/*
* resource.c
*
- * Read uncompressed and compressed metadata and file resources.
+ * Code for reading streams and resources, including compressed WIM resources.
*/
/*
- * Copyright (C) 2012 Eric Biggers
+ * Copyright (C) 2012, 2013 Eric Biggers
*
* This file is part of wimlib, a library for working with WIM files.
*
* wimlib; if not, see http://www.gnu.org/licenses/.
*/
-#include "config.h"
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
-#include <stdlib.h>
-#include <stdarg.h>
+#include "wimlib.h"
+#include "wimlib/assert.h"
+#include "wimlib/endianness.h"
+#include "wimlib/error.h"
+#include "wimlib/file_io.h"
+#include "wimlib/lookup_table.h"
+#include "wimlib/resource.h"
+#include "wimlib/sha1.h"
+#include "wimlib/wim.h"
+
+#ifdef __WIN32__
+/* for read_win32_file_prefix(), read_win32_encrypted_file_prefix() */
+# include "wimlib/win32.h"
+#endif
#ifdef WITH_NTFS_3G
-#include <ntfs-3g/attrib.h>
-#include <ntfs-3g/inode.h>
-#include <ntfs-3g/dir.h>
+/* for read_ntfs_file_prefix() */
+# include "wimlib/ntfs_3g.h"
#endif
-#include "wimlib_internal.h"
-#include "lookup_table.h"
-#include "io.h"
-#include "lzx.h"
-#include "xpress.h"
-#include "sha1.h"
-#include "dentry.h"
-#include <unistd.h>
-#include <errno.h>
#ifdef HAVE_ALLOCA_H
-#include <alloca.h>
#endif
-
+#include <errno.h>
+#include <fcntl.h>
+#include <stdlib.h>
+#include <unistd.h>
/*
- * Reads all or part of a compressed resource into an in-memory buffer.
+ * Compressed WIM resources
+ *
+ * A compressed resource in a WIM consists of a number of compressed chunks,
+ * each of which decompresses to a fixed chunk size (given in the WIM header;
+ * usually 32768) except possibly the last, which always decompresses to any
+ * remaining bytes. In addition, immediately before the chunks, a table (the
+ * "chunk table") provides the offset, in bytes relative to the end of the chunk
+ * table, of the start of each compressed chunk, except for the first chunk
+ * which is omitted as it always has an offset of 0. Therefore, a compressed
+ * resource with N chunks will have a chunk table with N - 1 entries.
+ *
+ * Additional information:
*
- * @fp: The FILE* for the WIM file.
- * @resource_compressed_size: The compressed size of the resource.
- * @resource_uncompressed_size: The uncompressed size of the resource.
- * @resource_offset: The offset of the start of the resource from
- * the start of the stream @fp.
- * @resource_ctype: The compression type of the resource.
- * @len: The number of bytes of uncompressed data to read from
- * the resource.
- * @offset: The offset of the bytes to read within the uncompressed
- * resource.
- * @contents_len: An array into which the uncompressed data is written.
- * It must be at least @len bytes long.
+ * - 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.
*
- * Returns zero on success, nonzero on failure.
+ * - The chunk table is included in the compressed size of the resource provided
+ * in the corresponding entry in the WIM's stream lookup table.
+ *
+ * - The compressed size of a chunk is never greater than the uncompressed size.
+ * From the compressor's point of view, chunks that would have compressed to a
+ * size greater than or equal to their original size are in fact stored
+ * uncompressed. From the decompresser's point of view, chunks with
+ * compressed size equal to their uncompressed size are in fact uncompressed.
+ *
+ * Furthermore, wimlib supports its own "pipable" WIM format, and for this the
+ * structure of compressed resources was modified to allow piped reading and
+ * writing. To make sequential writing possible, the chunk table is placed
+ * after the chunks rather than before the chunks, and to make sequential
+ * reading possible, each chunk is prefixed with a 4-byte header giving its
+ * compressed size as a 32-bit, unsigned, little-endian integer. Otherwise the
+ * details are the same.
*/
-static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
- u64 resource_uncompressed_size,
- u64 resource_offset, int resource_ctype,
- u64 len, u64 offset, u8 contents_ret[])
-{
- DEBUG2("comp size = %"PRIu64", uncomp size = %"PRIu64", "
- "res offset = %"PRIu64"",
- resource_compressed_size,
- resource_uncompressed_size,
- resource_offset);
- DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"",
- wimlib_get_compression_type_string(resource_ctype), len, offset);
- /* Trivial case */
- if (len == 0)
- return 0;
- int (*decompress)(const void *, uint, void *, uint);
- /* Set the appropriate decompress function. */
- if (resource_ctype == WIM_COMPRESSION_TYPE_LZX)
- decompress = lzx_decompress;
- else
- decompress = xpress_decompress;
-
- /* The structure of a compressed resource consists of a table of chunk
- * offsets followed by the chunks themselves. Each chunk consists of
- * compressed data, and there is one chunk for each WIM_CHUNK_SIZE =
- * 32768 bytes of the uncompressed file, with the last chunk having any
- * remaining bytes.
- *
- * The chunk offsets are measured relative to the end of the chunk
- * table. The first chunk is omitted from the table in the WIM file
- * because its offset is implicitly given by the fact that it directly
- * follows the chunk table and therefore must have an offset of 0.
- */
+struct data_range {
+ u64 offset;
+ u64 size;
+};
- /* Calculate how many chunks the resource conists of in its entirety. */
- u64 num_chunks = (resource_uncompressed_size + WIM_CHUNK_SIZE - 1) /
- WIM_CHUNK_SIZE;
- /* As mentioned, the first chunk has no entry in the chunk table. */
- u64 num_chunk_entries = num_chunks - 1;
-
-
- /* The index of the chunk that the read starts at. */
- u64 start_chunk = offset / WIM_CHUNK_SIZE;
- /* The byte offset at which the read starts, within the start chunk. */
- u64 start_chunk_offset = offset % WIM_CHUNK_SIZE;
-
- /* The index of the chunk that contains the last byte of the read. */
- u64 end_chunk = (offset + len - 1) / WIM_CHUNK_SIZE;
- /* The byte offset of the last byte of the read, within the end chunk */
- u64 end_chunk_offset = (offset + len - 1) % WIM_CHUNK_SIZE;
-
- /* Number of chunks that are actually needed to read the requested part
- * of the file. */
- u64 num_needed_chunks = end_chunk - start_chunk + 1;
-
- /* If the end chunk is not the last chunk, an extra chunk entry is
- * needed because we need to know the offset of the chunk after the last
- * chunk read to figure out the size of the last read chunk. */
- if (end_chunk != num_chunks - 1)
- num_needed_chunks++;
-
- /* Declare the chunk table. It will only contain offsets for the chunks
- * that are actually needed for this read. */
- u64 chunk_offsets[num_needed_chunks];
-
- /* Set the implicit offset of the first chunk if it is included in the
- * needed chunks.
- *
- * Note: M$'s documentation includes a picture that shows the first
- * chunk starting right after the chunk entry table, labeled as offset
- * 0x10. However, in the actual file format, the offset is measured
- * from the end of the chunk entry table, so the first chunk has an
- * offset of 0. */
- if (start_chunk == 0)
- chunk_offsets[0] = 0;
-
- /* According to M$'s documentation, if the uncompressed size of
- * the file is greater than 4 GB, the chunk entries are 8-byte
- * integers. Otherwise, they are 4-byte integers. */
- u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
- 8 : 4;
-
- /* Size of the full chunk table in the WIM file. */
- u64 chunk_table_size = chunk_entry_size * num_chunk_entries;
-
- /* Read the needed chunk offsets from the table in the WIM file. */
-
- /* Index, in the WIM file, of the first needed entry in the
- * chunk table. */
- u64 start_table_idx = (start_chunk == 0) ? 0 : start_chunk - 1;
-
- /* Number of entries we need to actually read from the chunk
- * table (excludes the implicit first chunk). */
- u64 num_needed_chunk_entries = (start_chunk == 0) ?
- num_needed_chunks - 1 : num_needed_chunks;
-
- /* Skip over unneeded chunk table entries. */
- u64 file_offset_of_needed_chunk_entries = resource_offset +
- start_table_idx * chunk_entry_size;
- if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) {
- ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
- "chunk table of compressed resource",
- file_offset_of_needed_chunk_entries);
- return WIMLIB_ERR_READ;
+/*
+ * read_compressed_wim_resource() -
+ *
+ * Read data from a compressed WIM resource.
+ *
+ * @rspec
+ * Specification of the compressed WIM resource to read from.
+ * @ranges
+ * Nonoverlapping, nonempty ranges of the uncompressed resource data to
+ * read, sorted by increasing offset.
+ * @num_ranges
+ * Number of ranges in @ranges; must be at least 1.
+ * @cb
+ * Callback function to feed the data being read. Each call provides the
+ * next chunk of the requested data, uncompressed. Each chunk will be of
+ * nonzero size and will not cross range boundaries, but otherwise will be
+ * of unspecified size.
+ * @cb_ctx
+ * Parameter to pass to @cb_ctx.
+ *
+ * Possible return values:
+ *
+ * WIMLIB_ERR_SUCCESS (0)
+ * WIMLIB_ERR_READ (errno set)
+ * WIMLIB_ERR_UNEXPECTED_END_OF_FILE (errno set to 0)
+ * WIMLIB_ERR_NOMEM (errno set to ENOMEM)
+ * WIMLIB_ERR_DECOMPRESSION (errno set to EINVAL)
+ *
+ * or other error code returned by the @cb function.
