/*
* resource.c
*
- * Read uncompressed and compressed metadata and file resources.
+ * Read uncompressed and compressed metadata and file resources from a WIM file.
*/
/*
- * Copyright (C) 2012 Eric Biggers
+ * Copyright (C) 2012, 2013 Eric Biggers
*
* This file is part of wimlib, a library for working with WIM files.
*
* wimlib; if not, see http://www.gnu.org/licenses/.
*/
-#include "config.h"
-
-#include <stdlib.h>
-#include <stdarg.h>
-
-#include "dentry.h"
-
#include "wimlib_internal.h"
+#include "dentry.h"
#include "lookup_table.h"
#include "buffer_io.h"
-#include "lzx.h"
-#include "xpress.h"
#include "sha1.h"
-#include <unistd.h>
-#include <errno.h>
-#ifdef WITH_NTFS_3G
-#include <time.h>
-#include <ntfs-3g/attrib.h>
-#include <ntfs-3g/inode.h>
-#include <ntfs-3g/dir.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
/*
- * 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"",
- resource_compressed_size,
- resource_uncompressed_size,
- resource_offset);
- DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"",
- wimlib_get_compression_type_string(resource_ctype), len, offset);
/* Trivial case */
if (len == 0)
return 0;
int (*decompress)(const void *, unsigned, void *, unsigned);
/* Set the appropriate decompress function. */
if (resource_ctype == WIMLIB_COMPRESSION_TYPE_LZX)
- decompress = lzx_decompress;
+ 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
* 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.
/* Skip over unneeded chunk table entries. */
u64 file_offset_of_needed_chunk_entries = resource_offset +
start_table_idx * chunk_entry_size;
- if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) {
- ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
- "chunk table of compressed resource",
- file_offset_of_needed_chunk_entries);
- return WIMLIB_ERR_READ;
- }
/* 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. */
/* 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_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
- "first chunk of compressed resource",
- file_offset_of_first_needed_chunk);
- return WIMLIB_ERR_READ;
- }
+ 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
* 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").",
- i, start_chunk, end_chunk);
-
/* Calculate the sizes of the compressed chunk and of the
* uncompressed chunk. */
unsigned compressed_chunk_size;
uncompressed_chunk_size = WIM_CHUNK_SIZE;
}
- DEBUG2("compressed_chunk_size = %u, "
- "uncompressed_chunk_size = %u",
- 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", start_offset,
- end_offset);
- DEBUG2("partial_chunk_size = %u", 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_WITH_ERRNO("Uncompressed partial "
- "chunk fseek() error");
- 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 {
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");
- else
- ERROR_WITH_ERRNO("Error reading compressed file resource");
- 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")",
- offset, len);
- return WIMLIB_ERR_READ;
- }
- if (fread(contents_ret, 1, len, fp) != len) {
- if (feof(fp)) {
- ERROR("Unexpected EOF in uncompressed file resource");
+ 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,
- len, offset);
+ /* 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 WIMLIB_ERR_READ;
+ cur_read_offset += compressed_chunk_size;
}
- return 0;
-}
-
+ ret = 0;
+out:
+ if (chunk_offsets_malloced)
+ FREE(chunk_offsets);
+ return ret;
+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;
/* 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;
}
-#ifdef WITH_FUSE
-static FILE *wim_get_fp(WIMStruct *w)
+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)
{
- pthread_mutex_lock(&w->fp_tab_mutex);
- FILE *fp;
-
- wimlib_assert(w->filename != NULL);
+ WIMStruct *wim;
+ int in_fd;
+ int ret;
