1 #ifndef _WIMLIB_LOOKUP_TABLE_H
2 #define _WIMLIB_LOOKUP_TABLE_H
3 #include "wimlib_internal.h"
8 /* Size of each lookup table entry in the WIM file. */
9 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
11 #define LOOKUP_FLAG_ADS_OK 0x00000001
12 #define LOOKUP_FLAG_DIRECTORY_OK 0x00000002
15 //#define LOOKUP_FLAG_FOLLOW_SYMLINKS 0x00000004
18 /* A lookup table that is used to translate the hash codes of dentries into the
19 * offsets and sizes of uncompressed or compressed file resources. It is
20 * implemented as a hash table. */
22 struct hlist_head *array;
30 struct ntfs_location {
32 char *stream_name_utf16;
33 u16 stream_name_utf16_num_chars;
34 struct _ntfs_volume **ntfs_vol_p;
35 bool is_reparse_point;
39 /* An enumerated type that identifies where the stream corresponding to this
40 * lookup table entry is actually located.
42 * If we open a WIM and read its lookup table, the location is set to
43 * RESOURCE_IN_WIM since all the streams will initially be located in the WIM.
44 * However, to deal with problems such as image capture and image mount, we
45 * allow the actual location of the stream to be somewhere else, such as an
48 enum resource_location {
49 /* The lookup table entry does not correspond to a stream (this state
50 * should exist only temporarily) */
51 RESOURCE_NONEXISTENT = 0,
53 /* The stream resource is located in a WIM file. The WIMStruct for the
54 * WIM file will be pointed to by the @wim member. */
57 /* The stream resource is located in an external file. The name of the
58 * file will be provided by @file_on_disk member. In addition, if
59 * @file_on_disk_fp is not NULL, it will be an open FILE * to the file.
61 RESOURCE_IN_FILE_ON_DISK,
63 /* The stream resource is located in an external file in the staging
64 * directory for a read-write mount. */
65 RESOURCE_IN_STAGING_FILE,
67 /* The stream resource is directly attached in an in-memory buffer
68 * pointed to by @attached_buffer. */
69 RESOURCE_IN_ATTACHED_BUFFER,
71 /* The stream resource is located in an NTFS volume. It is identified
72 * by volume, filename, data stream name, and by whether it is a reparse
73 * point or not. @ntfs_loc points to a structure containing this
75 RESOURCE_IN_NTFS_VOLUME,
79 * An entry in the lookup table in the WIM file.
81 * It is used to find data streams for files in the WIM.
83 * Metadata resources and reparse point data buffers will also have lookup table
84 * entries associated with the data.
86 * The lookup_table_entry for a given dentry or alternate stream entry in the
87 * WIM is found using the SHA1 message digest field.
89 struct lookup_table_entry {
91 /* List of lookup table entries in this hash bucket */
92 struct hlist_node hash_list;
94 /* Location and size of the stream in the WIM, whether it is compressed
95 * or not, and whether it's a metadata resource or not. This is an
97 struct resource_entry resource_entry;
99 /* Specifies which part of the split WIM the resource is located in.
100 * This is on on-disk field.
102 * In stand-alone WIMs, this must be 1.
104 * In split WIMs, every split WIM part has its own lookup table, and in
105 * read_lookup_table() it's currently expected that the part number of
106 * each lookup table entry in a split WIM part's lookup table is the
107 * same as the part number of that split WIM part. So this makes this
108 * field redundant since we store a pointer to the corresponding
109 * WIMStruct in the lookup table entry anyway.
113 /* See enum resource_location above */
114 u16 resource_location;
117 * Number of times this lookup table entry is referenced by dentries.
