#include <time.h>
#include <unistd.h>
-#include "wimlib_internal.h"
#include "dentry.h"
#include "io.h"
-#include "timestamp.h"
#include "lookup_table.h"
#include "sha1.h"
+#include "timestamp.h"
+#include "wimlib_internal.h"
/*
* Returns true if @dentry has the UTF-8 file name @name that has length
{
u64 length = (dentry->length + 7) & ~7;
for (u16 i = 0; i < dentry->num_ads; i++)
- length += ads_entry_length(&dentry->ads_entries[i]);
+ length += ads_entry_total_length(&dentry->ads_entries[i]);
return length;
}
}
-/* Makes all timestamp fields for the dentry be the current time. */
+/* Sets all the timestamp fields of the dentry to the current time. */
void dentry_update_all_timestamps(struct dentry *dentry)
{
u64 now = get_wim_timestamp();
dentry->last_write_time = now;
}
+/* Returns the alternate data stream entry belonging to @dentry that has the
+ * stream name @stream_name. */
struct ads_entry *dentry_get_ads_entry(struct dentry *dentry,
const char *stream_name)
{
ads_entry->lte_group_list.type = STREAM_TYPE_ADS;
}
-/* Add an alternate stream entry to a dentry and return a pointer to it, or NULL
- * on failure. */
+/*
+ * Add an alternate stream entry to a dentry and return a pointer to it, or NULL
+ * if memory could not be allocated.
+ */
struct ads_entry *dentry_add_ads(struct dentry *dentry, const char *stream_name)
{
u16 num_ads;
return new_entry;
}
+/* Remove an alternate data stream from a dentry.
+ *
+ * The corresponding lookup table entry for the stream is NOT changed.
+ *
+ * @dentry: The dentry
+ * @ads_entry: The alternate data stream entry (it MUST be one of the
+ * ads_entry's in the array dentry->ads_entries).
+ */
void dentry_remove_ads(struct dentry *dentry, struct ads_entry *ads_entry)
{
u16 idx;
child = dentry->children;
dentry->subdir_offset = *subdir_offset_p;
- if (child) {
+ if (child) {
/* Advance the subdir offset by the amount of space the children
* of this dentry take up. */
do {
} else {
/* On disk, childless directories have a valid subdir_offset
* that points to an 8-byte end-of-directory dentry. Regular
- * files have a subdir_offset of 0. */
+ * files or reparse points have a subdir_offset of 0. */
if (dentry_is_directory(dentry))
*subdir_offset_p += 8;
else
/* Returns the child of @dentry that has the file name @name.
* Returns NULL if no child has the name. */
struct dentry *get_dentry_child_with_name(const struct dentry *dentry,
- const char *name)
+ const char *name)
{
struct dentry *child;
size_t name_len;
/* Retrieves the dentry that has the UTF-8 @path relative to the dentry
* @cur_dir. Returns NULL if no dentry having the path is found. */
-static struct dentry *get_dentry_relative_path(struct dentry *cur_dir, const char *path)
+static struct dentry *get_dentry_relative_path(struct dentry *cur_dir,
+ const char *path)
{
struct dentry *child;
size_t base_len;
return get_dentry_relative_path(root, path);
}
-/* Returns the parent directory for the @path. */
+/* Returns the dentry that corresponds to the parent directory of @path, or NULL
+ * if the dentry is not found. */
struct dentry *get_parent_dentry(WIMStruct *w, const char *path)
{
size_t path_len = strlen(path);
return 0;
}
+/* We want to be able to show the names of the file attribute flags that are
+ * set. */
struct file_attr_flag {
u32 flag;
const char *name;
};
struct file_attr_flag file_attr_flags[] = {
- {FILE_ATTRIBUTE_READONLY, "READONLY"},
- {FILE_ATTRIBUTE_HIDDEN, "HIDDEN"},
- {FILE_ATTRIBUTE_SYSTEM, "SYSTEM"},
- {FILE_ATTRIBUTE_DIRECTORY, "DIRECTORY"},
- {FILE_ATTRIBUTE_ARCHIVE, "ARCHIVE"},
