4 #include "object-store.h"
12 #include "tree-walk.h"
17 static int dry_run
, quiet
, recover
, has_errors
, strict
;
18 static const char unpack_usage
[] = "git unpack-objects [-n] [-q] [-r] [--strict]";
20 /* We always read in 4kB chunks. */
21 static unsigned char buffer
[4096];
22 static unsigned int offset
, len
;
23 static off_t consumed_bytes
;
24 static off_t max_input_size
;
25 static git_hash_ctx ctx
;
26 static struct fsck_options fsck_options
= FSCK_OPTIONS_STRICT
;
27 static struct progress
*progress
;
30 * When running under --strict mode, objects whose reachability are
31 * suspect are kept in core without getting written in the object
39 static struct decoration obj_decorate
;
41 static struct obj_buffer
*lookup_object_buffer(struct object
*base
)
43 return lookup_decoration(&obj_decorate
, base
);
46 static void add_object_buffer(struct object
*object
, char *buffer
, unsigned long size
)
48 struct obj_buffer
*obj
;
49 obj
= xcalloc(1, sizeof(struct obj_buffer
));
52 if (add_decoration(&obj_decorate
, object
, obj
))
53 die("object %s tried to add buffer twice!", oid_to_hex(&object
->oid
));
57 * Make sure at least "min" bytes are available in the buffer, and
58 * return the pointer to the buffer.
60 static void *fill(int min
)
63 return buffer
+ offset
;
64 if (min
> sizeof(buffer
))
65 die("cannot fill %d bytes", min
);
67 the_hash_algo
->update_fn(&ctx
, buffer
, offset
);
68 memmove(buffer
, buffer
+ offset
, len
);
72 ssize_t ret
= xread(0, buffer
+ len
, sizeof(buffer
) - len
);
76 die_errno("read error on input");
83 static void use(int bytes
)
86 die("used more bytes than were available");
90 /* make sure off_t is sufficiently large not to wrap */
91 if (signed_add_overflows(consumed_bytes
, bytes
))
92 die("pack too large for current definition of off_t");
93 consumed_bytes
+= bytes
;
94 if (max_input_size
&& consumed_bytes
> max_input_size
)
95 die(_("pack exceeds maximum allowed size"));
96 display_throughput(progress
, consumed_bytes
);
99 static void *get_data(unsigned long size
)
102 void *buf
= xmallocz(size
);
104 memset(&stream
, 0, sizeof(stream
));
106 stream
.next_out
= buf
;
107 stream
.avail_out
= size
;
108 stream
.next_in
= fill(1);
109 stream
.avail_in
= len
;
110 git_inflate_init(&stream
);
113 int ret
= git_inflate(&stream
, 0);
114 use(len
- stream
.avail_in
);
115 if (stream
.total_out
== size
&& ret
== Z_STREAM_END
)
118 error("inflate returned %d", ret
);
125 stream
.next_in
= fill(1);
126 stream
.avail_in
= len
;
128 git_inflate_end(&stream
);
133 struct object_id base_oid
;
138 struct delta_info
*next
;
141 static struct delta_info
*delta_list
;
143 static void add_delta_to_list(unsigned nr
, const struct object_id
*base_oid
,
145 void *delta
, unsigned long size
)
147 struct delta_info
*info
= xmalloc(sizeof(*info
));
149 oidcpy(&info
->base_oid
, base_oid
);
150 info
->base_offset
= base_offset
;
154 info
->next
= delta_list
;
160 struct object_id oid
;
164 /* Remember to update object flag allocation in object.h */
165 #define FLAG_OPEN (1u<<20)
166 #define FLAG_WRITTEN (1u<<21)
168 static struct obj_info
*obj_list
;
169 static unsigned nr_objects
;
172 * Called only from check_object() after it verified this object
175 static void write_cached_object(struct object
*obj
, struct obj_buffer
*obj_buf
)
177 struct object_id oid
;
179 if (write_object_file(obj_buf
->buffer
, obj_buf
->size
,
180 type_name(obj
->type
), &oid
) < 0)
181 die("failed to write object %s", oid_to_hex(&obj
->oid
));
182 obj
->flags
|= FLAG_WRITTEN
;
186 * At the very end of the processing, write_rest() scans the objects
187 * that have reachability requirements and calls this function.
188 * Verify its reachability and validity recursively and write it out.
