/*
* resource.c
*
- * Read uncompressed and compressed metadata and file resources.
- *
- * Copyright (C) 2010 Carl Thijssen
- * Copyright (C) 2012 Eric Biggers
+ * Read uncompressed and compressed metadata and file resources from a WIM file.
+ */
+
+/*
+ * Copyright (C) 2012, 2013 Eric Biggers
*
- * wimlib - Library for working with WIM files
+ * This file is part of wimlib, a library for working with WIM files.
*
- * This library is free software; you can redistribute it and/or modify it under
- * the terms of the GNU Lesser General Public License as published by the Free
- * Software Foundation; either version 2.1 of the License, or (at your option) any
- * later version.
+ * wimlib is free software; you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation; either version 3 of the License, or (at your option) any later
+ * version.
*
- * This library is distributed in the hope that it will be useful, but WITHOUT ANY
- * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
- * PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
+ * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
+ * A PARTICULAR PURPOSE. See the GNU General Public License for more details.
*
- * You should have received a copy of the GNU Lesser General Public License along
- * with this library; if not, write to the Free Software Foundation, Inc., 59
- * Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ * You should have received a copy of the GNU General Public License along with
+ * wimlib; if not, see http://www.gnu.org/licenses/.
*/
-#include "wimlib_internal.h"
-#include "lookup_table.h"
-#include "io.h"
-#include "lzx.h"
-#include "xpress.h"
-#include "sha1.h"
-#include "dentry.h"
-#include <unistd.h>
-#include <errno.h>
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#include "wimlib.h"
+#include "wimlib/endianness.h"
+#include "wimlib/error.h"
+#include "wimlib/file_io.h"
+#include "wimlib/lookup_table.h"
+#include "wimlib/resource.h"
+#include "wimlib/sha1.h"
+
+#ifdef __WIN32__
+/* for read_win32_file_prefix(), read_win32_encrypted_file_prefix() */
+# include "wimlib/win32.h"
+#endif
-/* Used for buffering FILE IO */
-#define BUFFER_SIZE 4096
+#ifdef WITH_NTFS_3G
+/* for read_ntfs_file_prefix() */
+# include "wimlib/ntfs_3g.h"
+#endif
+
+#ifdef HAVE_ALLOCA_H
+# include <alloca.h>
+#endif
+#include <errno.h>
+#include <fcntl.h>
+#include <stdlib.h>
+#include <unistd.h>
-/*
- * Reads all or part of a compressed resource into an in-memory buffer.
+/*
+ * Compressed WIM resources
+ *
+ * A compressed resource in a WIM consists of a number of compressed chunks,
+ * each of which decompresses to a fixed chunk size (given in the WIM header;
+ * usually 32768) except possibly the last, which always decompresses to any
+ * remaining bytes. In addition, immediately before the chunks, a table (the
+ * "chunk table") provides the offset, in bytes relative to the end of the chunk
+ * table, of the start of each compressed chunk, except for the first chunk
+ * which is omitted as it always has an offset of 0. Therefore, a compressed
+ * resource with N chunks will have a chunk table with N - 1 entries.
+ *
+ * Additional information:
+ *
+ * - Entries in the chunk table are 4 bytes each, except if the uncompressed
+ * size of the resource is greater than 4 GiB, in which case the entries in
+ * the chunk table are 8 bytes each. In either case, the entries are unsigned
+ * little-endian integers.
*
- * @fp: The FILE* for the WIM file.
- * @resource_compressed_size: The compressed size of the resource.
- * @resource_uncompressed_size: The uncompressed size of the resource.
- * @resource_offset: The offset of the start of the resource from
- * the start of the stream @fp.
- * @resource_ctype: The compression type of the resource.
- * @len: The number of bytes of uncompressed data to read from
- * the resource.
- * @offset: The offset of the bytes to read within the uncompressed
- * resource.
- * @contents_len: An array into which the uncompressed data is written.
- * It must be at least @len bytes long.
+ * - The chunk table is included in the compressed size of the resource provided
+ * in the corresponding entry in the WIM's stream lookup table.
*
- * Returns zero on success, nonzero on failure.
+ * - The compressed size of a chunk is never greater than the uncompressed size.
+ * From the compressor's point of view, chunks that would have compressed to a
+ * size greater than or equal to their original size are in fact stored
+ * uncompressed. From the decompresser's point of view, chunks with
+ * compressed size equal to their uncompressed size are in fact uncompressed.
+ *
+ * Furthermore, wimlib supports its own "pipable" WIM format, and for this the
+ * structure of compressed resources was modified to allow piped reading and
+ * writing. To make sequential writing possible, the chunk table is placed
+ * after the chunks rather than before the chunks, and to make sequential
+ * reading possible, each chunk is prefixed with a 4-byte header giving its
+ * compressed size as a 32-bit, unsigned, little-endian integer. Otherwise the
+ * details are the same.
*/
-static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
- u64 resource_uncompressed_size,
- u64 resource_offset, int resource_ctype,
- u64 len, u64 offset, u8 contents_ret[])
+
+
+/* Decompress the specified chunk that uses the specified compression type
+ * @ctype, part of a WIM with default chunk size @wim_chunk_size. For LZX the
+ * separate @wim_chunk_size is needed because it determines the window size used
+ * for LZX compression. */
+static int
+decompress(const void *cchunk, unsigned clen, void *uchunk, unsigned ulen,
+ int ctype, u32 wim_chunk_size)
{
+ switch (ctype) {
+ case WIMLIB_COMPRESSION_TYPE_LZX:
+ return wimlib_lzx_decompress2(cchunk, clen,
+ uchunk, ulen, wim_chunk_size);
+ case WIMLIB_COMPRESSION_TYPE_XPRESS:
+ return wimlib_xpress_decompress(cchunk, clen,
+ uchunk, ulen);
+ case WIMLIB_COMPRESSION_TYPE_LZMS:
+ return wimlib_lzms_decompress(cchunk, clen, uchunk, ulen);
+ default:
+ wimlib_assert(0);
+ return -1;
+ }
+}
- DEBUG2("comp size = %"PRIu64", "
- "uncomp size = %"PRIu64", "
- "res offset = %"PRIu64"\n",
- resource_compressed_size,
- resource_uncompressed_size,
- resource_offset);
- DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"\n",
- wimlib_get_compression_type_string(resource_ctype),
- len, offset);
- /* Trivial case */
- if (len == 0)
- return 0;
-
- int (*decompress)(const void *, uint, void *, uint);
- /* Set the appropriate decompress function. */
- if (resource_ctype == WIM_COMPRESSION_TYPE_LZX)
- decompress = lzx_decompress;
- else
- decompress = xpress_decompress;
-
- /* The structure of a compressed resource consists of a table of chunk
- * offsets followed by the chunks themselves. Each chunk consists of
- * compressed data, and there is one chunk for each WIM_CHUNK_SIZE =
- * 32768 bytes of the uncompressed file, with the last chunk having any
- * remaining bytes.
- *
- * The chunk offsets are measured relative to the end of the chunk
- * table. The first chunk is omitted from the table in the WIM file
- * because its offset is implicitly given by the fact that it directly
- * follows the chunk table and therefore must have an offset of 0.
- */
-
- /* Calculate how many chunks the resource conists of in its entirety. */
- u64 num_chunks = (resource_uncompressed_size + WIM_CHUNK_SIZE - 1) /
- WIM_CHUNK_SIZE;
- /* As mentioned, the first chunk has no entry in the chunk table. */
- u64 num_chunk_entries = num_chunks - 1;
-
-
- /* The index of the chunk that the read starts at. */
- u64 start_chunk = offset / WIM_CHUNK_SIZE;
- /* The byte offset at which the read starts, within the start chunk. */
- u64 start_chunk_offset = offset % WIM_CHUNK_SIZE;
-
- /* The index of the chunk that contains the last byte of the read. */
- u64 end_chunk = (offset + len - 1) / WIM_CHUNK_SIZE;
- /* The byte offset of the last byte of the read, within the end chunk */
- u64 end_chunk_offset = (offset + len - 1) % WIM_CHUNK_SIZE;
-
- /* Number of chunks that are actually needed to read the requested part
- * of the file. */
- u64 num_needed_chunks = end_chunk - start_chunk + 1;
-
- /* If the end chunk is not the last chunk, an extra chunk entry is
- * needed because we need to know the offset of the chunk after the last
- * chunk read to figure out the size of the last read chunk. */
- if (end_chunk != num_chunks - 1)
- num_needed_chunks++;
-
- /* Declare the chunk table. It will only contain offsets for the chunks
- * that are actually needed for this read. */
- u64 chunk_offsets[num_needed_chunks];
-
- /* Set the implicit offset of the first chunk if it is included in the
- * needed chunks.
- *
- * Note: M$'s documentation includes a picture that shows the first
- * chunk starting right after the chunk entry table, labeled as offset
- * 0x10. However, in the actual file format, the offset is measured
- * from the end of the chunk entry table, so the first chunk has an
- * offset of 0. */
- if (start_chunk == 0)
- chunk_offsets[0] = 0;
-
- /* According to M$'s documentation, if the uncompressed size of
- * the file is greater than 4 GB, the chunk entries are 8-byte
- * integers. Otherwise, they are 4-byte integers. */
- u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
- 8 : 4;
-
- /* Size of the full chunk table in the WIM file. */
- u64 chunk_table_size = chunk_entry_size * num_chunk_entries;
-
- /* Read the needed chunk offsets from the table in the WIM file. */
-
- /* Index, in the WIM file, of the first needed entry in the
- * chunk table. */
- u64 start_table_idx = (start_chunk == 0) ? 0 : start_chunk - 1;
-
- /* Number of entries we need to actually read from the chunk
- * table (excludes the implicit first chunk). */
- u64 num_needed_chunk_entries = (start_chunk == 0) ?
- num_needed_chunks - 1 : num_needed_chunks;
-
- /* Skip over unneeded chunk table entries. */
- u64 file_offset_of_needed_chunk_entries = resource_offset +
- start_table_idx * chunk_entry_size;
- if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) {
- ERROR("Failed to seek to byte %"PRIu64" "
- "to read chunk table of compressed "
- "resource: %m\n",
- file_offset_of_needed_chunk_entries);
- return WIMLIB_ERR_READ;
+struct data_range {
+ u64 offset;
+ u64 size;
+};
+
+/* Alternate chunk table format for resources with WIM_RESHDR_FLAG_CONCAT set.