- 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;
+ 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;
+ }
+ } else {
+ /* Send data directly to a buffer */
+ if (full_pread(in_fd, ctx_or_buf, size, offset) != size)
+ goto read_error;
}
+ ret = 0;
}
- DEBUG("Opening extra file descriptor to `%s'", w->filename);
- fp = fopen(w->filename, "rb");
- if (!fp)
- ERROR_WITH_ERRNO("Failed to open `%s'", w->filename);
+ goto out;
+read_error:
+ ERROR_WITH_ERRNO("Error reading data from WIM");
+ ret = WIMLIB_ERR_READ;
out:
- pthread_mutex_unlock(&w->fp_tab_mutex);
- return fp;
+ if (ret) {
+ if (errno == 0)
+ errno = EIO;
+ }
+ return ret;
+}
+
+
+int
+read_partial_wim_resource_into_buf(const struct wim_lookup_table_entry *lte,
+ size_t size, u64 offset, void *buf)
+{
+ return read_partial_wim_resource(lte, size, NULL, buf, 0, offset);
}
-static int wim_release_fp(WIMStruct *w, FILE *fp)
+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 = 0;
- FILE **fp_tab;
+ return read_partial_wim_resource(lte, size, cb, ctx_or_buf, flags, 0);
+}
- 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;
+#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__ */
+
+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;
- fp_tab = REALLOC(w->fp_tab, sizeof(FILE*) * (w->num_allocated_fps + 4));
- if (!fp_tab) {
- ret = WIMLIB_ERR_NOMEM;
- goto out;
+ 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);
}
- 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:
- pthread_mutex_unlock(&w->fp_tab_mutex);
- return ret;
+ return 0;
}
-#endif
+
+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);
/*
- * Reads some data from the resource corresponding to a WIM lookup table entry.
+ * 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.
*
- * @lte: The WIM lookup table entry for the resource.
- * @buf: Buffer into which to write the data.
- * @size: Number of bytes to read.
- * @offset: Offset at which to start reading the resource.
+ * 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).
*
- * Returns zero on success, nonzero on failure.
+ * 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.
*/
-int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[],
- size_t size, u64 offset, int flags)
+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 ctype;
- int ret = 0;
- FILE *fp;
+ 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);
+}
- /* We shouldn't be allowing read over-runs in any part of the library.
- * */
- if (flags & WIMLIB_RESOURCE_FLAG_RAW)
- wimlib_assert(offset + size <= lte->resource_entry.size);
- else
- wimlib_assert(offset + size <= lte->resource_entry.original_size);
-
- switch (lte->resource_location) {
- case RESOURCE_IN_WIM:
- /* The resource is in a WIM file, and its WIMStruct is given by
- * the lte->wim member. The resource may be either compressed
- * or uncompressed. */
- wimlib_assert(lte->wim != NULL);
-
- #ifdef WITH_FUSE
- if (flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED) {
- fp = wim_get_fp(lte->wim);
- if (!fp)
- return WIMLIB_ERR_OPEN;
- } else
- #endif
- {
- wimlib_assert(!(flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED));
- wimlib_assert(lte->wim->fp != NULL);
- fp = lte->wim->fp;
- }
+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);
+}
- ctype = wim_resource_compression_type(lte);
+struct extract_ctx {
+ SHA_CTX sha_ctx;
+ consume_data_callback_t extract_chunk;
+ void *extract_chunk_arg;
+};
- wimlib_assert(ctype != WIMLIB_COMPRESSION_TYPE_NONE ||
- (lte->resource_entry.original_size ==
- lte->resource_entry.size));
+static int
+extract_chunk_sha1_wrapper(const void *chunk, size_t chunk_size,
+ void *_ctx)
+{
+ struct extract_ctx *ctx = _ctx;
- if ((flags & WIMLIB_RESOURCE_FLAG_RAW)
- || ctype == WIMLIB_COMPRESSION_TYPE_NONE)
- ret = read_uncompressed_resource(fp,
