* wimlib; if not, see http://www.gnu.org/licenses/.
*/
-#include "wimlib_internal.h"
-#include "dentry.h"
-#include "lookup_table.h"
-#include "buffer_io.h"
-#include "sha1.h"
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#include "wimlib.h"
+#include "wimlib/dentry.h"
+#include "wimlib/endianness.h"
+#include "wimlib/error.h"
+#include "wimlib/file_io.h"
+#include "wimlib/lookup_table.h"
+#include "wimlib/resource.h"
+#include "wimlib/sha1.h"
#ifdef __WIN32__
-# include "win32.h"
+/* for read_win32_file_prefix(), read_win32_encrypted_file_prefix() */
+# include "wimlib/win32.h"
#endif
-#include <errno.h>
-#include <stdarg.h>
-#include <stdlib.h>
-#include <unistd.h>
-#include <fcntl.h>
+#ifdef WITH_NTFS_3G
+/* for read_ntfs_file_prefix() */
+# include "wimlib/ntfs_3g.h"
+#endif
#ifdef HAVE_ALLOCA_H
# include <alloca.h>
#endif
+#include <errno.h>
+#include <fcntl.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <unistd.h>
/*
* Reads all or part of a compressed WIM resource.
* Returns zero on success, nonzero on failure.
*/
static int
-read_compressed_resource(filedes_t in_fd,
+read_compressed_resource(int in_fd,
u64 resource_compressed_size,
u64 resource_uncompressed_size,
u64 resource_offset,
/* Calculate how many chunks the resource consists of in its entirety.
* */
- u64 num_chunks = (resource_uncompressed_size + WIM_CHUNK_SIZE - 1) /
- WIM_CHUNK_SIZE;
+ u64 num_chunks = DIV_ROUND_UP(resource_uncompressed_size, WIM_CHUNK_SIZE);
+
/* As mentioned, the first chunk has no entry in the chunk table. */
u64 num_chunk_entries = num_chunks - 1;
if (end_chunk != num_chunks - 1)
num_needed_chunks++;
+ /* According to M$'s documentation, if the uncompressed size of
+ * the file is greater than 4 GB, the chunk entries are 8-byte
+ * integers. Otherwise, they are 4-byte integers. */
+ u64 chunk_entry_size = (resource_uncompressed_size >
+ (u64)1 << 32) ? 8 : 4;
+
+ /* Size of the full chunk table in the WIM file. */
+ u64 chunk_table_size = chunk_entry_size * num_chunk_entries;
+
/* Allocate the chunk table. It will only contain offsets for the
* chunks that are actually needed for this read. */
u64 *chunk_offsets;
bool chunk_offsets_malloced;
- if (num_needed_chunks < 1000) {
+ if (num_needed_chunks < 1024) {
chunk_offsets = alloca(num_needed_chunks * sizeof(u64));
chunk_offsets_malloced = false;
} else {
if (start_chunk == 0)
chunk_offsets[0] = 0;
- /* According to M$'s documentation, if the uncompressed size of
- * the file is greater than 4 GB, the chunk entries are 8-byte
- * integers. Otherwise, they are 4-byte integers. */
- u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
- 8 : 4;
-
- /* Size of the full chunk table in the WIM file. */
- u64 chunk_table_size = chunk_entry_size * num_chunk_entries;
/* Read the needed chunk offsets from the table in the WIM file. */
u64 file_offset_of_needed_chunk_entries = resource_offset +
start_table_idx * chunk_entry_size;
+ /* Allocate a buffer into which to read the raw chunk entries. */
+ void *chunk_tab_buf;
+ bool chunk_tab_buf_malloced = false;
+
/* Number of bytes we need to read from the chunk table. */
size_t size = num_needed_chunk_entries * chunk_entry_size;
+ if ((u64)size != num_needed_chunk_entries * chunk_entry_size) {
+ ERROR("Compressed read request too large to fit into memory!");
+ ret = WIMLIB_ERR_NOMEM;
+ goto out;
+ }
- /* 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 (size < 4096) {
+ chunk_tab_buf = alloca(size);
+ } else {
+ chunk_tab_buf = malloc(size);
+ if (!chunk_tab_buf) {
+ ERROR("Failed to allocate chunk table buffer of "
+ "size %zu bytes", size);
+ ret = WIMLIB_ERR_NOMEM;
+ goto out;
+ }
+ chunk_tab_buf_malloced = true;
+ }
if (full_pread(in_fd, chunk_tab_buf, size,
