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"
39 * Reads all or part of a compressed resource into an in-memory buffer.
41 * @fp: The FILE* for the WIM file.
42 * @resource_compressed_size: The compressed size of the resource.
43 * @resource_uncompressed_size: The uncompressed size of the resource.
44 * @resource_offset: The offset of the start of the resource from
45 * the start of the stream @fp.
46 * @resource_ctype: The compression type of the resource.
47 * @len: The number of bytes of uncompressed data to read from
49 * @offset: The offset of the bytes to read within the uncompressed
51 * @contents_len: An array into which the uncompressed data is written.
52 * It must be at least @len bytes long.
54 * Returns zero on success, nonzero on failure.
56 static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
57 u64 resource_uncompressed_size,
58 u64 resource_offset, int resource_ctype,
59 u64 len, u64 offset, u8 contents_ret[])
62 DEBUG2("comp size = %"PRIu64", "
63 "uncomp size = %"PRIu64", "
64 "res offset = %"PRIu64"\n",
65 resource_compressed_size,
66 resource_uncompressed_size,
68 DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"\n",
69 wimlib_get_compression_type_string(resource_ctype),
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("Failed to seek to byte %"PRIu64" "
161 "to read chunk table of compressed "
163 file_offset_of_needed_chunk_entries);
164 return WIMLIB_ERR_READ;
167 /* Number of bytes we need to read from the chunk table. */
168 size_t size = num_needed_chunk_entries * chunk_entry_size;
170 u8 chunk_tab_buf[size];
172 if (fread(chunk_tab_buf, 1, size, fp) != size)
175 /* Now fill in chunk_offsets from the entries we have read in
178 u64 *chunk_tab_p = chunk_offsets;
179 if (start_chunk == 0)
182 if (chunk_entry_size == 4) {
183 u32 *entries = (u32*)chunk_tab_buf;
184 while (num_needed_chunk_entries--)
185 *chunk_tab_p++ = to_le32(*entries++);
187 u64 *entries = (u64*)chunk_tab_buf;
188 while (num_needed_chunk_entries--)
189 *chunk_tab_p++ = to_le64(*entries++);
192 /* Done with the chunk table now. We must now seek to the first chunk
193 * that is needed for the read. */
195 u64 file_offset_of_first_needed_chunk = resource_offset +
196 chunk_table_size + chunk_offsets[0];
197 if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) {
198 ERROR("Failed to seek to byte %"PRIu64" "
199 "to read first chunk of compressed "
201 file_offset_of_first_needed_chunk);
202 return WIMLIB_ERR_READ;
205 /* Pointer to current position in the output buffer for uncompressed
207 u8 *out_p = (u8*)contents_ret;
209 /* Buffer for compressed data. While most compressed chunks will have a
210 * size much less than WIM_CHUNK_SIZE, WIM_CHUNK_SIZE - 1 is the maximum
211 * size in the worst-case. This assumption is valid only if chunks that
212 * happen to compress to more than the uncompressed size (i.e. a
213 * sequence of random bytes) are always stored uncompressed. But this seems
214 * to be the case in M$'s WIM files, even though it is undocumented. */
215 u8 compressed_buf[WIM_CHUNK_SIZE - 1];
218 /* Decompress all the chunks. */
219 for (u64 i = start_chunk; i <= end_chunk; i++) {
221 DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64")\n",
222 i, start_chunk, end_chunk);
224 /* Calculate the sizes of the compressed chunk and of the
225 * uncompressed chunk. */
226 uint compressed_chunk_size, uncompressed_chunk_size;
227 if (i != num_chunks - 1) {
228 /* All the chunks except the last one in the resource
229 * expand to WIM_CHUNK_SIZE uncompressed, and the amount
230 * of compressed data for the chunk is given by the
231 * difference of offsets in the chunk offset table. */
232 compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
233 chunk_offsets[i - start_chunk];
234 uncompressed_chunk_size = WIM_CHUNK_SIZE;
236 /* The last compressed chunk consists of the remaining
237 * bytes in the file resource, and the last uncompressed
238 * chunk has size equal to however many bytes are left-
239 * that is, the remainder of the uncompressed size when
240 * divided by WIM_CHUNK_SIZE.