- {FILE_ATTRIBUTE_DEVICE, "DEVICE"},
- {FILE_ATTRIBUTE_NORMAL, "NORMAL"},
- {FILE_ATTRIBUTE_TEMPORARY, "TEMPORARY"},
- {FILE_ATTRIBUTE_SPARSE_FILE, "SPARSE_FILE"},
- {FILE_ATTRIBUTE_REPARSE_POINT, "REPARSE_POINT"},
- {FILE_ATTRIBUTE_COMPRESSED, "COMPRESSED"},
- {FILE_ATTRIBUTE_OFFLINE, "OFFLINE"},
+ {FILE_ATTRIBUTE_READONLY, "READONLY"},
+ {FILE_ATTRIBUTE_HIDDEN, "HIDDEN"},
+ {FILE_ATTRIBUTE_SYSTEM, "SYSTEM"},
+ {FILE_ATTRIBUTE_DIRECTORY, "DIRECTORY"},
+ {FILE_ATTRIBUTE_ARCHIVE, "ARCHIVE"},
+ {FILE_ATTRIBUTE_DEVICE, "DEVICE"},
+ {FILE_ATTRIBUTE_NORMAL, "NORMAL"},
+ {FILE_ATTRIBUTE_TEMPORARY, "TEMPORARY"},
+ {FILE_ATTRIBUTE_SPARSE_FILE, "SPARSE_FILE"},
+ {FILE_ATTRIBUTE_REPARSE_POINT, "REPARSE_POINT"},
+ {FILE_ATTRIBUTE_COMPRESSED, "COMPRESSED"},
+ {FILE_ATTRIBUTE_OFFLINE, "OFFLINE"},
{FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,"NOT_CONTENT_INDEXED"},
- {FILE_ATTRIBUTE_ENCRYPTED, "ENCRYPTED"},
- {FILE_ATTRIBUTE_VIRTUAL, "VIRTUAL"},
+ {FILE_ATTRIBUTE_ENCRYPTED, "ENCRYPTED"},
+ {FILE_ATTRIBUTE_VIRTUAL, "VIRTUAL"},
};
-/* Prints a directory entry. @lookup_table is a pointer to the lookup table, or
- * NULL if the resource entry for the dentry is not to be printed. */
+/* Prints a directory entry. @lookup_table is a pointer to the lookup table, if
+ * available. If the dentry is unresolved and the lookup table is NULL, the
+ * lookup table entries will not be printed. Otherwise, they will be. */
int print_dentry(struct dentry *dentry, void *lookup_table)
{
const u8 *hash;
printf("Last Write Time = 0x%"PRIx64"\n");
#endif
+ /* Translate the timestamps into something readable */
time_t creat_time = wim_timestamp_to_unix(dentry->creation_time);
time_t access_time = wim_timestamp_to_unix(dentry->last_access_time);
time_t mod_time = wim_timestamp_to_unix(dentry->last_write_time);
- printf("Creation Time = %s", asctime(localtime(&creat_time)));
- printf("Last Access Time = %s", asctime(localtime(&access_time)));
- printf("Last Write Time = %s", asctime(localtime(&mod_time)));
+ printf("Creation Time = %s", asctime(gmtime(&creat_time)));
+ printf("Last Access Time = %s", asctime(gmtime(&access_time)));
+ printf("Last Write Time = %s", asctime(gmtime(&mod_time)));
printf("Reparse Tag = 0x%"PRIx32"\n", dentry->reparse_tag);
printf("Hard Link Group = 0x%"PRIx64"\n", dentry->hard_link);
return 0;
}
-static inline void dentry_common_init(struct dentry *dentry)
+/* Initializations done on every `struct dentry'. */
+static void dentry_common_init(struct dentry *dentry)
{
memset(dentry, 0, sizeof(struct dentry));
dentry->refcnt = 1;
/*
* Creates an unlinked directory entry.
*
- * @name: The base name of the new dentry.
- * @return: A pointer to the new dentry, or NULL if out of memory.
+ * @name: The UTF-8 filename of the new dentry.
+ *
+ * Returns a pointer to the new dentry, or NULL if out of memory.
*/
struct dentry *new_dentry(const char *name)
{
FREE(dentry->extracted_file);
}
+/* Frees a WIM dentry. */
void free_dentry(struct dentry *dentry)
{
wimlib_assert(dentry);
__destroy_dentry(dentry);
+ /* Don't destroy the ADS entries if they "belong" to a different dentry
+ * */
if (dentry->ads_entries_status != ADS_ENTRIES_USER)
dentry_free_ads_entries(dentry);
FREE(dentry);
}
-/* clones a dentry.
+/* Partically clones a dentry.
*
* Beware:
- * - memory for file names is not cloned
+ * - memory for file names is not cloned (the pointers are all set to NULL
+ * and the lengths are set to zero)
* - next, prev, and children pointers and not touched
- * - stream entries are not cloned.
+ * - stream entries are not cloned (pointer left untouched).
