/*
* 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.
*/
/*
#include "wimlib_internal.h"
#include "lookup_table.h"
-#include "io.h"
+#include "buffer_io.h"
#include <errno.h>
-struct lookup_table *new_lookup_table(size_t capacity)
+#ifdef WITH_FUSE
+#include <unistd.h>
+#endif
+
+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 = MALLOC(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()
{
- struct lookup_table_entry *lte;
-
- lte = CALLOC(1, sizeof(struct lookup_table_entry));
- if (!lte) {
+ struct wim_lookup_table_entry *lte;
+
+ lte = CALLOC(1, sizeof(struct wim_lookup_table_entry));
+ if (lte) {
+ lte->part_number = 1;
+ lte->refcnt = 1;
+ } else {
ERROR("Out of memory (tried to allocate %zu bytes for "
"lookup table entry)",
- sizeof(struct lookup_table_entry));
- return NULL;
+ sizeof(struct wim_lookup_table_entry));
}
-
- lte->part_number = 1;
- lte->refcnt = 1;
- INIT_LIST_HEAD(<e->lte_group_list);
return lte;
}
-void free_lookup_table_entry(struct lookup_table_entry *lte)
+struct wim_lookup_table_entry *
+clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
{
- if (lte) {
-#ifdef WITH_FUSE
- if (lte->staging_list.next)
- list_del(<e->staging_list);
+ struct wim_lookup_table_entry *new;
+
+ new = MALLOC(sizeof(*new));
+ if (!new)
+ return NULL;
+
+ memcpy(new, old, sizeof(*old));
+ new->extracted_file = NULL;
+ switch (new->resource_location) {
+ case RESOURCE_IN_STAGING_FILE:
+ case RESOURCE_IN_FILE_ON_DISK:
+ BUILD_BUG_ON((void*)&old->file_on_disk !=
+ (void*)&old->staging_file_name);
+ new->staging_file_name = STRDUP(old->staging_file_name);
+ if (!new->staging_file_name)
+ goto out_free;
+ break;
+ case RESOURCE_IN_ATTACHED_BUFFER:
+ new->attached_buffer = MALLOC(wim_resource_size(old));
+ if (!new->attached_buffer)
+ 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)
+ goto out_free;
+ memcpy(loc, old->ntfs_loc, sizeof(*loc));
+ loc->path_utf8 = NULL;
+ loc->stream_name_utf16 = 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)
+ goto out_free;
+ memcpy(loc->stream_name_utf16,
+ old->ntfs_loc->stream_name_utf16,
+ loc->stream_name_utf16_num_chars * 2);
+ }
+ break;
#endif
+ default:
+ break;
+ }
+ return new;
+out_free:
+ free_lookup_table_entry(new);
+ return NULL;
+}
+
+void free_lookup_table_entry(struct wim_lookup_table_entry *lte)
+{
+ if (lte) {
switch (lte->resource_location) {
case RESOURCE_IN_STAGING_FILE:
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->staging_file_name);
+ BUILD_BUG_ON((void*)<e->file_on_disk !=
+ (void*)<e->attached_buffer);
FREE(lte->file_on_disk);
break;
#ifdef WITH_NTFS_3G
default:
break;
}
- FREE(lte->extracted_file);
FREE(lte);
}
}
-static int do_free_lookup_table_entry(struct lookup_table_entry *entry,
+static int do_free_lookup_table_entry(struct wim_lookup_table_entry *entry,
void *ignore)
{
free_lookup_table_entry(entry);
}
-void free_lookup_table(struct lookup_table *table)
+void free_lookup_table(struct wim_lookup_table *table)
{
DEBUG2("Freeing lookup table");
if (table) {
* @table: A pointer to the lookup table.
