#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_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.
+ * @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
*
* 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,
+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[])
{
* 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. */
/* According to M$'s documentation, if the uncompressed size of
* the file is greater than 4 GB, the chunk entries are 8-byte
* integers. Otherwise, they are 4-byte integers. */
- u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
+ u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ?
8 : 4;
/* Size of the full chunk table in the WIM file. */
/* Number of entries we need to actually read from the chunk
* table (excludes the implicit first chunk). */
- u64 num_needed_chunk_entries = (start_chunk == 0) ?
+ u64 num_needed_chunk_entries = (start_chunk == 0) ?
num_needed_chunks - 1 : num_needed_chunks;
/* Skip over unneeded chunk table entries. */
- u64 file_offset_of_needed_chunk_entries = resource_offset +
+ u64 file_offset_of_needed_chunk_entries = resource_offset +
start_table_idx * chunk_entry_size;
if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) {
ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
/* Done with the chunk table now. We must now seek to the first chunk
* that is needed for the read. */
- u64 file_offset_of_first_needed_chunk = resource_offset +
+ u64 file_offset_of_first_needed_chunk = resource_offset +
chunk_table_size + chunk_offsets[0];
if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) {
ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
* expand to WIM_CHUNK_SIZE uncompressed, and the amount
* of compressed data for the chunk is given by the
* difference of offsets in the chunk offset table. */
- compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
+ compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] -
chunk_offsets[i - start_chunk];
uncompressed_chunk_size = WIM_CHUNK_SIZE;
} else {
* bytes in the file resource, and the last uncompressed
* chunk has size equal to however many bytes are left-
* that is, the remainder of the uncompressed size when
- * divided by WIM_CHUNK_SIZE.
+ * divided by WIM_CHUNK_SIZE.
*
* Note that the resource_compressed_size includes the
* chunk table, so the size of it must be subtracted. */
- compressed_chunk_size = resource_compressed_size -
+ compressed_chunk_size = resource_compressed_size -
chunk_table_size -
chunk_offsets[i - start_chunk];
- uncompressed_chunk_size = resource_uncompressed_size %
+ uncompressed_chunk_size = resource_uncompressed_size %
WIM_CHUNK_SIZE;
/* If the remainder is 0, the last chunk actually
end_offset = WIM_CHUNK_SIZE - 1;
u64 partial_chunk_size = end_offset + 1 - start_offset;
- bool is_partial_chunk = (partial_chunk_size !=
+ bool is_partial_chunk = (partial_chunk_size !=
uncompressed_chunk_size);
DEBUG2("start_offset = %u, end_offset = %u", start_offset,
return WIMLIB_ERR_READ;
}
}
- if (fread(out_p, 1, partial_chunk_size, fp) !=
+ if (fread(out_p, 1, partial_chunk_size, fp) !=
partial_chunk_size)
goto err;
} else {
int ret;
/* Read the compressed data into compressed_buf. */
- if (fread(compressed_buf, 1, compressed_chunk_size,
+ if (fread(compressed_buf, 1, compressed_chunk_size,
fp) != compressed_chunk_size)
goto err;
ret = decompress(compressed_buf,
compressed_chunk_size,
- uncompressed_buf,
+ uncompressed_buf,
uncompressed_chunk_size);
if (ret != 0)
return WIMLIB_ERR_DECOMPRESSION;
return WIMLIB_ERR_READ;
}
-/*
+/*
* Reads uncompressed data from an open file stream.
*/
int read_uncompressed_resource(FILE *fp, u64 offset, u64 len,
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;
}
int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[],
size_t size, u64 offset, bool raw)
{
+ int ctype;
+ int ret = 0;
+ FILE *fp;
+
/* We shouldn't be allowing read over-runs in any part of the library.