*/
struct dentry *clone_dentry(struct dentry *old)
{
static int do_free_dentry(struct dentry *dentry, void *__lookup_table)
{
struct lookup_table *lookup_table = __lookup_table;
+ struct lookup_table_entry *lte;
+ unsigned i;
+
if (lookup_table) {
- struct lookup_table_entry *lte;
- if (dentry->resolved)
- lte = dentry->lte;
- else
- lte = __lookup_resource(lookup_table, dentry->hash);
- lte_decrement_refcnt(lte, lookup_table);
+ for (i = 0; i <= dentry->num_ads; i++) {
+ lte = dentry_stream_lte(dentry, i, lookup_table);
+ lte_decrement_refcnt(lte, lookup_table);
+ }
}
wimlib_assert(dentry->refcnt != 0);
return ret;
}
+/*
+ * Changes the name of an alternate data stream */
int change_ads_name(struct ads_entry *entry, const char *new_name)
{
return get_names(&entry->stream_name, &entry->stream_name_utf8,
{
struct image_statistics *stats;
struct lookup_table_entry *lte;
- u16 i;
stats = arg;
else
++*stats->file_count;
- if (dentry->resolved)
- lte = dentry->lte;
- else
- lte = __lookup_resource(stats->lookup_table, dentry->hash);
- i = 0;
- while (1) {
+ for (unsigned i = 0; i <= dentry->num_ads; i++) {
+ lte = dentry_stream_lte(dentry, i, stats->lookup_table);
if (lte) {
- u64 size = lte->resource_entry.original_size;
- *stats->total_bytes += size;
+ *stats->total_bytes += wim_resource_size(lte);
if (++lte->out_refcnt == 1)
- *stats->hard_link_bytes += size;
+ *stats->hard_link_bytes += wim_resource_size(lte);
}
- if (i == dentry->num_ads)
- break;
- lte = __lookup_resource(stats->lookup_table,
- dentry->ads_entries[i].hash);
- i++;
}
-
return 0;
}
+/* Calculates some statistics about a dentry tree. */
void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *table,
u64 *dir_count_ret, u64 *file_count_ret,
u64 *total_bytes_ret,
for_dentry_in_tree(root, calculate_dentry_statistics, &stats);
}
+
+/*
+ * Reads the alternate data stream entries for a dentry.
+ *
+ * @p: Pointer to buffer that starts with the first alternate stream entry.
+ *
+ * @dentry: Dentry to load the alternate data streams into.
+ * @dentry->num_ads must have been set to the number of
+ * data streams that are expected.
+ *
+ * @remaining_size: Number of bytes of data remaining in the buffer pointed
+ * to by @p.
+ *
+ * The format of the on-disk alternate stream entries is as follows:
+ *
+ * struct ads_entry_on_disk {
+ * u64 length; // Length of the entry, in bytes
+ * u64 reserved; // Seems to be unused
+ * u8 hash[20]; // SHA1 message digest of the uncompressed stream
+ * u16 stream_name_len; // Length of the stream name, in bytes
+ * char stream_name[]; // Stream name in UTF-16LE, @stream_name_len bytes long,
+ * // not including null terminator
+ * u16 zero; // UTF-16 null terminator for the stream name, NOT
+ * // included in @stream_name_len
+ * };
+ *
+ * In addition, the entries are 8-byte aligned.
+ *
+ * Return 0 on success or nonzero on failure. On success, dentry->ads_entries
+ * is set to an array of `struct ads_entry's of length dentry->num_ads. On
+ * failure, @dentry is not modified.
+ */
static int read_ads_entries(const u8 *p, struct dentry *dentry,
u64 remaining_size)
{
- u16 num_ads = dentry->num_ads;
- struct ads_entry *ads_entries = CALLOC(num_ads, sizeof(struct ads_entry));
+ u16 num_ads;
+ struct ads_entry *ads_entries;
int ret;
+
+ num_ads = dentry->num_ads;
+ ads_entries = CALLOC(num_ads, sizeof(struct ads_entry));
if (!ads_entries) {
ERROR("Could not allocate memory for %"PRIu16" "
"alternate data stream entries", num_ads);
return WIMLIB_ERR_NOMEM;
}
- DEBUG2("Reading %"PRIu16" alternate data streams "
- "(remaining size = %"PRIu64")", num_ads, remaining_size);
for (u16 i = 0; i < num_ads; i++) {
struct ads_entry *cur_entry = &ads_entries[i];
u64 length;
size_t utf8_len;
const char *p_save = p;
+
/* Read the base stream entry, excluding the stream name. */
if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
ERROR("Stream entries go past end of metadata resource");
remaining_size -= WIM_ADS_ENTRY_DISK_SIZE;
p = get_u64(p, &length); /* ADS entry length */
-
- DEBUG2("ADS length = %"PRIu64, length);
-
p += 8; /* Unused */
p = get_bytes(p, SHA1_HASH_SIZE, (u8*)cur_entry->hash);
p = get_u16(p, &cur_entry->stream_name_len);
- DEBUG2("Stream name length = %u", cur_entry->stream_name_len);
-
cur_entry->stream_name = NULL;
cur_entry->stream_name_utf8 = NULL;
ret = WIMLIB_ERR_NOMEM;
goto out_free_ads_entries;
}
- p = p_save + ads_entry_length(cur_entry);
+ p = p_save + ads_entry_total_length(cur_entry);
}
dentry->ads_entries = ads_entries;
return 0;
}
/*
- * Reads a directory entry from the metadata resource.
+ * Reads a directory entry, including all alternate data stream entries that
+ * follow it, from the WIM image's metadata resource.
+ *
+ * @metadata_resource: Buffer containing the uncompressed metadata resource.
+ * @metadata_resource_len: Length of the metadata resource.