190 static int check_object(struct object
*obj
, int type
, void *data
, struct fsck_options
*options
)
192 struct obj_buffer
*obj_buf
;
197 if (obj
->flags
& FLAG_WRITTEN
)
200 if (type
!= OBJ_ANY
&& obj
->type
!= type
)
201 die("object type mismatch");
203 if (!(obj
->flags
& FLAG_OPEN
)) {
205 int type
= oid_object_info(the_repository
, &obj
->oid
, &size
);
206 if (type
!= obj
->type
|| type
<= 0)
207 die("object of unexpected type");
208 obj
->flags
|= FLAG_WRITTEN
;
212 obj_buf
= lookup_object_buffer(obj
);
214 die("Whoops! Cannot find object '%s'", oid_to_hex(&obj
->oid
));
215 if (fsck_object(obj
, obj_buf
->buffer
, obj_buf
->size
, &fsck_options
))
216 die("fsck error in packed object");
217 fsck_options
.walk
= check_object
;
218 if (fsck_walk(obj
, NULL
, &fsck_options
))
219 die("Error on reachable objects of %s", oid_to_hex(&obj
->oid
));
220 write_cached_object(obj
, obj_buf
);
224 static void write_rest(void)
227 for (i
= 0; i
< nr_objects
; i
++) {
229 check_object(obj_list
[i
].obj
, OBJ_ANY
, NULL
, NULL
);
233 static void added_object(unsigned nr
, enum object_type type
,
234 void *data
, unsigned long size
);
237 * Write out nr-th object from the list, now we know the contents
238 * of it. Under --strict, this buffers structured objects in-core,
239 * to be checked at the end.
241 static void write_object(unsigned nr
, enum object_type type
,
242 void *buf
, unsigned long size
)
245 if (write_object_file(buf
, size
, type_name(type
),
246 &obj_list
[nr
].oid
) < 0)
247 die("failed to write object");
248 added_object(nr
, type
, buf
, size
);
250 obj_list
[nr
].obj
= NULL
;
251 } else if (type
== OBJ_BLOB
) {
253 if (write_object_file(buf
, size
, type_name(type
),
254 &obj_list
[nr
].oid
) < 0)
255 die("failed to write object");
256 added_object(nr
, type
, buf
, size
);
259 blob
= lookup_blob(the_repository
, &obj_list
[nr
].oid
);
261 blob
->object
.flags
|= FLAG_WRITTEN
;
263 die("invalid blob object");
264 obj_list
[nr
].obj
= NULL
;
268 hash_object_file(the_hash_algo
, buf
, size
, type_name(type
),
270 added_object(nr
, type
, buf
, size
);
271 obj
= parse_object_buffer(the_repository
, &obj_list
[nr
].oid
,
275 die("invalid %s", type_name(type
));
276 add_object_buffer(obj
, buf
, size
);
277 obj
->flags
|= FLAG_OPEN
;
278 obj_list
[nr
].obj
= obj
;
282 static void resolve_delta(unsigned nr
, enum object_type type
,
283 void *base
, unsigned long base_size
,
284 void *delta
, unsigned long delta_size
)
287 unsigned long result_size
;
289 result
= patch_delta(base
, base_size
,
293 die("failed to apply delta");
295 write_object(nr
, type
, result
, result_size
);
299 * We now know the contents of an object (which is nr-th in the pack);
300 * resolve all the deltified objects that are based on it.
302 static void added_object(unsigned nr
, enum object_type type
,
303 void *data
, unsigned long size
)
305 struct delta_info
**p
= &delta_list
;
306 struct delta_info
*info
;
308 while ((info
= *p
) != NULL
) {
309 if (oideq(&info
->base_oid
, &obj_list
[nr
].oid
) ||
310 info
->base_offset
== obj_list
[nr
].offset
) {
313 resolve_delta(info
->nr
, type
, data
, size
,
314 info
->delta
, info
->size
);
322 static void unpack_non_delta_entry(enum object_type type
, unsigned long size
,
325 void *buf
= get_data(size
);
328 write_object(nr
, type
, buf
, size
);
333 static int resolve_against_held(unsigned nr
, const struct object_id
*base
,
334 void *delta_data
, unsigned long delta_size
)
337 struct obj_buffer
*obj_buffer
;
338 obj
= lookup_object(the_repository
, base
);
341 obj_buffer
= lookup_object_buffer(obj
);
344 resolve_delta(nr
, obj
->type
, obj_buffer
->buffer
,
345 obj_buffer
->size
, delta_data
, delta_size
);
349 static void unpack_delta_entry(enum object_type type
, unsigned long delta_size
,
352 void *delta_data
, *base
;
353 unsigned long base_size
;
354 struct object_id base_oid
;
356 if (type
== OBJ_REF_DELTA
) {
357 hashcpy(base_oid
.hash
, fill(the_hash_algo
->rawsz
));
358 use(the_hash_algo
->rawsz
);
359 delta_data
= get_data(delta_size
);
360 if (dry_run
|| !delta_data
) {
364 if (has_object_file(&base_oid
))
365 ; /* Ok we have this one */
366 else if (resolve_against_held(nr
, &base_oid
,
367 delta_data
, delta_size
))
368 return; /* we are done */
370 /* cannot resolve yet --- queue it */
371 oidclr(&obj_list
[nr
].oid
);
372 add_delta_to_list(nr
, &base_oid
, 0, delta_data
, delta_size
);
376 unsigned base_found
= 0;
377 unsigned char *pack
, c
;
379 unsigned lo
, mid
, hi
;
384 base_offset
= c
& 127;
387 if (!base_offset
|| MSB(base_offset
, 7))
388 die("offset value overflow for delta base object");
392 base_offset
= (base_offset
<< 7) + (c
& 127);
394 base_offset
= obj_list
[nr
].offset
- base_offset
;
395 if (base_offset
<= 0 || base_offset
>= obj_list
[nr
].offset
)
396 die("offset value out of bound for delta base object");
398 delta_data
= get_data(delta_size
);
399 if (dry_run
|| !delta_data
) {
406 mid
= lo
+ (hi
- lo
) / 2;
407 if (base_offset
< obj_list
[mid
].offset
) {
409 } else if (base_offset
> obj_list
[mid
].offset
) {
412 oidcpy(&base_oid
, &obj_list
[mid
].oid
);
413 base_found
= !is_null_oid(&base_oid
);
419 * The delta base object is itself a delta that
420 * has not been resolved yet.
