/*
* resource.c
*
- * Read uncompressed and compressed metadata and file resources from a WIM file.
+ * Code for reading blobs and resources, including compressed WIM resources.
*/
/*
- * Copyright (C) 2012, 2013 Eric Biggers
+ * Copyright (C) 2012, 2013, 2015 Eric Biggers
*
- * This file is part of wimlib, a library for working with WIM files.
+ * 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 3 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 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.
*
- * 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 General Public License along with
- * wimlib; if not, see http://www.gnu.org/licenses/.
+ * 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/.
*/
#ifdef HAVE_CONFIG_H
# include "config.h"
#endif
-#include "wimlib.h"
-#include "wimlib/dentry.h"
+#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/lookup_table.h"
+#include "wimlib/ntfs_3g.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
-
-#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 <stdarg.h>
-#include <stdlib.h>
-#include <unistd.h>
+#include "wimlib/wim.h"
+#include "wimlib/win32.h"
/*
- * Compressed resources
- *
- * A compressed resource in a WIM consists of a number of consecutive LZX or
- * XPRESS-compressed chunks, each of which decompresses to 32768 bytes of data,
- * 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.
+ * Compressed WIM resources
*
- * Additional information:
+ * A compressed resource in a WIM consists of a sequence of chunks. Each chunk
+ * decompresses to the same size except possibly for the last, which
+ * decompresses to the remaining size. Chunks that did not compress to less
+ * than their original size are stored uncompressed.
*
- * - 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.
+ * We support three variations on this resource format, independently of the
+ * compression type and chunk size which can vary as well:
*
- * - The chunk table is included in the compressed size of the resource provided
- * in the corresponding entry in the WIM's stream lookup table.
+ * - Original resource format: immediately before the compressed chunks, 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. Chunk table
+ * entries are 32-bit for resources < 4 GiB uncompressed and 64-bit for
+ * resources >= 4 GiB uncompressed.
*
- * - 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.
+ * - Solid resource format (distinguished by the use of WIM_RESHDR_FLAG_SOLID
+ * instead of WIM_RESHDR_FLAG_COMPRESSED): similar to the original format, but
+ * the resource begins with a 16-byte header which specifies the uncompressed
+ * size of the resource, the compression type, and the chunk size. (In the
+ * original format, these values were instead determined from outside the
+ * resource itself, from the blob table and the WIM file header.) In addition,
+ * in this format the entries in the chunk table contain compressed chunk
+ * sizes rather than offsets. As a consequence of this, the chunk table
+ * entries are always 32-bit and there is an entry for chunk 0.
*
- * 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 (less than or
- * equal to 32768). Otherwise the details are the same.
+ * - Pipable resource format (wimlib extension; all resources in a pipable WIM
+ * have this format): similar to the original format, but the chunk table is
+ * at the end of the resource rather than the beginning, and each compressed
+ * chunk is prefixed with its compressed size as a 32-bit integer. This
+ * format allows a resource to be written without rewinding.
*/
-typedef int (*decompress_func_t)(const void *, unsigned, void *, unsigned);
-static decompress_func_t
-get_decompress_func(int ctype)
-{
- switch (ctype) {
- case WIMLIB_COMPRESSION_TYPE_LZX:
- return wimlib_lzx_decompress;
- case WIMLIB_COMPRESSION_TYPE_XPRESS:
- return wimlib_xpress_decompress;
- default:
- wimlib_assert(0);
- return NULL;
- }
-}
+struct data_range {
+ u64 offset;
+ u64 size;
+};
/*
- * read_compressed_resource()-
+ * Read data from a compressed WIM resource.
*
- * Read data from a compressed resource being read from a seekable WIM file.
- * The resource may be either pipable or non-pipable.
+ * @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.
+ * @cbs
+ * Structure which provides the consume_chunk() callback to feed the data
+ * being read. Each call provides the next chunk of the requested data,
+ * uncompressed. Each chunk will be nonempty and will not cross range
+ * boundaries but otherwise will be of unspecified size.
*
- * @flags may be:
+ * Possible return values:
*
- * 0:
- * Just do a normal read, decompressing the data if necessary.
+ * WIMLIB_ERR_SUCCESS (0)
+ * WIMLIB_ERR_READ (errno set)
+ * WIMLIB_ERR_UNEXPECTED_END_OF_FILE (errno set to EINVAL)
+ * WIMLIB_ERR_NOMEM (errno set to ENOMEM)
+ * WIMLIB_ERR_DECOMPRESSION (errno set to EINVAL)
+ * WIMLIB_ERR_INVALID_CHUNK_SIZE (errno set to EINVAL)
*
- * WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS:
- * Read the raw contents of the compressed chunks of the compressed
- * resource. For pipable resources, this does *not* include the chunk
- * headers. If a callback function is being used, it will be called once
- * for each compressed chunk. For non-pipable resources, this mode
- * excludes the chunk table. For pipable resources, this mode excludes the
- * stream and chunk headers.
+ * or other error code returned by the cbs->consume_chunk() function.
*/
static int
-read_compressed_resource(const struct wim_lookup_table_entry * const lte,
- u64 size, const consume_data_callback_t cb,
- const u32 in_chunk_size, void * const ctx_or_buf,
- const int flags, const u64 offset)
+read_compressed_wim_resource(const struct wim_resource_descriptor * const rdesc,
+ const struct data_range * const ranges,
+ const size_t num_ranges,
+ const struct read_blob_callbacks *cbs)
{
int ret;
-
- const u32 orig_chunk_size = wim_resource_chunk_size(lte);
- const u32 orig_chunk_order = bsr32(orig_chunk_size);
-
- wimlib_assert(is_power_of_2(orig_chunk_size));
- wimlib_assert(cb == NULL || is_power_of_2(in_chunk_size));
-
- /* Currently, reading raw compressed chunks is only guaranteed to work
- * correctly when the full resource is requested. Furthermore, in such
- * cases the requested size is specified as the compressed size, but
- * here we change it to an uncompressed size to avoid confusing the rest
- * of this function. */
- if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
- wimlib_assert(offset == 0);
- wimlib_assert(size == lte->resource_entry.size);
- wimlib_assert(wim_resource_chunk_size(lte) == in_chunk_size);
- size = wim_resource_size(lte);
- }
-
- wimlib_assert(offset + size <= wim_resource_size(lte));
-
- /* Handle the trivial case. */
- if (size == 0)
- return 0;
-
u64 *chunk_offsets = NULL;
- u8 *out_buf = NULL;
- u8 *tmp_buf = NULL;
- void *compressed_buf = NULL;
+ u8 *ubuf = NULL;
+ void *cbuf = NULL;
bool chunk_offsets_malloced = false;
- bool out_buf_malloced = false;
- bool tmp_buf_malloced = false;
- bool compressed_buf_malloced = false;
- const size_t stack_max = 32768;
+ bool ubuf_malloced = false;
+ bool cbuf_malloced = false;
+ struct wimlib_decompressor *decompressor = NULL;
+
+ /* Sanity checks */
+ wimlib_assert(num_ranges != 0);
+ for (size_t i = 0; i < num_ranges; i++) {
+ wimlib_assert(ranges[i].offset + ranges[i].size > ranges[i].offset &&
+ 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 appropriate decompression function. */
- const decompress_func_t decompress =
- get_decompress_func(wim_resource_compression_type(lte));
+ /* 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 = <e->wim->in_fd;
+ 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);
+ }
- /* Calculate the number of chunks the resource is divided into. */
- const u64 num_chunks = wim_resource_chunks(lte);
+ if (unlikely(!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_cleanup;
+ }
- /* Calculate the number of entries in the chunk table; it's one less
- * than the number of chunks, since the first chunk has no entry. */
- const u64 num_chunk_entries = num_chunks - 1;
+ /* Get valid decompressor. */
+ if (likely(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 (unlikely(ret)) {
+ if (ret != WIMLIB_ERR_NOMEM)
+ errno = EINVAL;
+ goto out_cleanup;
+ }
+ }
- /* Calculate the 0-based index of the chunk at which the read starts.
