/*
* lookup_table.c
*
- * Lookup table, implemented as a hash table, that maps dentries to file
- * resources.
+ * Lookup table, implemented as a hash table, that maps SHA1 message digests to
+ * data streams; plus code to read and write the corresponding on-disk data.
*/
/*
- * Copyright (C) 2012 Eric Biggers
+ * Copyright (C) 2012, 2013 Eric Biggers
*
* This file is part of wimlib, a library for working with WIM files.
*
* along with wimlib; if not, see http://www.gnu.org/licenses/.
*/
-#include "wimlib_internal.h"
-#include "lookup_table.h"
-#include "io.h"
-#include <errno.h>
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#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/resource.h"
+#include "wimlib/util.h"
+#include "wimlib/write.h"
+#include <errno.h>
+#include <stdlib.h>
#ifdef WITH_FUSE
-#include <unistd.h>
+# include <unistd.h> /* for unlink() */
#endif
-struct lookup_table *new_lookup_table(size_t capacity)
+struct wim_lookup_table *
+new_lookup_table(size_t capacity)
{
- struct lookup_table *table;
+ struct wim_lookup_table *table;
struct hlist_head *array;
- table = MALLOC(sizeof(struct lookup_table));
- if (!table)
- goto err;
- array = CALLOC(capacity, sizeof(array[0]));
- if (!array) {
- FREE(table);
- goto err;
+ 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->num_entries = 0;
- table->capacity = capacity;
- table->array = array;
return table;
-err:
- ERROR("Failed to allocate memory for lookup table with capacity %zu",
- capacity);
- return NULL;
}
-struct lookup_table_entry *new_lookup_table_entry()
+struct wim_lookup_table_entry *
+new_lookup_table_entry(void)
{
- struct lookup_table_entry *lte;
+ struct wim_lookup_table_entry *lte;
- lte = CALLOC(1, sizeof(struct lookup_table_entry));
- if (lte) {
- lte->part_number = 1;
- lte->refcnt = 1;
- } else {
+ 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 lookup_table_entry));
+ sizeof(struct wim_lookup_table_entry));
+ return NULL;
}
+ lte->refcnt = 1;
+ BUILD_BUG_ON(RESOURCE_NONEXISTENT != 0);
return lte;
}
-struct lookup_table_entry *
-clone_lookup_table_entry(const struct lookup_table_entry *old)
+struct wim_lookup_table_entry *
+clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
{
- struct lookup_table_entry *new;
+ struct wim_lookup_table_entry *new;
- new = MALLOC(sizeof(*new));
- if (!new)
+ new = memdup(old, sizeof(struct wim_lookup_table_entry));
+ if (new == NULL)
return NULL;
- memcpy(new, old, sizeof(*old));
new->extracted_file = NULL;
switch (new->resource_location) {
- case RESOURCE_IN_STAGING_FILE:
+ case RESOURCE_IN_WIM:
+ list_add(&new->rspec_node, &new->rspec->stream_list);
+ break;
+
case RESOURCE_IN_FILE_ON_DISK:
- wimlib_assert((void*)&old->file_on_disk ==
- (void*)&old->staging_file_name);
- new->staging_file_name = STRDUP(old->staging_file_name);
- if (!new->staging_file_name)
+#ifdef __WIN32__
+ case RESOURCE_WIN32_ENCRYPTED:
+#endif
+#ifdef WITH_FUSE
+ case RESOURCE_IN_STAGING_FILE:
+ BUILD_BUG_ON((void*)&old->file_on_disk !=
+ (void*)&old->staging_file_name);
+#endif
+ new->file_on_disk = TSTRDUP(old->file_on_disk);
+ if (new->file_on_disk == NULL)
goto out_free;
break;
case RESOURCE_IN_ATTACHED_BUFFER:
- new->attached_buffer = MALLOC(wim_resource_size(old));
- if (!new->attached_buffer)
+ new->attached_buffer = memdup(old->attached_buffer, old->size);
+ if (new->attached_buffer == NULL)
goto out_free;
- memcpy(new->attached_buffer, old->attached_buffer,
- wim_resource_size(old));
break;
#ifdef WITH_NTFS_3G
case RESOURCE_IN_NTFS_VOLUME:
if (old->ntfs_loc) {
struct ntfs_location *loc;
- loc = MALLOC(sizeof(*loc));
- if (!loc)
+ loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location));
+ if (loc == NULL)
goto out_free;
- memcpy(loc, old->ntfs_loc, sizeof(*loc));
- loc->path_utf8 = NULL;
- loc->stream_name_utf16 = NULL;
+ loc->path = NULL;
+ loc->stream_name = NULL;
new->ntfs_loc = loc;
- loc->path_utf8 = STRDUP(old->ntfs_loc->path_utf8);
- if (!loc->path_utf8)
- goto out_free;
- loc->stream_name_utf16 = MALLOC(loc->stream_name_utf16_num_chars * 2);
- if (!loc->stream_name_utf16)
+ loc->path = STRDUP(old->ntfs_loc->path);
+ if (loc->path == NULL)
goto out_free;
- memcpy(loc->stream_name_utf16,
