompressed_resource != NULL.
656 * @out: The file stream to write the file resource to.
657 * @output_type: The compression type to use when writing the resource to
659 * @new_size_ret: A location into which the new compressed size of the file
660 * resource in returned.
662 static int recompress_resource(FILE *in, const u8 *uncompressed_resource,
663 u64 size, u64 original_size,
664 off_t offset, int input_ctype, FILE *out,
665 int output_ctype, u64 *new_size_ret)
668 int (*compress)(const void *, uint, void *, uint *);
669 if (output_ctype == WIM_COMPRESSION_TYPE_LZX)
670 compress = lzx_compress;
672 compress = xpress_compress;
674 u8 uncompressed_buf[WIM_CHUNK_SIZE];
675 u8 compressed_buf[WIM_CHUNK_SIZE - 1];
677 /* Determine how many compressed chunks the file needs to be divided
679 u64 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
681 u64 num_chunk_entries = num_chunks - 1;
683 /* Size of the chunk entries--- 8 bytes for files over 4GB, otherwise 4
685 uint chunk_entry_size = (original_size >= (u64)1 << 32) ? 8 : 4;
687 /* Array in which to construct the chunk offset table. */
688 u64 chunk_offsets[num_chunk_entries];
690 /* Offset of the start of the chunk table in the output file. */
691 off_t chunk_tab_offset = ftello(out);
693 /* Total size of the chunk table (as written to the file) */
694 u64 chunk_tab_size = chunk_entry_size * num_chunk_entries;
696 /* Reserve space for the chunk table. */
697 if (fwrite(chunk_offsets, 1, chunk_tab_size, out) != chunk_tab_size) {
698 ERROR("Failed to write chunk offset table: %m\n");
699 return WIMLIB_ERR_WRITE;
702 /* Read each chunk of the file, compress it, write it to the output
703 * file, and update th chunk offset table. */
704 u64 cur_chunk_offset = 0;
705 for (u64 i = 0; i < num_chunks; i++) {
707 u64 uncompressed_offset = i * WIM_CHUNK_SIZE;
708 u64 uncompressed_chunk_size = min(WIM_CHUNK_SIZE,
709 original_size - uncompressed_offset);
711 const u8 *uncompressed_p;
712 if (uncompressed_resource != NULL) {
713 uncompressed_p = uncompressed_resource +
717 /* Read chunk i of the file into uncompressed_buf. */
718 ret = read_resource(in, size, original_size, offset, input_ctype,
719 uncompressed_chunk_size,
724 uncompressed_p = uncompressed_buf;
728 chunk_offsets[i - 1] = cur_chunk_offset;
732 ret = compress(uncompressed_p, uncompressed_chunk_size,
733 compressed_buf, &compressed_len);
735 /* if compress() returned nonzero, the compressed chunk would
736 * have been at least as large as the uncompressed chunk. In
737 * this situation, the WIM format requires that the uncompressed
738 * chunk be written instead. */
739 const u8 *buf_to_write;
742 buf_to_write = compressed_buf;
743 len_to_write = compressed_len;
745 buf_to_write = uncompressed_p;
746 len_to_write = uncompressed_chunk_size;
749 if (fwrite(buf_to_write, 1, len_to_write, out) != len_to_write) {
750 ERROR("Failed to write compressed file resource: %m\n");
751 return WIMLIB_ERR_WRITE;
753 cur_chunk_offset += len_to_write;
756 /* The chunk offset after the last chunk, plus the size of the chunk
757 * table, gives the total compressed size of the resource. */
758 *new_size_ret = cur_chunk_offset + chunk_tab_size;
760 /* Now that all entries of the chunk table are determined, rewind the
761 * stream to where the chunk table was, and write it back out. */
763 if (fseeko(out, chunk_tab_offset, SEEK_SET) != 0) {
764 ERROR("Failed to seek to beginning of chunk table: %m\n");
765 return WIMLIB_ERR_READ;
768 if (chunk_entry_size == 8) {
769 array_to_le64(chunk_offsets, num_chunk_entries);
771 if (fwrite(chunk_offsets, 1, chunk_tab_size, out) !=
773 ERROR("Failed to write chunk table: %m\n");
774 return WIMLIB_ERR_WRITE;
777 u32 chunk_entries_small[num_chunk_entries];
