6 #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.
50 const void *(*read
)(struct input_stream
*, unsigned long *len
);
55 KHASH_INIT(odb_path_map
, const char * /* key: odb_path */,
56 struct object_directory
*, 1, fspathhash
, fspatheq
)
58 void prepare_alt_odb(struct repository
*r
);
59 int has_alt_odb(struct repository
*r
);
60 char *compute_alternate_path(const char *path
, struct strbuf
*err
);
61 struct object_directory
*find_odb(struct repository
*r
, const char *obj_dir
);
62 typedef int alt_odb_fn(struct object_directory
*, void *);
63 int foreach_alt_odb(alt_odb_fn
, void*);
64 typedef void alternate_ref_fn(const struct object_id
*oid
, void *);
65 void for_each_alternate_ref(alternate_ref_fn
, void *);
68 * Add the directory to the on-disk alternates file; the new entry will also
69 * take effect in the current process.
71 void add_to_alternates_file(const char *dir
);
74 * Add the directory to the in-memory list of alternates (along with any
75 * recursive alternates it points to), but do not modify the on-disk alternates
78 void add_to_alternates_memory(const char *dir
);
81 * Replace the current writable object directory with the specified temporary
82 * object directory; returns the former primary object directory.
84 struct object_directory
*set_temporary_primary_odb(const char *dir
, int will_destroy
);
87 * Restore a previous ODB replaced by set_temporary_main_odb.
89 void restore_primary_odb(struct object_directory
*restore_odb
, const char *old_path
);
92 * Populate and return the loose object cache array corresponding to the
95 struct oidtree
*odb_loose_cache(struct object_directory
*odb
,
96 const struct object_id
*oid
);
98 /* Empty the loose object cache for the specified object directory. */
99 void odb_clear_loose_cache(struct object_directory
*odb
);
101 /* Clear and free the specified object directory */
102 void free_object_directory(struct object_directory
*odb
);
105 struct hashmap_entry packmap_ent
;
106 struct packed_git
*next
;
107 struct list_head mru
;
108 struct pack_window
*windows
;
110 const void *index_data
;
112 uint32_t num_objects
;
114 struct oidset bad_objects
;
118 int index
; /* for builtin/pack-objects.c */
119 unsigned pack_local
:1,
127 unsigned char hash
[GIT_MAX_RAWSZ
];
128 struct revindex_entry
*revindex
;
129 const uint32_t *revindex_data
;
130 const uint32_t *revindex_map
;
131 size_t revindex_size
;
133 * mtimes_map points at the beginning of the memory mapped region of
134 * this pack's corresponding .mtimes file, and mtimes_size is the size
135 * of that .mtimes file
137 const uint32_t *mtimes_map
;
139 /* something like ".git/objects/pack/xxxxx.pack" */
140 char pack_name
[FLEX_ARRAY
]; /* more */
143 struct multi_pack_index
;
145 static inline int pack_map_entry_cmp(const void *cmp_data UNUSED
,
146 const struct hashmap_entry
*entry
,
147 const struct hashmap_entry
*entry2
,
150 const char *key
= keydata
;
151 const struct packed_git
*pg1
, *pg2
;
153 pg1
= container_of(entry
, const struct packed_git
, packmap_ent
);
154 pg2
= container_of(entry2
, const struct packed_git
, packmap_ent
);
156 return strcmp(pg1
->pack_name
, key
? key
: pg2
->pack_name
);
159 struct raw_object_store
{
161 * Set of all object directories; the main directory is first (and
162 * cannot be NULL after initialization). Subsequent directories are
165 struct object_directory
*odb
;
166 struct object_directory
**odb_tail
;
167 kh_odb_path_map_t
*odb_by_path
;
169 int loaded_alternates
;
172 * A list of alternate object directories loaded from the environment;
173 * this should not generally need to be accessed directly, but will
174 * populate the "odb" list when prepare_alt_odb() is run.
179 * Objects that should be substituted by other objects
180 * (see git-replace(1)).
182 struct oidmap
*replace_map
;
183 unsigned replace_map_initialized
: 1;
184 pthread_mutex_t replace_mutex
; /* protect object replace functions */
186 struct commit_graph
*commit_graph
;
187 unsigned commit_graph_attempted
: 1; /* if loading has been attempted */
192 * should only be accessed directly by packfile.c and midx.c
194 struct multi_pack_index
*multi_pack_index
;
199 * should only be accessed directly by packfile.c
202 struct packed_git
*packed_git
;
203 /* A most-recently-used ordered version of the packed_git list. */
204 struct list_head packed_git_mru
;
207 struct packed_git
**packs
;
212 * A map of packfiles to packed_git structs for tracking which
213 * packs have been loaded already.
215 struct hashmap pack_map
;
218 * A fast, rough count of the number of objects in the repository.
219 * These two fields are not meant for direct access. Use
220 * repo_approximate_object_count() instead.