- */
- const u64 start_chunk = offset >> orig_chunk_order;
+ const u32 chunk_order = fls32(chunk_size);
- /* Calculate the offset, within the start chunk, of the first byte of
- * the read. */
- const u32 start_offset_in_chunk = offset & (orig_chunk_size - 1);
+ /* 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 index of the chunk that contains the last byte of the
- * read. */
- const u64 end_chunk = (offset + size - 1) >> orig_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 offset, within the end chunk, of the last byte of the
- * read. */
- const u32 end_offset_in_chunk = (offset + size - 1) & (orig_chunk_size - 1);
+ /* 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 entries are actually needed to read the
- * requested part of the resource. Include an entry for the first chunk
- * even though that doesn't exist in the on-disk table, but take into
- * account that if the last chunk required for the read is not the last
- * chunk of the resource, an extra chunk entry is needed so that the
- * compressed size of the last chunk of the read can be determined. */
- const u64 num_alloc_chunk_entries = end_chunk - start_chunk +
- 1 + (end_chunk != num_chunks - 1);
+ /* 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 = (wim_resource_size(lte) > ((u64)1 << 32)) ? 8 : 4;
+ const u64 chunk_entry_size = get_chunk_entry_size(rdesc->uncompressed_size,
+ alt_chunk_table);
- /* Calculate the size, in bytes, of the full chunk table. */
+ /* Calculate the size of the chunk table in bytes. */
const u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
- /* Allocate a buffer to hold a subset of the chunk table. It will only
- * contain offsets for the chunks that are actually needed for this
- * read. For speed, allocate the buffer on the stack unless it's too
- * large. */
- if (num_alloc_chunk_entries <= stack_max) {
- chunk_offsets = alloca(num_alloc_chunk_entries * sizeof(u64));
- chunk_offsets_malloced = false;
- } else {
- chunk_offsets = MALLOC(num_alloc_chunk_entries * sizeof(u64));
- if (!chunk_offsets) {
- ERROR("Failed to allocate chunk table "
- "with %"PRIu64" entries", num_alloc_chunk_entries);
- return WIMLIB_ERR_NOMEM;
+ /* 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 num_chunk_entries_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;
+ num_chunk_entries_to_read = last_needed_chunk + 1;
+ } else {
+
+ num_chunk_entries_to_read = last_needed_chunk - read_start_chunk + 1;
+
+ /* The first chunk has no explicit chunk table entry. */
+ if (read_start_chunk == 0) {
+ num_chunk_entries_to_read--;
+ first_chunk_entry_to_read = 0;
+ } else {
+ first_chunk_entry_to_read = read_start_chunk - 1;
+ }
+
+ /* Unless we're reading the final chunk of the resource,
+ * we need the offset of the chunk following the last
+ * needed chunk so that the compressed size of the last
+ * needed chunk can be computed. */
+ if (last_needed_chunk < num_chunks - 1)
+ num_chunk_entries_to_read++;
}
- chunk_offsets_malloced = true;
- }
- /* Set the implicit offset of the first chunk if it's included in the
- * needed chunks. */
- if (start_chunk == 0)
- chunk_offsets[0] = 0;
-
- /* Calculate the index of the first needed entry in the chunk table. */
- const u64 start_table_idx = (start_chunk == 0) ? 0 : start_chunk - 1;
-
- /* Calculate the number of entries that need to be read from the chunk
- * table. */
- const u64 num_needed_chunk_entries = (start_chunk == 0) ?
- num_alloc_chunk_entries - 1 : num_alloc_chunk_entries;
-
- /* Calculate the number of bytes of data that need to be read from the
- * chunk table. */
- const size_t chunk_table_needed_size =
- num_needed_chunk_entries * chunk_entry_size;
- if ((u64)chunk_table_needed_size !=
- num_needed_chunk_entries * chunk_entry_size)
- {
- ERROR("Compressed read request too large to fit into memory!");
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_memory;
- }
+ const u64 chunk_offsets_alloc_size =
+ max(num_chunk_entries_to_read,
+ num_needed_chunk_offsets) * sizeof(chunk_offsets[0]);
- /* Calculate the byte offset, in the WIM file, of the first chunk table
- * entry to read. Take into account that if the WIM file is in the
- * special "pipable" format, then the chunk table is at the end of the
- * resource, not the beginning. */
- const u64 file_offset_of_needed_chunk_entries =
- lte->resource_entry.offset
- + (start_table_idx * chunk_entry_size)
- + (lte->is_pipable ? (lte->resource_entry.size - chunk_table_size) : 0);
-
- /* Read the needed chunk table entries into the end of the chunk_offsets
- * buffer. */
- void * const chunk_tab_data = (u8*)&chunk_offsets[num_alloc_chunk_entries] -
- chunk_table_needed_size;
- ret = full_pread(in_fd, chunk_tab_data, chunk_table_needed_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. Careful: chunk_offsets aliases chunk_tab_data, which
- * breaks C's aliasing rules when we read 32-bit integers and store
- * 64-bit integers. But since the operations are safe as long as the
- * compiler doesn't mess with their order, we use the gcc may_alias
- * extension to tell the compiler that loads from the 32-bit integers
- * may alias stores to the 64-bit integers. */
- {
- typedef le64 __attribute__((may_alias)) aliased_le64_t;
- typedef le32 __attribute__((may_alias)) aliased_le32_t;
- u64 * const chunk_offsets_p = chunk_offsets + (start_chunk == 0);
- u64 i;
-
- if (chunk_entry_size == 4) {
- aliased_le32_t *raw_entries = (aliased_le32_t*)chunk_tab_data;
- for (i = 0; i < num_needed_chunk_entries; i++)
- chunk_offsets_p[i] = le32_to_cpu(raw_entries[i]);
- } else {
- aliased_le64_t *raw_entries = (aliased_le64_t*)chunk_tab_data;
- for (i = 0; i < num_needed_chunk_entries; i++)
- chunk_offsets_p[i] = le64_to_cpu(raw_entries[i]);
+ if (unlikely((size_t)chunk_offsets_alloc_size != chunk_offsets_alloc_size)) {
+ errno = ENOMEM;
+ goto oom;
}
- }
- /* Calculate file offset of the first chunk that needs to be read.
- * Note: if the resource is pipable, the entries in the chunk table do
- * *not* include the chunk headers. */
- u64 cur_read_offset = lte->resource_entry.offset + chunk_offsets[0];
- if (!lte->is_pipable)
- cur_read_offset += chunk_table_size;
- else
- cur_read_offset += start_chunk * sizeof(struct pwm_chunk_hdr);
-
- /* If using a callback function, allocate a temporary buffer that will
- * be used to pass data to it. If writing directly to a buffer instead,
- * arrange to write data directly into it. */
- size_t out_buf_size;
- u8 *out_buf_end, *out_p;
- if (cb) {
- out_buf_size = max(in_chunk_size, orig_chunk_size);
- if (out_buf_size <= stack_max) {
- out_buf = alloca(out_buf_size);
+ if (likely(chunk_offsets_alloc_size <= STACK_MAX)) {
+ chunk_offsets = alloca(chunk_offsets_alloc_size);
} else {
- out_buf = MALLOC(out_buf_size);
- if (out_buf == NULL) {
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_memory;
- }
- out_buf_malloced = true;
+ chunk_offsets = MALLOC(chunk_offsets_alloc_size);
+ if (unlikely(!chunk_offsets))
+ goto oom;
+ chunk_offsets_malloced = true;
}
- } else {
- out_buf_size = size;
- out_buf = ctx_or_buf;
- }
- out_buf_end = out_buf + out_buf_size;
- out_p = out_buf;
- /* Unless the raw compressed data was requested, allocate a temporary
- * buffer for reading compressed chunks, each of which can be at most
- * orig_chunk_size - 1 bytes. This excludes compressed chunks that are
- * a full orig_chunk_size bytes, which are actually stored uncompressed.
