read_resource_prefix(): Remove comment about removed flag
[wimlib] / src / resource.c
index b525f22..4b91be6 100644 (file)
@@ -1,83 +1,75 @@
 /*
  * 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 "dentry.h"
 #include "lookup_table.h"
-#include "io.h"
-#include "lzx.h"
-#include "xpress.h"
+#include "buffer_io.h"
 #include "sha1.h"
-#include "dentry.h"
-#include <unistd.h>
+
+#ifdef __WIN32__
+#  include "win32.h"
+#endif
+
 #include <errno.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <fcntl.h>
 
-/* Used for buffering FILE IO */
-#define BUFFER_SIZE 4096
+#ifdef HAVE_ALLOCA_H
+#  include <alloca.h>
+#endif
 
-/* 
- * Reads all or part of a compressed resource into an in-memory buffer.
- *
- * @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.
+/*
+ * 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(int in_fd,
+                        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
@@ -88,10 +80,11 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
         * 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. */
@@ -118,9 +111,22 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
        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.
@@ -136,7 +142,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
        /* 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. */
@@ -150,27 +156,23 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
 
        /* 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;
-       }
 
        /* 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;
+       if (full_pread(in_fd, chunk_tab_buf, size,
+                      file_offset_of_needed_chunk_entries) != size)
+               goto read_error;
 
        /* Now fill in chunk_offsets from the entries we have read in
         * chunk_tab_buf. */
@@ -182,29 +184,26 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
        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 + 
-                               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;
-       }
+       u64 cur_read_offset = resource_offset + chunk_table_size + chunk_offsets[0];
 
        /* 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
@@ -212,24 +211,21 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
         * 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 {
@@ -237,15 +233,15 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
                         * 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
@@ -254,10 +250,6 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
                                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)
@@ -270,249 +262,93 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size,
                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;
-                               }
+                       /* Uncompressed chunk */
+                       if (full_pread(in_fd,
+                                      cb ? out_p + start_offset : out_p,
+                                      partial_chunk_size,
+                                      cur_read_offset + start_offset) != partial_chunk_size)
+                       {
+                               goto read_error;
                        }
-                       if (fread(out_p, 1, partial_chunk_size, fp) != 
-                                       partial_chunk_size)
-                               goto err;
                } else {
                        /* Compressed chunk */
-                       int ret;
 
                        /* Read the compressed data into compressed_buf. */
-                       if (fread(compressed_buf, 1, compressed_chunk_size, 
-                                               fp) != compressed_chunk_size)
-                               goto err;
+                       if (full_pread(in_fd,
+                                      compressed_buf,
+                                      compressed_chunk_size,
+                                      cur_read_offset) != compressed_chunk_size)
+                       {
+                               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;
-       }
-
-       return 0;
-
-err:
-       if (feof(fp))
-               ERROR("Unexpected EOF in compressed file resource\n");
-       else
-               ERROR("Error reading compressed file resource: %m\n");
-       return WIMLIB_ERR_READ;
-}
-
-/* 
- * 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")!\n", 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");
+               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 {
-                       ERROR("Failed to read %"PRIu64" bytes from "
-                                       "uncompressed resource at offset "
-                                       "%"PRIu64"\n", 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;
+                       /* 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 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);
+               cur_read_offset += compressed_chunk_size;
        }
-}
 
