*/
-/* Decompress the specified chunk that uses the specified compression type
- * @ctype, part of a WIM with default chunk size @wim_chunk_size. For LZX the
- * separate @wim_chunk_size is needed because it determines the window size used
- * for LZX compression. */
-static int
-decompress(const void *cchunk, unsigned clen, void *uchunk, unsigned ulen,
- int ctype, u32 wim_chunk_size)
-{
- switch (ctype) {
- case WIMLIB_COMPRESSION_TYPE_LZX:
- return wimlib_lzx_decompress2(cchunk, clen,
- uchunk, ulen, wim_chunk_size);
- case WIMLIB_COMPRESSION_TYPE_XPRESS:
- return wimlib_xpress_decompress(cchunk, clen,
- uchunk, ulen);
- case WIMLIB_COMPRESSION_TYPE_LZMS:
- return wimlib_lzms_decompress(cchunk, clen, uchunk, ulen);
- default:
- ERROR("Invalid compression format (%d)", ctype);
- return -1;
- }
-}
-
struct data_range {
u64 offset;
u64 size;
};
-/* Alternate chunk table format for resources with
- * WIM_RESHDR_FLAG_PACKED_STREAMS set. */
-struct alt_chunk_table_header_disk {
- /* Uncompressed size of the resource in bytes. */
- le64 res_usize;
-
- /* Number of bytes each compressed chunk decompresses into, except
- * possibly the last which decompresses into the remainder. */
- le32 chunk_size;
-
- /* Compression format used for compressed chunks:
- * 0 = None
- * 1 = LZX
- * 2 = XPRESS
- * 3 = LZMS */
- le32 compression_format;
-
- /* This header is directly followed by a table of compressed sizes of
- * the chunks. */
-} _packed_attribute;
-
/*
* read_compressed_wim_resource() -
*
* Number of ranges in @ranges; must be at least 1.
* @cb
* Callback function to feed the data being read. Each call provides the
- * next chunk of the requested data. Each chunk will be of nonzero size
- * and will not cross range boundaries, but otherwise is of unspecified
- * size.
+ * next chunk of the requested data, uncompressed. Each chunk will be of
+ * nonzero size and will not cross range boundaries, but otherwise will be
+ * of unspecified size.
* @cb_ctx
* Parameter to pass to @cb_ctx.
- * @raw_chunks_mode
- * If %true, this function will provide the raw compressed chunks of the
- * resource rather than the uncompressed data. In this mode, only a single
- * data range can be requested, and it must cover the entire uncompressed
- * resource.
*
* Possible return values:
*
const struct data_range * const ranges,
const size_t num_ranges,
const consume_data_callback_t cb,
- void * const cb_ctx,
- const bool raw_chunks_mode)
+ void * const cb_ctx)
{
int ret;
int errno_save;
bool chunk_offsets_malloced = false;
bool ubuf_malloced = false;
bool cbuf_malloced = false;
+ struct wimlib_decompressor *decompressor = NULL;
/* Sanity checks */
wimlib_assert(rspec != NULL);
for (size_t i = 0; i < num_ranges - 1; i++)
wimlib_assert(ranges[i].offset + ranges[i].size <= ranges[i + 1].offset);
- if (raw_chunks_mode) {
- wimlib_assert(num_ranges == 1);
- wimlib_assert(ranges[0].offset == 0);
- wimlib_assert(ranges[0].size == rspec->uncompressed_size);
- }
-
/* Get the offsets of the first and last bytes of the read. */
const u64 first_offset = ranges[0].offset;
const u64 last_offset = ranges[num_ranges - 1].offset + ranges[num_ranges - 1].size - 1;
/* Get the maximum size of uncompressed chunks in this resource, which
* we require be a power of 2. */
- u32 chunk_size;
u64 cur_read_offset = rspec->offset_in_wim;
- int ctype;
+ int ctype = rspec->compression_type;
+ u32 chunk_size = rspec->chunk_size;
if (alt_chunk_table) {
/* Alternate chunk table format. Its header specifies the chunk
- * size and compression format. */
- struct alt_chunk_table_header_disk hdr;
-
- ret = full_pread(in_fd, &hdr, sizeof(hdr), cur_read_offset);
- if (ret)
- goto read_error;
- cur_read_offset += sizeof(hdr);
-
- chunk_size = le32_to_cpu(hdr.chunk_size);
- ctype = le32_to_cpu(hdr.compression_format);
-
- /* Format numbers must be the same as in WIMGAPI to be
- * compatible. */
- BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
- BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_LZX != 1);
- BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_XPRESS != 2);
- BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_LZMS != 3);
- } else {
- /* "Normal" format: the maximum uncompressed chunk size and the
- * compression format default to those of the WIM itself. */
- chunk_size = rspec->wim->chunk_size;
- ctype = rspec->wim->compression_type;
+ * size and compression format. Note: it could be read here;
+ * however, the relevant data was already loaded into @rspec by
+ * read_wim_lookup_table(). */
