9 #include "thread-utils.h"
15 struct object_directory
{
16 struct object_directory
*next
;
19 * Used to store the results of readdir(3) calls when we are OK
20 * sacrificing accuracy due to races for speed. That includes
21 * object existence with OBJECT_INFO_QUICK, as well as
22 * our search for unique abbreviated hashes. Don't use it for tasks
23 * requiring greater accuracy!
25 * Be sure to call odb_load_loose_cache() before using.
27 uint32_t loose_objects_subdir_seen
[8]; /* 256 bits */
28 struct oidtree
*loose_objects_cache
;
31 * Path to the alternative object store. If this is a relative path,
32 * it is relative to the current working directory.
37 KHASH_INIT(odb_path_map
, const char * /* key: odb_path */,
38 struct object_directory
*, 1, fspathhash
, fspatheq
)
40 void prepare_alt_odb(struct repository
*r
);
41 char *compute_alternate_path(const char *path
, struct strbuf
*err
);
42 struct object_directory
*find_odb(struct repository
*r
, const char *obj_dir
);
43 typedef int alt_odb_fn(struct object_directory
*, void *);
44 int foreach_alt_odb(alt_odb_fn
, void*);
45 typedef void alternate_ref_fn(const struct object_id
*oid
, void *);
46 void for_each_alternate_ref(alternate_ref_fn
, void *);
49 * Add the directory to the on-disk alternates file; the new entry will also
50 * take effect in the current process.
52 void add_to_alternates_file(const char *dir
);
55 * Add the directory to the in-memory list of alternates (along with any
56 * recursive alternates it points to), but do not modify the on-disk alternates
59 void add_to_alternates_memory(const char *dir
);
62 * Populate and return the loose object cache array corresponding to the
65 struct oidtree
*odb_loose_cache(struct object_directory
*odb
,
66 const struct object_id
*oid
);
68 /* Empty the loose object cache for the specified object directory. */
69 void odb_clear_loose_cache(struct object_directory
*odb
);
72 struct hashmap_entry packmap_ent
;
73 struct packed_git
*next
;
75 struct pack_window
*windows
;
77 const void *index_data
;
81 struct oidset bad_objects
;
85 int index
; /* for builtin/pack-objects.c */
86 unsigned pack_local
:1,
93 unsigned char hash
[GIT_MAX_RAWSZ
];
94 struct revindex_entry
*revindex
;
95 const uint32_t *revindex_data
;
96 const uint32_t *revindex_map
;
98 /* something like ".git/objects/pack/xxxxx.pack" */
99 char pack_name
[FLEX_ARRAY
]; /* more */
102 struct multi_pack_index
;
104 static inline int pack_map_entry_cmp(const void *unused_cmp_data
,
105 const struct hashmap_entry
*entry
,
106 const struct hashmap_entry
*entry2
,
109 const char *key
= keydata
;
110 const struct packed_git
*pg1
, *pg2
;
112 pg1
= container_of(entry
, const struct packed_git
, packmap_ent
);
113 pg2
= container_of(entry2
, const struct packed_git
, packmap_ent
);
115 return strcmp(pg1
->pack_name
, key
? key
: pg2
->pack_name
);
118 struct raw_object_store
{
120 * Set of all object directories; the main directory is first (and
121 * cannot be NULL after initialization). Subsequent directories are
124 struct object_directory
*odb
;
125 struct object_directory
**odb_tail
;
126 kh_odb_path_map_t
*odb_by_path
;
128 int loaded_alternates
;
131 * A list of alternate object directories loaded from the environment;
132 * this should not generally need to be accessed directly, but will
133 * populate the "odb" list when prepare_alt_odb() is run.
138 * Objects that should be substituted by other objects
139 * (see git-replace(1)).
141 struct oidmap
*replace_map
;
142 unsigned replace_map_initialized
: 1;
143 pthread_mutex_t replace_mutex
; /* protect object replace functions */
145 struct commit_graph
*commit_graph
;
146 unsigned commit_graph_attempted
: 1; /* if loading has been attempted */
151 * should only be accessed directly by packfile.c and midx.c
153 struct multi_pack_index
*multi_pack_index
;
158 * should only be accessed directly by packfile.c
161 struct packed_git
*packed_git
;
162 /* A most-recently-used ordered version of the packed_git list. */
163 struct list_head packed_git_mru
;
166 struct packed_git
**packs
;
171 * A map of packfiles to packed_git structs for tracking which
172 * packs have been loaded already.
174 struct hashmap pack_map
;
177 * A fast, rough count of the number of objects in the repository.
178 * These two fields are not meant for direct access. Use
179 * approximate_object_count() instead.
181 unsigned long approximate_object_count
;
182 unsigned approximate_object_count_valid
: 1;
185 * Whether packed_git has already been populated with this repository's
188 unsigned packed_git_initialized
: 1;
191 struct raw_object_store
*raw_object_store_new(void);
192 void raw_object_store_clear(struct raw_object_store
*o
);
195 * Put in `buf` the name of the file in the local object database that
196 * would be used to store a loose object with the specified oid.