-
-/* 
- * 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;
-       }
+       ret = 0;
+out:
+       if (chunk_offsets_malloced)
+               FREE(chunk_offsets);
        return ret;
-}
-
-/* 
- * Copies the file resource specified by the lookup table entry @lte from the
- * input WIM, pointed to by the fp field of the WIMStruct, to the output WIM,
- * pointed to by the out_fp field of the WIMStruct.
- *
- * The output_resource_entry, out_refcnt, and part_number fields of @lte are
- * updated.
- *
- * Metadata resources are not copied (they are handled elsewhere for joining and
- * splitting).
- */
-int copy_resource(struct lookup_table_entry *lte, void *w)
-{
-       if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
-           !((WIMStruct*)w)->write_metadata) {
-               return 0;
-       }
-
-       FILE *in_fp = ((WIMStruct*)w)->fp;
-       FILE *out_fp = ((WIMStruct*)w)->out_fp;
-       int ret;
-       u64 size = lte->resource_entry.size;
-       u64 offset = lte->resource_entry.offset;
-       off_t new_offset = ftello(out_fp);
-
-       if (new_offset == -1)
-               return WIMLIB_ERR_WRITE;
-
-       ret = copy_between_files(in_fp, offset, out_fp, size);
-       if (ret != 0)
-               return ret;
 
-       memcpy(&lte->output_resource_entry, &lte->resource_entry, 
-                       sizeof(struct resource_entry));
-
-       lte->output_resource_entry.offset = new_offset;
-       lte->out_refcnt = lte->refcnt;
-       lte->part_number = ((WIMStruct*)w)->hdr.part_number;
-       return 0;
+read_error:
+       ERROR_WITH_ERRNO("Error reading compressed file resource");
+       ret = WIMLIB_ERR_READ;
+       goto out;
 }
 
 /* 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;
@@ -521,15 +357,29 @@ const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry)
        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);
@@ -538,656 +388,344 @@ u8 *put_resource_entry(u8 *p, const struct resource_entry *entry)
        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 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)
 {
-       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;
-       }
-}
-
-
-
-/*
- * Copies bytes between two file streams.
- *
- * Copies @len bytes from @in to @out, at the current position in @out, and at
- * an offset of @in_offset in @in.
- */
-int copy_between_files(FILE *in, off_t in_offset, FILE *out, size_t len)
-{
-       u8 buf[BUFFER_SIZE];
-       size_t n;
+       WIMStruct *wim;
+       int in_fd;
+       int ret;
 
-       if (fseeko(in, in_offset, SEEK_SET) != 0) {
-               ERROR("Failed to seek to byte %"PRIu64" of input file: %m\n",
-                               in_offset);
-               return WIMLIB_ERR_READ;
-       }
-       /* To reduce memory usage and improve speed, read and write BUFFER_SIZE
-        * bytes at a time. */
-       while (len != 0) {
-               n = min(len, BUFFER_SIZE);
-               if (fread(buf, 1, n, in) != n) {
-                       if (feof(in)) {
-                               ERROR("Unexpected EOF when copying data "
-                                               "between files\n");
-                       } else {
-                               ERROR("Error copying data between files: %m\n");
+       wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
+
+       wim = lte->wim;
+       in_fd = wim->in_fd;
+
+       if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED &&
+           !(flags & WIMLIB_RESOURCE_FLAG_RAW))
+       {
+               ret = read_compressed_resource(in_fd,
+                                              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 (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 = full_pread(in_fd, buf,
+                                                              bytes_to_read, offset);
+                               if (bytes_read != bytes_to_read)
+                                       goto read_error;
+                               ret = cb(buf, bytes_read, ctx_or_buf);
+                               if (ret)
+                                       goto out;
+                               size -= bytes_read;
+                               offset += bytes_read;
                        }
-                       return WIMLIB_ERR_READ;
-               }
-
-               if (fwrite(buf, 1, n, out) != n) {
-                       ERROR("Error copying data between files: %m\n");
-                       return WIMLIB_ERR_WRITE;
+               } else {
+                       /* Send data directly to a buffer */
+                       if (full_pread(in_fd, ctx_or_buf, size, offset) != size)
+                               goto read_error;
                }
-               len -= n;
+               ret = 0;
        }
-       return 0;
+       goto out;
+read_error:
+       ERROR_WITH_ERRNO("Error reading data from WIM");
+       ret = WIMLIB_ERR_READ;
+out:
+       if (ret) {
+               if (errno == 0)
+                       errno = EIO;
+       }
+       return ret;
 }
 