- old->ntfs_loc->stream_name_utf16,
- loc->stream_name_utf16_num_chars * 2);
+ if (loc->stream_name_nchars != 0) {
+ loc->stream_name = memdup(old->ntfs_loc->stream_name,
+ loc->stream_name_nchars * 2);
+ if (loc->stream_name == NULL)
+ goto out_free;
+ }
}
break;
#endif
return NULL;
}
-void free_lookup_table_entry(struct lookup_table_entry *lte)
+void
+free_lookup_table_entry(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:
- case RESOURCE_IN_FILE_ON_DISK:
- wimlib_assert(((void*)<e->file_on_disk ==
- (void*)<e->staging_file_name)
- && ((void*)<e->file_on_disk ==
- (void*)<e->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_utf8);
- FREE(lte->ntfs_loc->stream_name_utf16);
+ FREE(lte->ntfs_loc->path);
+ FREE(lte->ntfs_loc->stream_name);
FREE(lte->ntfs_loc);
}
break;
default:
break;
}
- FREE(lte->extracted_file);
FREE(lte);
}
}
-static int do_free_lookup_table_entry(struct lookup_table_entry *entry,
- void *ignore)
+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 lookup_table *table)
+void
+free_lookup_table(struct wim_lookup_table *table)
{
- DEBUG2("Freeing lookup table");
+ DEBUG("Freeing lookup table.");
if (table) {
if (table->array) {
for_lookup_table_entry(table,
}
}
+static void
+lookup_table_insert_raw(struct wim_lookup_table *table,
+ struct wim_lookup_table_entry *lte)
+{
+ size_t i = lte->hash_short % table->capacity;
+
+ hlist_add_head(<e->hash_list, &table->array[i]);
+}
+
+static void
+enlarge_lookup_table(struct wim_lookup_table *table)
+{
+ size_t old_capacity, new_capacity;
+ struct hlist_head *old_array, *new_array;
+ struct wim_lookup_table_entry *lte;
+ struct hlist_node *cur, *tmp;
+ size_t i;
+
+ old_capacity = table->capacity;
+ new_capacity = old_capacity * 2;
+ new_array = CALLOC(new_capacity, sizeof(struct hlist_head));
+ if (new_array == NULL)
+ return;
+ old_array = table->array;
+ table->array = new_array;
+ table->capacity = new_capacity;
+
+ for (i = 0; i < old_capacity; i++) {
+ hlist_for_each_entry_safe(lte, cur, tmp, &old_array[i], hash_list) {
+ hlist_del(<e->hash_list);
+ lookup_table_insert_raw(table, lte);
+ }
+ }
+ FREE(old_array);
+}
+
+
/*
* Inserts an entry into the lookup table.
*
* @table: A pointer to the lookup table.
- * @entry: A pointer to the entry to insert.
+ * @lte: A pointer to the entry to insert.
*/
-void lookup_table_insert(struct lookup_table *table,
- struct lookup_table_entry *lte)
+void
+lookup_table_insert(struct wim_lookup_table *table,
+ struct wim_lookup_table_entry *lte)
{
- size_t i = lte->hash_short % table->capacity;
- hlist_add_head(<e->hash_list, &table->array[i]);
-
- /* XXX Make the table grow when too many entries have been inserted. */
- table->num_entries++;
+ lookup_table_insert_raw(table, lte);
+ if (++table->num_entries > table->capacity)
+ enlarge_lookup_table(table);
}
-static void finalize_lte(struct lookup_table_entry *lte)
+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);
- wimlib_assert(lte->staging_list.next);
- wimlib_assert(lte->staging_list.prev);
- list_del(<e->staging_list);
+ list_del(<e->unhashed_list);
}
#endif
free_lookup_table_entry(lte);
* 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. */
-void lte_decrement_refcnt(struct lookup_table_entry *lte,
- struct lookup_table *table)
+void
+lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
+ struct wim_lookup_table *table)
{
wimlib_assert(lte != NULL);
wimlib_assert(lte->refcnt != 0);
if (--lte->refcnt == 0) {
- lookup_table_unlink(table, lte);
+ if (lte->unhashed)
+ list_del(<e->unhashed_list);
+ else
+ lookup_table_unlink(table, lte);
#ifdef WITH_FUSE
if (lte->num_opened_fds == 0)
#endif
}
#ifdef WITH_FUSE
-void lte_decrement_num_opened_fds(struct lookup_table_entry *lte)
+void
+lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
{
- wimlib_assert(lte != NULL);
- wimlib_assert(lte->num_opened_fds != 0);
- if (atomic_dec(<e->num_opened_fds) == 0 && lte->refcnt == 0)
- finalize_lte(lte);
+ if (lte->num_opened_fds != 0)
+ if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
+ finalize_lte(lte);
}
#endif
-/*
- * Calls a function on all the entries in the lookup table. Stop early and
- * return nonzero if any call to the function returns nonzero.