198 const char *loose_object_path(struct repository
*r
, struct strbuf
*buf
,
199 const struct object_id
*oid
);
201 void *map_loose_object(struct repository
*r
, const struct object_id
*oid
,
202 unsigned long *size
);
204 void *read_object_file_extended(struct repository
*r
,
205 const struct object_id
*oid
,
206 enum object_type
*type
,
207 unsigned long *size
, int lookup_replace
);
208 static inline void *repo_read_object_file(struct repository
*r
,
209 const struct object_id
*oid
,
210 enum object_type
*type
,
213 return read_object_file_extended(r
, oid
, type
, size
, 1);
215 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
216 #define read_object_file(oid, type, size) repo_read_object_file(the_repository, oid, type, size)
219 /* Read and unpack an object file into memory, write memory to an object file */
220 int oid_object_info(struct repository
*r
, const struct object_id
*, unsigned long *);
222 int hash_object_file(const struct git_hash_algo
*algo
, const void *buf
,
223 unsigned long len
, const char *type
,
224 struct object_id
*oid
);
226 int write_object_file(const void *buf
, unsigned long len
,
227 const char *type
, struct object_id
*oid
);
229 int hash_object_file_literally(const void *buf
, unsigned long len
,
230 const char *type
, struct object_id
*oid
,
234 * Add an object file to the in-memory object store, without writing it
237 * Callers are responsible for calling write_object_file to record the
238 * object in persistent storage before writing any other new objects
241 int pretend_object_file(void *, unsigned long, enum object_type
,
242 struct object_id
*oid
);
244 int force_object_loose(const struct object_id
*oid
, time_t mtime
);
247 * Open the loose object at path, check its hash, and return the contents,
248 * use the "oi" argument to assert things about the object, or e.g. populate its
249 * type, and size. If the object is a blob, then "contents" may return NULL,
250 * to allow streaming of large blobs.
252 * Returns 0 on success, negative on error (details may be written to stderr).
254 int read_loose_object(const char *path
,
255 const struct object_id
*expected_oid
,
256 struct object_id
*real_oid
,
258 struct object_info
*oi
);
260 /* Retry packed storage after checking packed and loose storage */
261 #define HAS_OBJECT_RECHECK_PACKED 1
264 * Returns 1 if the object exists. This function will not lazily fetch objects
265 * in a partial clone.
267 int has_object(struct repository
*r
, const struct object_id
*oid
,
271 * These macros and functions are deprecated. If checking existence for an
272 * object that is likely to be missing and/or whose absence is relatively
273 * inconsequential (or is consequential but the caller is prepared to handle
274 * it), use has_object(), which has better defaults (no lazy fetch in a partial
275 * clone and no rechecking of packed storage). In the unlikely event that a
276 * caller needs to assert existence of an object that it fully expects to
277 * exist, and wants to trigger a lazy fetch in a partial clone, use
278 * oid_object_info_extended() with a NULL struct object_info.
280 * These functions can be removed once all callers have migrated to
281 * has_object() and/or oid_object_info_extended().
283 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
284 #define has_sha1_file_with_flags(sha1, flags) repo_has_sha1_file_with_flags(the_repository, sha1, flags)
285 #define has_sha1_file(sha1) repo_has_sha1_file(the_repository, sha1)
287 int repo_has_object_file(struct repository
*r
, const struct object_id
*oid
);
288 int repo_has_object_file_with_flags(struct repository
*r
,
289 const struct object_id
*oid
, int flags
);
290 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
291 #define has_object_file(oid) repo_has_object_file(the_repository, oid)
292 #define has_object_file_with_flags(oid, flags) repo_has_object_file_with_flags(the_repository, oid, flags)
296 * Return true iff an alternate object database has a loose object
297 * with the specified name. This function does not respect replace
300 int has_loose_object_nonlocal(const struct object_id
*);
302 void assert_oid_type(const struct object_id
*oid
, enum object_type expect
);
305 * Enabling the object read lock allows multiple threads to safely call the
306 * following functions in parallel: repo_read_object_file(), read_object_file(),
307 * read_object_file_extended(), read_object_with_reference(), read_object(),
308 * oid_object_info() and oid_object_info_extended().
310 * obj_read_lock() and obj_read_unlock() may also be used to protect other
311 * section which cannot execute in parallel with object reading. Since the used
312 * lock is a recursive mutex, these sections can even contain calls to object
313 * reading functions. However, beware that in these cases zlib inflation won't
314 * be performed in parallel, losing performance.
316 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
317 * any of its callees end up calling it, this recursive call won't benefit from
318 * parallel inflation.