 
-/* 
- * Uncompresses a WIM file resource and writes it uncompressed to a file stream.
- *
- * @in:                    The file stream that contains the file resource.
- * @size:           The size of the resource in the input file.
- * @original_size:  The original (uncompressed) size of the resource. 
- * @offset:        The offset of the start of the resource in @in.
- * @input_ctype:    The compression type of the resource in @in.
- * @out:           The file stream to write the file resource to.
- */
-static int uncompress_resource(FILE *in, u64 size, u64 original_size,
-                              off_t offset, int input_ctype, FILE *out)
+int
+read_partial_wim_resource_into_buf(const struct wim_lookup_table_entry *lte,
+                                  size_t size, u64 offset, void *buf)
 {
-       int ret;
-       u8 buf[WIM_CHUNK_SIZE];
-       /* Determine how many compressed chunks the file is divided into. */
-       u64 num_chunks;
-       u64 i;
-       u64 uncompressed_offset;
-       u64 uncompressed_chunk_size;
-       
-       num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
-
-       for (i = 0; i < num_chunks; i++) {
-
-               uncompressed_offset = i * WIM_CHUNK_SIZE;
-               uncompressed_chunk_size = min(WIM_CHUNK_SIZE, 
-                                       original_size - uncompressed_offset);
-
-               ret = read_resource(in, size, original_size, offset, input_ctype, 
-                                       uncompressed_chunk_size, 
-                                       uncompressed_offset, buf);
-               if (ret != 0)
-                       return ret;
-
-               if (fwrite(buf, 1, uncompressed_chunk_size, out) != 
-                                               uncompressed_chunk_size) {
-                       ERROR("Failed to write file resource: %m\n");
-                       return WIMLIB_ERR_WRITE;
-               }
-       }
-       return 0;
+       return read_partial_wim_resource(lte, size, NULL, buf, 0, offset);
 }
 
