/*
* resource.c
*
- * Code for reading streams and resources, including compressed WIM resources.
+ * 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 <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 <stdlib.h>
-#include <unistd.h>
+#include "wimlib/wim.h"
+#include "wimlib/win32.h"
/*
* Compressed WIM resources
*
- * A compressed resource in a WIM consists of a number of compressed chunks,
- * each of which decompresses to a fixed chunk size (given in the WIM header;
- * usually 32768) except possibly the last, which always decompresses to any
- * remaining bytes. In addition, immediately before the chunks, a table (the
- * "chunk table") provides the offset, in bytes relative to the end of the chunk
- * table, of the start of each compressed chunk, except for the first chunk
- * which is omitted as it always has an offset of 0. Therefore, a compressed
- * resource with N chunks will have a chunk table with N - 1 entries.
- *
- * Additional information:
+ * 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. 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.
*/
-/* Decompress the specified chunk that uses the specified compression type
- * @ctype, part of a WIM with default chunk size @wim_chunk_size. For LZX the
- * separate @wim_chunk_size is needed because it determines the window size used
- * for LZX compression. */
-static int
-decompress(const void *cchunk, unsigned clen, void *uchunk, unsigned ulen,
- int ctype, u32 wim_chunk_size)
-{
- switch (ctype) {
- case WIMLIB_COMPRESSION_TYPE_LZX:
- return wimlib_lzx_decompress2(cchunk, clen,
- uchunk, ulen, wim_chunk_size);
- case WIMLIB_COMPRESSION_TYPE_XPRESS:
- return wimlib_xpress_decompress(cchunk, clen,
- uchunk, ulen);
- case WIMLIB_COMPRESSION_TYPE_LZMS:
- return wimlib_lzms_decompress(cchunk, clen, uchunk, ulen);
- default:
- ERROR("Invalid compression format (%d)", ctype);
- return -1;
- }
-}
-
struct data_range {
u64 offset;
u64 size;
};
/*
- * read_compressed_wim_resource() -
- *
* Read data from a compressed WIM resource.
*
- * @rspec
- * Specification of the compressed WIM resource to read from.
+ * @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.
+ * @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.
*
* Possible return values:
*
* 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_CHUNK_SIZE (errno set to EINVAL)
*
- * or other error code returned by the @cb function.
+ * or other error code returned by the cbs->consume_chunk() function.
*/
static int
-read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
+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)
+ const struct read_blob_callbacks *cbs)
{
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(rspec != NULL);
- wimlib_assert(resource_is_compressed(rspec));
- 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,
- rspec->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 <= rspec->uncompressed_size);
+ 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);
const u64 last_offset = ranges[num_ranges - 1].offset + ranges[num_ranges - 1].size - 1;
/* Get the file descriptor for the WIM. */
- struct filedes * const in_fd = &rspec->wim->in_fd;
+ 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 = !filedes_is_seekable(in_fd);
+ const bool is_pipe_read = (rdesc->is_pipable && !filedes_is_seekable(in_fd));
- /* Determine if the chunk table is in an altenate format. */
- const bool alt_chunk_table = (rspec->flags & WIM_RESHDR_FLAG_PACKED_STREAMS)
+ /* 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. */
- u32 chunk_size;
- u64 cur_read_offset = rspec->offset_in_wim;
- int ctype;
+ 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. */
- struct alt_chunk_table_header_disk hdr;
-
- ret = full_pread(in_fd, &hdr, sizeof(hdr), cur_read_offset);
- if (ret)
- goto read_error;
- cur_read_offset += sizeof(hdr);
-
- chunk_size = le32_to_cpu(hdr.chunk_size);
- ctype = le32_to_cpu(hdr.compression_format);
-
- /* Format numbers must be the same as in WIMGAPI to be
- * compatible. */
- BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
- BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_LZX != 1);
- BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_XPRESS != 2);
- BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_LZMS != 3);
- } else {
- /* "Normal" format: the maximum uncompressed chunk size and the
- * compression format default to those of the WIM itself. */
- chunk_size = rspec->wim->chunk_size;
- ctype = rspec->wim->compression_type;
+ * 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);
}
- if (!is_power_of_2(chunk_size)) {
+
+ if (unlikely(!is_power_of_2(chunk_size))) {
ERROR("Invalid compressed resource: "
- "expected power-of-2 chunk size (got %u)", chunk_size);
+ "expected power-of-2 chunk size (got %"PRIu32")",
+ chunk_size);
ret = WIMLIB_ERR_INVALID_CHUNK_SIZE;
- goto out_free_memory;
+ errno = EINVAL;
+ goto out_cleanup;
}
- const u32 chunk_order = bsr32(chunk_size);
+ /* 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;
+ }
+ }
+
+ const u32 chunk_order = fls32(chunk_size);
/* Calculate the total number of chunks the resource is divided into. */
- const u64 num_chunks = (rspec->uncompressed_size + chunk_size - 1) >> chunk_order;
+ 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 num_chunk_entries = (alt_chunk_table ? num_chunks : num_chunks - 1);
/* Set the size of each chunk table entry based on the resource's
- * uncompressed size. XXX: Does the alternate chunk table really
- * always have 4-byte entries? */
- const u64 chunk_entry_size =