- */
-int for_lookup_table_entry(struct lookup_table *table,
- int (*visitor)(struct lookup_table_entry *, void *),
- void *arg)
+/* 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. */
+int
+for_lookup_table_entry(struct wim_lookup_table *table,
+ int (*visitor)(struct wim_lookup_table_entry *, void *),
+ void *arg)
{
- struct lookup_table_entry *lte;
+ struct wim_lookup_table_entry *lte;
struct hlist_node *pos, *tmp;
int ret;
hash_list)
{
ret = visitor(lte, arg);
- if (ret != 0)
+ if (ret)
return ret;
}
}
return 0;
}
+/* qsort() callback that sorts streams (represented by `struct
+ * wim_lookup_table_entry's) into an order optimized for reading.
+ *
+ * Sorting is done primarily by resource location, then secondarily by a
+ * per-resource location order. For example, resources in WIM files are sorted
+ * primarily by part number, then secondarily by offset, as to implement optimal
+ * reading of either a standalone or split WIM. */
+static int
+cmp_streams_by_sequential_order(const void *p1, const void *p2)
+{
+ const struct wim_lookup_table_entry *lte1, *lte2;
+ int v;
+ WIMStruct *wim1, *wim2;
+
+ lte1 = *(const struct wim_lookup_table_entry**)p1;
+ lte2 = *(const struct wim_lookup_table_entry**)p2;
-/*
- * Reads the lookup table from a WIM file.
- */
-int read_lookup_table(WIMStruct *w)
+ v = (int)lte1->resource_location - (int)lte2->resource_location;
+
+ /* Different resource locations? */
+ if (v)
+ return v;
+
+ switch (lte1->resource_location) {
+ case RESOURCE_IN_WIM:
+ wim1 = lte1->rspec->wim;
+ wim2 = lte2->rspec->wim;
+
+ /* Different (possibly split) WIMs? */
+ if (wim1 != wim2) {
+ v = memcmp(wim1->hdr.guid, wim2->hdr.guid, WIM_GID_LEN);
+ if (v)
+ return v;
+ }
+
+ /* Different part numbers in the same WIM? */
+ v = (int)wim1->hdr.part_number - (int)wim2->hdr.part_number;
+ if (v)
+ return v;
+
+ if (lte1->rspec->offset_in_wim != lte2->rspec->offset_in_wim)
+ return cmp_u64(lte1->rspec->offset_in_wim,
+ lte2->rspec->offset_in_wim);
+
+ return cmp_u64(lte1->offset_in_res, lte2->offset_in_res);
+
+ case RESOURCE_IN_FILE_ON_DISK:
+#ifdef WITH_FUSE
+ case RESOURCE_IN_STAGING_FILE:
+#endif
+#ifdef __WIN32__
+ case RESOURCE_WIN32_ENCRYPTED:
+#endif
+ /* Compare files by path: just a heuristic that will place files
+ * in the same directory next to each other. */
+ return tstrcmp(lte1->file_on_disk, lte2->file_on_disk);
+#ifdef WITH_NTFS_3G
+ case RESOURCE_IN_NTFS_VOLUME:
+ return tstrcmp(lte1->ntfs_loc->path, lte2->ntfs_loc->path);
+#endif
+ default:
+ /* No additional sorting order defined for this resource
+ * location (e.g. RESOURCE_IN_ATTACHED_BUFFER); simply compare
+ * everything equal to each other. */
+ return 0;
+ }
+}
+
+int
+sort_stream_list(struct list_head *stream_list,
+ size_t list_head_offset,
+ int (*compar)(const void *, const void*))
{
- u64 num_entries;
- u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
- int ret;
- struct lookup_table *table;
- struct lookup_table_entry *cur_entry = NULL, *duplicate_entry;
+ struct list_head *cur;
+ struct wim_lookup_table_entry **array;
+ size_t i;
+ size_t array_size;
+ size_t num_streams = 0;
- DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
- w->hdr.lookup_table_res_entry.offset,
- w->hdr.lookup_table_res_entry.original_size);
+ list_for_each(cur, stream_list)
+ num_streams++;
- if (fseeko(w->fp, w->hdr.lookup_table_res_entry.offset, SEEK_SET) != 0)
- {
- ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
- "lookup table",
- w->hdr.lookup_table_res_entry.offset);
- return WIMLIB_ERR_READ;
+ if (num_streams <= 1)
+ return 0;
+
+ array_size = num_streams * sizeof(array[0]);
+ array = MALLOC(array_size);
+ if (array == NULL)
+ return WIMLIB_ERR_NOMEM;
+
+ cur = stream_list->next;
+ for (i = 0; i < num_streams; i++) {
+ array[i] = (struct wim_lookup_table_entry*)((u8*)cur -
+ list_head_offset);
+ cur = cur->next;
}
- num_entries = w->hdr.lookup_table_res_entry.original_size /
- WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
- table = new_lookup_table(num_entries * 2 + 1);
- if (!table)
+ qsort(array, num_streams, sizeof(array[0]), compar);
+
+ INIT_LIST_HEAD(stream_list);
+ for (i = 0; i < num_streams; i++) {
+ list_add_tail((struct list_head*)
+ ((u8*)array[i] + list_head_offset),
+ stream_list);
+ }
+ FREE(array);
+ return 0;
+}
+
+/* Sort the specified list of streams in an order optimized for reading. */
+int
+sort_stream_list_by_sequential_order(struct list_head *stream_list,
+ size_t list_head_offset)
+{
+ return sort_stream_list(stream_list, list_head_offset,
+ cmp_streams_by_sequential_order);
+}
+
+
+static int
+add_lte_to_array(struct wim_lookup_table_entry *lte,
+ void *_pp)
+{
+ struct wim_lookup_table_entry ***pp = _pp;
+ *(*pp)++ = lte;
+ return 0;
+}
+
+/* Iterate through the lookup table entries, but first sort them by stream