-/* 
- * Transfers a file resource between two files, writing it compressed.  The file
- * resource in the input file may be either compressed or uncompressed.
- * Alternatively, the input resource may be in-memory, but it must be
- * uncompressed.
- *
- * @in:                    The file stream that contains the file resource.  Ignored
- *                     if uncompressed_resource != NULL.
- * @uncompressed_resource:     If this pointer is not NULL, it points to an
- *                                     array of @original_size bytes that are
- *                                     the uncompressed input resource.
- * @size:           The size of the resource in the input file.
- * @original_size:  The original (uncompressed) size of the resource. 
- * @offset:        The offset of the start of the resource in @in.  Ignored
- *                     if uncompressed_resource != NULL.
- * @input_ctype:    The compression type of the resource in @in.  Ignored if
- *                     uncompressed_resource != NULL.
- * @out:           The file stream to write the file resource to.
- * @output_type:    The compression type to use when writing the resource to
- *                     @out.
- * @new_size_ret:   A location into which the new compressed size of the file
- *                     resource in returned.
- */
-static int recompress_resource(FILE *in, const u8 *uncompressed_resource, 
-                                       u64 size, u64 original_size,
-                                       off_t offset, int input_ctype, FILE *out,
-                                       int output_ctype, u64 *new_size_ret)
+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)
 {
-       int ret;
-       int (*compress)(const void *, uint, void *, uint *);
-       if (output_ctype == WIM_COMPRESSION_TYPE_LZX)
-               compress = lzx_compress;
-       else
-               compress = xpress_compress;
-
-       u8 uncompressed_buf[WIM_CHUNK_SIZE];
-       u8 compressed_buf[WIM_CHUNK_SIZE - 1];
-
-       /* Determine how many compressed chunks the file needs to be divided
-        * into. */
-       u64 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
-
-       u64 num_chunk_entries = num_chunks - 1;
-
-       /* Size of the chunk entries--- 8 bytes for files over 4GB, otherwise 4
-        * bytes */
-       uint chunk_entry_size = (original_size >= (u64)1 << 32) ?  8 : 4;
-
-       /* Array in which to construct the chunk offset table. */
-       u64 chunk_offsets[num_chunk_entries];
-
-       /* Offset of the start of the chunk table in the output file. */
-       off_t chunk_tab_offset = ftello(out);
-
-       /* Total size of the chunk table (as written to the file) */
-       u64 chunk_tab_size = chunk_entry_size * num_chunk_entries;
-
-       /* Reserve space for the chunk table. */
-       if (fwrite(chunk_offsets, 1, chunk_tab_size, out) != chunk_tab_size) {
-               ERROR("Failed to write chunk offset table: %m\n");
-               return WIMLIB_ERR_WRITE;
-       }
-
-       /* Read each chunk of the file, compress it, write it to the output
-        * file, and update th chunk offset table. */
-       u64 cur_chunk_offset = 0;
-       for (u64 i = 0; i < num_chunks; i++) {
-
-               u64 uncompressed_offset = i * WIM_CHUNK_SIZE;
-               u64 uncompressed_chunk_size = min(WIM_CHUNK_SIZE, 
-                                       original_size - uncompressed_offset);
-
-               const u8 *uncompressed_p;
-               if (uncompressed_resource != NULL) {
-                       uncompressed_p = uncompressed_resource + 
-                                                       uncompressed_offset;
-
-               } else {
-                       /* Read chunk i of the file into uncompressed_buf. */
-                       ret = read_resource(in, size, original_size, offset, input_ctype, 
-                                               uncompressed_chunk_size, 
-                                               uncompressed_offset, 
-                                               uncompressed_buf);
-                       if (ret != 0)
-                               return ret;
-                       uncompressed_p = uncompressed_buf;
-               }
-
-               if (i != 0)
-                       chunk_offsets[i - 1] = cur_chunk_offset;
-
-               uint compressed_len;
-
-               ret = compress(uncompressed_p, uncompressed_chunk_size, 
-                              compressed_buf, &compressed_len);
-
-               /* if compress() returned nonzero, the compressed chunk would
-                * have been at least as large as the uncompressed chunk.  In
-                * this situation, the WIM format requires that the uncompressed
-                * chunk be written instead. */
-               const u8 *buf_to_write;
-               uint len_to_write;
-               if (ret == 0) {
-                       buf_to_write = compressed_buf;
-                       len_to_write = compressed_len;
-               } else {
-                       buf_to_write = uncompressed_p;
-                       len_to_write = uncompressed_chunk_size;
-               }
-
-               if (fwrite(buf_to_write, 1, len_to_write, out) != len_to_write) {
-                       ERROR("Failed to write compressed file resource: %m\n");
-                       return WIMLIB_ERR_WRITE;
-               }
-               cur_chunk_offset += len_to_write;
-       }
+       return read_partial_wim_resource(lte, size, cb, ctx_or_buf, flags, 0);
+}
 
-       /* The chunk offset after the last chunk, plus the size of the chunk
-        * table, gives the total compressed size of the resource. */
-       *new_size_ret = cur_chunk_offset + chunk_tab_size;
 
-       /* Now that all entries of the chunk table are determined, rewind the
-        * stream to where the chunk table was, and write it back out. */
+#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;
 