- (rspec->uncompressed_size > (1ULL << 32) && !alt_chunk_table)
- ? 8 : 4;
+ * 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;
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)
+ if (unlikely((size_t)chunk_offsets_alloc_size != chunk_offsets_alloc_size)) {
+ errno = ENOMEM;
goto oom;
+ }
- if (chunk_offsets_alloc_size <= STACK_MAX) {
+ if (likely(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)
+ if (unlikely(!chunk_offsets))
goto oom;
chunk_offsets_malloced = true;
}
const u64 file_offset_of_needed_chunk_entries =
cur_read_offset
+ (first_chunk_entry_to_read * chunk_entry_size)
- + (rspec->is_pipable ? (rspec->size_in_wim - chunk_table_size) : 0);
+ + (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,
+ ret = full_pread(in_fd, chunk_table_data, chunk_table_size_to_read,
file_offset_of_needed_chunk_entries);
- if (ret)
+ 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 __attribute__((may_alias)) aliased_le64_t;
- typedef le32 __attribute__((may_alias)) aliased_le32_t;
+ 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) {
/* Set offset to beginning of first chunk to read. */
cur_read_offset += chunk_offsets[0];
- if (rspec->is_pipable)
+ if (rdesc->is_pipable)
cur_read_offset += read_start_chunk * sizeof(struct pwm_chunk_hdr);
else
cur_read_offset += chunk_table_size;
ubuf = alloca(chunk_size);
} else {
ubuf = MALLOC(chunk_size);
- if (ubuf == NULL)
+ if (unlikely(!ubuf))
goto oom;
ubuf_malloced = true;
}
cbuf = alloca(chunk_size - 1);
} else {
cbuf = MALLOC(chunk_size - 1);
- if (cbuf == NULL)
+ if (unlikely(!cbuf))
goto oom;
cbuf_malloced = true;
}
/* Calculate uncompressed size of next chunk. */
u32 chunk_usize;
- if ((i == num_chunks - 1) && (rspec->uncompressed_size & (chunk_size - 1)))
- chunk_usize = (rspec->uncompressed_size & (chunk_size - 1));
+ if ((i == num_chunks - 1) && (rdesc->uncompressed_size & (chunk_size - 1)))
+ chunk_usize = (rdesc->uncompressed_size & (chunk_size - 1));
else
chunk_usize = chunk_size;
ret = full_pread(in_fd, &chunk_hdr,
sizeof(chunk_hdr), cur_read_offset);
- if (ret)
+ if (unlikely(ret))
goto read_error;
chunk_csize = le32_to_cpu(chunk_hdr.compressed_size);
} else {
if (i == num_chunks - 1) {
- chunk_csize = rspec->size_in_wim -
+ chunk_csize = rdesc->size_in_wim -
chunk_table_full_size -
chunk_offsets[i - read_start_chunk];
- if (rspec->is_pipable)
+ 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) {
+ if (unlikely(chunk_csize == 0 || chunk_csize > chunk_usize)) {
ERROR("Invalid chunk size in compressed resource!");
errno = EINVAL;
ret = WIMLIB_ERR_DECOMPRESSION;
- goto out_free_memory;
+ goto out_cleanup;
}
- if (rspec->is_pipable)
+ if (rdesc->is_pipable)
cur_read_offset += sizeof(struct pwm_chunk_hdr);
/* Offsets in the uncompressed resource at which this chunk
u8 dummy;
ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
- if (ret)
+ if (unlikely(ret))
goto read_error;
}
} else {
read_buf,
chunk_csize,
cur_read_offset);
- if (ret)
+ if (unlikely(ret))
goto read_error;
if (read_buf == cbuf) {
- DEBUG("Decompressing chunk %"PRIu64" "
- "(csize=%"PRIu32" usize=%"PRIu32")",
- i, chunk_csize, chunk_usize);
- ret = decompress(cbuf,
- chunk_csize,
- ubuf,
- chunk_usize,
- ctype,
- chunk_size);
- if (ret) {
+ 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_free_memory;
+ goto out_cleanup;
}
}
cur_read_offset += chunk_csize;
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;
+ 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) {
}
if (is_pipe_read &&
- last_offset == rspec->uncompressed_size - 1 &&
+ last_offset == rdesc->uncompressed_size - 1 &&
chunk_table_size)
{
u8 dummy;
* returns. */
cur_read_offset += chunk_table_size;
ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
- if (ret)
+ if (unlikely(ret))
goto read_error;
}
ret = 0;
-out_free_memory:
- errno_save = errno;
+
+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 (ubuf_malloced)
FREE(ubuf);
if (cbuf_malloced)
FREE(cbuf);
- errno = errno_save;
return ret;
oom:
- ERROR("Not enough memory available to read size=%"PRIu64" bytes "
- "from compressed WIM resource!", last_offset - first_offset + 1);
- errno = ENOMEM;
+ ERROR("Out of memory while reading compressed WIM resource");
ret = WIMLIB_ERR_NOMEM;
- goto out_free_memory;
+ goto out_cleanup;
read_error:
- ERROR_WITH_ERRNO("Error reading compressed WIM resource!");
- goto out_free_memory;
+ ERROR_WITH_ERRNO("Error reading data from WIM file");
+ goto out_cleanup;
}
/* Read raw data from a file descriptor at the specified offset, feeding the
- * data it in chunks into the specified callback function. */
+ * data in nonempty chunks into the cbs->consume_chunk() function. */
static int
-read_raw_file_data(struct filedes *in_fd, u64 size,
- consume_data_callback_t cb, void *cb_ctx, u64 offset)
+read_raw_file_data(struct filedes *in_fd, u64 offset, u64 size,
+ const struct read_blob_callbacks *cbs)
{
u8 buf[BUFFER_SIZE];
size_t bytes_to_read;
while (size) {
bytes_to_read = min(sizeof(buf), size);
ret = full_pread(in_fd, buf, bytes_to_read, offset);
- if (ret) {
+ if (unlikely(ret)) {
ERROR_WITH_ERRNO("Read error");
return ret;
}
- ret = cb(buf, bytes_to_read, cb_ctx);
- if (ret)
+ ret = call_consume_chunk(buf, bytes_to_read, cbs);
+ if (unlikely(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. */
+/* 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)
{
- u8 **buf_p = _ctx;
+ void **buf_p = _ctx;
*buf_p = mempcpy(*buf_p, chunk, size);
return 0;
}
/*
- * read_partial_wim_resource()-
- *
- * Read a range of data from an uncompressed or compressed resource in a WIM
- * file.