778 for (u64 i = 0; i < num_chunk_entries; i++)
779 chunk_entries_small[i] = to_le32(chunk_offsets[i]);
780 if (fwrite(chunk_entries_small, 1, chunk_tab_size, out) !=
782 ERROR("Failed to write chunk table: %m\n");
783 return WIMLIB_ERR_WRITE;
787 if (fseeko(out, 0, SEEK_END) != 0) {
788 ERROR("Failed to seek to end of output file: %m\n");
789 return WIMLIB_ERR_WRITE;
795 int write_resource_from_memory(const u8 resource[], int out_ctype,
796 u64 resource_original_size, FILE *out,
797 u64 *resource_size_ret)
799 if (out_ctype == WIM_COMPRESSION_TYPE_NONE) {
800 if (fwrite(resource, 1, resource_original_size, out) !=
801 resource_original_size) {
802 ERROR("Failed to write resource of length "
804 resource_original_size);
805 return WIMLIB_ERR_WRITE;
807 *resource_size_ret = resource_original_size;
810 return recompress_resource(NULL, resource, resource_original_size,
811 resource_original_size, 0, 0, out, out_ctype,
818 * Transfers a file resource from a FILE* opened for reading to a FILE* opened
819 * for writing, possibly changing the compression type.
821 * @in: The FILE* that contains the file resource.
822 * @size: The (compressed) size of the file resource.
823 * @original_size: The uncompressed size of the file resource.
824 * @offset: The offset of the file resource in the input file.
825 * @input_ctype: The compression type of the file resource in the input
827 * @out: The FILE* for the output file. The file resource is
828 * written at the current position of @out.
829 * @output_ctype: The compression type to which the file resource will be
831 * @output_res_entry: A pointer to a resource entry that, upon successful
832 * return of this function, will have the size,
833 * original size, offset, and flags fields filled
834 * in for the file resource written to the output
837 static int transfer_file_resource(FILE *in, u64 size, u64 original_size,
838 off_t offset, int input_ctype, FILE *out,
840 struct resource_entry *output_res_entry)
844 /* Handle zero-length files */
845 if (original_size == 0) {
846 memset(output_res_entry, 0, sizeof(*output_res_entry));
850 /* Get current offset in the output file. */
851 output_res_entry->offset = ftello(out);
852 if (output_res_entry->offset == -1) {
853 ERROR("Failed to get output position: %m\n");
854 return WIMLIB_ERR_WRITE;
857 if (output_ctype == input_ctype) {
858 /* The same compression types; simply copy the resource. */
860 ret = copy_between_files(in, offset, out, size);
863 output_res_entry->size = size;
865 /* Different compression types. */
867 if (output_ctype == WIM_COMPRESSION_TYPE_NONE) {
868 /* Uncompress a compressed file resource */
869 ret = uncompress_resource(in, size,
870 original_size, offset,
874 output_res_entry->size = original_size;
877 /* Compress an uncompressed file resource, or compress a
878 * compressed file resource using a different
879 * compression type */
880 ret = recompress_resource(in, NULL, size, original_size,
881 offset, input_ctype, out,
882 output_ctype, &new_size);
885 output_res_entry->size = new_size;
890 output_res_entry->original_size = original_size;
891 if (output_ctype == WIM_COMPRESSION_TYPE_NONE)
892 output_res_entry->flags = 0;
894 output_res_entry->flags = WIM_RESHDR_FLAG_COMPRESSED;
899 * Reads the metadata metadata resource from the WIM file. The metadata
900 * resource consists of the security data, followed by the directory entry for
901 * the root directory, followed by all the other directory entries in the
902 * filesystem. The subdir_offset field of each directory entry gives the start
903 * of its child entries from the beginning of the metadata resource. An
904 * end-of-directory is signaled by a directory entry of length '0', really of
905 * length 8, because that's how long the 'length' field is.