+ */
+static int
+read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
+ const struct data_range * const ranges,
+ const size_t num_ranges,
+ const consume_data_callback_t cb,
+ void * const cb_ctx)
+{
+ int ret;
+ int errno_save;
+
+ u64 *chunk_offsets = NULL;
+ u8 *ubuf = NULL;
+ void *cbuf = NULL;
+ bool chunk_offsets_malloced = false;
+ bool ubuf_malloced = false;
+ bool cbuf_malloced = false;
+ struct wimlib_decompressor *decompressor = NULL;
+
+ /* Sanity checks */
+ wimlib_assert(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);
+ }
+ for (size_t i = 0; i < num_ranges - 1; i++)
+ wimlib_assert(ranges[i].offset + ranges[i].size <= ranges[i + 1].offset);
+
+ /* Get the offsets of the first and last bytes of the read. */
+ const u64 first_offset = ranges[0].offset;
+ const u64 last_offset = ranges[num_ranges - 1].offset + ranges[num_ranges - 1].size - 1;
+
+ /* Get the file descriptor for the WIM. */
+ struct filedes * const in_fd = &rspec->wim->in_fd;
+
+ /* Determine if we're reading a pipable resource from a pipe or not. */
+ const bool is_pipe_read = !filedes_is_seekable(in_fd);
+
+ /* Determine if the chunk table is in an altenate format. */
+ const bool alt_chunk_table = (rspec->flags & WIM_RESHDR_FLAG_PACKED_STREAMS)
+ && !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 = rspec->offset_in_wim;
+ int ctype = rspec->compression_type;
+ u32 chunk_size = rspec->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 @rspec by
+ * read_wim_lookup_table(). */
+ cur_read_offset += sizeof(struct alt_chunk_table_header_disk);
+ }
+
+ if (!is_power_of_2(chunk_size)) {
+ ERROR("Invalid compressed resource: "
+ "expected power-of-2 chunk size (got %"PRIu32")",
+ chunk_size);
+ ret = WIMLIB_ERR_INVALID_CHUNK_SIZE;
+ goto out_free_memory;
}
- /* Number of bytes we need to read from the chunk table. */
- size_t size = num_needed_chunk_entries * chunk_entry_size;
+ /* Get valid decompressor. */
+ if (ctype == rspec->wim->decompressor_ctype &&
+ chunk_size == rspec->wim->decompressor_max_block_size)
+ {
+ /* Cached decompressor. */
+ decompressor = rspec->wim->decompressor;
+ rspec->wim->decompressor_ctype = WIMLIB_COMPRESSION_TYPE_NONE;
+ rspec->wim->decompressor = NULL;
+ } else {
+ ret = wimlib_create_decompressor(ctype, chunk_size, NULL,
+ &decompressor);
+ if (ret)
+ goto out_free_memory;
+ }
- u8 chunk_tab_buf[size];
+ const u32 chunk_order = bsr32(chunk_size);
+
+ /* Calculate the total number of chunks the resource is divided into. */
+ const u64 num_chunks = (rspec->uncompressed_size + chunk_size - 1) >> chunk_order;
+
+ /* Calculate the 0-based indices of the first and last chunks containing
+ * data that needs to be passed to the callback. */
+ const u64 first_needed_chunk = first_offset >> chunk_order;
+ const u64 last_needed_chunk = last_offset >> chunk_order;
+
+ /* Calculate the 0-based index of the first chunk that actually needs to
+ * be read. This is normally first_needed_chunk, but for pipe reads we
+ * must always start from the 0th chunk. */
+ const u64 read_start_chunk = (is_pipe_read ? 0 : first_needed_chunk);
+
+ /* Calculate the number of chunk offsets that are needed for the chunks
+ * being read. */
+ const u64 num_needed_chunk_offsets =
+ last_needed_chunk - read_start_chunk + 1 +
+ (last_needed_chunk < num_chunks - 1);
+
+ /* Calculate the number of entries in the chunk table. Normally, it's
+ * one less than the number of chunks, since the first chunk has no
+ * entry. But in the alternate chunk table format, the chunk entries
+ * contain chunk sizes, not offsets, and there is one per chunk. */
+ const u64 num_chunk_entries = (alt_chunk_table ? num_chunks : num_chunks - 1);
+
+ /* Set the size of each chunk table entry based on the resource's
+ * uncompressed size. */
+ const u64 chunk_entry_size = get_chunk_entry_size(rspec->uncompressed_size,
+ alt_chunk_table);
+
+ /* Calculate the size of the chunk table in bytes. */
+ const u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
+
+ /* Calculate the size of the chunk table in bytes, including the header
+ * in the case of the alternate chunk table format. */
+ const u64 chunk_table_full_size =
+ (alt_chunk_table) ? chunk_table_size + sizeof(struct alt_chunk_table_header_disk)
+ : chunk_table_size;
+
+ if (!is_pipe_read) {
+ /* Read the needed chunk table entries into memory and use them
+ * to initialize the chunk_offsets array. */
+
+ u64 first_chunk_entry_to_read;
+ u64 last_chunk_entry_to_read;
+
+ if (alt_chunk_table) {
+ /* The alternate chunk table contains chunk sizes, not
+ * offsets, so we always must read all preceding entries
+ * in order to determine offsets. */
+ first_chunk_entry_to_read = 0;
+ last_chunk_entry_to_read = last_needed_chunk;
+ } else {
+ /* Here we must account for the fact that the first
+ * chunk has no explicit chunk table entry. */
- if (fread(chunk_tab_buf, 1, size, fp) != size)
- goto err;
+ if (read_start_chunk == 0)
+ first_chunk_entry_to_read = 0;
+ else
+ first_chunk_entry_to_read = read_start_chunk - 1;
- /* Now fill in chunk_offsets from the entries we have read in
- * chunk_tab_buf. */
+ if (last_needed_chunk == 0)
+ last_chunk_entry_to_read = 0;
+ else
+ last_chunk_entry_to_read = last_needed_chunk - 1;
- u64 *chunk_tab_p = chunk_offsets;
- if (start_chunk == 0)
- chunk_tab_p++;
+ if (last_needed_chunk < num_chunks - 1)
+ last_chunk_entry_to_read++;
+ }
- if (chunk_entry_size == 4) {
- u32 *entries = (u32*)chunk_tab_buf;
- while (num_needed_chunk_entries--)
- *chunk_tab_p++ = le32_to_cpu(*entries++);
- } else {
- u64 *entries = (u64*)chunk_tab_buf;
- while (num_needed_chunk_entries--)
- *chunk_tab_p++ = le64_to_cpu(*entries++);
- }
+ const u64 num_chunk_entries_to_read =
+ last_chunk_entry_to_read - first_chunk_entry_to_read + 1;
- /* Done with the chunk table now. We must now seek to the first chunk
- * that is needed for the read. */
+ const u64 chunk_offsets_alloc_size =
+ max(num_chunk_entries_to_read,
+ num_needed_chunk_offsets) * sizeof(chunk_offsets[0]);
- u64 file_offset_of_first_needed_chunk = resource_offset +
- chunk_table_size + chunk_offsets[0];
- if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) {
- ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
- "first chunk of compressed resource",
- file_offset_of_first_needed_chunk);
- return WIMLIB_ERR_READ;
- }
+ if ((size_t)chunk_offsets_alloc_size != chunk_offsets_alloc_size)
+ goto oom;
- /* Pointer to current position in the output buffer for uncompressed
- * data. */
- u8 *out_p = (u8*)contents_ret;
-
- /* Buffer for compressed data. While most compressed chunks will have a
- * size much less than WIM_CHUNK_SIZE, WIM_CHUNK_SIZE - 1 is the maximum
- * size in the worst-case. This assumption is valid only if chunks that
- * happen to compress to more than the uncompressed size (i.e. a
- * sequence of random bytes) are always stored uncompressed. But this seems
- * to be the case in M$'s WIM files, even though it is undocumented. */
- u8 compressed_buf[WIM_CHUNK_SIZE - 1];
-
-
- /* Decompress all the chunks. */
- for (u64 i = start_chunk; i <= end_chunk; i++) {
-
- DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64").",
- i, start_chunk, end_chunk);
-
- /* Calculate the sizes of the compressed chunk and of the
- * uncompressed chunk. */
- uint compressed_chunk_size, uncompressed_chunk_size;
- if (i != num_chunks - 1) {
- /* All the chunks except the last one in the resource
- * expand to WIM_CHUNK_SIZE uncompressed, and the amount
- * of compressed data for the chunk is given by the
- * difference of offsets in the chunk offset table. */
- compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
- chunk_offsets[i - start_chunk];
- uncompressed_chunk_size = WIM_CHUNK_SIZE;
+ if (chunk_offsets_alloc_size <= STACK_MAX) {
+ chunk_offsets = alloca(chunk_offsets_alloc_size);
} else {
- /* The last compressed chunk consists of the remaining
- * bytes in the file resource, and the last uncompressed
- * chunk has size equal to however many bytes are left-
- * that is, the remainder of the uncompressed size when
- * divided by WIM_CHUNK_SIZE.
- *
- * Note that the resource_compressed_size includes the
- * chunk table, so the size of it must be subtracted. */
- compressed_chunk_size = resource_compressed_size -
- chunk_table_size -
- chunk_offsets[i - start_chunk];
-
- uncompressed_chunk_size = resource_uncompressed_size %
- WIM_CHUNK_SIZE;
-
- /* If the remainder is 0, the last chunk actually
- * uncompresses to a full WIM_CHUNK_SIZE bytes. */
- if (uncompressed_chunk_size == 0)
- uncompressed_chunk_size = WIM_CHUNK_SIZE;
+ chunk_offsets = MALLOC(chunk_offsets_alloc_size);
+ if (chunk_offsets == NULL)
+ goto oom;
+ chunk_offsets_malloced = true;
}
- DEBUG2("compressed_chunk_size = %u, "
- "uncompressed_chunk_size = %u",
- compressed_chunk_size, uncompressed_chunk_size);
+ const size_t chunk_table_size_to_read =
+ num_chunk_entries_to_read * chunk_entry_size;
+ const u64 file_offset_of_needed_chunk_entries =
+ cur_read_offset
+ + (first_chunk_entry_to_read * chunk_entry_size)
+ + (rspec->is_pipable ? (rspec->size_in_wim - chunk_table_size) : 0);
- /* Figure out how much of this chunk we actually need to read */
- u64 start_offset;
- if (i == start_chunk)
- start_offset = start_chunk_offset;
- else
- start_offset = 0;
- u64 end_offset;
- if (i == end_chunk)
- end_offset = end_chunk_offset;
- else
- end_offset = WIM_CHUNK_SIZE - 1;
-
- u64 partial_chunk_size = end_offset + 1 - start_offset;
- bool is_partial_chunk = (partial_chunk_size !=
- uncompressed_chunk_size);
-
- DEBUG2("start_offset = %u, end_offset = %u", start_offset,
- end_offset);
- DEBUG2("partial_chunk_size = %u", partial_chunk_size);
-
- /* This is undocumented, but chunks can be uncompressed. This
- * appears to always be the case when the compressed chunk size
- * is equal to the uncompressed chunk size. */
- if (compressed_chunk_size == uncompressed_chunk_size) {
- /* Probably an uncompressed chunk */
-
- if (start_offset != 0) {
- if (fseeko(fp, start_offset, SEEK_CUR) != 0) {
- ERROR_WITH_ERRNO("Uncompressed partial "
- "chunk fseek() error");
- return WIMLIB_ERR_READ;
- }
+ void * const chunk_table_data =
+ (u8*)chunk_offsets +
+ chunk_offsets_alloc_size -
+ chunk_table_size_to_read;
+
+ ret = full_pread(in_fd, chunk_table_data, chunk_table_size_to_read,
+ file_offset_of_needed_chunk_entries);
+ if (ret)
+ goto read_error;
+
+ /* Now fill in chunk_offsets from the entries we have read in
+ * chunk_tab_data. We break aliasing rules here to avoid having
+ * to allocate yet another array. */
+ typedef le64 __attribute__((may_alias)) aliased_le64_t;
+ typedef le32 __attribute__((may_alias)) aliased_le32_t;
+ u64 * chunk_offsets_p = chunk_offsets;
+
+ if (alt_chunk_table) {
+ u64 cur_offset = 0;
+ aliased_le32_t *raw_entries = chunk_table_data;
+
+ for (size_t i = 0; i < num_chunk_entries_to_read; i++) {
+ u32 entry = le32_to_cpu(raw_entries[i]);
+ if (i >= read_start_chunk)
+ *chunk_offsets_p++ = cur_offset;
+ cur_offset += entry;
}
- if (fread(out_p, 1, partial_chunk_size, fp) !=
- partial_chunk_size)
- goto err;
+ if (last_needed_chunk < num_chunks - 1)
+ *chunk_offsets_p = cur_offset;
} else {
- /* Compressed chunk */
- int ret;
-
- /* Read the compressed data into compressed_buf. */
- if (fread(compressed_buf, 1, compressed_chunk_size,
- fp) != compressed_chunk_size)
- goto err;
-
- /* For partial chunks we must buffer the uncompressed
- * data because we don't need all of it. */
- if (is_partial_chunk) {
- u8 uncompressed_buf[uncompressed_chunk_size];
-
- ret = decompress(compressed_buf,
- compressed_chunk_size,
- uncompressed_buf,
- uncompressed_chunk_size);
- if (ret != 0)
- return WIMLIB_ERR_DECOMPRESSION;
- memcpy(out_p, uncompressed_buf + start_offset,
- partial_chunk_size);
+ 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 {
- ret = decompress(compressed_buf,
- compressed_chunk_size,
- out_p,
- uncompressed_chunk_size);
- if (ret != 0)
- return WIMLIB_ERR_DECOMPRESSION;
+ 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]);
}
}
- /* Advance the pointer into the uncompressed output data by the
- * number of uncompressed bytes that were written. */
- out_p += partial_chunk_size;
+ /* Set offset to beginning of first chunk to read. */
+ cur_read_offset += chunk_offsets[0];
+ if (rspec->is_pipable)
+ cur_read_offset += read_start_chunk * sizeof(struct pwm_chunk_hdr);
+ else
+ cur_read_offset += chunk_table_size;
}
- return 0;
-
-err:
- if (feof(fp))
- ERROR("Unexpected EOF in compressed file resource");
- else
- ERROR_WITH_ERRNO("Error reading compressed file resource");
- return WIMLIB_ERR_READ;
-}
+ /* Allocate buffer for holding the uncompressed data of each chunk. */
+ if (chunk_size <= STACK_MAX) {
+ ubuf = alloca(chunk_size);
+ } else {
+ ubuf = MALLOC(chunk_size);
+ if (ubuf == NULL)
+ goto oom;
+ ubuf_malloced = true;
+ }
-/*
- * Reads uncompressed data from an open file stream.