- */
- if (!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS)) {
- if (orig_chunk_size - 1 <= stack_max) {
- compressed_buf = alloca(orig_chunk_size - 1);
+ 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 (unlikely(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 {
- compressed_buf = MALLOC(orig_chunk_size - 1);
- if (compressed_buf == NULL) {
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_memory;
+ 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]);
}
- compressed_buf_malloced = true;
}
+
+ /* 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 yet another temporary buffer, this one for reading partial
- * chunks. */
- if (start_offset_in_chunk != 0 ||
- (end_offset_in_chunk != orig_chunk_size - 1 &&
- offset + size != wim_resource_size(lte)))
- {
- if (orig_chunk_size <= stack_max) {
- tmp_buf = alloca(orig_chunk_size);
- } else {
- tmp_buf = MALLOC(orig_chunk_size);
- if (tmp_buf == NULL) {
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_memory;
- }
- tmp_buf_malloced = true;
- }
+ /* 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 (unlikely(!ubuf))
+ goto oom;
+ ubuf_malloced = true;
}
- /* Read, and possibly decompress, each needed chunk, either writing the
- * data directly into the @ctx_or_buf buffer or passing it to the @cb
- * callback function. */
- for (u64 i = start_chunk; i <= end_chunk; i++) {
+ /* 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 (unlikely(!cbuf))
+ goto oom;
+ cbuf_malloced = true;
+ }
- /* If the resource is pipable, skip the chunk header. */
- if (lte->is_pipable)
- cur_read_offset += sizeof(struct pwm_chunk_hdr);
+ /* 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;
- /* Calculate the sizes of the compressed chunk and of the
- * uncompressed chunk. */
- u32 compressed_chunk_size;
- u32 uncompressed_chunk_size;
- if (i != num_chunks - 1) {
- /* Not the last chunk. Compressed size is given by
- * difference of chunk table entries; uncompressed size
- * is always the WIM chunk size. */
- compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
- chunk_offsets[i - start_chunk];
- uncompressed_chunk_size = orig_chunk_size;
- } else {
- /* Last chunk. Compressed size is the remaining size in
- * the compressed resource; uncompressed size is the
- * remaining size in the uncompressed resource. */
- compressed_chunk_size = lte->resource_entry.size -
- chunk_table_size -
- chunk_offsets[i - start_chunk];
- if (lte->is_pipable)
- compressed_chunk_size -= num_chunks *
- sizeof(struct pwm_chunk_hdr);
-
- if ((wim_resource_size(lte) & (orig_chunk_size - 1)) == 0)
- uncompressed_chunk_size = orig_chunk_size;
- else
- uncompressed_chunk_size = wim_resource_size(lte) &
- (orig_chunk_size - 1);
- }
+ /* Read and process each needed chunk. */
+ for (u64 i = read_start_chunk; i <= last_needed_chunk; i++) {
- /* Calculate how much of this chunk needs to be read. */
+ /* 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;
- u32 partial_chunk_size;
- u32 start_offset = 0;
- u32 end_offset = orig_chunk_size - 1;
+ /* Calculate compressed size of next chunk. */
+ u32 chunk_csize;
+ if (is_pipe_read) {
+ struct pwm_chunk_hdr chunk_hdr;
- if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
- partial_chunk_size = compressed_chunk_size;
+ ret = full_pread(in_fd, &chunk_hdr,
+ sizeof(chunk_hdr), cur_read_offset);
+ if (unlikely(ret))
+ goto read_error;
+ chunk_csize = le32_to_cpu(chunk_hdr.compressed_size);
} else {
- if (i == start_chunk)
- start_offset = start_offset_in_chunk;
+ 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 (unlikely(chunk_csize == 0 || chunk_csize > chunk_usize)) {
+ ERROR("Invalid chunk size in compressed resource!");
+ errno = EINVAL;
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ goto out_cleanup;
+ }
+ if (rdesc->is_pipable)
+ cur_read_offset += sizeof(struct pwm_chunk_hdr);
- if (i == end_chunk)
- end_offset = end_offset_in_chunk;
+ /* 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;
- partial_chunk_size = end_offset + 1 - start_offset;
- }
+ if (chunk_end_offset <= cur_range_pos) {
- if (compressed_chunk_size == uncompressed_chunk_size ||
- (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS))
- {
- /* Chunk stored uncompressed, or reading raw chunk data. */
- ret = full_pread(in_fd,
- out_p,
- partial_chunk_size,
- cur_read_offset + start_offset);
- if (ret)
- goto read_error;
+ /* 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 (unlikely(ret))
+ goto read_error;
+ }
} else {
- /* Compressed chunk and not doing raw read. */
- u8 *target;
- /* Read the compressed data into compressed_buf. */
- ret = full_pread(in_fd,
- compressed_buf,
- compressed_chunk_size,
- cur_read_offset);
- if (ret)
- goto read_error;
+ /* Read the chunk and feed data to the callback
+ * function. */
+ u8 *read_buf;
- /* For partial chunks we must buffer the uncompressed
- * data because we don't need all of it. */
- if (partial_chunk_size == uncompressed_chunk_size)
- target = out_p;
+ if (chunk_csize == chunk_usize)
+ read_buf = ubuf;
else
- target = tmp_buf;
-
- /* Decompress the chunk. */
- ret = (*decompress)(compressed_buf,
- compressed_chunk_size,
- target,
- uncompressed_chunk_size);
- if (ret) {
- ERROR("Failed to decompress data.");
- ret = WIMLIB_ERR_DECOMPRESSION;
- errno = EINVAL;
- goto out_free_memory;
- }
- if (partial_chunk_size != uncompressed_chunk_size)
- memcpy(out_p, tmp_buf + start_offset,
- partial_chunk_size);
- }
+ read_buf = cbuf;
- out_p += partial_chunk_size;
-
- if (cb) {
- /* Feed the data to the callback function. */
- wimlib_assert(offset == 0);
+ ret = full_pread(in_fd,
+ read_buf,
+ chunk_csize,
+ cur_read_offset);
+ if (unlikely(ret))
+ goto read_error;
- if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
- ret = cb(out_buf, out_p - out_buf, ctx_or_buf);
- if (ret)
- goto out_free_memory;
- out_p = out_buf;
-
- } else if (i == end_chunk || out_p == out_buf_end) {
- size_t bytes_sent;
- const u8 *p;
-
- for (p = out_buf; p != out_p; p += bytes_sent) {
- bytes_sent = min(in_chunk_size, out_p - p);
- ret = cb(p, bytes_sent, ctx_or_buf);
- if (ret)
- goto out_free_memory;
+ if (read_buf == cbuf) {
+ ret = wimlib_decompress(cbuf,
+ chunk_csize,
+ ubuf,
+ chunk_usize,
+ decompressor);
+ if (unlikely(ret)) {
+ ERROR("Failed to decompress data!");
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ errno = EINVAL;
+ goto out_cleanup;
}
- out_p = out_buf;
}
+ 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 = call_consume_chunk(&ubuf[start], size, cbs);
+ if (unlikely(ret))
+ goto out_cleanup;
+
+ 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);
}
- cur_read_offset += compressed_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 (unlikely(ret))
+ goto read_error;
+ }
ret = 0;
-out_free_memory:
+
+out_cleanup:
+ 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 (out_buf_malloced)
- FREE(out_buf);
- if (compressed_buf_malloced)
- FREE(compressed_buf);
- if (tmp_buf_malloced)
- FREE(tmp_buf);
+ if (ubuf_malloced)
+ FREE(ubuf);
+ if (cbuf_malloced)
+ FREE(cbuf);
return ret;
+oom:
+ ERROR("Out of memory while reading compressed WIM resource");
+ ret = WIMLIB_ERR_NOMEM;
+ goto out_cleanup;
+
read_error:
- ERROR_WITH_ERRNO("Error reading compressed file resource");
- goto out_free_memory;
+ ERROR_WITH_ERRNO("Error reading data from WIM file");
+ goto out_cleanup;
}
-/* Skip over the chunk table at the end of pipable, compressed resource being
- * read from a pipe. */
+/* Read raw data from a file descriptor at the specified offset, feeding the
+ * data in nonempty chunks into the cbs->consume_chunk() function. */
static int
-skip_chunk_table(const struct wim_lookup_table_entry *lte,
