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 * This is a temporary object store created by the tmp_objdir
32 * facility. Disable ref updates since the objects in the store
33 * might be discarded on rollback.
35 int disable_ref_updates
;
38 * This object store is ephemeral, so there is no need to fsync.
43 * Path to the alternative object store. If this is a relative path,
44 * it is relative to the current working directory.
49 KHASH_INIT(odb_path_map
, const char * /* key: odb_path */,
50 struct object_directory
*, 1, fspathhash
, fspatheq
)
52 void prepare_alt_odb(struct repository
*r
);
53 char *compute_alternate_path(const char *path
, struct strbuf
*err
);
54 struct object_directory
*find_odb(struct repository
*r
, const char *obj_dir
);
55 typedef int alt_odb_fn(struct object_directory
*, void *);
56 int foreach_alt_odb(alt_odb_fn
, void*);
57 typedef void alternate_ref_fn(const struct object_id
*oid
, void *);
58 void for_each_alternate_ref(alternate_ref_fn
, void *);
61 * Add the directory to the on-disk alternates file; the new entry will also
62 * take effect in the current process.
64 void add_to_alternates_file(const char *dir
);
67 * Add the directory to the in-memory list of alternates (along with any
68 * recursive alternates it points to), but do not modify the on-disk alternates
71 void add_to_alternates_memory(const char *dir
);
74 * Replace the current writable object directory with the specified temporary
75 * object directory; returns the former primary object directory.
77 struct object_directory
*set_temporary_primary_odb(const char *dir
, int will_destroy
);
80 * Restore a previous ODB replaced by set_temporary_main_odb.
82 void restore_primary_odb(struct object_directory
*restore_odb
, const char *old_path
);
85 * Populate and return the loose object cache array corresponding to the
88 struct oidtree
*odb_loose_cache(struct object_directory
*odb
,
89 const struct object_id
*oid
);
91 /* Empty the loose object cache for the specified object directory. */
92 void odb_clear_loose_cache(struct object_directory
*odb
);
94 /* Clear and free the specified object directory */
95 void free_object_directory(struct object_directory
*odb
);
98 struct hashmap_entry packmap_ent
;
99 struct packed_git
*next
;
100 struct list_head mru
;
101 struct pack_window
*windows
;
103 const void *index_data
;
105 uint32_t num_objects
;
107 struct oidset bad_objects
;
111 int index
; /* for builtin/pack-objects.c */
112 unsigned pack_local
:1,
119 unsigned char hash
[GIT_MAX_RAWSZ
];
120 struct revindex_entry
*revindex
;
121 const uint32_t *revindex_data
;
122 const uint32_t *revindex_map
;
123 size_t revindex_size
;
124 /* something like ".git/objects/pack/xxxxx.pack" */
125 char pack_name
[FLEX_ARRAY
]; /* more */
128 struct multi_pack_index
;
130 static inline int pack_map_entry_cmp(const void *unused_cmp_data
,
131 const struct hashmap_entry
*entry
,
132 const struct hashmap_entry
*entry2
,
135 const char *key
= keydata
;
136 const struct packed_git
*pg1
, *pg2
;
138 pg1
= container_of(entry
, const struct packed_git
, packmap_ent
);
139 pg2
= container_of(entry2
, const struct packed_git
, packmap_ent
);
141 return strcmp(pg1
->pack_name
, key
? key
: pg2
->pack_name
);
144 struct raw_object_store
{
146 * Set of all object directories; the main directory is first (and
147 * cannot be NULL after initialization). Subsequent directories are
150 struct object_directory
*odb
;
151 struct object_directory
**odb_tail
;
152 kh_odb_path_map_t
*odb_by_path
;
154 int loaded_alternates
;
157 * A list of alternate object directories loaded from the environment;
158 * this should not generally need to be accessed directly, but will
159 * populate the "odb" list when prepare_alt_odb() is run.
