X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Flookup_table.h;h=83796aae2d8bca57cf4566de9467755444da9cfa;hp=afc2117e91557061ff6e0bd8bee8fd551dac07fd;hb=394c5bd3292c0f3168416c0a5f25989e557b3cfc;hpb=1ca6a5d40c3a135c423de9c9b8ae5c6237f828c4 diff --git a/src/lookup_table.h b/src/lookup_table.h index afc2117e..83796aae 100644 --- a/src/lookup_table.h +++ b/src/lookup_table.h @@ -11,144 +11,251 @@ #define LOOKUP_FLAG_ADS_OK 0x00000001 #define LOOKUP_FLAG_DIRECTORY_OK 0x00000002 -/* Not yet used */ -//#define LOOKUP_FLAG_FOLLOW_SYMLINKS 0x00000004 +#ifdef __WIN32__ +#include +#endif -/* A lookup table that is used to translate the hash codes of dentries into the - * offsets and sizes of uncompressed or compressed file resources. It is - * implemented as a hash table. */ -struct lookup_table { +/* The lookup table of a WIM file maps SHA1 message digests to streams of data. + * Here, the in-memory structure is implemented as a hash table. + * + * Given a SHA1 message digest, the mapped-to stream is specified by an offset + * in the WIM, an uncompressed and compressed size, and resource flags (see + * 'struct resource_entry'). But, we associate additional information, such as + * a reference count, with each stream, so the actual mapping is from SHA1 + * message digests to 'struct wim_lookup_table_entry's, each of which contains + * an embedded 'struct resource_entry'. + * + * Note: Everything will break horribly if there is a SHA1 collision. + */ +struct wim_lookup_table { struct hlist_head *array; u64 num_entries; u64 capacity; + struct list_head *unhashed_streams; }; -struct wimlib_fd; - #ifdef WITH_NTFS_3G struct ntfs_location { - char *path_utf8; - char *stream_name_utf16; - u16 stream_name_utf16_num_chars; - struct _ntfs_volume **ntfs_vol_p; + tchar *path; + utf16lechar *stream_name; + u16 stream_name_nchars; + struct _ntfs_volume *ntfs_vol; bool is_reparse_point; }; #endif -/* - * An entry in the lookup table in the WIM file. +/* An enumerated type that identifies where the stream corresponding to this + * lookup table entry is actually located. * - * It is used to find data streams for files in the WIM. + * If we open a WIM and read its lookup table, the location is set to + * RESOURCE_IN_WIM since all the streams will initially be located in the WIM. + * However, to deal with problems such as image capture and image mount, we + * allow the actual location of the stream to be somewhere else, such as an + * external file. + */ +enum resource_location { + /* The lookup table entry does not correspond to a stream (this state + * should exist only temporarily) */ + RESOURCE_NONEXISTENT = 0, + + /* The stream resource is located in a WIM file. The WIMStruct for the + * WIM file will be pointed to by the @wim member. */ + RESOURCE_IN_WIM, + + /* The stream resource is located in an external file. The name of the + * file will be provided by @file_on_disk member. In addition, if + * @file_on_disk_fp is not NULL, it will be an open FILE * to the file. + * */ + RESOURCE_IN_FILE_ON_DISK, + + /* The stream resource is directly attached in an in-memory buffer + * pointed to by @attached_buffer. */ + RESOURCE_IN_ATTACHED_BUFFER, + +#ifdef WITH_FUSE + /* The stream resource is located in an external file in the staging + * directory for a read-write mount. */ + RESOURCE_IN_STAGING_FILE, +#endif + +#ifdef WITH_NTFS_3G + /* The stream resource is located in an NTFS volume. It is identified + * by volume, filename, data stream name, and by whether it is a reparse + * point or not. @ntfs_loc points to a structure containing this + * information. */ + RESOURCE_IN_NTFS_VOLUME, +#endif + +#ifdef __WIN32__ + /* Resource must be accessed using Win32 API (may be a named data + * stream) */ + RESOURCE_WIN32, + + /* Windows only: the file is on disk in the file named @file_on_disk, + * but the file is encrypted and must be read using special functions. + * */ + RESOURCE_WIN32_ENCRYPTED, +#endif + +}; + +/* + * An entry in the lookup table in the WIM file. * - * The lookup_table_entry for a given dentry in the WIM is found using the SHA1 - * message digest field. + * It is used to find data streams for files in the WIM. + * + * Metadata resources and reparse point data buffers will also have lookup table + * entries associated with the data. + * + * The lookup_table_entry for a given dentry or alternate stream entry in the + * WIM is found using the SHA1 message digest field. */ -struct lookup_table_entry { +struct wim_lookup_table_entry { /* List of lookup table entries in this hash bucket */ struct hlist_node hash_list; - /* @resource_entry is read from the lookup table in the WIM - * file; it says where to find the file resource in the WIM - * file, and whether it is compressed or not. */ + /* Location and size of the stream in the WIM, whether it is compressed + * or not, and whether it's a metadata resource or not. This is an + * on-disk field. */ struct resource_entry resource_entry; - /* Currently ignored; set to 1 in new lookup table entries. */ + /* Specifies which part of the split WIM the resource is located in. + * This is on on-disk field. + * + * In stand-alone WIMs, this must be 1. + * + * In split WIMs, every split WIM part has its own lookup table, and in + * read_lookup_table() it's currently expected that the part number of + * each lookup table entry in a split WIM part's lookup table is the + * same as the part number of that split WIM part. So this makes this + * field redundant since we store a pointer to the corresponding + * WIMStruct in the lookup table entry anyway. + */ u16 part_number; - /* If %true, this lookup table entry corresponds to a symbolic link - * reparse buffer. @symlink_reparse_data_buf will give the target of - * the symbolic link. */ - enum { - RESOURCE_NONEXISTENT = 0, - RESOURCE_IN_WIM, - RESOURCE_IN_FILE_ON_DISK, - RESOURCE_IN_STAGING_FILE, - RESOURCE_IN_ATTACHED_BUFFER, - RESOURCE_IN_NTFS_VOLUME, - } resource_location; - - /* Number of times this lookup table entry is referenced by dentries. */ + /* One of the `enum resource_location' values documented above. */ + u16 resource_location : 5; + + /* 1 if this stream is a unique size (only set while writing streams). */ + u8 unique_size : 1; + + /* 1 if this stream had a SHA1-message digest calculated for it yet? */ + u8 unhashed : 1; + + /* (On-disk field) + * Number of times this lookup table entry is referenced by dentries. + * Unfortunately, this field is not always set correctly in Microsoft's + * WIMs, so we have no choice but to fix it if more references to the + * lookup table entry are found than stated here. */ u32 refcnt; union { - /* SHA1 hash of the file resource pointed to by this lookup - * table entry */ + /* (On-disk field) SHA1 message digest of the stream referenced + * by this lookup table entry */ u8 hash[SHA1_HASH_SIZE]; - /* First 4 or 8 bytes of the SHA1 hash, used for inserting the - * entry into the hash table. Since the SHA1 hashes can be - * considered random, we don't really need the full 20 byte hash - * just to insert the entry in a hash table. */ + /* First 4 or 8 bytes of the SHA1 message digest, used for + * inserting the entry into the hash table. Since the SHA1 + * message digest can be considered random, we don't really need + * the full 20 byte hash just to insert the entry in a hash + * table. */ size_t hash_short; + + /* Unhashed entries only (unhashed == 1): these variables make + * it possible to find the pointer to this 'struct + * wim_lookup_table_entry' contained in either 'struct + * wim_ads_entry' or 'struct wim_inode'. There can be at most 1 + * such pointer, as we can only join duplicate streams after + * they have been hashed. */ + struct { + struct wim_inode *back_inode; + u32 back_stream_id; + }; }; - /* If @file_on_disk != NULL, the file resource indicated