118 * Unfortunately, this field is not always set correctly in Microsoft's
119 * WIMs, so we have no choice but to fix it if more references to the
120 * lookup table entry are found than stated here. */
124 /* (On-disk field) SHA1 message digest of the stream referenced
125 * by this lookup table entry */
126 u8 hash[SHA1_HASH_SIZE];
128 /* First 4 or 8 bytes of the SHA1 message digest, used for
129 * inserting the entry into the hash table. Since the SHA1
130 * message digest can be considered random, we don't really need
131 * the full 20 byte hash just to insert the entry in a hash
140 /* Pointers to somewhere where the stream is actually located. See the
141 * comments for the @resource_location field above. */
145 char *staging_file_name;
148 struct ntfs_location *ntfs_loc;
152 /* @file_on_disk_fp and @attr are both used to cache file/stream
153 * handles so we don't have re-open them on every read */
155 /* Valid iff resource_location == RESOURCE_IN_FILE_ON_DISK */
156 FILE *file_on_disk_fp;
158 /* Valid iff resource_location == RESOURCE_IN_NTFS_VOLUME */
159 struct _ntfs_attr *attr;
162 /* Pointer to inode that contains the opened file descriptors to
163 * this stream (valid iff resource_location ==
164 * RESOURCE_IN_STAGING_FILE) */
165 struct inode *lte_inode;
168 /* When a WIM file is written, out_refcnt starts at 0 and is incremented
169 * whenever the file resource pointed to by this lookup table entry
170 * needs to be written. The file resource only need to be written when
171 * out_refcnt is nonzero, since otherwise it is not referenced by any
178 /* When a WIM file is written, @output_resource_entry is filled
179 * in with the resource entry for the output WIM. This will not
180 * necessarily be the same as the @resource_entry since: - The
181 * stream may have a different offset in the new WIM - The
182 * stream may have a different compressed size in the new WIM if
183 * the compression type changed
185 struct resource_entry output_resource_entry;
187 struct list_head msg_list;
191 /* This field is used for the special hardlink or symlink image
192 * extraction mode. In these mode, all identical files are linked
193 * together, and @extracted_file will be set to the filename of the
194 * first extracted file containing this stream. */
195 char *extracted_file;
197 /* List of lookup table entries that correspond to streams that have
198 * been extracted to the staging directory when modifying a read-write
200 struct list_head staging_list;
202 /* Temporary field for creating a singly linked list. */
203 struct lookup_table_entry *next_lte_in_swm;
207 static inline u64 wim_resource_size(const struct lookup_table_entry *lte)
209 return lte->resource_entry.original_size;
212 static inline u64 wim_resource_chunks(const struct lookup_table_entry *lte)
214 return (wim_resource_size(lte) + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE;
218 wim_resource_compressed_size(const struct lookup_table_entry *lte)
220 return lte->resource_entry.size;
224 * XXX Probably should store the compression type directly in the lookup table
228 wim_resource_compression_type(const struct lookup_table_entry *lte)
230 if (!(lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
231 || lte->resource_location != RESOURCE_IN_WIM)
232 return WIM_COMPRESSION_TYPE_NONE;
233 return wimlib_get_compression_type(lte->wim);
237 extern struct lookup_table *new_lookup_table(size_t capacity);
239 extern void lookup_table_insert(struct lookup_table *table,
240 struct lookup_table_entry *lte);
242 /* Unlinks a lookup table entry from the table; does not free it. */
243 static inline void lookup_table_unlink(struct lookup_table *table,
244 struct lookup_table_entry *lte)
246 hlist_del(<e->hash_list);
247 table->num_entries--;
250 extern struct lookup_table_entry *new_lookup_table_entry();
252 extern struct lookup_table_entry *
253 clone_lookup_table_entry(const struct lookup_table_entry *lte);
255 extern int for_lookup_table_entry(struct lookup_table *table,
256 int (*visitor)(struct lookup_table_entry *, void *),
259 extern struct lookup_table_entry *
260 __lookup_resource(const struct lookup_table *table, const u8 hash[]);
262 extern int lookup_resource(WIMStruct *w, const char *path,
263 int lookup_flags, struct dentry **dentry_ret,
264 struct lookup_table_entry **lte_ret,
265 u16 *stream_idx_ret);
267 extern void lte_decrement_refcnt(struct lookup_table_entry *lte,
268 struct lookup_table *table);
270 extern void lte_decrement_num_opened_fds(struct lookup_table_entry *lte);