907 * @fp: The FILE* for the input WIM file.
908 * @res_entry: The resource entry for the metadata resource (a.k.a the metadata
910 * @wim_ctype: The compression type of the WIM file.
911 * @root_dentry_p: A pointer to a pointer to a struct dentry structure into which the
912 * root dentry is allocated and returned.
914 * @return: True on success, false on failure.
916 int read_metadata_resource(FILE *fp, const struct resource_entry *res_entry,
917 int wim_ctype, struct image_metadata *image_metadata)
923 struct dentry *dentry = NULL;
925 DEBUG("Reading metadata resource: length = %lu, offset = %lu\n",
926 res_entry->original_size, res_entry->offset);
928 if (res_entry->original_size < 8) {
929 ERROR("Expected at least 8 bytes for the metadata "
931 return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
934 /* Allocate memory for the uncompressed metadata resource. */
935 buf = MALLOC(res_entry->original_size);
938 ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
939 "metadata resource!\n",
940 res_entry->original_size);
941 return WIMLIB_ERR_NOMEM;
944 /* Determine the compression type of the metadata resource. */
945 ctype = resource_compression_type(wim_ctype, res_entry->flags);
947 /* Read the metadata resource into memory. (It may be compressed.) */
948 ret = read_full_resource(fp, res_entry->size,
949 res_entry->original_size, res_entry->offset,
954 DEBUG("Finished reading metadata resource into memory.\n");
957 dentry = MALLOC(sizeof(struct dentry));
959 ERROR("Failed to allocate %zu bytes for root dentry!\n",
960 sizeof(struct dentry));
961 ret = WIMLIB_ERR_NOMEM;
965 /* Read the root directory entry starts after security data, on an
966 * 8-byte aligned address.
968 * The security data starts with a 4-byte integer giving its total
971 /* Read the security data into a WIMSecurityData structure. */
972 #ifdef ENABLE_SECURITY_DATA
973 ret = read_security_data(buf, res_entry->original_size,
974 &image_metadata->security_data);
978 get_u32(buf, &dentry_offset);
979 if (dentry_offset == 0)
981 dentry_offset += (8 - dentry_offset % 8) % 8;
983 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 /* Now read the entire directory entry tree. */
993 ret = read_dentry_tree(buf, res_entry->original_size, dentry);
997 /* Calculate the full paths in the dentry tree. */
998 ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
1002 image_metadata->root_dentry = dentry;
1006 free_dentry_tree(dentry, NULL, false);
1012 /* Write the metadata resource for the current image. */
1013 int write_metadata_resource(WIMStruct *w)
1019 off_t subdir_offset;
1020 struct dentry *root;
1021 struct lookup_table_entry *lte;
1022 struct resource_entry *res_entry;
1023 off_t metadata_offset;
1024 u64 metadata_original_size;
1025 u64 metadata_compressed_size;
1027 u8 hash[WIM_HASH_SIZE];
1029 DEBUG("Writing metadata resource for image %u\n", w->current_image);
1032 root = wim_root_dentry(w);
1033 metadata_ctype = wimlib_get_compression_type(w);
1034 metadata_offset = ftello(out);
1035 if (metadata_offset == -1)
1036 return WIMLIB_ERR_WRITE;
1038 #ifdef ENABLE_SECURITY_DATA
1039 struct wim_security_data *sd = wim_security_data(w);
1041 subdir_offset = sd->total_length + root->length + 8;
1044 subdir_offset = 8 + root->length + 8;
1045 calculate_subdir_offsets(root, &subdir_offset);
1046 metadata_original_size = subdir_offset;
1047 buf = MALLOC(metadata_original_size);
1049 ERROR("Failed to allocate %"PRIu64" bytes for "
