# include "config.h"
#endif
+#include "wimlib/assert.h"
#include "wimlib/endianness.h"
#include "wimlib/error.h"
-#include "wimlib/file_io.h"
-#include "wimlib/glob.h"
#include "wimlib/lookup_table.h"
#include "wimlib/metadata.h"
-#include "wimlib/paths.h"
+#include "wimlib/ntfs_3g.h"
#include "wimlib/resource.h"
#include "wimlib/util.h"
#include "wimlib/write.h"
-#include <errno.h>
#include <stdlib.h>
-#ifdef WITH_FUSE
-# include <unistd.h> /* for unlink() */
-#endif
+#include <string.h>
+#include <unistd.h> /* for unlink() */
+
+/* WIM lookup table:
+ *
+ * This is a logical mapping from SHA1 message digests to the data streams
+ * contained in a WIM.
+ *
+ * Here it is implemented as a hash table.
+ *
+ * Note: Everything will break horribly if there is a SHA1 collision.
+ */
+struct wim_lookup_table {
+ struct hlist_head *array;
+ size_t num_entries;
+ size_t capacity;
+};
struct wim_lookup_table *
new_lookup_table(size_t capacity)
struct wim_lookup_table *table;
struct hlist_head *array;
- table = CALLOC(1, sizeof(struct wim_lookup_table));
- if (table) {
- array = CALLOC(capacity, sizeof(array[0]));
- if (array) {
- table->num_entries = 0;
- table->capacity = capacity;
- table->array = array;
- } else {
- FREE(table);
- table = NULL;
- ERROR("Failed to allocate memory for lookup table "
- "with capacity %zu", capacity);
- }
+ table = MALLOC(sizeof(struct wim_lookup_table));
+ if (table == NULL)
+ goto oom;
+
+ array = CALLOC(capacity, sizeof(array[0]));
+ if (array == NULL) {
+ FREE(table);
+ goto oom;
}
+
+ table->num_entries = 0;
+ table->capacity = capacity;
+ table->array = array;
return table;
+
+oom:
+ ERROR("Failed to allocate memory for lookup table "
+ "with capacity %zu", capacity);
+ return NULL;
+}
+
+static int
+do_free_lookup_table_entry(struct wim_lookup_table_entry *entry, void *ignore)
+{
+ free_lookup_table_entry(entry);
+ return 0;
+}
+
+void
+free_lookup_table(struct wim_lookup_table *table)
+{
+ DEBUG("Freeing lookup table.");
+ if (table == NULL)
+ return;
+
+ if (table->array) {
+ for_lookup_table_entry(table,
+ do_free_lookup_table_entry,
+ NULL);
+ FREE(table->array);
+ }
+ FREE(table);
}
struct wim_lookup_table_entry *
struct wim_lookup_table_entry *lte;
lte = CALLOC(1, sizeof(struct wim_lookup_table_entry));
- if (lte == NULL) {
- ERROR("Out of memory (tried to allocate %zu bytes for "
- "lookup table entry)",
- sizeof(struct wim_lookup_table_entry));
+ if (lte == NULL)
return NULL;
- }
+
lte->refcnt = 1;
+
+ /* lte->resource_location = RESOURCE_NONEXISTENT */
BUILD_BUG_ON(RESOURCE_NONEXISTENT != 0);
+
return lte;
}
break;
}
return new;
+
out_free:
free_lookup_table_entry(new);
return NULL;
}
void
-free_lookup_table_entry(struct wim_lookup_table_entry *lte)
+lte_put_resource(struct wim_lookup_table_entry *lte)
{
- if (lte) {
- switch (lte->resource_location) {
- case RESOURCE_IN_WIM:
- list_del(<e->rspec_node);
- if (list_empty(<e->rspec->stream_list))
- FREE(lte->rspec);
- break;
- case RESOURCE_IN_FILE_ON_DISK:
- #ifdef __WIN32__
- case RESOURCE_WIN32_ENCRYPTED:
- #endif
- #ifdef WITH_FUSE
- case RESOURCE_IN_STAGING_FILE:
- BUILD_BUG_ON((void*)<e->file_on_disk !=
- (void*)<e->staging_file_name);
- #endif
- case RESOURCE_IN_ATTACHED_BUFFER:
- BUILD_BUG_ON((void*)<e->file_on_disk !=
- (void*)<e->attached_buffer);
- FREE(lte->file_on_disk);
- break;
-#ifdef WITH_NTFS_3G
- case RESOURCE_IN_NTFS_VOLUME:
- if (lte->ntfs_loc) {
- FREE(lte->ntfs_loc->path);
- FREE(lte->ntfs_loc->stream_name);
- FREE(lte->ntfs_loc);
- }
- break;
+ switch (lte->resource_location) {
+ case RESOURCE_IN_WIM:
+ list_del(<e->rspec_node);
+ if (list_empty(<e->rspec->stream_list))
+ FREE(lte->rspec);
+ break;
+ case RESOURCE_IN_FILE_ON_DISK:
+#ifdef __WIN32__
+ case RESOURCE_WIN32_ENCRYPTED:
#endif
- default:
- break;
+#ifdef WITH_FUSE
+ case RESOURCE_IN_STAGING_FILE:
+ BUILD_BUG_ON((void*)<e->file_on_disk !=
+ (void*)<e->staging_file_name);
+#endif
+ case RESOURCE_IN_ATTACHED_BUFFER:
+ BUILD_BUG_ON((void*)<e->file_on_disk !=
+ (void*)<e->attached_buffer);
+ FREE(lte->file_on_disk);
+ break;
+#ifdef WITH_NTFS_3G
+ case RESOURCE_IN_NTFS_VOLUME:
+ if (lte->ntfs_loc) {
+ FREE(lte->ntfs_loc->path);
+ FREE(lte->ntfs_loc->stream_name);
+ FREE(lte->ntfs_loc);
}
+ break;
+#endif
+ default:
+ break;
+ }
+}
+
+void
+free_lookup_table_entry(struct wim_lookup_table_entry *lte)
+{
+ if (lte) {
+ lte_put_resource(lte);
FREE(lte);
}
}
-static int
-do_free_lookup_table_entry(struct wim_lookup_table_entry *entry, void *ignore)
+/* Should this stream be retained even if it has no references? */
+static bool
+should_retain_lte(const struct wim_lookup_table_entry *lte)
{
- free_lookup_table_entry(entry);
- return 0;
+ return lte->resource_location == RESOURCE_IN_WIM;
}
+static void
+finalize_lte(struct wim_lookup_table_entry *lte)
+{
+ if (!should_retain_lte(lte))
+ free_lookup_table_entry(lte);
+}
+/*
+ * Decrements the reference count for the lookup table entry @lte, which must be
+ * inserted in the stream lookup table @table.
+ *
+ * If the reference count reaches 0, this may cause @lte to be destroyed.
+ * However, we may retain entries with 0 reference count. This does not affect
+ * correctness, but it prevents the entries for valid streams in a WIM archive,
+ * which will continue to be present after appending to the file, from being
+ * lost merely because we dropped all references to them.