+ cur_read_offset += sizeof(struct alt_chunk_table_header_disk);
}
+
if (!is_power_of_2(chunk_size)) {
ERROR("Invalid compressed resource: "
- "expected power-of-2 chunk size (got %u)", chunk_size);
+ "expected power-of-2 chunk size (got %"PRIu32")",
+ chunk_size);
ret = WIMLIB_ERR_INVALID_CHUNK_SIZE;
goto out_free_memory;
}
+ /* Get valid decompressor. */
+ if (ctype == rspec->wim->decompressor_ctype &&
+ chunk_size == rspec->wim->decompressor_max_block_size)
+ {
+ /* Cached decompressor. */
+ decompressor = rspec->wim->decompressor;
+ rspec->wim->decompressor_ctype = WIMLIB_COMPRESSION_TYPE_NONE;
+ rspec->wim->decompressor = NULL;
+ } else {
+ ret = wimlib_create_decompressor(ctype, chunk_size, NULL,
+ &decompressor);
+ if (ret)
+ goto out_free_memory;
+ }
+
const u32 chunk_order = bsr32(chunk_size);
/* Calculate the total number of chunks the resource is divided into. */
const u64 num_chunk_entries = (alt_chunk_table ? num_chunks : num_chunks - 1);
/* Set the size of each chunk table entry based on the resource's
- * uncompressed size. XXX: Does the alternate chunk table really
- * always have 4-byte entries? */
- const u64 chunk_entry_size =
- (rspec->uncompressed_size > (1ULL << 32) && !alt_chunk_table)
- ? 8 : 4;
+ * uncompressed size. */
+ const u64 chunk_entry_size = get_chunk_entry_size(rspec->uncompressed_size,
+ alt_chunk_table);
/* Calculate the size of the chunk table in bytes. */
const u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
chunk_offsets_alloc_size -
chunk_table_size_to_read;
- ret = full_pread(in_fd, chunk_table_data, chunk_table_size,
+ ret = full_pread(in_fd, chunk_table_data, chunk_table_size_to_read,
file_offset_of_needed_chunk_entries);
if (ret)
goto read_error;
ubuf_malloced = true;
}
- /* Unless the raw compressed data was requested, allocate a temporary
- * buffer for reading compressed chunks, each of which can be at most
- * @chunk_size - 1 bytes. This excludes compressed chunks that are a
- * full @chunk_size bytes, which are actually stored uncompressed. */
- if (!raw_chunks_mode) {
- if (chunk_size - 1 <= STACK_MAX) {
- cbuf = alloca(chunk_size - 1);
- } else {
- cbuf = MALLOC(chunk_size - 1);
- if (cbuf == NULL)
- goto oom;
- cbuf_malloced = true;
- }
+ /* Allocate a temporary buffer for reading compressed chunks, each of
+ * which can be at most @chunk_size - 1 bytes. This excludes compressed
+ * chunks that are a full @chunk_size bytes, which are actually stored
+ * uncompressed. */
+ if (chunk_size - 1 <= STACK_MAX) {
+ cbuf = alloca(chunk_size - 1);
+ } else {
+ cbuf = MALLOC(chunk_size - 1);
+ if (cbuf == NULL)
+ goto oom;
+ cbuf_malloced = true;
}
/* Set current data range. */
* function. */
u8 *read_buf;
- if (chunk_csize == chunk_usize || raw_chunks_mode)
+ if (chunk_csize == chunk_usize)
read_buf = ubuf;
else
read_buf = cbuf;
if (read_buf == cbuf) {
DEBUG("Decompressing chunk %"PRIu64" "
- "(csize=%"PRIu64" usize=%"PRIu64"",
+ "(csize=%"PRIu32" usize=%"PRIu32")",
i, chunk_csize, chunk_usize);
- ret = decompress(cbuf,
- chunk_csize,
- ubuf,
- chunk_usize,
- ctype,
- chunk_size);
+ ret = wimlib_decompress(cbuf,
+ chunk_csize,
+ ubuf,
+ chunk_usize,
+ decompressor);
if (ret) {
ERROR("Failed to decompress data!");
ret = WIMLIB_ERR_DECOMPRESSION;
end = min(cur_range_end, chunk_end_offset) - chunk_start_offset;
size = end - start;
- if (raw_chunks_mode)
- ret = (*cb)(&ubuf[0], chunk_csize, cb_ctx);
- else
- ret = (*cb)(&ubuf[start], size, cb_ctx);
+ ret = (*cb)(&ubuf[start], size, cb_ctx);
if (ret)
goto out_free_memory;
goto read_error;
}
ret = 0;
+
out_free_memory:
errno_save = errno;
+ if (decompressor) {
+ wimlib_free_decompressor(rspec->wim->decompressor);
+ rspec->wim->decompressor = decompressor;
+ rspec->wim->decompressor_ctype = ctype;
+ rspec->wim->decompressor_max_block_size = chunk_size;
+ }
if (chunk_offsets_malloced)
FREE(chunk_offsets);
if (ubuf_malloced)
/* Read raw data from a file descriptor at the specified offset, feeding the
* data it in chunks into the specified callback function. */
static int
-read_raw_file_data(struct filedes *in_fd, u64 size,
- consume_data_callback_t cb, void *cb_ctx, u64 offset)
+read_raw_file_data(struct filedes *in_fd, u64 offset, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
u8 buf[BUFFER_SIZE];
size_t bytes_to_read;
* read_partial_wim_resource()-
*
* Read a range of data from an uncompressed or compressed resource in a WIM
- * file. Data is fed chunk-by-chunk into the callback function @cb, passing it
- * the argument @cb_ctx. The chunks are of unspecified size unless the
- * RAW_CHUNKS mode is requested.
