/*
* resource.c
*
- * Read uncompressed and compressed metadata and file resources.
- *
- * Copyright (C) 2010 Carl Thijssen
- * Copyright (C) 2012 Eric Biggers
- *
- * wimlib - Library for working with WIM files
+ * Code for reading blobs and resources, including compressed WIM resources.
+ */
+
+/*
+ * Copyright (C) 2012, 2013, 2015 Eric Biggers
*
- * This library is free software; you can redistribute it and/or modify it under
+ * This file 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
+ * 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.
+ * This file 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.
*
- * 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 Lesser General Public License
+ * along with this file; 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>
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
#include <errno.h>
+#include <fcntl.h>
+#include <unistd.h>
+
+#include "wimlib/alloca.h"
+#include "wimlib/assert.h"
+#include "wimlib/bitops.h"
+#include "wimlib/blob_table.h"
+#include "wimlib/endianness.h"
+#include "wimlib/error.h"
+#include "wimlib/file_io.h"
+#include "wimlib/resource.h"
+#include "wimlib/sha1.h"
+#include "wimlib/wim.h"
+
+#ifdef __WIN32__
+/* for read_winnt_stream_prefix(), read_win32_encrypted_file_prefix() */
+# include "wimlib/win32.h"
+#endif
+
+#ifdef WITH_NTFS_3G
+/* for read_ntfs_attribute_prefix() */
+# include "wimlib/ntfs_3g.h"
+#endif
-/*
- * 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.
*
- * @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.
+ * Additional information:
*
- * Returns zero on success, nonzero on failure.
+ * - 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.
+ *
+ * - The chunk table is included in the compressed size of the resource provided
+ * in the corresponding entry in the WIM's blob table.
+ *
+ * - 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[])
-{
- 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.
- */
+struct data_range {
+ u64 offset;
+ u64 size;
+};
- /* 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;
+/*
+ * read_compressed_wim_resource() -
+ *
+ * Read data from a compressed WIM resource.
+ *
+ * @rdesc
+ * Description of the compressed WIM resource to read from.
+ * @ranges
+ * Nonoverlapping, nonempty ranges of the uncompressed resource data to
+ * read, sorted by increasing offset.
+ * @num_ranges
+ * Number of ranges in @ranges; must be at least 1.
+ * @cb
+ * Callback function to feed the data being read. Each call provides the
+ * next chunk of the requested data, uncompressed. Each chunk will be of
+ * nonzero size and will not cross range boundaries, but otherwise will be
+ * of unspecified size.
+ * @cb_ctx
+ * Parameter to pass to @cb_ctx.
+ *
+ * Possible 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.
+ */
+static int
+read_compressed_wim_resource(const struct wim_resource_descriptor * const rdesc,
+ const struct data_range * const ranges,
+ const size_t num_ranges,
+ const consume_data_callback_t cb,
+ void * const cb_ctx)
+{
+ 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;
+ struct wimlib_decompressor *decompressor = NULL;
+
+ /* Sanity checks */
+ wimlib_assert(rdesc != NULL);
+ wimlib_assert(resource_is_compressed(rdesc));
+ wimlib_assert(cb != NULL);
+ wimlib_assert(num_ranges != 0);
+ for (size_t i = 0; i < num_ranges; i++) {
+ DEBUG("Range %zu/%zu: %"PRIu64"@+%"PRIu64" / %"PRIu64,
+ i + 1, num_ranges, ranges[i].size, ranges[i].offset,
+ rdesc->uncompressed_size);
+ 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 <= rdesc->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 = &rdesc->wim->in_fd;
+
+ /* Determine if we're reading a pipable resource from a pipe or not. */
+ const bool is_pipe_read = (rdesc->is_pipable && !filedes_is_seekable(in_fd));
+
+ /* Determine if the chunk table is in an alternate format. */
+ const bool alt_chunk_table = (rdesc->flags & WIM_RESHDR_FLAG_SOLID)
+ && !is_pipe_read;
+
+ /* Get the maximum size of uncompressed chunks in this resource, which
+ * we require be a power of 2. */
+ u64 cur_read_offset = rdesc->offset_in_wim;
+ int ctype = rdesc->compression_type;
+ u32 chunk_size = rdesc->chunk_size;
+ if (alt_chunk_table) {
+ /* Alternate chunk table format. Its header specifies the chunk
+ * size and compression format. Note: it could be read here;
+ * however, the relevant data was already loaded into @rdesc by
+ * read_blob_table(). */
+ cur_read_offset += sizeof(struct alt_chunk_table_header_disk);
}
- /* Number of bytes we need to read from the chunk table. */
- size_t size = num_needed_chunk_entries * chunk_entry_size;
+ if (!is_power_of_2(chunk_size)) {
+ ERROR("Invalid compressed resource: "
+ "expected power-of-2 chunk size (got %"PRIu32")",
+ chunk_size);
+ ret = WIMLIB_ERR_INVALID_CHUNK_SIZE;
+ errno = EINVAL;
+ goto out_free_memory;
+ }
- u8 chunk_tab_buf[size];
+ /* Get valid decompressor. */
+ if (ctype == rdesc->wim->decompressor_ctype &&
+ chunk_size == rdesc->wim->decompressor_max_block_size)
+ {
+ /* Cached decompressor. */
+ decompressor = rdesc->wim->decompressor;
+ rdesc->wim->decompressor_ctype = WIMLIB_COMPRESSION_TYPE_NONE;
+ rdesc->wim->decompressor = NULL;
+ } else {
+ ret = wimlib_create_decompressor(ctype, chunk_size,
+ &decompressor);
