/*
* resource.c
*
- * Read uncompressed and compressed metadata and file resources.
- *
- * Copyright (C) 2010 Carl Thijssen
- * Copyright (C) 2012 Eric Biggers
+ * Read uncompressed and compressed metadata and file resources from a WIM file.
+ */
+
+/*
+ * Copyright (C) 2012, 2013 Eric Biggers
*
- * wimlib - Library for working with WIM files
+ * This file is part of wimlib, a library for working with WIM files.
*
- * This library is free software; you can redistribute it and/or modify it under
- * the terms of the GNU Lesser General Public License as published by the Free
- * Software Foundation; either version 2.1 of the License, or (at your option) any
- * later version.
+ * wimlib is free software; you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation; either version 3 of the License, or (at your option) any later
+ * version.
*
- * This library is distributed in the hope that it will be useful, but WITHOUT ANY
- * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
- * PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
+ * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
+ * A PARTICULAR PURPOSE. See the GNU General Public License for more details.
*
- * You should have received a copy of the GNU Lesser General Public License along
- * with this library; if not, write to the Free Software Foundation, Inc., 59
- * Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ * You should have received a copy of the GNU General Public License along with
+ * wimlib; if not, see http://www.gnu.org/licenses/.
*/
#include "wimlib_internal.h"
-#include "io.h"
-#include "lzx.h"
-#include "xpress.h"
#include "dentry.h"
-#include <unistd.h>
+#include "lookup_table.h"
+#include "buffer_io.h"
+#include "sha1.h"
+
+#ifdef __WIN32__
+# include "win32.h"
+#endif
+
#include <errno.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#ifdef HAVE_ALLOCA_H
+# include <alloca.h>
+#endif
+
+/* Write @n bytes from @buf to the file descriptor @fd, retrying on internupt
+ * and on short writes.
+ *
+ * Returns short count and set errno on failure. */
+static ssize_t
+full_write(int fd, const void *buf, size_t n)
+{
+ const void *p = buf;
+ ssize_t ret;
+ ssize_t total = 0;
+
+ while (total != n) {
+ ret = write(fd, p, n);
+ if (ret < 0) {
+ if (errno == EINTR)
+ continue;
+ else
+ break;
+ }
+ total += ret;
+ p += ret;
+ }
+ return total;
+}
-/*
- * Reads all or part of a compressed resource into an in-memory buffer.
+/* Read @n bytes from the file descriptor @fd to the buffer @buf, retrying on
+ * internupt and on short reads.
*
- * @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.
+ * Returns short count and set errno on failure. */
+static size_t
+full_read(int fd, void *buf, size_t n)
+{
+ size_t bytes_remaining = n;
+ while (bytes_remaining) {
+ ssize_t bytes_read = read(fd, buf, bytes_remaining);
+ if (bytes_read < 0) {
+ if (errno == EINTR)
+ continue;
+ break;
+ }
+ bytes_remaining -= bytes_read;
+ buf += bytes_read;
+ }
+ return n - bytes_remaining;
+}
+
+/*
+ * Reads all or part of a compressed WIM resource.
*
* Returns zero on success, nonzero on failure.
*/
-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[])
+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,
+ consume_data_callback_t cb,
+ void *ctx_or_buf)
{
+ int ret;
- DEBUG2("comp size = %"PRIu64", "
- "uncomp size = %"PRIu64", "
- "res offset = %"PRIu64"\n",
- resource_compressed_size,
- resource_uncompressed_size,
- resource_offset);
- DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"\n",
- wimlib_get_compression_type_string(resource_ctype),
- len, offset);
/* Trivial case */
if (len == 0)
return 0;
- int (*decompress)(const void *, uint, void *, uint);
+ int (*decompress)(const void *, unsigned, void *, unsigned);
/* Set the appropriate decompress function. */
- if (resource_ctype == WIM_COMPRESSION_TYPE_LZX)
- decompress = lzx_decompress;
+ if (resource_ctype == WIMLIB_COMPRESSION_TYPE_LZX)
+ decompress = wimlib_lzx_decompress;
else
- decompress = xpress_decompress;
+ decompress = wimlib_xpress_decompress;
/* The structure of a compressed resource consists of a table of chunk
* offsets followed by the chunks themselves. Each chunk consists of
* 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.
+ * follows the chunk table and therefore must have an offset of 0.