+ * Read @size bytes at @offset in the WIM resource described by @rdesc and feed
+ * the data into the @cbs->consume_chunk callback function.
*
- * @rspec
- * Specification 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.
- *
- * 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)
+ * @offset and @size are assumed to have already been validated against the
+ * resource's uncompressed size.
*
- * 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.
*/
static int
-read_partial_wim_resource(const struct wim_resource_spec *rspec,
- u64 offset, u64 size,
- consume_data_callback_t cb, void *cb_ctx)
+read_partial_wim_resource(const struct wim_resource_descriptor *rdesc,
+ const u64 offset, const u64 size,
+ const struct read_blob_callbacks *cbs)
{
- /* Sanity checks. */
- wimlib_assert(offset + size >= offset);
- wimlib_assert(offset + size <= rspec->uncompressed_size);
-
- DEBUG("Reading %"PRIu64" @ %"PRIu64" from WIM resource "
- "%"PRIu64" => %"PRIu64" @ %"PRIu64,
- size, offset, rspec->uncompressed_size,
- rspec->size_in_wim, rspec->offset_in_wim);
-
- /* Trivial case. */
- if (size == 0)
- return 0;
-
- if (resource_is_compressed(rspec)) {
+ if (rdesc->flags & (WIM_RESHDR_FLAG_COMPRESSED |
+ WIM_RESHDR_FLAG_SOLID))
+ {
+ /* Compressed resource */
+ if (unlikely(!size))
+ return 0;
struct data_range range = {
.offset = offset,
.size = size,
};
- return read_compressed_wim_resource(rspec, &range, 1,
- cb, cb_ctx);
- } else {
- return read_raw_file_data(&rspec->wim->in_fd,
- size,
- cb,
- cb_ctx,
- rspec->offset_in_wim + offset);
+ return read_compressed_wim_resource(rdesc, &range, 1, cbs);
}
+
+ /* Uncompressed resource */
+ return read_raw_file_data(&rdesc->wim->in_fd,
+ rdesc->offset_in_wim + offset,
+ size, cbs);
}
-/* Read the specified range of uncompressed data from the specified stream,
- * which must be located into a WIM file, into the specified buffer. */
+/* 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_stream_into_buf(const struct wim_lookup_table_entry *lte,
- size_t size, u64 offset, void *_buf)
+read_partial_wim_blob_into_buf(const struct blob_descriptor *blob,
+ u64 offset, size_t size, void *buf)
{
- u8 *buf = _buf;
-
- wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
-
- return read_partial_wim_resource(lte->rspec,
- lte->offset_in_res + offset,
+ struct read_blob_callbacks cbs = {
+ .consume_chunk = bufferer_cb,
+ .ctx = &buf,
+ };
+ return read_partial_wim_resource(blob->rdesc,
+ blob->offset_in_res + offset,
size,
- bufferer_cb,
- &buf);
-}
-
-/* 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;
+ &cbs);
}
-/* Skip over the data of the specified stream, which must correspond to a full
- * WIM resource. */
+/* Skip over the data of the specified WIM resource. */
int
-skip_wim_stream(struct wim_lookup_table_entry *lte)
+skip_wim_resource(const struct wim_resource_descriptor *rdesc)
{
- wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
- wimlib_assert(!(lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS));
- DEBUG("Skipping stream (size=%"PRIu64")", lte->size);
- return read_partial_wim_resource(lte->rspec,
- 0,
- lte->rspec->uncompressed_size,
- skip_chunk_cb,
- NULL);
+ struct read_blob_callbacks cbs = {
+ };
+ return read_partial_wim_resource(rdesc, 0,
+ rdesc->uncompressed_size, &cbs);
}
static int
-read_wim_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
- consume_data_callback_t cb, void *cb_ctx)
+read_wim_blob_prefix(const struct blob_descriptor *blob, u64 size,
+ const struct read_blob_callbacks *cbs)
{
- return read_partial_wim_resource(lte->rspec, lte->offset_in_res, size,
- cb, cb_ctx);
+ return read_partial_wim_resource(blob->rdesc, blob->offset_in_res,
+ size, cbs);
}
-#ifndef __WIN32__
-/* This function handles reading stream data that is located in an external
- * file, such as a file that has been added to the WIM image through execution
- * of a wimlib_add_command.