- lte->resource_entry.offset + offset,
- size, buf);
- else
- ret = read_compressed_resource(fp,
- lte->resource_entry.size,
- lte->resource_entry.original_size,
- lte->resource_entry.offset,
- ctype, size, offset, buf);
- #ifdef WITH_FUSE
- if (flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED) {
- int ret2 = wim_release_fp(lte->wim, fp);
- if (ret == 0)
- ret = ret2;
- }
- #endif
- break;
- case RESOURCE_IN_STAGING_FILE:
- case RESOURCE_IN_FILE_ON_DISK:
- /* The resource is in some file on the external filesystem and
- * needs to be read uncompressed */
- wimlib_assert(lte->file_on_disk);
- wimlib_assert(<e->file_on_disk == <e->staging_file_name);
- /* Use existing file pointer if available; otherwise open one
- * temporarily */
- if (lte->file_on_disk_fp) {
- fp = lte->file_on_disk_fp;
- } else {
- fp = fopen(lte->file_on_disk, "rb");
- if (!fp) {
- ERROR_WITH_ERRNO("Failed to open the file "
- "`%s'", lte->file_on_disk);
- ret = WIMLIB_ERR_OPEN;
- break;
+ 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("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;
}
}
- ret = read_uncompressed_resource(fp, offset, size, buf);
- if (fp != lte->file_on_disk_fp)
- fclose(fp);
- break;
- case RESOURCE_IN_ATTACHED_BUFFER:
- /* The resource is directly attached uncompressed in an
- * in-memory buffer. */
- wimlib_assert(lte->attached_buffer != NULL);
- memcpy(buf, lte->attached_buffer + offset, size);
- break;
-#ifdef WITH_NTFS_3G
- case RESOURCE_IN_NTFS_VOLUME:
- wimlib_assert(lte->ntfs_loc != NULL);
- wimlib_assert(lte->attr != NULL);
- if (lte->ntfs_loc->is_reparse_point)
- offset += 8;
- if (ntfs_attr_pread(lte->attr, offset, size, buf) != size) {
- ERROR_WITH_ERRNO("Error reading NTFS attribute "
- "at `%s'",
- lte->ntfs_loc->path_utf8);
- ret = WIMLIB_ERR_NTFS_3G;
- }
- break;
-#endif
- default:
- wimlib_assert(0);
- ret = -1;
- break;
+ } else {
+ /* Don't do SHA1 */
+ ret = read_resource_prefix(lte, size, extract_chunk,
+ extract_chunk_arg, 0);
}
return ret;
}
-/*
- * Reads all the data from the resource corresponding to a WIM lookup table
- * entry.
- *
- * @lte: The WIM lookup table entry for the resource.
- * @buf: Buffer into which to write the data. It must be at least
- * wim_resource_size(lte) bytes long.
- *
- * Returns 0 on success; nonzero on failure.
- */
-int read_full_wim_resource(const struct lookup_table_entry *lte, u8 buf[],
- int flags)
+static int
+extract_wim_chunk_to_fd(const void *buf, size_t len, void *_fd_p)
{
- return read_wim_resource(lte, buf, wim_resource_size(lte), 0, flags);
+ 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;
+ }
}
-/* Like write_wim_resource(), but the resource is specified by a buffer of
- * uncompressed data rather a lookup table entry; also writes the SHA1 hash of
- * the buffer to @hash. */
-static int write_wim_resource_from_buffer(const u8 *buf, u64 buf_size,
- FILE *out_fp, int out_ctype,
- struct resource_entry *out_res_entry,
- u8 hash[SHA1_HASH_SIZE])
+int
+extract_wim_resource_to_fd(const struct wim_lookup_table_entry *lte,
+ int fd, u64 size)
{
- /* Set up a temporary lookup table entry to provide to
- * write_wim_resource(). */
- struct lookup_table_entry lte;
- int ret;
- lte.resource_entry.flags = 0;
- lte.resource_entry.original_size = buf_size;
- lte.resource_entry.size = buf_size;
- lte.resource_entry.offset = 0;
- lte.resource_location = RESOURCE_IN_ATTACHED_BUFFER;
- lte.attached_buffer = (u8*)buf;
-
- zero_out_hash(lte.hash);
- ret = write_wim_resource(<e, out_fp, out_ctype, out_res_entry, 0);
- if (ret != 0)
- return ret;
- copy_hash(hash, lte.hash);
- return 0;
+ return extract_wim_resource(lte, size, extract_wim_chunk_to_fd, &fd);
}
-/*
- * Extracts the first @size bytes of the WIM resource specified by @lte to the
- * open file descriptor @fd.