242 * Note that the resource_compressed_size includes the
243 * chunk table, so the size of it must be subtracted. */
244 compressed_chunk_size = resource_compressed_size -
246 chunk_offsets[i - start_chunk];
248 uncompressed_chunk_size = resource_uncompressed_size %
251 /* If the remainder is 0, the last chunk actually
252 * uncompresses to a full WIM_CHUNK_SIZE bytes. */
253 if (uncompressed_chunk_size == 0)
254 uncompressed_chunk_size = WIM_CHUNK_SIZE;
257 DEBUG2("compressed_chunk_size = %u, uncompressed_chunk_size = %u\n",
258 compressed_chunk_size, uncompressed_chunk_size);
261 /* Figure out how much of this chunk we actually need to read */
263 if (i == start_chunk)
264 start_offset = start_chunk_offset;
269 end_offset = end_chunk_offset;
271 end_offset = WIM_CHUNK_SIZE - 1;
273 u64 partial_chunk_size = end_offset + 1 - start_offset;
274 bool is_partial_chunk = (partial_chunk_size !=
275 uncompressed_chunk_size);
277 DEBUG2("start_offset = %u, end_offset = %u\n", start_offset,
279 DEBUG2("partial_chunk_size = %u\n", partial_chunk_size);
281 /* This is undocumented, but chunks can be uncompressed. This
282 * appears to always be the case when the compressed chunk size
283 * is equal to the uncompressed chunk size. */
284 if (compressed_chunk_size == uncompressed_chunk_size) {
285 /* Probably an uncompressed chunk */
287 if (start_offset != 0) {
288 if (fseeko(fp, start_offset, SEEK_CUR) != 0) {
289 ERROR("Uncompressed partial chunk "
290 "fseek() error: %m\n");
291 return WIMLIB_ERR_READ;
294 if (fread(out_p, 1, partial_chunk_size, fp) !=
298 /* Compressed chunk */
301 /* Read the compressed data into compressed_buf. */
302 if (fread(compressed_buf, 1, compressed_chunk_size,
303 fp) != compressed_chunk_size)
306 /* For partial chunks we must buffer the uncompressed
307 * data because we don't need all of it. */
308 if (is_partial_chunk) {
309 u8 uncompressed_buf[uncompressed_chunk_size];
311 ret = decompress(compressed_buf,
312 compressed_chunk_size,
314 uncompressed_chunk_size);
316 return WIMLIB_ERR_DECOMPRESSION;
317 memcpy(out_p, uncompressed_buf + start_offset,
320 DEBUG2("out_p = %p\n");
321 ret = decompress(compressed_buf,
322 compressed_chunk_size,
324 uncompressed_chunk_size);
326 return WIMLIB_ERR_DECOMPRESSION;
330 /* Advance the pointer into the uncompressed output data by the
331 * number of uncompressed bytes that were written. */
332 out_p += partial_chunk_size;
339 ERROR("Unexpected EOF in compressed file resource\n");
341 ERROR("Error reading compressed file resource: %m\n");
342 return WIMLIB_ERR_READ;
346 * Reads uncompressed data from an open file stream.
348 int read_uncompressed_resource(FILE *fp, u64 offset, u64 len,
351 if (fseeko(fp, offset, SEEK_SET) != 0) {
352 ERROR("Failed to seek to byte %"PRIu64" of input file "
353 "to read uncompressed resource "
354 "(len = %"PRIu64")!\n", offset, len);
355 return WIMLIB_ERR_READ;
357 if (fread(contents_ret, 1, len, fp) != len) {
359 ERROR("Unexpected EOF in uncompressed file resource!\n");
361 ERROR("Failed to read %"PRIu64" bytes from "
362 "uncompressed resource at offset "
363 "%"PRIu64"\n", len, offset);
365 return WIMLIB_ERR_READ;
372 * Reads a WIM resource.