320 void enable_obj_read_lock(void);
321 void disable_obj_read_lock(void);
323 extern int obj_read_use_lock
;
324 extern pthread_mutex_t obj_read_mutex
;
326 static inline void obj_read_lock(void)
328 if(obj_read_use_lock
)
329 pthread_mutex_lock(&obj_read_mutex
);
332 static inline void obj_read_unlock(void)
334 if(obj_read_use_lock
)
335 pthread_mutex_unlock(&obj_read_mutex
);
340 enum object_type
*typep
;
341 unsigned long *sizep
;
343 struct object_id
*delta_base_oid
;
344 struct strbuf
*type_name
;
357 * ... Nothing to expose in this case
360 * ... Nothing to expose in this case
364 struct packed_git
*pack
;
366 unsigned int is_delta
;
372 * Initializer for a "struct object_info" that wants no items. You may
373 * also memset() the memory to all-zeroes.
375 #define OBJECT_INFO_INIT { 0 }
377 /* Invoke lookup_replace_object() on the given hash */
378 #define OBJECT_INFO_LOOKUP_REPLACE 1
379 /* Allow reading from a loose object file of unknown/bogus type */
380 #define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
381 /* Do not retry packed storage after checking packed and loose storage */
382 #define OBJECT_INFO_QUICK 8
383 /* Do not check loose object */
384 #define OBJECT_INFO_IGNORE_LOOSE 16
386 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
389 #define OBJECT_INFO_SKIP_FETCH_OBJECT 32
391 * This is meant for bulk prefetching of missing blobs in a partial
392 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
394 #define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
396 int oid_object_info_extended(struct repository
*r
,
397 const struct object_id
*,
398 struct object_info
*, unsigned flags
);
401 * Iterate over the files in the loose-object parts of the object
402 * directory "path", triggering the following callbacks:
404 * - loose_object is called for each loose object we find.
406 * - loose_cruft is called for any files that do not appear to be
407 * loose objects. Note that we only look in the loose object
408 * directories "objects/[0-9a-f]{2}/", so we will not report
409 * "objects/foobar" as cruft.
411 * - loose_subdir is called for each top-level hashed subdirectory
412 * of the object directory (e.g., "$OBJDIR/f0"). It is called
413 * after the objects in the directory are processed.
415 * Any callback that is NULL will be ignored. Callbacks returning non-zero
416 * will end the iteration.
418 * In the "buf" variant, "path" is a strbuf which will also be used as a
419 * scratch buffer, but restored to its original contents before
420 * the function returns.
422 typedef int each_loose_object_fn(const struct object_id
*oid
,
425 typedef int each_loose_cruft_fn(const char *basename
,
428 typedef int each_loose_subdir_fn(unsigned int nr
,
431 int for_each_file_in_obj_subdir(unsigned int subdir_nr
,
433 each_loose_object_fn obj_cb
,
434 each_loose_cruft_fn cruft_cb
,
435 each_loose_subdir_fn subdir_cb
,
437 int for_each_loose_file_in_objdir(const char *path
,
438 each_loose_object_fn obj_cb
,
439 each_loose_cruft_fn cruft_cb
,
440 each_loose_subdir_fn subdir_cb
,
442 int for_each_loose_file_in_objdir_buf(struct strbuf
*path
,
443 each_loose_object_fn obj_cb
,
444 each_loose_cruft_fn cruft_cb
,
445 each_loose_subdir_fn subdir_cb
,
448 /* Flags for for_each_*_object() below. */
449 enum for_each_object_flags
{
450 /* Iterate only over local objects, not alternates. */
451 FOR_EACH_OBJECT_LOCAL_ONLY
= (1<<0),
453 /* Only iterate over packs obtained from the promisor remote. */
454 FOR_EACH_OBJECT_PROMISOR_ONLY
= (1<<1),
457 * Visit objects within a pack in packfile order rather than .idx order
459 FOR_EACH_OBJECT_PACK_ORDER
= (1<<2),
461 /* Only iterate over packs that are not marked as kept in-core. */
462 FOR_EACH_OBJECT_SKIP_IN_CORE_KEPT_PACKS
= (1<<3),
464 /* Only iterate over packs that do not have .keep files. */
465 FOR_EACH_OBJECT_SKIP_ON_DISK_KEPT_PACKS
= (1<<4),
469 * Iterate over all accessible loose objects without respect to
470 * reachability. By default, this includes both local and alternate objects.
471 * The order in which objects are visited is unspecified.
473 * Any flags specific to packs are ignored.
475 int for_each_loose_object(each_loose_object_fn
, void *,
476 enum for_each_object_flags flags
);
479 * Iterate over all accessible packed objects without respect to reachability.
480 * By default, this includes both local and alternate packs.
482 * Note that some objects may appear twice if they are found in multiple packs.
483 * Each pack is visited in an unspecified order. By default, objects within a
484 * pack are visited in pack-idx order (i.e., sorted by oid).
486 typedef int each_packed_object_fn(const struct object_id
*oid
,
487 struct packed_git
*pack
,
490 int for_each_object_in_pack(struct packed_git
*p
,
491 each_packed_object_fn
, void *data
,
492 enum for_each_object_flags flags
);
493 int for_each_packed_object(each_packed_object_fn
, void *,
494 enum for_each_object_flags flags
);
496 #endif /* OBJECT_STORE_H */