- *
- * Returns 0 on success; nonzero on failure.
- */
-int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd,
- u64 size)
+
+static int
+sha1_chunk(const void *buf, size_t len, void *ctx)
{
- u64 bytes_remaining = size;
- u8 buf[min(WIM_CHUNK_SIZE, bytes_remaining)];
- u64 offset = 0;
- int ret = 0;
- u8 hash[SHA1_HASH_SIZE];
-
- SHA_CTX ctx;
- sha1_init(&ctx);
-
- while (bytes_remaining) {
- u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE);
- ret = read_wim_resource(lte, buf, to_read, offset, 0);
- if (ret != 0)
- break;
- sha1_update(&ctx, buf, to_read);
- if (full_write(fd, buf, to_read) < to_read) {
- ERROR_WITH_ERRNO("Error extracting WIM resource");
- return WIMLIB_ERR_WRITE;
- }
- bytes_remaining -= to_read;
- offset += to_read;
- }
- sha1_final(hash, &ctx);
- if (!hashes_equal(hash, lte->hash)) {
- ERROR("Invalid checksum on a WIM resource "
- "(detected when extracting to external file)");
- ERROR("The following WIM resource is invalid:");
- print_lookup_table_entry(lte);
- return WIMLIB_ERR_INVALID_RESOURCE_HASH;
- }
+ sha1_update(ctx, buf, len);
return 0;
}
-/*
- * Extracts the WIM resource specified by @lte to the open file descriptor @fd.
- *
- * Returns 0 on success; nonzero on failure.
- */
-int extract_full_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd)
+/* Calculate the SHA1 message digest of a stream. */
+int
+sha1_resource(struct wim_lookup_table_entry *lte)
{
- return extract_wim_resource_to_fd(lte, fd, wim_resource_size(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;
}
/*
*
* 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 lookup_table_entry *lte, void *wim)
+int
+copy_resource(struct wim_lookup_table_entry *lte, void *wim)
{
WIMStruct *w = wim;
int ret;
- if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) &&
- !w->write_metadata)
- return 0;
-
- ret = write_wim_resource(lte, w->out_fp,
+ ret = write_wim_resource(lte, w->out_fd,
wim_resource_compression_type(lte),
<e->output_resource_entry, 0);
- if (ret != 0)
- return ret;
- lte->out_refcnt = lte->refcnt;
- lte->part_number = w->hdr.part_number;
- return 0;
-}
-
-/*
- * 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.
- * @wim_ctype: The compression type of the WIM file.
- * @imd: Pointer to the image metadata structure. Its `metadata_lte'
- * member specifies the lookup table entry for the metadata
- * resource. The rest of the image metadata entry will be filled
- * in by this function.
- *
- * @return: Zero on success, nonzero on failure.