+ */
void
-free_lookup_table(struct wim_lookup_table *table)
+lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
+ struct wim_lookup_table *table)
{
- DEBUG("Freeing lookup table.");
- if (table) {
- if (table->array) {
- for_lookup_table_entry(table,
- do_free_lookup_table_entry,
- NULL);
- FREE(table->array);
+ wimlib_assert(lte->refcnt != 0);
+
+ if (--lte->refcnt == 0) {
+ if (lte->unhashed) {
+ list_del(<e->unhashed_list);
+ #ifdef WITH_FUSE
+ /* If the stream has been extracted to a staging file
+ * for a FUSE mount, unlink the staging file. (Note
+ * that there still may be open file descriptors to it.)
+ * */
+ if (lte->resource_location == RESOURCE_IN_STAGING_FILE)
+ unlink(lte->staging_file_name);
+ #endif
+ } else {
+ if (!should_retain_lte(lte))
+ lookup_table_unlink(table, lte);
}
- FREE(table);
+
+ /* If FUSE mounts are enabled, we don't actually free the entry
+ * until the last file descriptor has been closed by
+ * lte_decrement_num_opened_fds(). */
+#ifdef WITH_FUSE
+ if (lte->num_opened_fds == 0)
+#endif
+ finalize_lte(lte);
}
}
+#ifdef WITH_FUSE
+void
+lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
+{
+ wimlib_assert(lte->num_opened_fds != 0);
+
+ if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
+ finalize_lte(lte);
+}
+#endif
+
static void
lookup_table_insert_raw(struct wim_lookup_table *table,
struct wim_lookup_table_entry *lte)
FREE(old_array);
}
-
-/*
- * Inserts an entry into the lookup table.
- *
- * @table: A pointer to the lookup table.
- * @lte: A pointer to the entry to insert.
- */
+/* Inserts an entry into the lookup table. */
void
lookup_table_insert(struct wim_lookup_table *table,
struct wim_lookup_table_entry *lte)
enlarge_lookup_table(table);
}
-static void
-finalize_lte(struct wim_lookup_table_entry *lte)
-{
- #ifdef WITH_FUSE
- if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
- unlink(lte->staging_file_name);
- list_del(<e->unhashed_list);
- }
- #endif
- free_lookup_table_entry(lte);
-}
-
-/* Decrements the reference count for the lookup table entry @lte. If its
- * reference count reaches 0, it is unlinked from the lookup table. If,
- * furthermore, the entry has no opened file descriptors associated with it, the
- * entry is freed. */
+/* Unlinks a lookup table entry from the table; does not free it. */
void
-lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
- struct wim_lookup_table *table)
+lookup_table_unlink(struct wim_lookup_table *table,
+ struct wim_lookup_table_entry *lte)
{
- wimlib_assert(lte != NULL);
- wimlib_assert(lte->refcnt != 0);
- if (--lte->refcnt == 0) {
- if (lte->unhashed)
- list_del(<e->unhashed_list);
- else
- lookup_table_unlink(table, lte);
- #ifdef WITH_FUSE
- if (lte->num_opened_fds == 0)
- #endif
- finalize_lte(lte);
- }
+ wimlib_assert(!lte->unhashed);
+ wimlib_assert(table->num_entries != 0);
+
+ hlist_del(<e->hash_list);
+ table->num_entries--;
}
-#ifdef WITH_FUSE
-void
-lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
+/* Given a SHA1 message digest, return the corresponding entry in the WIM's
+ * lookup table, or NULL if there is none. */
+struct wim_lookup_table_entry *
+lookup_stream(const struct wim_lookup_table *table, const u8 hash[])
{
- if (lte->num_opened_fds != 0)
- if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
- finalize_lte(lte);
+ size_t i;
+ struct wim_lookup_table_entry *lte;
+ struct hlist_node *pos;
+
+ i = *(size_t*)hash % table->capacity;
+ hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
+ if (hashes_equal(hash, lte->hash))
+ return lte;
+ return NULL;
}
-#endif
/* Calls a function on all the entries in the WIM lookup table. Stop early and
* return nonzero if any call to the function returns nonzero. */
#define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
+static int
+cmp_streams_by_offset_in_res(const void *p1, const void *p2)
+{
+ const struct wim_lookup_table_entry *lte1, *lte2;
+
+ lte1 = *(const struct wim_lookup_table_entry**)p1;
+ lte2 = *(const struct wim_lookup_table_entry**)p2;
+
+ return cmp_u64(lte1->offset_in_res, lte2->offset_in_res);
+}
+
/* Validate the size and location of a WIM resource. */
static int
-validate_resource(const struct wim_resource_spec *rspec)
+validate_resource(struct wim_resource_spec *rspec)
{
struct wim_lookup_table_entry *lte;
- u64 cur_offset;
+ bool out_of_order;
+ u64 expected_next_offset;
+ int ret;
- /* Verify that calculating the offset of the end of the resource doesn't
- * overflow. */
+ /* Verify that the resource itself has a valid offset and size. */
if (rspec->offset_in_wim + rspec->size_in_wim < rspec->size_in_wim)
- goto invalid;
+ goto invalid_due_to_overflow;
- /* Verify that each stream in the resource has a valid offset and size,
- * and that no streams overlap, and that the streams were added in order
- * of increasing offset. */
- cur_offset = 0;
+ /* Verify that each stream in the resource has a valid offset and size.
+ */
+ expected_next_offset = 0;
+ out_of_order = false;
list_for_each_entry(lte, &rspec->stream_list, rspec_node) {
if (lte->offset_in_res + lte->size < lte->size ||
- lte->offset_in_res + lte->size > rspec->uncompressed_size ||
- lte->offset_in_res < cur_offset)
- goto invalid;
+ lte->offset_in_res + lte->size > rspec->uncompressed_size)
+ goto invalid_due_to_overflow;
+
+ if (lte->offset_in_res >= expected_next_offset)
+ expected_next_offset = lte->offset_in_res + lte->size;
+ else
+ out_of_order = true;
+ }
+
+ /* If the streams were not located at strictly increasing positions (not
+ * allowing for overlap), sort them. Then make sure that none overlap.
+ */
+ if (out_of_order) {
+ ret = sort_stream_list(&rspec->stream_list,
+ offsetof(struct wim_lookup_table_entry,
+ rspec_node),
+ cmp_streams_by_offset_in_res);
+ if (ret)
+ return ret;
- cur_offset = lte->offset_in_res + lte->size;
+ expected_next_offset = 0;
+ list_for_each_entry(lte, &rspec->stream_list, rspec_node) {
+ if (lte->offset_in_res >= expected_next_offset)
+ expected_next_offset = lte->offset_in_res + lte->size;
+ else
+ goto invalid_due_to_overlap;
+ }
}
+
return 0;
-invalid:
+invalid_due_to_overflow:
+ ERROR("Invalid resource entry (offset overflow)");
+ return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
- ERROR("Invalid resource entry!");
+invalid_due_to_overlap:
+ ERROR("Invalid resource entry (streams in packed resource overlap)");
return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
}
+/* Validate the resource, or free it if unused. */
+static int
+finish_resource(struct wim_resource_spec *rspec)
+{
+ if (!list_empty(&rspec->stream_list)) {
+ /* This resource contains at least one stream. */
+ return validate_resource(rspec);
+ } else {
+ /* No streams are in this resource. Get rid of it. */
+ FREE(rspec);
+ return 0;
+ }
+}
+
/*
- * Reads the lookup table from a WIM file. Each entry specifies a stream that
- * the WIM file contains, along with its location and SHA1 message digest.