+ if (ret) {
+ if (ret != WIMLIB_ERR_NOMEM)
+ errno = EINVAL;
+ goto out_free_memory;
+ }
+ }
- if (fread(chunk_tab_buf, 1, size, fp) != size)
- goto err;
+ const u32 chunk_order = fls32(chunk_size);
+
+ /* Calculate the total number of chunks the resource is divided into. */
+ const u64 num_chunks = (rdesc->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. */
+ const u64 chunk_entry_size = get_chunk_entry_size(rdesc->uncompressed_size,
+ alt_chunk_table);
+
+ /* Calculate the size of the chunk table in bytes. */
+ const u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
+
+ /* Calculate the size of the chunk table in bytes, including the header
+ * in the case of the alternate chunk table format. */
+ 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. */
- /* Now fill in chunk_offsets from the entries we have read in
- * chunk_tab_buf. */
+ if (read_start_chunk == 0)
+ first_chunk_entry_to_read = 0;
+ else
+ first_chunk_entry_to_read = read_start_chunk - 1;
- u64 *chunk_tab_p = chunk_offsets;
- if (start_chunk == 0)
- chunk_tab_p++;
+ if (last_needed_chunk == 0)
+ last_chunk_entry_to_read = 0;
+ else
+ last_chunk_entry_to_read = last_needed_chunk - 1;
- if (chunk_entry_size == 4) {
- u32 *entries = (u32*)chunk_tab_buf;
- while (num_needed_chunk_entries--)
- *chunk_tab_p++ = to_le32(*entries++);
+ if (last_needed_chunk < num_chunks - 1)
+ last_chunk_entry_to_read++;
+ }
+
+ const u64 num_chunk_entries_to_read =
+ last_chunk_entry_to_read - first_chunk_entry_to_read + 1;
+
+ const u64 chunk_offsets_alloc_size =
+ max(num_chunk_entries_to_read,
+ num_needed_chunk_offsets) * sizeof(chunk_offsets[0]);
+
+ if ((size_t)chunk_offsets_alloc_size != chunk_offsets_alloc_size)
+ goto oom;
+
+ 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)
+ + (rdesc->is_pipable ? (rdesc->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_to_read,
+ 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 _may_alias_attribute aliased_le64_t;
+ typedef le32 _may_alias_attribute 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;
+ }
+ if (last_needed_chunk < num_chunks - 1)
+ *chunk_offsets_p = cur_offset;
+ } 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 (rdesc->is_pipable)
+ cur_read_offset += read_start_chunk * sizeof(struct pwm_chunk_hdr);
+ else
+ cur_read_offset += chunk_table_size;
+ }
+
+ /* Allocate buffer for holding the uncompressed data of each chunk. */
+ if (chunk_size <= STACK_MAX) {
+ ubuf = alloca(chunk_size);
} else {
- u64 *entries = (u64*)chunk_tab_buf;
- while (num_needed_chunk_entries--)
- *chunk_tab_p++ = to_le64(*entries++);
+ ubuf = MALLOC(chunk_size);
+ if (ubuf == NULL)
+ goto oom;
+ ubuf_malloced = true;
}
- /* 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 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 (chunk_size - 1 <= STACK_MAX) {
+ cbuf = alloca(chunk_size - 1);
+ } else {
+ cbuf = MALLOC(chunk_size - 1);
+ if (cbuf == NULL)
+ goto oom;
+ cbuf_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;
+ /* Set current data range. */
+ 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;
+
+ /* Read and process each needed chunk. */
+ for (u64 i = read_start_chunk; i <= last_needed_chunk; i++) {
+
+ /* Calculate uncompressed size of next chunk. */
+ u32 chunk_usize;
+ if ((i == num_chunks - 1) && (rdesc->uncompressed_size & (chunk_size - 1)))
+ chunk_usize = (rdesc->uncompressed_size & (chunk_size - 1));
+ else
+ 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 {
- /* 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;
+ if (i == num_chunks - 1) {
+ chunk_csize = rdesc->size_in_wim -
+ chunk_table_full_size -
+ chunk_offsets[i - read_start_chunk];
+ if (rdesc->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 (rdesc->is_pipable)
+ cur_read_offset += sizeof(struct pwm_chunk_hdr);
- DEBUG2("compressed_chunk_size = %u, uncompressed_chunk_size = %u\n",
- compressed_chunk_size, uncompressed_chunk_size);
+ /* Offsets in the uncompressed resource at which this chunk
+ * starts and ends. */
+ 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) {
- /* 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;
- 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;
- }
+ /* 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);
- } 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;
+ /* Read the chunk and feed data to the callback
+ * function. */
+ u8 *read_buf;
+
+ if (chunk_csize == chunk_usize)
+ read_buf = ubuf;
+ else
+ read_buf = cbuf;
+
+ ret = full_pread(in_fd,
+ read_buf,
+ chunk_csize,
+ cur_read_offset);
+ if (ret)
+ goto read_error;
+
+ if (read_buf == cbuf) {
+ DEBUG("Decompressing chunk %"PRIu64" "
+ "(csize=%"PRIu32" usize=%"PRIu32")",
+ i, chunk_csize, chunk_usize);
+ ret = wimlib_decompress(cbuf,
+ chunk_csize,
+ ubuf,
+ chunk_usize,
+ decompressor);
+ if (ret) {
+ ERROR("Failed to decompress data!");
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ errno = EINVAL;
+ goto out_free_memory;
+ }
}
+ cur_read_offset += chunk_csize;
+
+ /* At least one range requires data in this chunk. */
+ do {
+ size_t start, end, size;
+
+ /* Calculate how many bytes of data should be
+ * sent to the callback function, taking into
+ * account that data sent to the callback
+ * function must not overlap range boundaries.