- */
-int read_metadata_resource(WIMStruct *w, struct image_metadata *imd)
-{
- u8 *buf;
- u32 dentry_offset;
- int ret;
- struct dentry *dentry;
- struct inode_table inode_tab;
- const struct lookup_table_entry *metadata_lte;
- u64 metadata_len;
- struct hlist_head inode_list;
-
- metadata_lte = imd->metadata_lte;
- metadata_len = wim_resource_size(metadata_lte);
-
- DEBUG("Reading metadata resource: length = %"PRIu64", "
- "offset = %"PRIu64"", metadata_len,
- metadata_lte->resource_entry.offset);
-
- /* There is no way the metadata resource could possibly be less than (8
- * + WIM_DENTRY_DISK_SIZE) bytes, where the 8 is for security data (with
- * no security descriptors) and WIM_DENTRY_DISK_SIZE is for the root
- * dentry. */
- if (metadata_len < 8 + WIM_DENTRY_DISK_SIZE) {
- ERROR("Expected at least %u bytes for the metadata resource",
- 8 + WIM_DENTRY_DISK_SIZE);
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
- }
-
- if (sizeof(size_t) < 8 && metadata_len > 0xffffffff) {
- ERROR("Metadata resource is too large (%"PRIu64" bytes",
- metadata_len);
- return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
+ if (ret == 0) {
+ lte->out_refcnt = lte->refcnt;
+ lte->part_number = w->hdr.part_number;
}
-
- /* Allocate memory for the uncompressed metadata resource. */
- buf = MALLOC(metadata_len);
-
- if (!buf) {
- ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
- "metadata resource", metadata_len);
- return WIMLIB_ERR_NOMEM;
- }
-
- /* Read the metadata resource into memory. (It may be compressed.) */
- ret = read_full_wim_resource(metadata_lte, buf, 0);
- if (ret != 0)
- goto out_free_buf;
-
- DEBUG("Finished reading metadata resource into memory.");
-
- /* The root directory entry starts after security data, aligned on an
- * 8-byte boundary within the metadata resource.
- *
- * The security data starts with a 4-byte integer giving its total
- * length, so if we round that up to an 8-byte boundary that gives us
- * the offset of the root dentry.
- *
- * Here we read the security data into a wim_security_data structure,
- * and if successful, go ahead and calculate the offset in the metadata
- * resource of the root dentry. */
-
- wimlib_assert(imd->security_data == NULL);
- ret = read_security_data(buf, metadata_len, &imd->security_data);
- if (ret != 0)
- goto out_free_buf;
-
- dentry_offset = (imd->security_data->total_length + 7) & ~7;
-
- if (dentry_offset == 0) {
- ERROR("Integer overflow while reading metadata resource");
- ret = WIMLIB_ERR_INVALID_SECURITY_DATA;
- goto out_free_security_data;
- }
-
- DEBUG("Reading root dentry");
-
- /* Allocate memory for the root dentry and read it into memory */
- dentry = MALLOC(sizeof(struct dentry));
- if (!dentry) {
- ERROR("Failed to allocate %zu bytes for root dentry",
- sizeof(struct dentry));
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_security_data;
- }
-
- ret = read_dentry(buf, metadata_len, dentry_offset, dentry);
-
- /* This is the root dentry, so set its parent to itself. */
- dentry->parent = dentry;
-
- if (ret == 0 && dentry->length == 0) {
- ERROR("Metadata resource cannot begin with end-of-directory entry!");
- ret = WIMLIB_ERR_INVALID_DENTRY;
- }
-
- if (ret != 0) {
- FREE(dentry);
- goto out_free_security_data;
- }
-
- inode_add_dentry(dentry, dentry->d_inode);
-
- /* Now read the entire directory entry tree into memory. */
- DEBUG("Reading dentry tree");
- ret = read_dentry_tree(buf, metadata_len, dentry);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- /* Calculate the full paths in the dentry tree. */
- DEBUG("Calculating dentry full paths");
- ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- /* Build hash table that maps hard link group IDs to dentry sets */
- DEBUG("Building link group table");
- ret = init_inode_table(&inode_tab, 9001);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- for_dentry_in_tree(dentry, inode_table_insert, &inode_tab);
-
- DEBUG("Fixing inconsistencies in the hard link groups");
- ret = fix_inodes(&inode_tab, &inode_list);
- destroy_inode_table(&inode_tab);
- if (ret != 0)
- goto out_free_dentry_tree;
-