* */
if (raw)
else
wimlib_assert(offset + size <= lte->resource_entry.original_size);
- int ctype;
- int ret;
- FILE *fp;
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);
- wimlib_assert(lte->wim->fp);
+ wimlib_assert(lte->wim != NULL);
+ wimlib_assert(lte->wim->fp != NULL);
ctype = wim_resource_compression_type(lte);
wimlib_assert(ctype != WIM_COMPRESSION_TYPE_NONE ||
lte->resource_entry.size));
if (raw || ctype == WIM_COMPRESSION_TYPE_NONE)
- return read_uncompressed_resource(lte->wim->fp,
- lte->resource_entry.offset + offset,
- size, buf);
+ ret = read_uncompressed_resource(lte->wim->fp,
+ lte->resource_entry.offset + offset,
+ size, buf);
else
- return read_compressed_resource(lte->wim->fp,
- lte->resource_entry.size,
- lte->resource_entry.original_size,
- lte->resource_entry.offset,
- ctype, size, offset, buf);
+ ret = read_compressed_resource(lte->wim->fp,
+ lte->resource_entry.size,
+ lte->resource_entry.original_size,
+ lte->resource_entry.offset,
+ ctype, size, offset, buf);
break;
case RESOURCE_IN_STAGING_FILE:
case RESOURCE_IN_FILE_ON_DISK:
if (!fp) {
ERROR_WITH_ERRNO("Failed to open the file "
"`%s'", lte->file_on_disk);
- return WIMLIB_ERR_OPEN;
+ ret = WIMLIB_ERR_OPEN;
+ break;
}
}
ret = read_uncompressed_resource(fp, offset, size, buf);
if (fp != lte->file_on_disk_fp)
fclose(fp);
- return ret;
break;
case RESOURCE_IN_ATTACHED_BUFFER:
/* The resource is directly attached uncompressed in an
* in-memory buffer. */
- wimlib_assert(lte->attached_buffer);
+ wimlib_assert(lte->attached_buffer != NULL);
memcpy(buf, lte->attached_buffer + offset, size);
- return 0;
break;
#ifdef WITH_NTFS_3G
case RESOURCE_IN_NTFS_VOLUME:
- wimlib_assert(lte->ntfs_loc);
- if (lte->attr) {
+ wimlib_assert(lte->ntfs_loc != NULL);
+ wimlib_assert(lte->attr != NULL);
+ {
u64 adjusted_offset;
if (lte->ntfs_loc->is_reparse_point)
adjusted_offset = offset + 8;
else
adjusted_offset = offset;
- if (ntfs_attr_pread(lte->attr, offset, size, buf) == size) {
- return 0;
- } else {
+ if (ntfs_attr_pread(lte->attr, offset, size, buf) != size) {
ERROR_WITH_ERRNO("Error reading NTFS attribute "
"at `%s'",
lte->ntfs_loc->path_utf8);
- return WIMLIB_ERR_NTFS_3G;
+ ret = WIMLIB_ERR_NTFS_3G;
}
- } else {
- wimlib_assert(0);
+ break;
}
- break;
#endif
default:
- assert(0);
+ wimlib_assert(0);
+ ret = -1;
+ break;
}
+ return ret;
}
-/*
+/*
* Reads all the data from the resource corresponding to a WIM lookup table
* entry.
*
u64 offsets[0];
};
-/*
+/*
* Allocates and initializes a chunk table, and reserves space for it in the
* output file.
*/
struct chunk_table *chunk_tab = CALLOC(1, alloc_size);
int ret;
-
if (!chunk_tab) {
ERROR("Failed to allocate chunk table for %"PRIu64" byte "
"resource", size);
return ret;
}
-/*
+/*
* Compresses a chunk of a WIM resource.
*
* @chunk: Uncompressed data of the chunk.
* @chunk: Uncompressed data of the chunk.
* @chunk_size: Size of the chunk (<= WIM_CHUNK_SIZE)
* @out_fp: FILE * to write tho chunk to.