+ * Reads the lookup table from a WIM file. Usually, each entry specifies a
+ * stream that the WIM file contains, along with its location and SHA1 message
+ * digest.
+ *
+ * Saves lookup table entries for non-metadata streams in a hash table (set to
+ * wim->lookup_table), and saves the metadata entry for each image in a special
+ * per-image location (the wim->image_metadata array).
*
- * Saves lookup table entries for non-metadata streams in a hash table, and
- * saves the metadata entry for each image in a special per-image location (the
- * image_metadata array).
+ * This works for both version WIM_VERSION_DEFAULT (68864) and version
+ * WIM_VERSION_PACKED_STREAMS (3584) WIMs.
*
- * Return values:
+ * Possible return values:
* WIMLIB_ERR_SUCCESS (0)
* WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY
- * WIMLIB_ERR_RESOURCE_NOT_FOUND
+ * WIMLIB_ERR_NOMEM
*
- * Or an error code caused by failure to read the lookup table into memory.
+ * Or an error code caused by failure to read the lookup table from the WIM
+ * file.
*/
int
read_wim_lookup_table(WIMStruct *wim)
{
int ret;
- size_t i;
size_t num_entries;
- struct wim_lookup_table *table;
- struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
- struct wim_resource_spec *cur_rspec;
- void *buf;
- bool back_to_back_pack;
+ void *buf = NULL;
+ struct wim_lookup_table *table = NULL;
+ struct wim_lookup_table_entry *cur_entry = NULL;
+ struct wim_resource_spec *cur_rspec = NULL;
+ size_t num_duplicate_entries = 0;
+ size_t num_wrong_part_entries = 0;
+ u32 image_index = 0;
DEBUG("Reading lookup table.");
BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
- /* Calculate number of entries in the lookup table. */
+ /* Calculate the number of entries in the lookup table. */
num_entries = wim->hdr.lookup_table_reshdr.uncompressed_size /
sizeof(struct wim_lookup_table_entry_disk);
/* Allocate a hash table to map SHA1 message digests into stream
* specifications. This is the in-memory "lookup table". */
table = new_lookup_table(num_entries * 2 + 1);
- if (table == NULL) {
- ERROR("Not enough memory to read lookup table.");
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_buf;
- }
-
- /* Allocate and initalize stream entries from the raw lookup table
- * buffer. */
- wim->current_image = 0;
- cur_rspec = NULL;
- for (i = 0; i < num_entries; i++) {
+ if (!table)
+ goto oom;
+
+ /* Allocate and initalize stream entries ('struct
+ * wim_lookup_table_entry's) from the raw lookup table buffer. Each of
+ * these entries will point to a 'struct wim_resource_spec' that
+ * describes the underlying resource. In WIMs with version number
+ * WIM_VERSION_PACKED_STREAMS, a resource may contain multiple streams.
+ */
+ for (size_t i = 0; i < num_entries; i++) {
const struct wim_lookup_table_entry_disk *disk_entry =
&((const struct wim_lookup_table_entry_disk*)buf)[i];
- u16 part_number;
struct wim_reshdr reshdr;
+ u16 part_number;
+ struct wim_lookup_table_entry *duplicate_entry;
+ /* Get the resource header */
get_wim_reshdr(&disk_entry->reshdr, &reshdr);
DEBUG("reshdr: size_in_wim=%"PRIu64", "
"uncompressed_size=%"PRIu64", "
"offset_in_wim=%"PRIu64", "
- "flags=0x%02x",
+ "flags=0x%02x\n",
reshdr.size_in_wim, reshdr.uncompressed_size,
reshdr.offset_in_wim, reshdr.flags);
+ /* Ignore PACKED_STREAMS flag if it isn't supposed to be used in
+ * this WIM version */
if (wim->hdr.wim_version == WIM_VERSION_DEFAULT)
reshdr.flags &= ~WIM_RESHDR_FLAG_PACKED_STREAMS;
+ /* Allocate a 'struct wim_lookup_table_entry' */
cur_entry = new_lookup_table_entry();
- if (cur_entry == NULL) {
- ERROR("Not enough memory to read lookup table!");
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_lookup_table;
- }
+ if (!cur_entry)
+ goto oom;
+ /* Get the part number, reference count, and hash. */
part_number = le16_to_cpu(disk_entry->part_number);
cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
copy_hash(cur_entry->hash, disk_entry->hash);
+ /* Verify that the part number matches that of the underlying
+ * WIM file. */
if (part_number != wim->hdr.part_number) {
- WARNING("A lookup table entry in part %hu of the WIM "
- "points to part %hu (ignoring it)",
- wim->hdr.part_number, part_number);
- free_lookup_table_entry(cur_entry);
- continue;
+ num_wrong_part_entries++;
+ goto free_cur_entry_and_continue;
}
+ /* If resource is uncompressed, check for (unexpected) size
+ * mismatch. */
if (!(reshdr.flags & (WIM_RESHDR_FLAG_PACKED_STREAMS |
WIM_RESHDR_FLAG_COMPRESSED))) {
if (reshdr.uncompressed_size != reshdr.size_in_wim) {
- ERROR("Invalid resource entry!");
- ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
- goto out_free_cur_entry;
+ /* So ... This is an uncompressed resource, but
+ * its uncompressed size is NOT the same as its
+ * "compressed" size (size_in_wim). What to do
+ * with it?
+ *
+ * Based on a simple test, WIMGAPI seems to
+ * handle this as follows:
+ *
+ * if (size_in_wim > uncompressed_size) {
+ * Ignore uncompressed_size; use
+ * size_in_wim instead.
+ * } else {
+ * Honor uncompressed_size, but treat the
+ * part of the file data above size_in_wim
+ * as all zeros.
+ * }
+ *
+ * So we will do the same.
+ */
+ if (reshdr.size_in_wim > reshdr.uncompressed_size)
+ reshdr.uncompressed_size = reshdr.size_in_wim;
}
}
- back_to_back_pack = false;
- if (!(reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) ||
- cur_rspec == NULL ||
- (back_to_back_pack =
- ((reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) &&
- reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER &&
- cur_rspec != NULL &&
- cur_rspec->size_in_wim != 0)))
+ /*
+ * Possibly start a new resource.
+ *
+ * We need to start a new resource if:
+ *
+ * - There is no previous resource (cur_rspec).
+ *
+ * OR
+ *
+ * - The resource header did not have PACKED_STREAMS set, so it
+ * specifies a new, single-stream resource.
+ *
+ * OR
+ *
+ * - The resource header had PACKED_STREAMS set, and it's a
+ * special entry that specifies the resource itself as opposed
+ * to a stream, and we already encountered one such entry in
+ * the current resource. We will interpret this as the
+ * beginning of a new packed resource. (However, note that
+ * wimlib does not currently allow create WIMs with multiple
+ * packed resources, as to remain compatible with WIMGAPI.)