*/
- /* Calculate how many chunks the resource conists of in its entirety. */
+ /* Calculate how many chunks the resource consists 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. */
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];
+ /* Allocate the chunk table. It will only contain offsets for the
+ * chunks that are actually needed for this read. */
+ u64 *chunk_offsets;
+ bool chunk_offsets_malloced;
+ if (num_needed_chunks < 1000) {
+ chunk_offsets = alloca(num_needed_chunks * sizeof(u64));
+ chunk_offsets_malloced = false;
+ } else {
+ chunk_offsets = malloc(num_needed_chunks * sizeof(u64));
+ if (!chunk_offsets) {
+ ERROR("Failed to allocate chunk table "
+ "with %"PRIu64" entries", num_needed_chunks);
+ return WIMLIB_ERR_NOMEM;
+ }
+ chunk_offsets_malloced = true;
+ }
/* Set the implicit offset of the first chunk if it is included in the
* needed chunks.
/* 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) ?
+ u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
8 : 4;
/* Size of the full chunk table in the WIM file. */
/* 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) ?
+ 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 +
+ u64 file_offset_of_needed_chunk_entries = resource_offset +
start_table_idx * chunk_entry_size;
- if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) {
- ERROR("Failed to seek to byte %"PRIu64" "
- "to read chunk table of compressed "
- "resource: %m\n",
- file_offset_of_needed_chunk_entries);
- return WIMLIB_ERR_READ;
- }
+ if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET))
+ goto read_error;
/* Number of bytes we need to read from the chunk table. */
size_t size = num_needed_chunk_entries * chunk_entry_size;
- u8 chunk_tab_buf[size];
+ /* Read the raw data into the end of the chunk_offsets array to
+ * avoid allocating another array. */
+ void *chunk_tab_buf = (void*)&chunk_offsets[num_needed_chunks] - size;
if (fread(chunk_tab_buf, 1, size, fp) != size)
- goto err;
+ goto read_error;
/* Now fill in chunk_offsets from the entries we have read in
* chunk_tab_buf. */
if (chunk_entry_size == 4) {
u32 *entries = (u32*)chunk_tab_buf;
while (num_needed_chunk_entries--)
- *chunk_tab_p++ = to_le32(*entries++);
+ *chunk_tab_p++ = le32_to_cpu(*entries++);
} else {
u64 *entries = (u64*)chunk_tab_buf;
while (num_needed_chunk_entries--)
- *chunk_tab_p++ = to_le64(*entries++);
+ *chunk_tab_p++ = le64_to_cpu(*entries++);
}
/* Done with the chunk table now. We must now seek to the first chunk
* that is needed for the read. */
- u64 file_offset_of_first_needed_chunk = resource_offset +
+ u64 file_offset_of_first_needed_chunk = resource_offset +
chunk_table_size + chunk_offsets[0];
- if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) {
- ERROR("Failed to seek to byte %"PRIu64" "
- "to read first chunk of compressed "
- "resource: %m\n",
- file_offset_of_first_needed_chunk);
- return WIMLIB_ERR_READ;
- }
+ if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET))
+ goto read_error;
/* Pointer to current position in the output buffer for uncompressed
- * data. */
- u8 *out_p = (u8*)contents_ret;
+ * data. Alternatively, if using a callback function, we repeatedly
+ * fill a temporary buffer to feed data into the callback function. */
+ u8 *out_p;
+ if (cb)
+ out_p = alloca(WIM_CHUNK_SIZE);
+ else
+ out_p = ctx_or_buf;
/* 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
* 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];
-
+ void *compressed_buf = alloca(WIM_CHUNK_SIZE - 1);
/* Decompress all the chunks. */
for (u64 i = start_chunk; i <= end_chunk; i++) {
- DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64")\n",
- i, start_chunk, end_chunk);
-
/* Calculate the sizes of the compressed chunk and of the
* uncompressed chunk. */
- uint compressed_chunk_size, uncompressed_chunk_size;
+ unsigned compressed_chunk_size;
+ unsigned 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] -
+ compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
chunk_offsets[i - start_chunk];
uncompressed_chunk_size = WIM_CHUNK_SIZE;
} else {
* 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.