374 * @fp: The FILE* for the WIM file.
375 * @resource_size: The compressed size of the resource.
376 * @resource_original_size: The uncompressed size of the resource.
377 * @resource_offset: The offset of the resource in the stream @fp.
378 * @resource_ctype: The compression type of the resource.
379 * (WIM_COMPRESSION_TYPE_*)
380 * @len: How many bytes of the resource should be read.
381 * @offset: The offset within the resource at which the read
384 * To read the whole file resource, specify offset =
385 * 0 and len = resource_original_size, or call
386 * read_full_resource().
388 * @contents_ret: An array, that must have length at least @len,
389 * into which the uncompressed contents of
390 * the file resource starting at @offset and
391 * continuing for @len bytes will be written.
393 * @return: Zero on success, nonzero on failure. Failure may be due to
394 * being unable to read the data from the WIM file at the
395 * specified length and offset, or it may be due to the
396 * compressed data (if the data is compressed) being
399 int read_resource(FILE *fp, u64 resource_size, u64 resource_original_size,
400 u64 resource_offset, int resource_ctype, u64 len,
401 u64 offset, void *contents_ret)
403 if (resource_ctype == WIM_COMPRESSION_TYPE_NONE) {
404 if (resource_size != resource_original_size) {
405 ERROR("Resource with original size %"PRIu64" "
406 "bytes is marked as uncompressed, \n",
407 resource_original_size);
408 ERROR(" but its actual size is %"PRIu64" "
411 return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
413 return read_uncompressed_resource(fp,
414 resource_offset + offset,
417 return read_compressed_resource(fp, resource_size,
418 resource_original_size, resource_offset,
419 resource_ctype, len, offset, contents_ret);
425 * Extracts the first @size bytes file resource specified by @entry to the open
426 * file @fd. Returns nonzero on error.
429 * This function is somewhat redundant with uncompress_resource(). The
430 * main difference is that this function writes to a file descriptor using
431 * low-level calls to write() rather than to a FILE* with fwrite(); also this
432 * function allows only up to @size bytes to be extracted.
434 int extract_resource_to_fd(WIMStruct *w, const struct resource_entry *entry,
439 u8 buf[min(size, WIM_CHUNK_SIZE)];
447 num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
449 res_ctype = wim_resource_compression_type(w, entry);
451 for (i = 0; i < num_chunks; i++) {
452 if (i == num_chunks - 1) {
453 n = size % WIM_CHUNK_SIZE;
459 ret = read_resource(w->fp, entry->size, entry->original_size,
460 entry->offset, res_ctype, n, offset, buf);
464 if (full_write(fd, buf, n) != n)
465 return WIMLIB_ERR_WRITE;
472 * Copies the file resource specified by the lookup table entry @lte from the
473 * input WIM, pointed to by the fp field of the WIMStruct, to the output WIM,
474 * pointed to by the out_fp field of the WIMStruct.
476 * The output_resource_entry, out_refcnt, and part_number fields of @lte are
479 * Metadata resources are not copied (they are handled elsewhere for joining and
482 int copy_resource(struct lookup_table_entry *lte, void *w)
484 if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
485 !((WIMStruct*)w)->write_metadata) {
489 FILE *in_fp = ((WIMStruct*)w)->fp;
490 FILE *out_fp = ((WIMStruct*)w)->out_fp;
492 u64 size = lte->resource_entry.size;
493 u64 offset = lte->resource_entry.offset;
494 off_t new_offset = ftello(out_fp);
496 if (new_offset == -1)
497 return WIMLIB_ERR_WRITE;
499 ret = copy_between_files(in_fp, offset, out_fp, size);
503 memcpy(<e->output_resource_entry, <e->resource_entry,
504 sizeof(struct resource_entry));
506 lte->output_resource_entry.offset = new_offset;
507 lte->out_refcnt = lte->refcnt;
508 lte->part_number = ((WIMStruct*)w)->hdr.part_number;
512 /* Reads the contents of a struct resource_entry, as represented in the on-disk
513 * format, from the memory pointed to by @p, and fills in the fields of @entry.