by this lookup - * table entry is not in the WIM file, but rather a file on disk; this - * occurs for files that are added to the WIM. In that case, - * file_on_disk is the name of the file in the outside filesystem. - * It will not be compressed, and its size will be given by - * resource_entry.size and resource_entry.original_size. */ + /* When a WIM file is written, out_refcnt starts at 0 and is incremented + * whenever the stream pointed to by this lookup table entry needs to be + * written. The stream only need to be written when out_refcnt is + * nonzero, since otherwise it is not referenced by any dentries. */ + u32 out_refcnt; + + /* Pointers to somewhere where the stream is actually located. See the + * comments for the @resource_location field above. */ union { WIMStruct *wim; - char *file_on_disk; - char *staging_file_name; - u8 *attached_buffer; + tchar *file_on_disk; + void *attached_buffer; + #ifdef WITH_FUSE + tchar *staging_file_name; + #endif #ifdef WITH_NTFS_3G struct ntfs_location *ntfs_loc; #endif }; + + /* Actual reference count to this stream (only used while + * verifying an image). */ + u32 real_refcnt; + union { - struct lookup_table_entry *next_lte_in_swm; - FILE *file_on_disk_fp; - #ifdef WITH_NTFS_3G - struct _ntfs_attr *attr; + #ifdef WITH_FUSE + /* Number of times this stream has been opened (used only during + * mounting) */ + u16 num_opened_fds; #endif + + /* This field is used for the special hardlink or symlink image + * extraction mode. In these mode, all identical files are linked + * together, and @extracted_file will be set to the filename of the + * first extracted file containing this stream. */ + tchar *extracted_file; }; -#ifdef WITH_FUSE - /* File descriptors table for this data stream */ - u16 num_opened_fds; - u16 num_allocated_fds; - struct wimlib_fd **fds; -#endif - /* When a WIM file is written, out_refcnt starts at 0 and is incremented - * whenever the file resource pointed to by this lookup table entry - * needs to be written. Naturally, the file resource only need to be - * written when out_refcnt is 0. Incrementing it further is needed to - * find the correct reference count to write to the lookup table in the - * output file, which may be less than the regular refcnt if not all - * images in the WIM file are written. - * - * output_resource_entry is the struct resource_entry for the position of the - * file resource when written to the output file. */ - u32 out_refcnt; union { + /* When a WIM file is written, @output_resource_entry is filled + * in with the resource entry for the output WIM. This will not + * necessarily be the same as the @resource_entry since: - The + * stream may have a different offset in the new WIM - The + * stream may have a different compressed size in the new WIM if + * the compression type changed + */ struct resource_entry output_resource_entry; - char *extracted_file; + + struct list_head msg_list; + struct list_head inode_list; + + struct { + struct hlist_node hash_list_2; + + struct list_head write_streams_list; + }; }; - /* Circular linked list of streams that share the same lookup table - * entry - * - * This list of streams may include streams from different hard link - * sets that happen to be the same. */ - struct list_head lte_group_list; - - /* List of lookup table entries that correspond to streams that have - * been extracted to the staging directory when modifying a read-write - * mounted WIM. */ - struct list_head staging_list; + /* Temporary list fields */ + union { + struct list_head unhashed_list; + struct list_head swm_stream_list; + struct list_head extraction_list; + struct list_head export_stream_list; + }; }; -static inline u64 wim_resource_size(const struct lookup_table_entry *lte) +static inline u64 +wim_resource_size(const struct wim_lookup_table_entry *lte) { return lte->resource_entry.original_size; } static inline u64 -wim_resource_compressed_size(const struct lookup_table_entry *lte) +wim_resource_chunks(const struct wim_lookup_table_entry *lte) +{ + return (wim_resource_size(lte) + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; +} + +static inline u64 +wim_resource_compressed_size(const struct wim_lookup_table_entry *lte) { return lte->resource_entry.size; } @@ -158,202 +265,248 @@ wim_resource_compressed_size(const struct lookup_table_entry *lte) * entry */ static inline int -wim_resource_compression_type(const struct lookup_table_entry *lte) +wim_resource_compression_type(const struct wim_lookup_table_entry *lte) { if (!(lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) || lte->resource_location != RESOURCE_IN_WIM) - return WIM_COMPRESSION_TYPE_NONE; + return WIMLIB_COMPRESSION_TYPE_NONE; return wimlib_get_compression_type(lte->wim); } +static inline bool +lte_filename_valid(const struct wim_lookup_table_entry *lte) +{ + return lte->resource_location == RESOURCE_IN_FILE_ON_DISK + #ifdef __WIN32__ + || lte->resource_location == RESOURCE_WIN32 + || lte->resource_location == RESOURCE_WIN32_ENCRYPTED + #endif + #ifdef WITH_FUSE + || lte->resource_location == RESOURCE_IN_STAGING_FILE + #endif + ; +} + +extern struct wim_lookup_table * +new_lookup_table(size_t capacity); + +extern int +read_lookup_table(WIMStruct *w); -extern struct lookup_table *new_lookup_table(size_t capacity); +extern int +write_lookup_table(WIMStruct *w, int image, struct resource_entry *out_res_entry); -extern void lookup_table_insert(struct lookup_table *table, - struct lookup_table_entry *lte); +extern void +free_lookup_table(struct wim_lookup_table *table); + +extern void +lookup_table_insert(struct wim_lookup_table *table, struct wim_lookup_table_entry *lte); /* Unlinks a lookup table entry from the table; does not free it. */ -static inline void lookup_table_unlink(struct lookup_table *table, - struct lookup_table_entry *lte) +static inline void +lookup_table_unlink(struct wim_lookup_table *table, struct wim_lookup_table_entry *lte) { + wimlib_assert(!lte->unhashed); hlist_del(<e->hash_list); + wimlib_assert(table->num_entries != 0); table->num_entries--; } +extern struct wim_lookup_table_entry * +new_lookup_table_entry(); -extern struct lookup_table_entry * -lookup_table_decrement_refcnt(struct lookup_table* table, const u8 hash[]); - -extern struct lookup_table_entry * -lte_decrement_refcnt(struct lookup_table_entry *lte, - struct lookup_table *table); +extern struct wim_lookup_table_entry * +clone_lookup_table_entry(const struct wim_lookup_table_entry *lte); +extern void +print_lookup_table_entry(const struct wim_lookup_table_entry *entry, + FILE *out); -extern struct lookup_table_entry *new_lookup_table_entry(); +extern void +free_lookup_table_entry(struct wim_lookup_table_entry *lte); -extern int for_lookup_table_entry(struct lookup_table *table, - int (*visitor)(struct lookup_table_entry *, void *), - void *arg); +extern int +for_lookup_table_entry(struct wim_lookup_table *table, + int (*visitor)(struct wim_lookup_table_entry *, void *), + void *arg); -extern struct lookup_table_entry * -__lookup_resource(const struct lookup_table *table, const u8 hash[]); +extern int +cmp_streams_by_wim_position(const void *p1, const void *p2); -extern 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); - -extern int zero_out_refcnts(struct lookup_table_entry *entry, void *ignore); +extern int +for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table, + int (*visitor)(struct wim_lookup_table_entry *, + void *), + void *arg); -extern void print_lookup_table_entry(const struct lookup_table_entry *entry); +extern struct wim_lookup_table_entry * +__lookup_resource(const struct wim_lookup_table *table, const u8 hash[]); -extern int read_lookup_table(WIMStruct *w); +extern int +lookup_resource(WIMStruct *w, const tchar *path, + int lookup_flags, struct wim_dentry **dentry_ret, + struct