* @entry: 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]);
table->num_entries++;
}
-
+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->staging_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. */
-struct lookup_table_entry *
-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)
{
- if (lte) {
- wimlib_assert(lte->refcnt);
- if (--lte->refcnt == 0) {
- lookup_table_unlink(table, lte);
- #ifdef WITH_FUSE
- if (lte->num_opened_fds == 0)
- #endif
- {
- free_lookup_table_entry(lte);
- lte = NULL;
- }
- }
+ wimlib_assert(lte != NULL);
+ wimlib_assert(lte->refcnt != 0);
+ if (--lte->refcnt == 0) {
+ lookup_table_unlink(table, lte);
+ #ifdef WITH_FUSE
+ if (lte->num_opened_fds == 0)
+ #endif
+ finalize_lte(lte);
}
- return lte;
}
-/*
- * 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 *),
+#ifdef WITH_FUSE
+void lte_decrement_num_opened_fds(struct wim_lookup_table_entry *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 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;
*/
int read_lookup_table(WIMStruct *w)
{
- 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;
+ u64 num_entries;
+ u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
+ int ret;
+ struct wim_lookup_table *table;
+ struct wim_lookup_table_entry *cur_entry = NULL, *duplicate_entry;
+
+ if (resource_is_compressed(&w->hdr.lookup_table_res_entry)) {
+ ERROR("Didn't expect a compressed lookup table!");
+ ERROR("Ask the author to implement support for this.");
+ return WIMLIB_ERR_COMPRESSED_LOOKUP_TABLE;
+ }
DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
w->hdr.lookup_table_res_entry.offset,
w->hdr.lookup_table_res_entry.original_size);
- if (fseeko(w->fp, w->hdr.lookup_table_res_entry.offset, SEEK_SET) != 0) {
+ 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);
}
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);
w->hdr.part_number, cur_entry->part_number);
ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
goto out_free_cur_entry;
-
}
if (is_zero_hash(cur_entry->hash)) {
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) {
+ if (duplicate_entry
+ && !((duplicate_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
+ && cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
+ {
ERROR("The WIM lookup table contains two entries with the "
"same SHA1 message digest!");
ERROR("The first entry is:");
if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
&& (cur_entry->resource_entry.size !=
- cur_entry->resource_entry.original_size))
+ cur_entry->resource_entry.original_size))
{
ERROR("Found uncompressed resource with original size "
"not the same as compressed size");
ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
goto out_free_cur_entry;
}
+ if ((cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
+ && cur_entry->refcnt != 1)
+ {
+ ERROR("Found metadata resource with refcnt != 1:");
+ print_lookup_table_entry(cur_entry);
+ ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
+ goto out_free_cur_entry;
+ }
lookup_table_insert(table, cur_entry);
}
}
-/*
+/*
* Writes a lookup table entry to the output file.
*/
-int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out)
+int write_lookup_table_entry(struct wim_lookup_table_entry *lte, void *__out)
{
FILE *out;
u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
return 0;
}
+/* Writes the lookup table to the output file. */
+int write_lookup_table(struct wim_lookup_table *table, FILE *out,
+ struct resource_entry *out_res_entry)
+{
+ off_t start_offset, end_offset;
+ int ret;
+
+ start_offset = ftello(out);
+ if (start_offset == -1)
+ return WIMLIB_ERR_WRITE;
+
+ ret = for_lookup_table_entry(table, write_lookup_table_entry, out);
+ if (ret != 0)
+ return ret;
+
+ end_offset = ftello(out);
+ if (end_offset == -1)
+ return WIMLIB_ERR_WRITE;
+
+ out_res_entry->offset = start_offset;
+ out_res_entry->size = end_offset - start_offset;
+ out_res_entry->original_size = end_offset - start_offset;
+ out_res_entry->flags = WIM_RESHDR_FLAG_METADATA;
+
+ return 0;
+}
+
+
+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 wim_lookup_table_entry *lte, void *ignore)
+{
+ lte->out_refcnt = 0;
+ return 0;
+}
-int zero_out_refcnts(struct lookup_table_entry *entry, void *ignore)
+int lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *ignore)
{
- entry->out_refcnt = 0;
+ 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)
{
if (!lte) {
putchar('\n');
return;
}
- printf("Offset = %"PRIu64" bytes\n",
+ printf("Offset = %"PRIu64" bytes\n",
lte->resource_entry.offset);
- printf("Size = %"PRIu64" bytes\n",
+ printf("Size = %"PRIu64" bytes\n",
(u64)lte->resource_entry.size);
- printf("Original size = %"PRIu64" bytes\n",
+ 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);
putchar('\n');
}
-static int do_print_lookup_table_entry(struct lookup_table_entry *lte,
+static int do_print_lookup_table_entry(struct wim_lookup_table_entry *lte,
void *ignore)
{
print_lookup_table_entry(lte);
}
/*
- * Prints the lookup table of a WIM file.
+ * Prints the lookup table of a WIM file.
*/
WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
{
- for_lookup_table_entry(w->lookup_table,
+ for_lookup_table_entry(w->lookup_table,
do_print_lookup_table_entry,
NULL);
}
-/*
- * Looks up an entry in the lookup table.
