X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=include%2Fwimlib%2Fdentry.h;h=0871edc2d418894edc160743346ca3ae523091ce;hp=1e08ae3f86f26b546d1f5e2725ace362f44bc517;hb=391aa5222080e7544aa82905980f39870ffc6cba;hpb=e78ce6d7c0a0d93194c8fcf9398fc086b912f842 diff --git a/include/wimlib/dentry.h b/include/wimlib/dentry.h index 1e08ae3f..0871edc2 100644 --- a/include/wimlib/dentry.h +++ b/include/wimlib/dentry.h @@ -18,6 +18,7 @@ struct wim_lookup_table; struct wim_lookup_table_entry; struct wimfs_fd; struct wim_inode; +struct wim_security_data; /* Size of the struct wim_dentry up to and including the file_name_len. */ #define WIM_DENTRY_DISK_SIZE 102 @@ -125,12 +126,25 @@ ads_entries_have_same_name(const struct wim_ads_entry *entry_1, * dentry_tree_fix_inodes() in hardlink.c). */ struct wim_dentry { - /* The inode for this dentry */ + /* Pointer to the inode for this dentry. This will contain some + * information that was factored out of the on-disk WIM dentry as common + * to all dentries in a hard link group. */ struct wim_inode *d_inode; - /* Red-black tree of sibling dentries */ + /* Node for the parent's red-black tree of child dentries, sorted by + * case sensitive long name. */ struct rb_node rb_node; +#ifdef __WIN32__ + /* Node for the parent's red-black tree of child dentries, sorted by + * case insensitive long name. */ + struct rb_node rb_node_case_insensitive; + + /* List of dentries in a directory that have different case sensitive + * long names but share the same case insensitive long name */ + struct list_head case_insensitive_conflict_list; +#endif + /* Length of UTF-16LE encoded short filename, in bytes, not including * the terminating zero wide-character. */ u16 short_name_nbytes; @@ -143,58 +157,80 @@ struct wim_dentry { * including the terminating null character. */ u32 full_path_nbytes; - /* Does this dentry need to be extracted? */ + /* For extraction operations, a subtree of dentries will have this flag + * set so we can keep track of which dentries still need to be + * extracted. Otherwise this will always be 0. */ u8 needs_extraction : 1; - /* Only used during NTFS capture */ + /* For extraction operations, this flag will be set when a dentry in the + * tree being extracted is not being extracted for some reason (file + * type not supported by target filesystem or contains invalid + * characters). Otherwise this will always be 0. */ + u8 not_extracted : 1; + + /* When capturing from a NTFS volume using NTFS-3g, this flag is set on + * dentries that were created from a filename in the WIN32 or WIN32+DOS + * namespaces rather than the POSIX namespace. Otherwise this will + * always be 0. */ u8 is_win32_name : 1; - /* Temporary list */ + /* When verifying the dentry tree after reading it into memory, this + * flag will be set on all dentries in a hard link group that have a + * nonempty DOS name except one. This is because it is supposed to be + * illegal (on NTFS, at least) for a single inode to have multiple DOS + * names. */ + u8 dos_name_invalid : 1; + + /* Temporary list field used to make lists of dentries in a few places. + * */ struct list_head tmp_list; - /* List of dentries in the inode (hard link set) */ + /* Linked list node that places this dentry in the list of aliases for + * its inode (d_inode) */ struct list_head d_alias; /* The parent of this directory entry. */ struct wim_dentry *parent; - /* - * 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. - * - * 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. - */ + /* 'length' and 'subdir_offset' are only used while reading and writing + * this dentry; see the corresponding field in + * `struct wim_dentry_on_disk' for explanation. */ u64 length; - - /* 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; + /* These correspond to the two unused fields in the on-disk WIM dentry; + * we read them into memory so we can write them unchanged. These + * fields are set to 0 on new dentries. */ u64 d_unused_1; u64 d_unused_2; - /* Pointer to the UTF-16LE short filename (malloc()ed buffer) */ + /* Pointer to the UTF-16LE short filename (malloc()ed buffer), or NULL + * if this dentry has no short name. */ utf16lechar *short_name; - /* Pointer to the UTF-16LE filename (malloc()ed buffer). */ + /* Pointer to the UTF-16LE filename (malloc()ed buffer), or NULL if this + * dentry has no filename. */ utf16lechar *file_name; - /* Full path of this dentry */ + /* Full path to this dentry in the WIM, in platform-dependent tchars + * that can be printed without conversion. By default this field will + * be NULL and will only be calculated on-demand by the + * calculate_dentry_full_path() or dentry_full_path() functions. */ tchar *_full_path; + + /* (Extraction only) Actual