- */
-int read_uncompressed_resource(FILE *fp, u64 offset, u64 len,
- u8 contents_ret[])
-{
- if (fseeko(fp, offset, SEEK_SET) != 0) {
- ERROR("Failed to seek to byte %"PRIu64" of input file "
- "to read uncompressed resource (len = %"PRIu64")",
- offset, len);
- return WIMLIB_ERR_READ;
+ /* Allocate a temporary buffer for reading compressed chunks, each of
+ * which can be at most @chunk_size - 1 bytes. This excludes compressed
+ * chunks that are a full @chunk_size bytes, which are actually stored
+ * uncompressed. */
+ if (chunk_size - 1 <= STACK_MAX) {
+ cbuf = alloca(chunk_size - 1);
+ } else {
+ cbuf = MALLOC(chunk_size - 1);
+ if (cbuf == NULL)
+ goto oom;
+ cbuf_malloced = true;
}
- if (fread(contents_ret, 1, len, fp) != len) {
- if (feof(fp)) {
- ERROR("Unexpected EOF in uncompressed file resource");
+
+ /* Set current data range. */
+ const struct data_range *cur_range = ranges;
+ const struct data_range * const end_range = &ranges[num_ranges];
+ u64 cur_range_pos = cur_range->offset;
+ u64 cur_range_end = cur_range->offset + cur_range->size;
+
+ /* Read and process each needed chunk. */
+ for (u64 i = read_start_chunk; i <= last_needed_chunk; i++) {
+
+ /* Calculate uncompressed size of next chunk. */
+ u32 chunk_usize;
+ if ((i == num_chunks - 1) && (rspec->uncompressed_size & (chunk_size - 1)))
+ chunk_usize = (rspec->uncompressed_size & (chunk_size - 1));
+ else
+ chunk_usize = chunk_size;
+
+ /* Calculate compressed size of next chunk. */
+ u32 chunk_csize;
+ if (is_pipe_read) {
+ struct pwm_chunk_hdr chunk_hdr;
+
+ ret = full_pread(in_fd, &chunk_hdr,
+ sizeof(chunk_hdr), cur_read_offset);
+ if (ret)
+ goto read_error;
+ chunk_csize = le32_to_cpu(chunk_hdr.compressed_size);
} else {
- ERROR("Failed to read %"PRIu64" bytes from "
- "uncompressed resource at offset %"PRIu64,
- len, offset);
+ if (i == num_chunks - 1) {
+ chunk_csize = rspec->size_in_wim -
+ chunk_table_full_size -
+ chunk_offsets[i - read_start_chunk];
+ if (rspec->is_pipable)
+ chunk_csize -= num_chunks * sizeof(struct pwm_chunk_hdr);
+ } else {
+ chunk_csize = chunk_offsets[i + 1 - read_start_chunk] -
+ chunk_offsets[i - read_start_chunk];
+ }
+ }
+ if (chunk_csize == 0 || chunk_csize > chunk_usize) {
+ ERROR("Invalid chunk size in compressed resource!");
+ errno = EINVAL;
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ goto out_free_memory;
+ }
+ if (rspec->is_pipable)
+ cur_read_offset += sizeof(struct pwm_chunk_hdr);
+
+ /* Offsets in the uncompressed resource at which this chunk
+ * starts and ends. */
+ const u64 chunk_start_offset = i << chunk_order;
+ const u64 chunk_end_offset = chunk_start_offset + chunk_usize;
+
+ if (chunk_end_offset <= cur_range_pos) {
+
+ /* The next range does not require data in this chunk,
+ * so skip it. */
+ cur_read_offset += chunk_csize;
+ if (is_pipe_read) {
+ u8 dummy;
+
+ ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
+ if (ret)
+ goto read_error;
+ }
+ } else {
+
+ /* Read the chunk and feed data to the callback
+ * function. */
+ u8 *read_buf;
+
+ if (chunk_csize == chunk_usize)
+ read_buf = ubuf;
+ else
+ read_buf = cbuf;
+
+ ret = full_pread(in_fd,
+ read_buf,
+ chunk_csize,
+ cur_read_offset);
+ if (ret)
+ goto read_error;
+
+ if (read_buf == cbuf) {
+ DEBUG("Decompressing chunk %"PRIu64" "
+ "(csize=%"PRIu32" usize=%"PRIu32")",
+ i, chunk_csize, chunk_usize);
+ ret = wimlib_decompress(cbuf,
+ chunk_csize,
+ ubuf,
+ chunk_usize,
+ decompressor);
+ if (ret) {
+ ERROR("Failed to decompress data!");
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ errno = EINVAL;
+ goto out_free_memory;
+ }
+ }
+ cur_read_offset += chunk_csize;
+
+ /* At least one range requires data in this chunk. */
+ do {
+ size_t start, end, size;
+
+ /* Calculate how many bytes of data should be
+ * sent to the callback function, taking into
+ * account that data sent to the callback
+ * function must not overlap range boundaries.
+ */
+ start = cur_range_pos - chunk_start_offset;
+ end = min(cur_range_end, chunk_end_offset) - chunk_start_offset;
+ size = end - start;
+
+ ret = (*cb)(&ubuf[start], size, cb_ctx);
+
+ if (ret)
+ goto out_free_memory;
+
+ cur_range_pos += size;
+ if (cur_range_pos == cur_range_end) {
+ /* Advance to next range. */
+ if (++cur_range == end_range) {
+ cur_range_pos = ~0ULL;
+ } else {
+ cur_range_pos = cur_range->offset;
+ cur_range_end = cur_range->offset + cur_range->size;
+ }
+ }
+ } while (cur_range_pos < chunk_end_offset);
}
- return WIMLIB_ERR_READ;
}
- return 0;
-}
+ if (is_pipe_read &&
+ last_offset == rspec->uncompressed_size - 1 &&
+ chunk_table_size)
+ {
+ u8 dummy;
+ /* If reading a pipable resource from a pipe and the full data
+ * was requested, skip the chunk table at the end so that the
+ * file descriptor is fully clear of the resource after this
+ * returns. */
+ cur_read_offset += chunk_table_size;
+ ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
+ if (ret)
+ goto read_error;
+ }
+ ret = 0;
+
+out_free_memory:
+ errno_save = errno;
+ if (decompressor) {
+ wimlib_free_decompressor(rspec->wim->decompressor);
+ rspec->wim->decompressor = decompressor;
+ rspec->wim->decompressor_ctype = ctype;
+ rspec->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;
+ ret = WIMLIB_ERR_NOMEM;
+ goto out_free_memory;
+read_error:
+ ERROR_WITH_ERRNO("Error reading compressed WIM resource!");
+ goto out_free_memory;
+}
-/* Reads the contents of a struct resource_entry, as represented in the on-disk
- * format, from the memory pointed to by @p, and fills in the fields of @entry.
- * A pointer to the byte after the memory read at @p is returned. */
-const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry)
+/* Read raw data from a file descriptor at the specified offset, feeding the
+ * data it in chunks into the specified callback function. */
+static int
+read_raw_file_data(struct filedes *in_fd, u64 offset, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
- u64 size;
- u8 flags;
-
- p = get_u56(p, &size);
- p = get_u8(p, &flags);
- entry->size = size;
- entry->flags = flags;
-
- /* 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... */
- p = get_u64(p, &entry->offset);
- if (entry->offset & 0xc000000000000000ULL) {
- WARNING("Truncating offset in resource entry");
- entry->offset &= 0x3fffffffffffffffULL;
- }
- p = get_u64(p, &entry->original_size);
- if (entry->original_size & 0xc000000000000000ULL) {
- WARNING("Truncating original_size in resource entry");
- entry->original_size &= 0x3fffffffffffffffULL;
+ u8 buf[BUFFER_SIZE];
+ size_t bytes_to_read;
+ int ret;
+
+ while (size) {
+ bytes_to_read = min(sizeof(buf), size);
+ ret = full_pread(in_fd, buf, bytes_to_read, offset);
+ if (ret) {
+ ERROR_WITH_ERRNO("Read error");
+ return ret;
+ }
+ ret = cb(buf, bytes_to_read, cb_ctx);
+ if (ret)
+ return ret;
+ size -= bytes_to_read;
+ offset += bytes_to_read;
}
- return p;
+ return 0;
}
-/* Copies the struct resource_entry @entry to the memory pointed to by @p in the
- * on-disk format. A pointer to the byte after the memory written at @p is
- * returned. */
-u8 *put_resource_entry(u8 *p, const struct resource_entry *entry)
+/* A consume_data_callback_t implementation that simply concatenates all chunks
+ * into a buffer. */
+static int
+bufferer_cb(const void *chunk, size_t size, void *_ctx)
{
- p = put_u56(p, entry->size);
- p = put_u8(p, entry->flags);
- p = put_u64(p, entry->offset);
- p = put_u64(p, entry->original_size);
- return p;
+ u8 **buf_p = _ctx;
+
+ *buf_p = mempcpy(*buf_p, chunk, size);
+ return 0;
}
/*
- * Reads some data from the resource corresponding to a WIM lookup table entry.