+ */
+ start = cur_range_pos - chunk_start_offset;
+ end = min(cur_range_end, chunk_end_offset) - chunk_start_offset;
+ size = end - start;
+
+ ret = (*cb)(&ubuf[start], size, cb_ctx);
+
+ if (ret)
+ goto out_free_memory;
+
+ cur_range_pos += size;
+ if (cur_range_pos == cur_range_end) {
+ /* Advance to next range. */
+ 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 == rdesc->uncompressed_size - 1 &&
+ chunk_table_size)
+ {
+ u8 dummy;
+ /* If reading a pipable resource from a pipe and the full data
+ * was requested, skip the chunk table at the end so that the
+ * file descriptor is fully clear of the resource after this
+ * returns. */
+ 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 (decompressor) {
+ wimlib_free_decompressor(rdesc->wim->decompressor);
+ rdesc->wim->decompressor = decompressor;
+ rdesc->wim->decompressor_ctype = ctype;
+ rdesc->wim->decompressor_max_block_size = chunk_size;
+ }
+ 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 WIM 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 WIM 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[])
+static int
+fill_zeroes(u64 size, consume_data_callback_t cb, void *cb_ctx)
{
- 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 (unlikely(size)) {
+ u8 buf[min(size, BUFFER_SIZE)];
+
+ memset(buf, 0, sizeof(buf));
+
+ do {
+ size_t len;
+ int ret;
+
+ len = min(size, BUFFER_SIZE);
+ ret = cb(buf, len, cb_ctx);
+ if (ret)
+ return ret;
+ size -= len;
+ } while (size);
}
- if (fread(contents_ret, 1, len, fp) != len) {
- if (feof(fp)) {
- ERROR("Unexpected EOF in uncompressed file resource!\n");
- } else {
- ERROR("Failed to read %"PRIu64" bytes from "
- "uncompressed resource at offset "
- "%"PRIu64"\n", len, offset);
+ return 0;
+}
+
+/* Read raw data from a file descriptor at the specified offset, feeding the
+ * data it in chunks into the specified callback function. */
+static int
+read_raw_file_data(struct filedes *in_fd, u64 offset, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
+{
+ u8 buf[BUFFER_SIZE];
+ size_t bytes_to_read;
+ int ret;
+
+ while (size) {
+ bytes_to_read = min(sizeof(buf), size);
+ ret = full_pread(in_fd, buf, bytes_to_read, offset);
+ if (ret) {
+ ERROR_WITH_ERRNO("Read error");
+ return ret;
}
- return WIMLIB_ERR_READ;
+ ret = cb(buf, bytes_to_read, cb_ctx);
+ if (ret)
+ return ret;
+ size -= bytes_to_read;
+ offset += bytes_to_read;
}
return 0;
}
+/* A consume_data_callback_t implementation that simply concatenates all chunks
+ * into a buffer. */
+static int
+bufferer_cb(const void *chunk, size_t size, void *_ctx)
+{
+ u8 **buf_p = _ctx;
+
+ *buf_p = mempcpy(*buf_p, chunk, size);
+ return 0;
+}
-/*
- * Reads a WIM resource.
+/*
+ * read_partial_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.
+ * Read a range of data from an uncompressed or compressed resource in a WIM
+ * file.
*
- * To read the whole file resource, specify offset =
- * 0 and len = resource_original_size, or call
- * read_full_resource().
+ * @rdesc
+ * Description of the WIM resource to read from.
+ * @offset
+ * Offset within the uncompressed resource at which to start reading.
+ * @size
+ * Number of bytes to read.
+ * @cb
+ * Callback function to feed the data being read. Each call provides the
+ * next chunk of the requested data, uncompressed. Each chunk will be of
+ * nonzero size and will not cross range boundaries, but otherwise will be
+ * of unspecified size.
+ * @cb_ctx
+ * Parameter to pass to @cb_ctx.
*
- * @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 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)
*
- * @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.
+ * or other error code returned by the @cb function.
*/
-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)
+static int
+read_partial_wim_resource(const struct wim_resource_descriptor *rdesc,
+ u64 offset, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
- 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;
- }
- 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);
- }
-}
+ /* Sanity checks. */
+ wimlib_assert(offset + size >= offset);
+ wimlib_assert(offset + size <= rdesc->uncompressed_size);
+ DEBUG("Reading %"PRIu64" @ %"PRIu64" from WIM resource "
+ "%"PRIu64" => %"PRIu64" @ %"PRIu64,
+ size, offset, rdesc->uncompressed_size,
+ rdesc->size_in_wim, rdesc->offset_in_wim);
-/*
- * 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)
-{
- u64 num_chunks;
- u64 n;
- u8 buf[min(size, WIM_CHUNK_SIZE)];
- int res_ctype;
- u64 offset;
- u64 i;
- int ret;
+ /* Trivial case. */
+ if (size == 0)
+ return 0;
- 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;
+ if (resource_is_compressed(rdesc)) {
+ struct data_range range = {
+ .offset = offset,
+ .size = size,
+ };
+ return read_compressed_wim_resource(rdesc, &range, 1,
+ cb, cb_ctx);
+ } else {
+ /* Reading uncompressed resource. For completeness, handle the
+ * weird case where size_in_wim < uncompressed_size. */
+
+ u64 read_size;
+ u64 zeroes_size;
+ int ret;
+
+ if (likely(offset + size <= rdesc->size_in_wim) ||
+ rdesc->is_pipable)
+ {
+ read_size = size;
+ zeroes_size = 0;
+ } else {
+ if (offset >= rdesc->size_in_wim) {
+ read_size = 0;
+ zeroes_size = size;
+ } else {
+ read_size = rdesc->size_in_wim - offset;
+ zeroes_size = offset + size - rdesc->size_in_wim;
}
}
- ret = read_resource(w->fp, entry->size, entry->original_size,
- entry->offset, res_ctype, n, offset, buf);
- if (ret != 0)
+ ret = read_raw_file_data(&rdesc->wim->in_fd,
+ rdesc->offset_in_wim + offset,
+ read_size,
+ cb,
+ cb_ctx);
+ if (ret)
return ret;
- if (full_write(fd, buf, n) != n)
- return WIMLIB_ERR_WRITE;
- offset += n;
+ return fill_zeroes(zeroes_size, cb, cb_ctx);
}
- return ret;
}
-/*
- * 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.
- *
- * The output_resource_entry, out_refcnt, and part_number fields of @lte are
- * updated.
- *
- * Metadata resources are not copied (they are handled elsewhere for joining and
- * splitting).
- */
-int copy_resource(struct lookup_table_entry *lte, void *w)
+/* Read the specified range of uncompressed data from the specified blob, which
+ * must be located into a WIM file, into the specified buffer. */
+int
+read_partial_wim_blob_into_buf(const struct blob_descriptor *blob,
+ size_t size, u64 offset, void *_buf)
{
- if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
- !((WIMStruct*)w)->write_metadata) {
- return 0;
- }
-
- 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);
+ u8 *buf = _buf;
- if (new_offset == -1)
- return WIMLIB_ERR_WRITE;
+ wimlib_assert(blob->blob_location == BLOB_IN_WIM);
- ret = copy_between_files(in_fp, offset, out_fp, size);
- if (ret != 0)
- return ret;
-
- memcpy(<e->output_resource_entry, <e->resource_entry,
- sizeof(struct resource_entry));
+ return read_partial_wim_resource(blob->rdesc,
+ blob->offset_in_res + offset,
+ size,
+ bufferer_cb,
+ &buf);
+}
- lte->output_resource_entry.offset = new_offset;
- lte->out_refcnt = lte->refcnt;
- lte->part_number = ((WIMStruct*)w)->hdr.part_number;
+/* A consume_data_callback_t implementation that simply ignores the data
+ * received. */
+static int
+skip_chunk_cb(const void *chunk, size_t size, void *_ctx)
+{
return 0;
}
-/* 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)
+/* Skip over the data of the specified WIM resource. */
+int
+skip_wim_resource(struct wim_resource_descriptor *rdesc)
{
- 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;
+ DEBUG("Skipping resource (size=%"PRIu64")", rdesc->uncompressed_size);
+ return read_partial_wim_resource(rdesc, 0, rdesc->uncompressed_size,
+ skip_chunk_cb, NULL);
}
-/* 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)
+static int
+read_wim_blob_prefix(const struct blob_descriptor *blob, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
- 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(blob->rdesc, blob->offset_in_res, size,
+ cb, cb_ctx);
}
-/* 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)
+/* This function handles reading blob 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.