- struct filedes *in_fd)
+read_raw_file_data(struct filedes *in_fd, u64 offset, u64 size,
+ const struct read_blob_callbacks *cbs,
+ const tchar *filename)
{
- u64 num_chunk_entries = wim_resource_chunks(lte) - 1;
- u64 chunk_entry_size = (wim_resource_size(lte) > ((u64)1 << 32)) ? 8 : 4;
- u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
+ u8 buf[BUFFER_SIZE];
+ size_t bytes_to_read;
int ret;
- if (num_chunk_entries != 0) {
- u8 dummy;
- ret = full_pread(in_fd, &dummy, 1,
- in_fd->offset + chunk_table_size - 1);
- if (ret)
+ while (size) {
+ bytes_to_read = min(sizeof(buf), size);
+ ret = full_pread(in_fd, buf, bytes_to_read, offset);
+ if (unlikely(ret))
+ goto read_error;
+ ret = call_consume_chunk(buf, bytes_to_read, cbs);
+ if (unlikely(ret))
return ret;
+ size -= bytes_to_read;
+ offset += bytes_to_read;
}
return 0;
-}
-/* Read and decompress data from a compressed, pipable resource being read from
- * a pipe. */
-static int
-read_pipable_resource(const struct wim_lookup_table_entry *lte,
- u64 size, consume_data_callback_t cb,
- u32 in_chunk_size, void *ctx_or_buf,
- int flags, u64 offset)
-{
- struct filedes *in_fd;
- decompress_func_t decompress;
- int ret;
- const u32 orig_chunk_size = wim_resource_chunk_size(lte);
- u8 cchunk[orig_chunk_size - 1];
-
- size_t out_buf_size;
- u8 *out_buf, *out_buf_end, *out_p;
- if (cb) {
- out_buf_size = max(in_chunk_size, orig_chunk_size);
- out_buf = alloca(out_buf_size);
+read_error:
+ if (!filename) {
+ ERROR_WITH_ERRNO("Error reading data from WIM file");
+ } else if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE) {
+ ERROR("\"%"TS"\": File was concurrently truncated", filename);
+ ret = WIMLIB_ERR_CONCURRENT_MODIFICATION_DETECTED;
} else {
- out_buf_size = size;
- out_buf = ctx_or_buf;
+ ERROR_WITH_ERRNO("\"%"TS"\": Error reading data", filename);
}
- out_buf_end = out_buf + out_buf_size;
- out_p = out_buf;
-
- /* Get pointers to appropriate decompression function and the input file
- * descriptor. */
- decompress = get_decompress_func(wim_resource_compression_type(lte));
- in_fd = <e->wim->in_fd;
-
- /* This function currently assumes the entire resource is being read at
- * once and that the raw compressed data isn't being requested. This is
- * based on the fact that this function currently only gets called
- * during the operation of wimlib_extract_image_from_pipe(). */
- wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW));
- wimlib_assert(offset == 0);
- wimlib_assert(size == wim_resource_size(lte));
- wimlib_assert(in_fd->offset == lte->resource_entry.offset);
-
- u32 chunk_usize;
- for (offset = 0; offset < size; offset += chunk_usize) {
- struct pwm_chunk_hdr chunk_hdr;
- u32 chunk_csize;
-
- /* Calculate uncompressed size of next chunk. */
- chunk_usize = min(orig_chunk_size, size - offset);
-
- /* Read the compressed size of the next chunk from the chunk
- * header. */
- ret = full_read(in_fd, &chunk_hdr, sizeof(chunk_hdr));
- if (ret)
- goto read_error;
-
- chunk_csize = le32_to_cpu(chunk_hdr.compressed_size);
-
- if (chunk_csize > orig_chunk_size) {
- errno = EINVAL;
- ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
- goto invalid;
- }
-
- /* Read chunk data. */
- ret = full_read(in_fd, cchunk, chunk_csize);
- if (ret)
- goto read_error;
-
- if (flags & WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY)
- continue;
-
- /* Decompress chunk if needed. Uncompressed size same
- * as compressed size means the chunk is uncompressed.
- */
- if (chunk_csize == chunk_usize) {
- memcpy(out_p, cchunk, chunk_usize);
- } else {
- ret = (*decompress)(cchunk, chunk_csize,
- out_p, chunk_usize);
- if (ret) {
- errno = EINVAL;
- ret = WIMLIB_ERR_DECOMPRESSION;
- goto invalid;
- }
- }
- out_p += chunk_usize;
-
- /* Feed the uncompressed data into the callback function or copy
- * it into the provided buffer. */
- if (cb && (out_p == out_buf_end ||
- offset + chunk_usize == size))
- {
- size_t bytes_sent;
- const u8 *p;
+ return ret;
+}
- for (p = out_buf; p != out_p; p += bytes_sent) {
- bytes_sent = min(in_chunk_size, out_p - p);
- ret = cb(p, bytes_sent, ctx_or_buf);
- if (ret)
- return ret;
- }
- out_p = out_buf;
- }
- }
+/* A consume_chunk() implementation that simply concatenates all chunks into an
+ * in-memory buffer. */
+static int
+bufferer_cb(const void *chunk, size_t size, void *_ctx)
+{
+ void **buf_p = _ctx;
- ret = skip_chunk_table(lte, in_fd);
- if (ret)
- goto read_error;
+ *buf_p = mempcpy(*buf_p, chunk, size);
return 0;
-
-read_error:
- ERROR_WITH_ERRNO("Error reading compressed file resource");
- return ret;
-
-invalid:
- ERROR("Compressed file resource is invalid");
- return ret;
}
/*
- * read_partial_wim_resource()-
- *
- * Read a range of data from a uncompressed or compressed resource in a WIM
- * file. Data is written into a buffer or fed into a callback function, as
- * documented in read_resource_prefix().
- *
- * @flags can be:
+ * Read @size bytes at @offset in the WIM resource described by @rdesc and feed
+ * the data into the @cbs->consume_chunk callback function.
*
- * 0:
- * Just do a normal read, decompressing the data if necessary. @size and
- * @offset are interpreted relative to the uncompressed contents of the
- * stream.
+ * @offset and @size are assumed to have already been validated against the
+ * resource's uncompressed size.
*
- * WIMLIB_READ_RESOURCE_FLAG_RAW_FULL:
- * Only valid when the resource is compressed: Read the raw contents of
- * the compressed resource. If the resource is non-pipable, this includes
- * the chunk table as well as the compressed chunks. If the resource is
- * pipable, this includes the compressed chunks--- including the chunk
- * headers--- and the chunk table. The stream header is still *not*
- * included.
- *
- * In this mode, @offset is relative to the beginning of the raw contents
- * of the compressed resource--- that is, the chunk table if the resource
- * is non-pipable, or the header for the first compressed chunk if the
- * resource is pipable. @size is the number of raw bytes to read, which
- * must not overrun the end of the resource. For example, if @offset is 0,
- * then @size can be at most the raw size of the compressed resource
- * (@lte->resource_entry.size).
- *
- * WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS:
- * Only valid when the resource is compressed and is not being read from a
- * pipe: Read the raw contents of the compressed chunks of the compressed
- * resource. For pipable resources, this does *not* include the chunk
- * headers. If a callback function is being used, it will be called once
- * for each compressed chunk. The chunk table is excluded. Also, for
- * pipable resources, the stream and chunk headers are excluded. In this
- * mode, @size must be exactly the raw size of the compressed resource
- * (@lte->resource_entry.size) and @offset must be 0.
- *
- * WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY:
- * Only valid when the resource is being read from a pipe: Skip over the
- * requested data rather than feed it to the callback function or write it
- * into the buffer. No decompression is done.
- * WIMLIB_READ_RESOURCE_FLAG_RAW_* may not be combined with this flag.
- * @offset must be 0 and @size must be the uncompressed size of the
- * resource.
- *
- * 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)
- * WIMLIB_ERR_INVALID_PIPABLE_WIM (errno set to EINVAL)
- *
- * or other error code returned by the @cb function.
+ * Returns 0 on success; or the first nonzero value returned by the callback
+ * function; or a nonzero wimlib error code with errno set as well.