name to extract this dentry as, along with + * its length in tchars excluding the NULL terminator. This usually + * will be the same as file_name, with the character encoding converted + * if needed. But if file_name contains characters not accepted on the + * current platform, then this may be set slightly differently from + * file_name. This will be either NULL or a malloc()ed buffer that may + * alias file_name. */ + tchar *extraction_name; + size_t extraction_name_nchars; + + /* (Extraction only) List head for building a list of dentries that + * contain a certain stream. */ + struct list_head extraction_stream_list; }; #define rbnode_dentry(node) container_of(node, struct wim_dentry, rb_node) @@ -202,103 +238,142 @@ struct wim_dentry { /* * WIM inode. * - * As mentioned above, in the WIM file that is no on-disk analogue of a real - * inode, as most of these fields are duplicated in the dentries. + * As mentioned in the comment above `struct wim_dentry', in the WIM file that + * is no on-disk analogue of a real inode, as most of these fields are + * duplicated in the dentries. Instead, a `struct wim_inode' is something we + * create ourselves to simplify the handling of hard links. */ struct wim_inode { - /* Timestamps for the inode. The timestamps are the number of - * 100-nanosecond intervals that have elapsed since 12:00 A.M., January - * 1st, 1601, UTC. This is the same format used in NTFS inodes. */ - u64 i_creation_time; - u64 i_last_access_time; - u64 i_last_write_time; + /* If i_resolved == 0: + * SHA1 message digest of the contents of the unnamed-data stream + * of this inode. + * + * If i_resolved == 1: + * Pointer to the lookup table entry for the unnamed data stream + * of this inode, or NULL. + * + * i_hash corresponds to the 'unnamed_stream_hash' field of the `struct + * wim_dentry_on_disk' and the additional caveats documented about that + * field apply here (for example, the quirks regarding all-zero hashes). + */ + union { + u8 i_hash[SHA1_HASH_SIZE]; + struct wim_lookup_table_entry *i_lte; + }; - /* The file attributes associated with this inode. This is a bitwise OR - * of the FILE_ATTRIBUTE_* flags. */ + /* Corresponds to the 'attributes' field of `struct wim_dentry_on_disk'; + * bitwise OR of the FILE_ATTRIBUTE_* flags that give the attributes of + * this inode. */ u32 i_attributes; - /* 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 i_security_id; + /* Root of a red-black tree storing the child dentries of this inode, if + * any. Keyed by wim_dentry->file_name, case sensitively. */ + struct rb_root i_children; - /* %true iff the inode's lookup table entries has been resolved (i.e. - * the @lte field is valid, but the @hash field is not valid) - * - * (This is not an on-disk field.) */ - u8 i_resolved : 1; +#ifdef __WIN32__ + /* Root of a red-black tree storing the children of this inode, if any. + * Keyed by wim_dentry->file_name, case insensitively. */ + struct rb_root i_children_case_insensitive; +#endif - /* %true iff verify_inode() has run on this inode. */ - u8 i_verified : 1; + /* List of dentries that are aliases for this inode. There will be + * i_nlink dentries in this list. */ + struct list_head i_dentry; - u8 i_visited : 1; + /* Field to place this inode into a list. */ + union { + /* Hash list node- used in hardlink.c when the inodes are placed + * into a hash table keyed by inode number and optionally device + * number, in order to detect dentries that are aliases for the + * same inode. */ + struct hlist_node i_hlist; - /* Used only in NTFS-mode extraction */ - u8 i_dos_name_extracted : 1; + /* Normal list node- used to connect all the inodes of a WIM image + * into a single linked list referenced from the + * `struct wim_image_metadata' for that image. */ + struct list_head i_list; + }; - /* Set to 0 if reparse point fixups have been done. Otherwise set to 1. - * - * Note: this actually may reflect the SYMBOLIC_LINK_RELATIVE flag. */ - u16 i_not_rpfixed; + /* Number of dentries that are aliases for this inode. */ + u32 i_nlink; /* Number of alternate data streams associated with this inode */ u16 i_num_ads; - /* Unused/unknown fields that we just read into memory so we can - * re-write them unchanged. */ - u32 i_rp_unknown_1; - u16 i_rp_unknown_2; + /* Flag that indicates whether this inode's streams have been + * "resolved". By default, the inode starts as "unresolved", meaning + * that the i_hash field, along with the hash field of any associated + * wim_ads_entry's, are valid and should be used as keys in the WIM + * lookup table to find the associated `struct wim_lookup_table_entry'. + * But if the inode has been resolved, then each of these fields is + * replaced with a pointer directly to