+ */
+struct alt_chunk_table_header_disk {
+ /* Uncompressed size of the resource. */
+ le64 res_usize;
+
+ /* Number of bytes each compressed chunk decompresses into, except
+ * possibly the last which decompresses into the remainder. */
+ le32 chunk_size;
+
+ /* ??? */
+ le32 unknown;
+
+ /* This header is directly followed by a table of compressed sizes of
+ * the chunks. */
+} _packed_attribute;
+
+/* Read data from a compressed WIM resource. */
+static int
+read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
+ const struct data_range * const ranges,
+ const size_t num_ranges,
+ const consume_data_callback_t cb,
+ void * const cb_ctx,
+ const bool raw_chunks_mode)
+{
+ int ret;
+ int errno_save;
+
+ u64 *chunk_offsets = NULL;
+ u8 *ubuf = NULL;
+ void *cbuf = NULL;
+ bool chunk_offsets_malloced = false;
+ bool ubuf_malloced = false;
+ bool cbuf_malloced = false;
+
+ /* Sanity checks */
+ wimlib_assert(rspec != NULL);
+ wimlib_assert(rspec->ctype != WIMLIB_COMPRESSION_TYPE_NONE);
+ wimlib_assert(is_power_of_2(rspec->cchunk_size));
+ wimlib_assert(cb != NULL);
+ wimlib_assert(num_ranges != 0);
+ for (size_t i = 0; i < num_ranges; i++) {
+ wimlib_assert(ranges[i].size != 0);
+ wimlib_assert(ranges[i].offset + ranges[i].size >= ranges[i].size);
+ wimlib_assert(ranges[i].offset + ranges[i].size <= rspec->uncompressed_size);
+ }
+ for (size_t i = 0; i < num_ranges - 1; i++)
+ wimlib_assert(ranges[i].offset + ranges[i].size <= ranges[i + 1].offset);
+
+ /* Get the offsets of the first and last bytes of the read. */
+ const u64 first_offset = ranges[0].offset;
+ const u64 last_offset = ranges[num_ranges - 1].offset + ranges[num_ranges - 1].size - 1;
+
+ /* Get the file descriptor for the WIM. */
+ struct filedes * const in_fd = &rspec->wim->in_fd;
+
+ /* Determine if we're reading a pipable resource from a pipe or not. */
+ const bool is_pipe_read = !filedes_is_seekable(in_fd);
+
+ /* Determine if the chunk table is in an altenate format. */
+ const bool alt_chunk_table = (rspec->flags & WIM_RESHDR_FLAG_CONCAT) && !is_pipe_read;
+
+ /* Get the maximum size of uncompressed chunks in this resource, which
+ * we require be a power of 2. */
+ u32 chunk_size;
+ u64 cur_read_offset = rspec->offset_in_wim;
+ if (alt_chunk_table) {
+ /* Alternate chunk table format. */
+ struct alt_chunk_table_header_disk hdr;
+
+ ret = full_pread(in_fd, &hdr, sizeof(hdr), cur_read_offset);
+ if (ret)
+ goto read_error;
+ cur_read_offset += sizeof(hdr);
+
+ chunk_size = le32_to_cpu(hdr.chunk_size);
+
+ if (!is_power_of_2(chunk_size)) {
+ ERROR("Invalid compressed resource: "
+ "expected power-of-2 chunk size (got %u)", chunk_size);
+ ret = WIMLIB_ERR_INVALID_CHUNK_SIZE;
+ goto out_free_memory;
+ }
+ } else {
+ chunk_size = rspec->cchunk_size;
}
+ const u32 chunk_order = bsr32(chunk_size);
+
+ /* Calculate the total number of chunks the resource is divided into. */
+ const u64 num_chunks = (rspec->uncompressed_size + chunk_size - 1) >> chunk_order;
+
+ /* Calculate the 0-based indices of the first and last chunks containing
+ * data that needs to be passed to the callback. */
+ const u64 first_needed_chunk = first_offset >> chunk_order;
+ const u64 last_needed_chunk = last_offset >> chunk_order;
+
+ /* Calculate the 0-based index of the first chunk that actually needs to
+ * be read. This is normally first_needed_chunk, but for pipe reads we
+ * must always start from the 0th chunk. */
+ const u64 read_start_chunk = (is_pipe_read ? 0 : first_needed_chunk);
+
+ /* Calculate the number of chunk offsets that are needed for the chunks
+ * being read. */
+ const u64 num_needed_chunk_offsets =
+ last_needed_chunk - read_start_chunk + 1 +
+ (last_needed_chunk < num_chunks - 1);
+
+ /* Calculate the number of entries in the chunk table. Normally, it's
+ * one less than the number of chunks, since the first chunk has no
+ * entry. But in the alternate chunk table format, the chunk entries
+ * contain chunk sizes, not offsets, and there is one per chunk. */
+ const u64 num_chunk_entries = (alt_chunk_table ? num_chunks : num_chunks - 1);
+
+ /* Set the size of each chunk table entry based on the resource's
+ * uncompressed size. XXX: Does the alternate chunk table really
+ * always have 4-byte entries? */
+ const u64 chunk_entry_size =
+ (rspec->uncompressed_size > (1ULL << 32) && !alt_chunk_table)
+ ? 8 : 4;
+
+ /* Calculate the size of the chunk table in bytes. */
+ const u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
+
+ /* Includes header */
+ const u64 chunk_table_full_size =
+ (alt_chunk_table) ? chunk_table_size + sizeof(struct alt_chunk_table_header_disk)
+ : chunk_table_size;
+
+ if (!is_pipe_read) {
+ /* Read the needed chunk table entries into memory and use them
+ * to initialize the chunk_offsets array. */
+
+ u64 first_chunk_entry_to_read;
+ u64 last_chunk_entry_to_read;
+
+ if (alt_chunk_table) {
+ /* The alternate chunk table contains chunk sizes, not
+ * offsets, so we always must read all preceding entries
+ * in order to determine offsets. */
+ first_chunk_entry_to_read = 0;
+ last_chunk_entry_to_read = last_needed_chunk;
+ } else {
+ /* Here we must account for the fact that the first
+ * chunk has no explicit chunk table entry. */
+
+ if (read_start_chunk == 0)
+ first_chunk_entry_to_read = 0;
+ else
+ first_chunk_entry_to_read = read_start_chunk - 1;
- /* Number of bytes we need to read from the chunk table. */
- size_t size = num_needed_chunk_entries * chunk_entry_size;
+ if (last_needed_chunk == 0)
+ last_chunk_entry_to_read = 0;
+ else
+ last_chunk_entry_to_read = last_needed_chunk - 1;
- u8 chunk_tab_buf[size];
+ if (last_needed_chunk < num_chunks - 1)
+ last_chunk_entry_to_read++;
+ }
- if (fread(chunk_tab_buf, 1, size, fp) != size)
- goto err;
+ const u64 num_chunk_entries_to_read =
+ last_chunk_entry_to_read - first_chunk_entry_to_read + 1;
- /* Now fill in chunk_offsets from the entries we have read in
- * chunk_tab_buf. */
+ const u64 chunk_offsets_alloc_size =
+ max(num_chunk_entries_to_read,
+ num_needed_chunk_offsets) * sizeof(chunk_offsets[0]);
- u64 *chunk_tab_p = chunk_offsets;
- if (start_chunk == 0)
- chunk_tab_p++;
+ if ((size_t)chunk_offsets_alloc_size != chunk_offsets_alloc_size)
+ goto oom;
- if (chunk_entry_size == 4) {
- u32 *entries = (u32*)chunk_tab_buf;
- while (num_needed_chunk_entries--)
- *chunk_tab_p++ = to_le32(*entries++);
- } else {
- u64 *entries = (u64*)chunk_tab_buf;
- while (num_needed_chunk_entries--)
- *chunk_tab_p++ = to_le64(*entries++);
+ if (chunk_offsets_alloc_size <= STACK_MAX) {
+ chunk_offsets = alloca(chunk_offsets_alloc_size);
+ } else {
+ chunk_offsets = MALLOC(chunk_offsets_alloc_size);
+ if (chunk_offsets == NULL)
+ goto oom;
+ chunk_offsets_malloced = true;
+ }
+
+ const size_t chunk_table_size_to_read =
+ num_chunk_entries_to_read * chunk_entry_size;
+
+ const u64 file_offset_of_needed_chunk_entries =
+ cur_read_offset
+ + (first_chunk_entry_to_read * chunk_entry_size)
+ + (rspec->is_pipable ? (rspec->size_in_wim - chunk_table_size) : 0);
+
+ void * const chunk_table_data =
+ (u8*)chunk_offsets +
+ chunk_offsets_alloc_size -
+ chunk_table_size_to_read;
+
+ ret = full_pread(in_fd, chunk_table_data, chunk_table_size,
+ file_offset_of_needed_chunk_entries);
+ if (ret)
+ goto read_error;
+
+ /* Now fill in chunk_offsets from the entries we have read in
+ * chunk_tab_data. We break aliasing rules here to avoid having
+ * to allocate yet another array. */
+ typedef le64 __attribute__((may_alias)) aliased_le64_t;
+ typedef le32 __attribute__((may_alias)) aliased_le32_t;
+ u64 * chunk_offsets_p = chunk_offsets;
+
+ if (alt_chunk_table) {
+ u64 cur_offset = 0;
+ aliased_le32_t *raw_entries = chunk_table_data;
+
+ for (size_t i = 0; i < num_chunk_entries_to_read; i++) {
+ u32 entry = le32_to_cpu(raw_entries[i]);
+ if (i >= read_start_chunk)
+ *chunk_offsets_p++ = cur_offset;
+ cur_offset += entry;
+ }
+ } else {
+ if (read_start_chunk == 0)
+ *chunk_offsets_p++ = 0;
+
+ if (chunk_entry_size == 4) {
+ aliased_le32_t *raw_entries = chunk_table_data;
+ for (size_t i = 0; i < num_chunk_entries_to_read; i++)
+ *chunk_offsets_p++ = le32_to_cpu(raw_entries[i]);
+ } else {
+ aliased_le64_t *raw_entries = chunk_table_data;
+ for (size_t i = 0; i < num_chunk_entries_to_read; i++)
+ *chunk_offsets_p++ = le64_to_cpu(raw_entries[i]);
+ }
+ }
+
+ /* Set offset to beginning of first chunk to read. */
+ cur_read_offset += chunk_offsets[0];
+ if (rspec->is_pipable)
+ cur_read_offset += read_start_chunk * sizeof(struct pwm_chunk_hdr);
+ else
+ cur_read_offset += chunk_table_size;
}
- /* Done with the chunk table now. We must now seek to the first chunk
- * that is needed for the read. */
-
- u64 file_offset_of_first_needed_chunk = resource_offset +
- chunk_table_size + chunk_offsets[0];
- if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) {
- ERROR("Failed to seek to byte %"PRIu64" "
- "to read first chunk of compressed "
- "resource: %m\n",
- file_offset_of_first_needed_chunk);
- return WIMLIB_ERR_READ;
+ /* Allocate buffer for holding the uncompressed data of each chunk. */
+ if (chunk_size <= STACK_MAX) {
+ ubuf = alloca(chunk_size);
+ } else {
+ ubuf = MALLOC(chunk_size);
+ if (ubuf == NULL)
+ goto oom;
+ ubuf_malloced = true;
}
- /* Pointer to current position in the output buffer for uncompressed
- * data. */
- u8 *out_p = (u8*)contents_ret;
-
- /* Buffer for compressed data. While most compressed chunks will have a
- * size much less than WIM_CHUNK_SIZE, WIM_CHUNK_SIZE - 1 is the maximum
- * size in the worst-case. This assumption is valid only if chunks that
- * happen to compress to more than the uncompressed size (i.e. a
- * sequence of random bytes) are always stored uncompressed. But this seems
- * to be the case in M$'s WIM files, even though it is undocumented. */
- u8 compressed_buf[WIM_CHUNK_SIZE - 1];
-
-
- /* Decompress all the chunks. */
- for (u64 i = start_chunk; i <= end_chunk; i++) {
-
- DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64")\n",
- i, start_chunk, end_chunk);
-
- /* Calculate the sizes of the compressed chunk and of the
- * uncompressed chunk. */
- uint compressed_chunk_size, uncompressed_chunk_size;
- if (i != num_chunks - 1) {
- /* All the chunks except the last one in the resource
- * expand to WIM_CHUNK_SIZE uncompressed, and the amount
- * of compressed data for the chunk is given by the
- * difference of offsets in the chunk offset table. */
- compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
- chunk_offsets[i - start_chunk];
- uncompressed_chunk_size = WIM_CHUNK_SIZE;
+ /* Unless the raw compressed data was requested, allocate a temporary
+ * buffer for reading compressed chunks, each of which can be at most
+ * @chunk_size - 1 bytes. This excludes compressed chunks that are a
+ * full @chunk_size bytes, which are actually stored uncompressed. */
+ if (!raw_chunks_mode) {
+ if (chunk_size - 1 <= STACK_MAX) {
+ cbuf = alloca(chunk_size - 1);
} else {
- /* The last compressed chunk consists of the remaining
- * bytes in the file resource, and the last uncompressed
- * chunk has size equal to however many bytes are left-
- * that is, the remainder of the uncompressed size when
- * divided by WIM_CHUNK_SIZE.