514 * A pointer to the byte after the memory read at @p is returned. */
515 const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry)
520 p = get_u56(p, &size);
521 p = get_u8(p, &flags);
523 entry->flags = flags;
524 p = get_u64(p, &entry->offset);
525 p = get_u64(p, &entry->original_size);
529 /* Copies the struct resource_entry @entry to the memory pointed to by @p in the
530 * on-disk format. A pointer to the byte after the memory written at @p is
532 u8 *put_resource_entry(u8 *p, const struct resource_entry *entry)
534 p = put_u56(p, entry->size);
535 p = put_u8(p, entry->flags);
536 p = put_u64(p, entry->offset);
537 p = put_u64(p, entry->original_size);
541 /* Given the compression type for the WIM file as a whole as the flags field of
542 * a resource entry, returns the compression type for that resource entry. */
543 int resource_compression_type(int wim_ctype, int reshdr_flags)
545 if (wim_ctype == WIM_COMPRESSION_TYPE_NONE) {
546 return WIM_COMPRESSION_TYPE_NONE;
548 if (reshdr_flags & WIM_RESHDR_FLAG_COMPRESSED)
551 return WIM_COMPRESSION_TYPE_NONE;
558 * Copies bytes between two file streams.
560 * Copies @len bytes from @in to @out, at the current position in @out, and at
561 * an offset of @in_offset in @in.
563 int copy_between_files(FILE *in, off_t in_offset, FILE *out, size_t len)
568 if (fseeko(in, in_offset, SEEK_SET) != 0) {
569 ERROR("Failed to seek to byte %"PRIu64" of input file: %m\n",
571 return WIMLIB_ERR_READ;
573 /* To reduce memory usage and improve speed, read and write BUFFER_SIZE
574 * bytes at a time. */
576 n = min(len, BUFFER_SIZE);
577 if (fread(buf, 1, n, in) != n) {
579 ERROR("Unexpected EOF when copying data "
582 ERROR("Error copying data between files: %m\n");
584 return WIMLIB_ERR_READ;
587 if (fwrite(buf, 1, n, out) != n) {
588 ERROR("Error copying data between files: %m\n");
589 return WIMLIB_ERR_WRITE;
598 * Uncompresses a WIM file resource and writes it uncompressed to a file stream.
600 * @in: The file stream that contains the file resource.
601 * @size: The size of the resource in the input file.
602 * @original_size: The original (uncompressed) size of the resource.
603 * @offset: The offset of the start of the resource in @in.
604 * @input_ctype: The compression type of the resource in @in.
605 * @out: The file stream to write the file resource to.
607 static int uncompress_resource(FILE *in, u64 size, u64 original_size,
608 off_t offset, int input_ctype, FILE *out)
611 u8 buf[WIM_CHUNK_SIZE];
612 /* Determine how many compressed chunks the file is divided into. */
615 u64 uncompressed_offset;
616 u64 uncompressed_chunk_size;
618 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
620 for (i = 0; i < num_chunks; i++) {
622 uncompressed_offset = i * WIM_CHUNK_SIZE;
623 uncompressed_chunk_size = min(WIM_CHUNK_SIZE,
624 original_size - uncompressed_offset);
626 ret = read_resource(in, size, original_size, offset, input_ctype,
627 uncompressed_chunk_size,
628 uncompressed_offset, buf);
632 if (fwrite(buf, 1, uncompressed_chunk_size, out) !=
633 uncompressed_chunk_size) {
634 ERROR("Failed to write file resource: %m\n");
635 return WIMLIB_ERR_WRITE;
642 * Transfers a file resource between two files, writing it compressed. The file
643 * resource in the input file may be either compressed or uncompressed.
644 * Alternatively, the input resource may be in-memory, but it must be
647 * @in: The file stream that contains the file resource. Ignored
648 * if uncompressed_resource != NULL.
