]> git.ipfire.org Git - thirdparty/git.git/blob - object-store-ll.h
Always check the return value of `repo_read_object_file()`
[thirdparty/git.git] / object-store-ll.h
1 #ifndef OBJECT_STORE_LL_H
2 #define OBJECT_STORE_LL_H
3
4 #include "hashmap.h"
5 #include "object.h"
6 #include "list.h"
7 #include "thread-utils.h"
8 #include "oidset.h"
9
10 struct oidmap;
11 struct oidtree;
12 struct strbuf;
13
14 struct object_directory {
15 struct object_directory *next;
16
17 /*
18 * Used to store the results of readdir(3) calls when we are OK
19 * sacrificing accuracy due to races for speed. That includes
20 * object existence with OBJECT_INFO_QUICK, as well as
21 * our search for unique abbreviated hashes. Don't use it for tasks
22 * requiring greater accuracy!
23 *
24 * Be sure to call odb_load_loose_cache() before using.
25 */
26 uint32_t loose_objects_subdir_seen[8]; /* 256 bits */
27 struct oidtree *loose_objects_cache;
28
29 /*
30 * This is a temporary object store created by the tmp_objdir
31 * facility. Disable ref updates since the objects in the store
32 * might be discarded on rollback.
33 */
34 int disable_ref_updates;
35
36 /*
37 * This object store is ephemeral, so there is no need to fsync.
38 */
39 int will_destroy;
40
41 /*
42 * Path to the alternative object store. If this is a relative path,
43 * it is relative to the current working directory.
44 */
45 char *path;
46 };
47
48 struct input_stream {
49 const void *(*read)(struct input_stream *, unsigned long *len);
50 void *data;
51 int is_finished;
52 };
53
54 void prepare_alt_odb(struct repository *r);
55 int has_alt_odb(struct repository *r);
56 char *compute_alternate_path(const char *path, struct strbuf *err);
57 struct object_directory *find_odb(struct repository *r, const char *obj_dir);
58 typedef int alt_odb_fn(struct object_directory *, void *);
59 int foreach_alt_odb(alt_odb_fn, void*);
60 typedef void alternate_ref_fn(const struct object_id *oid, void *);
61 void for_each_alternate_ref(alternate_ref_fn, void *);
62
63 /*
64 * Add the directory to the on-disk alternates file; the new entry will also
65 * take effect in the current process.
66 */
67 void add_to_alternates_file(const char *dir);
68
69 /*
70 * Add the directory to the in-memory list of alternates (along with any
71 * recursive alternates it points to), but do not modify the on-disk alternates
72 * file.
73 */
74 void add_to_alternates_memory(const char *dir);
75
76 /*
77 * Replace the current writable object directory with the specified temporary
78 * object directory; returns the former primary object directory.
79 */
80 struct object_directory *set_temporary_primary_odb(const char *dir, int will_destroy);
81
82 /*
83 * Restore a previous ODB replaced by set_temporary_main_odb.
84 */
85 void restore_primary_odb(struct object_directory *restore_odb, const char *old_path);
86
87 /*
88 * Populate and return the loose object cache array corresponding to the
89 * given object ID.
90 */
91 struct oidtree *odb_loose_cache(struct object_directory *odb,
92 const struct object_id *oid);
93
94 /* Empty the loose object cache for the specified object directory. */
95 void odb_clear_loose_cache(struct object_directory *odb);
96
97 /* Clear and free the specified object directory */
98 void free_object_directory(struct object_directory *odb);
99
100 struct packed_git {
101 struct hashmap_entry packmap_ent;
102 struct packed_git *next;
103 struct list_head mru;
104 struct pack_window *windows;
105 off_t pack_size;
106 const void *index_data;
107 size_t index_size;
108 uint32_t num_objects;
109 size_t crc_offset;
110 struct oidset bad_objects;
111 int index_version;
112 time_t mtime;
113 int pack_fd;
114 int index; /* for builtin/pack-objects.c */
115 unsigned pack_local:1,
116 pack_keep:1,
117 pack_keep_in_core:1,
118 freshened:1,
119 do_not_close:1,
120 pack_promisor:1,
121 multi_pack_index:1,
122 is_cruft:1;
123 unsigned char hash[GIT_MAX_RAWSZ];
124 struct revindex_entry *revindex;
125 const uint32_t *revindex_data;
126 const uint32_t *revindex_map;
127 size_t revindex_size;
128 /*
129 * mtimes_map points at the beginning of the memory mapped region of
130 * this pack's corresponding .mtimes file, and mtimes_size is the size
131 * of that .mtimes file
132 */
133 const uint32_t *mtimes_map;
134 size_t mtimes_size;
135 /* something like ".git/objects/pack/xxxxx.pack" */
136 char pack_name[FLEX_ARRAY]; /* more */
137 };
138
139 struct multi_pack_index;
140
141 static inline int pack_map_entry_cmp(const void *cmp_data UNUSED,
142 const struct hashmap_entry *entry,
143 const struct hashmap_entry *entry2,
144 const void *keydata)
145 {
146 const char *key = keydata;
147 const struct packed_git *pg1, *pg2;
148
149 pg1 = container_of(entry, const struct packed_git, packmap_ent);
150 pg2 = container_of(entry2, const struct packed_git, packmap_ent);
151
152 return strcmp(pg1->pack_name, key ? key : pg2->pack_name);
153 }
154
155 struct raw_object_store {
156 /*
157 * Set of all object directories; the main directory is first (and
158 * cannot be NULL after initialization). Subsequent directories are
159 * alternates.
