9 #include "thread-utils.h"
11 struct object_directory {
12 struct object_directory *next;
15 * Used to store the results of readdir(3) calls when we are OK
16 * sacrificing accuracy due to races for speed. That includes
17 * object existence with OBJECT_INFO_QUICK, as well as
18 * our search for unique abbreviated hashes. Don't use it for tasks
19 * requiring greater accuracy!
21 * Be sure to call odb_load_loose_cache() before using.
23 char loose_objects_subdir_seen[256];
24 struct oid_array loose_objects_cache[256];
27 * Path to the alternative object store. If this is a relative path,
28 * it is relative to the current working directory.
33 void prepare_alt_odb(struct repository *r);
34 char *compute_alternate_path(const char *path, struct strbuf *err);
35 typedef int alt_odb_fn(struct object_directory *, void *);
36 int foreach_alt_odb(alt_odb_fn, void*);
37 typedef void alternate_ref_fn(const struct object_id *oid, void *);
38 void for_each_alternate_ref(alternate_ref_fn, void *);
41 * Add the directory to the on-disk alternates file; the new entry will also
42 * take effect in the current process.
44 void add_to_alternates_file(const char *dir);
47 * Add the directory to the in-memory list of alternates (along with any
48 * recursive alternates it points to), but do not modify the on-disk alternates
51 void add_to_alternates_memory(const char *dir);
54 * Populate and return the loose object cache array corresponding to the
57 struct oid_array *odb_loose_cache(struct object_directory *odb,
58 const struct object_id *oid);
60 /* Empty the loose object cache for the specified object directory. */
61 void odb_clear_loose_cache(struct object_directory *odb);
64 struct hashmap_entry packmap_ent;
65 struct packed_git *next;
67 struct pack_window *windows;
69 const void *index_data;
72 uint32_t num_bad_objects;
74 unsigned char *bad_object_sha1;
78 int index; /* for builtin/pack-objects.c */
79 unsigned pack_local:1,
86 unsigned char hash[GIT_MAX_RAWSZ];
87 struct revindex_entry *revindex;
88 const uint32_t *revindex_data;
89 const uint32_t *revindex_map;
91 /* something like ".git/objects/pack/xxxxx.pack" */
92 char pack_name[FLEX_ARRAY]; /* more */
95 struct multi_pack_index;
97 static inline int pack_map_entry_cmp(const void *unused_cmp_data,
98 const struct hashmap_entry *entry,
99 const struct hashmap_entry *entry2,
102 const char *key = keydata;
103 const struct packed_git *pg1, *pg2;
105 pg1 = container_of(entry, const struct packed_git, packmap_ent);
106 pg2 = container_of(entry2, const struct packed_git, packmap_ent);
108 return strcmp(pg1->pack_name, key ? key : pg2->pack_name);
111 struct raw_object_store {
113 * Set of all object directories; the main directory is first (and
114 * cannot be NULL after initialization). Subsequent directories are
117 struct object_directory *odb;
118 struct object_directory **odb_tail;
119 int loaded_alternates;
122 * A list of alternate object directories loaded from the environment;
123 * this should not generally need to be accessed directly, but will
124 * populate the "odb" list when prepare_alt_odb() is run.
129 * Objects that should be substituted by other objects
130 * (see git-replace(1)).
132 struct oidmap *replace_map;
133 unsigned replace_map_initialized : 1;
134 pthread_mutex_t replace_mutex; /* protect object replace functions */
136 struct commit_graph *commit_graph;
137 unsigned commit_graph_attempted : 1; /* if loading has been attempted */
142 * should only be accessed directly by packfile.c and midx.c
144 struct multi_pack_index *multi_pack_index;
149 * should only be accessed directly by packfile.c
152 struct packed_git *packed_git;
153 /* A most-recently-used ordered version of the packed_git list. */
154 struct list_head packed_git_mru;
157 * A map of packfiles to packed_git structs for tracking which
158 * packs have been loaded already.
160 struct hashmap pack_map;
163 * A fast, rough count of the number of objects in the repository.
164 * These two fields are not meant for direct access. Use
165 * approximate_object_count() instead.
167 unsigned long approximate_object_count;
168 unsigned approximate_object_count_valid : 1;
171 * Whether packed_git has already been populated with this repository's
174 unsigned packed_git_initialized : 1;
177 struct raw_object_store *raw_object_store_new(void);
178 void raw_object_store_clear(struct raw_object_store *o);
181 * Put in `buf` the name of the file in the local object database that
182 * would be used to store a loose object with the specified oid.