- *
- * By default, this function provides the uncompressed data of the resource, and
- * @offset and @size and interpreted relative to the uncompressed contents of
- * the resource. This behavior can be modified by either of the following
- * flags:
+ * file.
*
- * WIMLIB_READ_RESOURCE_FLAG_RAW_FULL:
- * Read @size bytes at @offset of the raw contents of the compressed
- * resource. In the case of pipable resources, this excludes the stream
- * header. Exclusive with WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS.
- *
- * WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS:
- * Read the raw compressed chunks of the compressed resource. @size must
- * be the full uncompressed size, @offset must be 0, and @cb_chunk_size
- * must be the resource chunk size.
+ * @rspec
+ * Specification of the WIM resource to read from.
+ * @offset
+ * Offset within the uncompressed resource at which to start reading.
+ * @size
+ * Number of bytes to read.
+ * @cb
+ * Callback function to feed the data being read. Each call provides the
+ * next chunk of the requested data, uncompressed. Each chunk will be of
+ * nonzero size and will not cross range boundaries, but otherwise will be
+ * of unspecified size.
+ * @cb_ctx
+ * Parameter to pass to @cb_ctx.
*
* Return values:
* WIMLIB_ERR_SUCCESS (0)
*/
static int
read_partial_wim_resource(const struct wim_resource_spec *rspec,
- u64 offset, u64 size, consume_data_callback_t cb,
- void *cb_ctx, int flags)
+ u64 offset, u64 size,
+ consume_data_callback_t cb, void *cb_ctx)
{
/* Sanity checks. */
- if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) {
- wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS));
- wimlib_assert(offset + size >= offset);
- wimlib_assert(offset + size <= rspec->size_in_wim);
- } else if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
- wimlib_assert(offset == 0);
- wimlib_assert(size == rspec->uncompressed_size);
- } else {
- wimlib_assert(offset + size >= offset);
- wimlib_assert(offset + size <= rspec->uncompressed_size);
- }
+ wimlib_assert(offset + size >= offset);
+ wimlib_assert(offset + size <= rspec->uncompressed_size);
DEBUG("Reading %"PRIu64" @ %"PRIu64" from WIM resource "
- "%"PRIu64" => %"PRIu64" @ %"PRIu64" (flags 0x%08x)",
+ "%"PRIu64" => %"PRIu64" @ %"PRIu64,
size, offset, rspec->uncompressed_size,
- rspec->size_in_wim, rspec->offset_in_wim, flags);
+ rspec->size_in_wim, rspec->offset_in_wim);
/* Trivial case. */
if (size == 0)
return 0;
- if ((flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) ||
- !resource_is_compressed(rspec))
- {
- return read_raw_file_data(&rspec->wim->in_fd,
- size,
- cb,
- cb_ctx,
- rspec->offset_in_wim + offset);
- } else {
- bool raw_chunks = (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS);
+ if (resource_is_compressed(rspec)) {
struct data_range range = {
.offset = offset,
.size = size,
};
return read_compressed_wim_resource(rspec, &range, 1,
- cb, cb_ctx, raw_chunks);
+ cb, cb_ctx);
+ } else {
+ return read_raw_file_data(&rspec->wim->in_fd,
+ rspec->offset_in_wim + offset,
+ size,
+ cb,
+ cb_ctx);
}
}
lte->offset_in_res + offset,
size,
bufferer_cb,
- &buf,
- 0);
+ &buf);
}
/* A consume_data_callback_t implementation that simply ignores the data
skip_wim_stream(struct wim_lookup_table_entry *lte)
{
wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
- wimlib_assert(!lte_is_partial(lte));
+ wimlib_assert(!(lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS));
+ DEBUG("Skipping stream (size=%"PRIu64")", lte->size);
return read_partial_wim_resource(lte->rspec,
0,
lte->rspec->uncompressed_size,
skip_chunk_cb,
- NULL,
- WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS);
+ NULL);
}
static int
read_wim_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
- consume_data_callback_t cb, void *cb_ctx, int flags)
+ consume_data_callback_t cb, void *cb_ctx)
{
- return read_partial_wim_resource(lte->rspec,
- lte->offset_in_res,
- size,
- cb,
- cb_ctx,
- flags);
+ return read_partial_wim_resource(lte->rspec, lte->offset_in_res, size,
+ cb, cb_ctx);
}
#ifndef __WIN32__
* encrypted), so Windows uses its own code for its equivalent case. */
static int
read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte, u64 size,
- consume_data_callback_t cb, void *cb_ctx,
- int _ignored_flags)
+ consume_data_callback_t cb, void *cb_ctx)
{
int ret;
int raw_fd;
return WIMLIB_ERR_OPEN;
}
filedes_init(&fd, raw_fd);
- ret = read_raw_file_data(&fd, size, cb, cb_ctx, 0);
+ ret = read_raw_file_data(&fd, 0, size, cb, cb_ctx);
filedes_close(&fd);
return ret;
}
* fact, already located in an in-memory buffer. */
static int
read_buffer_prefix(const struct wim_lookup_table_entry *lte,
- u64 size, consume_data_callback_t cb,
- void *cb_ctx, int _ignored_flags)
+ u64 size, consume_data_callback_t cb, void *cb_ctx)
{
wimlib_assert(size <= lte->size);
return (*cb)(lte->attached_buffer, size, cb_ctx);
typedef int (*read_stream_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
u64 size,
consume_data_callback_t cb,
- void *cb_ctx, int flags);
+ void *cb_ctx);
/*
* read_stream_prefix()-
* This function feeds the data to a callback function @cb in chunks of
* unspecified size.