+ * offset in the WIM. Caution: this is intended to be used when the stream
+ * offset field has actually been set. */
+int
+for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
+ int (*visitor)(struct wim_lookup_table_entry *,
+ void *),
+ void *arg)
+{
+ struct wim_lookup_table_entry **lte_array, **p;
+ size_t num_streams = table->num_entries;
+ int ret;
+
+ lte_array = MALLOC(num_streams * sizeof(lte_array[0]));
+ if (!lte_array)
return WIMLIB_ERR_NOMEM;
+ p = lte_array;
+ for_lookup_table_entry(table, add_lte_to_array, &p);
- while (num_entries--) {
- const u8 *p;
+ wimlib_assert(p == lte_array + num_streams);
+
+ qsort(lte_array, num_streams, sizeof(lte_array[0]),
+ cmp_streams_by_sequential_order);
+ ret = 0;
+ for (size_t i = 0; i < num_streams; i++) {
+ ret = visitor(lte_array[i], arg);
+ if (ret)
+ break;
+ }
+ FREE(lte_array);
+ return ret;
+}
+
+/* On-disk format of a WIM lookup table entry (stream entry). */
+struct wim_lookup_table_entry_disk {
+ /* Size, offset, and flags of the stream. */
+ struct wim_reshdr_disk reshdr;
+
+ /* Which part of the split WIM this stream is in; indexed from 1. */
+ le16 part_number;
+
+ /* Reference count of this stream over all WIM images. */
+ le32 refcnt;
+
+ /* SHA1 message digest of the uncompressed data of this stream, or
+ * optionally all zeroes if this stream is of zero length. */
+ u8 hash[SHA1_HASH_SIZE];
+} _packed_attribute;
+
+#define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
+
+/* Validate the size and location of a WIM resource. */
+static int
+validate_resource(const struct wim_resource_spec *rspec)
+{
+ struct wim_lookup_table_entry *lte;
+ u64 cur_offset;
+
+ /* Verify that calculating the offset of the end of the resource doesn't
+ * overflow. */
+ if (rspec->offset_in_wim + rspec->size_in_wim < rspec->size_in_wim)
+ goto invalid;
+
+ /* 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;
+ 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;
+
+ cur_offset = lte->offset_in_res + lte->size;
+ }
+ return 0;
+
+invalid:
+
+ ERROR("Invalid resource entry!");
+ return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
+}
+
+/*
+ * 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.
+ *
+ * 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).
+ *
+ * Return values:
+ * WIMLIB_ERR_SUCCESS (0)
+ * WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY
+ * WIMLIB_ERR_RESOURCE_NOT_FOUND
+ *
+ * Or an error code caused by failure to read the lookup table into memory.
+ */
+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;
+
+ DEBUG("Reading lookup table.");
+
+ /* Sanity check: lookup table entries are 50 bytes each. */
+ BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
+ WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
+
+ /* Calculate number of entries in the lookup table. */
+ num_entries = wim->hdr.lookup_table_reshdr.uncompressed_size /
+ sizeof(struct wim_lookup_table_entry_disk);
+
+ /* Read the lookup table into a buffer. */
+ ret = wim_reshdr_to_data(&wim->hdr.lookup_table_reshdr, wim, &buf);
+ if (ret)
+ goto out;
+
+ /* 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++) {
+ 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;
+
+ get_wim_reshdr(&disk_entry->reshdr, &reshdr);
+
+ DEBUG("reshdr: size_in_wim=%"PRIu64", "
+ "uncompressed_size=%"PRIu64", "
+ "offset_in_wim=%"PRIu64", "
+ "flags=0x%02x",
+ reshdr.size_in_wim, reshdr.uncompressed_size,
+ reshdr.offset_in_wim, reshdr.flags);
+
+ if (wim->hdr.wim_version == WIM_VERSION_DEFAULT)
+ reshdr.flags &= ~WIM_RESHDR_FLAG_PACKED_STREAMS;
- if (fread(buf, 1, sizeof(buf), w->fp) != sizeof(buf)) {
- if (feof(w->fp)) {
- ERROR("Unexpected EOF in WIM lookup table!");
- } else {
- ERROR_WITH_ERRNO("Error reading WIM lookup "
- "table");
- }
- ret = WIMLIB_ERR_READ;
- goto out;
- }
cur_entry = new_lookup_table_entry();
- if (!cur_entry) {
+ if (cur_entry == NULL) {
+ ERROR("Not enough memory to read lookup table!");
ret = WIMLIB_ERR_NOMEM;
- goto out;
+ goto err;
}
- cur_entry->wim = w;
- cur_entry->resource_location = RESOURCE_IN_WIM;
- p = get_resource_entry(buf, &cur_entry->resource_entry);
- p = get_u16(p, &cur_entry->part_number);
- p = get_u32(p, &cur_entry->refcnt);
- p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->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);
- if (cur_entry->part_number != w->hdr.part_number) {
- ERROR("A lookup table entry in part %hu of the WIM "
- "points to part %hu",
- w->hdr.part_number, cur_entry->part_number);
- ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
- goto out_free_cur_entry;
+ 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;
}
- if (is_zero_hash(cur_entry->hash)) {
- ERROR("The WIM lookup table contains an entry with a "
- "SHA1 message digest of all 0's");
- ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
- goto out_free_cur_entry;
- }
-
- /* Ordinarily, no two streams should share the same SHA1 message
- * digest. However, this constraint can be broken for metadata
- * resources--- two identical images will have the same metadata
- * resource, but their lookup table entries are not shared. */
- duplicate_entry = __lookup_resource(table, cur_entry->hash);
- if (duplicate_entry
- && !((duplicate_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
- && cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
+ 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 err;
+ }
+ }
+
+ 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)))
{
- ERROR("The WIM lookup table contains two entries with the "
- "same SHA1 message digest!");
- ERROR("The first entry is:");
- print_lookup_table_entry(duplicate_entry);
- ERROR("The second entry is:");
- print_lookup_table_entry(cur_entry);
- ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
- goto out_free_cur_entry;
- }
-
- if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
- && (cur_entry->resource_entry.size !=
- cur_entry->resource_entry.original_size))
+ /* Starting new run of streams that share the same WIM
+ * resource. */
+ struct wim_lookup_table_entry *prev_entry = NULL;
+
+ if (back_to_back_pack &&
+ !list_empty(&cur_rspec->stream_list))
+ {
+ prev_entry = list_entry(cur_rspec->stream_list.prev,
+ struct wim_lookup_table_entry,
+ rspec_node);
+ lte_unbind_wim_resource_spec(prev_entry);
+ }
+ if (cur_rspec != NULL) {
+ ret = validate_resource(cur_rspec);
+ if (ret)
+ goto err;
+ }
+
+ /* 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 err;
+ }
+ wim_res_hdr_to_spec(&reshdr, wim, cur_rspec);
+
+ if (prev_entry)
+ lte_bind_wim_resource_spec(prev_entry, cur_rspec);
+ }
+
+ if ((reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) &&
+ reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER)
{
- ERROR("Found uncompressed resource with original size "
- "not the same as compressed size");
- ERROR("The lookup table entry for the resource is as follows:");
- print_lookup_table_entry(cur_entry);
- ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
- goto out_free_cur_entry;
+ /* Found the specification for the packed resource.
+ * Transfer the values to the `struct
+ * wim_resource_spec', and discard the current stream
+ * since this lookup table entry did not, in fact,
+ * correspond to a "stream".
+ */
+
+ /* Uncompressed size of the resource pack 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 err;
+
+ 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_LZX != 1);
+ BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_XPRESS != 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;
+ }
+
+ if (is_zero_hash(cur_entry->hash)) {
+ free_lookup_table_entry(cur_entry);
+ continue;
+ }
+
+ if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
+ /* Continuing the pack with another stream. */
+ DEBUG("Continuing pack with stream: "
+ "%"PRIu64" uncompressed bytes @ "
+ "resource offset %"PRIu64")",
+ reshdr.size_in_wim, reshdr.offset_in_wim);
}
+
+ 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
+ * in-resource offset, not the in-WIM offset, and the
+ * size field is used for the uncompressed size, not the
+ * compressed size. */
+ cur_entry->offset_in_res = reshdr.offset_in_wim;
+ cur_entry->size = reshdr.size_in_wim;
+ cur_entry->flags = reshdr.flags;
+ } else {
+ /* Normal case: The stream corresponds one-to-one with
+ * the resource entry. */
+ cur_entry->offset_in_res = 0;
+ cur_entry->size = reshdr.uncompressed_size;
+ cur_entry->flags = reshdr.flags;
+ cur_rspec = NULL;
+ }
+
+ 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 err;
+ }
+
+ 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);
+ }
+ cur_entry = NULL;
+ /* Validate the last resource. */
+ if (cur_rspec != NULL) {
+ ret = validate_resource(cur_rspec);
+ if (ret)
+ goto err;
+ }
+
+ 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++)
+ put_image_metadata(wim->image_metadata[i], NULL);
+ wim->hdr.image_count = wim->current_image;
}
DEBUG("Done reading lookup table.");
- w->lookup_table = table;
- return 0;
-out_free_cur_entry:
- FREE(cur_entry);
-out:
+ wim->lookup_table = table;
+ ret = 0;
+ goto out_free_buf;
+
+err:
+ 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;
}
+static void
+put_wim_lookup_table_entry(struct wim_lookup_table_entry_disk *disk_entry,
+ const struct wim_reshdr *out_reshdr,
+ u16 part_number, u32 refcnt, const u8 *hash)
+{
+ put_wim_reshdr(out_reshdr, &disk_entry->reshdr);
+ disk_entry->part_number = cpu_to_le16(part_number);
+ disk_entry->refcnt = cpu_to_le32(refcnt);
+ copy_hash(disk_entry->hash, hash);
+}
-/*
- * Writes a lookup table entry to the output file.