160 */
161 struct object_directory *odb;
162 struct object_directory **odb_tail;
163 struct kh_odb_path_map *odb_by_path;
164
165 int loaded_alternates;
166
167 /*
168 * A list of alternate object directories loaded from the environment;
169 * this should not generally need to be accessed directly, but will
170 * populate the "odb" list when prepare_alt_odb() is run.
171 */
172 char *alternate_db;
173
174 /*
175 * Objects that should be substituted by other objects
176 * (see git-replace(1)).
177 */
178 struct oidmap *replace_map;
179 unsigned replace_map_initialized : 1;
180 pthread_mutex_t replace_mutex; /* protect object replace functions */
181
182 struct commit_graph *commit_graph;
183 unsigned commit_graph_attempted : 1; /* if loading has been attempted */
184
185 /*
186 * private data
187 *
188 * should only be accessed directly by packfile.c and midx.c
189 */
190 struct multi_pack_index *multi_pack_index;
191
192 /*
193 * private data
194 *
195 * should only be accessed directly by packfile.c
196 */
197
198 struct packed_git *packed_git;
199 /* A most-recently-used ordered version of the packed_git list. */
200 struct list_head packed_git_mru;
201
202 struct {
203 struct packed_git **packs;
204 unsigned flags;
205 } kept_pack_cache;
206
207 /*
208 * A map of packfiles to packed_git structs for tracking which
209 * packs have been loaded already.
210 */
211 struct hashmap pack_map;
212
213 /*
214 * A fast, rough count of the number of objects in the repository.
215 * These two fields are not meant for direct access. Use
216 * repo_approximate_object_count() instead.
217 */
218 unsigned long approximate_object_count;
219 unsigned approximate_object_count_valid : 1;
220
221 /*
222 * Whether packed_git has already been populated with this repository's
223 * packs.
224 */
225 unsigned packed_git_initialized : 1;
226 };
227
228 struct raw_object_store *raw_object_store_new(void);
229 void raw_object_store_clear(struct raw_object_store *o);
230
231 /*
232 * Put in `buf` the name of the file in the local object database that
233 * would be used to store a loose object with the specified oid.
234 */
235 const char *loose_object_path(struct repository *r, struct strbuf *buf,
236 const struct object_id *oid);
237
238 void *map_loose_object(struct repository *r, const struct object_id *oid,
239 unsigned long *size);
240
241 void *repo_read_object_file(struct repository *r,
242 const struct object_id *oid,
243 enum object_type *type,
244 unsigned long *size);
245
246 /* Read and unpack an object file into memory, write memory to an object file */
247 int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
248
249 void hash_object_file(const struct git_hash_algo *algo, const void *buf,
250 unsigned long len, enum object_type type,
251 struct object_id *oid);
252
253 int write_object_file_flags(const void *buf, unsigned long len,
254 enum object_type type, struct object_id *oid,
255 unsigned flags);
256 static inline int write_object_file(const void *buf, unsigned long len,
257 enum object_type type, struct object_id *oid)
258 {
259 return write_object_file_flags(buf, len, type, oid, 0);
260 }
261
262 int write_object_file_literally(const void *buf, unsigned long len,
263 const char *type, struct object_id *oid,
264 unsigned flags);
265 int stream_loose_object(struct input_stream *in_stream, size_t len,
266 struct object_id *oid);
267
268 /*
269 * Add an object file to the in-memory object store, without writing it
270 * to disk.
271 *
272 * Callers are responsible for calling write_object_file to record the
273 * object in persistent storage before writing any other new objects
274 * that reference it.