222 unsigned long approximate_object_count
;
223 unsigned approximate_object_count_valid
: 1;
226 * Whether packed_git has already been populated with this repository's
229 unsigned packed_git_initialized
: 1;
232 struct raw_object_store
*raw_object_store_new(void);
233 void raw_object_store_clear(struct raw_object_store
*o
);
236 * Put in `buf` the name of the file in the local object database that
237 * would be used to store a loose object with the specified oid.
239 const char *loose_object_path(struct repository
*r
, struct strbuf
*buf
,
240 const struct object_id
*oid
);
242 void *map_loose_object(struct repository
*r
, const struct object_id
*oid
,
243 unsigned long *size
);
245 void *repo_read_object_file(struct repository
*r
,
246 const struct object_id
*oid
,
247 enum object_type
*type
,
248 unsigned long *size
);
250 /* Read and unpack an object file into memory, write memory to an object file */
251 int oid_object_info(struct repository
*r
, const struct object_id
*, unsigned long *);
253 void hash_object_file(const struct git_hash_algo
*algo
, const void *buf
,
254 unsigned long len
, enum object_type type
,
255 struct object_id
*oid
);
257 int write_object_file_flags(const void *buf
, unsigned long len
,
258 enum object_type type
, struct object_id
*oid
,
260 static inline int write_object_file(const void *buf
, unsigned long len
,
261 enum object_type type
, struct object_id
*oid
)
263 return write_object_file_flags(buf
, len
, type
, oid
, 0);
266 int write_object_file_literally(const void *buf
, unsigned long len
,
267 const char *type
, struct object_id
*oid
,
269 int stream_loose_object(struct input_stream
*in_stream
, size_t len
,
270 struct object_id
*oid
);
273 * Add an object file to the in-memory object store, without writing it
276 * Callers are responsible for calling write_object_file to record the
277 * object in persistent storage before writing any other new objects
280 int pretend_object_file(void *, unsigned long, enum object_type
,
281 struct object_id
*oid
);
283 int force_object_loose(const struct object_id
*oid
, time_t mtime
);
287 enum object_type
*typep
;
288 unsigned long *sizep
;
290 struct object_id
*delta_base_oid
;
291 struct strbuf
*type_name
;
304 * ... Nothing to expose in this case
307 * ... Nothing to expose in this case
311 struct packed_git
*pack
;
313 unsigned int is_delta
;
319 * Initializer for a "struct object_info" that wants no items. You may
320 * also memset() the memory to all-zeroes.
322 #define OBJECT_INFO_INIT { 0 }
324 /* Invoke lookup_replace_object() on the given hash */
325 #define OBJECT_INFO_LOOKUP_REPLACE 1
326 /* Allow reading from a loose object file of unknown/bogus type */
327 #define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
328 /* Do not retry packed storage after checking packed and loose storage */
329 #define OBJECT_INFO_QUICK 8
331 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
334 #define OBJECT_INFO_SKIP_FETCH_OBJECT 16
336 * This is meant for bulk prefetching of missing blobs in a partial
337 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
339 #define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
341 /* Die if object corruption (not just an object being missing) was detected. */
342 #define OBJECT_INFO_DIE_IF_CORRUPT 32
344 int oid_object_info_extended(struct repository
*r
,
345 const struct object_id
*,
346 struct object_info
*, unsigned flags
);
349 * Open the loose object at path, check its hash, and return the contents,
350 * use the "oi" argument to assert things about the object, or e.g. populate its
351 * type, and size. If the object is a blob, then "contents" may return NULL,
352 * to allow streaming of large blobs.
354 * Returns 0 on success, negative on error (details may be written to stderr).
356 int read_loose_object(const char *path
,
357 const struct object_id
*expected_oid
,
358 struct object_id
*real_oid
,
360 struct object_info
*oi
);
362 /* Retry packed storage after checking packed and loose storage */
363 #define HAS_OBJECT_RECHECK_PACKED 1
366 * Returns 1 if the object exists. This function will not lazily fetch objects
367 * in a partial clone.
369 int has_object(struct repository
*r
, const struct object_id
*oid
,
373 * These macros and functions are deprecated. If checking existence for an
374 * object that is likely to be missing and/or whose absence is relatively
375 * inconsequential (or is consequential but the caller is prepared to handle
376 * it), use has_object(), which has better defaults (no lazy fetch in a partial
377 * clone and no rechecking of packed storage). In the unlikely event that a
378 * caller needs to assert existence of an object that it fully expects to
379 * exist, and wants to trigger a lazy fetch in a partial clone, use
380 * oid_object_info_extended() with a NULL struct object_info.
382 * These functions can be removed once all callers have migrated to
383 * has_object() and/or oid_object_info_extended().
385 int repo_has_object_file(struct repository
*r
, const struct object_id
*oid
);
386 int repo_has_object_file_with_flags(struct repository
*r
,
387 const struct object_id
*oid
, int flags
);
390 * Return true iff an alternate object database has a loose object
391 * with the specified name. This function does not respect replace
394 int has_loose_object_nonlocal(const struct object_id
*);
396 int has_loose_object(const struct object_id
*);
399 * format_object_header() is a thin wrapper around s xsnprintf() that
400 * writes the initial "<type> <obj-len>" part of the loose object
401 * header. It returns the size that snprintf() returns + 1.