*/
-int
-read_partial_wim_resource(const struct wim_lookup_table_entry *lte,
- u64 size, consume_data_callback_t cb,
- u32 in_chunk_size,
- void *ctx_or_buf, int flags, u64 offset)
+static int
+read_partial_wim_resource(const struct wim_resource_descriptor *rdesc,
+ const u64 offset, const u64 size,
+ const struct read_blob_callbacks *cbs)
{
- struct filedes *in_fd;
- int ret;
-
- /* Make sure the resource is actually located in a WIM file and is not
- * somewhere else. */
- wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
-
- /* If a callback was specified, in_chunk_size must be a power of 2 (and
- * not 0). */
- wimlib_assert(cb == NULL || is_power_of_2(in_chunk_size));
-
- /* If a callback was specified, offset must be zero. */
- wimlib_assert(cb == NULL || offset == 0);
-
- /* Retrieve input file descriptor for the WIM file. */
- in_fd = <e->wim->in_fd;
-
- /* Don't allow raw reads (either full or chunks) of uncompressed
- * resources. */
- wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW) ||
- resource_is_compressed(<e->resource_entry));
-
- /* Don't allow seek-only reads unless reading from a pipe; also don't
- * allow combining SEEK_ONLY with either RAW flag. */
- wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY) ||
- (!filedes_is_seekable(in_fd) &&
- !(flags & WIMLIB_READ_RESOURCE_FLAG_RAW)));
-
- DEBUG("Reading WIM resource: %"PRIu64" @ +%"PRIu64" "
- "from %"PRIu64" @ +%"PRIu64" (readflags 0x%08x, resflags 0x%02x%s)",
- size, offset,
- lte->resource_entry.original_size, lte->resource_entry.offset,
- flags, lte->resource_entry.flags,
- (lte->is_pipable ? ", pipable" : ""));
-
- if ((flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) ||
- !resource_is_compressed(<e->resource_entry))
+ if (rdesc->flags & (WIM_RESHDR_FLAG_COMPRESSED |
+ WIM_RESHDR_FLAG_SOLID))
{
- /* Reading raw resource contents or reading uncompressed
- * resource. */
- wimlib_assert(offset + size <= lte->resource_entry.size);
- offset += lte->resource_entry.offset;
- if (flags & WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY) {
- if (lte->resource_entry.size != 0) {
- u8 dummy;
- ret = full_pread(in_fd, &dummy, 1,
- offset + lte->resource_entry.size - 1);
- if (ret)
- goto read_error;
- }
- } else if (cb) {
- /* Send data to callback function */
- u8 buf[min(in_chunk_size, size)];
- while (size) {
- size_t bytes_to_read = min(in_chunk_size, size);
- ret = full_pread(in_fd, buf, bytes_to_read,
- offset);
- if (ret)
- goto read_error;
- ret = cb(buf, bytes_to_read, ctx_or_buf);
- if (ret)
- goto out;
- size -= bytes_to_read;
- offset += bytes_to_read;
- }
- } else {
- /* Send data directly to a buffer */
- ret = full_pread(in_fd, ctx_or_buf, size, offset);
- if (ret)
- goto read_error;
- }
- ret = 0;
- } else if (lte->is_pipable && !filedes_is_seekable(in_fd)) {
- /* Reading compressed, pipable resource from pipe. */
- ret = read_pipable_resource(lte, size, cb,
- in_chunk_size,
- ctx_or_buf, flags, offset);
- } else {
- /* Reading compressed, possibly pipable resource from seekable
- * file. */
- ret = read_compressed_resource(lte, size, cb,
- in_chunk_size,
- ctx_or_buf, flags, offset);
+ /* Compressed resource */
+ if (unlikely(!size))
+ return 0;
+ struct data_range range = {
+ .offset = offset,
+ .size = size,
+ };
+ return read_compressed_wim_resource(rdesc, &range, 1, cbs);
}
- goto out;
-read_error:
- ERROR_WITH_ERRNO("Error reading data from WIM");
-out:
- return ret;
+ /* Uncompressed resource */
+ return read_raw_file_data(&rdesc->wim->in_fd,
+ rdesc->offset_in_wim + offset,
+ size, cbs, NULL);
}
+/* Read the specified range of uncompressed data from the specified blob, which
+ * must be located in a WIM file, into the specified buffer. */
+int
+read_partial_wim_blob_into_buf(const struct blob_descriptor *blob,
+ u64 offset, size_t size, void *buf)
+{
+ struct read_blob_callbacks cbs = {
+ .consume_chunk = bufferer_cb,
+ .ctx = &buf,
+ };
+ return read_partial_wim_resource(blob->rdesc,
+ blob->offset_in_res + offset,
+ size,
+ &cbs);
+}
+/* Skip over the data of the specified WIM resource. */
int
-read_partial_wim_resource_into_buf(const struct wim_lookup_table_entry *lte,
- size_t size, u64 offset, void *buf)
+skip_wim_resource(const struct wim_resource_descriptor *rdesc)
{
- return read_partial_wim_resource(lte, size, NULL, 0, buf, 0, offset);
+ struct read_blob_callbacks cbs = {
+ };
+ return read_partial_wim_resource(rdesc, 0,
+ rdesc->uncompressed_size, &cbs);
}
static int
-read_wim_resource_prefix(const struct wim_lookup_table_entry *lte,
- u64 size,
- consume_data_callback_t cb,
- u32 in_chunk_size,
- void *ctx_or_buf,
- int flags)
+read_wim_blob_prefix(const struct blob_descriptor *blob, u64 size,
+ const struct read_blob_callbacks *cbs)
{
- return read_partial_wim_resource(lte, size, cb, in_chunk_size,
- ctx_or_buf, flags, 0);
+ return read_partial_wim_resource(blob->rdesc, blob->offset_in_res,
+ size, cbs);
}
-
-#ifndef __WIN32__
+/* 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 wim_lookup_table_entry *lte,
- u64 size,
- consume_data_callback_t cb,
- u32 in_chunk_size,
- void *ctx_or_buf,
- int _ignored_flags)
+read_file_on_disk_prefix(const struct blob_descriptor *blob, u64 size,
+ const struct read_blob_callbacks *cbs)
{
- const tchar *filename = lte->file_on_disk;
int ret;
- struct filedes fd;
int raw_fd;
- u8 *out_buf;
- bool out_buf_malloced;
- const size_t stack_max = 32768;
-
- DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"",
- size, lte->file_on_disk);
+ struct filedes fd;
- raw_fd = open(filename, O_RDONLY);
- if (raw_fd < 0) {
- ERROR_WITH_ERRNO("Can't open \"%"TS"\"", filename);
+ raw_fd = topen(blob->file_on_disk, O_BINARY | O_RDONLY);
+ if (unlikely(raw_fd < 0)) {
+ ERROR_WITH_ERRNO("Can't open \"%"TS"\"", blob->file_on_disk);
return WIMLIB_ERR_OPEN;
}
filedes_init(&fd, raw_fd);
- out_buf_malloced = false;
- if (cb) {
- /* Send data to callback function */
- if (in_chunk_size <= stack_max) {
- out_buf = alloca(in_chunk_size);
- } else {
- out_buf = MALLOC(in_chunk_size);
- if (out_buf == NULL) {
- ret = WIMLIB_ERR_NOMEM;
- goto out_close;
- }
- out_buf_malloced = true;
- }
-
- size_t bytes_to_read;
- while (size) {
- bytes_to_read = min(in_chunk_size, size);
- ret = full_read(&fd, out_buf, bytes_to_read);
- if (ret)
- goto read_error;
- ret = cb(out_buf, bytes_to_read, ctx_or_buf);
- if (ret)
- goto out_close;
- size -= bytes_to_read;
- }
- } else {
- /* Send data directly to a buffer */
- ret = full_read(&fd, ctx_or_buf, size);
- if (ret)
- goto read_error;
- }
- ret = 0;
- goto out_close;
-
-read_error:
- ERROR_WITH_ERRNO("Error reading \"%"TS"\"", filename);
-out_close:
+ ret = read_raw_file_data(&fd, 0, size, cbs, blob->file_on_disk);
filedes_close(&fd);
- if (out_buf_malloced)
- FREE(out_buf);
return ret;
}
-#endif /* !__WIN32__ */
+#ifdef WITH_FUSE
static int
-read_buffer_prefix(const struct wim_lookup_table_entry *lte,
- u64 size, consume_data_callback_t cb,
- u32 in_chunk_size,
- void *ctx_or_buf, int _ignored_flags)
+read_staging_file_prefix(const struct blob_descriptor *blob, u64 size,
+ const struct read_blob_callbacks *cbs)
{
+ int raw_fd;
+ struct filedes fd;
+ int ret;
- if (cb) {
- int ret;
- u32 chunk_size;
-
- for (u64 offset = 0; offset < size; offset += chunk_size) {
- chunk_size = min(in_chunk_size, size - offset);
- ret = cb((const u8*)lte->attached_buffer + offset,
- chunk_size, ctx_or_buf);
- if (ret)
- return ret;
- }
- } else {
- memcpy(ctx_or_buf, lte->attached_buffer, size);
+ raw_fd = openat(blob->staging_dir_fd, blob->staging_file_name,
+ O_RDONLY | O_NOFOLLOW);
+ if (unlikely(raw_fd < 0)) {
+ ERROR_WITH_ERRNO("Can't open staging file \"%s\"",
+ blob->staging_file_name);
+ return WIMLIB_ERR_OPEN;
}
- return 0;
+ filedes_init(&fd, raw_fd);
+ ret = read_raw_file_data(&fd, 0, size, cbs, blob->staging_file_name);
+ 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, const struct read_blob_callbacks *cbs)
+{
+ if (unlikely(!size))
+ return 0;
+ return call_consume_chunk(blob->attached_buffer, size, cbs);
}
-typedef int (*read_resource_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
- u64 size,
- consume_data_callback_t cb,
- u32 in_chunk_size,
- void *ctx_or_buf,
- int flags);
+typedef int (*read_blob_prefix_handler_t)(const struct blob_descriptor *blob,
+ u64 size,
+ const struct read_blob_callbacks *cbs);
/*
- * read_resource_prefix()-
+ * Read the first @size bytes from a generic "blob", which may be located in any
+ * one of several locations, such as in a WIM resource (possibly compressed), in
+ * an external file, or directly in an in-memory buffer. The blob data will be
+ * fed to the cbs->consume_chunk() callback function in chunks that are nonempty
+ * but otherwise are of unspecified size.