184 const char *loose_object_path(struct repository *r, struct strbuf *buf,
185 const struct object_id *oid);
187 void *map_loose_object(struct repository *r, const struct object_id *oid,
188 unsigned long *size);
190 void *read_object_file_extended(struct repository *r,
191 const struct object_id *oid,
192 enum object_type *type,
193 unsigned long *size, int lookup_replace);
194 static inline void *repo_read_object_file(struct repository *r,
195 const struct object_id *oid,
196 enum object_type *type,
199 return read_object_file_extended(r, oid, type, size, 1);
201 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
202 #define read_object_file(oid, type, size) repo_read_object_file(the_repository, oid, type, size)
205 /* Read and unpack an object file into memory, write memory to an object file */
206 int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
208 int hash_object_file(const struct git_hash_algo *algo, const void *buf,
209 unsigned long len, const char *type,
210 struct object_id *oid);
212 int write_object_file(const void *buf, unsigned long len,
213 const char *type, struct object_id *oid);
215 int hash_object_file_literally(const void *buf, unsigned long len,
216 const char *type, struct object_id *oid,
220 * Add an object file to the in-memory object store, without writing it
223 * Callers are responsible for calling write_object_file to record the
224 * object in persistent storage before writing any other new objects
227 int pretend_object_file(void *, unsigned long, enum object_type,
228 struct object_id *oid);
230 int force_object_loose(const struct object_id *oid, time_t mtime);
233 * Open the loose object at path, check its hash, and return the contents,
234 * type, and size. If the object is a blob, then "contents" may return NULL,
235 * to allow streaming of large blobs.
237 * Returns 0 on success, negative on error (details may be written to stderr).
239 int read_loose_object(const char *path,
240 const struct object_id *expected_oid,
241 enum object_type *type,
245 /* Retry packed storage after checking packed and loose storage */
246 #define HAS_OBJECT_RECHECK_PACKED 1
249 * Returns 1 if the object exists. This function will not lazily fetch objects
250 * in a partial clone.
252 int has_object(struct repository *r, const struct object_id *oid,
256 * These macros and functions are deprecated. If checking existence for an
257 * object that is likely to be missing and/or whose absence is relatively
258 * inconsequential (or is consequential but the caller is prepared to handle
259 * it), use has_object(), which has better defaults (no lazy fetch in a partial
260 * clone and no rechecking of packed storage). In the unlikely event that a
261 * caller needs to assert existence of an object that it fully expects to
262 * exist, and wants to trigger a lazy fetch in a partial clone, use
263 * oid_object_info_extended() with a NULL struct object_info.
265 * These functions can be removed once all callers have migrated to
266 * has_object() and/or oid_object_info_extended().
268 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
269 #define has_sha1_file_with_flags(sha1, flags) repo_has_sha1_file_with_flags(the_repository, sha1, flags)
270 #define has_sha1_file(sha1) repo_has_sha1_file(the_repository, sha1)
272 int repo_has_object_file(struct repository *r, const struct object_id *oid);
273 int repo_has_object_file_with_flags(struct repository *r,
274 const struct object_id *oid, int flags);
275 #ifndef NO_THE_REPOSITORY_COMPATIBILITY_MACROS
276 #define has_object_file(oid) repo_has_object_file(the_repository, oid)
277 #define has_object_file_with_flags(oid, flags) repo_has_object_file_with_flags(the_repository, oid, flags)
281 * Return true iff an alternate object database has a loose object
282 * with the specified name. This function does not respect replace
285 int has_loose_object_nonlocal(const struct object_id *);
287 void assert_oid_type(const struct object_id *oid, enum object_type expect);
290 * Enabling the object read lock allows multiple threads to safely call the
291 * following functions in parallel: repo_read_object_file(), read_object_file(),
292 * read_object_file_extended(), read_object_with_reference(), read_object(),
293 * oid_object_info() and oid_object_info_extended().
295 * obj_read_lock() and obj_read_unlock() may also be used to protect other
296 * section which cannot execute in parallel with object reading. Since the used
297 * lock is a recursive mutex, these sections can even contain calls to object
298 * reading functions. However, beware that in these cases zlib inflation won't
299 * be performed in parallel, losing performance.
301 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
302 * any of its callees end up calling it, this recursive call won't benefit from
303 * parallel inflation.