+ * read_partial_wim_resource()-
+ *
+ * Read a range of data from an uncompressed or compressed resource in a WIM
+ * file.
+ *
+ * @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.
*
- * @lte: The WIM lookup table entry for the resource.
- * @buf: Buffer into which to write the data.
- * @size: Number of bytes to read.
- * @offset: Offset at which to start reading the resource.
- * @raw: If %true, compressed data is read literally rather than being
- * decompressed first.
+ * 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)
*
- * Returns zero on success, nonzero on failure.
+ * or other error code returned by the @cb function.
*/
-int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[],
- size_t size, u64 offset, bool raw)
+static int
+read_partial_wim_resource(const struct wim_resource_spec *rspec,
+ u64 offset, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
- int ctype;
- int ret = 0;
- FILE *fp;
-
- /* We shouldn't be allowing read over-runs in any part of the library.
- * */
- if (raw)
- wimlib_assert(offset + size <= lte->resource_entry.size);
- else
- wimlib_assert(offset + size <= lte->resource_entry.original_size);
-
- switch (lte->resource_location) {
- case RESOURCE_IN_WIM:
- /* The resource is in a WIM file, and its WIMStruct is given by
- * the lte->wim member. The resource may be either compressed
- * or uncompressed. */
- wimlib_assert(lte->wim != NULL);
- wimlib_assert(lte->wim->fp != NULL);
- ctype = wim_resource_compression_type(lte);
-
- wimlib_assert(ctype != WIM_COMPRESSION_TYPE_NONE ||
- (lte->resource_entry.original_size ==
- lte->resource_entry.size));
-
- if (raw || ctype == WIM_COMPRESSION_TYPE_NONE)
- ret = read_uncompressed_resource(lte->wim->fp,
- lte->resource_entry.offset + offset,
- size, buf);
- else
- ret = read_compressed_resource(lte->wim->fp,
- lte->resource_entry.size,
- lte->resource_entry.original_size,
- lte->resource_entry.offset,
- ctype, size, offset, buf);
- break;
- case RESOURCE_IN_STAGING_FILE:
- case RESOURCE_IN_FILE_ON_DISK:
- /* The resource is in some file on the external filesystem and
- * needs to be read uncompressed */
- wimlib_assert(lte->file_on_disk);
- wimlib_assert(<e->file_on_disk == <e->staging_file_name);
- /* Use existing file pointer if available; otherwise open one
- * temporarily */
- if (lte->file_on_disk_fp) {
- fp = lte->file_on_disk_fp;
- } else {
- fp = fopen(lte->file_on_disk, "rb");
- if (!fp) {
- ERROR_WITH_ERRNO("Failed to open the file "
- "`%s'", lte->file_on_disk);
- ret = WIMLIB_ERR_OPEN;
- break;
- }
- }
- ret = read_uncompressed_resource(fp, offset, size, buf);
- if (fp != lte->file_on_disk_fp)
- fclose(fp);
- break;
- case RESOURCE_IN_ATTACHED_BUFFER:
- /* The resource is directly attached uncompressed in an
- * in-memory buffer. */
- wimlib_assert(lte->attached_buffer != NULL);
- memcpy(buf, lte->attached_buffer + offset, size);
- break;
-#ifdef WITH_NTFS_3G
- case RESOURCE_IN_NTFS_VOLUME:
- wimlib_assert(lte->ntfs_loc != NULL);
- wimlib_assert(lte->attr != NULL);
- {
- if (lte->ntfs_loc->is_reparse_point)
- offset += 8;
- if (ntfs_attr_pread(lte->attr, offset, size, buf) != size) {
- ERROR_WITH_ERRNO("Error reading NTFS attribute "
- "at `%s'",
- lte->ntfs_loc->path_utf8);
- ret = WIMLIB_ERR_NTFS_3G;
- }
- break;
- }
-#endif
- default:
- wimlib_assert(0);
- ret = -1;
- break;
+ /* 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)) {
+ 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,
+ rspec->offset_in_wim + offset,
+ size,
+ cb,
+ cb_ctx);
}
- return ret;
}
-/*
- * Reads all the data from the resource corresponding to a WIM lookup table
- * entry.
- *
- * @lte: The WIM lookup table entry for the resource.
- * @buf: Buffer into which to write the data. It must be at least
- * wim_resource_size(lte) bytes long.
- *
- * Returns 0 on success; nonzero on failure.
- */
-int read_full_wim_resource(const struct lookup_table_entry *lte, u8 buf[])
+/* Read the specified range of uncompressed data from the specified stream,
+ * which must be located into 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)
+{
+ u8 *buf = _buf;
+
+ wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
+
+ return read_partial_wim_resource(lte->rspec,
+ lte->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 read_wim_resource(lte, buf, wim_resource_size(lte), 0, false);
+ return 0;
}
-/* Chunk table that's located at the beginning of each compressed resource in
- * the WIM. (This is not the on-disk format; the on-disk format just has an
- * array of offsets.) */
-struct chunk_table {
- off_t file_offset;
- u64 num_chunks;
- u64 original_resource_size;
- u64 bytes_per_chunk_entry;
- u64 table_disk_size;
- u64 cur_offset;
- u64 *cur_offset_p;
- u64 offsets[0];
-};
+/* Skip over the data of the specified stream, which must correspond to a full
+ * WIM resource. */
+int
+skip_wim_stream(struct wim_lookup_table_entry *lte)
+{
+ 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);
+}
-/*
- * Allocates and initializes a chunk table, and reserves space for it in the
- * output file.
- */
static int
-begin_wim_resource_chunk_tab(const struct lookup_table_entry *lte,
- FILE *out_fp,
- off_t file_offset,
- struct chunk_table **chunk_tab_ret)
+read_wim_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
+{
+ return read_partial_wim_resource(lte->rspec, lte->offset_in_res, size,
+ cb, cb_ctx);
+}
+
+#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 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)
{
- u64 size = wim_resource_size(lte);
- u64 num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
- size_t alloc_size = sizeof(struct chunk_table) + num_chunks * sizeof(u64);
- struct chunk_table *chunk_tab = CALLOC(1, alloc_size);
int ret;
+ int raw_fd;
+ struct filedes fd;
- if (!chunk_tab) {
- ERROR("Failed to allocate chunk table for %"PRIu64" byte "
- "resource", size);
- ret = WIMLIB_ERR_NOMEM;
- goto out;
- }
- chunk_tab->file_offset = file_offset;
- chunk_tab->num_chunks = num_chunks;
- chunk_tab->original_resource_size = size;
- chunk_tab->bytes_per_chunk_entry = (size >= (1ULL << 32)) ? 8 : 4;
- chunk_tab->table_disk_size = chunk_tab->bytes_per_chunk_entry *
- (num_chunks - 1);
- chunk_tab->cur_offset = 0;
- chunk_tab->cur_offset_p = chunk_tab->offsets;
-
- if (fwrite(chunk_tab, 1, chunk_tab->table_disk_size, out_fp) !=
- chunk_tab->table_disk_size) {
- ERROR_WITH_ERRNO("Failed to write chunk table in compressed "
- "file resource");
- ret = WIMLIB_ERR_WRITE;
- goto out;
- }
+ wimlib_assert(size <= lte->size);
- ret = 0;
-out:
- *chunk_tab_ret = chunk_tab;
+ DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", size, lte->file_on_disk);
+
+ raw_fd = open(lte->file_on_disk, O_BINARY | O_RDONLY);
+ if (raw_fd < 0) {
+ ERROR_WITH_ERRNO("Can't open \"%"TS"\"", lte->file_on_disk);
+ return WIMLIB_ERR_OPEN;
+ }
+ filedes_init(&fd, raw_fd);
+ ret = read_raw_file_data(&fd, 0, size, cb, cb_ctx);
+ filedes_close(&fd);
return ret;
}
+#endif /* !__WIN32__ */
-/*
- * Compresses a chunk of a WIM resource.
- *
- * @chunk: Uncompressed data of the chunk.
- * @chunk_size: Size of the uncompressed chunk in bytes.
- * @compressed_chunk: Pointer to output buffer of size at least
- * (@chunk_size - 1) bytes.
- * @compressed_chunk_len_ret: Pointer to an unsigned int into which the size
- * of the compressed chunk will be
- * returned.
- * @ctype: Type of compression to use. Must be WIM_COMPRESSION_TYPE_LZX
- * or WIM_COMPRESSION_TYPE_XPRESS.
- *
- * Returns zero if compressed succeeded, and nonzero if the chunk could not be
- * compressed to any smaller than @chunk_size. This function cannot fail for
- * any other reasons.
- */
-static int compress_chunk(const u8 chunk[], unsigned chunk_size,
- u8 compressed_chunk[],
- unsigned *compressed_chunk_len_ret,
- int ctype)
+/* This function handles the trivial case of reading stream data that is, in
+ * fact, already located in an in-memory buffer. */
+static int
+read_buffer_prefix(const struct wim_lookup_table_entry *lte,
+ u64 size, consume_data_callback_t cb, void *cb_ctx)
{
- int (*compress)(const void *, unsigned, void *, unsigned *);
- switch (ctype) {
- case WIM_COMPRESSION_TYPE_LZX:
- compress = lzx_compress;
- break;
- case WIM_COMPRESSION_TYPE_XPRESS:
- compress = xpress_compress;
- break;
- default:
- wimlib_assert(0);
- break;
- }
- return (*compress)(chunk, chunk_size, compressed_chunk,
- compressed_chunk_len_ret);
+ wimlib_assert(size <= lte->size);
+ return (*cb)(lte->attached_buffer, size, cb_ctx);
}
+typedef int (*read_stream_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
+ u64 size,
+ consume_data_callback_t cb,
+ void *cb_ctx);
+
/*
- * Writes a chunk of a WIM resource to an output file.
+ * read_stream_prefix()-
*
- * @chunk: Uncompressed data of the chunk.