-       if (fseeko(out, chunk_tab_offset, SEEK_SET) != 0) {
-               ERROR("Failed to seek to beginning of chunk table: %m\n");
-               return WIMLIB_ERR_READ;
+       fd = open(filename, O_RDONLY);
+       if (fd < 0) {
+               ERROR_WITH_ERRNO("Can't open \"%"TS"\"", filename);
+               return WIMLIB_ERR_OPEN;
        }
-
-       if (chunk_entry_size == 8) {
-               array_to_le64(chunk_offsets, num_chunk_entries);
-
-               if (fwrite(chunk_offsets, 1, chunk_tab_size, out) != 
-                               chunk_tab_size) {
-                       ERROR("Failed to write chunk table: %m\n");
-                       return WIMLIB_ERR_WRITE;
+       if (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 {
-               u32 chunk_entries_small[num_chunk_entries];
-               for (u64 i = 0; i < num_chunk_entries; i++)
-                       chunk_entries_small[i] = to_le32(chunk_offsets[i]);
-               if (fwrite(chunk_entries_small, 1, chunk_tab_size, out) != 
-                               chunk_tab_size) {
-                       ERROR("Failed to write chunk table: %m\n");
-                       return WIMLIB_ERR_WRITE;
-               }
-       }
-
-       if (fseeko(out, 0, SEEK_END) != 0) {
-               ERROR("Failed to seek to end of output file: %m\n");
-               return WIMLIB_ERR_WRITE;
+               /* Send data directly to a buffer */
+               bytes_read = full_read(fd, ctx_or_buf, size);
+               if (bytes_read != size)
+                       goto read_error;
        }
-
-       return 0;
+       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__ */
 
-int write_resource_from_memory(const u8 resource[], int out_ctype,
-                              u64 resource_original_size, FILE *out,
-                              u64 *resource_size_ret)
+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)
 {
-       if (out_ctype == WIM_COMPRESSION_TYPE_NONE) {
-               if (fwrite(resource, 1, resource_original_size, out) != 
-                                       resource_original_size) {
-                       ERROR("Failed to write resource of length "
-                                       "%"PRIu64": %m\n", 
-                                       resource_original_size);
-                       return WIMLIB_ERR_WRITE;
+       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;
                }
-               *resource_size_ret = resource_original_size;
-               return 0;
        } else {
-               return recompress_resource(NULL, resource, resource_original_size,
-                               resource_original_size, 0, 0, out, out_ctype, 
-                                                       resource_size_ret);
+               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);
 
-/* 
- * Transfers a file resource from a FILE* opened for reading to a FILE* opened
- * for writing, possibly changing the compression type. 
+/*
+ * 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).
  *
- * @in:                        The FILE* that contains the file resource.
- * @size:              The (compressed) size of the file resource.
- * @original_size:     The uncompressed size of the file resource.
- * @offset:            The offset of the file resource in the input file.
- * @input_ctype:       The compression type of the file resource in the input
- *                             file.
- * @out:               The FILE* for the output file.  The file resource is 
- *                             written at the current position of @out.
- * @output_ctype:      The compression type to which the file resource will be
- *                             converted.
- * @output_res_entry:  A pointer to a resource entry that, upon successful
- *                             return of this function,  will have the size,
- *                             original size, offset, and flags fields filled
- *                             in for the file resource written to the output
- *                             file.
+ * 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_RAW if the raw compressed data is to be supplied
+ *     instead of the uncompressed data.
+ * Otherwise, the @flags are ignored.
  */
-static int transfer_file_resource(FILE *in, u64 size, u64 original_size, 
-                                 off_t offset, int input_ctype, FILE *out, 
-                                 int output_ctype, 
-                                 struct resource_entry *output_res_entry)
+int
+read_resource_prefix(const struct wim_lookup_table_entry *lte,
+                    u64 size, consume_data_callback_t cb, void *ctx_or_buf,
+                    int flags)
 {
-       int ret;
-
-       /* Handle zero-length files */
-       if (original_size == 0) {
-               memset(output_res_entry, 0, sizeof(*output_res_entry));
-               return 0;
-       }
-
-       /* Get current offset in the output file. */
-       output_res_entry->offset = ftello(out);
-       if (output_res_entry->offset == -1) {
-               ERROR("Failed to get output position: %m\n");
-               return WIMLIB_ERR_WRITE;
-       }
+       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);
+}
 
-       if (output_ctype == input_ctype) {
-               /* The same compression types; simply copy the resource. */
+int
+read_full_resource_into_buf(const struct wim_lookup_table_entry *lte,
+                           void *buf)
+{
+       return read_resource_prefix(lte, wim_resource_size(lte), NULL, buf, 0);
+}
 