*
- * Read the first @size bytes from a generic "resource", which may be located in
- * the WIM (compressed or uncompressed), in an external file, or directly in an
- * in-memory buffer.
- *
- * Feed 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 @size bytes).
- *
- * When using a callback function, it is called with chunks up to 32768 bytes in
- * size until the resource is exhausted.
- *
- * If the resource is located in a WIM file, @flags can be set as documented in
- * read_partial_wim_resource(). Otherwise @flags are ignored.
+ * 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 cbs->consume_chunk() returned nonzero
+ * in which case that error code will be returned.
*/
-int
-read_resource_prefix(const struct wim_lookup_table_entry *lte,
- u64 size, consume_data_callback_t cb, u32 in_chunk_size,
- void *ctx_or_buf, int flags)
+static int
+read_blob_prefix(const struct blob_descriptor *blob, u64 size,
+ const struct read_blob_callbacks *cbs)
{
- static const read_resource_prefix_handler_t handlers[] = {
- [RESOURCE_IN_WIM] = read_wim_resource_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,
+ 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
- [RESOURCE_IN_STAGING_FILE] = read_file_on_disk_prefix,
+ [BLOB_IN_STAGING_FILE] = read_staging_file_prefix,
#endif
#ifdef WITH_NTFS_3G
- [RESOURCE_IN_NTFS_VOLUME] = read_ntfs_file_prefix,
+ [BLOB_IN_NTFS_VOLUME] = read_ntfs_attribute_prefix,
#endif
#ifdef __WIN32__
- [RESOURCE_WIN32_ENCRYPTED] = read_win32_encrypted_file_prefix,
+ [BLOB_IN_WINDOWS_FILE] = read_windows_file_prefix,
#endif
};
- wimlib_assert(lte->resource_location < ARRAY_LEN(handlers)
- && handlers[lte->resource_location] != NULL);
- wimlib_assert(cb == NULL || in_chunk_size > 0);
- return handlers[lte->resource_location](lte, size, cb, in_chunk_size, ctx_or_buf, flags);
+ 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, cbs);
+}
+
+/* Read the full data of the specified blob, passing the data into the specified
+ * callbacks (all of which are optional). */
+int
+read_blob_with_cbs(struct blob_descriptor *blob,
+ const struct read_blob_callbacks *cbs)
+{
+ int ret;
+
+ ret = call_begin_blob(blob, cbs);
+ if (unlikely(ret))
+ return ret;
+
+ ret = read_blob_prefix(blob, blob->size, cbs);
+
+ return call_end_blob(blob, ret, cbs);
}
+/* Read the full uncompressed data of the specified blob into the specified
+ * buffer, which must have space for at least blob->size bytes. The SHA-1
+ * message digest is *not* checked. */
int
-read_full_resource_into_buf(const struct wim_lookup_table_entry *lte,
- void *buf)
+read_blob_into_buf(const struct blob_descriptor *blob, void *buf)
{
- return read_resource_prefix(lte, wim_resource_size(lte), NULL, 0, buf, 0);
+ struct read_blob_callbacks cbs = {
+ .consume_chunk = bufferer_cb,
+ .ctx = &buf,
+ };
+ return read_blob_prefix(blob, blob->size, &cbs);
}
+/* Retrieve the full uncompressed data of the specified blob. A buffer large
+ * enough hold the data is allocated and returned in @buf_ret. The SHA-1
+ * message digest is *not* checked. */
int
-read_full_resource_into_alloc_buf(const struct wim_lookup_table_entry *lte,
- void **buf_ret)
+read_blob_into_alloc_buf(const struct blob_descriptor *blob, void **buf_ret)
{
int ret;
void *buf;
- if ((size_t)lte->resource_entry.original_size !=
- lte->resource_entry.original_size)
- {
- ERROR("Can't read %"PRIu64" byte resource into "
- "memory", lte->resource_entry.original_size);
+ if (unlikely((size_t)blob->size != blob->size)) {
+ ERROR("Can't read %"PRIu64" byte blob into memory", blob->size);
return WIMLIB_ERR_NOMEM;
}
- buf = MALLOC(lte->resource_entry.original_size);
- if (!buf)
+ buf = MALLOC(blob->size);
+ if (unlikely(!buf))
return WIMLIB_ERR_NOMEM;
- ret = read_full_resource_into_buf(lte, buf);
- if (ret) {
+ ret = read_blob_into_buf(blob, buf);
+ if (unlikely(ret)) {
FREE(buf);
return ret;
}
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
-res_entry_to_data(const struct resource_entry *res_entry,
- WIMStruct *wim, void **buf_ret)
+wim_reshdr_to_data(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ void **buf_ret)
{
+ struct wim_resource_descriptor rdesc;
+ struct blob_descriptor blob;
+
+ wim_reshdr_to_desc_and_blob(reshdr, wim, &rdesc, &blob);
+
+ return read_blob_into_alloc_buf(&blob, buf_ret);
+}
+
+/* Calculate the SHA-1 message digest of the uncompressed data of the specified
+ * WIM resource. */
+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;
- struct wim_lookup_table_entry *lte;
- lte = new_lookup_table_entry();
- if (!lte)
- return WIMLIB_ERR_NOMEM;
+ wim_reshdr_to_desc_and_blob(reshdr, wim, &rdesc, &blob);
+ blob.unhashed = 1;
- copy_resource_entry(<e->resource_entry, res_entry);
- lte->unhashed = 1;
- lte->part_number = wim->hdr.part_number;
- lte_init_wim(lte, wim);
+ ret = sha1_blob(&blob);
+ if (unlikely(ret))
+ return ret;
- ret = read_full_resource_into_alloc_buf(lte, buf_ret);
- free_lookup_table_entry(lte);
- return ret;
+ copy_hash(hash, blob.hash);
+ return 0;
+}
+
+struct blobifier_context {
+ struct read_blob_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)
+{
+ struct list_head *cur;
+
+ cur = (struct list_head*)((u8*)blob + list_head_offset);
+
+ return (struct blob_descriptor*)((u8*)cur->next - list_head_offset);
+}
+
+/* A consume_chunk() implementation that translates raw resource data into
+ * blobs, calling the begin_blob, consume_chunk, and end_blob callbacks as
+ * appropriate. */
+static int
+blobifier_cb(const void *chunk, size_t size, void *_ctx)
+{
+ struct blobifier_context *ctx = _ctx;
+ int ret;
+
+ wimlib_assert(ctx->cur_blob != NULL);
+ wimlib_assert(size <= ctx->cur_blob->size - ctx->cur_blob_offset);
+
+ if (ctx->cur_blob_offset == 0) {
+ /* Starting a new blob. */
+ ret = call_begin_blob(ctx->cur_blob, &ctx->cbs);
+ if (ret)
+ return ret;
+ }
+
+ ctx->cur_blob_offset += size;
+
+ ret = call_consume_chunk(chunk, size, &ctx->cbs);
+ if (ret)
+ return ret;
+
+ if (ctx->cur_blob_offset == ctx->cur_blob->size) {
+ /* Finished reading all the data for a blob. */