+/* 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, void *cb_ctx)
+read_file_on_disk_prefix(const struct blob_descriptor *blob, u64 size,
+ const struct read_blob_callbacks *cbs)
{
int ret;
int raw_fd;
struct filedes fd;
- wimlib_assert(size <= lte->size);
+ 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);
+ ret = read_raw_file_data(&fd, 0, size, cbs);
+ filedes_close(&fd);
+ return ret;
+}
- DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", size, lte->file_on_disk);
+#ifdef WITH_FUSE
+static int
+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;
- raw_fd = open(lte->file_on_disk, O_BINARY | O_RDONLY);
- if (raw_fd < 0) {
- ERROR_WITH_ERRNO("Can't open \"%"TS"\"", lte->file_on_disk);
+ 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;
}
filedes_init(&fd, raw_fd);
- ret = read_raw_file_data(&fd, size, cb, cb_ctx, 0);
+ ret = read_raw_file_data(&fd, 0, size, cbs);
filedes_close(&fd);
return ret;
}
-#endif /* !__WIN32__ */
+#endif
-/* This function handles the trivial case of reading stream data that is, in
- * fact, already located in an in-memory buffer. */
+/* 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 wim_lookup_table_entry *lte,
- u64 size, consume_data_callback_t cb, void *cb_ctx)
+read_buffer_prefix(const struct blob_descriptor *blob,
+ u64 size, const struct read_blob_callbacks *cbs)
{
- wimlib_assert(size <= lte->size);
- return (*cb)(lte->attached_buffer, size, cb_ctx);
+ if (unlikely(!size))
+ return 0;
+ return call_consume_chunk(blob->attached_buffer, size, cbs);
}
-typedef int (*read_stream_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
- u64 size,
- consume_data_callback_t cb,
- void *cb_ctx);
+typedef int (*read_blob_prefix_handler_t)(const struct blob_descriptor *blob,
+ u64 size,
+ const struct read_blob_callbacks *cbs);
/*
- * read_stream_prefix()-
- *
- * Reads the first @size bytes from a generic "stream", which may be located in
- * any one of several locations, such as in a WIM file (compressed or
- * uncompressed), in an external file, or directly in an in-memory buffer.
- *
- * This function feeds the data to a callback function @cb in chunks of
- * unspecified size.
+ * 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.
*
* Returns 0 on success; nonzero on error. A nonzero value will be returned if
- * the stream data cannot be successfully read (for a number of different
- * reasons, depending on the stream location), or if @cb returned nonzero in
- * which case that error code will be returned.
+ * 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_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
- consume_data_callback_t cb, void *cb_ctx)
+static int
+read_blob_prefix(const struct blob_descriptor *blob, u64 size,
+ const struct read_blob_callbacks *cbs)
{
- static const read_stream_prefix_handler_t handlers[] = {
- [RESOURCE_IN_WIM] = read_wim_stream_prefix,
- #ifdef __WIN32__
- [RESOURCE_IN_FILE_ON_DISK] = read_win32_file_prefix,
- #else
- [RESOURCE_IN_FILE_ON_DISK] = read_file_on_disk_prefix,
- #endif
- [RESOURCE_IN_ATTACHED_BUFFER] = read_buffer_prefix,
+ 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_WINNT_FILE_ON_DISK] = read_winnt_stream_prefix,
+ [BLOB_WIN32_ENCRYPTED] = read_win32_encrypted_file_prefix,
#endif
};
- wimlib_assert(lte->resource_location < ARRAY_LEN(handlers)
- && handlers[lte->resource_location] != NULL);
- return handlers[lte->resource_location](lte, size, cb, cb_ctx);
+ 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 uncompressed data of the specified stream into the specified
- * buffer, which must have space for at least lte->size bytes. */
+/* Read the full data of the specified blob, passing the data into the specified
+ * callbacks (all of which are optional). */
int
-read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *_buf)
+read_blob_with_cbs(struct blob_descriptor *blob,
+ const struct read_blob_callbacks *cbs)
{
- u8 *buf = _buf;
- return read_stream_prefix(lte, lte->size, bufferer_cb, &buf);
+ 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_blob_into_buf(const struct blob_descriptor *blob, void *buf)
+{
+ 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 stream. A buffer large
- * enough hold the data is allocated and returned in @buf_ret. */
+/* 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_stream_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->size != lte->size) {
- ERROR("Can't read %"PRIu64" byte stream into "
- "memory", lte->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->size);
- if (buf == NULL)
+ buf = MALLOC(blob->size);
+ if (unlikely(!buf))
return WIMLIB_ERR_NOMEM;
- ret = read_full_stream_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 the specified WIM resource. A buffer
- * large enough hold the data is allocated and returned in @buf_ret. */