+ * @offset: Offset of this directory entry in the metadata resource.
+ * @dentry: A `struct dentry' that will be filled in by this function.
+ *
+ * Return 0 on success or nonzero on failure. On failure, @dentry have been
+ * modified, bu it will be left with no pointers to any allocated buffers.
+ * On success, the dentry->length field must be examined. If zero, this was a
+ * special "end of directory" dentry and not a real dentry. If nonzero, this
+ * was a real dentry.
*/
int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
u64 offset, struct dentry *dentry)
return WIMLIB_ERR_INVALID_DENTRY;
}
- /* Before reading the whole entry, we need to read just the length.
- * This is because an entry of length 8 (that is, just the length field)
+ /* Before reading the whole dentry, we need to read just the length.
+ * This is because a dentry of length 8 (that is, just the length field)
* terminates the list of sibling directory entries. */
p = get_u64(&metadata_resource[offset], &dentry->length);
/* A zero length field (really a length of 8, since that's how big the
* directory entry is...) indicates that this is the end of directory
* dentry. We do not read it into memory as an actual dentry, so just
- * return true in that case. */
+ * return successfully in that case. */
if (dentry->length == 0)
return 0;
+ /* If the dentry does not overflow the metadata resource buffer and is
+ * not too short, read the rest of it (excluding the alternate data
+ * streams, but including the file name and short name variable-length
+ * fields) into memory. */
if (offset + dentry->length >= metadata_resource_len) {
ERROR("Directory entry at offset %"PRIu64" and with size "
"%"PRIu64" ends past the end of the metadata resource "
return WIMLIB_ERR_INVALID_DENTRY;
}
- /* If it is a recognized length, read the rest of the directory entry.
- * Note: The root directory entry has no name, and its length does not
- * include the short name length field. */
if (dentry->length < WIM_DENTRY_DISK_SIZE) {
ERROR("Directory entry has invalid length of %"PRIu64" bytes",
dentry->length);
/* By the way, the reparse_reserved field does not actually exist (at
* least when the file is not a reparse point) */
-
p = get_u16(p, &dentry->num_ads);
p = get_u16(p, &short_name_len);
p = get_u16(p, &file_name_len);
+ /* We now know the length of the file name and short name. These should
+ * be included in the dentry length, but make sure the numbers are
+ * consistent. */
calculated_size = WIM_DENTRY_DISK_SIZE + file_name_len + 2 +
short_name_len;
p = get_bytes(p, short_name_len, short_name);
- /* Some directory entries inexplicibly have a little over 70 bytes of
+ /* Some directory entries inexplicably have a little over 70 bytes of
* extra data. The exact amount of data seems to be 72 bytes, but it is
- * aligned on the next 8-byte boundary. Here's an example of the
- * aligned data:
+ * aligned on the next 8-byte boundary. It does NOT seem to be
+ * alternate data stream entries. Here's an example of the aligned
+ * data:
*
* 01000000 40000000 6c786bba c58ede11 b0bb0026 1870892a b6adb76f
* e63a3e46 8fca8653 0d2effa1 6c786bba c58ede11 b0bb0026 1870892a
}
#endif
+ /*
+ * Read the alternate data streams, if present. dentry->num_ads tells
+ * us how many they are, and they will directly follow the dentry
+ * on-disk.
+ *
+ * Note that each alternate data stream entry begins on an 8-byte
+ * aligned boundary, and the alternate data stream entries are NOT
+ * included in the dentry->length field for some reason.
+ */
if (dentry->num_ads != 0) {
calculated_size = (calculated_size + 7) & ~7;
if (calculated_size > metadata_resource_len - offset) {
goto out_free_short_name;
}
+ /* We've read all the data for this dentry. Set the names and their
+ * lengths, and we've done. */
dentry->file_name = file_name;
dentry->file_name_utf8 = file_name_utf8;
dentry->short_name = short_name;
}
/*
- * Writes a dentry to an output buffer.
+ * Writes a WIM dentry to an output buffer.
*
* @dentry: The dentry structure.
* @p: The memory location to write the data to.