+ *
+ * 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. */
+static int
+read_file_on_disk_prefix(const struct blob_descriptor *blob, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
- 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;
+ int ret;
+ int raw_fd;
+ struct filedes fd;
+
+ DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", size, blob->file_on_disk);
+
+ raw_fd = topen(blob->file_on_disk, O_BINARY | O_RDONLY);
+ if (raw_fd < 0) {
+ ERROR_WITH_ERRNO("Can't open \"%"TS"\"", blob->file_on_disk);
+ return WIMLIB_ERR_OPEN;
}
+ filedes_init(&fd, raw_fd);
+ ret = read_raw_file_data(&fd, 0, size, cb, cb_ctx);
+ filedes_close(&fd);
+ return ret;
}
+#ifdef WITH_FUSE
+static int
+read_staging_file_prefix(const struct blob_descriptor *blob, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
+{
+ int raw_fd;
+ struct filedes fd;
+ int ret;
+
+ DEBUG("Reading %"PRIu64" bytes from staging file \"%s\"",
+ size, blob->staging_file_name);
+ raw_fd = openat(blob->staging_dir_fd, blob->staging_file_name,
+ O_RDONLY | O_NOFOLLOW);
+ if (raw_fd < 0) {
+ ERROR_WITH_ERRNO("Can't open staging file \"%s\"",
+ blob->staging_file_name);
+ return WIMLIB_ERR_OPEN;
+ }
+ filedes_init(&fd, raw_fd);
+ ret = read_raw_file_data(&fd, 0, size, cb, cb_ctx);
+ filedes_close(&fd);
+ return ret;
+}
+#endif
+
+/* This function handles the trivial case of reading blob data that is, in fact,
+ * already located in an in-memory buffer. */
+static int
+read_buffer_prefix(const struct blob_descriptor *blob,
+ u64 size, consume_data_callback_t cb, void *cb_ctx)
+{
+ return (*cb)(blob->attached_buffer, size, cb_ctx);
+}
+
+typedef int (*read_blob_prefix_handler_t)(const struct blob_descriptor *blob,
+ u64 size,
+ consume_data_callback_t cb,
+ void *cb_ctx);
/*
- * Copies bytes between two file streams.
+ * read_blob_prefix()-
+ *
+ * Reads the first @size bytes from a generic "blob", 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 to a callback function @cb in chunks of
+ * unspecified size.
*
- * Copies @len bytes from @in to @out, at the current position in @out, and at
- * an offset of @in_offset in @in.
+ * Returns 0 on success; nonzero on error. A nonzero value will be returned if
+ * the blob data cannot be successfully read (for a number of different reasons,
+ * depending on the blob location), or if @cb returned nonzero in which case
+ * that error code will be returned.
*/
-int copy_between_files(FILE *in, off_t in_offset, FILE *out, size_t len)
+static int
+read_blob_prefix(const struct blob_descriptor *blob, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
- u8 buf[BUFFER_SIZE];
- size_t n;
+ static const read_blob_prefix_handler_t handlers[] = {
+ [BLOB_IN_WIM] = read_wim_blob_prefix,
+ [BLOB_IN_FILE_ON_DISK] = read_file_on_disk_prefix,
+ [BLOB_IN_ATTACHED_BUFFER] = read_buffer_prefix,
+ #ifdef WITH_FUSE
+ [BLOB_IN_STAGING_FILE] = read_staging_file_prefix,
+ #endif
+ #ifdef WITH_NTFS_3G
+ [BLOB_IN_NTFS_VOLUME] = read_ntfs_attribute_prefix,
+ #endif
+ #ifdef __WIN32__
+ [BLOB_IN_WINNT_FILE_ON_DISK] = read_winnt_stream_prefix,
+ [BLOB_WIN32_ENCRYPTED] = read_win32_encrypted_file_prefix,
+ #endif
+ };
+ wimlib_assert(blob->blob_location < ARRAY_LEN(handlers)
+ && handlers[blob->blob_location] != NULL);
+ wimlib_assert(size <= blob->size);
+ return handlers[blob->blob_location](blob, size, cb, cb_ctx);
+}
- 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;
+/* Read the full uncompressed data of the specified blob into the specified
+ * buffer, which must have space for at least blob->size bytes. */
+int
+read_full_blob_into_buf(const struct blob_descriptor *blob, void *_buf)
+{
+ u8 *buf = _buf;
+ return read_blob_prefix(blob, blob->size, bufferer_cb, &buf);
+}
+
+/* Retrieve the full uncompressed data of the specified blob. A buffer large
+ * enough hold the data is allocated and returned in @buf_ret. */
+int
+read_full_blob_into_alloc_buf(const struct blob_descriptor *blob, void **buf_ret)
+{
+ int ret;
+ void *buf;
+
+ if ((size_t)blob->size != blob->size) {
+ ERROR("Can't read %"PRIu64" byte blob into memory", blob->size);
+ return WIMLIB_ERR_NOMEM;
}
- /* 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;
- }
- if (fwrite(buf, 1, n, out) != n) {
- ERROR("Error copying data between files: %m\n");
- return WIMLIB_ERR_WRITE;
- }
- len -= n;
+ buf = MALLOC(blob->size);
+ if (buf == NULL)
+ return WIMLIB_ERR_NOMEM;
+
+ ret = read_full_blob_into_buf(blob, buf);
+ if (ret) {
+ FREE(buf);
+ return ret;
}
+
+ *buf_ret = buf;
return 0;
}
+/* Retrieve the full uncompressed data of a WIM resource specified as a raw
+ * `wim_reshdr' and the corresponding WIM file. A buffer large enough hold the
+ * data is allocated and returned in @buf_ret. */
+int
+wim_reshdr_to_data(const struct wim_reshdr *reshdr, WIMStruct *wim, void **buf_ret)
+{
+ struct wim_resource_descriptor rdesc;
+ struct blob_descriptor blob;
-/*
- * Uncompresses a WIM file resource and writes it uncompressed to a file stream.