*
- * If the stream is located in a WIM file, @flags can be set as documented in
- * read_partial_wim_resource(). Otherwise @flags are ignored.
- *
* Returns 0 on success; nonzero on error. A nonzero value will be returned if
* the stream data cannot be successfully read (for a number of different
* reasons, depending on the stream location), or if @cb returned nonzero in
* which case that error code will be returned.
*/
-int
+static int
read_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
- consume_data_callback_t cb, void *cb_ctx, int flags)
+ consume_data_callback_t cb, void *cb_ctx)
{
static const read_stream_prefix_handler_t handlers[] = {
[RESOURCE_IN_WIM] = read_wim_stream_prefix,
};
wimlib_assert(lte->resource_location < ARRAY_LEN(handlers)
&& handlers[lte->resource_location] != NULL);
- return handlers[lte->resource_location](lte, size, cb, cb_ctx, flags);
+ return handlers[lte->resource_location](lte, size, cb, cb_ctx);
}
/* Read the full uncompressed data of the specified stream into the specified
read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *_buf)
{
u8 *buf = _buf;
- return read_stream_prefix(lte, lte->size, bufferer_cb, &buf, 0);
+ return read_stream_prefix(lte, lte->size, bufferer_cb, &buf);
}
/* Retrieve the full uncompressed data of the specified stream. A buffer large
if (lte == NULL)
return WIMLIB_ERR_NOMEM;
- lte->unhashed = 1;
lte_bind_wim_resource_spec(lte, rspec);
lte->flags = rspec->flags;
lte->size = rspec->uncompressed_size;
struct streamifier_context {
struct read_stream_list_callbacks cbs;
struct wim_lookup_table_entry *cur_stream;
+ struct wim_lookup_table_entry *next_stream;
u64 cur_stream_offset;
struct wim_lookup_table_entry *final_stream;
size_t list_head_offset;
};
+static struct wim_lookup_table_entry *
+next_stream(struct wim_lookup_table_entry *lte, size_t list_head_offset)
+{
+ struct list_head *cur;
+
+ cur = (struct list_head*)((u8*)lte + list_head_offset);
+
+ return (struct wim_lookup_table_entry*)((u8*)cur->next - list_head_offset);
+}
+
/* A consume_data_callback_t implementation that translates raw resource data
* into streams, calling the begin_stream, consume_chunk, and end_stream
* callback functions as appropriate. */
if (ctx->cur_stream_offset == ctx->cur_stream->size) {
/* Finished reading all the data for a stream. */
+
+ ctx->cur_stream_offset = 0;
+
DEBUG("End stream (size=%"PRIu64").", ctx->cur_stream->size);
ret = (*ctx->cbs.end_stream)(ctx->cur_stream, 0,
ctx->cbs.end_stream_ctx);
if (ret)
return ret;
- if (ctx->cur_stream != ctx->final_stream) {
- /* Advance to next stream. */
- struct list_head *cur, *next;
-
- cur = (struct list_head *)
- ((u8*)ctx->cur_stream + ctx->list_head_offset);
- next = cur->next;
-
- ctx->cur_stream = (struct wim_lookup_table_entry *)
- ((u8*)next - ctx->list_head_offset);
-
- ctx->cur_stream_offset = 0;
- } else {
- /* No more streams. */
- ctx->cur_stream = NULL;
+ /* Advance to next stream. */
+ ctx->cur_stream = ctx->next_stream;
+ if (ctx->cur_stream != NULL) {
+ if (ctx->cur_stream != ctx->final_stream)
+ ctx->next_stream = next_stream(ctx->cur_stream,
+ ctx->list_head_offset);
+ else
+ ctx->next_stream = NULL;
}
}
return 0;
struct hasher_context {
SHA_CTX sha_ctx;
+ int flags;
struct read_stream_list_callbacks cbs;
};
sha1_final(hash, &ctx->sha_ctx);
if (lte->unhashed) {
- /* No SHA1 message digest was previously present for the stream.