+
+ ctx->cur_blob_offset = 0;
+
+ ret = call_end_blob(ctx->cur_blob, 0, &ctx->cbs);
+ 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;
}
-struct extract_ctx {
+struct hasher_context {
SHA_CTX sha_ctx;
- consume_data_callback_t extract_chunk;
- void *extract_chunk_arg;
+ int flags;
+ struct read_blob_callbacks cbs;
};
+/* 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;
+
+ sha1_init(&ctx->sha_ctx);
+
+ return call_begin_blob(blob, &ctx->cbs);
+}
+
+/* A consume_chunk() implementation that continues calculating the SHA-1 message
+ * digest of the blob being read, then optionally passes the data on to another
+ * consume_chunk() implementation. This allows checking the SHA-1 message
+ * digest of a blob being extracted, for example. */
static int
-extract_chunk_sha1_wrapper(const void *chunk, size_t chunk_size,
- void *_ctx)
+hasher_consume_chunk(const void *chunk, size_t size, void *_ctx)
{
- struct extract_ctx *ctx = _ctx;
+ struct hasher_context *ctx = _ctx;
- sha1_update(&ctx->sha_ctx, chunk, chunk_size);
- return ctx->extract_chunk(chunk, chunk_size, ctx->extract_chunk_arg);
+ sha1_update(&ctx->sha_ctx, chunk, size);
+
+ return call_consume_chunk(chunk, size, &ctx->cbs);
}
-/* Extracts the first @size bytes of a WIM resource to somewhere. In the
- * process, the SHA1 message digest of the resource is checked if the full
- * resource is being extracted.
+static int
+report_sha1_mismatch_error(const struct blob_descriptor *blob,
+ const u8 actual_hash[SHA1_HASH_SIZE])
+{
+ tchar expected_hashstr[SHA1_HASH_SIZE * 2 + 1];
+ tchar actual_hashstr[SHA1_HASH_SIZE * 2 + 1];
+
+ wimlib_assert(blob->blob_location != BLOB_NONEXISTENT);
+ wimlib_assert(blob->blob_location != BLOB_IN_ATTACHED_BUFFER);
+
+ sprint_hash(blob->hash, expected_hashstr);
+ sprint_hash(actual_hash, actual_hashstr);
+
+ if (blob_is_in_file(blob)) {
+ ERROR("A file was concurrently modified!\n"
+ " Path: \"%"TS"\"\n"
+ " Expected SHA-1: %"TS"\n"
+ " Actual SHA-1: %"TS"\n",
+ blob_file_path(blob), expected_hashstr, actual_hashstr);
+ return WIMLIB_ERR_CONCURRENT_MODIFICATION_DETECTED;
+ } else if (blob->blob_location == BLOB_IN_WIM) {
+ const struct wim_resource_descriptor *rdesc = blob->rdesc;
+ ERROR("A WIM resource is corrupted!\n"
+ " WIM file: \"%"TS"\"\n"
+ " Blob uncompressed size: %"PRIu64"\n"
+ " Resource offset in WIM: %"PRIu64"\n"
+ " Resource uncompressed size: %"PRIu64"\n"
+ " Resource size in WIM: %"PRIu64"\n"
+ " Resource flags: 0x%x%"TS"\n"
+ " Resource compression type: %"TS"\n"
+ " Resource compression chunk size: %"PRIu32"\n"
+ " Expected SHA-1: %"TS"\n"
+ " Actual SHA-1: %"TS"\n",
+ rdesc->wim->filename,
+ blob->size,
+ rdesc->offset_in_wim,
+ rdesc->uncompressed_size,
+ rdesc->size_in_wim,
+ (unsigned int)rdesc->flags,
+ (rdesc->is_pipable ? T(", pipable") : T("")),
+ wimlib_get_compression_type_string(
+ rdesc->compression_type),
+ rdesc->chunk_size,
+ expected_hashstr, actual_hashstr);
+ return WIMLIB_ERR_INVALID_RESOURCE_HASH;
+ } else {
+ ERROR("File data was concurrently modified!\n"
+ " Location ID: %d\n"
+ " Expected SHA-1: %"TS"\n"
+ " Actual SHA-1: %"TS"\n",
+ (int)blob->blob_location,
+ expected_hashstr, actual_hashstr);
+ return WIMLIB_ERR_CONCURRENT_MODIFICATION_DETECTED;
+ }
+}
+
+/* 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 (unlikely(status)) {
+ /* Error occurred; the full blob may not have been read. */
+ ret = status;
+ goto out_next_cb;
+ }
+
+ /* Retrieve the final SHA-1 message digest. */
+ sha1_final(hash, &ctx->sha_ctx);
+
+ /* Set the SHA-1 message digest of the blob, or compare the calculated
+ * value with stored value. */
+ if (blob->unhashed) {
+ if (ctx->flags & COMPUTE_MISSING_BLOB_HASHES)
+ copy_hash(blob->hash, hash);
+ } else if ((ctx->flags & VERIFY_BLOB_HASHES) &&
+ unlikely(!hashes_equal(hash, blob->hash)))
+ {
+ ret = report_sha1_mismatch_error(blob, hash);
+ goto out_next_cb;
+ }
+ ret = 0;
+out_next_cb:
+ return call_end_blob(blob, ret, &ctx->cbs);
+}
+
+/* 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. */
+int
+read_blob_with_sha1(struct blob_descriptor *blob,
+ const struct read_blob_callbacks *cbs)
+{
+ struct hasher_context hasher_ctx = {
+ .flags = VERIFY_BLOB_HASHES | COMPUTE_MISSING_BLOB_HASHES,
+ .cbs = *cbs,
+ };
+ struct read_blob_callbacks hasher_cbs = {
+ .begin_blob = hasher_begin_blob,
+ .consume_chunk = hasher_consume_chunk,
+ .end_blob = hasher_end_blob,
+ .ctx = &hasher_ctx,
+ };
+ return read_blob_with_cbs(blob, &hasher_cbs);
+}
+
+static int
+read_blobs_in_solid_resource(struct blob_descriptor *first_blob,
+ struct blob_descriptor *last_blob,
+ size_t blob_count,
+ size_t list_head_offset,
+ const struct read_blob_callbacks *sink_cbs)
+{
+ struct data_range *ranges;
+ bool ranges_malloced;
+ struct blob_descriptor *cur_blob;
+ size_t i;
+ int ret;
+ u64 ranges_alloc_size;
+
+ /* Setup data ranges array (one range per blob to read); this way
+ * read_compressed_wim_resource() does not need to be aware of blobs.
+ */
+
+ ranges_alloc_size = (u64)blob_count * sizeof(ranges[0]);
+
+ if (unlikely((size_t)ranges_alloc_size != ranges_alloc_size))
+ goto oom;
+
+ if (ranges_alloc_size <= STACK_MAX) {
+ ranges = alloca(ranges_alloc_size);
+ ranges_malloced = false;
+ } else {
+ ranges = MALLOC(ranges_alloc_size);
+ if (unlikely(!ranges))
+ goto oom;
+ 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;
+ }
+
+ 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,
+ };
+ struct read_blob_callbacks cbs = {
+ .consume_chunk = blobifier_cb,
+ .ctx = &blobifier_ctx,
+ };
+
+ ret = read_compressed_wim_resource(first_blob->rdesc, ranges,
+ blob_count, &cbs);
+
+ if (ranges_malloced)
+ FREE(ranges);
+
+ if (unlikely(ret && blobifier_ctx.cur_blob_offset != 0)) {
+ ret = call_end_blob(blobifier_ctx.cur_blob, ret,
+ &blobifier_ctx.cbs);
+ }
+ return ret;
+
+oom:
+ ERROR("Too many blobs in one resource!");
+ return WIMLIB_ERR_NOMEM;
+}
+
+/*
+ * 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:
+ *
+ * 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.