- * @out_ctype: Compression type to use when writing the chunk (ignored if no
+ * @out_ctype: Compression type to use when writing the chunk (ignored if no
* chunk table provided)
* @chunk_tab: Pointer to chunk table being created. It is updated with the
* offset of the chunk we write.
*chunk_tab->cur_offset_p++ = chunk_tab->cur_offset;
chunk_tab->cur_offset += out_chunk_size;
}
-
+
if (fwrite(out_chunk, 1, out_chunk_size, out_fp) != out_chunk_size) {
ERROR_WITH_ERRNO("Failed to write WIM resource chunk");
return WIMLIB_ERR_WRITE;
return 0;
}
-/*
+/*
* Finishes a WIM chunk tale and writes it to the output file at the correct
* offset.
*
* the same as the compression type of the WIM resource we
* need to read, we simply copy the data (i.e. we do not
* uncompress it, then compress it again).
- * @out_res_entry: If non-NULL, a resource entry that is filled in with the
+ * @out_res_entry: If non-NULL, a resource entry that is filled in with the
* offset, original size, compressed size, and compression flag
* of the output resource.
*
u64 original_size;
u64 old_compressed_size;
u64 new_compressed_size;
- u64 offset = 0;
- int ret = 0;
+ u64 offset;
+ int ret;
struct chunk_table *chunk_tab = NULL;
bool raw;
off_t file_offset;
"stream");
return WIMLIB_ERR_WRITE;
}
-
+
/* Are the compression types the same? If so, do a raw copy (copy
* without decompressing and recompressing the data). */
raw = (wim_resource_compression_type(lte) == out_ctype
/* While there are still bytes remaining in the WIM resource, read a
* chunk of the resource, update SHA1, then write that chunk using the
* desired compression type. */
+ offset = 0;
do {
u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE);
ret = read_wim_resource(lte, buf, to_read, offset, raw);
/* Raw copy: The new compressed size is the same as the old compressed
* size
- *
+ *
* Using WIM_COMPRESSION_TYPE_NONE: The new compressed size is the
* original size
*
if (ftruncate(fileno(out_fp), file_offset + out_res_entry->size) != 0) {
ERROR_WITH_ERRNO("Failed to truncate output WIM file");
ret = WIMLIB_ERR_WRITE;
+ goto out_fclose;
+ }
+ } else {
+ if (out_res_entry) {
+ out_res_entry->size = new_compressed_size;
+ out_res_entry->original_size = original_size;
+ out_res_entry->offset = file_offset;
+ out_res_entry->flags = lte->resource_entry.flags
+ & ~WIM_RESHDR_FLAG_COMPRESSED;
+ if (out_ctype != WIM_COMPRESSION_TYPE_NONE)
+ out_res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
}
- goto out_fclose;
- }
- wimlib_assert(new_compressed_size <= original_size || raw);
- if (out_res_entry) {
- out_res_entry->size = new_compressed_size;
- out_res_entry->original_size = original_size;
- out_res_entry->offset = file_offset;
- out_res_entry->flags = lte->resource_entry.flags
- & ~WIM_RESHDR_FLAG_COMPRESSED;
- if (out_ctype != WIM_COMPRESSION_TYPE_NONE)
- out_res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED;
}
+ ret = 0;
out_fclose:
if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK
- && lte->file_on_disk_fp) {
+ && lte->file_on_disk_fp) {
fclose(lte->file_on_disk_fp);
lte->file_on_disk_fp = NULL;
}
if (lte->attr) {
ntfs_attr_close(lte->attr);
lte->attr = NULL;
- } if (ni) {
- ntfs_inode_close(ni);
}
+ if (ni)
+ ntfs_inode_close(ni);
}
#endif
out:
struct resource_entry *out_res_entry,
u8 hash[SHA1_HASH_SIZE])
{
- /* Set up a temporary lookup table entry that we provide to
+ /* Set up a temporary lookup table entry to provide to
* write_wim_resource(). */
struct lookup_table_entry lte;
int ret;
return 0;
}
-/*
+/*
* Extracts the first @size bytes of the WIM resource specified by @lte to the
* open file descriptor @fd.