- * @chunk_size: Size of the chunk (<= WIM_CHUNK_SIZE)
- * @out_fp: FILE * to write tho chunk to.
- * @out_ctype: Compression type to use when writing the chunk (ignored if no
- * chunk table provided)
- * @chunk_tab: Pointer to chunk table being created. It is updated with the
- * offset of the chunk we write.
+ * 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.
*
- * Returns 0 on success; nonzero on failure.
+ * This function feeds the data to a callback function @cb in chunks 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.
*/
-static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size,
- FILE *out_fp, int out_ctype,
- struct chunk_table *chunk_tab)
+static int
+read_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
- const u8 *out_chunk;
- unsigned out_chunk_size;
+ static const read_stream_prefix_handler_t handlers[] = {
+ [RESOURCE_IN_WIM] = read_wim_stream_prefix,
+ #ifdef __WIN32__
+ [RESOURCE_IN_FILE_ON_DISK] = read_win32_file_prefix,
+ #else
+ [RESOURCE_IN_FILE_ON_DISK] = read_file_on_disk_prefix,
+ #endif
+ [RESOURCE_IN_ATTACHED_BUFFER] = read_buffer_prefix,
+ #ifdef WITH_FUSE
+ [RESOURCE_IN_STAGING_FILE] = read_file_on_disk_prefix,
+ #endif
+ #ifdef WITH_NTFS_3G
+ [RESOURCE_IN_NTFS_VOLUME] = read_ntfs_file_prefix,
+ #endif
+ #ifdef __WIN32__
+ [RESOURCE_WIN32_ENCRYPTED] = read_win32_encrypted_file_prefix,
+ #endif
+ };
+ wimlib_assert(lte->resource_location < ARRAY_LEN(handlers)
+ && handlers[lte->resource_location] != NULL);
+ return handlers[lte->resource_location](lte, size, cb, cb_ctx);
+}
- wimlib_assert(chunk_size <= WIM_CHUNK_SIZE);
+/* Read the full uncompressed data of the specified stream into the specified
+ * buffer, which must have space for at least lte->size bytes. */
+int
+read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *_buf)
+{
+ u8 *buf = _buf;
+ return read_stream_prefix(lte, lte->size, bufferer_cb, &buf);
+}
- if (!chunk_tab) {
- out_chunk = chunk;
- out_chunk_size = chunk_size;
- } else {
- u8 *compressed_chunk = alloca(chunk_size);
- int ret;
+/* Retrieve the full uncompressed data of the specified stream. A buffer large
+ * enough hold the data is allocated and returned in @buf_ret. */
+int
+read_full_stream_into_alloc_buf(const struct wim_lookup_table_entry *lte,
+ void **buf_ret)
+{
+ int ret;
+ void *buf;
- ret = compress_chunk(chunk, chunk_size, compressed_chunk,
- &out_chunk_size, out_ctype);
- if (ret == 0) {
- out_chunk = compressed_chunk;
- } else {
- out_chunk = chunk;
- out_chunk_size = chunk_size;
- }
- *chunk_tab->cur_offset_p++ = chunk_tab->cur_offset;
- chunk_tab->cur_offset += out_chunk_size;
+ if ((size_t)lte->size != lte->size) {
+ ERROR("Can't read %"PRIu64" byte stream into "
+ "memory", lte->size);
+ return WIMLIB_ERR_NOMEM;
}
- if (fwrite(out_chunk, 1, out_chunk_size, out_fp) != out_chunk_size) {
- ERROR_WITH_ERRNO("Failed to write WIM resource chunk");
- return WIMLIB_ERR_WRITE;
+ buf = MALLOC(lte->size);
+ if (buf == NULL)
+ return WIMLIB_ERR_NOMEM;
+
+ ret = read_full_stream_into_buf(lte, buf);
+ if (ret) {
+ FREE(buf);
+ return ret;
}
+
+ *buf_ret = buf;
return 0;
}
-/*
- * Finishes a WIM chunk tale and writes it to the output file at the correct
- * offset.
- *
- * The final size of the full compressed resource is returned in the
- * @compressed_size_p.
- */
+/* 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
-finish_wim_resource_chunk_tab(struct chunk_table *chunk_tab,
- FILE *out_fp, u64 *compressed_size_p)
+wim_resource_spec_to_data(struct wim_resource_spec *rspec, void **buf_ret)
{
- size_t bytes_written;
- if (fseeko(out_fp, chunk_tab->file_offset, SEEK_SET) != 0) {
- ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" of output "
- "WIM file", chunk_tab->file_offset);
- return WIMLIB_ERR_WRITE;
- }
+ int ret;
+ struct wim_lookup_table_entry *lte;
- if (chunk_tab->bytes_per_chunk_entry == 8) {
- array_cpu_to_le64(chunk_tab->offsets, chunk_tab->num_chunks);
- } else {
- for (u64 i = 0; i < chunk_tab->num_chunks; i++)
- ((u32*)chunk_tab->offsets)[i] =
- cpu_to_le32(chunk_tab->offsets[i]);
- }
- bytes_written = fwrite((u8*)chunk_tab->offsets +
- chunk_tab->bytes_per_chunk_entry,
- 1, chunk_tab->table_disk_size, out_fp);
- if (bytes_written != chunk_tab->table_disk_size) {
- ERROR_WITH_ERRNO("Failed to write chunk table in compressed "
- "file resource");
- return WIMLIB_ERR_WRITE;
- }
- if (fseeko(out_fp, 0, SEEK_END) != 0) {
- ERROR_WITH_ERRNO("Failed to seek to end of output WIM file");
- return WIMLIB_ERR_WRITE;
- }
- *compressed_size_p = chunk_tab->cur_offset + chunk_tab->table_disk_size;
- return 0;
+ 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;
+
+ ret = read_full_stream_into_alloc_buf(lte, buf_ret);
+
+ lte_unbind_wim_resource_spec(lte);
+ free_lookup_table_entry(lte);
+ return ret;
}
-/*
- * Writes a WIM resource to a FILE * opened for writing. The resource may be
- * written uncompressed or compressed depending on the @out_ctype parameter.
- *
- * If by chance the resource compresses to more than the original size (this may
- * happen with random data or files than are pre-compressed), the resource is
- * instead written uncompressed (and this is reflected in the @out_res_entry by
- * removing the WIM_RESHDR_FLAG_COMPRESSED flag).
- *
- * @lte: The lookup table entry for the WIM resource.
- * @out_fp: The FILE * to write the resource to.
- * @out_ctype: The compression type of the resource to write. Note: if this is
- * the same as the compression type of the WIM resource we
- * need to read, we simply copy the data (i.e. we do not
- * uncompress it, then compress it again).
- * @out_res_entry: If non-NULL, a resource entry that is filled in with the
- * offset, original size, compressed size, and compression flag
- * of the output resource.
- *
- * Returns 0 on success; nonzero on failure.
- */
-static int write_wim_resource(struct lookup_table_entry *lte,
- FILE *out_fp, int out_ctype,
- struct resource_entry *out_res_entry)
+/* 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. */
+int
+wim_reshdr_to_data(const struct wim_reshdr *reshdr, WIMStruct *wim, void **buf_ret)
{
- u64 bytes_remaining;
- u64 original_size;
- u64 old_compressed_size;
- u64 new_compressed_size;
- u64 offset;
+ 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);
+}
+
+int
+wim_reshdr_to_hash(const struct wim_reshdr *reshdr, WIMStruct *wim,
+ u8 hash[SHA1_HASH_SIZE])
+{
+ struct wim_resource_spec rspec;
int ret;
- struct chunk_table *chunk_tab = NULL;
- bool raw;
- off_t file_offset;
-#ifdef WITH_NTFS_3G
- ntfs_inode *ni = NULL;
-#endif
+ struct wim_lookup_table_entry *lte;
- wimlib_assert(lte);
+ wim_res_hdr_to_spec(reshdr, wim, &rspec);
- /* Original size of the resource */
- original_size = wim_resource_size(lte);
+ lte = new_lookup_table_entry();
+ if (lte == NULL)
+ return WIMLIB_ERR_NOMEM;
- /* Compressed size of the resource (as it exists now) */
- old_compressed_size = wim_resource_compressed_size(lte);
+ lte_bind_wim_resource_spec(lte, &rspec);
+ lte->flags = rspec.flags;
+ lte->size = rspec.uncompressed_size;
+ lte->offset_in_res = 0;
+ lte->unhashed = 1;
- /* Current offset in output file */
- file_offset = ftello(out_fp);
- if (file_offset == -1) {
- ERROR_WITH_ERRNO("Failed to get offset in output "
- "stream");
- return WIMLIB_ERR_WRITE;
- }
+ ret = sha1_stream(lte);
- /* Are the compression types the same? If so, do a raw copy (copy
- * without decompressing and recompressing the data). */
- raw = (wim_resource_compression_type(lte) == out_ctype
- && out_ctype != WIM_COMPRESSION_TYPE_NONE);
- if (raw)
- bytes_remaining = old_compressed_size;
- else
- bytes_remaining = original_size;
+ lte_unbind_wim_resource_spec(lte);
+ copy_hash(hash, lte->hash);
+ free_lookup_table_entry(lte);
+ return ret;
+}
- /* Empty resource; nothing needs to be done, so just return success. */
- if (bytes_remaining == 0)
- return 0;
+struct streamifier_context {
+ struct read_stream_list_callbacks cbs;
+ struct wim_lookup_table_entry *cur_stream;
+ struct wim_lookup_table_entry *next_stream;
+ u64 cur_stream_offset;
+ struct wim_lookup_table_entry *final_stream;
+ size_t list_head_offset;
+};
- /* Buffer for reading chunks for the resource */
- u8 buf[min(WIM_CHUNK_SIZE, bytes_remaining)];
+static struct wim_lookup_table_entry *
+next_stream(struct wim_lookup_table_entry *lte, size_t list_head_offset)
+{
+ struct list_head *cur;
- /* If we are writing a compressed resource and not doing a raw copy, we
- * need to initialize the chunk table */
- if (out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw) {
- ret = begin_wim_resource_chunk_tab(lte, out_fp, file_offset,
- &chunk_tab);
- if (ret != 0)
- goto out;
- }
+ cur = (struct list_head*)((u8*)lte + list_head_offset);
- /* If the WIM resource is in an external file, open a FILE * to it so we
- * don't have to open a temporary one in read_wim_resource() for each
- * chunk. */
- if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK
- && !lte->file_on_disk_fp)
- {
- wimlib_assert(lte->file_on_disk);
- lte->file_on_disk_fp = fopen(lte->file_on_disk, "rb");
- if (!lte->file_on_disk_fp) {
- ERROR_WITH_ERRNO("Failed to open the file `%s' for "
- "reading", lte->file_on_disk);
- ret = WIMLIB_ERR_OPEN;
- goto out;
- }
- }
-#ifdef WITH_NTFS_3G
- else if (lte->resource_location == RESOURCE_IN_NTFS_VOLUME
- && !lte->attr)
- {
- struct ntfs_location *loc = lte->ntfs_loc;
- wimlib_assert(loc);
- ni = ntfs_pathname_to_inode(*loc->ntfs_vol_p, NULL, loc->path_utf8);
- if (!ni) {
- ERROR_WITH_ERRNO("Failed to open inode `%s' in NTFS "
- "volume", loc->path_utf8);
- ret = WIMLIB_ERR_NTFS_3G;
- goto out;
- }
- lte->attr = ntfs_attr_open(ni,
- loc->is_reparse_point ? AT_REPARSE_POINT : AT_DATA,
- (ntfschar*)loc->stream_name_utf16,
- loc->stream_name_utf16_num_chars);
- if (!lte->attr) {
- ERROR_WITH_ERRNO("Failed to open attribute of `%s' in "
- "NTFS volume", loc->path_utf8);
- ret = WIMLIB_ERR_NTFS_3G;
- goto out_fclose;
- }
+ return (struct wim_lookup_table_entry*)((u8*)cur->next - 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 int
+streamifier_cb(const void *chunk, size_t size, void *_ctx)
+{
+ struct streamifier_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);
+
+ 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 (ret)
+ return ret;
}
-#endif
- /* If we aren't doing a raw copy, we will compute the SHA1 message
- * digest of the resource as we read it, and verify it's the same as the
- * hash given in the lookup table entry once we've finished reading the
- * resource. */
- SHA_CTX ctx;
- if (!raw)
- sha1_init(&ctx);
-
- /* While there are still bytes remaining in the WIM resource, read a
- * chunk of the resource, update SHA1, then write that chunk using the
- * desired compression type. */
- offset = 0;
- do {
- u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE);
- ret = read_wim_resource(lte, buf, to_read, offset, raw);
- if (ret != 0)
- goto out_fclose;
- if (!raw)
- sha1_update(&ctx, buf, to_read);
- ret = write_wim_resource_chunk(buf, to_read, out_fp,
- out_ctype, chunk_tab);
- if (ret != 0)
- goto out_fclose;
- bytes_remaining -= to_read;
- offset += to_read;
- } while (bytes_remaining);
-
- /* Raw copy: The new compressed size is the same as the old compressed
- * size
- *
- * Using WIM_COMPRESSION_TYPE_NONE: The new compressed size is the
- * original size
- *
- * Using a different compression type: Call
- * finish_wim_resource_chunk_tab() and it will provide the new
- * compressed size.