275 */
276 int pretend_object_file(void *, unsigned long, enum object_type,
277 struct object_id *oid);
278
279 int force_object_loose(const struct object_id *oid, time_t mtime);
280
281 struct object_info {
282 /* Request */
283 enum object_type *typep;
284 unsigned long *sizep;
285 off_t *disk_sizep;
286 struct object_id *delta_base_oid;
287 struct strbuf *type_name;
288 void **contentp;
289
290 /* Response */
291 enum {
292 OI_CACHED,
293 OI_LOOSE,
294 OI_PACKED,
295 OI_DBCACHED
296 } whence;
297 union {
298 /*
299 * struct {
300 * ... Nothing to expose in this case
301 * } cached;
302 * struct {
303 * ... Nothing to expose in this case
304 * } loose;
305 */
306 struct {
307 struct packed_git *pack;
308 off_t offset;
309 unsigned int is_delta;
310 } packed;
311 } u;
312 };
313
314 /*
315 * Initializer for a "struct object_info" that wants no items. You may
316 * also memset() the memory to all-zeroes.
317 */
318 #define OBJECT_INFO_INIT { 0 }
319
320 /* Invoke lookup_replace_object() on the given hash */
321 #define OBJECT_INFO_LOOKUP_REPLACE 1
322 /* Allow reading from a loose object file of unknown/bogus type */
323 #define OBJECT_INFO_ALLOW_UNKNOWN_TYPE 2
324 /* Do not retry packed storage after checking packed and loose storage */
325 #define OBJECT_INFO_QUICK 8
326 /*
327 * Do not attempt to fetch the object if missing (even if fetch_is_missing is
328 * nonzero).
329 */
330 #define OBJECT_INFO_SKIP_FETCH_OBJECT 16
331 /*
332 * This is meant for bulk prefetching of missing blobs in a partial
333 * clone. Implies OBJECT_INFO_SKIP_FETCH_OBJECT and OBJECT_INFO_QUICK
334 */
335 #define OBJECT_INFO_FOR_PREFETCH (OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_QUICK)
336
337 /* Die if object corruption (not just an object being missing) was detected. */
338 #define OBJECT_INFO_DIE_IF_CORRUPT 32
339
340 int oid_object_info_extended(struct repository *r,
341 const struct object_id *,
342 struct object_info *, unsigned flags);
343
344 /*
345 * Open the loose object at path, check its hash, and return the contents,
346 * use the "oi" argument to assert things about the object, or e.g. populate its
347 * type, and size. If the object is a blob, then "contents" may return NULL,
348 * to allow streaming of large blobs.
349 *
350 * Returns 0 on success, negative on error (details may be written to stderr).
351 */
352 int read_loose_object(const char *path,
353 const struct object_id *expected_oid,
354 struct object_id *real_oid,
355 void **contents,
356 struct object_info *oi);
357
358 /* Retry packed storage after checking packed and loose storage */
359 #define HAS_OBJECT_RECHECK_PACKED 1
360
361 /*
362 * Returns 1 if the object exists. This function will not lazily fetch objects
363 * in a partial clone.
364 */
365 int has_object(struct repository *r, const struct object_id *oid,
366 unsigned flags);
367
368 /*
369 * These macros and functions are deprecated. If checking existence for an
370 * object that is likely to be missing and/or whose absence is relatively
371 * inconsequential (or is consequential but the caller is prepared to handle
372 * it), use has_object(), which has better defaults (no lazy fetch in a partial
373 * clone and no rechecking of packed storage). In the unlikely event that a
374 * caller needs to assert existence of an object that it fully expects to
375 * exist, and wants to trigger a lazy fetch in a partial clone, use
376 * oid_object_info_extended() with a NULL struct object_info.
377 *
378 * These functions can be removed once all callers have migrated to
379 * has_object() and/or oid_object_info_extended().
380 */
381 int repo_has_object_file(struct repository *r, const struct object_id *oid);
382 int repo_has_object_file_with_flags(struct repository *r,
383 const struct object_id *oid, int flags);
384
385 /*
386 * Return true iff an alternate object database has a loose object
387 * with the specified name. This function does not respect replace
388 * references.
389 */
390 int has_loose_object_nonlocal(const struct object_id *);
391
392 int has_loose_object(const struct object_id *);
393
394 /**
395 * format_object_header() is a thin wrapper around s xsnprintf() that
396 * writes the initial "<type> <obj-len>" part of the loose object
397 * header. It returns the size that snprintf() returns + 1.
398 */
399 int format_object_header(char *str, size_t size, enum object_type type,
400 size_t objsize);
401
402 void assert_oid_type(const struct object_id *oid, enum object_type expect);
403
404 /*
405 * Enabling the object read lock allows multiple threads to safely call the
406 * following functions in parallel: repo_read_object_file(),
407 * read_object_with_reference(), oid_object_info() and oid_object_info_extended().
408 *
409 * obj_read_lock() and obj_read_unlock() may also be used to protect other
410 * section which cannot execute in parallel with object reading. Since the used
411 * lock is a recursive mutex, these sections can even contain calls to object
412 * reading functions. However, beware that in these cases zlib inflation won't
413 * be performed in parallel, losing performance.