if (p - orig_p < dentry->length)
p = put_zeroes(p, dentry->length - (p - orig_p));
+ /* Align to 8-byte boundary */
p = put_zeroes(p, (8 - dentry->length % 8) % 8);
+ /* Write the alternate data streams, if there are any. Please see
+ * read_ads_entries() for comments about the format of the on-disk
+ * alternate data stream entries. */
for (u16 i = 0; i < dentry->num_ads; i++) {
- p = put_u64(p, ads_entry_length(&dentry->ads_entries[i]));
+ p = put_u64(p, ads_entry_total_length(&dentry->ads_entries[i]));
p = put_u64(p, 0); /* Unused */
if (dentry->resolved && dentry->ads_entries[i].lte)
hash = dentry->ads_entries[i].lte->hash;
return p;
}
-/* Recursive function that writes a dentry tree rooted at @tree, not including
- * @tree itself, which has already been written, except in the case of the root
- * dentry, which is written right away, along with an end-of-directory entry. */
-u8 *write_dentry_tree(const struct dentry *tree, u8 *p)
+/* Recursive function that writes a dentry tree rooted at @parent, not including
+ * @parent itself, which has already been written. */
+static u8 *write_dentry_tree_recursive(const struct dentry *parent, u8 *p)
{
const struct dentry *child;
- if (dentry_is_root(tree)) {
- p = write_dentry(tree, p);
-
- /* write end of directory entry */
- p = put_u64(p, 0);
- } else {
- /* Nothing to do for non-directories */
- if (!dentry_is_directory(tree))
- return p;
- }
+ /* Nothing to do if this dentry has no children. */
+ if (parent->subdir_offset == 0)
+ return p;
- /* Write child dentries and end-of-directory entry. */
- child = tree->children;
+ /* Write child dentries and end-of-directory entry.
+ *
+ * Note: we need to write all of this dentry's children before
+ * recursively writing the directory trees rooted at each of the child
+ * dentries, since the on-disk dentries for a dentry's children are
+ * always located at consecutive positions in the metadata resource! */
+ child = parent->children;
if (child) {
do {
p = write_dentry(child, p);
child = child->next;
- } while (child != tree->children);
+ } while (child != parent->children);
}
/* write end of directory entry */
/* Recurse on children. */
if (child) {
do {
- p = write_dentry_tree(child, p);
+ p = write_dentry_tree_recursive(child, p);
child = child->next;
- } while (child != tree->children);
+ } while (child != parent->children);
}
return p;
}
+/* Writes a directory tree to the metadata resource.
+ *
+ * @root: Root of the dentry tree.
+ * @p: Pointer to a buffer with enough space for the dentry tree.
+ *
+ * Returns pointer to the byte after the last byte we wrote.
+ */
+u8 *write_dentry_tree(const struct dentry *root, u8 *p)
+{
+ wimlib_assert(dentry_is_root(root));
+
+ /* If we're the root dentry, we have no parent that already
+ * wrote us, so we need to write ourselves. */
+ p = write_dentry(root, p);
+
+ /* Write end of directory entry after the root dentry just to be safe;
+ * however the root dentry obviously cannot have any siblings. */
+ p = put_u64(p, 0);
+
+ /* Recursively write the rest of the dentry tree. */
+ return write_dentry_tree_recursive(root, p);
+}
+
/* Reads the children of a dentry, and all their children, ..., etc. from the
* metadata resource and into the dentry tree.
*
* @metadata_resource: An array that contains the uncompressed metadata
* resource for the WIM file.
- * @metadata_resource_len: The length of @metadata_resource.
- * @dentry: A pointer to a struct dentry that is the root of the directory
+ *
+ * @metadata_resource_len: The length of the uncompressed metadata resource, in
+ * bytes.
+ *
+ * @dentry: A pointer to a `struct dentry' that is the root of the directory
* tree and has already been read from the metadata resource. It
* does not need to be the real root because this procedure is
* called recursively.
struct dentry cur_child;
int ret;
- /* If @dentry is a regular file, nothing more needs to be done for this
- * branch. */
+ /*
+ * If @dentry has no child dentries, nothing more needs to be done for
+ * this branch. This is the case for regular files, symbolic links, and
+ * *possibly* empty directories (although an empty directory may also
+ * have one child dentry that is the special end-of-directory dentry)
+ */
if (cur_offset == 0)
return 0;
break;
/* Check for end of directory. */
- if (cur_child.length == 0) {
- ret = 0;
+ if (cur_child.length == 0)
break;
- }
/* Not end of directory. Allocate this child permanently and
* link it to the parent and previous child. */
cur_offset += dentry_total_length(child);
}
- /* Link last child to first one, and set parent's
- * children pointer to the first child. */
+ /* Link last child to first one, and set parent's children pointer to
+ * the first child. */
if (prev_child) {
prev_child->next = first_child;
first_child->prev = prev_child;
/* Size of the struct dentry up to and including the file_name_len. */
#define WIM_DENTRY_DISK_SIZE 102
+/* Size of on-disk WIM alternate data stream entry, in bytes, up to and
+ * including the stream length field (see below). */
#define WIM_ADS_ENTRY_DISK_SIZE 38
+
/*
* Reparse tags documented at
* http://msdn.microsoft.com/en-us/library/dd541667(v=prot.10).aspx
struct lookup_table_entry;
-/* Alternate data stream entry */
+/* Alternate data stream entry.