- *
- * @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.
- */
-static int uncompress_resource(FILE *in, u64 size, u64 original_size,
- off_t offset, int input_ctype, FILE *out)
+ wim_res_hdr_to_desc(reshdr, wim, &rdesc);
+ blob_set_is_located_in_nonsolid_wim_resource(&blob, &rdesc);
+
+ return read_full_blob_into_alloc_buf(&blob, buf_ret);
+}
+
+int
+wim_reshdr_to_hash(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ u8 hash[SHA1_HASH_SIZE])
{
+ struct wim_resource_descriptor rdesc;
+ struct blob_descriptor blob;
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;
- if (fwrite(buf, 1, uncompressed_chunk_size, out) !=
- uncompressed_chunk_size) {
- ERROR("Failed to write file resource: %m\n");
- return WIMLIB_ERR_WRITE;
- }
- }
+ wim_res_hdr_to_desc(reshdr, wim, &rdesc);
+ blob_set_is_located_in_nonsolid_wim_resource(&blob, &rdesc);
+ blob.unhashed = 1;
+
+ ret = sha1_blob(&blob);
+ if (ret)
+ return ret;
+ copy_hash(hash, blob.hash);
return 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)
+struct blobifier_context {
+ struct read_blob_list_callbacks cbs;
+ struct blob_descriptor *cur_blob;
+ struct blob_descriptor *next_blob;
+ u64 cur_blob_offset;
+ struct blob_descriptor *final_blob;
+ size_t list_head_offset;
+};
+
+static struct blob_descriptor *
+next_blob(struct blob_descriptor *blob, size_t list_head_offset)
{
- int ret;
- int (*compress)(const void *, uint, void *, uint *);
- if (output_ctype == WIM_COMPRESSION_TYPE_LZX)
- compress = lzx_compress;
- else
- compress = xpress_compress;
+ struct list_head *cur;
- u8 uncompressed_buf[WIM_CHUNK_SIZE];
- u8 compressed_buf[WIM_CHUNK_SIZE - 1];
+ cur = (struct list_head*)((u8*)blob + list_head_offset);
- /* Determine how many compressed chunks the file needs to be divided
- * into. */
- u64 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
+ return (struct blob_descriptor*)((u8*)cur->next - list_head_offset);
+}
- u64 num_chunk_entries = num_chunks - 1;
+/* A consume_data_callback_t implementation that translates raw resource data
+ * into blobs, calling the begin_blob, consume_chunk, and end_blob callback
+ * functions as appropriate. */
+static int
+blobifier_cb(const void *chunk, size_t size, void *_ctx)
+{
+ struct blobifier_context *ctx = _ctx;
+ int ret;
- /* 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;
+ DEBUG("%zu bytes passed to blobifier", size);
- /* Array in which to construct the chunk offset table. */
- u64 chunk_offsets[num_chunk_entries];
+ wimlib_assert(ctx->cur_blob != NULL);
+ wimlib_assert(size <= ctx->cur_blob->size - ctx->cur_blob_offset);
- /* Offset of the start of the chunk table in the output file. */
- off_t chunk_tab_offset = ftello(out);
+ if (ctx->cur_blob_offset == 0) {
- /* Total size of the chunk table (as written to the file) */
- u64 chunk_tab_size = chunk_entry_size * num_chunk_entries;
+ /* Starting a new blob. */
+ DEBUG("Begin new blob (size=%"PRIu64").", ctx->cur_blob->size);
- /* 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;
+ ret = (*ctx->cbs.begin_blob)(ctx->cur_blob,
+ ctx->cbs.begin_blob_ctx);
+ if (ret)
+ return ret;
}
- /* 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++) {
+ /* Consume the chunk. */
+ ret = (*ctx->cbs.consume_chunk)(chunk, size,
+ ctx->cbs.consume_chunk_ctx);
+ ctx->cur_blob_offset += size;
+ if (ret)
+ return ret;
- u64 uncompressed_offset = i * WIM_CHUNK_SIZE;
- u64 uncompressed_chunk_size = min(WIM_CHUNK_SIZE,
- original_size - uncompressed_offset);
+ if (ctx->cur_blob_offset == ctx->cur_blob->size) {
+ /* Finished reading all the data for a blob. */
- const u8 *uncompressed_p;
- if (uncompressed_resource != NULL) {
- uncompressed_p = uncompressed_resource +
- uncompressed_offset;
+ ctx->cur_blob_offset = 0;
- } 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;
+ DEBUG("End blob (size=%"PRIu64").", ctx->cur_blob->size);
+ ret = (*ctx->cbs.end_blob)(ctx->cur_blob, 0,
+ ctx->cbs.end_blob_ctx);
+ if (ret)
+ return ret;
+
+ /* Advance to next blob. */
+ ctx->cur_blob = ctx->next_blob;
+ if (ctx->cur_blob != NULL) {
+ if (ctx->cur_blob != ctx->final_blob)
+ ctx->next_blob = next_blob(ctx->cur_blob,
+ ctx->list_head_offset);
+ else
+ ctx->next_blob = NULL;
}
+ }
+ return 0;
+}
- if (i != 0)
- chunk_offsets[i - 1] = cur_chunk_offset;
+struct hasher_context {
+ SHA_CTX sha_ctx;
+ int flags;
+ struct read_blob_list_callbacks cbs;
+};
- uint compressed_len;
+/* Callback for starting to read a blob while calculating its SHA-1 message
+ * digest. */
+static int
+hasher_begin_blob(struct blob_descriptor *blob, void *_ctx)
+{
+ struct hasher_context *ctx = _ctx;
- ret = compress(uncompressed_p, uncompressed_chunk_size,
- compressed_buf, &compressed_len);
+ sha1_init(&ctx->sha_ctx);
- /* 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;
- }
+ if (ctx->cbs.begin_blob == NULL)
+ return 0;
+ else
+ return (*ctx->cbs.begin_blob)(blob, ctx->cbs.begin_blob_ctx);
+}
- 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;
- }
+/* A consume_data_callback_t implementation that continues calculating the SHA-1