-static int
-wim_resource_spec_to_data(struct wim_resource_spec *rspec, void **buf_ret)
+/* 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)
{
- int ret;
- struct wim_lookup_table_entry *lte;
-
- lte = new_lookup_table_entry();
- if (lte == NULL)
- return WIMLIB_ERR_NOMEM;
-
- lte_bind_wim_resource_spec(lte, rspec);
- lte->flags = rspec->flags;
- lte->size = rspec->uncompressed_size;
- lte->offset_in_res = 0;
+ struct wim_resource_descriptor rdesc;
+ struct blob_descriptor blob;
- ret = read_full_stream_into_alloc_buf(lte, buf_ret);
+ wim_reshdr_to_desc_and_blob(reshdr, wim, &rdesc, &blob);
- lte_unbind_wim_resource_spec(lte);
- free_lookup_table_entry(lte);
- return ret;
+ return read_blob_into_alloc_buf(&blob, buf_ret);
}
-/* Retrieve the full uncompressed data of a WIM resource specified as a raw
- * `wim_reshdr' and the corresponding WIM file. A large enough hold the data is
- * allocated and returned in @buf_ret. */
+/* Calculate the SHA-1 message digest of the uncompressed data of the specified
+ * WIM resource. */
int
-wim_reshdr_to_data(const struct wim_reshdr *reshdr, WIMStruct *wim, void **buf_ret)
+wim_reshdr_to_hash(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ u8 hash[SHA1_HASH_SIZE])
{
- DEBUG("offset_in_wim=%"PRIu64", size_in_wim=%"PRIu64", "
- "uncompressed_size=%"PRIu64,
- reshdr->offset_in_wim, reshdr->size_in_wim,
- reshdr->uncompressed_size);
-
- struct wim_resource_spec rspec;
- wim_res_hdr_to_spec(reshdr, wim, &rspec);
- return wim_resource_spec_to_data(&rspec, buf_ret);
+ struct wim_resource_descriptor rdesc;
+ struct blob_descriptor blob;
+ int ret;
+
+ wim_reshdr_to_desc_and_blob(reshdr, wim, &rdesc, &blob);
+ blob.unhashed = 1;
+
+ ret = sha1_blob(&blob);
+ if (unlikely(ret))
+ return ret;
+
+ copy_hash(hash, blob.hash);
+ return 0;
}
-struct streamifier_context {
- struct read_stream_list_callbacks cbs;
- struct wim_lookup_table_entry *cur_stream;
- u64 cur_stream_offset;
- struct wim_lookup_table_entry *final_stream;
+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;
};
-/* A consume_data_callback_t implementation that translates raw resource data
- * into streams, calling the begin_stream, consume_chunk, and end_stream
- * callback functions as appropriate. */
+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
-streamifier_cb(const void *chunk, size_t size, void *_ctx)
+blobifier_cb(const void *chunk, size_t size, void *_ctx)
{
- struct streamifier_context *ctx = _ctx;
+ struct blobifier_context *ctx = _ctx;
int ret;
- DEBUG("%zu bytes passed to streamifier", size);
-
- wimlib_assert(ctx->cur_stream != NULL);
- wimlib_assert(size <= ctx->cur_stream->size - ctx->cur_stream_offset);
+ wimlib_assert(ctx->cur_blob != NULL);
+ wimlib_assert(size <= ctx->cur_blob->size - ctx->cur_blob_offset);
- if (ctx->cur_stream_offset == 0) {
- /* Starting a new stream. */
- DEBUG("Begin new stream (size=%"PRIu64").",
- ctx->cur_stream->size);
- ret = (*ctx->cbs.begin_stream)(ctx->cur_stream, true,
- ctx->cbs.begin_stream_ctx);
+ if (ctx->cur_blob_offset == 0) {
+ /* Starting a new blob. */
+ ret = call_begin_blob(ctx->cur_blob, &ctx->cbs);
if (ret)
return ret;
}
- /* Consume the chunk. */
- ret = (*ctx->cbs.consume_chunk)(chunk, size,
- ctx->cbs.consume_chunk_ctx);
+ ctx->cur_blob_offset += size;
+
+ ret = call_consume_chunk(chunk, size, &ctx->cbs);
if (ret)
return ret;
- ctx->cur_stream_offset += size;
-
- if (ctx->cur_stream_offset == ctx->cur_stream->size) {
- /* Finished reading all the data for a stream. */
- struct list_head *cur, *next;
+ if (ctx->cur_blob_offset == ctx->cur_blob->size) {
+ /* Finished reading all the data for a blob. */
- cur = (struct list_head *)
- ((u8*)ctx->cur_stream + ctx->list_head_offset);
- next = cur->next;
+ ctx->cur_blob_offset = 0;
- DEBUG("End stream (size=%"PRIu64").", ctx->cur_stream->size);
- ret = (*ctx->cbs.end_stream)(ctx->cur_stream, 0,
- ctx->cbs.end_stream_ctx);
+ ret = call_end_blob(ctx->cur_blob, 0, &ctx->cbs);
if (ret)
return ret;
- if (ctx->cur_stream != ctx->final_stream) {
- /* Advance to next stream. */
- ctx->cur_stream = (struct wim_lookup_table_entry *)
- ((u8*)next - ctx->list_head_offset);
-
- ctx->cur_stream_offset = 0;
- } else {
- /* No more streams. */
- ctx->cur_stream = NULL;
+ /* 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 hasher_context {
SHA_CTX sha_ctx;
int flags;
- struct read_stream_list_callbacks cbs;
+ struct read_blob_callbacks cbs;
};
-/* Callback for starting to read a stream while calculating its SHA1 message
+/* Callback for starting to read a blob while calculating its SHA-1 message
* digest. */
static int
-hasher_begin_stream(struct wim_lookup_table_entry *lte, bool is_partial_res,
- void *_ctx)
+hasher_begin_blob(struct blob_descriptor *blob, void *_ctx)
{
struct hasher_context *ctx = _ctx;