+ * 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 -
+ compressed_chunk_size = resource_compressed_size -
chunk_table_size -
chunk_offsets[i - start_chunk];
- uncompressed_chunk_size = resource_uncompressed_size %
+ uncompressed_chunk_size = resource_uncompressed_size %
WIM_CHUNK_SIZE;
/* If the remainder is 0, the last chunk actually
uncompressed_chunk_size = WIM_CHUNK_SIZE;
}
- DEBUG2("compressed_chunk_size = %u, uncompressed_chunk_size = %u\n",
- compressed_chunk_size, uncompressed_chunk_size);
-
-
/* Figure out how much of this chunk we actually need to read */
u64 start_offset;
if (i == start_chunk)
else
end_offset = WIM_CHUNK_SIZE - 1;
- u64 partial_chunk_size = end_offset + 1 - start_offset;
- bool is_partial_chunk = (partial_chunk_size !=
- uncompressed_chunk_size);
-
- DEBUG2("start_offset = %u, end_offset = %u\n", start_offset,
- end_offset);
- DEBUG2("partial_chunk_size = %u\n", partial_chunk_size);
+ unsigned partial_chunk_size = end_offset + 1 - start_offset;
+ bool is_partial_chunk = (partial_chunk_size != uncompressed_chunk_size);
/* This is undocumented, but chunks can be uncompressed. This
* appears to always be the case when the compressed chunk size
* is equal to the uncompressed chunk size. */
if (compressed_chunk_size == uncompressed_chunk_size) {
- /* Probably an uncompressed chunk */
-
- if (start_offset != 0) {
- if (fseeko(fp, start_offset, SEEK_CUR) != 0) {
- ERROR("Uncompressed partial chunk "
- "fseek() error: %m\n");
- return WIMLIB_ERR_READ;
- }
- }
- if (fread(out_p, 1, partial_chunk_size, fp) !=
- partial_chunk_size)
- goto err;
+ /* Uncompressed chunk */
+ if (start_offset != 0)
+ if (fseeko(fp, start_offset, SEEK_CUR))
+ goto read_error;
+ if (fread(cb ? out_p + start_offset : out_p,
+ 1, partial_chunk_size, fp) != partial_chunk_size)
+ goto read_error;
} else {
/* Compressed chunk */
- int ret;
/* Read the compressed data into compressed_buf. */
- if (fread(compressed_buf, 1, compressed_chunk_size,
+ if (fread(compressed_buf, 1, compressed_chunk_size,
fp) != compressed_chunk_size)
- goto err;
+ goto read_error;
- /* For partial chunks we must buffer the uncompressed
- * data because we don't need all of it. */
- if (is_partial_chunk) {
+ /* For partial chunks and when writing directly to a
+ * buffer, we must buffer the uncompressed data because
+ * we don't need all of it. */
+ if (is_partial_chunk && !cb) {
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;
+ compressed_chunk_size,
+ uncompressed_buf,
+ uncompressed_chunk_size);
+ if (ret) {
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ goto out;
+ }
memcpy(out_p, uncompressed_buf + start_offset,
- partial_chunk_size);
+ partial_chunk_size);
} else {
- DEBUG2("out_p = %p\n");
ret = decompress(compressed_buf,
- compressed_chunk_size,
- out_p,
- uncompressed_chunk_size);
- if (ret != 0)
- return WIMLIB_ERR_DECOMPRESSION;
+ compressed_chunk_size,
+ out_p,
+ uncompressed_chunk_size);
+ if (ret) {
+ ret = WIMLIB_ERR_DECOMPRESSION;
+ goto out;
+ }
}
}
-
- /* Advance the pointer into the uncompressed output data by the
- * number of uncompressed bytes that were written. */
- out_p += partial_chunk_size;
+ if (cb) {
+ /* Feed the data to the callback function */
+ ret = cb(out_p + start_offset,
+ partial_chunk_size, ctx_or_buf);
+ if (ret)
+ goto out;
+ } else {
+ /* No callback function provided; we are writing
+ * directly to a buffer. Advance the pointer into this
+ * buffer by the number of uncompressed bytes that were
+ * written. */
+ out_p += partial_chunk_size;
+ }
}
- return 0;
+ ret = 0;
+out:
+ if (chunk_offsets_malloced)
+ FREE(chunk_offsets);
+ return ret;
-err:
+read_error:
if (feof(fp))
- ERROR("Unexpected EOF in compressed file resource\n");
+ ERROR("Unexpected EOF in compressed file resource");
else
- ERROR("Error reading compressed file resource: %m\n");
- return WIMLIB_ERR_READ;
+ ERROR_WITH_ERRNO("Error reading compressed file resource");
+ ret = WIMLIB_ERR_READ;
+ goto out;
}
-/*
+/*
* Reads uncompressed data from an open file stream.