164 * Objects that should be substituted by other objects
165 * (see git-replace(1)).
167 struct oidmap
*replace_map
;
168 unsigned replace_map_initialized
: 1;
169 pthread_mutex_t replace_mutex
; /* protect object replace functions */
171 struct commit_graph
*commit_graph
;
172 unsigned commit_graph_attempted
: 1; /* if loading has been attempted */
177 * should only be accessed directly by packfile.c and midx.c
179 struct multi_pack_index
*multi_pack_index
;
184 * should only be accessed directly by packfile.c
187 struct packed_git
*packed_git
;
188 /* A most-recently-used ordered version of the packed_git list. */
189 struct list_head packed_git_mru
;
192 struct packed_git
**packs
;
197 * A map of packfiles to packed_git structs for tracking which
198 * packs have been loaded already.
200 struct hashmap pack_map
;
203 * A fast, rough count of the number of objects in the repository.
204 * These two fields are not meant for direct access. Use
205 * approximate_object_count() instead.
207 unsigned long approximate_object_count
;
208 unsigned approximate_object_count_valid
: 1;
211 * Whether packed_git has already been populated with this repository's
214 unsigned packed_git_initialized
: 1;
217 struct raw_object_store
*raw_object_store_new(void);
218 void raw_object_store_clear(struct raw_object_store
*o
);
221 * Put in `buf` the name of the file in the local object database that
222 * would be used to store a loose object with the specified oid.
224 const char *loose_object_path(struct repository
*r
, struct strbuf
*buf
,
225 const struct object_id
*oid
);
227 void *map_loose_object(struct repository
*r
, const struct object_id
*oid
,
228 unsigned long *size
);
230 void *read_object_file_extended(struct repository
*r
,
231 const struct object_id
*oid
,
232 enum object_type
*type
,
233 unsigned long *size
, int lookup_replace
);
234 static inline void *repo_read_object_file(struct repository
*r
,
235 const struct object_id
*oid
,
236 enum object_type
*type
,
239 return read_object_file_extended(r
, oid
, type
, size
, 1);
241 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
242 #define read_object_file(oid, type, size) repo_read_object_file(the_repository, oid, type, size)
245 /* Read and unpack an object file into memory, write memory to an object file */
246 int oid_object_info(struct repository
*r
, const struct object_id
*, unsigned long *);
248 void hash_object_file(const struct git_hash_algo
*algo
, const void *buf
,
249 unsigned long len
, enum object_type type
,
250 struct object_id
*oid
);
252 int write_object_file_flags(const void *buf
, unsigned long len
,
253 enum object_type type
, struct object_id
*oid
,
255 static inline int write_object_file(const void *buf
, unsigned long len
,
256 enum object_type type
, struct object_id
*oid
)
258 return write_object_file_flags(buf
, len
, type
, oid
, 0);
261 int write_object_file_literally(const void *buf
, unsigned long len
,
262 const char *type
, struct object_id
*oid
,
266 * Add an object file to the in-memory object store, without writing it
269 * Callers are responsible for calling write_object_file to record the
270 * object in persistent storage before writing any other new objects
273 int pretend_object_file(void *, unsigned long, enum object_type
,
274 struct object_id
*oid
);
276 int force_object_loose(const struct object_id
*oid
, time_t mtime
);
279 * Open the loose object at path, check its hash, and return the contents,
280 * use the "oi" argument to assert things about the object, or e.g. populate its
281 * type, and size. If the object is a blob, then "contents" may return NULL,
282 * to allow streaming of large blobs.
284 * Returns 0 on success, negative on error (details may be written to stderr).
286 int read_loose_object(const char *path
,
287 const struct object_id
*expected_oid
,
288 struct object_id
*real_oid
,
290 struct object_info
*oi
);
292 /* Retry packed storage after checking packed and loose storage */
293 #define HAS_OBJECT_RECHECK_PACKED 1
296 * Returns 1 if the object exists. This function will not lazily fetch objects
297 * in a partial clone.