sha1_init(&ctx->sha_ctx);
- if (ctx->cbs.begin_stream == NULL)
- return 0;
- else
- return (*ctx->cbs.begin_stream)(lte, is_partial_res,
- ctx->cbs.begin_stream_ctx);
+ return call_begin_blob(blob, &ctx->cbs);
}
-/* A consume_data_callback_t implementation that continues calculating the SHA1
- * message digest of the stream being read, then optionally passes the data on
- * to another consume_data_callback_t implementation. This allows checking the
- * SHA1 message digest of a stream being extracted, for example. */
+/* 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
hasher_consume_chunk(const void *chunk, size_t size, void *_ctx)
{
struct hasher_context *ctx = _ctx;
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);
+
+ return call_consume_chunk(chunk, size, &ctx->cbs);
}
-/* Callback for finishing reading a stream while calculating its SHA1 message
+/* Callback for finishing reading a blob while calculating its SHA-1 message
* digest. */
static int
-hasher_end_stream(struct wim_lookup_table_entry *lte, int status, void *_ctx)
+hasher_end_blob(struct blob_descriptor *blob, int status, void *_ctx)
{
struct hasher_context *ctx = _ctx;
u8 hash[SHA1_HASH_SIZE];
int ret;
- if (status) {
- /* Error occurred; the full stream may not have been read. */
+ if (unlikely(status)) {
+ /* Error occurred; the full blob may not have been read. */
ret = status;
goto out_next_cb;
}
- /* Retrieve the final SHA1 message digest. */
+ /* Retrieve the final SHA-1 message digest. */
sha1_final(hash, &ctx->sha_ctx);
- if (lte->unhashed) {
- if (ctx->flags & COMPUTE_MISSING_STREAM_HASHES) {
- /* No SHA1 message digest was previously present for the
- * stream. Set it to the one just calculated. */
- DEBUG("Set SHA1 message digest for stream "
- "(size=%"PRIu64").", lte->size);
- copy_hash(lte->hash, hash);
- }
- } else {
- if (ctx->flags & VERIFY_STREAM_HASHES) {
- /* The stream already had a SHA1 message digest present. Verify
- * that it is the same as the calculated value. */
- if (!hashes_equal(hash, lte->hash)) {
- if (wimlib_print_errors) {
- ERROR("Invalid SHA1 message digest "
- "on the following WIM stream:");
- print_lookup_table_entry(lte, stderr);
- }
- ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
- errno = EINVAL;
- goto out_next_cb;
- }
- DEBUG("SHA1 message digest okay for "
- "stream (size=%"PRIu64").", lte->size);
+ /* 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)))
+ {
+ 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", got SHA-1=%"TS")",
+ expected_hashstr, actual_hashstr);
}
+ ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
+ goto out_next_cb;
}
ret = 0;
out_next_cb:
- if (ctx->cbs.end_stream == NULL)
- return ret;
- else
- return (*ctx->cbs.end_stream)(lte, ret, ctx->cbs.end_stream_ctx);
+ 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_full_stream_with_cbs(struct wim_lookup_table_entry *lte,
- const struct read_stream_list_callbacks *cbs)
+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;
- ret = (*cbs->begin_stream)(lte, false, cbs->begin_stream_ctx);
- if (ret)
- return ret;
+ /* Setup data ranges array (one range per blob to read); this way
+ * read_compressed_wim_resource() does not need to be aware of blobs.
+ */
- ret = read_stream_prefix(lte, lte->size, cbs->consume_chunk,
- cbs->consume_chunk_ctx);
+ ranges_alloc_size = (u64)blob_count * sizeof(ranges[0]);
- return (*cbs->end_stream)(lte, ret, cbs->end_stream_ctx);
-}
+ if (unlikely((size_t)ranges_alloc_size != ranges_alloc_size))
+ goto oom;
-/* Read the full data of the specified stream, passing the data into the
- * specified callbacks (all of which are optional) and either checking or
- * computing the SHA1 message digest of the stream. */
-static int
-read_full_stream_with_sha1(struct wim_lookup_table_entry *lte,
- const struct read_stream_list_callbacks *cbs)
-{
- struct hasher_context hasher_ctx = {
- .flags = VERIFY_STREAM_HASHES | COMPUTE_MISSING_STREAM_HASHES,
- .cbs = *cbs,
- };
- struct read_stream_list_callbacks hasher_cbs = {
- .begin_stream = hasher_begin_stream,
- .begin_stream_ctx = &hasher_ctx,
- .consume_chunk = hasher_consume_chunk,
- .consume_chunk_ctx = &hasher_ctx,
- .end_stream = hasher_end_stream,
- .end_stream_ctx = &hasher_ctx,
+ 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,
};
- return read_full_stream_with_cbs(lte, &hasher_cbs);
+
+ 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 streams, each of which may be in any supported location (e.g.
- * in a WIM or in an external file). Unlike read_stream_prefix() or the
- * functions which call it, this function optimizes the case where multiple
- * streams are packed into a single compressed WIM resource and reads them all
- * consecutively, only decompressing the data one time.
+ * 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.