*/
-int read_uncompressed_resource(FILE *fp, u64 offset, u64 len,
- u8 contents_ret[])
+int
+read_uncompressed_resource(FILE *fp, u64 offset, u64 len, void *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")!\n", offset, len);
+ "to read uncompressed resource (len = %"PRIu64")",
+ offset, len);
return WIMLIB_ERR_READ;
}
if (fread(contents_ret, 1, len, fp) != len) {
if (feof(fp)) {
- ERROR("Unexpected EOF in uncompressed file resource!\n");
+ ERROR("Unexpected EOF in uncompressed file resource");
} else {
ERROR("Failed to read %"PRIu64" bytes from "
- "uncompressed resource at offset "
- "%"PRIu64"\n", len, offset);
+ "uncompressed resource at offset %"PRIu64,
+ len, offset);
}
return WIMLIB_ERR_READ;
}
return 0;
}
-/*
- * Reads a WIM resource.
- *
- * @fp: The FILE* for the WIM file.
- * @resource_size: The compressed size of the resource.
- * @resource_original_size: The uncompressed size of the resource.
- * @resource_offset: The offset of the resource in the stream @fp.
- * @resource_ctype: The compression type of the resource.
- * (WIM_COMPRESSION_TYPE_*)
- * @len: How many bytes of the resource should be read.
- * @offset: The offset within the resource at which the read
- * will occur.
- *
- * To read the whole file resource, specify offset =
- * 0 and len = resource_original_size, or call
- * read_full_resource().
- *
- * @contents_ret: An array, that must have length at least @len,
- * into which the uncompressed contents of
- * the file resource starting at @offset and
- * continuing for @len bytes will be written.
- *
- * @return: Zero on success, nonzero on failure. Failure may be due to
- * being unable to read the data from the WIM file at the
- * specified length and offset, or it may be due to the
- * compressed data (if the data is compressed) being
- * invalid.
- */
-int read_resource(FILE *fp, u64 resource_size, u64 resource_original_size,
- u64 resource_offset, int resource_ctype, u64 len,
- u64 offset, void *contents_ret)
-{
- if (resource_ctype == WIM_COMPRESSION_TYPE_NONE) {
- if (resource_size != resource_original_size) {
- ERROR("Resource with original size %"PRIu64" "
- "bytes is marked as uncompressed, \n",
- resource_original_size);
- ERROR(" but its actual size is %"PRIu64" "
- "bytes!\n",
- resource_size);
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
- }
- return read_uncompressed_resource(fp,
- resource_offset + offset,
- len, contents_ret);
- } else {
- return read_compressed_resource(fp, resource_size,
- resource_original_size, resource_offset,
- resource_ctype, len, offset, contents_ret);
- }
-}
-
-
-/*
- * Extracts the first @size bytes file resource specified by @entry to the open
- * file @fd. Returns nonzero on error.
- *
- * XXX
- * This function is somewhat redundant with uncompress_resource(). The
- * main difference is that this function writes to a file descriptor using
- * low-level calls to write() rather than to a FILE* with fwrite(); also this
- * function allows only up to @size bytes to be extracted.
- */
-int extract_resource_to_fd(WIMStruct *w, const struct resource_entry *entry,
- int fd, u64 size)
-{
- u64 num_chunks;
- u64 n;
- u8 buf[min(size, WIM_CHUNK_SIZE)];
- int res_ctype;
- u64 offset;
- u64 i;
- int ret;
-
- errno = 0;
-
- num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
- n = WIM_CHUNK_SIZE;
- res_ctype = wim_resource_compression_type(w, entry);
- offset = 0;
- for (i = 0; i < num_chunks; i++) {
- if (i == num_chunks - 1) {
- n = size % WIM_CHUNK_SIZE;
- if (n == 0) {
- n = WIM_CHUNK_SIZE;
- }
- }
-
- ret = read_resource(w->fp, entry->size, entry->original_size,
- entry->offset, res_ctype, n, offset, buf);
- if (ret != 0)
- return ret;
-
- if (full_write(fd, buf, n) != n)
- return WIMLIB_ERR_WRITE;
- offset += n;
- }
- return ret;
-}
-
/* 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)
+const void *
+get_resource_entry(const void *p, struct resource_entry *entry)