- */
-int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out)
+int
+write_wim_lookup_table_from_stream_list(struct list_head *stream_list,
+ struct filedes *out_fd,
+ u16 part_number,
+ struct wim_reshdr *out_reshdr,
+ int write_resource_flags)
{
- FILE *out;
- u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
- u8 *p;
+ size_t table_size;
+ struct wim_lookup_table_entry *lte;
+ struct wim_lookup_table_entry_disk *table_buf;
+ struct wim_lookup_table_entry_disk *table_buf_ptr;
+ int ret;
+ u64 prev_res_offset_in_wim = ~0ULL;
- out = __out;
+ table_size = 0;
+ list_for_each_entry(lte, stream_list, lookup_table_list) {
+ table_size += sizeof(struct wim_lookup_table_entry_disk);
- /* Don't write entries that have not had file resources or metadata
- * resources written for them. */
- if (lte->out_refcnt == 0)
- return 0;
+ if (lte->out_reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
+ lte->out_res_offset_in_wim != prev_res_offset_in_wim)
+ {
+ table_size += sizeof(struct wim_lookup_table_entry_disk);
+ prev_res_offset_in_wim = lte->out_res_offset_in_wim;
+ }
+ }
- if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
- DEBUG("Writing metadata entry at %lu (orig size = %zu)",
- ftello(out), lte->output_resource_entry.original_size);
+ DEBUG("Writing WIM lookup table (size=%zu, offset=%"PRIu64")",
+ table_size, out_fd->offset);
- p = put_resource_entry(buf, <e->output_resource_entry);
- p = put_u16(p, lte->part_number);
- p = put_u32(p, lte->out_refcnt);
- p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
- if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
- ERROR_WITH_ERRNO("Failed to write lookup table entry");
- return WIMLIB_ERR_WRITE;
+ table_buf = MALLOC(table_size);
+ if (table_buf == NULL) {
+ ERROR("Failed to allocate %zu bytes for temporary lookup table",
+ table_size);
+ return WIMLIB_ERR_NOMEM;
}
- return 0;
-}
+ table_buf_ptr = table_buf;
+
+ prev_res_offset_in_wim = ~0ULL;
+ list_for_each_entry(lte, stream_list, lookup_table_list) {
+
+ put_wim_lookup_table_entry(table_buf_ptr++,
+ <e->out_reshdr,
+ part_number,
+ lte->out_refcnt,
+ lte->hash);
+ if (lte->out_reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
+ lte->out_res_offset_in_wim != prev_res_offset_in_wim)
+ {
+ /* Put the main resource entry for the pack. */
+
+ struct wim_reshdr reshdr;
+
+ reshdr.offset_in_wim = lte->out_res_offset_in_wim;
+ reshdr.size_in_wim = lte->out_res_size_in_wim;
+ reshdr.uncompressed_size = WIM_PACK_MAGIC_NUMBER;
+ reshdr.flags = WIM_RESHDR_FLAG_PACKED_STREAMS;
+
+ DEBUG("Putting main entry for pack: "
+ "size_in_wim=%"PRIu64", "
+ "offset_in_wim=%"PRIu64", "
+ "uncompressed_size=%"PRIu64,
+ reshdr.size_in_wim,
+ reshdr.offset_in_wim,
+ reshdr.uncompressed_size);
+
+ put_wim_lookup_table_entry(table_buf_ptr++,
+ &reshdr,
+ part_number,
+ 1, zero_hash);
+ prev_res_offset_in_wim = lte->out_res_offset_in_wim;
+ }
+ }
+ wimlib_assert((u8*)table_buf_ptr - (u8*)table_buf == table_size);
+
+ /* Write the lookup table uncompressed. Although wimlib can handle a
+ * compressed lookup table, MS software cannot. */
+ ret = write_wim_resource_from_buffer(table_buf,
+ table_size,
+ WIM_RESHDR_FLAG_METADATA,
+ out_fd,
+ WIMLIB_COMPRESSION_TYPE_NONE,
+ 0,
+ out_reshdr,
+ NULL,
+ write_resource_flags);
+ FREE(table_buf);
+ DEBUG("ret=%d", ret);
+ return ret;
+}
-int lte_zero_real_refcnt(struct lookup_table_entry *lte, void *ignore)
+int
+lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
{
lte->real_refcnt = 0;
return 0;
}
-int lte_zero_out_refcnt(struct lookup_table_entry *lte, void *ignore)
+int
+lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
{
lte->out_refcnt = 0;
return 0;
}
-int lte_free_extracted_file(struct lookup_table_entry *lte, void *ignone)
+int
+lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
{
- FREE(lte->extracted_file);
- lte->extracted_file = NULL;
+ if (lte->extracted_file != NULL) {
+ FREE(lte->extracted_file);
+ lte->extracted_file = NULL;
+ }
return 0;
}
-void print_lookup_table_entry(const struct lookup_table_entry *lte)
+void
+print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
{
- if (!lte) {
- putchar('\n');
+ if (lte == NULL) {
+ tputc(T('\n'), out);
return;
}
- printf("Offset = %"PRIu64" bytes\n",
- lte->resource_entry.offset);
- printf("Size = %"PRIu64" bytes\n",
- (u64)lte->resource_entry.size);
- printf("Original size = %"PRIu64" bytes\n",
- lte->resource_entry.original_size);
- printf("Part Number = %hu\n", lte->part_number);
- printf("Reference Count = %u\n", lte->refcnt);
- printf("Hash = 0x");
- print_hash(lte->hash);
- putchar('\n');
- printf("Flags = ");
- u8 flags = lte->resource_entry.flags;
+
+