- * Set it to the one just calculated. */
- DEBUG("Set SHA1 message digest for stream (size=%"PRIu64").", lte->size);
- copy_hash(lte->hash, hash);
+ if (ctx->flags & COMPUTE_MISSING_STREAM_HASHES) {
+ /* No SHA1 message digest was previously present for the
+ * stream. Set it to the one just calculated. */
+ DEBUG("Set SHA1 message digest for stream "
+ "(size=%"PRIu64").", lte->size);
+ copy_hash(lte->hash, hash);
+ }
} else {
- /* The stream already had a SHA1 message digest present. Verify
- * that it is the same as the calculated value. */
- if (!hashes_equal(hash, lte->hash)) {
- if (wimlib_print_errors) {
- ERROR("Invalid SHA1 message digest "
- "on the following WIM stream:");
- print_lookup_table_entry(lte, stderr);
- if (lte->resource_location == RESOURCE_IN_WIM)
- ERROR("The WIM file appears to be corrupt!");
+ if (ctx->flags & VERIFY_STREAM_HASHES) {
+ /* The stream already had a SHA1 message digest present. Verify
+ * that it is the same as the calculated value. */
+ if (!hashes_equal(hash, lte->hash)) {
+ if (wimlib_print_errors) {
+ ERROR("Invalid SHA1 message digest "
+ "on the following WIM stream:");
+ print_lookup_table_entry(lte, stderr);
+ }
+ ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
+ errno = EINVAL;
+ goto out_next_cb;
}
- ret = WIMLIB_ERR_INVALID_RESOURCE_HASH;
- errno = EINVAL;
- goto out_next_cb;
+ DEBUG("SHA1 message digest okay for "
+ "stream (size=%"PRIu64").", lte->size);
}
- DEBUG("SHA1 message digest okay for stream (size=%"PRIu64").", lte->size);
}
ret = 0;
out_next_cb:
return (*ctx->cbs.end_stream)(lte, ret, ctx->cbs.end_stream_ctx);
}
+static int
+read_full_stream_with_cbs(struct wim_lookup_table_entry *lte,
+ const struct read_stream_list_callbacks *cbs)
+{
+ int ret;
+
+ ret = (*cbs->begin_stream)(lte, false, cbs->begin_stream_ctx);
+ if (ret)
+ return ret;
+
+ ret = read_stream_prefix(lte, lte->size, cbs->consume_chunk,
+ cbs->consume_chunk_ctx);
+
+ return (*cbs->end_stream)(lte, ret, cbs->end_stream_ctx);
+}
+
/* Read the full data of the specified stream, passing the data into the
* specified callbacks (all of which are optional) and either checking or
* computing the SHA1 message digest of the stream. */
read_full_stream_with_sha1(struct wim_lookup_table_entry *lte,
const struct read_stream_list_callbacks *cbs)
{
- int ret;
-
struct hasher_context hasher_ctx = {
+ .flags = VERIFY_STREAM_HASHES | COMPUTE_MISSING_STREAM_HASHES,
.cbs = *cbs,
};
+ struct read_stream_list_callbacks hasher_cbs = {
+ .begin_stream = hasher_begin_stream,
+ .begin_stream_ctx = &hasher_ctx,
+ .consume_chunk = hasher_consume_chunk,
+ .consume_chunk_ctx = &hasher_ctx,
+ .end_stream = hasher_end_stream,
+ .end_stream_ctx = &hasher_ctx,
- ret = hasher_begin_stream(lte, false, &hasher_ctx);
- if (ret)
- return ret;
-
- ret = read_stream_prefix(lte, lte->size, hasher_consume_chunk,
- &hasher_ctx, 0);
-
- return hasher_end_stream(lte, ret, &hasher_ctx);
+ };
+ return read_full_stream_with_cbs(lte, &hasher_cbs);
}
-struct rechunkifier_context {
- u8 *buffer;
- u32 buffer_filled;
- u32 cb_chunk_size;
-
- const struct data_range *ranges;
- size_t num_ranges;
- size_t cur_range;
- u64 range_bytes_remaining;
-
- consume_data_callback_t cb;
- void *cb_ctx;
-};
-
-/* Wrapper callback for adjusting the data chunk size. */
static int
-rechunkifier_cb(const void *chunk, size_t size, void *_ctx)
+read_packed_streams(struct wim_lookup_table_entry *first_stream,
+ struct wim_lookup_table_entry *last_stream,
+ u64 stream_count,
+ size_t list_head_offset,
+ const struct read_stream_list_callbacks *sink_cbs)
{
- struct rechunkifier_context *ctx = _ctx;
- const u8 *chunkptr = chunk;
- size_t bytes_to_copy;
+ struct data_range *ranges;
+ bool ranges_malloced;
+ struct wim_lookup_table_entry *cur_stream;
+ size_t i;
int ret;
+ u64 ranges_alloc_size;
- wimlib_assert(ctx->cur_range != ctx->num_ranges);
+ DEBUG("Reading %"PRIu64" streams combined in same WIM resource",
+ stream_count);
- while (size) {
+ /* Setup data ranges array (one range per stream to read); this way
+ * read_compressed_wim_resource() does not need to be aware of streams.