+ */
+ if (likely(!cur_rspec) ||
+ !(reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) ||
+ (reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER &&
+ cur_rspec->size_in_wim != 0))
{
- /* Starting new run of streams that share the same WIM
- * resource. */
- struct wim_lookup_table_entry *prev_entry = NULL;
-
- if (back_to_back_pack) {
- prev_entry = list_entry(cur_rspec->stream_list.prev,
- struct wim_lookup_table_entry,
- rspec_node);
- lte_unbind_wim_resource_spec(prev_entry);
- cur_rspec->uncompressed_size -= prev_entry->size;
- }
- if (cur_rspec != NULL) {
- ret = validate_resource(cur_rspec);
+ /* Finish previous resource (if existent) */
+ if (cur_rspec) {
+ ret = finish_resource(cur_rspec);
+ cur_rspec = NULL;
if (ret)
- goto out_free_cur_entry;
+ goto out;
}
/* Allocate the resource specification and initialize it
* with values from the current stream entry. */
cur_rspec = MALLOC(sizeof(*cur_rspec));
- if (cur_rspec == NULL) {
- ERROR("Not enough memory to read lookup table!");
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_cur_entry;
- }
+ if (!cur_rspec)
+ goto oom;
+
wim_res_hdr_to_spec(&reshdr, wim, cur_rspec);
/* If this is a packed run, the current stream entry may
cur_rspec->uncompressed_size = 0;
cur_rspec->offset_in_wim = 0;
}
-
- if (prev_entry) {
- lte_bind_wim_resource_spec(prev_entry, cur_rspec);
- cur_rspec->uncompressed_size = prev_entry->size;
- }
}
- if ((reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) &&
- reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER)
+ /* Now cur_rspec != NULL. */
+
+ /* Checked for packed resource specification. */
+ if (unlikely((reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) &&
+ reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER))
{
/* Found the specification for the packed resource.
* Transfer the values to the `struct
* since this lookup table entry did not, in fact,
* correspond to a "stream". */
+ /* The uncompressed size of the packed resource is
+ * actually stored in the header of the resource itself.
+ * Read it, and also grab the chunk size and compression
+ * type (which are not necessarily the defaults from the
+ * WIM header). */
+ struct alt_chunk_table_header_disk hdr;
+
+ ret = full_pread(&wim->in_fd, &hdr,
+ sizeof(hdr), reshdr.offset_in_wim);
+ if (ret)
+ goto out;
+
+ cur_rspec->uncompressed_size = le64_to_cpu(hdr.res_usize);
cur_rspec->offset_in_wim = reshdr.offset_in_wim;
cur_rspec->size_in_wim = reshdr.size_in_wim;
cur_rspec->flags = reshdr.flags;
+
+ /* Compression format numbers must be the same as in
+ * WIMGAPI to be compatible here. */
+ BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
+ BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_XPRESS != 1);
+ BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_LZX != 2);
+ BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_LZMS != 3);
+ cur_rspec->compression_type = le32_to_cpu(hdr.compression_format);
+
+ cur_rspec->chunk_size = le32_to_cpu(hdr.chunk_size);
+
DEBUG("Full pack is %"PRIu64" compressed bytes "
"at file offset %"PRIu64" (flags 0x%02x)",
cur_rspec->size_in_wim,
cur_rspec->offset_in_wim,
cur_rspec->flags);
- free_lookup_table_entry(cur_entry);
- continue;
+ goto free_cur_entry_and_continue;
}
- if (is_zero_hash(cur_entry->hash)) {
- free_lookup_table_entry(cur_entry);
- continue;
- }
+ /* Ignore entries with all zeroes in the hash field. */
+ if (is_zero_hash(cur_entry->hash))
+ goto free_cur_entry_and_continue;
- if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
- /* Continuing the pack with another stream. */
- DEBUG("Continuing packed run with stream: "
- "%"PRIu64" uncompressed bytes @ resource offset %"PRIu64")",
- reshdr.size_in_wim, reshdr.offset_in_wim);
- cur_rspec->uncompressed_size += reshdr.size_in_wim;
+ if (reshdr.flags & WIM_RESHDR_FLAG_METADATA) {
+
+ /* Lookup table entry for a metadata resource. */
+
+ /* Metadata entries with no references must be ignored.
+ * See, for example, the WinPE WIMs from the WAIK v2.1.
+ */
+ if (cur_entry->refcnt == 0)
+ goto free_cur_entry_and_continue;
+
+ if (cur_entry->refcnt != 1) {
+ /* We don't currently support this case due to
+ * the complications of multiple images sharing
+ * the same metadata resource or a metadata
+ * resource also being referenced by files.
+ */
+ ERROR("Found metadata resource with refcnt != 1");
+ ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
+ goto out;
+ }
+
+ if (wim->hdr.part_number != 1) {
+ WARNING("Ignoring metadata resource found in a "
+ "non-first part of the split WIM");
+ goto free_cur_entry_and_continue;
+ }
+
+ /* The number of entries in the lookup table with
+ * WIM_RESHDR_FLAG_METADATA set should be the same as
+ * the image_count field in the WIM header. */
+ if (image_index == wim->hdr.image_count) {
+ WARNING("Found more metadata resources than images");
+ goto free_cur_entry_and_continue;
+ }
+
+ /* Notice very carefully: We are assigning the metadata
+ * resources to images in the same order in which their
+ * lookup table entries occur on disk. (This is also
+ * the behavior of Microsoft's software.) In
+ * particular, this overrides the actual locations of
+ * the metadata resources themselves in the WIM file as
+ * well as any information written in the XML data. */
+ DEBUG("Found metadata resource for image %"PRIu32" at "
+ "offset %"PRIu64".",
+ image_index + 1,
+ reshdr.offset_in_wim);
+
+ wim->image_metadata[image_index++]->metadata_lte = cur_entry;
+ } else {
+ /* Lookup table entry for a non-metadata stream. */
+
+ /* Ignore this stream if it's a duplicate. */
+ duplicate_entry = lookup_stream(table, cur_entry->hash);
+ if (duplicate_entry) {
+ num_duplicate_entries++;
+ goto free_cur_entry_and_continue;
+ }
+
+ /* Insert the stream into the in-memory lookup table,
+ * keyed by its SHA1 message digest. */
+ lookup_table_insert(table, cur_entry);
}
+ /* Add the stream to the current resource specification. */
lte_bind_wim_resource_spec(cur_entry, cur_rspec);
if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
/* In packed runs, the offset field is used for
cur_entry->offset_in_res = reshdr.offset_in_wim;
cur_entry->size = reshdr.size_in_wim;
cur_entry->flags = reshdr.flags;
+ /* cur_rspec stays the same */
+
} else {
/* Normal case: The stream corresponds one-to-one with
* the resource entry. */
cur_entry->flags = reshdr.flags;
cur_rspec = NULL;
}
+ continue;
- if (cur_entry->flags & WIM_RESHDR_FLAG_METADATA) {
- /* Lookup table entry for a metadata resource */
-
- /* Metadata entries with no references must be ignored;
- * see for example the WinPE WIMs from the WAIK v2.1.