- */
- if (raw) {
- new_compressed_size = old_compressed_size;
- } else {
- if (out_ctype == WIM_COMPRESSION_TYPE_NONE)
- new_compressed_size = original_size;
- else {
- ret = finish_wim_resource_chunk_tab(chunk_tab, out_fp,
- &new_compressed_size);
- if (ret != 0)
- goto out_fclose;
+ /* Consume the chunk. */
+ ret = (*ctx->cbs.consume_chunk)(chunk, size,
+ ctx->cbs.consume_chunk_ctx);
+ 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. */
+
+ ctx->cur_stream_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);
+ if (ret)
+ return ret;
+
+ /* Advance to next stream. */
+ ctx->cur_stream = ctx->next_stream;
+ if (ctx->cur_stream != NULL) {
+ if (ctx->cur_stream != ctx->final_stream)
+ ctx->next_stream = next_stream(ctx->cur_stream,
+ ctx->list_head_offset);
+ else
+ ctx->next_stream = NULL;
}
}
+ return 0;
+}
- /* Verify SHA1 message digest of the resource, unless we are doing a raw
- * write (in which case we never even saw the uncompressed data). Or,
- * if the hash we had before is all 0's, just re-set it to be the new
- * hash. */
- if (!raw) {
- u8 md[SHA1_HASH_SIZE];
- sha1_final(md, &ctx);
- if (is_zero_hash(lte->hash)) {
- copy_hash(lte->hash, md);
- } else if (!hashes_equal(md, lte->hash)) {
- ERROR("WIM resource has incorrect hash!");
- if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK) {
- ERROR("We were reading it from `%s'; maybe it changed "
- "while we were reading it.",
- lte->file_on_disk);
- }
- ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
- goto out_fclose;
- }
+struct hasher_context {
+ SHA_CTX sha_ctx;
+ int flags;
+ struct read_stream_list_callbacks cbs;
+};
+
+/* Callback for starting to read a stream while calculating its SHA1 message
+ * digest. */
+static int
+hasher_begin_stream(struct wim_lookup_table_entry *lte, bool is_partial_res,
+ 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);
+}
+
+/* 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. */
+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);
+}
+
+static void
+get_sha1_string(const u8 md[SHA1_HASH_SIZE], tchar *str)
+{
+ for (size_t i = 0; i < SHA1_HASH_SIZE; i++)
+ str += tsprintf(str, T("%02x"), md[i]);
+}
+
+/* Callback for finishing reading a stream while calculating its SHA1 message
+ * digest. */
+static int
+hasher_end_stream(struct wim_lookup_table_entry *lte, 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. */
+ ret = status;
+ goto out_next_cb;
}
- if (!raw && new_compressed_size >= original_size &&
- out_ctype != WIM_COMPRESSION_TYPE_NONE)
- {
- /* Oops! We compressed the resource to larger than the original
- * size. Write the resource uncompressed instead. */
- if (fseeko(out_fp, file_offset, SEEK_SET) != 0) {
- ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" "
- "of output WIM file", file_offset);
- ret = WIMLIB_ERR_WRITE;
- goto out_fclose;
- }
- ret = write_wim_resource(lte, out_fp, WIM_COMPRESSION_TYPE_NONE,
- out_res_entry);
- if (ret != 0)
- goto out_fclose;
- if (fflush(out_fp) != 0) {
- ERROR_WITH_ERRNO("Failed to flush output WIM file");
- ret = WIMLIB_ERR_WRITE;
- goto out_fclose;
- }
- if (ftruncate(fileno(out_fp), file_offset + out_res_entry->size) != 0) {
- ERROR_WITH_ERRNO("Failed to truncate output WIM file");
- ret = WIMLIB_ERR_WRITE;
- goto out_fclose;
+ /* Retrieve the final SHA1 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 (out_res_entry) {
- out_res_entry->size = new_compressed_size;
- out_res_entry->original_size = original_size;
- out_res_entry->offset = file_offset;
- out_res_entry->flags = lte->resource_entry.flags
- & ~WIM_RESHDR_FLAG_COMPRESSED;
- if (out_ctype != WIM_COMPRESSION_TYPE_NONE)
- out_res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
+ 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) {
+ tchar expected_hashstr[SHA1_HASH_SIZE * 2 + 1];
+ tchar actual_hashstr[SHA1_HASH_SIZE * 2 + 1];
+ get_sha1_string(lte->hash, expected_hashstr);
+ get_sha1_string(hash, actual_hashstr);
+ ERROR("The stream is corrupted!\n"
+ " (Expected SHA1=%"TS",\n"
+ " got SHA1=%"TS")",
+ expected_hashstr, actual_hashstr);
+ }
+ ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
+ errno = EINVAL;
+ goto out_next_cb;
+ }
+ DEBUG("SHA1 message digest okay for "
+ "stream (size=%"PRIu64").", lte->size);
}
}
ret = 0;
-out_fclose:
- if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK
- && lte->file_on_disk_fp) {
- fclose(lte->file_on_disk_fp);
- lte->file_on_disk_fp = NULL;
- }
-#ifdef WITH_NTFS_3G
- else if (lte->resource_location == RESOURCE_IN_NTFS_VOLUME) {
- if (lte->attr) {
- ntfs_attr_close(lte->attr);
- lte->attr = NULL;
- }
- if (ni)
- ntfs_inode_close(ni);
- }
-#endif
-out:
- FREE(chunk_tab);
- return ret;
+out_next_cb:
+ if (ctx->cbs.end_stream == NULL)
+ return ret;
+ else
+ return (*ctx->cbs.end_stream)(lte, ret, ctx->cbs.end_stream_ctx);
}
-/* Like write_wim_resource(), but the resource is specified by a buffer of
- * uncompressed data rather a lookup table entry; also writes the SHA1 hash of
- * the buffer to @hash. */
-static int write_wim_resource_from_buffer(const u8 *buf, u64 buf_size,
- FILE *out_fp, int out_ctype,
- struct resource_entry *out_res_entry,
- u8 hash[SHA1_HASH_SIZE])
+static int
+read_full_stream_with_cbs(struct wim_lookup_table_entry *lte,
+ const struct read_stream_list_callbacks *cbs)
{
- /* Set up a temporary lookup table entry to provide to
- * write_wim_resource(). */
- struct lookup_table_entry lte;
int ret;
- lte.resource_entry.flags = 0;
- lte.resource_entry.original_size = buf_size;
- lte.resource_entry.size = buf_size;
- lte.resource_entry.offset = 0;
- lte.resource_location = RESOURCE_IN_ATTACHED_BUFFER;
- lte.attached_buffer = (u8*)buf;
-
- zero_out_hash(lte.hash);
- ret = write_wim_resource(<e, out_fp, out_ctype, out_res_entry);
- if (ret != 0)
+
+ ret = (*cbs->begin_stream)(lte, false, cbs->begin_stream_ctx);
+ if (ret)
return ret;
- copy_hash(hash, lte.hash);
- return 0;
-}
-/*
- * Extracts the first @size bytes of the WIM resource specified by @lte to the
- * open file descriptor @fd.
- *
- * Returns 0 on success; nonzero on failure.
- */
-int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd,
- u64 size)
-{
- u64 bytes_remaining = size;
- u8 buf[min(WIM_CHUNK_SIZE, bytes_remaining)];
- u64 offset = 0;
- int ret = 0;
- u8 hash[SHA1_HASH_SIZE];
+ ret = read_stream_prefix(lte, lte->size, cbs->consume_chunk,
+ cbs->consume_chunk_ctx);
- SHA_CTX ctx;
- sha1_init(&ctx);
-
- while (bytes_remaining) {
- u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE);
- ret = read_wim_resource(lte, buf, to_read, offset, false);
- if (ret != 0)
- break;
- sha1_update(&ctx, buf, to_read);
- if (full_write(fd, buf, to_read) < to_read) {
- ERROR_WITH_ERRNO("Error extracting WIM resource");
- return WIMLIB_ERR_WRITE;
- }
- bytes_remaining -= to_read;
- offset += to_read;
- }
- sha1_final(hash, &ctx);
- if (!hashes_equal(hash, lte->hash)) {
- ERROR("Invalid checksum on a WIM resource "
- "(detected when extracting to external file)");
- ERROR("The following WIM resource is invalid:");
- print_lookup_table_entry(lte);
- return WIMLIB_ERR_INVALID_RESOURCE_HASH;
- }
- return 0;
+ return (*cbs->end_stream)(lte, ret, cbs->end_stream_ctx);
}
-/*
- * Extracts the WIM resource specified by @lte to the open file descriptor @fd.