+ */
- /* Append more data to the buffer. */
- bytes_to_copy = size;
+ ranges_alloc_size = stream_count * sizeof(ranges[0]);
- if (bytes_to_copy > ctx->cb_chunk_size - ctx->buffer_filled)
- bytes_to_copy = ctx->cb_chunk_size - ctx->buffer_filled;
+ if (unlikely((size_t)ranges_alloc_size != ranges_alloc_size)) {
+ ERROR("Too many streams in one resource!");
+ return WIMLIB_ERR_NOMEM;
+ }
+ if (likely(ranges_alloc_size <= STACK_MAX)) {
+ ranges = alloca(ranges_alloc_size);
+ ranges_malloced = false;
+ } else {
+ ranges = MALLOC(ranges_alloc_size);
+ if (ranges == NULL) {
+ ERROR("Too many streams in one resource!");
+ return WIMLIB_ERR_NOMEM;
+ }
+ ranges_malloced = true;
+ }
- if (bytes_to_copy > ctx->range_bytes_remaining - ctx->buffer_filled)
- bytes_to_copy = ctx->range_bytes_remaining - ctx->buffer_filled;
+ for (i = 0, cur_stream = first_stream;
+ i < stream_count;
+ i++, cur_stream = next_stream(cur_stream, list_head_offset))
+ {
+ ranges[i].offset = cur_stream->offset_in_res;
+ ranges[i].size = cur_stream->size;
+ }
- memcpy(&ctx->buffer[ctx->buffer_filled], chunkptr, bytes_to_copy);
+ struct streamifier_context streamifier_ctx = {
+ .cbs = *sink_cbs,
+ .cur_stream = first_stream,
+ .next_stream = next_stream(first_stream, list_head_offset),
+ .cur_stream_offset = 0,
+ .final_stream = last_stream,
+ .list_head_offset = list_head_offset,
+ };
- ctx->buffer_filled += bytes_to_copy;
- chunkptr += bytes_to_copy;
- size -= bytes_to_copy;
- ctx->range_bytes_remaining -= bytes_to_copy;
+ ret = read_compressed_wim_resource(first_stream->rspec,
+ ranges,
+ stream_count,
+ streamifier_cb,
+ &streamifier_ctx);
- if (ctx->buffer_filled == ctx->cb_chunk_size ||
- ctx->range_bytes_remaining == 0)
- {
- /* Maximum chunk size reached, or current range ended.
- * Call the next consume_data_callback_t and empty the
- * buffer */
- ret = (*ctx->cb)(ctx->buffer, ctx->buffer_filled, ctx->cb_ctx);
- if (ret)
- return ret;
- ctx->buffer_filled = 0;
+ if (ranges_malloced)
+ FREE(ranges);
- if (ctx->range_bytes_remaining == 0 &&
- ++ctx->cur_range != ctx->num_ranges)
- ctx->range_bytes_remaining = ctx->ranges[ctx->cur_range].size;
+ if (ret) {
+ if (streamifier_ctx.cur_stream_offset != 0) {
+ ret = (*streamifier_ctx.cbs.end_stream)
+ (streamifier_ctx.cur_stream,
+ ret,
+ streamifier_ctx.cbs.end_stream_ctx);
}
}
- return 0;
+ return ret;
}
/*
* @list_head_offset
* Offset of the `struct list_head' within each `struct
* wim_lookup_table_entry' that makes up the @stream_list.
- * @cb_chunk_size
- * Size of chunks to provide to @consume_chunk. For a given stream, all
- * the chunks will be this size, except possibly the last which will be the
- * remainder. If @
- * @cbs TODO
+ * @cbs
+ * Callback functions to accept the stream data.
+ * @flags
+ * Bitwise OR of zero or more of the following flags:
+ *
+ * VERIFY_STREAM_HASHES:
+ * For all streams being read that have already had SHA1 message
+ * digests computed, calculate the SHA1 message digest of the read
+ * data and compare it with the previously computed value. If they
+ * do not match, return WIMLIB_ERR_INVALID_RESOURCE_HASH.
+ *
+ * COMPUTE_MISSING_STREAM_HASHES
+ * For all streams being read that have not yet had their SHA1
+ * message digests computed, calculate and save their SHA1 message
+ * digests.
+ *
+ * STREAM_LIST_ALREADY_SORTED
+ * @stream_list is already sorted in sequential order for reading.
+ *
+ * The callback functions are allowed to delete the current stream from the list
+ * if necessary.