+ *
+ * We read this from disk in the read_ads_entries() function; see that function
+ * for more explanation. */
struct ads_entry {
union {
/* SHA-1 message digest of stream contents */
struct stream_list_head lte_group_list;
};
-static inline u64 ads_entry_length(const struct ads_entry *entry)
+/* Returns the total length of a WIM alternate data stream entry on-disk,
+ * including the stream name, the null terminator, AND the padding after the
+ * entry to align the next one (or the next dentry) on an 8-byte boundary. */
+static inline u64 ads_entry_total_length(const struct ads_entry *entry)
{
u64 len = WIM_ADS_ENTRY_DISK_SIZE + entry->stream_name_len + 2;
return (len + 7) & ~7;
}
-/* In-memory structure for a WIM directory entry. There is a directory tree for
- * each image in the WIM. */
+/*
+ * In-memory structure for a WIM directory entry (dentry). There is a directory
+ * tree for each image in the WIM.
+ *
+ * Please note that this is a directory entry and not an inode. Since NTFS
+ * allows hard links, it's possible for a NTFS inode to correspond to multiple
+ * WIM dentries. The @hard_link field tells you the number of the NTFS inode
+ * that the dentry corresponds to.
+ *
+ * Unfortunately, WIM files do not have an analogue to an inode; instead certain
+ * information, such as file attributes, the security descriptor, and file
+ * streams is replicated in each hard-linked dentry, even though this
+ * information really is associated with an inode.
+ *
+ * Confusingly, it's also possible for stream information to be missing from a
+ * dentry in a hard link set, in which case the stream information needs to be
+ * gotten from one of the other dentries in the hard link set. In addition, it
+ * is possible for dentries to have inconsistent security IDs, file attributes,
+ * or file streams when they share the same hard link ID (don't even ask. I
+ * hope that Microsoft may have fixed this problem, since I've only noticed it
+ * in the 'install.wim' for Windows 7). For those dentries, we have to use the
+ * conflicting fields to split up the hard link groups.
+ */
struct dentry {
/* The parent of this directory entry. */
struct dentry *parent;
/* Linked list of sibling directory entries. */
struct dentry *next;
-
struct dentry *prev;
/* Pointer to a child of this directory entry. */
struct dentry *children;
- /* Size of directory entry on disk, in bytes. Typical size is around
+ /*
+ * Size of directory entry on disk, in bytes. Typical size is around
* 104 to 120 bytes.
*
* It is possible for the length field to be 0. This situation, which
*/
u64 length;
- /* The file attributes associated with this file. */
+ /* The file attributes associated with this file. This is a bitwise OR
+ * of the FILE_ATTRIBUTE_* flags. */
u32 attributes;
/* The index of the security descriptor in the WIM image's table of
* If -1, no security information exists for this file. */
int32_t security_id;
- /* The offset, from the start of the WIM metadata resource for this
- * image, of this directory entry's child files. 0 if the directory
- * entry has no children (as in the case of regular files or reparse
- * points). */
+ /* The offset, from the start of the uncompressed WIM metadata resource
+ * for this image, of this dentry's child dentries. 0 if the directory
+ * entry has no children, which is the case for regular files or reparse
+ * points. */
u64 subdir_offset;
/* Timestamps for the dentry. The timestamps are the number of
* 100-nanosecond intervals that have elapsed since 12:00 A.M., January
- * 1st, 1601, UTC. */
+ * 1st, 1601, UTC. This is the same format used in NTFS inodes. */
u64 creation_time;
u64 last_access_time;
u64 last_write_time;
- /* true if the dentry's lookup table entry has been resolved (i.e. the
- * @lte field is valid, but the @hash field is not valid) */
+ /* %true iff the dentry's lookup table entry has been resolved (i.e. the
+ * @lte field is valid, but the @hash field is not valid)
+ *
+ * (This is not an on-disk field.) */
bool resolved;
/* A hash of the file's contents, or a pointer to the lookup table entry
* read_dentry() function. */
//u32 reparse_reserved;
+ /* If the file is part of a hard link set, all the directory entries in
+ * the set will share the same value for this field.
+ *
+ * Unfortunately, in some WIMs it is NOT the case that all dentries that
+ * share this field are actually in the same hard link set, although the
+ * WIMs that wimlib writes maintain this restriction. */
+ u64 hard_link;
+
/* Number of alternate data streams associated with this file. */
u16 num_ads;
* the terminating zero byte. */
u16 file_name_utf8_len;
- /* Pointer to the short filename */
+ /* Pointer to the short filename (malloc()ed buffer) */
char *short_name;
- /* Pointer to the filename. */
+ /* Pointer to the filename (malloc()ed buffer). */
char *file_name;
- /* Pointer to the filename converted to UTF-8. */
+ /* Pointer to the filename converted to UTF-8 (malloc()ed buffer). */
char *file_name_utf8;
- /* Full path to this dentry. */
+ /* Full path to this dentry (malloc()ed buffer). */
char *full_path_utf8;
u32 full_path_utf8_len;
- /* Alternate stream entries for this dentry. */
+ /* Alternate stream entries for this dentry (malloc()ed buffer). */
struct ads_entry *ads_entries;
union {
u32 num_times_opened;
};
- /* If the file is part of a hard link set, all the directory entries in
- * the set will share the same value for this field.