file_offset_of_needed_chunk_entries) != size)
chunk_tab_p++;
if (chunk_entry_size == 4) {
- u32 *entries = (u32*)chunk_tab_buf;
+ le32 *entries = (le32*)chunk_tab_buf;
while (num_needed_chunk_entries--)
*chunk_tab_p++ = le32_to_cpu(*entries++);
} else {
- u64 *entries = (u64*)chunk_tab_buf;
+ le64 *entries = (le64*)chunk_tab_buf;
while (num_needed_chunk_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. */
+ /* Done reading the chunk table now. Now calculate the file offset for
+ * the first byte of compressed data we need to read. */
u64 cur_read_offset = resource_offset + chunk_table_size + chunk_offsets[0];
out:
if (chunk_offsets_malloced)
FREE(chunk_offsets);
+ if (chunk_tab_buf_malloced)
+ FREE(chunk_tab_buf);
return ret;
read_error:
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 void *
-get_resource_entry(const void *p, struct resource_entry *entry)
+/* Translates a WIM resource entry from the on-disk format to an in-memory
+ * format. */
+void
+get_resource_entry(const struct resource_entry_disk *disk_entry,
+ struct resource_entry *entry)
{
- u64 size;
- u8 flags;
-
- p = get_u56(p, &size);
- p = get_u8(p, &flags);
- entry->size = size;
- entry->flags = flags;
+ /* Note: disk_entry may not be 8 byte aligned--- in that case, the
+ * offset and original_size members will be unaligned. (This should be
+ * okay since `struct resource_entry_disk' is declared as packed.) */
+
+ /* Read the size and flags into a bitfield portably... */
+ entry->size = (((u64)disk_entry->size[0] << 0) |
+ ((u64)disk_entry->size[1] << 8) |
+ ((u64)disk_entry->size[2] << 16) |
+ ((u64)disk_entry->size[3] << 24) |
+ ((u64)disk_entry->size[4] << 32) |
+ ((u64)disk_entry->size[5] << 40) |
+ ((u64)disk_entry->size[6] << 48));
+ entry->flags = disk_entry->flags;
+ entry->offset = le64_to_cpu(disk_entry->offset);
+ entry->original_size = le64_to_cpu(disk_entry->original_size);
/* 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. */
-void *
-put_resource_entry(void *p, const struct resource_entry *entry)
+/* Translates a WIM resource entry from an in-memory format into the on-disk
+ * format. */
+void
+put_resource_entry(const struct resource_entry *entry,
+ struct resource_entry_disk *disk_entry)
{
- p = put_u56(p, entry->size);
- p = put_u8(p, entry->flags);
- p = put_u64(p, entry->offset);
- p = put_u64(p, entry->original_size);
- return p;
+ /* Note: disk_entry may not be 8 byte aligned--- in that case, the
+ * offset and original_size members will be unaligned. (This should be
+ * okay since `struct resource_entry_disk' is declared as packed.) */
+ u64 size = entry->size;
+
+ disk_entry->size[0] = size >> 0;
+ disk_entry->size[1] = size >> 8;
+ disk_entry->size[2] = size >> 16;
+ disk_entry->size[3] = size >> 24;
+ disk_entry->size[4] = size >> 32;
+ disk_entry->size[5] = size >> 40;
+ disk_entry->size[6] = size >> 48;
+ disk_entry->flags = entry->flags;
+ disk_entry->offset = cpu_to_le64(entry->offset);
+ disk_entry->original_size = cpu_to_le64(entry->original_size);
}
static int
u64 offset)
{
WIMStruct *wim;
- filedes_t in_fd;
+ int in_fd;
int ret;
wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
* size until the resource is exhausted.
*
* If the resource is located in a WIM file, @flags can be:
- * * WIMLIB_RESOURCE_FLAG_THREADSAFE_READ if it must be safe to access the resource
- * concurrently by multiple threads.
* * WIMLIB_RESOURCE_FLAG_RAW if the raw compressed data is to be supplied
* instead of the uncompressed data.
* Otherwise, the @flags are ignored.
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
+ if (wimlib_print_errors) {
+ 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!");
+ }
ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
}
}