403 int format_object_header(char *str
, size_t size
, enum object_type type
,
406 void assert_oid_type(const struct object_id
*oid
, enum object_type expect
);
409 * Enabling the object read lock allows multiple threads to safely call the
410 * following functions in parallel: repo_read_object_file(),
411 * read_object_with_reference(), oid_object_info() and oid_object_info_extended().
413 * obj_read_lock() and obj_read_unlock() may also be used to protect other
414 * section which cannot execute in parallel with object reading. Since the used
415 * lock is a recursive mutex, these sections can even contain calls to object
416 * reading functions. However, beware that in these cases zlib inflation won't
417 * be performed in parallel, losing performance.
419 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
420 * any of its callees end up calling it, this recursive call won't benefit from
421 * parallel inflation.
423 void enable_obj_read_lock(void);
424 void disable_obj_read_lock(void);
426 extern int obj_read_use_lock
;
427 extern pthread_mutex_t obj_read_mutex
;
429 static inline void obj_read_lock(void)
431 if(obj_read_use_lock
)
432 pthread_mutex_lock(&obj_read_mutex
);
435 static inline void obj_read_unlock(void)
437 if(obj_read_use_lock
)
438 pthread_mutex_unlock(&obj_read_mutex
);
442 * Iterate over the files in the loose-object parts of the object
443 * directory "path", triggering the following callbacks:
445 * - loose_object is called for each loose object we find.
447 * - loose_cruft is called for any files that do not appear to be
448 * loose objects. Note that we only look in the loose object
449 * directories "objects/[0-9a-f]{2}/", so we will not report
450 * "objects/foobar" as cruft.
452 * - loose_subdir is called for each top-level hashed subdirectory
453 * of the object directory (e.g., "$OBJDIR/f0"). It is called
454 * after the objects in the directory are processed.
456 * Any callback that is NULL will be ignored. Callbacks returning non-zero
457 * will end the iteration.
459 * In the "buf" variant, "path" is a strbuf which will also be used as a
460 * scratch buffer, but restored to its original contents before
461 * the function returns.
463 typedef int each_loose_object_fn(const struct object_id
*oid
,
466 typedef int each_loose_cruft_fn(const char *basename
,
469 typedef int each_loose_subdir_fn(unsigned int nr
,
472 int for_each_file_in_obj_subdir(unsigned int subdir_nr
,
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(const char *path
,
479 each_loose_object_fn obj_cb
,
480 each_loose_cruft_fn cruft_cb
,
481 each_loose_subdir_fn subdir_cb
,
483 int for_each_loose_file_in_objdir_buf(struct strbuf
*path
,
484 each_loose_object_fn obj_cb
,
485 each_loose_cruft_fn cruft_cb
,
486 each_loose_subdir_fn subdir_cb
,
489 /* Flags for for_each_*_object() below. */
490 enum for_each_object_flags
{
491 /* Iterate only over local objects, not alternates. */
492 FOR_EACH_OBJECT_LOCAL_ONLY
= (1<<0),
494 /* Only iterate over packs obtained from the promisor remote. */
495 FOR_EACH_OBJECT_PROMISOR_ONLY
= (1<<1),
498 * Visit objects within a pack in packfile order rather than .idx order
500 FOR_EACH_OBJECT_PACK_ORDER
= (1<<2),
502 /* Only iterate over packs that are not marked as kept in-core. */
503 FOR_EACH_OBJECT_SKIP_IN_CORE_KEPT_PACKS
= (1<<3),
505 /* Only iterate over packs that do not have .keep files. */
506 FOR_EACH_OBJECT_SKIP_ON_DISK_KEPT_PACKS
= (1<<4),
510 * Iterate over all accessible loose objects without respect to
511 * reachability. By default, this includes both local and alternate objects.
512 * The order in which objects are visited is unspecified.
514 * Any flags specific to packs are ignored.
516 int for_each_loose_object(each_loose_object_fn
, void *,
517 enum for_each_object_flags flags
);
520 * Iterate over all accessible packed objects without respect to reachability.
521 * By default, this includes both local and alternate packs.
523 * Note that some objects may appear twice if they are found in multiple packs.
524 * Each pack is visited in an unspecified order. By default, objects within a
525 * pack are visited in pack-idx order (i.e., sorted by oid).
527 typedef int each_packed_object_fn(const struct object_id
*oid
,
528 struct packed_git
*pack
,
531 int for_each_object_in_pack(struct packed_git
*p
,
532 each_packed_object_fn
, void *data
,
533 enum for_each_object_flags flags
);
534 int for_each_packed_object(each_packed_object_fn
, void *,
535 enum for_each_object_flags flags
);
537 #endif /* OBJECT_STORE_H */