- *
- * Note that the resource_compressed_size includes the
- * chunk table, so the size of it must be subtracted. */
- compressed_chunk_size = resource_compressed_size -
- chunk_table_size -
- chunk_offsets[i - start_chunk];
-
- uncompressed_chunk_size = resource_uncompressed_size %
- WIM_CHUNK_SIZE;
-
- /* If the remainder is 0, the last chunk actually
- * uncompresses to a full WIM_CHUNK_SIZE bytes. */
- if (uncompressed_chunk_size == 0)
- uncompressed_chunk_size = WIM_CHUNK_SIZE;
+ cbuf = MALLOC(chunk_size - 1);
+ if (cbuf == NULL)
+ goto oom;
+ cbuf_malloced = true;
}
+ }
- DEBUG2("compressed_chunk_size = %u, uncompressed_chunk_size = %u\n",
- compressed_chunk_size, uncompressed_chunk_size);
+ /* Read and process each needed chunk. */
+ const struct data_range *cur_range = ranges;
+ const struct data_range * const end_range = &ranges[num_ranges];
+ u64 cur_range_pos = cur_range->offset;
+ u64 cur_range_end = cur_range->offset + cur_range->size;
+ for (u64 i = read_start_chunk; i <= last_needed_chunk; i++) {
- /* Figure out how much of this chunk we actually need to read */
- u64 start_offset;
- if (i == start_chunk)
- start_offset = start_chunk_offset;
- else
- start_offset = 0;
- u64 end_offset;
- if (i == end_chunk)
- end_offset = end_chunk_offset;
+ /* Calculate uncompressed size of next chunk. */
+ u32 chunk_usize;
+ if ((i == num_chunks - 1) && (rspec->uncompressed_size & (chunk_size - 1)))
+ chunk_usize = (rspec->uncompressed_size & (chunk_size - 1));
else
- end_offset = WIM_CHUNK_SIZE - 1;
-
- u64 partial_chunk_size = end_offset + 1 - start_offset;
- bool is_partial_chunk = (partial_chunk_size !=
- uncompressed_chunk_size);
-
- DEBUG2("start_offset = %u, end_offset = %u\n", start_offset,
- end_offset);
- DEBUG2("partial_chunk_size = %u\n", partial_chunk_size);
-
- /* This is undocumented, but chunks can be uncompressed. This
- * appears to always be the case when the compressed chunk size
- * is equal to the uncompressed chunk size. */
- if (compressed_chunk_size == uncompressed_chunk_size) {
- /* Probably an uncompressed chunk */
-
- if (start_offset != 0) {
- if (fseeko(fp, start_offset, SEEK_CUR) != 0) {
- ERROR("Uncompressed partial chunk "
- "fseek() error: %m\n");
- return WIMLIB_ERR_READ;
- }
+ chunk_usize = chunk_size;
+
+ /* Calculate compressed size of next chunk. */
+ u32 chunk_csize;
+ if (is_pipe_read) {
+ struct pwm_chunk_hdr chunk_hdr;
+
+ ret = full_pread(in_fd, &chunk_hdr,
+ sizeof(chunk_hdr), cur_read_offset);
+ if (ret)
+ goto read_error;
+ chunk_csize = le32_to_cpu(chunk_hdr.compressed_size);
+ } else {
+ if (i == num_chunks - 1) {
+ chunk_csize = rspec->size_in_wim -
+ chunk_table_full_size -
+ chunk_offsets[i - read_start_chunk];
+ if (rspec->is_pipable)
+ chunk_csize -= num_chunks * sizeof(struct pwm_chunk_hdr);
+ } else {
+ chunk_csize = chunk_offsets[i + 1 - read_start_chunk] -
+ chunk_offsets[i - read_start_chunk];
+ }
+ }
+ if (chunk_csize == 0 || chunk_csize > chunk_usize) {
+ ERROR("Invalid chunk size in compressed resource!");
+ errno = EINVAL;
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ goto out_free_memory;
+ }
+ if (rspec->is_pipable)
+ cur_read_offset += sizeof(struct pwm_chunk_hdr);
+
+ /* Uncompressed offsets */
+ const u64 chunk_start_offset = i << chunk_order;
+ const u64 chunk_end_offset = chunk_start_offset + chunk_usize;
+
+ if (chunk_end_offset <= cur_range_pos) {
+
+ /* The next range does not require data in this chunk,
+ * so skip it. */
+
+ cur_read_offset += chunk_csize;
+ if (is_pipe_read) {
+ u8 dummy;
+
+ ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
+ if (ret)
+ goto read_error;
}
- if (fread(out_p, 1, partial_chunk_size, fp) !=
- partial_chunk_size)
- goto err;
} else {
- /* Compressed chunk */
- int ret;
-
- /* Read the compressed data into compressed_buf. */
- if (fread(compressed_buf, 1, compressed_chunk_size,
- fp) != compressed_chunk_size)
- goto err;
-
- /* For partial chunks we must buffer the uncompressed
- * data because we don't need all of it. */
- if (is_partial_chunk) {
- u8 uncompressed_buf[uncompressed_chunk_size];
-
- ret = decompress(compressed_buf,
- compressed_chunk_size,
- uncompressed_buf,
- uncompressed_chunk_size);
- if (ret != 0)
- return WIMLIB_ERR_DECOMPRESSION;
- memcpy(out_p, uncompressed_buf + start_offset,
- partial_chunk_size);
+
+ /* Read the chunk and feed data to the callback
+ * function. */
+ u8 *cb_buf;
+
+ ret = full_pread(in_fd,
+ cbuf,
+ chunk_csize,
+ cur_read_offset);
+ if (ret)
+ goto read_error;
+
+ if (chunk_csize != chunk_usize && !raw_chunks_mode) {
+ ret = decompress(cbuf,
+ chunk_csize,
+ ubuf,
+ chunk_usize,
+ rspec->ctype,
+ chunk_size);
+ if (ret) {
+ ERROR("Failed to decompress data!");
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ errno = EINVAL;
+ goto out_free_memory;
+ }
+ cb_buf = ubuf;
} else {
- DEBUG2("out_p = %p\n");
- ret = decompress(compressed_buf,
- compressed_chunk_size,
- out_p,
- uncompressed_chunk_size);
- if (ret != 0)
- return WIMLIB_ERR_DECOMPRESSION;
+ cb_buf = cbuf;
}
+ cur_read_offset += chunk_csize;
+
+ /* At least one range requires data in this chunk.
+ * However, the data fed to the callback function must
+ * not overlap range boundaries. */
+ do {
+ size_t start, end, size;
+
+ start = cur_range_pos - chunk_start_offset;
+ end = min(cur_range_end, chunk_end_offset) - chunk_start_offset;
+ size = end - start;
+
+ if (raw_chunks_mode)
+ ret = (*cb)(&cb_buf[0], chunk_csize, cb_ctx);
+ else
+ ret = (*cb)(&cb_buf[start], size, cb_ctx);
+
+ if (ret)
+ goto out_free_memory;
+
+ cur_range_pos += size;
+ if (cur_range_pos == cur_range_end) {
+ if (++cur_range == end_range) {
+ cur_range_pos = ~0ULL;
+ } else {
+ cur_range_pos = cur_range->offset;
+ cur_range_end = cur_range->offset + cur_range->size;
+ }
+ }
+ } while (cur_range_pos < chunk_end_offset);
}
+ }
- /* Advance the pointer into the uncompressed output data by the
- * number of uncompressed bytes that were written. */
- out_p += partial_chunk_size;
+ if (is_pipe_read
+ && last_offset == rspec->uncompressed_size - 1
+ && chunk_table_size)
+ {
+ u8 dummy;
+ /* Skip chunk table at end of pipable resource. */
+
+ cur_read_offset += chunk_table_size;
+ ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
+ if (ret)
+ goto read_error;
}
+ ret = 0;
+out_free_memory:
+ errno_save = errno;
+ if (chunk_offsets_malloced)
+ FREE(chunk_offsets);
+ if (ubuf_malloced)
+ FREE(ubuf);
+ if (cbuf_malloced)
+ FREE(cbuf);
+ errno = errno_save;
+ return ret;
- return 0;
+oom:
+ ERROR("Not enough memory available to read size=%"PRIu64" bytes "
+ "from compressed resource!", last_offset - first_offset + 1);
+ errno = ENOMEM;
+ ret = WIMLIB_ERR_NOMEM;
+ goto out_free_memory;
-err:
- if (feof(fp))
- ERROR("Unexpected EOF in compressed file resource\n");
- else
- ERROR("Error reading compressed file resource: %m\n");
- return WIMLIB_ERR_READ;
+read_error:
+ ERROR_WITH_ERRNO("Error reading compressed file resource!");
+ goto out_free_memory;
}
-/*
- * Reads uncompressed data from an open file stream.
- */
-int read_uncompressed_resource(FILE *fp, u64 offset, u64 len,
- u8 contents_ret[])
+/* Read raw data from a file descriptor at the specified offset. */
+static int
+read_raw_file_data(struct filedes *in_fd, u64 size, consume_data_callback_t cb,
+ u32 cb_chunk_size, void *ctx_or_buf, u64 offset)
{
- if (fseeko(fp, offset, SEEK_SET) != 0) {
- ERROR("Failed to seek to byte %"PRIu64" of input file "
- "to read uncompressed resource "
- "(len = %"PRIu64")!\n", offset, len);
- return WIMLIB_ERR_READ;
- }
- if (fread(contents_ret, 1, len, fp) != len) {
- if (feof(fp)) {
- ERROR("Unexpected EOF in uncompressed file resource!\n");
+ int ret;
+ u8 *tmp_buf;
+ bool tmp_buf_malloced = false;
+
+ if (cb) {
+ /* Send data to callback function in chunks. */
+ if (cb_chunk_size <= STACK_MAX) {
+ tmp_buf = alloca(cb_chunk_size);
} else {
- ERROR("Failed to read %"PRIu64" bytes from "
- "uncompressed resource at offset "
- "%"PRIu64"\n", len, offset);
+ tmp_buf = MALLOC(cb_chunk_size);
+ if (tmp_buf == NULL) {
+ ret = WIMLIB_ERR_NOMEM;
+ goto out;
+ }
+ tmp_buf_malloced = true;
}
- return WIMLIB_ERR_READ;
- }
- return 0;
-}
-
-/*
- * Reads a WIM resource.
- *
- * @fp: The FILE* for the WIM file.
- * @resource_size: The compressed size of the resource.
- * @resource_original_size: The uncompressed size of the resource.
- * @resource_offset: The offset of the resource in the stream @fp.
- * @resource_ctype: The compression type of the resource.
- * (WIM_COMPRESSION_TYPE_*)
- * @len: How many bytes of the resource should be read.
- * @offset: The offset within the resource at which the read
- * will occur.