273 extern int lte_zero_out_refcnt(struct lookup_table_entry *entry, void *ignore);
274 extern int lte_zero_real_refcnt(struct lookup_table_entry *entry, void *ignore);
275 extern int lte_zero_extracted_file(struct lookup_table_entry *lte, void *ignore);
276 extern int lte_free_extracted_file(struct lookup_table_entry *lte, void *ignore);
278 extern void print_lookup_table_entry(const struct lookup_table_entry *entry);
280 extern int read_lookup_table(WIMStruct *w);
282 extern void free_lookup_table(struct lookup_table *table);
284 extern int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out);
286 extern void free_lookup_table_entry(struct lookup_table_entry *lte);
288 extern int dentry_resolve_ltes(struct dentry *dentry, void *__table);
289 extern int dentry_unresolve_ltes(struct dentry *dentry, void *ignore);
291 /* Writes the lookup table to the output file. */
292 static inline int write_lookup_table(struct lookup_table *table, FILE *out)
294 return for_lookup_table_entry(table, write_lookup_table_entry, out);
297 /* Unlinks and frees an entry from a lookup table. */
298 static inline void lookup_table_remove(struct lookup_table *table,
299 struct lookup_table_entry *lte)
301 lookup_table_unlink(table, lte);
302 free_lookup_table_entry(lte);
305 static inline struct resource_entry* wim_metadata_resource_entry(WIMStruct *w)
307 return &w->image_metadata[
308 w->current_image - 1].metadata_lte->resource_entry;
311 static inline struct lookup_table_entry *
312 inode_stream_lte_resolved(const struct inode *inode, unsigned stream_idx)
314 wimlib_assert(inode->resolved);
315 wimlib_assert(stream_idx <= inode->num_ads);
319 return inode->ads_entries[stream_idx - 1].lte;
322 static inline struct lookup_table_entry *
323 inode_stream_lte_unresolved(const struct inode *inode, unsigned stream_idx,
324 const struct lookup_table *table)
326 wimlib_assert(!inode->resolved);
327 wimlib_assert(stream_idx <= inode->num_ads);
331 return __lookup_resource(table, inode->hash);
333 return __lookup_resource(table,
335 stream_idx - 1].hash);
338 * Returns the lookup table entry for stream @stream_idx of the inode, where
339 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
340 * corresponds to an alternate data stream.
342 * This works for both resolved and un-resolved dentries.
344 static inline struct lookup_table_entry *
345 inode_stream_lte(const struct inode *inode, unsigned stream_idx,
346 const struct lookup_table *table)
349 return inode_stream_lte_resolved(inode, stream_idx);
351 return inode_stream_lte_unresolved(inode, stream_idx, table);
355 static inline const u8 *inode_stream_hash_unresolved(const struct inode *inode,
358 wimlib_assert(!inode->resolved);
359 wimlib_assert(stream_idx <= inode->num_ads);
363 return inode->ads_entries[stream_idx - 1].hash;
367 static inline const u8 *inode_stream_hash_resolved(const struct inode *inode,
370 struct lookup_table_entry *lte;
371 lte = inode_stream_lte_resolved(inode, stream_idx);
379 * Returns the hash for stream @stream_idx of the inode, where stream_idx = 0
380 * means the default un-named file stream, and stream_idx >= 1 corresponds to an
381 * alternate data stream.
383 * This works for both resolved and un-resolved dentries.
385 static inline const u8 *inode_stream_hash(const struct inode *inode,
389 return inode_stream_hash_resolved(inode, stream_idx);
391 return inode_stream_hash_unresolved(inode, stream_idx);
394 static inline u16 inode_stream_name_len(const struct inode *inode,
397 wimlib_assert(stream_idx <= inode->num_ads);
401 return inode->ads_entries[stream_idx - 1].stream_name_len;
404 static inline struct lookup_table_entry *
405 inode_unnamed_lte_resolved(const struct inode *inode)
407 wimlib_assert(inode->resolved);
408 for (unsigned i = 0; i <= inode->num_ads; i++)
409 if (inode_stream_name_len(inode, i) == 0 &&
410 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
411 return inode_stream_lte_resolved(inode, i);
415 static inline struct lookup_table_entry *
416 inode_unnamed_lte_unresolved(const struct inode *inode,
417 const struct lookup_table *table)
419 wimlib_assert(!inode->resolved);
420 for (unsigned i = 0; i <= inode->num_ads; i++)
421 if (inode_stream_name_len(inode, i) == 0 &&
422 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
423 return inode_stream_lte_unresolved(inode, i, table);
427 extern struct lookup_table_entry *
428 inode_unnamed_lte(const struct inode *inode,
429 const struct lookup_table *table);