649 * @uncompressed_resource: If this pointer is not NULL, it points to an
650 * array of @original_size bytes that are
651 * the uncompressed input resource.
652 * @size: The size of the resource in the input file.
653 * @original_size: The original (uncompressed) size of the resource.
654 * @offset: The offset of the start of the resource in @in. Ignored
655 * if uncompressed_resource != NULL.
656 * @input_ctype: The compression type of the resource in @in. Ignored if
657 * uncompressed_resource != NULL.
658 * @out: The file stream to write the file resource to.
659 * @output_type: The compression type to use when writing the resource to
661 * @new_size_ret: A location into which the new compressed size of the file
662 * resource in returned.
664 static int recompress_resource(FILE *in, const u8 *uncompressed_resource,
665 u64 size, u64 original_size,
666 off_t offset, int input_ctype, FILE *out,
667 int output_ctype, u64 *new_size_ret)
670 int (*compress)(const void *, uint, void *, uint *);
671 if (output_ctype == WIM_COMPRESSION_TYPE_LZX)
672 compress = lzx_compress;
674 compress = xpress_compress;
676 u8 uncompressed_buf[WIM_CHUNK_SIZE];
677 u8 compressed_buf[WIM_CHUNK_SIZE - 1];
679 /* Determine how many compressed chunks the file needs to be divided
681 u64 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
683 u64 num_chunk_entries = num_chunks - 1;
685 /* Size of the chunk entries--- 8 bytes for files over 4GB, otherwise 4
687 uint chunk_entry_size = (original_size >= (u64)1 << 32) ? 8 : 4;
689 /* Array in which to construct the chunk offset table. */
690 u64 chunk_offsets[num_chunk_entries];
692 /* Offset of the start of the chunk table in the output file. */
693 off_t chunk_tab_offset = ftello(out);
695 /* Total size of the chunk table (as written to the file) */
696 u64 chunk_tab_size = chunk_entry_size * num_chunk_entries;
698 /* Reserve space for the chunk table. */
699 if (fwrite(chunk_offsets, 1, chunk_tab_size, out) != chunk_tab_size) {
700 ERROR("Failed to write chunk offset table: %m\n");
701 return WIMLIB_ERR_WRITE;
704 /* Read each chunk of the file, compress it, write it to the output
705 * file, and update th chunk offset table. */
706 u64 cur_chunk_offset = 0;
707 for (u64 i = 0; i < num_chunks; i++) {
709 u64 uncompressed_offset = i * WIM_CHUNK_SIZE;
710 u64 uncompressed_chunk_size = min(WIM_CHUNK_SIZE,
711 original_size - uncompressed_offset);
713 const u8 *uncompressed_p;
714 if (uncompressed_resource != NULL) {
715 uncompressed_p = uncompressed_resource +
719 /* Read chunk i of the file into uncompressed_buf. */
720 ret = read_resource(in, size, original_size, offset, input_ctype,
721 uncompressed_chunk_size,
726 uncompressed_p = uncompressed_buf;
730 chunk_offsets[i - 1] = cur_chunk_offset;
734 ret = compress(uncompressed_p, uncompressed_chunk_size,
735 compressed_buf, &compressed_len);
737 /* if compress() returned nonzero, the compressed chunk would
738 * have been at least as large as the uncompressed chunk. In
739 * this situation, the WIM format requires that the uncompressed
740 * chunk be written instead. */
741 const u8 *buf_to_write;
744 buf_to_write = compressed_buf;
745 len_to_write = compressed_len;
747 buf_to_write = uncompressed_p;
748 len_to_write = uncompressed_chunk_size;
751 if (fwrite(buf_to_write, 1, len_to_write, out) != len_to_write) {
752 ERROR("Failed to write compressed file resource: %m\n");
753 return WIMLIB_ERR_WRITE;
755 cur_chunk_offset += len_to_write;
758 /* The chunk offset after the last chunk, plus the size of the chunk
759 * table, gives the total compressed size of the resource. */
760 *new_size_ret = cur_chunk_offset + chunk_tab_size;
762 /* Now that all entries of the chunk table are determined, rewind the
763 * stream to where the chunk table was, and write it back out. */
765 if (fseeko(out, chunk_tab_offset, SEEK_SET) != 0) {