{
u64 size;
u8 flags;
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;
+ }
return p;
}
/* 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)
+void *
+put_resource_entry(void *p, const struct resource_entry *entry)
{
p = put_u56(p, entry->size);
p = put_u8(p, entry->flags);
return p;
}
-/* Given the compression type for the WIM file as a whole as the flags field of
- * a resource entry, returns the compression type for that resource entry. */
-int resource_compression_type(int wim_ctype, int reshdr_flags)
+static FILE *
+wim_get_fp(WIMStruct *w)
{
- if (wim_ctype == WIM_COMPRESSION_TYPE_NONE) {
- return WIM_COMPRESSION_TYPE_NONE;
- } else {
- if (reshdr_flags & WIM_RESHDR_FLAG_COMPRESSED)
- return wim_ctype;
- else
- return WIM_COMPRESSION_TYPE_NONE;
+#if defined(WITH_FUSE) || defined(ENABLE_MULTITHREADED_COMPRESSION)
+ pthread_mutex_lock(&w->fp_tab_mutex);
+ FILE *fp;
+
+ wimlib_assert(w->filename != NULL);
+
+ for (size_t i = 0; i < w->num_allocated_fps; i++) {
+ if (w->fp_tab[i]) {
+ fp = w->fp_tab[i];
+ w->fp_tab[i] = NULL;
+ goto out_unlock;
+ }
}
+ DEBUG("Opening extra file descriptor to `%"TS"'", w->filename);
+ fp = tfopen(w->filename, T("rb"));
+ if (!fp)
+ ERROR_WITH_ERRNO("Failed to open `%"TS"'", w->filename);
+out_unlock:
+ pthread_mutex_unlock(&w->fp_tab_mutex);
+#else /* WITH_FUSE || ENABLE_MULTITHREADED_COMPRESSION */
+ fp = w->fp;
+#endif /* !WITH_FUSE && !ENABLE_MULTITHREADED_COMPRESSION */
+ return fp;
}
-/*
- * 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.
- *
- * @fp: The FILE* for the input WIM file.
- * @res_entry: The resource entry for the metadata resource (a.k.a the metadata
- * for the metadata)
- * @wim_ctype: The compression type of the WIM file.
- * @root_dentry_p: A pointer to a pointer to a struct dentry structure into which the
- * root dentry is allocated and returned.
- *
- * @return: True on success, false on failure.
- */
-int read_metadata_resource(FILE *fp, const struct resource_entry *res_entry,
- int wim_ctype, struct dentry **root_dentry_p)
+static int
+wim_release_fp(WIMStruct *w, FILE *fp)
{
- u8 *buf;
- int ctype;
- u32 dentry_offset;
+ int ret = 0;
+#if defined(WITH_FUSE) || defined(ENABLE_MULTITHREADED_COMPRESSION)
+ FILE **fp_tab;
+
+ pthread_mutex_lock(&w->fp_tab_mutex);
+
+ for (size_t i = 0; i < w->num_allocated_fps; i++) {
+ if (w->fp_tab[i] == NULL) {
+ w->fp_tab[i] = fp;
+ goto out_unlock;
+ }
+ }
+
+ fp_tab = REALLOC(w->fp_tab, sizeof(FILE*) * (w->num_allocated_fps + 4));
+ if (!fp_tab) {
+ ret = WIMLIB_ERR_NOMEM;
+ fclose(fp);
+ goto out_unlock;
+ }
+ w->fp_tab = fp_tab;
+ memset(&w->fp_tab[w->num_allocated_fps], 0, 4 * sizeof(FILE*));
+ w->fp_tab[w->num_allocated_fps] = fp;
+ w->num_allocated_fps += 4;
+out_unlock:
+ pthread_mutex_unlock(&w->fp_tab_mutex);
+#endif /* WITH_FUSE || ENABLE_MULTITHREADED_COMPRESSION */
+ return ret;
+}
+
+static int
+read_partial_wim_resource(const struct wim_lookup_table_entry *lte,
+ u64 size,
+ consume_data_callback_t cb,
+ void *ctx_or_buf,
+ int flags,
+ u64 offset)
+{
+ FILE *wim_fp;
+ WIMStruct *wim;
int ret;
- struct dentry *dentry;
- DEBUG("Reading metadata resource: length = %lu, offset = %lu\n",
- res_entry->original_size, res_entry->offset);
+ wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
- if (res_entry->original_size < 8) {
- ERROR("Expected at least 8 bytes for the metadata "
- "resource!\n");
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
+ wim = lte->wim;
+ if (flags & WIMLIB_RESOURCE_FLAG_THREADSAFE_READ) {
+ wim_fp = wim_get_fp(wim);
+ if (!wim_fp) {
+ ret = WIMLIB_ERR_READ;
+ goto out;
+ }
+ } else {
+ wim_fp = lte->wim->fp;
}
- /* Allocate memory for the uncompressed metadata resource. */
- buf = MALLOC(res_entry->original_size);
+ if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED &&
+ !(flags & WIMLIB_RESOURCE_FLAG_RAW))
+ {
+ ret = read_compressed_resource(wim_fp,