*
* Returns 0 on success; a nonzero error code on failure. Failure can occur due
* to an error reading the data or due to an error status being returned by any
int
read_stream_list(struct list_head *stream_list,
size_t list_head_offset,
- u32 cb_chunk_size,
- const struct read_stream_list_callbacks *cbs)
+ const struct read_stream_list_callbacks *cbs,
+ int flags)
{
int ret;
struct list_head *cur, *next;
struct wim_lookup_table_entry *lte;
+ struct hasher_context *hasher_ctx;
+ struct read_stream_list_callbacks *sink_cbs;
- ret = sort_stream_list_by_sequential_order(stream_list, list_head_offset);
- if (ret)
- return ret;
+ if (!(flags & STREAM_LIST_ALREADY_SORTED)) {
+ ret = sort_stream_list_by_sequential_order(stream_list, list_head_offset);
+ if (ret)
+ return ret;
+ }
+
+ if (flags & (VERIFY_STREAM_HASHES | COMPUTE_MISSING_STREAM_HASHES)) {
+ hasher_ctx = alloca(sizeof(*hasher_ctx));
+ *hasher_ctx = (struct hasher_context) {
+ .flags = flags,
+ .cbs = *cbs,
+ };
+ sink_cbs = alloca(sizeof(*sink_cbs));
+ *sink_cbs = (struct read_stream_list_callbacks) {
+ .begin_stream = hasher_begin_stream,
+ .begin_stream_ctx = hasher_ctx,
+ .consume_chunk = hasher_consume_chunk,
+ .consume_chunk_ctx = hasher_ctx,
+ .end_stream = hasher_end_stream,
+ .end_stream_ctx = hasher_ctx,
+ };
+ } else {
+ sink_cbs = (struct read_stream_list_callbacks*)cbs;
+ }
for (cur = stream_list->next, next = cur->next;
cur != stream_list;
{
lte = (struct wim_lookup_table_entry*)((u8*)cur - list_head_offset);
- if (lte_is_partial(lte)) {
+ if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
+ lte->size != lte->rspec->uncompressed_size)
+ {
struct wim_lookup_table_entry *lte_next, *lte_last;
struct list_head *next2;
- size_t stream_count;
+ u64 stream_count;
/* The next stream is a proper sub-sequence of a WIM
* resource. See if there are other streams in the same
* first stream in the resource that needs to be
* read and @lte_last specifies the last stream
* in the resource that needs to be read. */
-
- DEBUG("Reading %zu streams combined in same "
- "WIM resource", stream_count);
-
next = next2;
-
- struct data_range ranges[stream_count];
-
- {
- struct list_head *next3;
- size_t i;
- struct wim_lookup_table_entry *lte_cur;
-
- next3 = cur;
- for (i = 0; i < stream_count; i++) {
- lte_cur = (struct wim_lookup_table_entry*)
- ((u8*)next3 - list_head_offset);
- ranges[i].offset = lte_cur->offset_in_res;
- ranges[i].size = lte_cur->size;
- next3 = next3->next;
- }
- }
-
- /* Set up a chain of callbacks.
- *
- * The first level is the
- * streamifier_cb,
- * which takes in chunks of data and divides
- * them into the constituent streams.
- *
- * The second level are the SHA1 message digest
- * callbacks, which checksum each stream.
- *
- * rechunkifier_cb handles dividing the read
- * data into chunks of maximum size
- * @cb_chunk_size. If @cb_chunk_size is 0, then
- * this callback is not needed.
- *
- * Finally, the last level of callbacks are
- * @cbs, passed as arguments to this function.