766 ERROR("Failed to seek to beginning of chunk table: %m\n");
767 return WIMLIB_ERR_READ;
770 if (chunk_entry_size == 8) {
771 array_to_le64(chunk_offsets, num_chunk_entries);
773 if (fwrite(chunk_offsets, 1, chunk_tab_size, out) !=
775 ERROR("Failed to write chunk table: %m\n");
776 return WIMLIB_ERR_WRITE;
779 u32 chunk_entries_small[num_chunk_entries];
780 for (u64 i = 0; i < num_chunk_entries; i++)
781 chunk_entries_small[i] = to_le32(chunk_offsets[i]);
782 if (fwrite(chunk_entries_small, 1, chunk_tab_size, out) !=
784 ERROR("Failed to write chunk table: %m\n");
785 return WIMLIB_ERR_WRITE;
789 if (fseeko(out, 0, SEEK_END) != 0) {
790 ERROR("Failed to seek to end of output file: %m\n");
791 return WIMLIB_ERR_WRITE;
797 int write_resource_from_memory(const u8 resource[], int out_ctype,
798 u64 resource_original_size, FILE *out,
799 u64 *resource_size_ret)
801 if (out_ctype == WIM_COMPRESSION_TYPE_NONE) {
802 if (fwrite(resource, 1, resource_original_size, out) !=
803 resource_original_size) {
804 ERROR("Failed to write resource of length "
806 resource_original_size);
807 return WIMLIB_ERR_WRITE;
809 *resource_size_ret = resource_original_size;
812 return recompress_resource(NULL, resource, resource_original_size,
813 resource_original_size, 0, 0, out, out_ctype,
820 * Transfers a file resource from a FILE* opened for reading to a FILE* opened
821 * for writing, possibly changing the compression type.
823 * @in: The FILE* that contains the file resource.
824 * @size: The (compressed) size of the file resource.
825 * @original_size: The uncompressed size of the file resource.
826 * @offset: The offset of the file resource in the input file.
827 * @input_ctype: The compression type of the file resource in the input
829 * @out: The FILE* for the output file. The file resource is
830 * written at the current position of @out.
831 * @output_ctype: The compression type to which the file resource will be
833 * @output_res_entry: A pointer to a resource entry that, upon successful
834 * return of this function, will have the size,
835 * original size, offset, and flags fields filled
836 * in for the file resource written to the output
839 static int transfer_file_resource(FILE *in, u64 size, u64 original_size,
840 off_t offset, int input_ctype, FILE *out,
842 struct resource_entry *output_res_entry)
846 /* Handle zero-length files */
847 if (original_size == 0) {
848 memset(output_res_entry, 0, sizeof(*output_res_entry));
852 /* Get current offset in the output file. */
853 output_res_entry->offset = ftello(out);
854 if (output_res_entry->offset == -1) {
855 ERROR("Failed to get output position: %m\n");
856 return WIMLIB_ERR_WRITE;
859 if (output_ctype == input_ctype) {
860 /* The same compression types; simply copy the resource. */
862 ret = copy_between_files(in, offset, out, 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, size,
872 original_size, offset,
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, NULL, size, original_size,
883 offset, input_ctype, out,
884 output_ctype, &new_size);
887 output_res_entry->size = new_size;
892 output_res_entry->original_size = original_size;
893 if (output_ctype == WIM_COMPRESSION_TYPE_NONE)
894 output_res_entry->flags = 0;
896 output_res_entry->flags = WIM_RESHDR_FLAG_COMPRESSED;
901 * Reads the metadata metadata resource from the WIM file. The metadata
902 * resource consists of the security data, followed by the directory entry for
903 * the root directory, followed by all the other directory entries in the
904 * filesystem. The subdir_offset field of each directory entry gives the start
905 * of its child entries from the beginning of the metadata resource. An
906 * end-of-directory is signaled by a directory entry of length '0', really of
907 * length 8, because that's how long the 'length' field is.