+ lte->resource_entry.size,
+ lte->resource_entry.original_size,
+ lte->resource_entry.offset,
+ wimlib_get_compression_type(wim),
+ size,
+ offset,
+ cb,
+ ctx_or_buf);
+ } else {
+ offset += lte->resource_entry.offset;
- if (!buf) {
- ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
- "metadata resource!\n",
- res_entry->original_size);
- return WIMLIB_ERR_NOMEM;
+ if (fseeko(wim_fp, offset, SEEK_SET)) {
+ ERROR_WITH_ERRNO("Failed to seek to offset %"PRIu64
+ " in WIM", offset);
+ ret = WIMLIB_ERR_READ;
+ goto out_release_fp;
+ }
+ if (cb) {
+ /* Send data to callback function */
+ u8 buf[min(WIM_CHUNK_SIZE, size)];
+ while (size) {
+ size_t bytes_to_read = min(WIM_CHUNK_SIZE, size);
+ size_t bytes_read = fread(buf, 1, bytes_to_read, wim_fp);
+
+ if (bytes_read != bytes_to_read)
+ goto read_error;
+ ret = cb(buf, bytes_read, ctx_or_buf);
+ if (ret)
+ goto out_release_fp;
+ size -= bytes_read;
+ }
+ } else {
+ /* Send data directly to a buffer */
+ if (fread(ctx_or_buf, 1, size, wim_fp) != size)
+ goto read_error;
+ }
+ ret = 0;
+ }
+ goto out_release_fp;
+read_error:
+ if (ferror(wim_fp))
+ ERROR_WITH_ERRNO("Error reading data from WIM");
+ else
+ ERROR("Unexpected EOF in WIM!");
+ ret = WIMLIB_ERR_READ;
+out_release_fp:
+ if (flags & WIMLIB_RESOURCE_FLAG_THREADSAFE_READ) {
+ int ret2 = wim_release_fp(wim, wim_fp);
+ if (ret == 0)
+ ret = ret2;
}
+out:
+ if (ret) {
+ if (errno == 0)
+ errno = EIO;
+ }
+ return ret;
+}
- /* Determine the compression type of the metadata resource. */
- ctype = resource_compression_type(wim_ctype, res_entry->flags);
- /* Read the metadata resource into memory. (It may be compressed.) */
- ret = read_full_resource(fp, res_entry->size,
- res_entry->original_size, res_entry->offset,
- ctype, buf);
- if (ret != 0)
- goto err1;
+int
+read_partial_wim_resource_into_buf(const struct wim_lookup_table_entry *lte,
+ size_t size, u64 offset, void *buf,
+ bool threadsafe)
+{
+ return read_partial_wim_resource(lte, size, NULL, buf,
+ threadsafe ? WIMLIB_RESOURCE_FLAG_THREADSAFE_READ : 0,
+ offset);
+}
- DEBUG("Finished reading metadata resource into memory.\n");
+static int
+read_wim_resource_prefix(const struct wim_lookup_table_entry *lte,
+ u64 size,
+ consume_data_callback_t cb,
+ void *ctx_or_buf,
+ int flags)
+{
+ return read_partial_wim_resource(lte, size, cb, ctx_or_buf, flags, 0);
+}
-#if 0
- /* Read the security data into a WIMSecurityData structure. */
- if (!read_security_data(buf, res_entry->original_size, sd))
- goto err1;
-#endif
- dentry = MALLOC(sizeof(struct dentry));
- if (!dentry) {
- ERROR("Failed to allocate %zu bytes for root dentry!\n",
- sizeof(struct dentry));
- ret = WIMLIB_ERR_NOMEM;
- goto err1;
+#ifndef __WIN32__
+static int
+read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte,
+ u64 size,
+ consume_data_callback_t cb,
+ void *ctx_or_buf,
+ int _ignored_flags)
+{
+ const tchar *filename = lte->file_on_disk;
+ int ret;
+ int fd;
+ size_t bytes_read;
+
+ fd = open(filename, O_RDONLY);
+ if (fd < 0) {
+ ERROR_WITH_ERRNO("Can't open \"%"TS"\"", filename);
+ return WIMLIB_ERR_OPEN;
+ }
+ if (cb) {
+ /* Send data to callback function */
+ u8 buf[min(WIM_CHUNK_SIZE, size)];
+ size_t bytes_to_read;
+ while (size) {
+ bytes_to_read = min(WIM_CHUNK_SIZE, size);
+ bytes_read = full_read(fd, buf, bytes_to_read);
+ if (bytes_read != bytes_to_read)
+ goto read_error;
+ ret = cb(buf, bytes_read, ctx_or_buf);
+ if (ret)
+ goto out_close;
+ size -= bytes_read;
+ }
+ } else {
+ /* Send data directly to a buffer */
+ bytes_read = full_read(fd, ctx_or_buf, size);
+ if (bytes_read != size)
+ goto read_error;
}
+ ret = 0;
+ goto out_close;
+read_error:
+ ERROR_WITH_ERRNO("Error reading \"%"TS"\"", filename);
+ ret = WIMLIB_ERR_READ;
+out_close:
+ close(fd);
+ return ret;
+}
+#endif /* !__WIN32__ */
- /* Read the root directory entry starts after security data, on an
- * 8-byte aligned address.