- */
-
- struct rechunkifier_context *rechunkifier_ctx = NULL;
- consume_data_callback_t last_cb;
- void *last_cb_ctx;
-
- if (cb_chunk_size != 0) {
- rechunkifier_ctx = alloca(sizeof(*rechunkifier_ctx));
- *rechunkifier_ctx = (struct rechunkifier_context) {
- .buffer = MALLOC(cb_chunk_size),
- .buffer_filled = 0,
- .cb_chunk_size = cb_chunk_size,
- .ranges = ranges,
- .num_ranges = stream_count,
- .cur_range = 0,
- .range_bytes_remaining = ranges[0].size,
- .cb = cbs->consume_chunk,
- .cb_ctx = cbs->consume_chunk_ctx,
- };
-
- if (rechunkifier_ctx->buffer == NULL)
- return WIMLIB_ERR_NOMEM;
- last_cb = rechunkifier_cb;
- last_cb_ctx = rechunkifier_ctx;
- } else {
- rechunkifier_ctx = NULL;
- last_cb = cbs->consume_chunk;
- last_cb_ctx = cbs->consume_chunk_ctx;
- }
-
- struct hasher_context hasher_ctx = {
- .cbs = {
- .begin_stream = cbs->begin_stream,
- .begin_stream_ctx = cbs->begin_stream_ctx,
- .consume_chunk = last_cb,
- .consume_chunk_ctx = last_cb_ctx,
- .end_stream = cbs->end_stream,
- .end_stream_ctx = cbs->end_stream_ctx,
- },
- };
-
- struct streamifier_context streamifier_ctx = {
- .cbs = {
- .begin_stream = hasher_begin_stream,
- .begin_stream_ctx = &hasher_ctx,
- .consume_chunk = hasher_consume_chunk,
- .consume_chunk_ctx = &hasher_ctx,
- .end_stream = hasher_end_stream,
- .end_stream_ctx = &hasher_ctx,
- },
- .cur_stream = lte,
- .cur_stream_offset = 0,
- .final_stream = lte_last,
- .list_head_offset = list_head_offset,
- };
-
- ret = read_compressed_wim_resource(lte->rspec,
- ranges,
- stream_count,
- streamifier_cb,
- &streamifier_ctx,
- false);
- if (rechunkifier_ctx != NULL)
- FREE(rechunkifier_ctx->buffer);
-
- if (ret) {
- if (streamifier_ctx.cur_stream_offset != 0) {
- ret = (*streamifier_ctx.cbs.end_stream)
- (streamifier_ctx.cur_stream,
- ret,
- streamifier_ctx.cbs.end_stream_ctx);
- }
+ ret = read_packed_streams(lte, lte_last,
+ stream_count,
+ list_head_offset,
+ sink_cbs);
+ if (ret)
return ret;
- }
continue;
}
}
- ret = read_full_stream_with_sha1(lte, cbs);
- if (ret > 0)
+ ret = read_full_stream_with_cbs(lte, sink_cbs);
+ if (ret && ret != BEGIN_STREAM_STATUS_SKIP_STREAM)
return ret;
}
return 0;
extract_stream(struct wim_lookup_table_entry *lte, u64 size,
consume_data_callback_t extract_chunk, void *extract_chunk_arg)
{
+ wimlib_assert(size <= lte->size);
if (size == lte->size) {
/* Do SHA1. */
struct read_stream_list_callbacks cbs = {
} else {
/* Don't do SHA1. */
return read_stream_prefix(lte, size, extract_chunk,
- extract_chunk_arg, 0);
+ extract_chunk_arg);
}
}
return extract_stream(lte, size, extract_chunk_to_fd, fd);
}
+/* Extract the full uncompressed contents of the specified stream to the
+ * specified file descriptor. */
+int
+extract_full_stream_to_fd(struct wim_lookup_table_entry *lte,
+ struct filedes *fd)
+{
+ return extract_stream_to_fd(lte, fd, lte->size);
+}
+
/* Calculate the SHA1 message digest of a stream and store it in @lte->hash. */
int
sha1_stream(struct wim_lookup_table_entry *lte)
}
/* Convert a short WIM resource header to a stand-alone WIM resource
- * specification. */
+ * specification.
+ *
+ * Note: for packed resources some fields still need to be overridden.
+ */
void
wim_res_hdr_to_spec(const struct wim_reshdr *reshdr, WIMStruct *wim,
- struct wim_resource_spec *spec)
+ struct wim_resource_spec *rspec)
{
- spec->wim = wim;
- spec->offset_in_wim = reshdr->offset_in_wim;
- spec->size_in_wim = reshdr->size_in_wim;
- spec->uncompressed_size = reshdr->uncompressed_size;
- INIT_LIST_HEAD(&spec->stream_list);
- spec->flags = reshdr->flags;
- spec->is_pipable = wim_is_pipable(wim);
+ rspec->wim = wim;
+ rspec->offset_in_wim = reshdr->offset_in_wim;
+ rspec->size_in_wim = reshdr->size_in_wim;
+ rspec->uncompressed_size = reshdr->uncompressed_size;
+ INIT_LIST_HEAD(&rspec->stream_list);
+ rspec->flags = reshdr->flags;
+ rspec->is_pipable = wim_is_pipable(wim);
+ if (rspec->flags & WIM_RESHDR_FLAG_COMPRESSED) {
+ rspec->compression_type = wim->compression_type;
+ rspec->chunk_size = wim->chunk_size;
+ } else {
+ rspec->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
+ rspec->chunk_size = 0;
+ }
}
/* Convert a stand-alone resource specification to a WIM resource header. */
/* Translates a WIM resource header from the on-disk format into an in-memory
* format. */
-int
+void
get_wim_reshdr(const struct wim_reshdr_disk *disk_reshdr,
struct wim_reshdr *reshdr)
{
((u64)disk_reshdr->size_in_wim[6] << 48));
reshdr->uncompressed_size = le64_to_cpu(disk_reshdr->uncompressed_size);
reshdr->flags = disk_reshdr->flags;
-
- /* Avoid possible overflows. */
- if (reshdr->offset_in_wim & 0xc000000000000000ULL)
- return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
-
- if (reshdr->uncompressed_size & 0xc000000000000000ULL)
- return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
-
- return 0;
}
/* Translates a WIM resource header from an in-memory format into the on-disk
git://wimlib.net / wimlib / blobdiff