- *
+ *
* Returns 0 on success; nonzero on failure.
*/
int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd,
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)
return extract_wim_resource_to_fd(lte, fd, wim_resource_size(lte));
}
-/*
+/*
* 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.
return 0;
ret = write_wim_resource(lte, w->out_fp,
- wim_resource_compression_type(lte),
+ wim_resource_compression_type(lte),
<e->output_resource_entry);
if (ret != 0)
return ret;
return 0;
}
-/*
+/*
* Writes a dentry's resources, including the main file resource as well as all
- * alternate data streams, to the output file.
+ * alternate data streams, to the output file.
*
* @dentry: The dentry for the file.
* @wim_p: A pointer to the WIMStruct containing @dentry.
*
- * @return zero on success, nonzero on failure.
+ * @return zero on success, nonzero on failure.
*/
int write_dentry_resources(struct dentry *dentry, void *wim_p)
{
return ret;
}
-/*
+/*
* Reads the metadata metadata resource from the WIM file. The metadata
* resource consists of the security data, followed by the directory entry for
* the root directory, followed by all the other directory entries in the
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;
+ }
+
/* Allocate memory for the uncompressed metadata resource. */
buf = MALLOC(metadata_len);
* 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;
- get_u32(buf, &dentry_offset);
- if (dentry_offset == 0)
- dentry_offset = 8;
- dentry_offset = (dentry_offset + 7) & ~7;
+ 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;
+ }
/* Allocate memory for the root dentry and read it into memory */
dentry = MALLOC(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 pointers correctly. */
struct lookup_table_entry *lte;
u64 metadata_original_size;
const struct wim_security_data *sd;
- const unsigned random_tail_len = 20;
DEBUG("Writing metadata resource for image %d", w->current_image);
/* We do not allow the security data pointer to be NULL, although it may
* point to an empty security data with no entries. */
- wimlib_assert(sd);
+ wimlib_assert(root != NULL);
+ wimlib_assert(sd != NULL);
/* Offset of first child of the root dentry. It's equal to:
* - The total length of the security data, rounded to the next 8-byte
calculate_subdir_offsets(root, &subdir_offset);
/* Total length of the metadata resource (uncompressed) */
- metadata_original_size = subdir_offset + random_tail_len;
+ metadata_original_size = subdir_offset;
/* Allocate a buffer to contain the uncompressed metadata resource */
buf = MALLOC(metadata_original_size);
p = write_security_data(sd, buf);
/* Write the dentry tree into the resource buffer */
- DEBUG("Writing dentry tree.");
p = write_dentry_tree(root, p);
- /*
- * Append 20 random bytes to the metadata resource so that we don't have
- * identical metadata resources if we happen to append exactly the same
- * image twice without any changes in timestamps. If this were to
- * happen, it would cause confusion about the number and order of images
- * in the WIM.
- */
- randomize_byte_array(p, random_tail_len);
-
/* We MUST have exactly filled the buffer; otherwise we calculated its
* size incorrectly or wrote the data incorrectly. */
- wimlib_assert(p - buf + random_tail_len == metadata_original_size);
+ wimlib_assert(p - buf == metadata_original_size);
/* Get the lookup table entry for the metadata resource so we can update
* it. */
lte = wim_metadata_lookup_table_entry(w);
+ wimlib_assert(lte != NULL);
+
/* Write the metadata resource to the output WIM using the proper
* compression type. The lookup table entry for the metadata resource
* is updated. */
/* 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);
- /* We do not allow a metadata resource to be referenced multiple times,
- * and the 20 random bytes appended to it should make it extremely
- * likely for each metadata resource to be unique, even if the exact
- * same image is captured. */
wimlib_assert(lte->out_refcnt == 0);
lte->out_refcnt = 1;