- *
- * Unfortunately, in some WIMs it is NOT the case that all dentries that
- * share this field are actually in the same hard link set, although the
- * WIMs that wimlib writes maintain this restriction. */
- u64 hard_link;
-
enum {
/* This dentry is the owner of its ads_entries, although it may
* be in a hard link set */
/* List of dentries sharing the same lookup table entry */
struct stream_list_head lte_group_list;
- /* Path to extracted file on disk (used during extraction only) */
+ /* Path to extracted file on disk (used during extraction only)
+ * (malloc()ed buffer) */
char *extracted_file;
};
-/* Return the number of dentries in the hard link group */
-static inline size_t dentry_link_group_size(const struct dentry *dentry)
-{
- const struct list_head *cur = &dentry->link_group_list;
- size_t size = 0;
- wimlib_assert(cur != NULL);
- do {
- size++;
- cur = cur->next;
- } while (cur != &dentry->link_group_list);
- return size;
-}
-
extern struct ads_entry *dentry_get_ads_entry(struct dentry *dentry,
const char *stream_name);
extern u8 *write_dentry_tree(const struct dentry *tree, u8 *p);
-/* Inline utility functions for dentries */
+/* Return the number of dentries in the hard link group */
+static inline size_t dentry_link_group_size(const struct dentry *dentry)
+{
+ const struct list_head *cur = &dentry->link_group_list;
+ size_t size = 0;
+ wimlib_assert(cur != NULL);
+ do {
+ size++;
+ cur = cur->next;
+ } while (cur != &dentry->link_group_list);
+ return size;
+}
static inline bool dentry_is_root(const struct dentry *dentry)
{
struct wim_security_data *sd;
struct link_group_table *lgt;
const struct lookup_table_entry *metadata_lte;
- const struct resource_entry *res_entry;
+ u64 metadata_len;
+ u64 metadata_offset;
metadata_lte = imd->metadata_lte;
- res_entry = &metadata_lte->resource_entry;
+ metadata_len = wim_resource_size(metadata_lte);
+ metadata_offset = metadata_lte->resource_entry.offset;
DEBUG("Reading metadata resource: length = %"PRIu64", "
- "offset = %"PRIu64"",
- res_entry->original_size, res_entry->offset);
-
- if (res_entry->original_size < 8) {
- ERROR("Expected at least 8 bytes for the metadata resource");
+ "offset = %"PRIu64"", metadata_len, metadata_offset);
+
+ /* There is no way the metadata resource could possibly be less than (8
+ * + WIM_DENTRY_DISK_SIZE) bytes, where the 8 is for security data (with
+ * no security descriptors) and WIM_DENTRY_DISK_SIZE is for the root
+ * dentry. */
+ if (metadata_len < 8 + WIM_DENTRY_DISK_SIZE) {
+ ERROR("Expected at least %zu bytes for the metadata resource",
+ 8 + WIM_DENTRY_DISK_SIZE);
return WIMLIB_ERR_INVALID_RESOURCE_SIZE;
}
/* Allocate memory for the uncompressed metadata resource. */
- buf = MALLOC(res_entry->original_size);
+ buf = MALLOC(metadata_len);
if (!buf) {
ERROR("Failed to allocate %"PRIu64" bytes for uncompressed "
- "metadata resource", res_entry->original_size);
+ "metadata resource", metadata_len);
return WIMLIB_ERR_NOMEM;
}
- /* Determine the compression type of the metadata resource. */
-
/* Read the metadata resource into memory. (It may be compressed.) */
ret = read_full_wim_resource(metadata_lte, buf);
if (ret != 0)
DEBUG("Finished reading metadata resource into memory.");
- /* The root directory entry starts after security data, on an 8-byte
- * aligned address.
+ /* The root directory entry starts after security data, aligned on an
+ * 8-byte boundary within the metadata resource.
*
* The security data starts with a 4-byte integer giving its total
- * length. */
+ * length, so if we round that up to an 8-byte boundary that gives us
+ * the offset of the root dentry.