+ 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)
- fputs("WIM_RESHDR_FLAG_COMPRESSED, ", stdout);
+ tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
if (flags & WIM_RESHDR_FLAG_FREE)
- fputs("WIM_RESHDR_FLAG_FREE, ", stdout);
+ tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
if (flags & WIM_RESHDR_FLAG_METADATA)
- fputs("WIM_RESHDR_FLAG_METADATA, ", stdout);
+ tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
if (flags & WIM_RESHDR_FLAG_SPANNED)
- fputs("WIM_RESHDR_FLAG_SPANNED, ", stdout);
- putchar('\n');
+ 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->wim->filename) {
- printf("WIM file = `%s'\n",
- lte->wim->filename);
+ 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:
- printf("File on Disk = `%s'\n", lte->file_on_disk);
+ tfprintf(out, T("File on Disk = `%"TS"'\n"),
+ lte->file_on_disk);
break;
+#ifdef WITH_FUSE
case RESOURCE_IN_STAGING_FILE:
- printf("Staging File = `%s'\n", lte->staging_file_name);
+ tfprintf(out, T("Staging File = `%"TS"'\n"),
+ lte->staging_file_name);
break;
+#endif
default:
break;
}
- putchar('\n');
+ tputc(T('\n'), out);
}
-static int do_print_lookup_table_entry(struct lookup_table_entry *lte,
- void *ignore)
+void
+lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
+ struct wimlib_resource_entry *wentry)
{
- print_lookup_table_entry(lte);
- return 0;
+ 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;
}
-/*
- * Prints the lookup table of a WIM file.
- */
-WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
+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)
{
- for_lookup_table_entry(w->lookup_table,
- do_print_lookup_table_entry,
- NULL);
+ 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);
}
-/*
- * Looks up an entry in the lookup table.
- */
-struct lookup_table_entry *
-__lookup_resource(const struct lookup_table *table, const u8 hash[])
+/* 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 lookup_table_entry *lte;
+ struct wim_lookup_table_entry *lte;
struct hlist_node *pos;
wimlib_assert(table != NULL);
*
* This is only for pre-resolved inodes.
*/
-int lookup_resource(WIMStruct *w, const char *path,
- int lookup_flags,
- struct dentry **dentry_ret,
- struct lookup_table_entry **lte_ret,
- u16 *stream_idx_ret)
-{
- struct dentry *dentry;
- struct lookup_table_entry *lte;
+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 char *stream_name = NULL;
- struct inode *inode;
- char *p = NULL;
+ 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 = (char*)stream_name - 1;
- *p = '\0';
+ p = (tchar*)stream_name - 1;
+ *p = T('\0');
}
}
- dentry = get_dentry(w, path);
+ dentry = get_dentry(wim, path, WIMLIB_CASE_SENSITIVE);
if (p)
- *p = ':';
+ *p = T(':');
if (!dentry)
- return -ENOENT;
+ return -errno;
inode = dentry->d_inode;
- wimlib_assert(inode->resolved);
+ 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 ads_entry *ads_entry;
+ struct wim_ads_entry *ads_entry;
u16 ads_idx;
ads_entry = inode_get_ads_entry(inode, stream_name,
&ads_idx);
return -ENOENT;
}
} else {
- lte = inode->lte;
- stream_idx = 0;
+ lte = inode_unnamed_stream_resolved(inode, &stream_idx);
}
out:
if (dentry_ret)
}
#endif
-static void inode_resolve_ltes(struct inode *inode, struct lookup_table *table)
+int
+resource_not_found_error(const struct wim_inode *inode, const u8 *hash)
{
- struct lookup_table_entry *lte;
-
- wimlib_assert(!inode->resolved);
+ 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 the default file stream */
- lte = __lookup_resource(table, inode->hash);
- inode->lte = lte;
- inode->resolved = true;
+/*
+ * 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 */
- for (u16 i = 0; i < inode->num_ads; i++) {
- struct ads_entry *cur_entry = &inode->ads_entries[i];
- lte = __lookup_resource(table, cur_entry->hash);
- cur_entry->lte = lte;
+ /* 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);
}
-static void inode_unresolve_ltes(struct inode *inode)
+void
+inode_unresolve_ltes(struct wim_inode *inode)
{
- wimlib_assert(inode->resolved);
- if (inode->lte)
- copy_hash(inode->hash, inode->lte->hash);
- else
- zero_out_hash(inode->hash);
-
- for (u16 i = 0; i < inode->num_ads; i++) {
- if (inode->ads_entries[i].lte)
- copy_hash(inode->ads_entries[i].hash,
- inode->ads_entries[i].lte->hash);
+ if (inode->i_resolved) {
+ if (inode->i_lte)
+ copy_hash(inode->i_hash, inode->i_lte->hash);
else
- zero_out_hash(inode->ads_entries[i].hash);
+ 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;
}
- inode->resolved = false;
}
-/* Resolve a dentry'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. A circular
- * linked list of streams sharing the same lookup table entry is created.