- *
- * To read the whole file resource, specify offset =
- * 0 and len = resource_original_size, or call
- * read_full_resource().
- *
- * @contents_ret: An array, that must have length at least @len,
- * into which the uncompressed contents of
- * the file resource starting at @offset and
- * continuing for @len bytes will be written.
- *
- * @return: Zero on success, nonzero on failure. Failure may be due to
- * being unable to read the data from the WIM file at the
- * specified length and offset, or it may be due to the
- * compressed data (if the data is compressed) being
- * invalid.
- */
-int read_resource(FILE *fp, u64 resource_size, u64 resource_original_size,
- u64 resource_offset, int resource_ctype, u64 len,
- u64 offset, void *contents_ret)
-{
- if (resource_ctype == WIM_COMPRESSION_TYPE_NONE) {
- if (resource_size != resource_original_size) {
- ERROR("Resource with original size %"PRIu64" "
- "bytes is marked as uncompressed, \n",
- resource_original_size);
- ERROR(" but its actual size is %"PRIu64" "
- "bytes!\n",
- resource_size);
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
+ while (size) {
+ size_t bytes_to_read = min(cb_chunk_size, size);
+ ret = full_pread(in_fd, tmp_buf, bytes_to_read,
+ offset);
+ if (ret)
+ goto read_error;
+ ret = cb(tmp_buf, bytes_to_read, ctx_or_buf);
+ if (ret)
+ goto out;
+ size -= bytes_to_read;
+ offset += bytes_to_read;
}
- return read_uncompressed_resource(fp,
- resource_offset + offset,
- len, contents_ret);
} else {
- return read_compressed_resource(fp, resource_size,
- resource_original_size, resource_offset,
- resource_ctype, len, offset, contents_ret);
+ /* Read data directly into buffer. */
+ ret = full_pread(in_fd, ctx_or_buf, size, offset);
+ if (ret)
+ goto read_error;
}
+ ret = 0;
+ goto out;
+
+read_error:
+ ERROR_WITH_ERRNO("Read error");
+out:
+ if (tmp_buf_malloced)
+ FREE(tmp_buf);
+ return ret;
}
+static int
+bufferer_cb(const void *chunk, size_t size, void *_ctx)
+{
+ u8 **buf_p = _ctx;
-/*
- * Extracts the first @size bytes file resource specified by @entry to the open
- * file @fd. Returns nonzero on error.
- *
- * XXX
- * This function is somewhat redundant with uncompress_resource(). The
- * main difference is that this function writes to a file descriptor using
- * low-level calls to write() rather than to a FILE* with fwrite(); also this
- * function allows only up to @size bytes to be extracted.
- */
-int extract_resource_to_fd(WIMStruct *w, const struct resource_entry *entry,
- int fd, u64 size)
+ *buf_p = mempcpy(*buf_p, chunk, size);
+ return 0;
+}
+
+struct rechunker_context {
+ u8 *buffer;
+ u32 buffer_filled;
+ u32 cb_chunk_size;
+
+ const struct data_range *ranges;
+ size_t num_ranges;
+ size_t cur_range;
+ u64 range_bytes_remaining;
+
+ consume_data_callback_t cb;
+ void *cb_ctx;
+};
+
+static int
+rechunker_cb(const void *chunk, size_t size, void *_ctx)
{
- u64 num_chunks;
- u64 n;
- u8 buf[min(size, WIM_CHUNK_SIZE)];
- int res_ctype;
- u64 offset;
- u64 i;
+ struct rechunker_context *ctx = _ctx;
+ const u8 *chunkptr = chunk;
+ size_t bytes_to_copy;
int ret;
- errno = 0;
-
- num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
- n = WIM_CHUNK_SIZE;
- res_ctype = wim_resource_compression_type(w, entry);
- offset = 0;
- for (i = 0; i < num_chunks; i++) {
- if (i == num_chunks - 1) {
- n = size % WIM_CHUNK_SIZE;
- if (n == 0) {
- n = WIM_CHUNK_SIZE;
- }
- }
+ wimlib_assert(ctx->cur_range != ctx->num_ranges);
- ret = read_resource(w->fp, entry->size, entry->original_size,
- entry->offset, res_ctype, n, offset, buf);
- if (ret != 0)
- return ret;
+ while (size) {
+ bytes_to_copy = size;
+
+ if (bytes_to_copy > ctx->cb_chunk_size - ctx->buffer_filled)
+ bytes_to_copy = ctx->cb_chunk_size - ctx->buffer_filled;
+
+ if (bytes_to_copy > ctx->range_bytes_remaining - ctx->buffer_filled)
+ bytes_to_copy = ctx->range_bytes_remaining - ctx->buffer_filled;
- if (full_write(fd, buf, n) != n)
- return WIMLIB_ERR_WRITE;
- offset += n;
+ memcpy(&ctx->buffer[ctx->buffer_filled], chunkptr, bytes_to_copy);
+
+ ctx->buffer_filled += bytes_to_copy;
+ chunkptr += bytes_to_copy;
+ size -= bytes_to_copy;
+ ctx->range_bytes_remaining -= bytes_to_copy;
+
+ if (ctx->buffer_filled == ctx->cb_chunk_size ||
+ ctx->range_bytes_remaining == 0)
+ {
+ ret = (*ctx->cb)(ctx->buffer, ctx->buffer_filled, ctx->cb_ctx);
+ if (ret)
+ return ret;
+ ctx->buffer_filled = 0;
+
+ if (ctx->range_bytes_remaining == 0 &&
+ ++ctx->cur_range != ctx->num_ranges)
+ ctx->range_bytes_remaining = ctx->ranges[ctx->cur_range].size;
+ }
}
- return ret;
+ return 0;
}
-/*
- * Copies the file resource specified by the lookup table entry @lte from the
- * input WIM, pointed to by the fp field of the WIMStruct, to the output WIM,
- * pointed to by the out_fp field of the WIMStruct.
+/*
+ * read_partial_wim_resource()-
+ *
+ * Read a range of data from an uncompressed or compressed resource in a WIM
+ * file. Data is written into a buffer or fed into a callback function, as
+ * documented in read_stream_prefix().
*
- * The output_resource_entry, out_refcnt, and part_number fields of @lte are
- * updated.
+ * By default, this function provides the uncompressed data of the resource, and
+ * @size and @offset and interpreted relative to the uncompressed contents of
+ * the resource. This behavior can be modified by either of the following
+ * flags:
*
- * Metadata resources are not copied (they are handled elsewhere for joining and
- * splitting).
+ * WIMLIB_READ_RESOURCE_FLAG_RAW_FULL:
+ * Read @size bytes at @offset of the raw contents of the compressed
+ * resource. In the case of pipable resources, this excludes the stream
+ * header. Exclusive with WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS.
+ *
+ * WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS:
+ * Read the raw compressed chunks of the compressed resource. @size must
+ * be the full uncompressed size, @offset must be 0, and @cb_chunk_size
+ * must be the resource chunk size.
+ *
+ * Return values:
+ * WIMLIB_ERR_SUCCESS (0)
+ * WIMLIB_ERR_READ (errno set)
+ * WIMLIB_ERR_UNEXPECTED_END_OF_FILE (errno set to 0)
+ * WIMLIB_ERR_NOMEM (errno set to ENOMEM)
+ * WIMLIB_ERR_DECOMPRESSION (errno set to EINVAL)
+ *
+ * or other error code returned by the @cb function.
*/
-int copy_resource(struct lookup_table_entry *lte, void *w)
+int
+read_partial_wim_resource(const struct wim_lookup_table_entry *lte,
+ u64 size, consume_data_callback_t cb,
+ u32 cb_chunk_size, void *ctx_or_buf,
+ int flags, u64 offset)
{
- if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
- !((WIMStruct*)w)->write_metadata) {
- return 0;
+ const struct wim_resource_spec *rspec;
+ struct filedes *in_fd;
+
+ /* Verify parameters. */
+ wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
+ rspec = lte->rspec;
+ in_fd = &rspec->wim->in_fd;
+ if (cb)
+ wimlib_assert(is_power_of_2(cb_chunk_size));
+ if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
+ /* Raw chunks mode is subject to the restrictions noted. */
+ wimlib_assert(!lte_is_partial(lte));
+ wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL));
+ wimlib_assert(cb_chunk_size == rspec->cchunk_size);
+ wimlib_assert(size == lte->size);
+ wimlib_assert(offset == 0);
+ } else if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) {
+ /* Raw full mode: read must not overrun end of store size. */
+ wimlib_assert(!lte_is_partial(lte));
+ wimlib_assert(offset + size >= size &&
+ offset + size <= rspec->size_in_wim);
+ } else {
+ /* Normal mode: read must not overrun end of original size. */
+ wimlib_assert(offset + size >= size &&
+ offset + size <= lte->size);
}
- FILE *in_fp = ((WIMStruct*)w)->fp;
- FILE *out_fp = ((WIMStruct*)w)->out_fp;
- int ret;
- u64 size = lte->resource_entry.size;
- u64 offset = lte->resource_entry.offset;
- off_t new_offset = ftello(out_fp);
+ DEBUG("Reading WIM resource: %"PRIu64" @ +%"PRIu64"[+%"PRIu64"] "
+ "from %"PRIu64"(%"PRIu64") @ +%"PRIu64" "
+ "(readflags 0x%08x, resflags 0x%02x%s)",
+ size, offset, lte->offset_in_res,
+ rspec->size_in_wim,
+ rspec->uncompressed_size,
+ rspec->offset_in_wim,
+ flags, lte->flags,
+ (rspec->is_pipable ? ", pipable" : ""));
+
+ if ((flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) ||
+ rspec->ctype == WIMLIB_COMPRESSION_TYPE_NONE)
+ {
+ return read_raw_file_data(in_fd,
+ size,
+ cb,
+ cb_chunk_size,
+ ctx_or_buf,
+ rspec->offset_in_wim + lte->offset_in_res + offset);
+ } else {
+ bool raw_chunks;
+ struct data_range range;
+ consume_data_callback_t internal_cb;
+ void *internal_cb_ctx;
+ u8 *buf;
+ bool rechunker_buf_malloced = false;
+ struct rechunker_context *rechunker_ctx;
+ int ret;
+
+ if (size == 0)
+ return 0;
+
+ range.offset = lte->offset_in_res + offset;
+ range.size = size;
+ raw_chunks = !!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS);
+
+ if (cb != NULL &&
+ cb_chunk_size == rspec->cchunk_size &&
+ !(rspec->flags & WIM_RESHDR_FLAG_CONCAT))
+ {
+ internal_cb = cb;
+ internal_cb_ctx = ctx_or_buf;
+ } else if (cb == NULL) {
+ buf = ctx_or_buf;
+ internal_cb = bufferer_cb;
+ internal_cb_ctx = &buf;
+ } else {
+ rechunker_ctx = alloca(sizeof(struct rechunker_context));
+
+ if (cb_chunk_size <= STACK_MAX) {
+ rechunker_ctx->buffer = alloca(cb_chunk_size);
+ } else {
+ rechunker_ctx->buffer = MALLOC(cb_chunk_size);
+ if (rechunker_ctx->buffer == NULL)
+ return WIMLIB_ERR_NOMEM;
+ rechunker_buf_malloced = true;
+ }
+ rechunker_ctx->buffer_filled = 0;
+ rechunker_ctx->cb_chunk_size = cb_chunk_size;
- if (new_offset == -1)
- return WIMLIB_ERR_WRITE;
+ rechunker_ctx->ranges = ⦥
+ rechunker_ctx->num_ranges = 1;
+ rechunker_ctx->cur_range = 0;
+ rechunker_ctx->range_bytes_remaining = range.size;
- ret = copy_between_files(in_fp, offset, out_fp, size);
- if (ret != 0)
- return ret;
+ rechunker_ctx->cb = cb;
+ rechunker_ctx->cb_ctx = ctx_or_buf;
- memcpy(<e->output_resource_entry, <e->resource_entry,
- sizeof(struct resource_entry));
+ internal_cb = rechunker_cb;
+ internal_cb_ctx = rechunker_ctx;
+ }
- lte->output_resource_entry.offset = new_offset;
- lte->out_refcnt = lte->refcnt;
- lte->part_number = ((WIMStruct*)w)->hdr.part_number;
- return 0;
-}
+ ret = read_compressed_wim_resource(rspec, &range, 1,
+ internal_cb, internal_cb_ctx,
+ raw_chunks);
+ if (rechunker_buf_malloced)
+ FREE(rechunker_ctx->buffer);
-/* Reads the contents of a struct resource_entry, as represented in the on-disk
- * format, from the memory pointed to by @p, and fills in the fields of @entry.
