X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fdentry.h;h=5b2aeab957fa6709790934b31d6cf089748105f1;hp=16cf5f37bba23200891fac458fca6e6f21385b79;hb=c155e206bf806dd639b57922c3eef0b428c985df;hpb=b3f993d2ece8fda5c989dc8c0311732fb20e3233 diff --git a/src/dentry.h b/src/dentry.h index 16cf5f37..5b2aeab9 100644 --- a/src/dentry.h +++ b/src/dentry.h @@ -4,6 +4,7 @@ #include "util.h" #include "config.h" #include "list.h" +#include "sha1.h" #include @@ -14,11 +15,10 @@ typedef struct WIMStruct WIMStruct; /* 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 -#ifndef WIM_HASH_SIZE -#define WIM_HASH_SIZE 20 -#endif /* * Reparse tags documented at @@ -52,10 +52,21 @@ typedef struct WIMStruct WIMStruct; #define FILE_ATTRIBUTE_ENCRYPTED 0x00004000 #define FILE_ATTRIBUTE_VIRTUAL 0x00010000 -/* Alternate data stream entry */ +struct lookup_table_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 { - /* SHA-1 message digest of stream contents */ - u8 hash[WIM_HASH_SIZE]; + union { + /* SHA-1 message digest of stream contents */ + u8 hash[SHA1_HASH_SIZE]; + + /* The corresponding lookup table entry (only for resolved + * streams) */ + struct lookup_table_entry *lte; + }; /* Length of stream name (UTF-16) */ u16 stream_name_len; @@ -68,11 +79,19 @@ struct ads_entry { /* Stream name (UTF-8) */ char *stream_name_utf8; + + /* Doubly linked list of streams that share the same lookup table entry */ + 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; + u64 len = WIM_ADS_ENTRY_DISK_SIZE; + if (entry->stream_name_len) + len += entry->stream_name_len + 2; return (len + 7) & ~7; } @@ -92,53 +111,98 @@ static inline bool ads_entry_has_name(const struct ads_entry *entry, } -/* In-memory structure for a 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, in bytes. Typical size is around 104 to 120 - * bytes. */ - /* It is possible for the length field to be 0. This situation, which + /* + * 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 * is undocumented, indicates the end of a list of sibling nodes in a * directory. It also means the real length is 8, because the dentry - * included only the length field, but that takes up 8 bytes. */ + * included only the length field, but that takes up 8 bytes. + * + * The length here includes the base directory entry on disk as well as + * the long and short filenames. It does NOT include any alternate + * stream entries that may follow the directory entry, even though the + * size of those needs to be considered. The length SHOULD be 8-byte + * aligned, although we don't require it to be. We do require the + * length to be large enough to hold the file name(s) of the dentry; + * additionally, a warning is issued if this field is larger than the + * aligned size. + */ 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 node in the security table that contains this file's - * security information. If -1, no security information exists for this - * file. */ + /* The index of the security descriptor in the WIM image's table of + * security descriptors that contains this file's security information. + * If -1, no security information exists for this file. */ int32_t security_id; - /* The offset, from the start of the metadata section, of this directory - * entry's child files. 0 if the directory entry has no children. */ + /* 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; - /* Reserved for future disuse. Currently ignoring these fields. */ - u64 unused1; - u64 unused2; - - /* Timestamps for the entry. The timestamps are the number of + /* 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; - /* A hash of the file's contents. */ - u8 hash[WIM_HASH_SIZE]; + /* %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 + * for this dentry if the lookup table entries have been resolved. + * + * More specifically, this is for the un-named default file stream, as + * opposed to the alternate (named) file streams, which may have their + * own lookup table entries. */ + union { + u8 hash[SHA1_HASH_SIZE]; + struct lookup_table_entry *lte; + }; /* Identity of a reparse point. See * http://msdn.microsoft.com/en-us/library/windows/desktop/aa365503(v=vs.85).aspx @@ -152,6 +216,14 @@ struct dentry { * 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; @@ -167,81 +239,58 @@ struct dentry { * 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 { /* Number of references to the dentry tree itself, as in multiple * WIMStructs */ - int refcnt; + u32 refcnt; /* Number of times this dentry has been opened (only for * directories!) */ 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. */ - u64 hard_link; - enum { /* This dentry is the owner of its ads_entries, although it may * be in a hard link set */ - GROUP_INDEPENDENT, + ADS_ENTRIES_DEFAULT = 0, /* This dentry is the owner of the ads_entries in the hard link * set */ - GROUP_MASTER, + ADS_ENTRIES_OWNER, /* This dentry shares its ads_entries with a dentry in the hard - * link set that has GROUP_MASTER set. */ - GROUP_SLAVE - } link_group_master_status; + * link set that has ADS_ENTRIES_OWNER set. */ + ADS_ENTRIES_USER + } ads_entries_status; /* List of dentries in the hard link set */ struct list_head link_group_list; - /* Path to extracted file on disk (used during extraction only) */ + /* 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) + * (malloc()ed buffer) */ char *extracted_file; }; -/* Return hash of the "unnamed" (default) data stream. */ -static inline const u8 *dentry_hash(const struct dentry *dentry) -{ - /* If there are alternate data streams, the dentry hash field is zeroed - * out, and we need to find the hash in the un-named data stream (should - * be the first one, but check them in order just in case, and fall back - * to the dentry hash field if we can't find an unnamed data stream). */ - for (u16 i = 0; i < dentry->num_ads; i++) - if (dentry->ads_entries[i].stream_name_len == 0) - return dentry->ads_entries[i].hash; - return dentry->hash; -} - -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; - 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); @@ -257,9 +306,6 @@ extern u64 dentry_total_length(const struct dentry *dentry); extern void stbuf_to_dentry(const struct stat *stbuf, struct dentry *dentry); -extern int dentry_to_stbuf(const struct dentry *dentry, struct stat *stbuf, - const struct lookup_table *table); - extern int for_dentry_in_tree(struct dentry *root, int (*visitor)(struct dentry*, void*), void *args); @@ -293,12 +339,9 @@ extern struct dentry *new_dentry(const char *name); extern void dentry_free_ads_entries(struct dentry *dentry); extern void free_dentry(struct dentry *dentry); extern void put_dentry(struct dentry *dentry); -extern int share_dentry_ads(struct dentry *master, - struct dentry *slave); extern struct dentry *clone_dentry(struct dentry *old); extern void free_dentry_tree(struct dentry *root, - struct lookup_table *lookup_table, - bool lt_decrement_refcnt); + struct lookup_table *lookup_table); extern int increment_dentry_refcnt(struct dentry *dentry, void *ignore); extern int decrement_dentry_refcnt(struct dentry *dentry, void *ignore); @@ -312,14 +355,26 @@ extern void calculate_dir_tree_statistics(struct dentry *root, extern int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len, u64 offset, struct dentry *dentry); +extern int verify_dentry(struct dentry *dentry, void *wim); + extern int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, struct dentry *dentry); extern u8 *write_dentry_tree(const struct dentry *tree, u8 *p); -/* Inline utility functions for WIMDentries */ - +/* 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) {