-               ret = copy_between_files(in, offset, out, size);
-               if (ret != 0)
-                       return ret;
-               output_res_entry->size = size;
-       } else {
-               /* Different compression types. */
-
-               if (output_ctype == WIM_COMPRESSION_TYPE_NONE) {
-                       /* Uncompress a compressed file resource */
-                       ret = uncompress_resource(in, size,
-                                               original_size, offset, 
-                                               input_ctype, out);
-                       if (ret != 0)
-                               return ret;
-                       output_res_entry->size = original_size;
-               } else {
-                       u64 new_size;
-                       /* Compress an uncompressed file resource, or compress a
-                        * compressed file resource using a different
-                        * compression type */
-                       ret = recompress_resource(in, NULL, size, original_size,
-                                               offset, input_ctype, out, 
-                                               output_ctype, &new_size);
-                       if (ret != 0)
-                               return ret;
-                       output_res_entry->size = new_size;
-               }
+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;
 
-       output_res_entry->original_size = original_size;
-       if (output_ctype == WIM_COMPRESSION_TYPE_NONE)
-               output_res_entry->flags = 0;
-       else
-               output_res_entry->flags = WIM_RESHDR_FLAG_COMPRESSED;
-       return 0;
+       sha1_update(&ctx->sha_ctx, chunk, chunk_size);
+       return ctx->extract_chunk(chunk, chunk_size, ctx->extract_chunk_arg);
 }
 
-/* 
- * 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.
+/* 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.
  *
- * @return:    True on success, false on failure.
- */
-int read_metadata_resource(FILE *fp, const struct resource_entry *res_entry,
-                          int wim_ctype, struct image_metadata *image_metadata)
+ * @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)
 {
-       u8 *buf;
-       int ctype;
-       u32 dentry_offset;
        int ret;
-       struct dentry *dentry = NULL;
-
-       DEBUG("Reading metadata resource: length = %lu, offset = %lu\n",
-                       res_entry->original_size, res_entry->offset);
-
-       if (res_entry->original_size < 8) {
-               ERROR("Expected at least 8 bytes for the metadata "
-                               "resource!\n");
-               return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
+       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("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;
+}
 
-       /* Allocate memory for the uncompressed metadata resource. */
-       buf = MALLOC(res_entry->original_size);
-
-       if (!buf) {
-               ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
-                               "metadata resource!\n",
-                               res_entry->original_size);
-               return WIMLIB_ERR_NOMEM;
+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;
        }
+}
 
-       /* Determine the compression type of the metadata resource. */
-       ctype = resource_compression_type(wim_ctype, res_entry->flags);
-
-       /* Read the metadata resource into memory.  (It may be compressed.) */
-       ret = read_full_resource(fp, res_entry->size, 
-                                res_entry->original_size, res_entry->offset, 
-                                ctype, buf);
-       if (ret != 0)
-               goto err1;
-
-       DEBUG("Finished reading metadata resource into memory.\n");
-
+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);
+}
 
-       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;
-       }
 
-       /* Read the root directory entry starts after security data, on an
-        * 8-byte aligned address. 
-        *
-        * The security data starts with a 4-byte integer giving its total
-        * length. */
-
-       /* Read the security data into a WIMSecurityData structure. */
-#ifdef ENABLE_SECURITY_DATA
-       ret = read_security_data(buf, res_entry->original_size, 
-                                &image_metadata->security_data);
-       if (ret != 0)
-               goto err1;
-#endif
-       get_u32(buf, &dentry_offset);
-       if (dentry_offset == 0)
-               dentry_offset = 8;
-       dentry_offset += (8 - dentry_offset % 8) % 8;
-               
-       ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry);
-       if (ret != 0)
-               goto err1;
-
-       /* This is the root dentry, so set its pointers correctly. */
-       dentry->parent = dentry;
-       dentry->next   = dentry;
-       dentry->prev   = dentry;
-
-       /* Now read the entire directory entry tree. */
-       ret = read_dentry_tree(buf, res_entry->original_size, dentry);
-       if (ret != 0)
-               goto err2;
-
-       /* Calculate the full paths in the dentry tree. */
-       ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
-       if (ret != 0)
-               goto err2;
-
-       image_metadata->root_dentry = dentry;
-       FREE(buf);
-       return ret;
-err2:
-       free_dentry_tree(dentry, NULL, false);
-err1:
-       FREE(buf);
-       return ret;
+static int
+sha1_chunk(const void *buf, size_t len, void *ctx)
+{
+       sha1_update(ctx, buf, len);
+       return 0;
 }
 