299 int has_object(struct repository
*r
, const struct object_id
*oid
,
303 * These macros and functions are deprecated. If checking existence for an
304 * object that is likely to be missing and/or whose absence is relatively
305 * inconsequential (or is consequential but the caller is prepared to handle
306 * it), use has_object(), which has better defaults (no lazy fetch in a partial
307 * clone and no rechecking of packed storage). In the unlikely event that a
308 * caller needs to assert existence of an object that it fully expects to
309 * exist, and wants to trigger a lazy fetch in a partial clone, use
310 * oid_object_info_extended() with a NULL struct object_info.
312 * These functions can be removed once all callers have migrated to
313 * has_object() and/or oid_object_info_extended().
315 int repo_has_object_file(struct repository
*r
, const struct object_id
*oid
);
316 int repo_has_object_file_with_flags(struct repository
*r
,
317 const struct object_id
*oid
, int flags
);
318 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
319 #define has_object_file(oid) repo_has_object_file(the_repository, oid)
320 #define has_object_file_with_flags(oid, flags) repo_has_object_file_with_flags(the_repository, oid, flags)
324 * Return true iff an alternate object database has a loose object
325 * with the specified name. This function does not respect replace
328 int has_loose_object_nonlocal(const struct object_id
*);
331 * format_object_header() is a thin wrapper around s xsnprintf() that
332 * writes the initial "<type> <obj-len>" part of the loose object
333 * header. It returns the size that snprintf() returns + 1.
335 int format_object_header(char *str
, size_t size
, enum object_type type
,
338 void assert_oid_type(const struct object_id
*oid
, enum object_type expect
);
341 * Enabling the object read lock allows multiple threads to safely call the
342 * following functions in parallel: repo_read_object_file(), read_object_file(),
343 * read_object_file_extended(), read_object_with_reference(), read_object(),
344 * oid_object_info() and oid_object_info_extended().
346 * obj_read_lock() and obj_read_unlock() may also be used to protect other
347 * section which cannot execute in parallel with object reading. Since the used
348 * lock is a recursive mutex, these sections can even contain calls to object
349 * reading functions. However, beware that in these cases zlib inflation won't
350 * be performed in parallel, losing performance.
352 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
353 * any of its callees end up calling it, this recursive call won't benefit from
354 * parallel inflation.
356 void enable_obj_read_lock(void);
357 void disable_obj_read_lock(void);
359 extern int obj_read_use_lock
;
360 extern pthread_mutex_t obj_read_mutex
;
362 static inline void obj_read_lock(void)
364 if(obj_read_use_lock
)
365 pthread_mutex_lock(&obj_read_mutex
);
368 static inline void obj_read_unlock(void)
370 if(obj_read_use_lock
)
371 pthread_mutex_unlock(&obj_read_mutex
);
376 enum object_type
*typep
;
377 unsigned long *sizep
;
379 struct object_id
*delta_base_oid
;
380 struct strbuf
*type_name
;
393 * ... Nothing to expose in this case
396 * ... Nothing to expose in this case
400 struct packed_git
*pack
;
402 unsigned int is_delta
;
408 * Initializer for a "struct object_info" that wants no items. You may
409 * also memset() the memory to all-zeroes.
411 #define OBJECT_INFO_INIT { 0 }
413 /* Invoke lookup_replace_object() on the given hash */
414 #define OBJECT_INFO_LOOKUP_REPLACE 1
415 /* Allow reading from a loose object file of unknown/bogus type */
416 #define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
417 /* Do not retry packed storage after checking packed and loose storage */
418 #define OBJECT_INFO_QUICK 8
419 /* Do not check loose object */
420 #define OBJECT_INFO_IGNORE_LOOSE 16
422 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
425 #define OBJECT_INFO_SKIP_FETCH_OBJECT 32
427 * This is meant for bulk prefetching of missing blobs in a partial
428 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
430 #define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
432 int oid_object_info_extended(struct repository
*r
,
433 const struct object_id
*,
434 struct object_info
*, unsigned flags
);
437 * Iterate over the files in the loose-object parts of the object
438 * directory "path", triggering the following callbacks:
440 * - loose_object is called for each loose object we find.