+ * message digest of the blob being read, then optionally passes the data on to
+ * another consume_data_callback_t implementation. This allows checking the
+ * SHA-1 message digest of a blob being extracted, for example. */
+static int
+hasher_consume_chunk(const void *chunk, size_t size, void *_ctx)
+{
+ struct hasher_context *ctx = _ctx;
- /* 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;
+ sha1_update(&ctx->sha_ctx, chunk, size);
+ if (ctx->cbs.consume_chunk == NULL)
+ return 0;
+ else
+ return (*ctx->cbs.consume_chunk)(chunk, size, ctx->cbs.consume_chunk_ctx);
+}
- /* Now that all entries of the chunk table are determined, rewind the
- * stream to where the chunk table was, and write it back out. */
+/* Callback for finishing reading a blob while calculating its SHA-1 message
+ * digest. */
+static int
+hasher_end_blob(struct blob_descriptor *blob, int status, void *_ctx)
+{
+ struct hasher_context *ctx = _ctx;
+ u8 hash[SHA1_HASH_SIZE];
+ int ret;
- if (fseeko(out, chunk_tab_offset, SEEK_SET) != 0) {
- ERROR("Failed to seek to beginning of chunk table: %m\n");
- return WIMLIB_ERR_READ;
+ if (status) {
+ /* Error occurred; the full blob may not have been read. */
+ ret = status;
+ goto out_next_cb;
}
- if (chunk_entry_size == 8) {
- array_to_le64(chunk_offsets, num_chunk_entries);
+ /* Retrieve the final SHA-1 message digest. */
+ sha1_final(hash, &ctx->sha_ctx);
- if (fwrite(chunk_offsets, 1, chunk_tab_size, out) !=
- chunk_tab_size) {
- ERROR("Failed to write chunk table: %m\n");
- return WIMLIB_ERR_WRITE;
+ if (blob->unhashed) {
+ if (ctx->flags & COMPUTE_MISSING_BLOB_HASHES) {
+ /* No SHA-1 message digest was previously present for the
+ * blob. Set it to the one just calculated. */
+ DEBUG("Set SHA-1 message digest for blob "
+ "(size=%"PRIu64").", blob->size);
+ copy_hash(blob->hash, hash);
}
} 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;
+ if (ctx->flags & VERIFY_BLOB_HASHES) {
+ /* The blob already had a SHA-1 message digest present.
+ * Verify that it is the same as the calculated value.
+ */
+ if (!hashes_equal(hash, blob->hash)) {
+ if (wimlib_print_errors) {
+ tchar expected_hashstr[SHA1_HASH_SIZE * 2 + 1];
+ tchar actual_hashstr[SHA1_HASH_SIZE * 2 + 1];
+ sprint_hash(blob->hash, expected_hashstr);
+ sprint_hash(hash, actual_hashstr);
+ ERROR("The data is corrupted!\n"
+ " (Expected SHA-1=%"TS",\n"
+ " got SHA-1=%"TS")",
+ expected_hashstr, actual_hashstr);
+ }
+ ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
+ errno = EINVAL;
+ goto out_next_cb;
+ }
+ DEBUG("SHA-1 message digest okay for "
+ "blob (size=%"PRIu64").", blob->size);
}
}
+ ret = 0;
+out_next_cb:
+ if (ctx->cbs.end_blob == NULL)
+ return ret;
+ else
+ return (*ctx->cbs.end_blob)(blob, ret, ctx->cbs.end_blob_ctx);
+}
- if (fseeko(out, 0, SEEK_END) != 0) {
- ERROR("Failed to seek to end of output file: %m\n");
- return WIMLIB_ERR_WRITE;
- }
+static int
+read_full_blob_with_cbs(struct blob_descriptor *blob,
+ const struct read_blob_list_callbacks *cbs)
+{
+ int ret;
- return 0;
+ ret = (*cbs->begin_blob)(blob, cbs->begin_blob_ctx);
+ if (ret)
+ return ret;
+
+ ret = read_blob_prefix(blob, blob->size, cbs->consume_chunk,
+ cbs->consume_chunk_ctx);
+
+ return (*cbs->end_blob)(blob, ret, cbs->end_blob_ctx);
}
-int write_resource_from_memory(const u8 resource[], int out_ctype,
- u64 resource_original_size, FILE *out,
- u64 *resource_size_ret)
+/* Read the full data of the specified blob, passing the data into the specified
+ * callbacks (all of which are optional) and either checking or computing the
+ * SHA-1 message digest of the blob. */
+static int
+read_full_blob_with_sha1(struct blob_descriptor *blob,
+ const struct read_blob_list_callbacks *cbs)
{
- 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);
- }
+ struct hasher_context hasher_ctx = {
+ .flags = VERIFY_BLOB_HASHES | COMPUTE_MISSING_BLOB_HASHES,
+ .cbs = *cbs,
+ };
+ struct read_blob_list_callbacks hasher_cbs = {
+ .begin_blob = hasher_begin_blob,
+ .begin_blob_ctx = &hasher_ctx,
+ .consume_chunk = hasher_consume_chunk,
+ .consume_chunk_ctx = &hasher_ctx,
+ .end_blob = hasher_end_blob,
+ .end_blob_ctx = &hasher_ctx,
+ };
+ return read_full_blob_with_cbs(blob, &hasher_cbs);
}
-
-/*
- * 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)
+static int
+read_blobs_in_solid_resource(struct blob_descriptor *first_blob,
+ struct blob_descriptor *last_blob,
+ u64 blob_count,
+ size_t list_head_offset,
+ const struct read_blob_list_callbacks *sink_cbs)
{
+ struct data_range *ranges;
+ bool ranges_malloced;
+ struct blob_descriptor *cur_blob;
+ size_t i;
int ret;
+ u64 ranges_alloc_size;
- /* Handle zero-length files */
- if (original_size == 0) {
- memset(output_res_entry, 0, sizeof(*output_res_entry));
- return 0;
- }
+ DEBUG("Reading %"PRIu64" blobs combined in same WIM resource",
+ blob_count);
- /* 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;
- }
+ /* Setup data ranges array (one range per blob to read); this way
+ * read_compressed_wim_resource() does not need to be aware of blobs.