- *
- * Returns 0 on success; nonzero on failure.
- */
-int extract_full_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd)
+/* 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)
{
- return extract_wim_resource_to_fd(lte, fd, wim_resource_size(lte));
+ 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,
+
+ };
+ return read_full_stream_with_cbs(lte, &hasher_cbs);
}
-/*
- * Copies the file resource specified by the lookup table entry @lte from the
- * input WIM to the output WIM that has its FILE * given by
- * ((WIMStruct*)wim)->out_fp.
- *
- * The output_resource_entry, out_refcnt, and part_number fields of @lte are
- * updated.
- *
- * Metadata resources are not copied (they are handled elsewhere for joining and
- * splitting).
- */
-int copy_resource(struct lookup_table_entry *lte, void *wim)
+static int
+read_packed_streams(struct wim_lookup_table_entry *first_stream,
+ struct wim_lookup_table_entry *last_stream,
+ u64 stream_count,
+ size_t list_head_offset,
+ const struct read_stream_list_callbacks *sink_cbs)
{
- WIMStruct *w = wim;
+ struct data_range *ranges;
+ bool ranges_malloced;
+ struct wim_lookup_table_entry *cur_stream;
+ size_t i;
int ret;
+ u64 ranges_alloc_size;
- if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
- !w->write_metadata)
- return 0;
+ DEBUG("Reading %"PRIu64" streams combined in same WIM resource",
+ stream_count);
- ret = write_wim_resource(lte, w->out_fp,
- wim_resource_compression_type(lte),
- <e->output_resource_entry);
- if (ret != 0)
- return ret;
- lte->out_refcnt = lte->refcnt;
- lte->part_number = w->hdr.part_number;
- return 0;
-}
+ /* Setup data ranges array (one range per stream to read); this way
+ * read_compressed_wim_resource() does not need to be aware of streams.
+ */
-/*
- * Writes a dentry's resources, including the main file resource as well as all
- * alternate data streams, to the output file.
- *
- * @dentry: The dentry for the file.
- * @wim_p: A pointer to the WIMStruct containing @dentry.
- *
- * @return zero on success, nonzero on failure.
- */
-int write_dentry_resources(struct dentry *dentry, void *wim_p)
-{
- WIMStruct *w = wim_p;
- int ret = 0;
- struct lookup_table_entry *lte;
- int ctype = wimlib_get_compression_type(w);
-
- if (w->write_flags & WIMLIB_WRITE_FLAG_VERBOSE) {
- wimlib_assert(dentry->full_path_utf8);
- printf("Writing streams for `%s'\n", dentry->full_path_utf8);
+ ranges_alloc_size = stream_count * sizeof(ranges[0]);
+
+ if (unlikely((size_t)ranges_alloc_size != ranges_alloc_size)) {
+ ERROR("Too many streams in one resource!");
+ return WIMLIB_ERR_NOMEM;
+ }
+ if (likely(ranges_alloc_size <= STACK_MAX)) {
+ ranges = alloca(ranges_alloc_size);
+ ranges_malloced = false;
+ } else {
+ ranges = MALLOC(ranges_alloc_size);
+ if (ranges == NULL) {
+ ERROR("Too many streams in one resource!");
+ return WIMLIB_ERR_NOMEM;
+ }
+ ranges_malloced = true;
+ }
+
+ for (i = 0, cur_stream = first_stream;
+ i < stream_count;
+ i++, cur_stream = next_stream(cur_stream, list_head_offset))
+ {
+ ranges[i].offset = cur_stream->offset_in_res;
+ ranges[i].size = cur_stream->size;
}
- for (unsigned i = 0; i <= dentry->d_inode->num_ads; i++) {
- lte = inode_stream_lte(dentry->d_inode, i, w->lookup_table);
- if (lte && ++lte->out_refcnt == 1) {
- ret = write_wim_resource(lte, w->out_fp, ctype,
- <e->output_resource_entry);
- if (ret != 0)
- break;
+ struct streamifier_context streamifier_ctx = {
+ .cbs = *sink_cbs,
+ .cur_stream = first_stream,
+ .next_stream = next_stream(first_stream, list_head_offset),
+ .cur_stream_offset = 0,
+ .final_stream = last_stream,
+ .list_head_offset = list_head_offset,
+ };
+
+ ret = read_compressed_wim_resource(first_stream->rspec,
+ ranges,
+ stream_count,
+ streamifier_cb,
+ &streamifier_ctx);
+
+ if (ranges_malloced)
+ FREE(ranges);
+
+ 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);
}
}
return ret;
}
/*
- * Reads the metadata metadata resource from the WIM file. The metadata
- * resource consists of the security data, followed by the directory entry for
- * the root directory, followed by all the other directory entries in the
- * filesystem. The subdir_offset field of each directory entry gives the start
- * of its child entries from the beginning of the metadata resource. An
- * end-of-directory is signaled by a directory entry of length '0', really of
- * length 8, because that's how long the 'length' field is.
+ * Read a list of 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.
+ *
+ * @stream_list
+ * List of streams (represented as `struct wim_lookup_table_entry's) to
+ * read.
+ * @list_head_offset
+ * Offset of the `struct list_head' within each `struct
+ * wim_lookup_table_entry' that makes up the @stream_list.
+ * @cbs
+ * Callback functions to accept the stream data.
+ * @flags
+ * Bitwise OR of zero or more of the following flags:
+ *
+ * VERIFY_STREAM_HASHES:
+ * For all streams being read that have already had SHA1 message
+ * digests computed, calculate the SHA1 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_STREAM_HASHES
+ * For all streams being read that have not yet had their SHA1
+ * message digests computed, calculate and save their SHA1 message
+ * digests.
+ *
+ * STREAM_LIST_ALREADY_SORTED
+ * @stream_list is already sorted in sequential order for reading.
*
- * @fp: The FILE* for the input WIM file.
- * @wim_ctype: The compression type of the WIM file.
- * @imd: Pointer to the image metadata structure. Its `metadata_lte'
- * member specifies the lookup table entry for the metadata
- * resource. The rest of the image metadata entry will be filled
- * in by this function.
+ * The callback functions are allowed to delete the current stream from the list
+ * if necessary.
*
- * @return: Zero on success, nonzero on failure.
+ * Returns 0 on success; a nonzero error code on failure. Failure can occur due
+ * to an error reading the data or due to an error status being returned by any
+ * of the callback functions.
*/
-int read_metadata_resource(WIMStruct *w, struct image_metadata *imd)
+int
+read_stream_list(struct list_head *stream_list,
+ size_t list_head_offset,
+ const struct read_stream_list_callbacks *cbs,
+ int flags)
{
- u8 *buf;
- u32 dentry_offset;
int ret;
- struct dentry *dentry;
- struct inode_table inode_tab;
- const struct lookup_table_entry *metadata_lte;
- u64 metadata_len;
- u64 metadata_offset;
- struct hlist_head inode_list;
-
- metadata_lte = imd->metadata_lte;
- metadata_len = wim_resource_size(metadata_lte);
- metadata_offset = metadata_lte->resource_entry.offset;
-
- DEBUG("Reading metadata resource: length = %"PRIu64", "
- "offset = %"PRIu64"", metadata_len, metadata_offset);
-
- /* There is no way the metadata resource could possibly be less than (8
- * + WIM_DENTRY_DISK_SIZE) bytes, where the 8 is for security data (with
- * no security descriptors) and WIM_DENTRY_DISK_SIZE is for the root
- * dentry. */
- if (metadata_len < 8 + WIM_DENTRY_DISK_SIZE) {
- ERROR("Expected at least %u bytes for the metadata resource",
- 8 + WIM_DENTRY_DISK_SIZE);
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
+ struct list_head *cur, *next;
+ struct wim_lookup_table_entry *lte;
+ struct hasher_context *hasher_ctx;
+ struct read_stream_list_callbacks *sink_cbs;
+
+ if (!(flags & STREAM_LIST_ALREADY_SORTED)) {
+ ret = sort_stream_list_by_sequential_order(stream_list, list_head_offset);
+ if (ret)
+ return ret;
}
- if (sizeof(size_t) < 8 && metadata_len > 0xffffffff) {
- ERROR("Metadata resource is too large (%"PRIu64" bytes",
- metadata_len);
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
+ if (flags & (VERIFY_STREAM_HASHES | COMPUTE_MISSING_STREAM_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,
+ };
+ } else {
+ sink_cbs = (struct read_stream_list_callbacks*)cbs;
}
- /* Allocate memory for the uncompressed metadata resource. */
- buf = MALLOC(metadata_len);
+ for (cur = stream_list->next, next = cur->next;
+ cur != stream_list;
+ cur = next, next = cur->next)
+ {
+ lte = (struct wim_lookup_table_entry*)((u8*)cur - list_head_offset);
- if (!buf) {
- ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
- "metadata resource", metadata_len);
- return WIMLIB_ERR_NOMEM;
- }
+ if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
+ lte->size != lte->rspec->uncompressed_size)
+ {
- /* Read the metadata resource into memory. (It may be compressed.) */
- ret = read_full_wim_resource(metadata_lte, buf);
- if (ret != 0)
- goto out_free_buf;
-
- DEBUG("Finished reading metadata resource into memory.");
-
- /* The root directory entry starts after security data, aligned on an
- * 8-byte boundary within the metadata resource.
- *
- * The security data starts with a 4-byte integer giving its total
- * length, so if we round that up to an 8-byte boundary that gives us
- * the offset of the root dentry.