*
- * @stream_list
- * List of streams (represented as `struct wim_lookup_table_entry's) to
- * read.
+ * @blob_list
+ * List of blobs to read.
* @list_head_offset
- * Offset of the `struct list_head' within each `struct
- * wim_lookup_table_entry' that makes up the @stream_list.
+ * Offset of the `struct list_head' within each `struct blob_descriptor'
+ * that makes up the @blob_list.
* @cbs
- * Callback functions to accept the stream data.
+ * 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.
*
+ * 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_stream_list(struct list_head *stream_list,
- size_t list_head_offset,
- const struct read_stream_list_callbacks *cbs,
- int flags)
+read_blob_list(struct list_head *blob_list, size_t list_head_offset,
+ const struct read_blob_callbacks *cbs, int flags)
{
int ret;
struct list_head *cur, *next;
- struct wim_lookup_table_entry *lte;
+ struct blob_descriptor *blob;
struct hasher_context *hasher_ctx;
- struct read_stream_list_callbacks *sink_cbs;
+ struct read_blob_callbacks *sink_cbs;
- if (!(flags & STREAM_LIST_ALREADY_SORTED)) {
- ret = sort_stream_list_by_sequential_order(stream_list, list_head_offset);
+ 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_STREAM_HASHES | COMPUTE_MISSING_STREAM_HASHES)) {
+ 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_stream_list_callbacks) {
- .begin_stream = hasher_begin_stream,
- .begin_stream_ctx = hasher_ctx,
- .consume_chunk = hasher_consume_chunk,
- .consume_chunk_ctx = hasher_ctx,
- .end_stream = hasher_end_stream,
- .end_stream_ctx = hasher_ctx,
+ *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_stream_list_callbacks*)cbs;
+ sink_cbs = (struct read_blob_callbacks *)cbs;
}
- for (cur = stream_list->next, next = cur->next;
- cur != stream_list;
+ for (cur = blob_list->next, next = cur->next;
+ cur != blob_list;
cur = next, next = cur->next)
{
- lte = (struct wim_lookup_table_entry*)((u8*)cur - list_head_offset);
+ blob = (struct blob_descriptor*)((u8*)cur - list_head_offset);
- if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
- lte->size != lte->rspec->uncompressed_size)
+ if (blob->blob_location == BLOB_IN_WIM &&
+ blob->size != blob->rdesc->uncompressed_size)
{
-
- struct wim_lookup_table_entry *lte_next, *lte_last;
+ struct blob_descriptor *blob_next, *blob_last;
struct list_head *next2;
- size_t stream_count;
+ size_t blob_count;
- /* The next stream is a proper sub-sequence of a WIM
- * resource. See if there are other streams in the same
+ /* 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_stream_list_by_sequential_order() sorted the
- * streams by offset in the WIM, this can be determined
- * by simply scanning forward in the list. */
+ * 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. */
- lte_last = lte;
- stream_count = 1;
+ blob_last = blob;
+ blob_count = 1;
for (next2 = next;
- next2 != stream_list
- && (lte_next = (struct wim_lookup_table_entry*)
+ next2 != blob_list
+ && (blob_next = (struct blob_descriptor*)
((u8*)next2 - list_head_offset),
- lte_next->resource_location == RESOURCE_IN_WIM
- && lte_next->rspec == lte->rspec);
+ blob_next->blob_location == BLOB_IN_WIM
+ && blob_next->rdesc == blob->rdesc);
next2 = next2->next)
{
- lte_last = lte_next;
- stream_count++;
+ blob_last = blob_next;
+ blob_count++;
}
- if (stream_count > 1) {
- /* Reading multiple streams combined into a
- * single WIM resource. They are in the stream
- * list, sorted by offset; @lte specifies the
- * first stream in the resource that needs to be
- * read and @lte_last specifies the last stream
- * in the resource that needs to be read. */
-
- DEBUG("Reading %zu streams combined in same "
- "WIM resource", stream_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;
-
- struct data_range ranges[stream_count];
-
- {
- struct list_head *next3;
- size_t i;
- struct wim_lookup_table_entry *lte_cur;
-
- next3 = cur;
- for (i = 0; i < stream_count; i++) {
- lte_cur = (struct wim_lookup_table_entry*)
- ((u8*)next3 - list_head_offset);
- ranges[i].offset = lte_cur->offset_in_res;
- ranges[i].size = lte_cur->size;
- next3 = next3->next;
- }
- }
-
- struct streamifier_context streamifier_ctx = {
- .cbs = *sink_cbs,
- .cur_stream = lte,
- .cur_stream_offset = 0,
- .final_stream = lte_last,
- .list_head_offset = list_head_offset,
- };
-
- ret = read_compressed_wim_resource(lte->rspec,
- ranges,
- stream_count,
- streamifier_cb,
- &streamifier_ctx);
-
- if (ret) {
- if (streamifier_ctx.cur_stream_offset != 0) {
- ret = (*streamifier_ctx.cbs.end_stream)
- (streamifier_ctx.cur_stream,
- ret,
- streamifier_ctx.cbs.end_stream_ctx);
- }
+ ret = read_blobs_in_solid_resource(blob, blob_last,
+ blob_count,
+ list_head_offset,
+ sink_cbs);
+ if (ret)
return ret;
- }
continue;
}
}
- ret = read_full_stream_with_cbs(lte, sink_cbs);
- if (ret && ret != BEGIN_STREAM_STATUS_SKIP_STREAM)
+ ret = read_blob_with_cbs(blob, sink_cbs);
+ if (unlikely(ret && ret != BEGIN_BLOB_STATUS_SKIP_BLOB))
return ret;
}
return 0;
}
-/* Extract the first @size bytes of the specified stream.