- *
- * The security data starts with a 4-byte integer giving its total
- * length. */
- get_u32(buf, &dentry_offset);
- dentry_offset += (8 - dentry_offset % 8) % 8;
-
- ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry);
- if (ret != 0)
- goto err1;
-
- /* This is the root dentry, so set its pointers correctly. */
- dentry->parent = dentry;
- dentry->next = dentry;
- dentry->prev = dentry;
-
- /* Now read the entire directory entry tree. */
- ret = read_dentry_tree(buf, res_entry->original_size, dentry);
- if (ret != 0)
- goto err2;
-
- /* Calculate the full paths in the dentry tree. */
- ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
- if (ret != 0)
- goto err2;
-
- *root_dentry_p = dentry;
- FREE(buf);
+static int
+read_buffer_prefix(const struct wim_lookup_table_entry *lte,
+ u64 size, consume_data_callback_t cb,
+ void *ctx_or_buf, int _ignored_flags)
+{
+ const void *inbuf = lte->attached_buffer;
+ int ret;
+
+ if (cb) {
+ while (size) {
+ size_t chunk_size = min(WIM_CHUNK_SIZE, size);
+ ret = cb(inbuf, chunk_size, ctx_or_buf);
+ if (ret)
+ return ret;
+ size -= chunk_size;
+ inbuf += chunk_size;
+ }
+ } else {
+ memcpy(ctx_or_buf, inbuf, size);
+ }
+ return 0;
+}
+
+typedef int (*read_resource_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
+ u64 size,
+ consume_data_callback_t cb,
+ void *ctx_or_buf,
+ int flags);
+
+/*
+ * Read the first @size bytes from a generic "resource", which may be located in
+ * the WIM (compressed or uncompressed), in an external file, or directly in an
+ * in-memory buffer.
+ *
+ * Feed the data either to a callback function (cb != NULL, passing it
+ * ctx_or_buf), or write it directly into a buffer (cb == NULL, ctx_or_buf
+ * specifies the buffer, which must have room for @size bytes).
+ *
+ * When using a callback function, it is called with chunks up to 32768 bytes in
+ * size until the resource is exhausted.
+ *
+ * If the resource is located in a WIM file, @flags can be:
+ * * WIMLIB_RESOURCE_FLAG_THREADSAFE_READ if it must be safe to access the resource
+ * concurrently by multiple threads.
+ * * WIMLIB_RESOURCE_FLAG_RAW if the raw compressed data is to be supplied
+ * instead of the uncompressed data.
+ * Otherwise, the @flags are ignored.