+ *
+ * Here we read the security data into a wim_security_data structure,
+ * and if successful, go ahead and calculate the offset in the metadata
+ * resource of the root dentry. */
- /* Read the security data into a wim_security_data structure. */
- ret = read_security_data(buf, res_entry->original_size, &sd);
+ ret = read_security_data(buf, metadata_len, &sd);
if (ret != 0)
goto out_free_buf;
+ get_u32(buf, &dentry_offset);
+ if (dentry_offset == 0)
+ dentry_offset = 8;
+ dentry_offset = (dentry_offset + 7) & ~7;
+
+ /* Allocate memory for the root dentry and read it into memory */
dentry = MALLOC(sizeof(struct dentry));
if (!dentry) {
ERROR("Failed to allocate %zu bytes for root dentry",
ret = WIMLIB_ERR_NOMEM;
goto out_free_security_data;
}
-
- get_u32(buf, &dentry_offset);
- if (dentry_offset == 0)
- dentry_offset = 8;
- dentry_offset = (dentry_offset + 7) & ~7;
- ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry);
+ ret = read_dentry(buf, metadata_len, dentry_offset, dentry);
+
/* This is the root dentry, so set its pointers correctly. */
dentry->parent = dentry;
dentry->next = dentry;
if (ret != 0)
goto out_free_dentry_tree;
+ /* Now read the entire directory entry tree into memory. */
DEBUG("Reading dentry tree");
- /* Now read the entire directory entry tree. */
- ret = read_dentry_tree(buf, res_entry->original_size, dentry);
+ ret = read_dentry_tree(buf, metadata_len, dentry);
if (ret != 0)
goto out_free_dentry_tree;
- DEBUG("Calculating dentry full paths");
/* Calculate the full paths in the dentry tree. */
+ DEBUG("Calculating dentry full paths");
ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL);
if (ret != 0)
goto out_free_dentry_tree;
- DEBUG("Building link group table");
/* Build hash table that maps hard link group IDs to dentry sets */
+ DEBUG("Building link group table");
lgt = new_link_group_table(9001);
if (!lgt)
goto out_free_dentry_tree;
struct dentry *root;
struct lookup_table_entry *lte, *duplicate_lte;
u64 metadata_original_size;
-
- /*
- * We append 20 random bytes to the metadata resource so that we don't
- * have identical metadata resources if we happen to append exactly the
- * same image twice without any changes in timestamps. If this were to
- * happen, it would cause confusion about the number and order of images
- * in the WIM.
- */
+ const struct wim_security_data *sd;
const unsigned random_tail_len = 20;
DEBUG("Writing metadata resource for image %d", w->current_image);
root = wim_root_dentry(w);
+ sd = wim_security_data(w);
- const struct wim_security_data *sd = wim_security_data(w);
+ /* We do not allow the security data pointer to be NULL, although it may
+ * point to an empty security data with no entries. */
wimlib_assert(sd);
- subdir_offset = sd->total_length + root->length + 8;
+
+ /* Offset of first child of the root dentry. It's equal to:
+ * - The total length of the security data, rounded to the next 8-byte
+ * boundary,
+ * - plus the total length of the root dentry,
+ * - plus 8 bytes for an end-of-directory entry following the root
+ * dentry (shouldn't really be needed, but just in case...)
+ */
+ subdir_offset = ((sd->total_length + 7) & ~7) + dentry_total_length(root) + 8;
+
+ /* Calculate the subdirectory offsets for the entire dentry tree. */
calculate_subdir_offsets(root, &subdir_offset);
+
+ /* Total length of the metadata resource (uncompressed) */
metadata_original_size = subdir_offset + random_tail_len;
+
+ /* Allocate a buffer to contain the uncompressed metadata resource */
buf = MALLOC(metadata_original_size);
if (!buf) {
ERROR("Failed to allocate %"PRIu64" bytes for "
return WIMLIB_ERR_NOMEM;
}
+ /* Write the security data into the resource buffer */
p = write_security_data(sd, buf);
+ /* Write the dentry tree into the resource buffer */
DEBUG("Writing dentry tree.");
p = write_dentry_tree(root, p);
+
+ /*
+ * Append 20 random bytes to the metadata resource so that we don't have
+ * identical metadata resources if we happen to append exactly the same
+ * image twice without any changes in timestamps. If this were to
+ * happen, it would cause confusion about the number and order of images
+ * in the WIM.
+ */
randomize_byte_array(p, random_tail_len);
+
+ /* We MUST have exactly filled the buffer; otherwise we calculated its
+ * size incorrectly or wrote the data incorrectly. */
wimlib_assert(p - buf + random_tail_len == metadata_original_size);
+ /* Get the lookup table entry for the metadata resource so we can update
+ * it. */
lte = wim_metadata_lookup_table_entry(w);
+ /* Write the metadata resource to the output WIM using the proper
+ * compression type. The lookup table entry for the metadata resource
+ * is updated. */
ret = write_wim_resource_from_buffer(buf, metadata_original_size,
w->out_fp,
wimlib_get_compression_type(w),
<e->output_resource_entry,
lte->hash);
+ if (ret != 0)
+ goto out;
+
+ /* It's very likely the SHA1 message digest of the metadata resource, so
+ * re-insert the lookup table entry into the lookup table. */
lookup_table_unlink(w->lookup_table, lte);
lookup_table_insert(w->lookup_table, lte);
+
+ /* We do not allow a metadata resource to be referenced multiple times,
+ * and the 20 random bytes appended to it should make it extremely
+ * likely for each metadata resource to be unique, even if the exact
+ * same image is captured. */
wimlib_assert(lte->out_refcnt == 0);
lte->out_refcnt = 1;
+
+ /* Make sure that the resource entry is written marked with the metadata
+ * flag. */
lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;
out:
+ /* All the data has been written to the new WIM; no need for the buffer
+ * anymore */
FREE(buf);
return ret;
}