- if (!w->all_images_verified) {
- DEBUG("Running miscellaneous verifications on the dentry tree");
- for_lookup_table_entry(w->lookup_table, lte_zero_real_refcnt, NULL);
- ret = for_dentry_in_tree(dentry, verify_dentry, w);
- if (ret != 0)
- goto out_free_dentry_tree;
- }
-
- DEBUG("Done reading image metadata");
-
- imd->root_dentry = dentry;
- imd->inode_list = inode_list;
- goto out_free_buf;
-out_free_dentry_tree:
- free_dentry_tree(dentry, NULL);
-out_free_security_data:
- free_security_data(imd->security_data);
- imd->security_data = NULL;
-out_free_buf:
- FREE(buf);
- return ret;
-}
-
-static void recalculate_security_data_length(struct wim_security_data *sd)
-{
- u32 total_length = sizeof(u64) * sd->num_entries + 2 * sizeof(u32);
- for (u32 i = 0; i < sd->num_entries; i++)
- total_length += sd->sizes[i];
- sd->total_length = total_length;
-}
-
-/* Write the metadata resource for the current WIM image. */
-int write_metadata_resource(WIMStruct *w)
-{
- u8 *buf;
- u8 *p;
- int ret;
- u64 subdir_offset;
- struct dentry *root;
- struct lookup_table_entry *lte;
- u64 metadata_original_size;
- struct wim_security_data *sd;
-
- DEBUG("Writing metadata resource for image %d (offset = %"PRIu64")",
- w->current_image, ftello(w->out_fp));
-
- root = wim_root_dentry(w);
- sd = wim_security_data(w);
-
- /* Offset of first child of the root dentry. It's equal to:
- * - The total length of the security data, rounded to the next 8-byte
- * boundary,
- * - plus the total length of the root dentry,
- * - plus 8 bytes for an end-of-directory entry following the root
- * dentry (shouldn't really be needed, but just in case...)
- */
- recalculate_security_data_length(sd);
- subdir_offset = (((u64)sd->total_length + 7) & ~7) +
- dentry_correct_total_length(root) + 8;
-
- /* Calculate the subdirectory offsets for the entire dentry tree. */
- calculate_subdir_offsets(root, &subdir_offset);
-
- /* Total length of the metadata resource (uncompressed) */
- metadata_original_size = subdir_offset;
-
- /* Allocate a buffer to contain the uncompressed metadata resource */
- buf = MALLOC(metadata_original_size);
- if (!buf) {
- ERROR("Failed to allocate %"PRIu64" bytes for "
- "metadata resource", metadata_original_size);
- return WIMLIB_ERR_NOMEM;
- }
-
- /* Write the security data into the resource buffer */
- p = write_security_data(sd, buf);
-
- /* Write the dentry tree into the resource buffer */
- p = write_dentry_tree(root, p);
-
- /* We MUST have exactly filled the buffer; otherwise we calculated its
- * size incorrectly or wrote the data incorrectly. */
- wimlib_assert(p - buf == metadata_original_size);
-
- /* Get the lookup table entry for the metadata resource so we can update
- * it. */
- lte = w->image_metadata[w->current_image - 1].metadata_lte;
-
- /* Write the metadata resource to the output WIM using the proper
- * compression type. The lookup table entry for the metadata resource
- * is updated. */
- ret = write_wim_resource_from_buffer(buf, metadata_original_size,
- w->out_fp,
- wimlib_get_compression_type(w),
- <e->output_resource_entry,
- lte->hash);
- if (ret != 0)
- goto out;
-
- /* It's very likely the SHA1 message digest of the metadata resource
- * changed, so re-insert the lookup table entry into the lookup table.
- *
- * We do not check for other lookup table entries having the same SHA1
- * message digest. It's possible for 2 absolutely identical images to
- * be added, therefore causing 2 identical metadata resources to be in
- * the WIM. However, in this case, it's expected for 2 separate lookup
- * table entries to be created, even though this doesn't make a whole
- * lot of sense since they will share the same SHA1 message digest.
- * */
- lookup_table_unlink(w->lookup_table, lte);
- lookup_table_insert(w->lookup_table, lte);
- lte->out_refcnt = 1;
-
- /* Make sure that the lookup table entry for this metadata resource is
- * marked with the metadata flag. */
- lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;
-out:
- /* All the data has been written to the new WIM; no need for the buffer
- * anymore */
- FREE(buf);
return ret;
}