909 * @fp: The FILE* for the input WIM file.
910 * @res_entry: The resource entry for the metadata resource (a.k.a the metadata
912 * @wim_ctype: The compression type of the WIM file.
913 * @root_dentry_p: A pointer to a pointer to a struct dentry structure into which the
914 * root dentry is allocated and returned.
916 * @return: True on success, false on failure.
918 int read_metadata_resource(FILE *fp, const struct resource_entry *res_entry,
919 int wim_ctype, struct image_metadata *image_metadata)
925 struct dentry *dentry = NULL;
927 DEBUG("Reading metadata resource: length = %lu, offset = %lu\n",
928 res_entry->original_size, res_entry->offset);
930 if (res_entry->original_size < 8) {
931 ERROR("Expected at least 8 bytes for the metadata "
933 return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
936 /* Allocate memory for the uncompressed metadata resource. */
937 buf = MALLOC(res_entry->original_size);
940 ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
941 "metadata resource!\n",
942 res_entry->original_size);
943 return WIMLIB_ERR_NOMEM;
946 /* Determine the compression type of the metadata resource. */
947 ctype = resource_compression_type(wim_ctype, res_entry->flags);
949 /* Read the metadata resource into memory. (It may be compressed.) */
950 ret = read_full_resource(fp, res_entry->size,
951 res_entry->original_size, res_entry->offset,
956 DEBUG("Finished reading metadata resource into memory.\n");
959 dentry = MALLOC(sizeof(struct dentry));
961 ERROR("Failed to allocate %zu bytes for root dentry!\n",
962 sizeof(struct dentry));
963 ret = WIMLIB_ERR_NOMEM;
967 /* Read the root directory entry starts after security data, on an
968 * 8-byte aligned address.
970 * The security data starts with a 4-byte integer giving its total
973 /* Read the security data into a WIMSecurityData structure. */
974 #ifdef ENABLE_SECURITY_DATA
975 ret = read_security_data(buf, res_entry->original_size,
976 &image_metadata->security_data);
980 get_u32(buf, &dentry_offset);
981 if (dentry_offset == 0)
983 dentry_offset += (8 - dentry_offset % 8) % 8;
985 ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry);
989 /* This is the root dentry, so set its pointers correctly. */
990 dentry->parent = dentry;
991 dentry->next = dentry;
992 dentry->prev = dentry;
994 /* Now read the entire directory entry tree. */
995 ret = read_dentry_tree(buf, res_entry->original_size, dentry);
999 /* Calculate the full paths in the dentry tree. */
1000 ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
1004 image_metadata->root_dentry = dentry;
1008 free_dentry_tree(dentry, NULL, false);
1014 /* Write the metadata resource for the current image. */
1015 int write_metadata_resource(WIMStruct *w)
1022 struct dentry *root;
1023 struct lookup_table_entry *lte;
1024 struct resource_entry *res_entry;
1025 off_t metadata_offset;
1026 u64 metadata_original_size;
1027 u64 metadata_compressed_size;
1029 u8 hash[WIM_HASH_SIZE];
1031 DEBUG("Writing metadata resource for image %u\n", w->current_image);
1034 root = wim_root_dentry(w);
1035 metadata_ctype = wimlib_get_compression_type(w);
1036 metadata_offset = ftello(out);
1037 if (metadata_offset == -1)
1038 return WIMLIB_ERR_WRITE;
1040 #ifdef ENABLE_SECURITY_DATA
1041 struct wim_security_data *sd = wim_security_data(w);
1043 subdir_offset = sd->total_length + root->length + 8;
1046 subdir_offset = 8 + root->length + 8;
1047 calculate_subdir_offsets(root, &subdir_offset);
1048 metadata_original_size = subdir_offset;
1049 buf = MALLOC(metadata_original_size);