- *
- * Here we read the security data into a wim_security_data structure,
- * and if successful, go ahead and calculate the offset in the metadata
- * resource of the root dentry. */
-
- wimlib_assert(imd->security_data == NULL);
- ret = read_security_data(buf, metadata_len, &imd->security_data);
- if (ret != 0)
- goto out_free_buf;
-
- dentry_offset = (imd->security_data->total_length + 7) & ~7;
-
- if (dentry_offset == 0) {
- ERROR("Integer overflow while reading metadata resource");
- ret = WIMLIB_ERR_INVALID_SECURITY_DATA;
- goto out_free_security_data;
+ struct wim_lookup_table_entry *lte_next, *lte_last;
+ struct list_head *next2;
+ u64 stream_count;
+
+ /* The next stream is a proper sub-sequence of a WIM
+ * resource. See if there are other streams in the same
+ * resource that need to be read. Since
+ * sort_stream_list_by_sequential_order() sorted the
+ * streams by offset in the WIM, this can be determined
+ * by simply scanning forward in the list. */
+
+ lte_last = lte;
+ stream_count = 1;
+ for (next2 = next;
+ next2 != stream_list
+ && (lte_next = (struct wim_lookup_table_entry*)
+ ((u8*)next2 - list_head_offset),
+ lte_next->resource_location == RESOURCE_IN_WIM
+ && lte_next->rspec == lte->rspec);
+ next2 = next2->next)
+ {
+ lte_last = lte_next;
+ stream_count++;
+ }
+ if (stream_count > 1) {
+ /* Reading multiple streams combined into a
+ * single WIM resource. They are in the stream
+ * list, sorted by offset; @lte specifies the
+ * first stream in the resource that needs to be
+ * read and @lte_last specifies the last stream
+ * in the resource that needs to be read. */
+ next = next2;
+ ret = read_packed_streams(lte, lte_last,
+ stream_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)
+ return ret;
}
+ return 0;
+}
- /* Allocate memory for the root dentry and read it into memory */
- dentry = MALLOC(sizeof(struct dentry));
- if (!dentry) {
- ERROR("Failed to allocate %zu bytes for root dentry",
- sizeof(struct dentry));
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_security_data;
+/* 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)
+{
+ wimlib_assert(size <= lte->size);
+ 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);
}
+}
- ret = read_dentry(buf, metadata_len, dentry_offset, dentry);
-
- /* This is the root dentry, so set its pointers correctly. */
- dentry->parent = dentry;
- dentry->next = dentry;
- dentry->prev = dentry;
- if (ret != 0)
- goto out_free_dentry_tree;
- inode_add_dentry(dentry, dentry->d_inode);
-
- /* Now read the entire directory entry tree into memory. */
- DEBUG("Reading dentry tree");
- ret = read_dentry_tree(buf, metadata_len, dentry);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- /* Calculate the full paths in the dentry tree. */
- DEBUG("Calculating dentry full paths");
- ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- /* Build hash table that maps hard link group IDs to dentry sets */
- DEBUG("Building link group table");
- ret = init_inode_table(&inode_tab, 9001);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- for_dentry_in_tree(dentry, inode_table_insert, &inode_tab);
-
- DEBUG("Fixing inconsistencies in the hard link groups");
- ret = fix_inodes(&inode_tab, &inode_list);
- destroy_inode_table(&inode_tab);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- DEBUG("Running miscellaneous verifications on the dentry tree");
- for_lookup_table_entry(w->lookup_table, lte_zero_real_refcnt, NULL);
- ret = for_dentry_in_tree(dentry, verify_dentry, w);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- DEBUG("Done reading image metadata");
-
- imd->root_dentry = dentry;
- imd->inode_list = inode_list;
- goto out_free_buf;
-out_free_dentry_tree:
- free_dentry_tree(dentry, NULL);
-out_free_security_data:
- free_security_data(imd->security_data);
- imd->security_data = NULL;
-out_free_buf:
- FREE(buf);
- return ret;
+/* A consume_data_callback_t implementation that writes the chunk of data to a
+ * file descriptor. */
+int
+extract_chunk_to_fd(const void *chunk, size_t size, void *_fd_p)
+{
+ 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;
}
-/* Write the metadata resource for the current WIM image. */
-int write_metadata_resource(WIMStruct *w)
+/* Extract the first @size bytes of the specified stream to the specified file
+ * descriptor. */
+int
+extract_stream_to_fd(struct wim_lookup_table_entry *lte,
+ struct filedes *fd, u64 size)
{
- u8 *buf;
- u8 *p;
- int ret;
- u64 subdir_offset;
- struct dentry *root;
- struct lookup_table_entry *lte;
- u64 metadata_original_size;
- const struct wim_security_data *sd;
-
- DEBUG("Writing metadata resource for image %d", w->current_image);
-
- root = wim_root_dentry(w);
- sd = wim_security_data(w);
-
- /* We do not allow the security data pointer to be NULL, although it may
- * point to an empty security data with no entries. */
- wimlib_assert(root != NULL);
- wimlib_assert(sd != NULL);
-
- /* Offset of first child of the root dentry. It's equal to:
- * - The total length of the security data, rounded to the next 8-byte
- * boundary,
- * - plus the total length of the root dentry,
- * - plus 8 bytes for an end-of-directory entry following the root
- * dentry (shouldn't really be needed, but just in case...)
- */
- subdir_offset = ((sd->total_length + 7) & ~7) +
- dentry_correct_total_length(root) + 8;
+ return extract_stream(lte, size, extract_chunk_to_fd, fd);
+}
- /* Calculate the subdirectory offsets for the entire dentry tree. */
- calculate_subdir_offsets(root, &subdir_offset);
+/* Extract the full uncompressed contents of the specified stream to the
+ * specified file descriptor. */
+int
+extract_full_stream_to_fd(struct wim_lookup_table_entry *lte,
+ struct filedes *fd)
+{
+ return extract_stream_to_fd(lte, fd, lte->size);
+}
- /* Total length of the metadata resource (uncompressed) */
- metadata_original_size = subdir_offset;
+/* Calculate the SHA1 message digest of a stream and store it in @lte->hash. */
+int
+sha1_stream(struct wim_lookup_table_entry *lte)
+{
+ wimlib_assert(lte->unhashed);
+ struct read_stream_list_callbacks cbs = {
+ };
+ return read_full_stream_with_sha1(lte, &cbs);
+}
- /* Allocate a buffer to contain the uncompressed metadata resource */
- buf = MALLOC(metadata_original_size);
- if (!buf) {
- ERROR("Failed to allocate %"PRIu64" bytes for "
- "metadata resource", metadata_original_size);
- return WIMLIB_ERR_NOMEM;
+/* Convert a short WIM resource header to a stand-alone WIM resource
+ * specification.
+ *
+ * Note: for packed 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 *rspec)
+{
+ rspec->wim = wim;
+ rspec->offset_in_wim = reshdr->offset_in_wim;
+ rspec->size_in_wim = reshdr->size_in_wim;
+ rspec->uncompressed_size = reshdr->uncompressed_size;
+ INIT_LIST_HEAD(&rspec->stream_list);
+ rspec->flags = reshdr->flags;
+ rspec->is_pipable = wim_is_pipable(wim);
+ if (rspec->flags & WIM_RESHDR_FLAG_COMPRESSED) {
+ rspec->compression_type = wim->compression_type;
+ rspec->chunk_size = wim->chunk_size;
+ } else {
+ rspec->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
+ rspec->chunk_size = 0;
}
+}
- /* Write the security data into the resource buffer */
- p = write_security_data(sd, buf);
-
- /* Write the dentry tree into the resource buffer */
- p = write_dentry_tree(root, p);
-
- /* We MUST have exactly filled the buffer; otherwise we calculated its
- * size incorrectly or wrote the data incorrectly. */
- wimlib_assert(p - buf == metadata_original_size);
-
- /* Get the lookup table entry for the metadata resource so we can update
- * it. */
- lte = wim_metadata_lookup_table_entry(w);
-
- wimlib_assert(lte != NULL);
-
- /* Write the metadata resource to the output WIM using the proper
- * compression type. The lookup table entry for the metadata resource
- * is updated. */
- ret = write_wim_resource_from_buffer(buf, metadata_original_size,
- w->out_fp,
- wimlib_get_compression_type(w),
- <e->output_resource_entry,
- lte->hash);
- if (ret != 0)
- goto out;
-
- /* It's very likely the SHA1 message digest of the metadata resource
- * changed, so re-insert the lookup table entry into the lookup table.
- *
- * We do not check for other lookup table entries having the same SHA1
- * message digest. It's possible for 2 absolutely identical images to
- * be added, therefore causing 2 identical metadata resources to be in
- * the WIM. However, in this case, it's expected for 2 separate lookup
- * table entries to be created, even though this doesn't make a whole
- * lot of sense since they will share the same SHA1 message digest.
- * */
- lookup_table_unlink(w->lookup_table, lte);
- lookup_table_insert(w->lookup_table, lte);
-
- wimlib_assert(lte->out_refcnt == 0);
- lte->out_refcnt = 1;
-
- /* Make sure that the resource entry is written marked with the metadata
- * flag. */
- lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;
-out:
- /* All the data has been written to the new WIM; no need for the buffer
- * anymore */
- FREE(buf);
- return ret;
+/* Convert a stand-alone resource specification to a WIM resource header. */
+void
+wim_res_spec_to_hdr(const struct wim_resource_spec *rspec,
+ struct wim_reshdr *reshdr)
+{
+ reshdr->offset_in_wim = rspec->offset_in_wim;
+ reshdr->size_in_wim = rspec->size_in_wim;
+ reshdr->flags = rspec->flags;
+ reshdr->uncompressed_size = rspec->uncompressed_size;
+}
+
+/* Translates a WIM resource header from the on-disk format into an in-memory
+ * format. */
+void
+get_wim_reshdr(const struct wim_reshdr_disk *disk_reshdr,
+ struct wim_reshdr *reshdr)
+{
+ reshdr->offset_in_wim = le64_to_cpu(disk_reshdr->offset_in_wim);
+ reshdr->size_in_wim = (((u64)disk_reshdr->size_in_wim[0] << 0) |
+ ((u64)disk_reshdr->size_in_wim[1] << 8) |
+ ((u64)disk_reshdr->size_in_wim[2] << 16) |
+ ((u64)disk_reshdr->size_in_wim[3] << 24) |
+ ((u64)disk_reshdr->size_in_wim[4] << 32) |
+ ((u64)disk_reshdr->size_in_wim[5] << 40) |
+ ((u64)disk_reshdr->size_in_wim[6] << 48));
+ reshdr->uncompressed_size = le64_to_cpu(disk_reshdr->uncompressed_size);
+ reshdr->flags = disk_reshdr->flags;
+}
+
+/* Translates a WIM resource header from an in-memory format into the on-disk
+ * format. */
+void
+put_wim_reshdr(const struct wim_reshdr *reshdr,
+ struct wim_reshdr_disk *disk_reshdr)
+{
+ disk_reshdr->size_in_wim[0] = reshdr->size_in_wim >> 0;
+ disk_reshdr->size_in_wim[1] = reshdr->size_in_wim >> 8;
+ disk_reshdr->size_in_wim[2] = reshdr->size_in_wim >> 16;
+ disk_reshdr->size_in_wim[3] = reshdr->size_in_wim >> 24;
+ disk_reshdr->size_in_wim[4] = reshdr->size_in_wim >> 32;
+ disk_reshdr->size_in_wim[5] = reshdr->size_in_wim >> 40;
+ disk_reshdr->size_in_wim[6] = reshdr->size_in_wim >> 48;
+ disk_reshdr->flags = reshdr->flags;
+ disk_reshdr->offset_in_wim = cpu_to_le64(reshdr->offset_in_wim);
+ disk_reshdr->uncompressed_size = cpu_to_le64(reshdr->uncompressed_size);
}
git://wimlib.net / wimlib / blobdiff