305 void enable_obj_read_lock(void);
306 void disable_obj_read_lock(void);
308 extern int obj_read_use_lock;
309 extern pthread_mutex_t obj_read_mutex;
311 static inline void obj_read_lock(void)
313 if(obj_read_use_lock)
314 pthread_mutex_lock(&obj_read_mutex);
317 static inline void obj_read_unlock(void)
319 if(obj_read_use_lock)
320 pthread_mutex_unlock(&obj_read_mutex);
325 enum object_type *typep;
326 unsigned long *sizep;
328 struct object_id *delta_base_oid;
329 struct strbuf *type_name;
342 * ... Nothing to expose in this case
345 * ... Nothing to expose in this case
349 struct packed_git *pack;
351 unsigned int is_delta;
357 * Initializer for a "struct object_info" that wants no items. You may
358 * also memset() the memory to all-zeroes.
360 #define OBJECT_INFO_INIT {NULL}
362 /* Invoke lookup_replace_object() on the given hash */
363 #define OBJECT_INFO_LOOKUP_REPLACE 1
364 /* Allow reading from a loose object file of unknown/bogus type */
365 #define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
366 /* Do not retry packed storage after checking packed and loose storage */
367 #define OBJECT_INFO_QUICK 8
368 /* Do not check loose object */
369 #define OBJECT_INFO_IGNORE_LOOSE 16
371 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
374 #define OBJECT_INFO_SKIP_FETCH_OBJECT 32
376 * This is meant for bulk prefetching of missing blobs in a partial
377 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
379 #define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
381 int oid_object_info_extended(struct repository *r,
382 const struct object_id *,
383 struct object_info *, unsigned flags);
386 * Iterate over the files in the loose-object parts of the object
387 * directory "path", triggering the following callbacks:
389 * - loose_object is called for each loose object we find.
391 * - loose_cruft is called for any files that do not appear to be
392 * loose objects. Note that we only look in the loose object
393 * directories "objects/[0-9a-f]{2}/", so we will not report
394 * "objects/foobar" as cruft.
396 * - loose_subdir is called for each top-level hashed subdirectory
397 * of the object directory (e.g., "$OBJDIR/f0"). It is called
398 * after the objects in the directory are processed.
400 * Any callback that is NULL will be ignored. Callbacks returning non-zero
401 * will end the iteration.
403 * In the "buf" variant, "path" is a strbuf which will also be used as a
404 * scratch buffer, but restored to its original contents before
405 * the function returns.
407 typedef int each_loose_object_fn(const struct object_id *oid,
410 typedef int each_loose_cruft_fn(const char *basename,
413 typedef int each_loose_subdir_fn(unsigned int nr,
416 int for_each_file_in_obj_subdir(unsigned int subdir_nr,
418 each_loose_object_fn obj_cb,
419 each_loose_cruft_fn cruft_cb,
420 each_loose_subdir_fn subdir_cb,
422 int for_each_loose_file_in_objdir(const char *path,
423 each_loose_object_fn obj_cb,
424 each_loose_cruft_fn cruft_cb,
425 each_loose_subdir_fn subdir_cb,
427 int for_each_loose_file_in_objdir_buf(struct strbuf *path,
428 each_loose_object_fn obj_cb,
429 each_loose_cruft_fn cruft_cb,
430 each_loose_subdir_fn subdir_cb,
433 /* Flags for for_each_*_object() below. */
434 enum for_each_object_flags {
435 /* Iterate only over local objects, not alternates. */
436 FOR_EACH_OBJECT_LOCAL_ONLY = (1<<0),
438 /* Only iterate over packs obtained from the promisor remote. */
439 FOR_EACH_OBJECT_PROMISOR_ONLY = (1<<1),
442 * Visit objects within a pack in packfile order rather than .idx order
444 FOR_EACH_OBJECT_PACK_ORDER = (1<<2),
448 * Iterate over all accessible loose objects without respect to
449 * reachability. By default, this includes both local and alternate objects.
450 * The order in which objects are visited is unspecified.
452 * Any flags specific to packs are ignored.
454 int for_each_loose_object(each_loose_object_fn, void *,
455 enum for_each_object_flags flags);
458 * Iterate over all accessible packed objects without respect to reachability.
459 * By default, this includes both local and alternate packs.
461 * Note that some objects may appear twice if they are found in multiple packs.
462 * Each pack is visited in an unspecified order. By default, objects within a
463 * pack are visited in pack-idx order (i.e., sorted by oid).
465 typedef int each_packed_object_fn(const struct object_id *oid,
466 struct packed_git *pack,
469 int for_each_object_in_pack(struct packed_git *p,
470 each_packed_object_fn, void *data,
471 enum for_each_object_flags flags);
472 int for_each_packed_object(each_packed_object_fn, void *,
473 enum for_each_object_flags flags);
475 #endif /* OBJECT_STORE_H */