- * */
- if (cur_entry->refcnt == 0) {
- free_lookup_table_entry(cur_entry);
- continue;
- }
-
- if (cur_entry->refcnt != 1) {
- if (wimlib_print_errors) {
- ERROR("Found metadata resource with refcnt != 1:");
- print_lookup_table_entry(cur_entry, stderr);
- }
- ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
- goto out_free_cur_entry;
- }
-
- if (wim->hdr.part_number != 1) {
- WARNING("Ignoring metadata resource found in a "
- "non-first part of the split WIM");
- free_lookup_table_entry(cur_entry);
- continue;
- }
- if (wim->current_image == wim->hdr.image_count) {
- WARNING("The WIM header says there are %u images "
- "in the WIM, but we found more metadata "
- "resources than this (ignoring the extra)",
- wim->hdr.image_count);
- free_lookup_table_entry(cur_entry);
- continue;
- }
-
- /* Notice very carefully: We are assigning the metadata
- * resources in the exact order mirrored by their lookup
- * table entries on disk, which is the behavior of
- * Microsoft's software. In particular, this overrides
- * the actual locations of the metadata resources
- * themselves in the WIM file as well as any information
- * written in the XML data. */
- DEBUG("Found metadata resource for image %u at "
- "offset %"PRIu64".",
- wim->current_image + 1,
- cur_entry->rspec->offset_in_wim);
- wim->image_metadata[
- wim->current_image++]->metadata_lte = cur_entry;
- continue;
- }
-
- /* Lookup table entry for a stream that is not a metadata
- * resource. */
- duplicate_entry = lookup_resource(table, cur_entry->hash);
- if (duplicate_entry) {
- if (wimlib_print_errors) {
- WARNING("The WIM lookup table contains two entries with the "
- "same SHA1 message digest!");
- WARNING("The first entry is:");
- print_lookup_table_entry(duplicate_entry, stderr);
- WARNING("The second entry is:");
- print_lookup_table_entry(cur_entry, stderr);
- }
- free_lookup_table_entry(cur_entry);
- continue;
- }
-
- /* Finally, insert the stream into the lookup table, keyed by
- * its SHA1 message digest. */
- lookup_table_insert(table, cur_entry);
+ free_cur_entry_and_continue:
+ free_lookup_table_entry(cur_entry);
}
+ cur_entry = NULL;
/* Validate the last resource. */
- if (cur_rspec != NULL) {
- ret = validate_resource(cur_rspec);
+ if (cur_rspec) {
+ ret = finish_resource(cur_rspec);
+ cur_rspec = NULL;
if (ret)
- goto out_free_lookup_table;
+ goto out;
}
- if (wim->hdr.part_number == 1 && wim->current_image != wim->hdr.image_count) {
- WARNING("The header of \"%"TS"\" says there are %u images in\n"
- " the WIM, but we only found %d metadata resources! Acting as if\n"
- " the header specified only %d images instead.",
- wim->filename, wim->hdr.image_count,
- wim->current_image, wim->current_image);
- for (int i = wim->current_image; i < wim->hdr.image_count; i++)
+ if (wim->hdr.part_number == 1 && image_index != wim->hdr.image_count) {
+ WARNING("Could not find metadata resources for all images");
+ for (u32 i = image_index; i < wim->hdr.image_count; i++)
put_image_metadata(wim->image_metadata[i], NULL);
- wim->hdr.image_count = wim->current_image;
+ wim->hdr.image_count = image_index;
+ }
+
+ if (num_duplicate_entries > 0) {
+ WARNING("Ignoring %zu duplicate streams in the WIM lookup table",
+ num_duplicate_entries);
}
+
+ if (num_wrong_part_entries > 0) {
+ WARNING("Ignoring %zu streams with wrong part number",
+ num_wrong_part_entries);
+ }
+
DEBUG("Done reading lookup table.");
wim->lookup_table = table;
+ table = NULL;
ret = 0;
- goto out_free_buf;
-
-out_free_cur_entry:
- FREE(cur_entry);
-out_free_lookup_table:
+ goto out;
+oom:
+ ERROR("Not enough memory to read lookup table!");
+ ret = WIMLIB_ERR_NOMEM;
+out:
+ if (cur_rspec && list_empty(&cur_rspec->stream_list))
+ FREE(cur_rspec);
+ free_lookup_table_entry(cur_entry);
free_lookup_table(table);
-out_free_buf:
FREE(buf);
-out:
- wim->current_image = 0;
return ret;
}
return 0;
}
-void
-print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
+/* Allocate a stream entry for the contents of the buffer, or re-use an existing
+ * entry in @lookup_table for the same stream. */
+struct wim_lookup_table_entry *
+new_stream_from_data_buffer(const void *buffer, size_t size,
+ struct wim_lookup_table *lookup_table)
{
- if (lte == NULL) {
- tputc(T('\n'), out);
- return;
+ u8 hash[SHA1_HASH_SIZE];
+ struct wim_lookup_table_entry *lte, *existing_lte;
+
+ sha1_buffer(buffer, size, hash);
+ existing_lte = lookup_stream(lookup_table, hash);
+ if (existing_lte) {
+ wimlib_assert(existing_lte->size == size);
+ lte = existing_lte;
+ lte->refcnt++;
+ } else {
+ void *buffer_copy;
+ lte = new_lookup_table_entry();
+ if (lte == NULL)
+ return NULL;
+ buffer_copy = memdup(buffer, size);
+ if (buffer_copy == NULL) {
+ free_lookup_table_entry(lte);
+ return NULL;
+ }
+ lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER;
+ lte->attached_buffer = buffer_copy;
+ lte->size = size;
+ copy_hash(lte->hash, hash);
+ lookup_table_insert(lookup_table, lte);
}
-
-
- tprintf(T("Uncompressed size = %"PRIu64" bytes\n"),
- lte->size);
- if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
- tprintf(T("Offset = %"PRIu64" bytes\n"),
- lte->offset_in_res);
-
- tprintf(T("Raw uncompressed size = %"PRIu64" bytes\n"),
- lte->rspec->uncompressed_size);
-
- tprintf(T("Raw compressed size = %"PRIu64" bytes\n"),
- lte->rspec->size_in_wim);
-
- tprintf(T("Raw offset = %"PRIu64" bytes\n"),
- lte->rspec->offset_in_wim);
- } else if (lte->resource_location == RESOURCE_IN_WIM) {
- tprintf(T("Compressed size = %"PRIu64" bytes\n"),
- lte->rspec->size_in_wim);
-
- tprintf(T("Offset = %"PRIu64" bytes\n"),
- lte->rspec->offset_in_wim);
- }
-
- tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
-
- if (lte->unhashed) {
- tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
- lte->back_inode, lte->back_stream_id);
- } else {
- tfprintf(out, T("Hash = 0x"));
- print_hash(lte->hash, out);
- tputc(T('\n'), out);
- }
-
- tfprintf(out, T("Flags = "));
- u8 flags = lte->flags;
- if (flags & WIM_RESHDR_FLAG_COMPRESSED)
- tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
- if (flags & WIM_RESHDR_FLAG_FREE)
- tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
- if (flags & WIM_RESHDR_FLAG_METADATA)
- tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
- if (flags & WIM_RESHDR_FLAG_SPANNED)
- tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
- if (flags & WIM_RESHDR_FLAG_PACKED_STREAMS)
- tfputs(T("WIM_RESHDR_FLAG_PACKED_STREAMS, "), out);
- tputc(T('\n'), out);
- switch (lte->resource_location) {
- case RESOURCE_IN_WIM:
- if (lte->rspec->wim->filename) {
- tfprintf(out, T("WIM file = `%"TS"'\n"),
- lte->rspec->wim->filename);
- }
- break;
-#ifdef __WIN32__
- case RESOURCE_WIN32_ENCRYPTED:
-#endif
- case RESOURCE_IN_FILE_ON_DISK:
- tfprintf(out, T("File on Disk = `%"TS"'\n"),
- lte->file_on_disk);
- break;
-#ifdef WITH_FUSE
- case RESOURCE_IN_STAGING_FILE:
- tfprintf(out, T("Staging File = `%"TS"'\n"),
- lte->staging_file_name);
- break;
-#endif
- default:
- break;
- }
- tputc(T('\n'), out);
-}
-
-void
-lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
- struct wimlib_resource_entry *wentry)
-{
- memset(wentry, 0, sizeof(*wentry));
-
- wentry->uncompressed_size = lte->size;
- if (lte->resource_location == RESOURCE_IN_WIM) {
- wentry->part_number = lte->rspec->wim->hdr.part_number;
- if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
- wentry->compressed_size = 0;
- wentry->offset = lte->offset_in_res;
- } else {
- wentry->compressed_size = lte->rspec->size_in_wim;
- wentry->offset = lte->rspec->offset_in_wim;
- }
- wentry->raw_resource_offset_in_wim = lte->rspec->offset_in_wim;
- /*wentry->raw_resource_uncompressed_size = lte->rspec->uncompressed_size;*/
- wentry->raw_resource_compressed_size = lte->rspec->size_in_wim;
- }
- copy_hash(wentry->sha1_hash, lte->hash);
- wentry->reference_count = lte->refcnt;
- wentry->is_compressed = (lte->flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
- wentry->is_metadata = (lte->flags & WIM_RESHDR_FLAG_METADATA) != 0;
- wentry->is_free = (lte->flags & WIM_RESHDR_FLAG_FREE) != 0;
- wentry->is_spanned = (lte->flags & WIM_RESHDR_FLAG_SPANNED) != 0;
- wentry->packed = (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) != 0;
-}
-
-struct iterate_lte_context {
- wimlib_iterate_lookup_table_callback_t cb;
- void *user_ctx;
-};
-
-static int
-do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
-{
- struct iterate_lte_context *ctx = _ctx;
- struct wimlib_resource_entry entry;
-
- lte_to_wimlib_resource_entry(lte, &entry);
- return (*ctx->cb)(&entry, ctx->user_ctx);
-}
-
-/* API function documented in wimlib.h */
-WIMLIBAPI int
-wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
- wimlib_iterate_lookup_table_callback_t cb,
- void *user_ctx)
-{
- struct iterate_lte_context ctx = {
- .cb = cb,
- .user_ctx = user_ctx,
- };
- if (wim->hdr.part_number == 1) {
- int ret;
- for (int i = 0; i < wim->hdr.image_count; i++) {
- ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
- &ctx);
- if (ret)
- return ret;
- }
- }
- return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);
-}
-
-/* Given a SHA1 message digest, return the corresponding entry in the WIM's
- * lookup table, or NULL if there is none. */
-struct wim_lookup_table_entry *
-lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
-{
- size_t i;
- struct wim_lookup_table_entry *lte;
- struct hlist_node *pos;
-
- wimlib_assert(table != NULL);
- wimlib_assert(hash != NULL);
-
- i = *(size_t*)hash % table->capacity;
- hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
- if (hashes_equal(hash, lte->hash))
- return lte;
- return NULL;
-}
-
-#ifdef WITH_FUSE
-/*
- * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
- * given a path name.
- *
- * This is only for pre-resolved inodes.
- */
-int
-wim_pathname_to_stream(WIMStruct *wim,
- const tchar *path,
- int lookup_flags,
- struct wim_dentry **dentry_ret,
- struct wim_lookup_table_entry **lte_ret,
- u16 *stream_idx_ret)
-{
- struct wim_dentry *dentry;
- struct wim_lookup_table_entry *lte;
- u16 stream_idx;
- const tchar *stream_name = NULL;
- struct wim_inode *inode;
- tchar *p = NULL;
-
- if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
- stream_name = path_stream_name(path);
- if (stream_name) {
- p = (tchar*)stream_name - 1;
- *p = T('\0');
- }
- }
-
- dentry = get_dentry(wim, path);
- if (p)
- *p = T(':');
- if (!dentry)
- return -errno;
-
- inode = dentry->d_inode;
-
- if (!inode->i_resolved)
- if (inode_resolve_ltes(inode, wim->lookup_table, false))
- return -EIO;
-
- if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
- && inode_is_directory(inode))
- return -EISDIR;
-
- if (stream_name) {
- struct wim_ads_entry *ads_entry;
- u16 ads_idx;
- ads_entry = inode_get_ads_entry(inode, stream_name,
- &ads_idx);
- if (ads_entry) {
- stream_idx = ads_idx + 1;
- lte = ads_entry->lte;
- goto out;
- } else {
- return -ENOENT;
- }
- } else {
- lte = inode_unnamed_stream_resolved(inode, &stream_idx);
- }
-out:
- if (dentry_ret)
- *dentry_ret = dentry;
- if (lte_ret)
- *lte_ret = lte;
- if (stream_idx_ret)
- *stream_idx_ret = stream_idx;
- return 0;
-}
-#endif
-
-int
-resource_not_found_error(const struct wim_inode *inode, const u8 *hash)
-{
- if (wimlib_print_errors) {
- ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
- tfprintf(stderr, T(" SHA-1 message digest of missing resource:\n "));
- print_hash(hash, stderr);
- tputc(T('\n'), stderr);
- }
- return WIMLIB_ERR_RESOURCE_NOT_FOUND;
-}
-
-/*
- * Resolve an inode's lookup table entries.
- *
- * This replaces the SHA1 hash fields (which are used to lookup an entry in the
- * lookup table) with pointers directly to the lookup table entries.
- *
- * If @force is %false:
- * If any needed SHA1 message digests are not found in the lookup table,
- * WIMLIB_ERR_RESOURCE_NOT_FOUND is returned and the inode is left
- * unmodified.
- * If @force is %true:
- * If any needed SHA1 message digests are not found in the lookup table,
- * new entries are allocated and inserted into the lookup table.