1050 "metadata resource\n", metadata_original_size);
1051 return WIMLIB_ERR_NOMEM;
1053 #ifdef ENABLE_SECURITY_DATA
1054 /* Write the security data. */
1055 p = write_security_data(sd, buf);
1057 p = put_u32(buf, 8); /* Total length of security data. */
1058 p = put_u32(p, 0); /* Number of security data entries. */
1061 DEBUG("Writing dentry tree.\n");
1062 p = write_dentry_tree(root, p);
1064 /* Like file resources, the lookup table entry for a metadata resource
1065 * uses for the hash code a SHA1 message digest of its uncompressed
1067 sha1_buffer(buf, metadata_original_size, hash);
1069 ret = write_resource_from_memory(buf,
1071 metadata_original_size,
1073 &metadata_compressed_size);
1078 /* Update the lookup table entry, including the hash and output resource
1079 * entry fields, for this image's metadata resource. */
1080 lte = wim_metadata_lookup_table_entry(w);
1081 res_entry = <e->output_resource_entry;
1083 if (memcmp(hash, lte->hash, WIM_HASH_SIZE) != 0) {
1084 lookup_table_unlink(w->lookup_table, lte);
1085 memcpy(lte->hash, hash, WIM_HASH_SIZE);
1086 lookup_table_insert(w->lookup_table, lte);
1088 res_entry->original_size = metadata_original_size;
1089 res_entry->offset = metadata_offset;
1090 res_entry->size = metadata_compressed_size;
1091 res_entry->flags = WIM_RESHDR_FLAG_METADATA;
1092 if (metadata_ctype != WIM_COMPRESSION_TYPE_NONE)
1093 res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
1098 * Writes a file resource to the output file.
1100 * @dentry: The dentry for the file resource.
1101 * @wim_p: A pointer to the WIMStruct. The fields of interest to this
1102 * function are the input and output file streams and the lookup table.
1103 * @return zero on success, nonzero on failure.
1105 int write_file_resource(struct dentry *dentry, void *wim_p)
1110 struct lookup_table_entry *lte;
1113 struct resource_entry *output_res_entry;
1120 /* Directories don't need file resources. */
1121 if (dentry_is_directory(dentry))
1124 /* Get the lookup entry for the file resource. */
1125 lte = wim_lookup_resource(w, dentry);
1129 /* No need to write file resources twice. (This indicates file
1130 * resources that are part of a hard link set.) */
1131 if (++lte->out_refcnt != 1)
1134 out_wim_ctype = wimlib_get_compression_type(w);
1135 output_res_entry = <e->output_resource_entry;
1137 /* do not write empty resources */
1138 if (lte->resource_entry.original_size == 0)
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. */
1143 if (lte->file_on_disk) {
1145 /* Read from disk (uncompressed) */
1147 len = lte->resource_entry.original_size;
1149 in = fopen(lte->file_on_disk, "rb");
1151 ERROR("Failed to open the file `%s': %m\n",
1153 return WIMLIB_ERR_OPEN;
1157 puts(lte->file_on_disk);
1159 ret = transfer_file_resource(in, len, len, 0,
1160 WIM_COMPRESSION_TYPE_NONE, out,
1161 out_wim_ctype, output_res_entry);
1165 /* Read from input WIM (possibly compressed) */
1167 /* It may be a different WIM file, in the case of
1168 * exporting images from one WIM file to another */
1169 if (lte->other_wim_fp) {
1170 /* Different WIM file. */
1171 in = lte->other_wim_fp;
1172 in_wim_ctype = lte->other_wim_ctype;
1174 /* Same WIM file. */
1176 in_wim_ctype = out_wim_ctype;
1178 int input_res_ctype = resource_compression_type(
1180 lte->resource_entry.flags);
1182 ret = transfer_file_resource(in,
1183 lte->resource_entry.size,
1184 lte->resource_entry.original_size,
1185 lte->resource_entry.offset,