442 * - loose_cruft is called for any files that do not appear to be
443 * loose objects. Note that we only look in the loose object
444 * directories "objects/[0-9a-f]{2}/", so we will not report
445 * "objects/foobar" as cruft.
447 * - loose_subdir is called for each top-level hashed subdirectory
448 * of the object directory (e.g., "$OBJDIR/f0"). It is called
449 * after the objects in the directory are processed.
451 * Any callback that is NULL will be ignored. Callbacks returning non-zero
452 * will end the iteration.
454 * In the "buf" variant, "path" is a strbuf which will also be used as a
455 * scratch buffer, but restored to its original contents before
456 * the function returns.
458 typedef int each_loose_object_fn(const struct object_id
*oid
,
461 typedef int each_loose_cruft_fn(const char *basename
,
464 typedef int each_loose_subdir_fn(unsigned int nr
,
467 int for_each_file_in_obj_subdir(unsigned int subdir_nr
,
469 each_loose_object_fn obj_cb
,
470 each_loose_cruft_fn cruft_cb
,
471 each_loose_subdir_fn subdir_cb
,
473 int for_each_loose_file_in_objdir(const char *path
,
474 each_loose_object_fn obj_cb
,
475 each_loose_cruft_fn cruft_cb
,
476 each_loose_subdir_fn subdir_cb
,
478 int for_each_loose_file_in_objdir_buf(struct strbuf
*path
,
479 each_loose_object_fn obj_cb
,
480 each_loose_cruft_fn cruft_cb
,
481 each_loose_subdir_fn subdir_cb
,
484 /* Flags for for_each_*_object() below. */
485 enum for_each_object_flags
{
486 /* Iterate only over local objects, not alternates. */
487 FOR_EACH_OBJECT_LOCAL_ONLY
= (1<<0),
489 /* Only iterate over packs obtained from the promisor remote. */
490 FOR_EACH_OBJECT_PROMISOR_ONLY
= (1<<1),
493 * Visit objects within a pack in packfile order rather than .idx order
495 FOR_EACH_OBJECT_PACK_ORDER
= (1<<2),
497 /* Only iterate over packs that are not marked as kept in-core. */
498 FOR_EACH_OBJECT_SKIP_IN_CORE_KEPT_PACKS
= (1<<3),
500 /* Only iterate over packs that do not have .keep files. */
501 FOR_EACH_OBJECT_SKIP_ON_DISK_KEPT_PACKS
= (1<<4),
505 * Iterate over all accessible loose objects without respect to
506 * reachability. By default, this includes both local and alternate objects.
507 * The order in which objects are visited is unspecified.
509 * Any flags specific to packs are ignored.
511 int for_each_loose_object(each_loose_object_fn
, void *,
512 enum for_each_object_flags flags
);
515 * Iterate over all accessible packed objects without respect to reachability.
516 * By default, this includes both local and alternate packs.
518 * Note that some objects may appear twice if they are found in multiple packs.
519 * Each pack is visited in an unspecified order. By default, objects within a
520 * pack are visited in pack-idx order (i.e., sorted by oid).
522 typedef int each_packed_object_fn(const struct object_id
*oid
,
523 struct packed_git
*pack
,
526 int for_each_object_in_pack(struct packed_git
*p
,
527 each_packed_object_fn
, void *data
,
528 enum for_each_object_flags flags
);
529 int for_each_packed_object(each_packed_object_fn
, void *,
530 enum for_each_object_flags flags
);
532 #endif /* OBJECT_STORE_H */