1051 ERROR("Failed to allocate %"PRIu64" bytes for "
1052 "metadata resource\n", metadata_original_size);
1053 return WIMLIB_ERR_NOMEM;
1055 #ifdef ENABLE_SECURITY_DATA
1056 /* Write the security data. */
1057 p = write_security_data(sd, buf);
1059 p = put_u32(buf, 8); /* Total length of security data. */
1060 p = put_u32(p, 0); /* Number of security data entries. */
1063 DEBUG("Writing dentry tree.\n");
1064 p = write_dentry_tree(root, p);
1066 /* Like file resources, the lookup table entry for a metadata resource
1067 * uses for the hash code a SHA1 message digest of its uncompressed
1069 sha1_buffer(buf, metadata_original_size, hash);
1071 ret = write_resource_from_memory(buf,
1073 metadata_original_size,
1075 &metadata_compressed_size);
1080 /* Update the lookup table entry, including the hash and output resource
1081 * entry fields, for this image's metadata resource. */
1082 lte = wim_metadata_lookup_table_entry(w);
1083 res_entry = <e->output_resource_entry;
1085 if (memcmp(hash, lte->hash, WIM_HASH_SIZE) != 0) {
1086 lookup_table_unlink(w->lookup_table, lte);
1087 memcpy(lte->hash, hash, WIM_HASH_SIZE);
1088 lookup_table_insert(w->lookup_table, lte);
1090 res_entry->original_size = metadata_original_size;
1091 res_entry->offset = metadata_offset;
1092 res_entry->size = metadata_compressed_size;
1093 res_entry->flags = WIM_RESHDR_FLAG_METADATA;
1094 if (metadata_ctype != WIM_COMPRESSION_TYPE_NONE)
1095 res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
1100 * Writes a file resource to the output file.
1102 * @dentry: The dentry for the file resource.
1103 * @wim_p: A pointer to the WIMStruct. The fields of interest to this
1104 * function are the input and output file streams and the lookup table.
1105 * @return zero on success, nonzero on failure.
1107 int write_file_resource(struct dentry *dentry, void *wim_p)
1112 struct lookup_table_entry *lte;
1115 struct resource_entry *output_res_entry;
1122 /* Directories don't need file resources. */
1123 if (dentry_is_directory(dentry))
1126 /* Get the lookup entry for the file resource. */
1127 lte = wim_lookup_resource(w, dentry);
1131 /* No need to write file resources twice. (This indicates file
1132 * resources that are part of a hard link set.) */
1133 if (++lte->out_refcnt != 1)
1136 out_wim_ctype = wimlib_get_compression_type(w);
1137 output_res_entry = <e->output_resource_entry;
1139 /* do not write empty resources */
1140 if (lte->resource_entry.original_size == 0)
1143 /* Figure out if we can read the resource from the WIM file, or
1144 * if we have to read it from the filesystem outside. */
1145 if (lte->file_on_disk) {
1147 /* Read from disk (uncompressed) */
1149 len = lte->resource_entry.original_size;
1151 in = fopen(lte->file_on_disk, "rb");
1153 ERROR("Failed to open the file `%s': %m\n",
1155 return WIMLIB_ERR_OPEN;
1159 puts(lte->file_on_disk);
1161 ret = transfer_file_resource(in, len, len, 0,
1162 WIM_COMPRESSION_TYPE_NONE, out,
1163 out_wim_ctype, output_res_entry);
1167 /* Read from input WIM (possibly compressed) */
1169 /* It may be a different WIM file, in the case of
1170 * exporting images from one WIM file to another */
1171 if (lte->other_wim_fp) {
1172 /* Different WIM file. */
1173 in = lte->other_wim_fp;
1174 in_wim_ctype = lte->other_wim_ctype;
1176 /* Same WIM file. */
1178 in_wim_ctype = out_wim_ctype;
1180 int input_res_ctype = resource_compression_type(
1182 lte->resource_entry.flags);
1184 ret = transfer_file_resource(in,
1185 lte->resource_entry.size,
1186 lte->resource_entry.original_size,
1187 lte->resource_entry.offset,