- */
-struct lookup_table_entry *
-__lookup_resource(const struct lookup_table *table, const u8 hash[])
+/* 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);
+ 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 NULL;
}
-/*
+#ifdef WITH_FUSE
+/*
* Finds the dentry, lookup table entry, and stream index for a WIM file stream,
* given a path name.
*
*/
int lookup_resource(WIMStruct *w, const char *path,
int lookup_flags,
- struct dentry **dentry_ret,
- struct lookup_table_entry **lte_ret,
- unsigned *stream_idx_ret)
+ struct wim_dentry **dentry_ret,
+ struct wim_lookup_table_entry **lte_ret,
+ u16 *stream_idx_ret)
{
- struct dentry *dentry;
- struct lookup_table_entry *lte;
- unsigned stream_idx;
+ struct wim_dentry *dentry;
+ struct wim_lookup_table_entry *lte;
+ u16 stream_idx;
const char *stream_name = NULL;
- struct inode *inode;
+ struct wim_inode *inode;
char *p = NULL;
if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
if (!dentry)
return -ENOENT;
- inode = dentry->inode;
+ inode = dentry->d_inode;
- wimlib_assert(inode->resolved);
+ wimlib_assert(inode->i_resolved);
if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
&& inode_is_directory(inode))
return -EISDIR;
- lte = inode->lte;
- stream_idx = 0;
if (stream_name) {
- size_t stream_name_len = strlen(stream_name);
- for (u16 i = 0; i < inode->num_ads; i++) {
- if (ads_entry_has_name(inode->ads_entries[i],
- stream_name,
- stream_name_len))
- {
- stream_idx = i + 1;
- lte = inode->ads_entries[i]->lte;
- goto out;
- }
+ 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;
}
- return -ENOENT;
+ } else {
+ lte = inode->i_lte;
+ stream_idx = 0;
}
out:
if (dentry_ret)
*stream_idx_ret = stream_idx;
return 0;
}
+#endif
-static int inode_resolve_ltes(struct inode *inode, struct lookup_table *table)
-{
- struct lookup_table_entry *lte;
-
- /* Resolve the default file stream */
- lte = __lookup_resource(table, inode->hash);
- if (lte)
- list_add(&inode->lte_group_list.list, <e->lte_group_list);
- else
- INIT_LIST_HEAD(&inode->lte_group_list.list);
- inode->lte = lte;
- inode->lte_group_list.type = STREAM_TYPE_NORMAL;
- inode->resolved = true;
-
- /* 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);
- if (lte)
- list_add(&cur_entry->lte_group_list.list,
- <e->lte_group_list);
- else
- INIT_LIST_HEAD(&cur_entry->lte_group_list.list);
- cur_entry->lte = lte;
- cur_entry->lte_group_list.type = STREAM_TYPE_ADS;
- }
- return 0;
-}
-
-/* Resolve a dentry's lookup table entries
+/* 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. A circular
* This function always succeeds; unresolved lookup table entries are given a
* NULL pointer.
*/
-int dentry_resolve_ltes(struct dentry *dentry, void *table)
+void inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table)
{
- if (dentry->inode->resolved)
- return 0;
- else
- return inode_resolve_ltes(dentry->inode, table);
+
+ if (!inode->i_resolved) {
+ struct wim_lookup_table_entry *lte;
+ /* Resolve the default file stream */
+ lte = __lookup_resource(table, inode->i_hash);
+ inode->i_lte = lte;
+ inode->i_resolved = 1;
+
+ /* Resolve the alternate data streams */
+ for (u16 i = 0; i < inode->i_num_ads; i++) {
+ struct wim_ads_entry *cur_entry = &inode->i_ads_entries[i];
+ lte = __lookup_resource(table, cur_entry->hash);
+ cur_entry->lte = lte;
+ }
+ }
}
+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 dentries.
+ */
+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);
+}
-/* Return the lookup table entry for the unnamed data stream of a inode, or
+/* 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
- * inode. So we need to check the alternate data streams too.
+ * 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 inode may appear to have than one unnamed stream, but if
- * the SHA1 message digest is all 0's then the corresponding stream does not
+ * 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 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);
}
+static int lte_add_stream_size(struct wim_lookup_table_entry *lte,
+ void *total_bytes_p)
+{
+ *(u64*)total_bytes_p += lte->resource_entry.size;
+ return 0;
+}
+
+u64 lookup_table_total_stream_size(struct wim_lookup_table *table)
+{
+ u64 total_size = 0;
+ for_lookup_table_entry(table, lte_add_stream_size, &total_size);
+ return total_size;
+}
git://wimlib.net / wimlib / blobdiff