+ *
+ * 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.
*
- * @extract_chunk is a function that is called to extract each chunk of the
- * resource. */
+ * 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
-extract_wim_resource(const struct wim_lookup_table_entry *lte,
- u64 size,
- consume_data_callback_t extract_chunk,
- void *extract_chunk_arg)
+read_blob_list(struct list_head *blob_list, size_t list_head_offset,
+ const struct read_blob_callbacks *cbs, int flags)
{
int ret;
- if (size == wim_resource_size(lte)) {
- /* 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_resource_prefix(lte, size,
- extract_chunk_sha1_wrapper,
- wim_resource_chunk_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 resource:");
- 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;
+ struct list_head *cur, *next;
+ struct blob_descriptor *blob;
+ struct hasher_context *hasher_ctx;
+ struct read_blob_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;
+ }
+
+ 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_callbacks) {
+ .begin_blob = hasher_begin_blob,
+ .consume_chunk = hasher_consume_chunk,
+ .end_blob = hasher_end_blob,
+ .ctx = hasher_ctx,
+ };
+ } else {
+ sink_cbs = (struct read_blob_callbacks *)cbs;
+ }
+
+ 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;
+ size_t 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;
}
}
- } else {
- /* Don't do SHA1 */
- ret = read_resource_prefix(lte, size, extract_chunk,
- wim_resource_chunk_size(lte),
- extract_chunk_arg, 0);
+
+ ret = read_blob_with_cbs(blob, sink_cbs);
+ if (unlikely(ret && ret != BEGIN_BLOB_STATUS_SKIP_BLOB))
+ return ret;
}
- return ret;
+ return 0;
}
static int
-extract_wim_chunk_to_fd(const void *buf, size_t len, void *_fd_p)
+extract_chunk_to_fd(const void *chunk, size_t size, void *_fd)
{
- struct filedes *fd = _fd_p;
- int ret = full_write(fd, buf, len);
- if (ret)
+ struct filedes *fd = _fd;
+ int ret = full_write(fd, chunk, size);
+ if (unlikely(ret))
ERROR_WITH_ERRNO("Error writing to file descriptor");
return ret;
}
+/* Extract the first @size bytes of the specified blob to the specified file
+ * descriptor. This does *not* check the SHA-1 message digest. */
int
-extract_wim_resource_to_fd(const struct wim_lookup_table_entry *lte,
- struct filedes *fd, u64 size)
+extract_blob_prefix_to_fd(struct blob_descriptor *blob, u64 size,
+ struct filedes *fd)
{
- return extract_wim_resource(lte, size, extract_wim_chunk_to_fd, fd);
+ struct read_blob_callbacks cbs = {
+ .consume_chunk = extract_chunk_to_fd,
+ .ctx = fd,
+ };
+ return read_blob_prefix(blob, size, &cbs);
}
-
-static int
-sha1_chunk(const void *buf, size_t len, void *ctx)
+/* Extract the full uncompressed contents of the specified blob to the specified
+ * file descriptor. This checks the SHA-1 message digest. */
+int
+extract_blob_to_fd(struct blob_descriptor *blob, struct filedes *fd)
{
- sha1_update(ctx, buf, len);
- return 0;
+ struct read_blob_callbacks cbs = {
+ .consume_chunk = extract_chunk_to_fd,
+ .ctx = fd,
+ };
+ return read_blob_with_sha1(blob, &cbs);
}
-/* Calculate the SHA1 message digest of a stream. */
+/* Calculate the SHA-1 message digest of a blob and store it in @blob->hash. */
int
-sha1_resource(struct wim_lookup_table_entry *lte)
+sha1_blob(struct blob_descriptor *blob)
{
- int ret;
- SHA_CTX sha_ctx;
-
- sha1_init(&sha_ctx);
- ret = read_resource_prefix(lte, wim_resource_size(lte),
- sha1_chunk, wim_resource_chunk_size(lte),
- &sha_ctx, 0);
- if (ret == 0)
- sha1_final(lte->hash, &sha_ctx);
- return ret;
+ struct read_blob_callbacks cbs = {
+ };
+ return read_blob_with_sha1(blob, &cbs);
}
-/* Translates a WIM resource entry from the on-disk format to an in-memory
- * format. */
+/*
+ * 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
-get_resource_entry(const struct resource_entry_disk *disk_entry,
- struct resource_entry *entry)
+wim_reshdr_to_desc(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ struct wim_resource_descriptor *rdesc)
{
- /* Note: disk_entry may not be 8 byte aligned--- in that case, the
- * offset and original_size members will be unaligned. (This should be
- * okay since `struct resource_entry_disk' is declared as packed.) */
-
- /* Read the size and flags into a bitfield portably... */
- entry->size = (((u64)disk_entry->size[0] << 0) |
- ((u64)disk_entry->size[1] << 8) |
- ((u64)disk_entry->size[2] << 16) |
- ((u64)disk_entry->size[3] << 24) |
- ((u64)disk_entry->size[4] << 32) |
- ((u64)disk_entry->size[5] << 40) |
- ((u64)disk_entry->size[6] << 48));
- entry->flags = disk_entry->flags;
- entry->offset = le64_to_cpu(disk_entry->offset);
- entry->original_size = le64_to_cpu(disk_entry->original_size);
-
- /* offset and original_size are truncated to 62 bits to avoid possible
- * overflows, when converting to a signed 64-bit integer (off_t) or when
- * adding size or original_size. This is okay since no one would ever
- * actually have a WIM bigger than 4611686018427387903 bytes... */
- if (entry->offset & 0xc000000000000000ULL) {
- WARNING("Truncating offset in resource entry");
- entry->offset &= 0x3fffffffffffffffULL;
- }
- if (entry->original_size & 0xc000000000000000ULL) {
- WARNING("Truncating original_size in resource entry");
- entry->original_size &= 0x3fffffffffffffffULL;
+ 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;
}
}
-/* Translates a WIM resource entry from an in-memory format into the on-disk
- * format. */
+/*
+ * Convert the short WIM resource header @reshdr to a stand-alone WIM resource
+ * descriptor @rdesc, then set @blob to consist of that entire resource. This
+ * should only be used for non-solid resources!
+ */
+void
+wim_reshdr_to_desc_and_blob(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ struct wim_resource_descriptor *rdesc,
+ struct blob_descriptor *blob)
+{
+ wim_reshdr_to_desc(reshdr, wim, rdesc);
+ blob->size = rdesc->uncompressed_size;
+ blob_set_is_located_in_wim_resource(blob, rdesc, 0);
+}
+
+/* Import a WIM resource header from the on-disk 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;
+}
+
+/* Export a WIM resource header to the on-disk format. */
void
-put_resource_entry(const struct resource_entry *entry,
- struct resource_entry_disk *disk_entry)
+put_wim_reshdr(const struct wim_reshdr *reshdr,
+ struct wim_reshdr_disk *disk_reshdr)
{
- /* Note: disk_entry may not be 8 byte aligned--- in that case, the
- * offset and original_size members will be unaligned. (This should be
- * okay since `struct resource_entry_disk' is declared as packed.) */
- u64 size = entry->size;
-
- disk_entry->size[0] = size >> 0;
- disk_entry->size[1] = size >> 8;
- disk_entry->size[2] = size >> 16;
- disk_entry->size[3] = size >> 24;
- disk_entry->size[4] = size >> 32;
- disk_entry->size[5] = size >> 40;
- disk_entry->size[6] = size >> 48;
- disk_entry->flags = entry->flags;
- disk_entry->offset = cpu_to_le64(entry->offset);
- disk_entry->original_size = cpu_to_le64(entry->original_size);
+ 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);
}