- * A pointer to the byte after the memory read at @p is returned. */
-const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry)
-{
- u64 size;
- u8 flags;
-
- p = get_u56(p, &size);
- p = get_u8(p, &flags);
- entry->size = size;
- entry->flags = flags;
- p = get_u64(p, &entry->offset);
- p = get_u64(p, &entry->original_size);
- return p;
+ return ret;
+ }
}
-/* Copies the struct resource_entry @entry to the memory pointed to by @p in the
- * on-disk format. A pointer to the byte after the memory written at @p is
- * returned. */
-u8 *put_resource_entry(u8 *p, const struct resource_entry *entry)
+int
+read_partial_wim_stream_into_buf(const struct wim_lookup_table_entry *lte,
+ size_t size, u64 offset, void *buf)
{
- p = put_u56(p, entry->size);
- p = put_u8(p, entry->flags);
- p = put_u64(p, entry->offset);
- p = put_u64(p, entry->original_size);
- return p;
+ return read_partial_wim_resource(lte, size, NULL, 0, buf, 0, offset);
}
-/* Given the compression type for the WIM file as a whole as the flags field of
- * a resource entry, returns the compression type for that resource entry. */
-int resource_compression_type(int wim_ctype, int reshdr_flags)
+static int
+read_wim_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
+ consume_data_callback_t cb, u32 cb_chunk_size,
+ void *ctx_or_buf, int flags)
{
- if (wim_ctype == WIM_COMPRESSION_TYPE_NONE) {
- return WIM_COMPRESSION_TYPE_NONE;
- } else {
- if (reshdr_flags & WIM_RESHDR_FLAG_COMPRESSED)
- return wim_ctype;
- else
- return WIM_COMPRESSION_TYPE_NONE;
- }
+ return read_partial_wim_resource(lte, size, cb, cb_chunk_size,
+ ctx_or_buf, flags, 0);
}
-
-
-/*
- * Copies bytes between two file streams.
+#ifndef __WIN32__
+/* This function handles reading stream data that is located in an external
+ * file, such as a file that has been added to the WIM image through execution
+ * of a wimlib_add_command.
*
- * Copies @len bytes from @in to @out, at the current position in @out, and at
- * an offset of @in_offset in @in.
+ * This assumes the file can be accessed using the standard POSIX open(),
+ * read(), and close(). On Windows this will not necessarily be the case (since
+ * the file may need FILE_FLAG_BACKUP_SEMANTICS to be opened, or the file may be
+ * encrypted), so Windows uses its own code for its equivalent case.
*/
-int copy_between_files(FILE *in, off_t in_offset, FILE *out, size_t len)
+static int
+read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte, u64 size,
+ consume_data_callback_t cb, u32 cb_chunk_size,
+ void *ctx_or_buf, int _ignored_flags)
{
- u8 buf[BUFFER_SIZE];
- size_t n;
+ int ret;
+ int raw_fd;
+ struct filedes fd;
- if (fseeko(in, in_offset, SEEK_SET) != 0) {
- ERROR("Failed to seek to byte %"PRIu64" of input file: %m\n",
- in_offset);
- return WIMLIB_ERR_READ;
- }
- /* To reduce memory usage and improve speed, read and write BUFFER_SIZE
- * bytes at a time. */
- while (len != 0) {
- n = min(len, BUFFER_SIZE);
- if (fread(buf, 1, n, in) != n) {
- if (feof(in)) {
- ERROR("Unexpected EOF when copying data "
- "between files\n");
- } else {
- ERROR("Error copying data between files: %m\n");
- }
- return WIMLIB_ERR_READ;
- }
+ wimlib_assert(size <= lte->size);
+ DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", size, lte->file_on_disk);
- if (fwrite(buf, 1, n, out) != n) {
- ERROR("Error copying data between files: %m\n");
- return WIMLIB_ERR_WRITE;
+ raw_fd = open(lte->file_on_disk, O_BINARY | O_RDONLY);
+ if (raw_fd < 0) {
+ ERROR_WITH_ERRNO("Can't open \"%"TS"\"", lte->file_on_disk);
+ return WIMLIB_ERR_OPEN;
+ }
+ filedes_init(&fd, raw_fd);
+ ret = read_raw_file_data(&fd, size, cb, cb_chunk_size, ctx_or_buf, 0);
+ filedes_close(&fd);
+ return ret;
+}
+#endif /* !__WIN32__ */
+
+/* This function handles the trivial case of reading stream data that is, in
+ * fact, already located in an in-memory buffer. */
+static int
+read_buffer_prefix(const struct wim_lookup_table_entry *lte,
+ u64 size, consume_data_callback_t cb,
+ u32 cb_chunk_size, void *ctx_or_buf, int _ignored_flags)
+{
+ wimlib_assert(size <= lte->size);
+
+ if (cb) {
+ /* Feed the data into the callback function in
+ * appropriately-sized chunks. */
+ int ret;
+ u32 chunk_size;
+
+ for (u64 offset = 0; offset < size; offset += chunk_size) {
+ chunk_size = min(cb_chunk_size, size - offset);
+ ret = cb((const u8*)lte->attached_buffer + offset,
+ chunk_size, ctx_or_buf);
+ if (ret)
+ return ret;
}
- len -= n;
+ } else {
+ /* Copy the data directly into the specified buffer. */
+ memcpy(ctx_or_buf, lte->attached_buffer, size);
}
return 0;
}
+typedef int (*read_stream_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
+ u64 size, consume_data_callback_t cb,
+ u32 cb_chunk_size, void *ctx_or_buf,
+ int flags);
-/*
- * Uncompresses a WIM file resource and writes it uncompressed to a file stream.
+/*
+ * read_stream_prefix()-
+ *
+ * Reads the first @size bytes from a generic "stream", which may be located in
+ * any one of several locations, such as in a WIM file (compressed or
+ * uncompressed), in an external file, or directly in an in-memory buffer.
+ *
+ * This function feeds the data either to a callback function (@cb != NULL,
+ * passing it @ctx_or_buf), or write it directly into a buffer (@cb == NULL,
+ * @ctx_or_buf specifies the buffer, which must have room for at least @size
+ * bytes).
+ *
+ * When (@cb != NULL), @cb_chunk_size specifies the maximum size of data chunks
+ * to feed the callback function. @cb_chunk_size must be positive, and if the
+ * stream is in a WIM file, must be a power of 2. All chunks, except possibly
+ * the last one, will be this size. If (@cb == NULL), @cb_chunk_size is
+ * ignored.
+ *
+ * If the stream is located in a WIM file, @flags can be set as documented in
+ * read_partial_wim_resource(). Otherwise @flags are ignored.
*
- * @in: The file stream that contains the file resource.
- * @size: The size of the resource in the input file.
- * @original_size: The original (uncompressed) size of the resource.
- * @offset: The offset of the start of the resource in @in.
- * @input_ctype: The compression type of the resource in @in.
- * @out: The file stream to write the file resource to.
+ * Returns 0 on success; nonzero on error. A nonzero value will be returned if
+ * the stream data cannot be successfully read (for a number of different
+ * reasons, depending on the stream location), or if a callback function was
+ * specified and it returned nonzero.
*/
-static int uncompress_resource(FILE *in, u64 size, u64 original_size,
- off_t offset, int input_ctype, FILE *out)
+int
+read_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
+ consume_data_callback_t cb, u32 cb_chunk_size,
+ void *ctx_or_buf, int flags)
{
- int ret;
- u8 buf[WIM_CHUNK_SIZE];
- /* Determine how many compressed chunks the file is divided into. */
- u64 num_chunks;
- u64 i;
- u64 uncompressed_offset;
- u64 uncompressed_chunk_size;
-
- num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
-
- for (i = 0; i < num_chunks; i++) {
-
- uncompressed_offset = i * WIM_CHUNK_SIZE;
- uncompressed_chunk_size = min(WIM_CHUNK_SIZE,
- original_size - uncompressed_offset);
-
- ret = read_resource(in, size, original_size, offset, input_ctype,
- uncompressed_chunk_size,
- uncompressed_offset, buf);
- if (ret != 0)
- return ret;
+ /* This function merely verifies several preconditions, then passes
+ * control to an appropriate function for understanding each possible
+ * stream location. */
+ static const read_stream_prefix_handler_t handlers[] = {
+ [RESOURCE_IN_WIM] = read_wim_stream_prefix,
+ #ifdef __WIN32__
+ [RESOURCE_IN_FILE_ON_DISK] = read_win32_file_prefix,
+ #else
+ [RESOURCE_IN_FILE_ON_DISK] = read_file_on_disk_prefix,
+ #endif
+ [RESOURCE_IN_ATTACHED_BUFFER] = read_buffer_prefix,
+ #ifdef WITH_FUSE
+ [RESOURCE_IN_STAGING_FILE] = read_file_on_disk_prefix,
+ #endif
+ #ifdef WITH_NTFS_3G
+ [RESOURCE_IN_NTFS_VOLUME] = read_ntfs_file_prefix,
+ #endif
+ #ifdef __WIN32__
+ [RESOURCE_WIN32_ENCRYPTED] = read_win32_encrypted_file_prefix,
+ #endif
+ };
+ wimlib_assert(lte->resource_location < ARRAY_LEN(handlers)
+ && handlers[lte->resource_location] != NULL);
+ wimlib_assert(cb == NULL || cb_chunk_size > 0);
+ return handlers[lte->resource_location](lte, size, cb, cb_chunk_size,
+ ctx_or_buf, flags);
+}
- if (fwrite(buf, 1, uncompressed_chunk_size, out) !=
- uncompressed_chunk_size) {
- ERROR("Failed to write file resource: %m\n");
- return WIMLIB_ERR_WRITE;
- }
- }
- return 0;
+/* Read the full uncompressed data of the specified stream into the specified
+ * buffer, which must have space for at least lte->size bytes. */
+int
+read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *buf)
+{
+ return read_stream_prefix(lte, lte->size, NULL, 0, buf, 0);
}
-/*
- * Transfers a file resource between two files, writing it compressed. The file
- * resource in the input file may be either compressed or uncompressed.
- * Alternatively, the input resource may be in-memory, but it must be
- * uncompressed.
- *
- * @in: The file stream that contains the file resource. Ignored
- * if uncompressed_resource != NULL.