- */
-int
-inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table,
- bool force)
-{
- const u8 *hash;
-
- if (!inode->i_resolved) {
- struct wim_lookup_table_entry *lte, *ads_lte;
-
- /* Resolve the default file stream */
- lte = NULL;
- hash = inode->i_hash;
- if (!is_zero_hash(hash)) {
- lte = lookup_resource(table, hash);
- if (!lte) {
- if (force) {
- lte = new_lookup_table_entry();
- if (!lte)
- return WIMLIB_ERR_NOMEM;
- copy_hash(lte->hash, hash);
- lookup_table_insert(table, lte);
- } else {
- goto resource_not_found;
- }
- }
- }
-
- /* Resolve the alternate data streams */
- struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
- for (u16 i = 0; i < inode->i_num_ads; i++) {
- struct wim_ads_entry *cur_entry;
-
- ads_lte = NULL;
- cur_entry = &inode->i_ads_entries[i];
- hash = cur_entry->hash;
- if (!is_zero_hash(hash)) {
- ads_lte = lookup_resource(table, hash);
- if (!ads_lte) {
- if (force) {
- ads_lte = new_lookup_table_entry();
- if (!ads_lte)
- return WIMLIB_ERR_NOMEM;
- copy_hash(ads_lte->hash, hash);
- lookup_table_insert(table, ads_lte);
- } else {
- goto resource_not_found;
- }
- }
- }
- ads_ltes[i] = ads_lte;
- }
- inode->i_lte = lte;
- for (u16 i = 0; i < inode->i_num_ads; i++)
- inode->i_ads_entries[i].lte = ads_ltes[i];
- inode->i_resolved = 1;
- }
- return 0;
-
-resource_not_found:
- return resource_not_found_error(inode, hash);
-}
-
-void
-inode_unresolve_ltes(struct wim_inode *inode)
-{
- if (inode->i_resolved) {
- if (inode->i_lte)
- copy_hash(inode->i_hash, inode->i_lte->hash);
- else
- zero_out_hash(inode->i_hash);
-
- for (u16 i = 0; i < inode->i_num_ads; i++) {
- if (inode->i_ads_entries[i].lte)
- copy_hash(inode->i_ads_entries[i].hash,
- inode->i_ads_entries[i].lte->hash);
- else
- zero_out_hash(inode->i_ads_entries[i].hash);
- }
- inode->i_resolved = 0;
- }
-}
-
-/*
- * Returns the lookup table entry for stream @stream_idx of the inode, where
- * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
- * corresponds to an alternate data stream.
- *
- * This works for both resolved and un-resolved inodes.
- */
-struct wim_lookup_table_entry *
-inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
- const struct wim_lookup_table *table)
-{
- if (inode->i_resolved)
- return inode_stream_lte_resolved(inode, stream_idx);
- else
- return inode_stream_lte_unresolved(inode, stream_idx, table);
-}
-
-struct wim_lookup_table_entry *
-inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret)
-{
- wimlib_assert(inode->i_resolved);
- for (unsigned i = 0; i <= inode->i_num_ads; i++) {
- if (inode_stream_name_nbytes(inode, i) == 0 &&
- !is_zero_hash(inode_stream_hash_resolved(inode, i)))
- {
- *stream_idx_ret = i;
- return inode_stream_lte_resolved(inode, i);
- }
- }
- *stream_idx_ret = 0;
- return NULL;
-}
-
-struct wim_lookup_table_entry *
-inode_unnamed_lte_resolved(const struct wim_inode *inode)
-{
- u16 stream_idx;
- return inode_unnamed_stream_resolved(inode, &stream_idx);
-}
-
-struct wim_lookup_table_entry *
-inode_unnamed_lte_unresolved(const struct wim_inode *inode,
- const struct wim_lookup_table *table)
-{
- wimlib_assert(!inode->i_resolved);
- for (unsigned i = 0; i <= inode->i_num_ads; i++) {
- if (inode_stream_name_nbytes(inode, i) == 0 &&
- !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
- {
- return inode_stream_lte_unresolved(inode, i, table);
- }
- }
- return NULL;
-}
-
-/* Return the lookup table entry for the unnamed data stream of an inode, or
- * NULL if there is none.
- *
- * You'd think this would be easier than it actually is, since the unnamed data
- * stream should be the one referenced from the inode itself. Alas, if there
- * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
- * data stream in one of the alternate data streams instead of inside the WIM
- * dentry itself. So we need to check the alternate data streams too.
- *
- * Also, note that a dentry may appear to have more than one unnamed stream, but
- * if the SHA1 message digest is all 0's then the corresponding stream does not
- * really "count" (this is the case for the inode's own file stream when the
- * file stream that should be there is actually in one of the alternate stream
- * entries.). This is despite the fact that we may need to extract such a
- * missing entry as an empty file or empty named data stream.
- */
-struct wim_lookup_table_entry *
-inode_unnamed_lte(const struct wim_inode *inode,
- const struct wim_lookup_table *table)
-{
- if (inode->i_resolved)
- return inode_unnamed_lte_resolved(inode);
- else
- return inode_unnamed_lte_unresolved(inode, table);
-}
-
-/* Returns the SHA1 message digest of the unnamed data stream of a WIM inode, or
- * 'zero_hash' if the unnamed data stream is missing has all zeroes in its SHA1
- * message digest field. */
-const u8 *
-inode_unnamed_stream_hash(const struct wim_inode *inode)
-{
- const u8 *hash;
-
- for (unsigned i = 0; i <= inode->i_num_ads; i++) {
- if (inode_stream_name_nbytes(inode, i) == 0) {
- hash = inode_stream_hash(inode, i);
- if (!is_zero_hash(hash))
- return hash;
- }
- }
- return zero_hash;
-}
-
-struct wim_lookup_table_entry **
-retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
-{
- wimlib_assert(lte->unhashed);
- struct wim_inode *inode = lte->back_inode;
- u32 stream_id = lte->back_stream_id;
- if (stream_id == 0)
- return &inode->i_lte;
- else
- for (u16 i = 0; i < inode->i_num_ads; i++)
- if (inode->i_ads_entries[i].stream_id == stream_id)
- return &inode->i_ads_entries[i].lte;
- wimlib_assert(0);
- return NULL;
-}
+ return lte;
+}
/* Calculate the SHA1 message digest of a stream and move it from the list of
* unhashed streams to the stream lookup table, possibly joining it with an
return ret;
/* Look for a duplicate stream */
- duplicate_lte = lookup_resource(lookup_table, lte->hash);
+ duplicate_lte = lookup_stream(lookup_table, lte->hash);
list_del(<e->unhashed_list);
if (duplicate_lte) {
/* We have a duplicate stream. Transfer the reference counts
return 0;
}
-static int
-lte_clone_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
-{
- struct wim_lookup_table *lookup_table = _lookup_table;