+ */
- if (output_ctype == input_ctype) {
- /* The same compression types; simply copy the resource. */
+ ranges_alloc_size = blob_count * sizeof(ranges[0]);
- ret = copy_between_files(in, offset, out, size);
- if (ret != 0)
- return ret;
- output_res_entry->size = size;
+ if (unlikely((size_t)ranges_alloc_size != ranges_alloc_size)) {
+ ERROR("Too many blobs in one resource!");
+ return WIMLIB_ERR_NOMEM;
+ }
+ if (likely(ranges_alloc_size <= STACK_MAX)) {
+ ranges = alloca(ranges_alloc_size);
+ ranges_malloced = false;
} 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;
+ ranges = MALLOC(ranges_alloc_size);
+ if (ranges == NULL) {
+ ERROR("Too many blobs in one resource!");
+ return WIMLIB_ERR_NOMEM;
}
+ ranges_malloced = true;
+ }
+ for (i = 0, cur_blob = first_blob;
+ i < blob_count;
+ i++, cur_blob = next_blob(cur_blob, list_head_offset))
+ {
+ ranges[i].offset = cur_blob->offset_in_res;
+ ranges[i].size = cur_blob->size;
}
- 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;
- return 0;
+ struct blobifier_context blobifier_ctx = {
+ .cbs = *sink_cbs,
+ .cur_blob = first_blob,
+ .next_blob = next_blob(first_blob, list_head_offset),
+ .cur_blob_offset = 0,
+ .final_blob = last_blob,
+ .list_head_offset = list_head_offset,
+ };
+
+ ret = read_compressed_wim_resource(first_blob->rdesc,
+ ranges,
+ blob_count,
+ blobifier_cb,
+ &blobifier_ctx);
+
+ if (ranges_malloced)
+ FREE(ranges);
+
+ if (ret) {
+ if (blobifier_ctx.cur_blob_offset != 0) {
+ ret = (*blobifier_ctx.cbs.end_blob)
+ (blobifier_ctx.cur_blob,
+ ret,
+ blobifier_ctx.cbs.end_blob_ctx);
+ }
+ }
+ return ret;
}
-/*
- * 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 blobs, each of which may be in any supported location (e.g.
+ * in a WIM or in an external file). This function optimizes the case where
+ * multiple blobs are combined into a single solid compressed WIM resource by
+ * reading the blobs in sequential order, only decompressing the solid resource
+ * one time.
+ *
+ * @blob_list
+ * List of blobs to read.
+ * @list_head_offset
+ * Offset of the `struct list_head' within each `struct blob_descriptor' that makes up
+ * the @blob_list.
+ * @cbs
+ * Callback functions to accept the blob data.
+ * @flags
+ * Bitwise OR of zero or more of the following flags:
*
- * @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.
+ * VERIFY_BLOB_HASHES:
+ * For all blobs being read that have already had SHA-1 message
+ * digests computed, calculate the SHA-1 message digest of the read
+ * data and compare it with the previously computed value. If they
+ * do not match, return WIMLIB_ERR_INVALID_RESOURCE_HASH.
*
- * @return: True on success, false on failure.
+ * COMPUTE_MISSING_BLOB_HASHES
+ * For all blobs being read that have not yet had their SHA-1
+ * message digests computed, calculate and save their SHA-1 message
+ * digests.
+ *
+ * BLOB_LIST_ALREADY_SORTED
+ * @blob_list is already sorted in sequential order for reading.
+ *
+ * The callback functions are allowed to delete the current blob from the list
+ * if necessary.
+ *
+ * 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_blob_list(struct list_head *blob_list,
+ size_t list_head_offset,
+ const struct read_blob_list_callbacks *cbs,
+ int flags)
{
- u8 *buf;
- int ctype;
- u32 dentry_offset;
int ret;
- struct dentry *dentry = NULL;
-
- DEBUG("Reading metadata resource: length = %lu, offset = %lu\n",
- res_entry->original_size, res_entry->offset);
-
- if (res_entry->original_size < 8) {
- ERROR("Expected at least 8 bytes for the metadata "
- "resource!\n");
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
+ struct list_head *cur, *next;
+ struct blob_descriptor *blob;
+ struct hasher_context *hasher_ctx;
+ struct read_blob_list_callbacks *sink_cbs;
+
+ if (!(flags & BLOB_LIST_ALREADY_SORTED)) {
+ ret = sort_blob_list_by_sequential_order(blob_list, list_head_offset);
+ if (ret)
+ return ret;
}
- /* Allocate memory for the uncompressed metadata resource. */
- buf = MALLOC(res_entry->original_size);
-
- if (!buf) {
- ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
- "metadata resource!\n",
- res_entry->original_size);
- return WIMLIB_ERR_NOMEM;
+ if (flags & (VERIFY_BLOB_HASHES | COMPUTE_MISSING_BLOB_HASHES)) {
+ hasher_ctx = alloca(sizeof(*hasher_ctx));
+ *hasher_ctx = (struct hasher_context) {
+ .flags = flags,
+ .cbs = *cbs,
+ };
+ sink_cbs = alloca(sizeof(*sink_cbs));
+ *sink_cbs = (struct read_blob_list_callbacks) {
+ .begin_blob = hasher_begin_blob,
+ .begin_blob_ctx = hasher_ctx,
+ .consume_chunk = hasher_consume_chunk,
+ .consume_chunk_ctx = hasher_ctx,
+ .end_blob = hasher_end_blob,
+ .end_blob_ctx = hasher_ctx,
+ };
+ } else {
+ sink_cbs = (struct read_blob_list_callbacks*)cbs;
}
- /* Determine the compression type of the metadata resource. */
- ctype = resource_compression_type(wim_ctype, res_entry->flags);
-
- /* 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;
-
- DEBUG("Finished reading metadata resource into memory.\n");
-
+ for (cur = blob_list->next, next = cur->next;
+ cur != blob_list;
+ cur = next, next = cur->next)
+ {
+ blob = (struct blob_descriptor*)((u8*)cur - list_head_offset);
+
+ if (blob->blob_location == BLOB_IN_WIM &&
+ blob->size != blob->rdesc->uncompressed_size)
+ {
+ struct blob_descriptor *blob_next, *blob_last;
+ struct list_head *next2;
+ u64 blob_count;
+
+ /* The next blob is a proper sub-sequence of a WIM
+ * resource. See if there are other blobs in the same
+ * resource that need to be read. Since
+ * sort_blob_list_by_sequential_order() sorted the blobs
+ * by offset in the WIM, this can be determined by
+ * simply scanning forward in the list. */
+
+ blob_last = blob;
+ blob_count = 1;
+ for (next2 = next;
+ next2 != blob_list
+ && (blob_next = (struct blob_descriptor*)
+ ((u8*)next2 - list_head_offset),
+ blob_next->blob_location == BLOB_IN_WIM
+ && blob_next->rdesc == blob->rdesc);
+ next2 = next2->next)
+ {
+ blob_last = blob_next;
+ blob_count++;
+ }
+ if (blob_count > 1) {
+ /* Reading multiple blobs combined into a single
+ * WIM resource. They are in the blob list,
+ * sorted by offset; @blob specifies the first
+ * blob in the resource that needs to be read
+ * and @blob_last specifies the last blob in the
+ * resource that needs to be read. */
+ next = next2;
+ ret = read_blobs_in_solid_resource(blob, blob_last,
+ blob_count,
+ list_head_offset,
+ sink_cbs);
+ if (ret)
+ return ret;
+ continue;
+ }
+ }
- 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;
+ ret = read_full_blob_with_cbs(blob, sink_cbs);
+ if (ret && ret != BEGIN_BLOB_STATUS_SKIP_BLOB)
+ return ret;
}
-
- /* 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);
- return ret;
+ return 0;
}
-/* Write the metadata resource for the current image. */
-int write_metadata_resource(WIMStruct *w)
+/*
+ * Extract the first @size bytes of the specified blob.