- * @uncompressed_resource: If this pointer is not NULL, it points to an
- * array of @original_size bytes that are
- * the uncompressed input resource.
- * @size: The size of the resource in the input file.
- * @original_size: The original (uncompressed) size of the resource.
- * @offset: The offset of the start of the resource in @in. Ignored
- * if uncompressed_resource != NULL.
- * @input_ctype: The compression type of the resource in @in. Ignored if
- * uncompressed_resource != NULL.
- * @out: The file stream to write the file resource to.
- * @output_type: The compression type to use when writing the resource to
- * @out.
- * @new_size_ret: A location into which the new compressed size of the file
- * resource in returned.
- */
-static int recompress_resource(FILE *in, const u8 *uncompressed_resource,
- u64 size, u64 original_size,
- off_t offset, int input_ctype, FILE *out,
- int output_ctype, u64 *new_size_ret)
+/* Read the full uncompressed data of the specified stream. A buffer sufficient
+ * to hold the data is allocated and returned in @buf_ret. */
+int
+read_full_stream_into_alloc_buf(const struct wim_lookup_table_entry *lte,
+ void **buf_ret)
{
int ret;
- int (*compress)(const void *, uint, void *, uint *);
- if (output_ctype == WIM_COMPRESSION_TYPE_LZX)
- compress = lzx_compress;
- else
- compress = xpress_compress;
-
- u8 uncompressed_buf[WIM_CHUNK_SIZE];
- u8 compressed_buf[WIM_CHUNK_SIZE - 1];
-
- /* Determine how many compressed chunks the file needs to be divided
- * into. */
- u64 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
-
- u64 num_chunk_entries = num_chunks - 1;
-
- /* Size of the chunk entries--- 8 bytes for files over 4GB, otherwise 4
- * bytes */
- uint chunk_entry_size = (original_size >= (u64)1 << 32) ? 8 : 4;
+ void *buf;
- /* Array in which to construct the chunk offset table. */
- u64 chunk_offsets[num_chunk_entries];
-
- /* Offset of the start of the chunk table in the output file. */
- off_t chunk_tab_offset = ftello(out);
-
- /* Total size of the chunk table (as written to the file) */
- u64 chunk_tab_size = chunk_entry_size * num_chunk_entries;
-
- /* Reserve space for the chunk table. */
- if (fwrite(chunk_offsets, 1, chunk_tab_size, out) != chunk_tab_size) {
- ERROR("Failed to write chunk offset table: %m\n");
- return WIMLIB_ERR_WRITE;
+ if ((size_t)lte->size != lte->size) {
+ ERROR("Can't read %"PRIu64" byte stream into "
+ "memory", lte->size);
+ return WIMLIB_ERR_NOMEM;
}
- /* Read each chunk of the file, compress it, write it to the output
- * file, and update th chunk offset table. */
- u64 cur_chunk_offset = 0;
- for (u64 i = 0; i < num_chunks; i++) {
-
- u64 uncompressed_offset = i * WIM_CHUNK_SIZE;
- u64 uncompressed_chunk_size = min(WIM_CHUNK_SIZE,
- original_size - uncompressed_offset);
-
- const u8 *uncompressed_p;
- if (uncompressed_resource != NULL) {
- uncompressed_p = uncompressed_resource +
- uncompressed_offset;
-
- } else {
- /* Read chunk i of the file into uncompressed_buf. */
- ret = read_resource(in, size, original_size, offset, input_ctype,
- uncompressed_chunk_size,
- uncompressed_offset,
- uncompressed_buf);
- if (ret != 0)
- return ret;
- uncompressed_p = uncompressed_buf;
- }
-
- if (i != 0)
- chunk_offsets[i - 1] = cur_chunk_offset;
-
- uint compressed_len;
-
- ret = compress(uncompressed_p, uncompressed_chunk_size,
- compressed_buf, &compressed_len);
-
- /* if compress() returned nonzero, the compressed chunk would
- * have been at least as large as the uncompressed chunk. In
- * this situation, the WIM format requires that the uncompressed
- * chunk be written instead. */
- const u8 *buf_to_write;
- uint len_to_write;
- if (ret == 0) {
- buf_to_write = compressed_buf;
- len_to_write = compressed_len;
- } else {
- buf_to_write = uncompressed_p;
- len_to_write = uncompressed_chunk_size;
- }
+ buf = MALLOC(lte->size);
+ if (buf == NULL)
+ return WIMLIB_ERR_NOMEM;
- if (fwrite(buf_to_write, 1, len_to_write, out) != len_to_write) {
- ERROR("Failed to write compressed file resource: %m\n");
- return WIMLIB_ERR_WRITE;
- }
- cur_chunk_offset += len_to_write;
+ ret = read_full_stream_into_buf(lte, buf);
+ if (ret) {
+ FREE(buf);
+ return ret;
}
- /* The chunk offset after the last chunk, plus the size of the chunk
- * table, gives the total compressed size of the resource. */
- *new_size_ret = cur_chunk_offset + chunk_tab_size;
+ *buf_ret = buf;
+ return 0;
+}
- /* Now that all entries of the chunk table are determined, rewind the
- * stream to where the chunk table was, and write it back out. */
+/* Retrieve the full uncompressed data of the specified WIM resource. */
+static int
+wim_resource_spec_to_data(struct wim_resource_spec *rspec, void **buf_ret)
+{
+ int ret;
+ struct wim_lookup_table_entry *lte;
- if (fseeko(out, chunk_tab_offset, SEEK_SET) != 0) {
- ERROR("Failed to seek to beginning of chunk table: %m\n");
- return WIMLIB_ERR_READ;
- }
+ lte = new_lookup_table_entry();
+ if (lte == NULL)
+ return WIMLIB_ERR_NOMEM;
- if (chunk_entry_size == 8) {
- array_to_le64(chunk_offsets, num_chunk_entries);
+ lte->unhashed = 1;
+ lte_bind_wim_resource_spec(lte, rspec);
+ lte->flags = rspec->flags;
+ lte->size = rspec->uncompressed_size;
+ lte->offset_in_res = 0;
- if (fwrite(chunk_offsets, 1, chunk_tab_size, out) !=
- chunk_tab_size) {
- ERROR("Failed to write chunk table: %m\n");
- return WIMLIB_ERR_WRITE;
- }
- } else {
- u32 chunk_entries_small[num_chunk_entries];
- for (u64 i = 0; i < num_chunk_entries; i++)
- chunk_entries_small[i] = to_le32(chunk_offsets[i]);
- if (fwrite(chunk_entries_small, 1, chunk_tab_size, out) !=
- chunk_tab_size) {
- ERROR("Failed to write chunk table: %m\n");
- return WIMLIB_ERR_WRITE;
- }
- }
+ ret = read_full_stream_into_alloc_buf(lte, buf_ret);
- if (fseeko(out, 0, SEEK_END) != 0) {
- ERROR("Failed to seek to end of output file: %m\n");
- return WIMLIB_ERR_WRITE;
- }
-
- return 0;
+ lte_unbind_wim_resource_spec(lte);
+ free_lookup_table_entry(lte);
+ return ret;
}
-int write_resource_from_memory(const u8 resource[], int out_ctype,
- u64 resource_original_size, FILE *out,
- u64 *resource_size_ret)
+/* Retrieve the full uncompressed data of the specified WIM resource. */
+int
+wim_reshdr_to_data(const struct wim_reshdr *reshdr, WIMStruct *wim, void **buf_ret)
{
- if (out_ctype == WIM_COMPRESSION_TYPE_NONE) {
- if (fwrite(resource, 1, resource_original_size, out) !=
- resource_original_size) {
- ERROR("Failed to write resource of length "
- "%"PRIu64": %m\n",
- resource_original_size);
- return WIMLIB_ERR_WRITE;
- }
- *resource_size_ret = resource_original_size;
- return 0;
- } else {
- return recompress_resource(NULL, resource, resource_original_size,
- resource_original_size, 0, 0, out, out_ctype,
- resource_size_ret);
- }
-}
+ DEBUG("offset_in_wim=%"PRIu64", size_in_wim=%"PRIu64", "
+ "uncompressed_size=%"PRIu64,
+ reshdr->offset_in_wim, reshdr->size_in_wim, reshdr->uncompressed_size);
+ struct wim_resource_spec rspec;
+ wim_res_hdr_to_spec(reshdr, wim, &rspec);
+ return wim_resource_spec_to_data(&rspec, buf_ret);
+}
-/*
- * Transfers a file resource from a FILE* opened for reading to a FILE* opened
- * for writing, possibly changing the compression type.
- *
- * @in: The FILE* that contains the file resource.
- * @size: The (compressed) size of the file resource.
- * @original_size: The uncompressed size of the file resource.
- * @offset: The offset of the file resource in the input file.
- * @input_ctype: The compression type of the file resource in the input
- * file.
- * @out: The FILE* for the output file. The file resource is
- * written at the current position of @out.
- * @output_ctype: The compression type to which the file resource will be
- * converted.
- * @output_res_entry: A pointer to a resource entry that, upon successful
- * return of this function, will have the size,
- * original size, offset, and flags fields filled
- * in for the file resource written to the output
- * file.
- */
-static int transfer_file_resource(FILE *in, u64 size, u64 original_size,
- off_t offset, int input_ctype, FILE *out,
- int output_ctype,
- struct resource_entry *output_res_entry)
+struct read_stream_list_ctx {
+ read_stream_list_begin_stream_t begin_stream;
+ consume_data_callback_t consume_chunk;
+ read_stream_list_end_stream_t end_stream;
+ void *begin_stream_ctx;
+ void *consume_chunk_ctx;
+ void *end_stream_ctx;
+ struct wim_lookup_table_entry *cur_stream;
+ u64 cur_stream_offset;
+ struct wim_lookup_table_entry *final_stream;
+ size_t list_head_offset;
+};
+
+static int
+read_stream_list_wrapper_cb(const void *chunk, size_t size, void *_ctx)
{
+ struct read_stream_list_ctx *ctx = _ctx;
int ret;
- /* Handle zero-length files */
- if (original_size == 0) {
- memset(output_res_entry, 0, sizeof(*output_res_entry));
- return 0;
+ if (ctx->cur_stream_offset == 0) {
+ /* Starting a new stream. */
+ ret = (*ctx->begin_stream)(ctx->cur_stream, ctx->begin_stream_ctx);
+ if (ret)
+ return ret;
}
- /* Get current offset in the output file. */
- output_res_entry->offset = ftello(out);
- if (output_res_entry->offset == -1) {
- ERROR("Failed to get output position: %m\n");
- return WIMLIB_ERR_WRITE;
- }
+ ret = (*ctx->consume_chunk)(chunk, size, ctx->consume_chunk_ctx);
+ if (ret)
+ return ret;
- if (output_ctype == input_ctype) {
- /* The same compression types; simply copy the resource. */
+ ctx->cur_stream_offset += size;
- ret = copy_between_files(in, offset, out, size);
- if (ret != 0)
+ if (ctx->cur_stream_offset == ctx->cur_stream->size) {
+ /* Finished reading all the data for a stream; advance
+ * to the next one. */
+ ret = (*ctx->end_stream)(ctx->cur_stream, ctx->end_stream_ctx);
+ if (ret)
return ret;
- output_res_entry->size = size;
- } else {
- /* Different compression types. */
-
- if (output_ctype == WIM_COMPRESSION_TYPE_NONE) {
- /* Uncompress a compressed file resource */
- ret = uncompress_resource(in, size,
- original_size, offset,
- input_ctype, out);
- if (ret != 0)
- return ret;
- output_res_entry->size = original_size;
- } else {
- u64 new_size;
- /* Compress an uncompressed file resource, or compress a
- * compressed file resource using a different
- * compression type */
- ret = recompress_resource(in, NULL, size, original_size,
- offset, input_ctype, out,
- output_ctype, &new_size);
- if (ret != 0)
- return ret;
- output_res_entry->size = new_size;
- }
- }
+ if (ctx->cur_stream == ctx->final_stream)
+ return 0;
+
+ struct list_head *cur = (struct list_head *)
+ ((u8*)ctx->cur_stream + ctx->list_head_offset);
+ struct list_head *next = cur->next;
- output_res_entry->original_size = original_size;
- if (output_ctype == WIM_COMPRESSION_TYPE_NONE)
- output_res_entry->flags = 0;
- else
- output_res_entry->flags = WIM_RESHDR_FLAG_COMPRESSED;
+ ctx->cur_stream = (struct wim_lookup_table_entry *)
+ ((u8*)next - ctx->list_head_offset);
+
+ ctx->cur_stream_offset = 0;
+ }
return 0;
}
-/*
- * Reads the metadata metadata resource from the WIM file. The metadata
- * resource consists of the security data, followed by the directory entry for
- * the root directory, followed by all the other directory entries in the
- * filesystem. The subdir_offset field of each directory entry gives the start
- * of its child entries from the beginning of the metadata resource. An
- * end-of-directory is signaled by a directory entry of length '0', really of
- * length 8, because that's how long the 'length' field is.