-
- if (lookup_resource(lookup_table, lte->hash))
- return 0; /* Resource already present. */
-
- lte = clone_lookup_table_entry(lte);
- if (lte == NULL)
- return WIMLIB_ERR_NOMEM;
- lte->out_refcnt = 1;
- lookup_table_insert(lookup_table, lte);
- return 0;
-}
-
-static int
-lte_delete_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
-{
- struct wim_lookup_table *lookup_table = _lookup_table;
-
- if (lte->out_refcnt) {
- lookup_table_unlink(lookup_table, lte);
- free_lookup_table_entry(lte);
- }
- return 0;
-}
-
-/* API function documented in wimlib.h */
-WIMLIBAPI int
-wimlib_reference_resources(WIMStruct *wim,
- WIMStruct **resource_wims, unsigned num_resource_wims,
- int ref_flags)
+void
+lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
+ struct wimlib_resource_entry *wentry)
{
- int ret;
- unsigned i;
-
- if (wim == NULL)
- return WIMLIB_ERR_INVALID_PARAM;
-
- if (num_resource_wims != 0 && resource_wims == NULL)
- return WIMLIB_ERR_INVALID_PARAM;
-
- for (i = 0; i < num_resource_wims; i++)
- if (resource_wims[i] == NULL)
- return WIMLIB_ERR_INVALID_PARAM;
-
- for_lookup_table_entry(wim->lookup_table, lte_zero_out_refcnt, NULL);
+ memset(wentry, 0, sizeof(*wentry));
- for (i = 0; i < num_resource_wims; i++) {
- ret = for_lookup_table_entry(resource_wims[i]->lookup_table,
- lte_clone_if_new,
- wim->lookup_table);
- if (ret)
- goto out_rollback;
+ wentry->uncompressed_size = lte->size;
+ if (lte->resource_location == RESOURCE_IN_WIM) {
+ wentry->part_number = lte->rspec->wim->hdr.part_number;
+ if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
+ wentry->compressed_size = 0;
+ wentry->offset = lte->offset_in_res;
+ } else {
+ wentry->compressed_size = lte->rspec->size_in_wim;
+ wentry->offset = lte->rspec->offset_in_wim;
+ }
+ wentry->raw_resource_offset_in_wim = lte->rspec->offset_in_wim;
+ /*wentry->raw_resource_uncompressed_size = lte->rspec->uncompressed_size;*/
+ wentry->raw_resource_compressed_size = lte->rspec->size_in_wim;
}
- return 0;
-
-out_rollback:
- for_lookup_table_entry(wim->lookup_table, lte_delete_if_new,
- wim->lookup_table);
- return ret;
+ copy_hash(wentry->sha1_hash, lte->hash);
+ wentry->reference_count = lte->refcnt;
+ wentry->is_compressed = (lte->flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
+ wentry->is_metadata = (lte->flags & WIM_RESHDR_FLAG_METADATA) != 0;
+ wentry->is_free = (lte->flags & WIM_RESHDR_FLAG_FREE) != 0;
+ wentry->is_spanned = (lte->flags & WIM_RESHDR_FLAG_SPANNED) != 0;
+ wentry->packed = (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) != 0;
}
-static int
-reference_resource_paths(WIMStruct *wim,
- const tchar * const *resource_wimfiles,
- unsigned num_resource_wimfiles,
- int ref_flags,
- int open_flags,
- wimlib_progress_func_t progress_func)
-{
- WIMStruct **resource_wims;
- unsigned i;
- int ret;
-
- resource_wims = CALLOC(num_resource_wimfiles, sizeof(resource_wims[0]));
- if (!resource_wims)
- return WIMLIB_ERR_NOMEM;
-
- for (i = 0; i < num_resource_wimfiles; i++) {
- DEBUG("Referencing resources from path \"%"TS"\"",
- resource_wimfiles[i]);
- ret = wimlib_open_wim(resource_wimfiles[i], open_flags,
- &resource_wims[i], progress_func);
- if (ret)
- goto out_free_resource_wims;
- }
-
- ret = wimlib_reference_resources(wim, resource_wims,
- num_resource_wimfiles, ref_flags);
- if (ret)
- goto out_free_resource_wims;
-
- for (i = 0; i < num_resource_wimfiles; i++)
- list_add_tail(&resource_wims[i]->subwim_node, &wim->subwims);
-
- ret = 0;
- goto out_free_array;
-
-out_free_resource_wims:
- for (i = 0; i < num_resource_wimfiles; i++)
- wimlib_free(resource_wims[i]);
-out_free_array:
- FREE(resource_wims);
- return ret;
-}
+struct iterate_lte_context {
+ wimlib_iterate_lookup_table_callback_t cb;
+ void *user_ctx;
+};
static int
-reference_resource_glob(WIMStruct *wim, const tchar *refglob,
- int ref_flags, int open_flags,
- wimlib_progress_func_t progress_func)
+do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
{
- glob_t globbuf;
- int ret;
-
- /* Note: glob() is replaced in Windows native builds. */
- ret = tglob(refglob, GLOB_ERR | GLOB_NOSORT, NULL, &globbuf);
- if (ret) {
- if (ret == GLOB_NOMATCH) {
- if (ref_flags & WIMLIB_REF_FLAG_GLOB_ERR_ON_NOMATCH) {
- ERROR("Found no files for glob \"%"TS"\"", refglob);
- return WIMLIB_ERR_GLOB_HAD_NO_MATCHES;
- } else {
- return reference_resource_paths(wim,
- &refglob,
- 1,
- ref_flags,
- open_flags,
- progress_func);
- }
- } else {
- ERROR_WITH_ERRNO("Failed to process glob \"%"TS"\"", refglob);
- if (ret == GLOB_NOSPACE)
- return WIMLIB_ERR_NOMEM;
- else
- return WIMLIB_ERR_READ;
- }
- }
+ struct iterate_lte_context *ctx = _ctx;
+ struct wimlib_resource_entry entry;
- ret = reference_resource_paths(wim,
- (const tchar * const *)globbuf.gl_pathv,
- globbuf.gl_pathc,
- ref_flags,
- open_flags,
- progress_func);
- globfree(&globbuf);
- return ret;
+ lte_to_wimlib_resource_entry(lte, &entry);
+ return (*ctx->cb)(&entry, ctx->user_ctx);
}
/* API function documented in wimlib.h */
WIMLIBAPI int
-wimlib_reference_resource_files(WIMStruct *wim,
- const tchar * const * resource_wimfiles_or_globs,
- unsigned count,
- int ref_flags,
- int open_flags,
- wimlib_progress_func_t progress_func)
+wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
+ wimlib_iterate_lookup_table_callback_t cb,
+ void *user_ctx)
{
- unsigned i;
- int ret;
+ if (flags != 0)
+ return WIMLIB_ERR_INVALID_PARAM;
- if (ref_flags & WIMLIB_REF_FLAG_GLOB_ENABLE) {
- for (i = 0; i < count; i++) {
- ret = reference_resource_glob(wim,
- resource_wimfiles_or_globs[i],
- ref_flags,
- open_flags,
- progress_func);
+ struct iterate_lte_context ctx = {
+ .cb = cb,
+ .user_ctx = user_ctx,
+ };
+ if (wim->hdr.part_number == 1) {
+ int ret;
+ for (int i = 0; i < wim->hdr.image_count; i++) {
+ ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
+ &ctx);
if (ret)
return ret;
}
- return 0;
- } else {
- return reference_resource_paths(wim, resource_wimfiles_or_globs,
- count, ref_flags,
- open_flags, progress_func);
}
+ return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);
}