-/* Write the metadata resource for the current image. */
-int write_metadata_resource(WIMStruct *w)
+/* Calculate the SHA1 message digest of a stream. */
+int
+sha1_resource(struct wim_lookup_table_entry *lte)
 {
-       FILE *out;
-       u8 *buf;
-       u8 *p;
        int ret;
-       off_t subdir_offset;
-       struct dentry *root;
-       struct lookup_table_entry *lte;
-       struct resource_entry *res_entry;
-       off_t metadata_offset;
-       u64 metadata_original_size;
-       u64 metadata_compressed_size;
-       int metadata_ctype;
-       u8  hash[WIM_HASH_SIZE];
-
-       DEBUG("Writing metadata resource for image %u\n", w->current_image);
-
-       out = w->out_fp;
-       root = wim_root_dentry(w);
-       metadata_ctype = wimlib_get_compression_type(w);
-       metadata_offset = ftello(out);
-       if (metadata_offset == -1)
-               return WIMLIB_ERR_WRITE;
+       SHA_CTX sha_ctx;
 
-       #ifdef ENABLE_SECURITY_DATA
-       struct wim_security_data *sd = wim_security_data(w);
-       if (sd)
-               subdir_offset = sd->total_length + root->length + 8;
-       else
-       #endif
-               subdir_offset = 8 + root->length + 8;
-       calculate_subdir_offsets(root, &subdir_offset);
-       metadata_original_size = subdir_offset;
-       buf = MALLOC(metadata_original_size);
-       if (!buf) {
-               ERROR("Failed to allocate %"PRIu64" bytes for "
-                               "metadata resource\n", metadata_original_size);
-               return WIMLIB_ERR_NOMEM;
-       }
-       #ifdef ENABLE_SECURITY_DATA
-       /* Write the security data. */
-       p = write_security_data(sd, buf);
-       #else
-       p = put_u32(buf, 8); /* Total length of security data. */
-       p = put_u32(p, 0); /* Number of security data entries. */
-       #endif
-
-       DEBUG("Writing dentry tree.\n");
-       p = write_dentry_tree(root, p);
-
-       /* Like file resources, the lookup table entry for a metadata resource
-        * uses for the hash code a SHA1 message digest of its uncompressed
-        * contents. */
-       sha1_buffer(buf, metadata_original_size, hash);
-
-       ret = write_resource_from_memory(buf, 
-                                        metadata_ctype,
-                                        metadata_original_size, 
-                                        out,
-                                        &metadata_compressed_size);
-       FREE(buf);
-       if (ret != 0)
-               return ret;
-
-       /* Update the lookup table entry, including the hash and output resource
-        * entry fields, for this image's metadata resource.  */
-       lte = wim_metadata_lookup_table_entry(w);
-       res_entry = &lte->output_resource_entry;
-       lte->out_refcnt++;
-       if (memcmp(hash, lte->hash, WIM_HASH_SIZE) != 0) {
-               lookup_table_unlink(w->lookup_table, lte);
-               memcpy(lte->hash, hash, WIM_HASH_SIZE);
-               lookup_table_insert(w->lookup_table, lte);
-       }
-       res_entry->original_size = metadata_original_size;
-       res_entry->offset        = metadata_offset;
-       res_entry->size          = metadata_compressed_size;
-       res_entry->flags         = WIM_RESHDR_FLAG_METADATA;
-       if (metadata_ctype != WIM_COMPRESSION_TYPE_NONE)
-               res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
-       return 0;
+       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;
 }
 