+/*
+ * Read a list of streams, each of which may be in any supported location (e.g.
+ * in a WIM or in an external file). Unlike read_stream_prefix() or the
+ * functions which call it, this function optimizes the case where multiple
+ * streams are packed into a single compressed WIM resource and reads them all
+ * consecutively, only decompressing the data one time.
*
- * @fp: The FILE* for the input WIM file.
- * @res_entry: The resource entry for the metadata resource (a.k.a the metadata
- * for the metadata)
- * @wim_ctype: The compression type of the WIM file.
- * @root_dentry_p: A pointer to a pointer to a struct dentry structure into which the
- * root dentry is allocated and returned.
+ * @stream_list
+ * List of streams (represented as `struct wim_lookup_table_entry's) to
+ * read.
+ * @list_head_offset
+ * Offset of the `struct list_head' within each `struct
+ * wim_lookup_table_entry' that makes up the @stream_list.
+ * @begin_stream
+ * Callback for starting to process a stream.
+ * @consume_chunk
+ * Callback for receiving a chunk of stream data.
+ * @end_stream
+ * Callback for finishing the processing of a stream.
+ * @cb_chunk_size
+ * Size of chunks to provide to @consume_chunk. For a given stream, all
+ * the chunks will be this size, except possibly the last which will be the
+ * remainder.
+ * @cb_ctx
+ * Parameter to pass to the callback functions.
*
- * @return: True on success, false on failure.
+ * Returns 0 on success; a nonzero error code on failure. Failure can occur due
+ * to an error reading the data or due to an error status being returned by any
+ * of the callback functions.
*/
-int read_metadata_resource(FILE *fp, const struct resource_entry *res_entry,
- int wim_ctype, struct image_metadata *image_metadata)
+int
+read_stream_list(struct list_head *stream_list,
+ size_t list_head_offset,
+ read_stream_list_begin_stream_t begin_stream,
+ consume_data_callback_t consume_chunk,
+ read_stream_list_end_stream_t end_stream,
+ u32 cb_chunk_size,
+ void *cb_ctx)
{
- u8 *buf;
- int ctype;
- u32 dentry_offset;
int ret;
- struct dentry *dentry = NULL;
+ struct list_head *cur, *next;
+ struct wim_lookup_table_entry *lte;
- DEBUG("Reading metadata resource: length = %lu, offset = %lu\n",
- res_entry->original_size, res_entry->offset);
+ ret = sort_stream_list_by_sequential_order(stream_list, list_head_offset);
+ if (ret)
+ return ret;
- if (res_entry->original_size < 8) {
- ERROR("Expected at least 8 bytes for the metadata "
- "resource!\n");
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
- }
+ for (cur = stream_list->next, next = cur->next;
+ cur != stream_list;
+ cur = next, next = cur->next)
+ {
+ lte = (struct wim_lookup_table_entry*)((u8*)cur - list_head_offset);
+
+ if (lte_is_partial(lte)) {
+
+ struct wim_lookup_table_entry *lte_next, *lte_last;
+ struct list_head *next2;
+ size_t stream_count;
+
+ /* The next stream is a proper sub-sequence of a WIM
+ * resource. See if there are other streams in the same
+ * resource that need to be read. Since
+ * sort_stream_list_by_sequential_order() sorted the
+ * streams by offset in the WIM, this can be determined
+ * by simply scanning forward in the list. */
+
+ lte_last = lte;
+ stream_count = 1;
+ for (next2 = next;
+ next2 != stream_list
+ && (lte_next = (struct wim_lookup_table_entry*)
+ ((u8*)next2 - list_head_offset),
+ lte_next->resource_location == RESOURCE_IN_WIM
+ && lte_next->rspec == lte->rspec);
+ next2 = next2->next)
+ {
+ lte_last = lte_next;
+ stream_count++;
+ }
+ if (stream_count > 1) {
+ /* Reading multiple streams combined into a
+ * single WIM resource. They are in the stream
+ * list, sorted by offset; @lte specifies the
+ * first stream in the resource that needs to be
+ * read and @lte_last specifies the last stream
+ * in the resource that needs to be read. */
+
+ next = next2;
+
+ struct data_range ranges[stream_count];
+
+ {
+ struct list_head *next3;
+ size_t i;
+ struct wim_lookup_table_entry *lte_cur;
+
+ next3 = cur;
+ for (i = 0; i < stream_count; i++) {
+ lte_cur = (struct wim_lookup_table_entry*)
+ ((u8*)next3 - list_head_offset);
+ ranges[i].offset = lte_cur->offset_in_res;
+ ranges[i].size = lte_cur->size;
+ next3 = next3->next;
+ }
+ }
- /* Allocate memory for the uncompressed metadata resource. */
- buf = MALLOC(res_entry->original_size);
+ struct rechunker_context rechunker_ctx = {
+ .buffer = MALLOC(cb_chunk_size),
+ .buffer_filled = 0,
+ .cb_chunk_size = cb_chunk_size,
+ .ranges = ranges,
+ .num_ranges = stream_count,
+ .cur_range = 0,
+ .range_bytes_remaining = ranges[0].size,
+ .cb = consume_chunk,
+ .cb_ctx = cb_ctx,
+ };
+
+ if (rechunker_ctx.buffer == NULL)
+ return WIMLIB_ERR_NOMEM;
+
+ struct read_stream_list_ctx ctx = {
+ .begin_stream = begin_stream,
+ .begin_stream_ctx = cb_ctx,
+ .consume_chunk = rechunker_cb,
+ .consume_chunk_ctx = &rechunker_ctx,
+ .end_stream = end_stream,
+ .end_stream_ctx = cb_ctx,
+ .cur_stream = lte,
+ .cur_stream_offset = 0,
+ .final_stream = lte_last,
+ .list_head_offset = list_head_offset,
+ };
+
+ ret = read_compressed_wim_resource(lte->rspec,
+ ranges,
+ stream_count,
+ read_stream_list_wrapper_cb,
+ &ctx,
+ false);
+ FREE(rechunker_ctx.buffer);
+ if (ret)
+ return ret;
+ continue;
+ }
+ }
+ ret = (*begin_stream)(lte, cb_ctx);
+ if (ret)
+ return ret;
- if (!buf) {
- ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
- "metadata resource!\n",
- res_entry->original_size);
- return WIMLIB_ERR_NOMEM;
- }
+ ret = read_stream_prefix(lte, lte->size, consume_chunk,
+ cb_chunk_size, cb_ctx, 0);
+ if (ret)
+ return ret;
- /* Determine the compression type of the metadata resource. */
- ctype = resource_compression_type(wim_ctype, res_entry->flags);
+ ret = (*end_stream)(lte, cb_ctx);
+ if (ret)
+ return ret;
+ }
+ return 0;
+}
- /* Read the metadata resource into memory. (It may be compressed.) */
- ret = read_full_resource(fp, res_entry->size,
- res_entry->original_size, res_entry->offset,
- ctype, buf);
- if (ret != 0)
- goto err1;
+struct extract_ctx {
+ SHA_CTX sha_ctx;
+ consume_data_callback_t extract_chunk;
+ void *extract_chunk_arg;
+};
- DEBUG("Finished reading metadata resource into memory.\n");
+static int
+extract_chunk_sha1_wrapper(const void *chunk, size_t chunk_size, void *_ctx)
+{
+ struct extract_ctx *ctx = _ctx;
+ sha1_update(&ctx->sha_ctx, chunk, chunk_size);
+ return ctx->extract_chunk(chunk, chunk_size, ctx->extract_chunk_arg);
+}
- dentry = MALLOC(sizeof(struct dentry));
- if (!dentry) {
- ERROR("Failed to allocate %zu bytes for root dentry!\n",
- sizeof(struct dentry));
- ret = WIMLIB_ERR_NOMEM;
- goto err1;
+/* Extracts the first @size bytes of a stream to somewhere. In the process, the
+ * SHA1 message digest of the uncompressed stream is checked if the full stream
+ * is being extracted.
+ *
+ * @extract_chunk is a function that will be called to extract each chunk of the
+ * stream. */
+int
+extract_stream(const struct wim_lookup_table_entry *lte, u64 size,
+ consume_data_callback_t extract_chunk, void *extract_chunk_arg)
+{
+ int ret;
+ if (size == lte->size) {
+ /* Do SHA1 */
+ struct extract_ctx ctx;
+ ctx.extract_chunk = extract_chunk;
+ ctx.extract_chunk_arg = extract_chunk_arg;
+ sha1_init(&ctx.sha_ctx);
+ ret = read_stream_prefix(lte, size,
+ extract_chunk_sha1_wrapper,
+ lte_cchunk_size(lte),
+ &ctx, 0);
+ if (ret == 0) {
+ u8 hash[SHA1_HASH_SIZE];
+ sha1_final(hash, &ctx.sha_ctx);
+ if (!hashes_equal(hash, lte->hash)) {
+ if (wimlib_print_errors) {
+ ERROR("Invalid SHA1 message digest "
+ "on the following WIM stream:");
+ print_lookup_table_entry(lte, stderr);
+ if (lte->resource_location == RESOURCE_IN_WIM)
+ ERROR("The WIM file appears to be corrupt!");
+ }
+ ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
+ }
+ }
+ } else {
+ /* Don't do SHA1 */
+ ret = read_stream_prefix(lte, size, extract_chunk,
+ lte_cchunk_size(lte),
+ extract_chunk_arg, 0);
}
-
- /* Read the root directory entry starts after security data, on an
- * 8-byte aligned address.