+/*
+ * 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 function always succeeds; unresolved lookup table entries are given a
- * NULL pointer.
+ * This works for both resolved and un-resolved inodes.
*/
-int dentry_resolve_ltes(struct dentry *dentry, void *table)
+struct wim_lookup_table_entry *
+inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
+ const struct wim_lookup_table *table)
{
- if (!dentry->d_inode->resolved)
- inode_resolve_ltes(dentry->d_inode, table);
- return 0;
+ if (inode->i_resolved)
+ return inode_stream_lte_resolved(inode, stream_idx);
+ else
+ return inode_stream_lte_unresolved(inode, stream_idx, table);
}
-int dentry_unresolve_ltes(struct dentry *dentry, void *ignore)
+struct wim_lookup_table_entry *
+inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret)
{
- if (dentry->d_inode->resolved)
- inode_unresolve_ltes(dentry->d_inode);
- return 0;
+ 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
* 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 lookup_table_entry *
-inode_unnamed_lte(const struct inode *inode,
- const struct lookup_table *table)
+struct wim_lookup_table_entry *
+inode_unnamed_lte(const struct wim_inode *inode,
+ const struct wim_lookup_table *table)
{
- if (inode->resolved)
+ 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;
+}
+
+/* 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
+ * existing lookup table entry for an identical stream.
+ *
+ * @lte: An unhashed lookup table entry.
+ * @lookup_table: Lookup table for the WIM.
+ * @lte_ret: On success, write a pointer to the resulting lookup table
+ * entry to this location. This will be the same as @lte
+ * if it was inserted into the lookup table, or different if
+ * a duplicate stream was found.
+ *
+ * Returns 0 on success; nonzero if there is an error reading the stream.
+ */
+int
+hash_unhashed_stream(struct wim_lookup_table_entry *lte,
+ struct wim_lookup_table *lookup_table,
+ struct wim_lookup_table_entry **lte_ret)
+{
+ int ret;
+ struct wim_lookup_table_entry *duplicate_lte;
+ struct wim_lookup_table_entry **back_ptr;
+
+ wimlib_assert(lte->unhashed);
+
+ /* back_ptr must be saved because @back_inode and @back_stream_id are in
+ * union with the SHA1 message digest and will no longer be valid once
+ * the SHA1 has been calculated. */
+ back_ptr = retrieve_lte_pointer(lte);
+
+ ret = sha1_stream(lte);
+ if (ret)
+ return ret;
+
+ /* Look for a duplicate stream */
+ duplicate_lte = lookup_resource(lookup_table, lte->hash);
+ list_del(<e->unhashed_list);
+ if (duplicate_lte) {
+ /* We have a duplicate stream. Transfer the reference counts
+ * from this stream to the duplicate and update the reference to
+ * this stream (in an inode or ads_entry) to point to the
+ * duplicate. The caller is responsible for freeing @lte if
+ * needed. */
+ wimlib_assert(!(duplicate_lte->unhashed));
+ wimlib_assert(duplicate_lte->size == lte->size);
+ duplicate_lte->refcnt += lte->refcnt;
+ lte->refcnt = 0;
+ *back_ptr = duplicate_lte;
+ lte = duplicate_lte;
+ } else {
+ /* No duplicate stream, so we need to insert this stream into
+ * the lookup table and treat it as a hashed stream. */
+ lookup_table_insert(lookup_table, lte);
+ lte->unhashed = 0;
+ }
+ *lte_ret = lte;
+ 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)
+{
+ 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);
+
+ 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;
+ }
+ return 0;
+
+out_rollback:
+ for_lookup_table_entry(wim->lookup_table, lte_delete_if_new,
+ wim->lookup_table);
+ return ret;
+}
+
+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;
+}
+
+static int
+reference_resource_glob(WIMStruct *wim, const tchar *refglob,
+ int ref_flags, int open_flags,
+ wimlib_progress_func_t progress_func)
+{
+ 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;
+ }
+ }
+
+ ret = reference_resource_paths(wim,
+ (const tchar * const *)globbuf.gl_pathv,
+ globbuf.gl_pathc,
+ ref_flags,
+ open_flags,
+ progress_func);
+ globfree(&globbuf);
+ return ret;
+}
+
+/* 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)
+{
+ unsigned i;
+ int ret;
+
+ 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);
+ if (ret)
+ return ret;
+ }
+ return 0;
+ } else {
+ return reference_resource_paths(wim, resource_wimfiles_or_globs,
+ count, ref_flags,
+ open_flags, progress_func);
+ }
+}
git://wimlib.net / wimlib / blobdiff