wim_lookup_table_entry **lte_ret, u16 *stream_idx_ret); -extern void free_lookup_table(struct lookup_table *table); +extern void +lte_decrement_refcnt(struct wim_lookup_table_entry *lte, + struct wim_lookup_table *table); +#ifdef WITH_FUSE +extern void +lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte); +#endif -extern int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out); +extern int +lte_zero_out_refcnt(struct wim_lookup_table_entry *entry, void *ignore); -extern void free_lookup_table_entry(struct lookup_table_entry *lte); +extern int +lte_zero_real_refcnt(struct wim_lookup_table_entry *entry, void *ignore); -extern int dentry_resolve_ltes(struct dentry *dentry, void *__table); +extern int +lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *ignore); -/* Writes the lookup table to the output file. */ -static inline int write_lookup_table(struct lookup_table *table, FILE *out) -{ - return for_lookup_table_entry(table, write_lookup_table_entry, out); -} +extern void +inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table); -/* Unlinks and frees an entry from a lookup table. */ -static inline void lookup_table_remove(struct lookup_table *table, - struct lookup_table_entry *lte) -{ - lookup_table_unlink(table, lte); - free_lookup_table_entry(lte); -} +extern void +inode_unresolve_ltes(struct wim_inode *inode); -static inline struct resource_entry* wim_metadata_resource_entry(WIMStruct *w) -{ - return &w->image_metadata[ - w->current_image - 1].metadata_lte->resource_entry; -} +extern int +write_lookup_table_entry(struct wim_lookup_table_entry *lte, void *__out); -static inline struct lookup_table_entry * -dentry_stream_lte_resolved(const struct dentry *dentry, unsigned stream_idx) +static inline struct wim_lookup_table_entry * +inode_stream_lte_resolved(const struct wim_inode *inode, unsigned stream_idx) { - wimlib_assert(dentry->resolved); - wimlib_assert(stream_idx <= dentry->num_ads); + wimlib_assert(inode->i_resolved); + wimlib_assert(stream_idx <= inode->i_num_ads); if (stream_idx == 0) - return dentry->lte; + return inode->i_lte; else - return dentry->ads_entries[stream_idx - 1].lte; + return inode->i_ads_entries[stream_idx - 1].lte; } -static inline struct lookup_table_entry * -dentry_stream_lte_unresolved(const struct dentry *dentry, unsigned stream_idx, - const struct lookup_table *table) +static inline struct wim_lookup_table_entry * +inode_stream_lte_unresolved(const struct wim_inode *inode, unsigned stream_idx, + const struct wim_lookup_table *table) { - wimlib_assert(!dentry->resolved); - wimlib_assert(stream_idx <= dentry->num_ads); + wimlib_assert(!inode->i_resolved); + wimlib_assert(stream_idx <= inode->i_num_ads); if (!table) return NULL; if (stream_idx == 0) - return __lookup_resource(table, dentry->hash); + return __lookup_resource(table, inode->i_hash); else return __lookup_resource(table, - dentry->ads_entries[ + inode->i_ads_entries[ stream_idx - 1].hash); } -/* - * Returns the lookup table entry for stream @stream_idx of the dentry, 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. - */ -static inline struct lookup_table_entry * -dentry_stream_lte(const struct dentry *dentry, unsigned stream_idx, - const struct lookup_table *table) -{ - if (dentry->resolved) - return dentry_stream_lte_resolved(dentry, stream_idx); - else - return dentry_stream_lte_unresolved(dentry, stream_idx, table); -} +extern struct wim_lookup_table_entry * +inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx, + const struct wim_lookup_table *table); -static inline const u8 *dentry_stream_hash_unresolved(const struct dentry *dentry, - unsigned stream_idx) +static inline const u8 * +inode_stream_hash_unresolved(const struct wim_inode *inode, unsigned stream_idx) { - wimlib_assert(!dentry->resolved); - wimlib_assert(stream_idx <= dentry->num_ads); + wimlib_assert(!inode->i_resolved); + wimlib_assert(stream_idx <= inode->i_num_ads); if (stream_idx == 0) - return dentry->hash; + return inode->i_hash; else - return dentry->ads_entries[stream_idx - 1].hash; + return inode->i_ads_entries[stream_idx - 1].hash; } -static inline u16 dentry_stream_name_len(const struct dentry *dentry, - unsigned stream_idx) -{ - wimlib_assert(stream_idx <= dentry->num_ads); - if (stream_idx == 0) - return 0; - else - return dentry->ads_entries[stream_idx - 1].stream_name_len; -} -static inline const u8 *dentry_stream_hash_resolved(const struct dentry *dentry, - unsigned stream_idx) +static inline const u8 * +inode_stream_hash_resolved(const struct wim_inode *inode, unsigned stream_idx) { - struct lookup_table_entry *lte; - lte = dentry_stream_lte_resolved(dentry, stream_idx); + struct wim_lookup_table_entry *lte; + lte = inode_stream_lte_resolved(inode, stream_idx); if (lte) return lte->hash; else - return NULL; + return zero_hash; } -/* - * Returns the hash for stream @stream_idx of the dentry, where stream_idx = 0 +/* + * Returns the hash 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. */ -static inline const u8 *dentry_stream_hash(const struct dentry *dentry, - unsigned stream_idx) +static inline const u8 * +inode_stream_hash(const struct wim_inode *inode, unsigned stream_idx) +{ + if (inode->i_resolved) + return inode_stream_hash_resolved(inode, stream_idx); + else + return inode_stream_hash_unresolved(inode, stream_idx); +} + +static inline u16 +inode_stream_name_nbytes(const struct wim_inode *inode, unsigned stream_idx) { - if (dentry->resolved) - return dentry_stream_hash_resolved(dentry, stream_idx); + wimlib_assert(stream_idx <= inode->i_num_ads); + if (stream_idx == 0) + return 0; else - return dentry_stream_hash_unresolved(dentry, stream_idx); + return inode->i_ads_entries[stream_idx - 1].stream_name_nbytes; } -static inline struct lookup_table_entry * -dentry_unnamed_lte_resolved(const struct dentry *dentry) +static inline struct wim_lookup_table_entry * +inode_unnamed_lte_resolved(const struct wim_inode *inode) { - wimlib_assert(dentry->resolved); - for (unsigned i = 0; i <= dentry->num_ads; i++) - if (dentry_stream_name_len(dentry, i) == 0 && - !is_zero_hash(dentry_stream_hash_resolved(dentry, i))) - return dentry_stream_lte_resolved(dentry, i); + 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))) + { + return inode_stream_lte_resolved(inode, i); + } + } return NULL; } -static inline struct lookup_table_entry * -dentry_unnamed_lte_unresolved(const struct dentry *dentry, - const struct lookup_table *table) +static inline struct wim_lookup_table_entry * +inode_unnamed_lte_unresolved(const struct wim_inode *inode, + const struct wim_lookup_table *table) { - wimlib_assert(!dentry->resolved); - for (unsigned i = 0; i <= dentry->num_ads; i++) - if (dentry_stream_name_len(dentry, i) == 0 && - !is_zero_hash(dentry_stream_hash_unresolved(dentry, i))) - return dentry_stream_lte_unresolved(dentry, i, 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; } -extern struct lookup_table_entry * -dentry_unnamed_lte(const struct dentry *dentry, - const struct lookup_table *table); +extern struct wim_lookup_table_entry * +inode_unnamed_lte(const struct wim_inode *inode, const struct wim_lookup_table *table); + +extern u64 +lookup_table_total_stream_size(struct wim_lookup_table *table); + + +static inline void +lookup_table_insert_unhashed(struct wim_lookup_table *table, + struct wim_lookup_table_entry *lte, + struct wim_inode *back_inode, + u32 back_stream_id) +{ + lte->unhashed = 1; + lte->back_inode = back_inode; + lte->back_stream_id = back_stream_id; + list_add_tail(<e->unhashed_list, table->unhashed_streams); +} + +extern int +hash_unhashed_stream(struct wim_lookup_table_entry *lte, + struct wim_lookup_table *lookup_table, + struct wim_lookup_table_entry **lte_ret); + +extern struct wim_lookup_table_entry ** +retrieve_lte_pointer(struct wim_lookup_table_entry *lte); #endif