- *
- * The security data starts with a 4-byte integer giving its total
- * length. */
-
- /* Read the security data into a WIMSecurityData structure. */
-#ifdef ENABLE_SECURITY_DATA
- ret = read_security_data(buf, res_entry->original_size,
- &image_metadata->security_data);
- if (ret != 0)
- goto err1;
-#endif
- get_u32(buf, &dentry_offset);
- if (dentry_offset == 0)
- dentry_offset = 8;
- dentry_offset += (8 - dentry_offset % 8) % 8;
-
- ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry);
- if (ret != 0)
- goto err1;
-
- /* This is the root dentry, so set its pointers correctly. */
- dentry->parent = dentry;
- dentry->next = dentry;
- dentry->prev = dentry;
-
- /* Now read the entire directory entry tree. */
- ret = read_dentry_tree(buf, res_entry->original_size, dentry);
- if (ret != 0)
- goto err2;
-
- /* Calculate the full paths in the dentry tree. */
- ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
- if (ret != 0)
- goto err2;
-
- image_metadata->root_dentry = dentry;
- FREE(buf);
return ret;
-err2:
- free_dentry_tree(dentry, NULL, false);
-err1:
- FREE(buf);
+}
+
+static int
+extract_wim_chunk_to_fd(const void *buf, size_t len, void *_fd_p)
+{
+ struct filedes *fd = _fd_p;
+ int ret = full_write(fd, buf, len);
+ if (ret)
+ ERROR_WITH_ERRNO("Error writing to file descriptor");
return ret;
}
-/* Write the metadata resource for the current image. */
-int write_metadata_resource(WIMStruct *w)
+/* Extract the first @size bytes of the specified stream to the specified file
+ * descriptor. If @size is the full size of the stream, its SHA1 message digest
+ * is also checked. */
+int
+extract_stream_to_fd(const struct wim_lookup_table_entry *lte,
+ struct filedes *fd, u64 size)
{
- FILE *out;
- u8 *buf;
- u8 *p;
- int ret;
- off_t subdir_offset;
- struct dentry *root;
- struct lookup_table_entry *lte;
- struct resource_entry *res_entry;
- off_t metadata_offset;
- u64 metadata_original_size;
- u64 metadata_compressed_size;
- int metadata_ctype;
- u8 hash[WIM_HASH_SIZE];
-
- DEBUG("Writing metadata resource for image %u\n", w->current_image);
-
- out = w->out_fp;
- root = wim_root_dentry(w);
- metadata_ctype = wimlib_get_compression_type(w);
- metadata_offset = ftello(out);
- if (metadata_offset == -1)
- return WIMLIB_ERR_WRITE;
-
- #ifdef ENABLE_SECURITY_DATA
- struct wim_security_data *sd = wim_security_data(w);
- if (sd)
- subdir_offset = sd->total_length + root->length + 8;
- else
- #endif
- subdir_offset = 8 + root->length + 8;
- calculate_subdir_offsets(root, &subdir_offset);
- metadata_original_size = subdir_offset;
- buf = MALLOC(metadata_original_size);
- if (!buf) {
- ERROR("Failed to allocate %"PRIu64" bytes for "
- "metadata resource\n", metadata_original_size);
- return WIMLIB_ERR_NOMEM;
- }
- #ifdef ENABLE_SECURITY_DATA
- /* Write the security data. */
- p = write_security_data(sd, buf);
- #else
- p = put_u32(buf, 8); /* Total length of security data. */
- p = put_u32(p, 0); /* Number of security data entries. */
- #endif
+ return extract_stream(lte, size, extract_wim_chunk_to_fd, fd);
+}
- DEBUG("Writing dentry tree.\n");
- p = write_dentry_tree(root, p);
-
- /* Like file resources, the lookup table entry for a metadata resource
- * uses for the hash code a SHA1 message digest of its uncompressed
- * contents. */
- sha1_buffer(buf, metadata_original_size, hash);
-
- ret = write_resource_from_memory(buf,
- metadata_ctype,
- metadata_original_size,
- out,
- &metadata_compressed_size);
- FREE(buf);
- if (ret != 0)
- return ret;
- /* Update the lookup table entry, including the hash and output resource
- * entry fields, for this image's metadata resource. */
- lte = wim_metadata_lookup_table_entry(w);
- res_entry = <e->output_resource_entry;
- lte->out_refcnt++;
- if (memcmp(hash, lte->hash, WIM_HASH_SIZE) != 0) {
- lookup_table_unlink(w->lookup_table, lte);
- memcpy(lte->hash, hash, WIM_HASH_SIZE);
- lookup_table_insert(w->lookup_table, lte);
- }
- res_entry->original_size = metadata_original_size;
- res_entry->offset = metadata_offset;
- res_entry->size = metadata_compressed_size;
- res_entry->flags = WIM_RESHDR_FLAG_METADATA;
- if (metadata_ctype != WIM_COMPRESSION_TYPE_NONE)
- res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
+static int
+sha1_chunk(const void *buf, size_t len, void *ctx)
+{
+ sha1_update(ctx, buf, len);
return 0;
}
-/*
- * Writes a file resource to the output file.
- *
- * @dentry: The dentry for the file resource.
- * @wim_p: A pointer to the WIMStruct. The fields of interest to this
- * function are the input and output file streams and the lookup table.
- * @return zero on success, nonzero on failure.
- */
-int write_file_resource(struct dentry *dentry, void *wim_p)
+/* Calculate the SHA1 message digest of a stream, storing it in @lte->hash. */
+int
+sha1_stream(struct wim_lookup_table_entry *lte)
{
- WIMStruct *w;
- FILE *out;
- FILE *in;
- struct lookup_table_entry *lte;
- int in_wim_ctype;
- int out_wim_ctype;
- int input_res_ctype;
- struct resource_entry *input_res_entry;
- struct resource_entry *output_res_entry;
- u64 len;
int ret;
+ SHA_CTX sha_ctx;
- w = wim_p;
- out = w->out_fp;
-
- /* Directories don't need file resources. */
- if (dentry_is_directory(dentry))
- return 0;
-
- /* Get the lookup entry for the file resource. */
- lte = wim_lookup_resource(w, dentry);
- if (!lte)
- return 0;
-
- /* No need to write file resources twice. (This indicates file
- * resources that are part of a hard link set.) */
- if (++lte->out_refcnt != 1)
- return 0;
+ sha1_init(&sha_ctx);
+ ret = read_stream_prefix(lte, lte->size,
+ sha1_chunk, lte_cchunk_size(lte),
+ &sha_ctx, 0);
+ if (ret == 0)
+ sha1_final(lte->hash, &sha_ctx);
- out_wim_ctype = wimlib_get_compression_type(w);
- output_res_entry = <e->output_resource_entry;
-
- /* Figure out if we can read the resource from the WIM file, or
- * if we have to read it from the filesystem outside. */
- if (lte->file_on_disk) {
-
- /* Read from disk (uncompressed) */
-
- len = lte->resource_entry.original_size;
+ return ret;
+}
- in = fopen(lte->file_on_disk, "rb");
- if (!in) {
- ERROR("Failed to open the file `%s': %m\n",
- lte->file_on_disk);
- return WIMLIB_ERR_OPEN;
- }
+/* Convert a WIM resource header to a stand-alone resource specification. */
+void
+wim_res_hdr_to_spec(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ struct wim_resource_spec *spec)
+{
+ spec->wim = wim;
+ spec->offset_in_wim = reshdr->offset_in_wim;
+ spec->size_in_wim = reshdr->size_in_wim;
+ spec->uncompressed_size = reshdr->uncompressed_size;
+ INIT_LIST_HEAD(&spec->lte_list);
+ spec->flags = reshdr->flags;
+ spec->is_pipable = wim_is_pipable(wim);
+ if (spec->flags & (WIM_RESHDR_FLAG_COMPRESSED | WIM_RESHDR_FLAG_CONCAT)) {
+ spec->ctype = wim->compression_type;
+ spec->cchunk_size = wim->chunk_size;
+ } else {
+ spec->ctype = WIMLIB_COMPRESSION_TYPE_NONE;
+ spec->cchunk_size = 0;
+ }
+}
- if (w->verbose)
- puts(lte->file_on_disk);
+/* Convert a stand-alone resource specification to a WIM resource header. */
+void
+wim_res_spec_to_hdr(const struct wim_resource_spec *rspec,
+ struct wim_reshdr *reshdr)
+{
+ reshdr->offset_in_wim = rspec->offset_in_wim;
+ reshdr->size_in_wim = rspec->size_in_wim;
+ reshdr->flags = rspec->flags;
+ reshdr->uncompressed_size = rspec->uncompressed_size;
+}
- ret = transfer_file_resource(in, len, len, 0,
- WIM_COMPRESSION_TYPE_NONE, out,
- out_wim_ctype, output_res_entry);
- fclose(in);
- } else {
+/* Translates a WIM resource header from the on-disk format into an in-memory
+ * format. */
+int
+get_wim_reshdr(const struct wim_reshdr_disk *disk_reshdr,
+ struct wim_reshdr *reshdr)
+{
+ reshdr->offset_in_wim = le64_to_cpu(disk_reshdr->offset_in_wim);
+ reshdr->size_in_wim = (((u64)disk_reshdr->size_in_wim[0] << 0) |
+ ((u64)disk_reshdr->size_in_wim[1] << 8) |
+ ((u64)disk_reshdr->size_in_wim[2] << 16) |
+ ((u64)disk_reshdr->size_in_wim[3] << 24) |
+ ((u64)disk_reshdr->size_in_wim[4] << 32) |
+ ((u64)disk_reshdr->size_in_wim[5] << 40) |
+ ((u64)disk_reshdr->size_in_wim[6] << 48));
+ reshdr->uncompressed_size = le64_to_cpu(disk_reshdr->uncompressed_size);
+ reshdr->flags = disk_reshdr->flags;
+
+ /* Truncate numbers to 62 bits to avoid possible overflows. */
+ if (reshdr->offset_in_wim & 0xc000000000000000ULL)
+ return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
+
+ if (reshdr->uncompressed_size & 0xc000000000000000ULL)
+ return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
- /* Read from input WIM (possibly compressed) */
+ return 0;
+}
- /* It may be a different WIM file, in the case of
- * exporting images from one WIM file to another */
- if (lte->other_wim_fp) {
- /* Different WIM file. */
- in = lte->other_wim_fp;
- in_wim_ctype = lte->other_wim_ctype;
- } else {
- /* Same WIM file. */
- in = w->fp;
- in_wim_ctype = out_wim_ctype;
- }
- input_res_entry = <e->resource_entry;
- input_res_ctype = resource_compression_type(
- in_wim_ctype,
- input_res_entry->flags);
-
- ret = transfer_file_resource(in,
- input_res_entry->size,
- input_res_entry->original_size,
- input_res_entry->offset,
- input_res_ctype,
- out,
- out_wim_ctype,
- output_res_entry);
- }
- return ret;
+/* Translates a WIM resource header from an in-memory format into the on-disk
+ * format. */
+void
+put_wim_reshdr(const struct wim_reshdr *reshdr,
+ struct wim_reshdr_disk *disk_reshdr)
+{
+ disk_reshdr->size_in_wim[0] = reshdr->size_in_wim >> 0;
+ disk_reshdr->size_in_wim[1] = reshdr->size_in_wim >> 8;
+ disk_reshdr->size_in_wim[2] = reshdr->size_in_wim >> 16;
+ disk_reshdr->size_in_wim[3] = reshdr->size_in_wim >> 24;
+ disk_reshdr->size_in_wim[4] = reshdr->size_in_wim >> 32;
+ disk_reshdr->size_in_wim[5] = reshdr->size_in_wim >> 40;
+ disk_reshdr->size_in_wim[6] = reshdr->size_in_wim >> 48;
+ disk_reshdr->flags = reshdr->flags;
+ disk_reshdr->offset_in_wim = cpu_to_le64(reshdr->offset_in_wim);
+ disk_reshdr->uncompressed_size = cpu_to_le64(reshdr->uncompressed_size);
}