+ */
+int
+read_resource_prefix(const struct wim_lookup_table_entry *lte,
+ u64 size, consume_data_callback_t cb, void *ctx_or_buf,
+ int flags)
+{
+ static const read_resource_prefix_handler_t handlers[] = {
+ [RESOURCE_IN_WIM] = read_wim_resource_prefix,
+ #ifndef __WIN32__
+ [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] = read_win32_file_prefix,
+ [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, ctx_or_buf, flags);
+}
+
+int
+read_full_resource_into_buf(const struct wim_lookup_table_entry *lte,
+ void *buf, bool thread_safe)
+{
+ return read_resource_prefix(lte,
+ wim_resource_size(lte),
+ NULL, buf,
+ thread_safe ? WIMLIB_RESOURCE_FLAG_THREADSAFE_READ : 0);
+}
+
+struct extract_ctx {
+ SHA_CTX sha_ctx;
+ consume_data_callback_t extract_chunk;
+ void *extract_chunk_arg;
+};
+
+static int
+extract_chunk_sha1_wrapper(const void *chunk, size_t chunk_size,
+ void *_ctx)
+{
+ struct extract_ctx *ctx = _ctx;
+
+ sha1_update(&ctx->sha_ctx, chunk, chunk_size);
+ return ctx->extract_chunk(chunk, chunk_size, ctx->extract_chunk_arg);
+}
+
+/* Extracts the first @size bytes of a WIM resource to somewhere. In the
+ * process, the SHA1 message digest of the resource is checked if the full
+ * resource is being extracted.
+ *
+ * @extract_chunk is a function that is called to extract each chunk of the
+ * resource. */
+int
+extract_wim_resource(const struct wim_lookup_table_entry *lte,
+ u64 size,
+ consume_data_callback_t extract_chunk,
+ void *extract_chunk_arg)
+{
+ int ret;
+ if (size == wim_resource_size(lte)) {
+ /* Do SHA1 */
+ struct extract_ctx ctx;
+ ctx.extract_chunk = extract_chunk;
+ ctx.extract_chunk_arg = extract_chunk_arg;
+ sha1_init(&ctx.sha_ctx);
+ ret = read_resource_prefix(lte, size,
+ extract_chunk_sha1_wrapper,
+ &ctx, 0);
+ if (ret == 0) {
+ u8 hash[SHA1_HASH_SIZE];
+ sha1_final(hash, &ctx.sha_ctx);
+ if (!hashes_equal(hash, lte->hash)) {
+ #ifdef ENABLE_ERROR_MESSAGES
+ ERROR_WITH_ERRNO("Invalid SHA1 message digest "
+ "on the following WIM resource:");
+ print_lookup_table_entry(lte, stderr);
+ if (lte->resource_location == RESOURCE_IN_WIM)
+ ERROR("The WIM file appears to be corrupt!");
+ #endif
+ ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
+ }
+ }
+ } else {
+ /* Don't do SHA1 */
+ ret = read_resource_prefix(lte, size, extract_chunk,
+ extract_chunk_arg, 0);
+ }
return ret;
-err2:
- free_dentry_tree(dentry, NULL, false);
-err1:
- FREE(buf);
+}
+
+static int
+extract_wim_chunk_to_fd(const void *buf, size_t len, void *_fd_p)
+{
+ int fd = *(int*)_fd_p;
+ ssize_t ret = full_write(fd, buf, len);
+ if (ret < len) {
+ ERROR_WITH_ERRNO("Error writing to file descriptor");
+ return WIMLIB_ERR_WRITE;
+ } else {
+ return 0;
+ }
+}
+
+int
+extract_wim_resource_to_fd(const struct wim_lookup_table_entry *lte,
+ int fd, u64 size)
+{
+ return extract_wim_resource(lte, size, extract_wim_chunk_to_fd, &fd);
+}
+
+
+static int
+sha1_chunk(const void *buf, size_t len, void *ctx)
+{
+ sha1_update(ctx, buf, len);
+ return 0;
+}
+
+/* Calculate the SHA1 message digest of a stream. */
+int
+sha1_resource(struct wim_lookup_table_entry *lte)
+{
+ int ret;
+ SHA_CTX sha_ctx;
+
+ sha1_init(&sha_ctx);
+ ret = read_resource_prefix(lte, wim_resource_size(lte),
+ sha1_chunk, &sha_ctx, 0);
+ if (ret == 0)
+ sha1_final(lte->hash, &sha_ctx);
return ret;
}
+/*
+ * 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.
+ *
+ * (This function is confusing and should be refactored somehow.)
+ */
+int
+copy_resource(struct wim_lookup_table_entry *lte, void *wim)
+{
+ WIMStruct *w = wim;
+ int ret;
+ ret = write_wim_resource(lte, w->out_fp,
+ wim_resource_compression_type(lte),
+ <e->output_resource_entry, 0);
+ if (ret == 0) {
+ lte->out_refcnt = lte->refcnt;
+ lte->part_number = w->hdr.part_number;
+ }
+ return ret;
+}