- *
- * If @size specifies the full uncompressed size of the stream, then the SHA1
- * message digest of the uncompressed stream is checked while being extracted.
- *
- * The uncompressed data of the resource is passed in chunks of unspecified size
- * to the @extract_chunk function, passing it @extract_chunk_arg. */
-int
-extract_stream(struct wim_lookup_table_entry *lte, u64 size,
- consume_data_callback_t extract_chunk, void *extract_chunk_arg)
+static int
+extract_chunk_to_fd(const void *chunk, size_t size, void *_fd)
{
- if (size == lte->size) {
- /* Do SHA1. */
- struct read_stream_list_callbacks cbs = {
- .consume_chunk = extract_chunk,
- .consume_chunk_ctx = extract_chunk_arg,
- };
- return read_full_stream_with_sha1(lte, &cbs);
- } else {
- /* Don't do SHA1. */
- return read_stream_prefix(lte, size, extract_chunk,
- extract_chunk_arg);
- }
+ struct filedes *fd = _fd;
+ int ret = full_write(fd, chunk, size);
+ if (unlikely(ret))
+ ERROR_WITH_ERRNO("Error writing to file descriptor");
+ return ret;
}
-/* A consume_data_callback_t implementation that writes the chunk of data to a
- * file descriptor. */
+/* 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_chunk_to_fd(const void *chunk, size_t size, void *_fd_p)
+extract_blob_prefix_to_fd(struct blob_descriptor *blob, u64 size,
+ struct filedes *fd)
{
- struct filedes *fd = _fd_p;
-
- int ret = full_write(fd, chunk, size);
- if (ret) {
- ERROR_WITH_ERRNO("Error writing to file descriptor");
- return ret;
- }
- return 0;
+ struct read_blob_callbacks cbs = {
+ .consume_chunk = extract_chunk_to_fd,
+ .ctx = fd,
+ };
+ return read_blob_prefix(blob, size, &cbs);
}
-/* Extract the first @size bytes of the specified stream to the specified file
- * descriptor. */
+/* Extract the full uncompressed contents of the specified blob to the specified
+ * file descriptor. This checks the SHA-1 message digest. */
int
-extract_stream_to_fd(struct wim_lookup_table_entry *lte,
- struct filedes *fd, u64 size)
+extract_blob_to_fd(struct blob_descriptor *blob, struct filedes *fd)
{
- return extract_stream(lte, size, extract_chunk_to_fd, fd);
+ 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 and store it in @lte->hash. */
+/* Calculate the SHA-1 message digest of a blob and store it in @blob->hash. */
int
-sha1_stream(struct wim_lookup_table_entry *lte)
+sha1_blob(struct blob_descriptor *blob)
{
- wimlib_assert(lte->unhashed);
- struct read_stream_list_callbacks cbs = {
+ struct read_blob_callbacks cbs = {
};
- return read_full_stream_with_sha1(lte, &cbs);
+ return read_blob_with_sha1(blob, &cbs);
}
-/* Convert a short WIM resource header to a stand-alone WIM resource
- * specification. */
+/*
+ * 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_spec(const struct wim_reshdr *reshdr, WIMStruct *wim,
- struct wim_resource_spec *spec)
+wim_reshdr_to_desc(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ struct wim_resource_descriptor *rdesc)
{
- spec->wim = wim;
- spec->offset_in_wim = reshdr->offset_in_wim;
- spec->size_in_wim = reshdr->size_in_wim;
- spec->uncompressed_size = reshdr->uncompressed_size;
- INIT_LIST_HEAD(&spec->stream_list);
- spec->flags = reshdr->flags;
- spec->is_pipable = wim_is_pipable(wim);
+ 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;
+ }
}
-/* Convert a stand-alone resource specification to a WIM resource header. */
+/*
+ * 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_res_spec_to_hdr(const struct wim_resource_spec *rspec,
- struct wim_reshdr *reshdr)
+wim_reshdr_to_desc_and_blob(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ struct wim_resource_descriptor *rdesc,
+ struct blob_descriptor *blob)
{
- reshdr->offset_in_wim = rspec->offset_in_wim;
- reshdr->size_in_wim = rspec->size_in_wim;
- reshdr->flags = rspec->flags;
- reshdr->uncompressed_size = rspec->uncompressed_size;
+ wim_reshdr_to_desc(reshdr, wim, rdesc);
+ blob->size = rdesc->uncompressed_size;
+ blob_set_is_located_in_wim_resource(blob, rdesc, 0);
}
-/* Translates a WIM resource header from the on-disk format into an in-memory
- * format. */
+/* 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->flags = disk_reshdr->flags;
}
-/* Translates a WIM resource header from an in-memory format into the on-disk
- * format. */
+/* Export a WIM resource header to the on-disk format. */
void
put_wim_reshdr(const struct wim_reshdr *reshdr,
struct wim_reshdr_disk *disk_reshdr)
git://wimlib.net / wimlib / blobdiff