-/* 
- * Writes a file resource to the output file. 
+/*
+ * 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.
  *
- * @dentry:  The dentry for the file resource.
- * @wim_p:  A pointer to the WIMStruct.  The fields of interest to this
- *     function are the input and output file streams and the lookup table. 
- * @return zero on success, nonzero on failure. 
+ * (This function is confusing and should be refactored somehow.)
  */
-int write_file_resource(struct dentry *dentry, void *wim_p)
+int
+copy_resource(struct wim_lookup_table_entry *lte, void *wim)
 {
-       WIMStruct *w;
-       FILE *out;
-       FILE *in;
-       struct lookup_table_entry *lte;
-       int in_wim_ctype;
-       int out_wim_ctype;
-       int input_res_ctype;
-       struct resource_entry *input_res_entry;
-       struct resource_entry *output_res_entry;
-       u64 len;
+       WIMStruct *w = wim;
        int ret;
 
-       w = wim_p;
-       out = w->out_fp;
-
-       /* Directories don't need file resources. */
-       if (dentry_is_directory(dentry))
-               return 0;
-
-       /* Get the lookup entry for the file resource. */
-       lte = wim_lookup_resource(w, dentry);
-       if (!lte)
-               return 0;
-
-       /* No need to write file resources twice.  (This indicates file
-        * resources that are part of a hard link set.) */
-       if (++lte->out_refcnt != 1)
-               return 0;
-
-       out_wim_ctype = wimlib_get_compression_type(w);
-       output_res_entry = &lte->output_resource_entry;
-
-       /* Figure out if we can read the resource from the WIM file, or
-        * if we have to read it from the filesystem outside. */
-       if (lte->file_on_disk) {
-
-               /* Read from disk (uncompressed) */
-
-               len = lte->resource_entry.original_size;
-
-               in = fopen(lte->file_on_disk, "rb");
-               if (!in) {
-                       ERROR("Failed to open the file `%s': %m\n",
-                                       lte->file_on_disk);
-                       return WIMLIB_ERR_OPEN;
-               }
-
-               if (w->verbose)
-                       puts(lte->file_on_disk);
-
-               ret = transfer_file_resource(in, len, len, 0,
-                                            WIM_COMPRESSION_TYPE_NONE, out, 
-                                            out_wim_ctype, output_res_entry);
-               fclose(in);
-       } else {
-
-               /* Read from input WIM (possibly compressed) */
-
-               /* It may be a different WIM file, in the case of
-                * exporting images from one WIM file to another */
-               if (lte->other_wim_fp) {
-                       /* Different WIM file. */
-                       in = lte->other_wim_fp;
-                       in_wim_ctype = lte->other_wim_ctype;
-               } else {
-                       /* Same WIM file. */
-                       in = w->fp;
-                       in_wim_ctype = out_wim_ctype;
-               }
-               input_res_entry = &lte->resource_entry;
-               input_res_ctype = resource_compression_type(
-                                       in_wim_ctype, 
-                                       input_res_entry->flags);
-
-               ret = transfer_file_resource(in, 
-                                       input_res_entry->size,
-                                       input_res_entry->original_size, 
-                                       input_res_entry->offset,
-                                       input_res_ctype, 
-                                       out, 
-                                       out_wim_ctype,
-                                       output_res_entry);
+       ret = write_wim_resource(lte, w->out_fd,
+                                wim_resource_compression_type(lte),
+                                &lte->output_resource_entry, 0);
+       if (ret == 0) {
+               lte->out_refcnt = lte->refcnt;
+               lte->part_number = w->hdr.part_number;
        }
        return ret;
 }