414 *
415 * TODO: oid_object_info_extended()'s call stack has a recursive behavior. If
416 * any of its callees end up calling it, this recursive call won't benefit from
417 * parallel inflation.
418 */
419 void enable_obj_read_lock(void);
420 void disable_obj_read_lock(void);
421
422 extern int obj_read_use_lock;
423 extern pthread_mutex_t obj_read_mutex;
424
425 static inline void obj_read_lock(void)
426 {
427 if(obj_read_use_lock)
428 pthread_mutex_lock(&obj_read_mutex);
429 }
430
431 static inline void obj_read_unlock(void)
432 {
433 if(obj_read_use_lock)
434 pthread_mutex_unlock(&obj_read_mutex);
435 }
436
437 /*
438 * Iterate over the files in the loose-object parts of the object
439 * directory "path", triggering the following callbacks:
440 *
441 * - loose_object is called for each loose object we find.
442 *
443 * - loose_cruft is called for any files that do not appear to be
444 * loose objects. Note that we only look in the loose object
445 * directories "objects/[0-9a-f]{2}/", so we will not report
446 * "objects/foobar" as cruft.
447 *
448 * - loose_subdir is called for each top-level hashed subdirectory
449 * of the object directory (e.g., "$OBJDIR/f0"). It is called
450 * after the objects in the directory are processed.
451 *
452 * Any callback that is NULL will be ignored. Callbacks returning non-zero
453 * will end the iteration.
454 *
455 * In the "buf" variant, "path" is a strbuf which will also be used as a
456 * scratch buffer, but restored to its original contents before
457 * the function returns.
458 */
459 typedef int each_loose_object_fn(const struct object_id *oid,
460 const char *path,
461 void *data);
462 typedef int each_loose_cruft_fn(const char *basename,
463 const char *path,
464 void *data);
465 typedef int each_loose_subdir_fn(unsigned int nr,
466 const char *path,
467 void *data);
468 int for_each_file_in_obj_subdir(unsigned int subdir_nr,
469 struct strbuf *path,
470 each_loose_object_fn obj_cb,
471 each_loose_cruft_fn cruft_cb,
472 each_loose_subdir_fn subdir_cb,
473 void *data);
474 int for_each_loose_file_in_objdir(const char *path,
475 each_loose_object_fn obj_cb,
476 each_loose_cruft_fn cruft_cb,
477 each_loose_subdir_fn subdir_cb,
478 void *data);
479 int for_each_loose_file_in_objdir_buf(struct strbuf *path,
480 each_loose_object_fn obj_cb,
481 each_loose_cruft_fn cruft_cb,
482 each_loose_subdir_fn subdir_cb,
483 void *data);
484
485 /* Flags for for_each_*_object() below. */
486 enum for_each_object_flags {
487 /* Iterate only over local objects, not alternates. */
488 FOR_EACH_OBJECT_LOCAL_ONLY = (1<<0),
489
490 /* Only iterate over packs obtained from the promisor remote. */
491 FOR_EACH_OBJECT_PROMISOR_ONLY = (1<<1),
492
493 /*
494 * Visit objects within a pack in packfile order rather than .idx order
495 */
496 FOR_EACH_OBJECT_PACK_ORDER = (1<<2),
497
498 /* Only iterate over packs that are not marked as kept in-core. */
499 FOR_EACH_OBJECT_SKIP_IN_CORE_KEPT_PACKS = (1<<3),
500
501 /* Only iterate over packs that do not have .keep files. */
502 FOR_EACH_OBJECT_SKIP_ON_DISK_KEPT_PACKS = (1<<4),
503 };
504
505 /*
506 * Iterate over all accessible loose objects without respect to
507 * reachability. By default, this includes both local and alternate objects.
508 * The order in which objects are visited is unspecified.
509 *
510 * Any flags specific to packs are ignored.
511 */
512 int for_each_loose_object(each_loose_object_fn, void *,
513 enum for_each_object_flags flags);
514
515 /*
516 * Iterate over all accessible packed objects without respect to reachability.
517 * By default, this includes both local and alternate packs.
518 *
519 * Note that some objects may appear twice if they are found in multiple packs.
520 * Each pack is visited in an unspecified order. By default, objects within a
521 * pack are visited in pack-idx order (i.e., sorted by oid).
522 */
523 typedef int each_packed_object_fn(const struct object_id *oid,
524 struct packed_git *pack,
525 uint32_t pos,
526 void *data);
527 int for_each_object_in_pack(struct packed_git *p,
528 each_packed_object_fn, void *data,
529 enum for_each_object_flags flags);
530 int for_each_packed_object(each_packed_object_fn, void *,
531 enum for_each_object_flags flags);
532
533 #endif /* OBJECT_STORE_LL_H */