the appropriate `struct + * wim_lookup_table_entry', or NULL if the stream is empty. */ + u8 i_resolved : 1; - /* If i_resolved == 0: - * SHA1 message digest of the contents of the unnamed-data stream - * of this inode, or all zeroes if this inode has no unnamed data - * stream, or optionally all zeroes if this inode has an empty - * unnamed data stream. - * - * If i_resolved == 1: - * Pointer to the lookup table entry for the unnamed data stream - * of this inode, or NULL if this inode has no unnamed data stream, - * or optionally all zeroes if this inode has an empty unnamed data - * stream. - */ - union { - u8 i_hash[SHA1_HASH_SIZE]; - struct wim_lookup_table_entry *i_lte; - }; + /* Flag used to mark this inode as visited; this is used when visiting + * all the inodes in a dentry tree exactly once. It will be 0 by + * default and must be cleared following the tree traversal, even in + * error paths. */ + u8 i_visited : 1; + + /* For NTFS-3g extraction: Set after the DOS name for this inode has + * been extracted. */ + u8 i_dos_name_extracted : 1; + + /* Pointer to a malloc()ed array of i_num_ads alternate data stream + * entries for this inode. */ + struct wim_ads_entry *i_ads_entries; + + /* Creation time, last access time, and last write time for this inode, in + * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601. They + * should correspond to the times gotten by calling GetFileTime() on + * Windows. */ + u64 i_creation_time; + u64 i_last_access_time; + u64 i_last_write_time; + + /* Corresponds to 'security_id' in `struct wim_dentry_on_disk': The + * index of this inode's security descriptor in the WIM image's table of + * security descriptors, or -1. Note: in verify_inode(), called + * whenever a WIM image is loaded, out-of-bounds indices are set to -1, + * so the extraction code does not need to do bounds checks. */ + int32_t i_security_id; /* Identity of a reparse point. See * http://msdn.microsoft.com/en-us/library/windows/desktop/aa365503(v=vs.85).aspx * for what a reparse point is. */ u32 i_reparse_tag; - /* Number of dentries that reference this inode */ - u32 i_nlink; + /* Unused/unknown fields that we just read into memory so we can + * re-write them unchanged. */ + u32 i_rp_unknown_1; + u16 i_rp_unknown_2; - /* Alternate data stream entries. */ - struct wim_ads_entry *i_ads_entries; + /* Corresponds to not_rpfixed in `struct wim_dentry_on_disk': Set to 0 + * if reparse point fixups have been done. Otherwise set to 1. Note: + * this actually may reflect the SYMBOLIC_LINK_RELATIVE flag. + */ + u16 i_not_rpfixed; - /* Inode number */ + /* Inode number; corresponds to hard_link_group_id in the `struct + * wim_dentry_on_disk'. */ u64 i_ino; - /* Device number, used only during image capture */ - u64 i_devno; - - /* List of dentries that reference this inode (there should be - * link_count of them) */ - struct list_head i_dentry; - union { - struct hlist_node i_hlist; - struct list_head i_list; + /* Device number, used only during image capture, so we can + * identify hard linked files by the combination of inode number + * and device number (rather than just inode number, which could + * be ambigious if the captured tree spans a mountpoint). Set + * to 0 otherwise. */ + u64 i_devno; + + /* Used only during image extraction: pointer to the first path + * (malloc()ed buffer) at which this inode has been extracted. + * Freed and set to NULL after the extraction is done (either + * success or failure). */ + tchar *i_extracted_file; }; - tchar *i_extracted_file; - - /* Root of a red-black tree storing the children of this inode (if - * non-empty, implies the inode is a directory, although that is also - * noted in the @attributes field.) */ - struct rb_root i_children; - /* Next alternate data stream ID to be assigned */ u32 i_next_stream_id; @@ -323,6 +398,9 @@ struct wim_inode { #define inode_first_dentry(inode) \ container_of(inode->i_dentry.next, struct wim_dentry, d_alias) +#define inode_first_full_path(inode) \ + dentry_full_path(inode_first_dentry(inode)) + static inline bool dentry_is_first_in_inode(const struct wim_dentry *dentry) { @@ -387,6 +465,9 @@ print_dentry(struct wim_dentry *dentry, void *lookup_table); extern int print_dentry_full_path(struct wim_dentry *entry, void *ignore); +extern int +calculate_dentry_full_path(struct wim_dentry *dentry); + extern int calculate_dentry_tree_full_paths(struct wim_dentry *root); @@ -570,7 +651,7 @@ inode_ref_streams(struct wim_inode *inode); extern int dentry_tree_fix_inodes(struct wim_dentry *root, struct list_head *inode_list); -extern int -verify_dentry(struct wim_dentry *dentry, void *wim); +int +verify_inode(struct wim_inode *inode, const struct wim_security_data *sd); #endif