+ *
+ * If @size specifies the full uncompressed size of the blob, then the SHA-1
+ * message digest of the uncompressed blob is checked while being extracted.
+ *
+ * The uncompressed data of the blob is passed in chunks of unspecified size to
+ * the @extract_chunk function, passing it @extract_chunk_arg.
+ */
+int
+extract_blob(struct blob_descriptor *blob, u64 size,
+ consume_data_callback_t extract_chunk, void *extract_chunk_arg)
{
- 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;
+ wimlib_assert(size <= blob->size);
+ if (size == blob->size) {
+ /* Do SHA-1. */
+ struct read_blob_list_callbacks cbs = {
+ .consume_chunk = extract_chunk,
+ .consume_chunk_ctx = extract_chunk_arg,
+ };
+ return read_full_blob_with_sha1(blob, &cbs);
+ } else {
+ /* Don't do SHA-1. */
+ return read_blob_prefix(blob, size, extract_chunk,
+ extract_chunk_arg);
}
- #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
+}
- 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;
+/* A consume_data_callback_t implementation that writes the chunk of data to a
+ * file descriptor. */
+static int
+extract_chunk_to_fd(const void *chunk, size_t size, void *_fd_p)
+{
+ struct filedes *fd = _fd_p;
- /* 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);
+ int ret = full_write(fd, chunk, size);
+ if (ret) {
+ ERROR_WITH_ERRNO("Error writing to file descriptor");
+ return ret;
}
- 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;
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)
+/* Extract the first @size bytes of the specified blob to the specified file
+ * descriptor. */
+int
+extract_blob_to_fd(struct blob_descriptor *blob, struct filedes *fd, u64 size)
{
- WIMStruct *w;
- FILE *out;
- FILE *in;
- struct lookup_table_entry *lte;
- int in_wim_ctype;
- int out_wim_ctype;
- struct resource_entry *output_res_entry;
- u64 len;
- int ret;
-
- 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;
-
- out_wim_ctype = wimlib_get_compression_type(w);
- output_res_entry = <e->output_resource_entry;
-
- /* do not write empty resources */
- if (lte->resource_entry.original_size == 0)
- return 0;
-
- /* 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 extract_blob(blob, size, extract_chunk_to_fd, fd);
+}
- 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;
- }
+/* Extract the full uncompressed contents of the specified blob to the specified
+ * file descriptor. */
+int
+extract_full_blob_to_fd(struct blob_descriptor *blob, struct filedes *fd)
+{
+ return extract_blob_to_fd(blob, fd, blob->size);
+}
- if (w->verbose)
- puts(lte->file_on_disk);
+/* Calculate the SHA-1 message digest of a blob and store it in @blob->hash. */
+int
+sha1_blob(struct blob_descriptor *blob)
+{
+ wimlib_assert(blob->unhashed);
+ struct read_blob_list_callbacks cbs = {
+ };
+ return read_full_blob_with_sha1(blob, &cbs);
+}
- ret = transfer_file_resource(in, len, len, 0,
- WIM_COMPRESSION_TYPE_NONE, out,
- out_wim_ctype, output_res_entry);
- fclose(in);
+/*
+ * Convert a short WIM resource header to a stand-alone WIM resource descriptor.
+ *
+ * Note: for solid resources some fields still need to be overridden.
+ */
+void
+wim_res_hdr_to_desc(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ struct wim_resource_descriptor *rdesc)
+{
+ rdesc->wim = wim;
+ rdesc->offset_in_wim = reshdr->offset_in_wim;
+ rdesc->size_in_wim = reshdr->size_in_wim;
+ rdesc->uncompressed_size = reshdr->uncompressed_size;
+ INIT_LIST_HEAD(&rdesc->blob_list);
+ rdesc->flags = reshdr->flags;
+ rdesc->is_pipable = wim_is_pipable(wim);
+ if (rdesc->flags & WIM_RESHDR_FLAG_COMPRESSED) {
+ rdesc->compression_type = wim->compression_type;
+ rdesc->chunk_size = wim->chunk_size;
} else {
+ rdesc->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
+ rdesc->chunk_size = 0;
+ }
+}
- /* Read from input WIM (possibly compressed) */
+/* Convert a stand-alone resource descriptor to a WIM resource header. */
+void
+wim_res_desc_to_hdr(const struct wim_resource_descriptor *rdesc,
+ struct wim_reshdr *reshdr)
+{
+ reshdr->offset_in_wim = rdesc->offset_in_wim;
+ reshdr->size_in_wim = rdesc->size_in_wim;
+ reshdr->flags = rdesc->flags;
+ reshdr->uncompressed_size = rdesc->uncompressed_size;
+}
- /* 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;
- }
- int input_res_ctype = resource_compression_type(
- in_wim_ctype,
- lte->resource_entry.flags);
-
- ret = transfer_file_resource(in,
- lte->resource_entry.size,
- lte->resource_entry.original_size,
- lte->resource_entry.offset,
- input_res_ctype,
- out,
- out_wim_ctype,
- output_res_entry);
- }
- return ret;
+/* Translates a WIM resource header from the on-disk format into an in-memory
+ * format. */
+void
+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;
+}
+
+/* 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);
}