422 oidclr(&obj_list
[nr
].oid
);
423 add_delta_to_list(nr
, &null_oid
, base_offset
, delta_data
, delta_size
);
428 if (resolve_against_held(nr
, &base_oid
, delta_data
, delta_size
))
431 base
= read_object_file(&base_oid
, &type
, &base_size
);
433 error("failed to read delta-pack base object %s",
434 oid_to_hex(&base_oid
));
440 resolve_delta(nr
, type
, base
, base_size
, delta_data
, delta_size
);
444 static void unpack_one(unsigned nr
)
448 unsigned long size
, c
;
449 enum object_type type
;
451 obj_list
[nr
].offset
= consumed_bytes
;
463 size
+= (c
& 0x7f) << shift
;
472 unpack_non_delta_entry(type
, size
, nr
);
476 unpack_delta_entry(type
, size
, nr
);
479 error("bad object type %d", type
);
487 static void unpack_all(void)
490 struct pack_header
*hdr
= fill(sizeof(struct pack_header
));
492 nr_objects
= ntohl(hdr
->hdr_entries
);
494 if (ntohl(hdr
->hdr_signature
) != PACK_SIGNATURE
)
495 die("bad pack file");
496 if (!pack_version_ok(hdr
->hdr_version
))
497 die("unknown pack file version %"PRIu32
,
498 ntohl(hdr
->hdr_version
));
499 use(sizeof(struct pack_header
));
502 progress
= start_progress(_("Unpacking objects"), nr_objects
);
503 obj_list
= xcalloc(nr_objects
, sizeof(*obj_list
));
504 for (i
= 0; i
< nr_objects
; i
++) {
506 display_progress(progress
, i
+ 1);
508 stop_progress(&progress
);
511 die("unresolved deltas left after unpacking");
514 int cmd_unpack_objects(int argc
, const char **argv
, const char *prefix
)
517 struct object_id oid
;
519 read_replace_refs
= 0;
521 git_config(git_default_config
, NULL
);
525 for (i
= 1 ; i
< argc
; i
++) {
526 const char *arg
= argv
[i
];
529 if (!strcmp(arg
, "-n")) {
533 if (!strcmp(arg
, "-q")) {
537 if (!strcmp(arg
, "-r")) {
541 if (!strcmp(arg
, "--strict")) {
545 if (skip_prefix(arg
, "--strict=", &arg
)) {
547 fsck_set_msg_types(&fsck_options
, arg
);
550 if (starts_with(arg
, "--pack_header=")) {
551 struct pack_header
*hdr
;
554 hdr
= (struct pack_header
*)buffer
;
555 hdr
->hdr_signature
= htonl(PACK_SIGNATURE
);
556 hdr
->hdr_version
= htonl(strtoul(arg
+ 14, &c
, 10));
559 hdr
->hdr_entries
= htonl(strtoul(c
+ 1, &c
, 10));
565 if (skip_prefix(arg
, "--max-input-size=", &arg
)) {
566 max_input_size
= strtoumax(arg
, NULL
, 10);
572 /* We don't take any non-flag arguments now.. Maybe some day */
575 the_hash_algo
->init_fn(&ctx
);
577 the_hash_algo
->update_fn(&ctx
, buffer
, offset
);
578 the_hash_algo
->final_fn(oid
.hash
, &ctx
);
581 if (fsck_finish(&fsck_options
))
582 die(_("fsck error in pack objects"));
584 if (!hasheq(fill(the_hash_algo
->rawsz
), oid
.hash
))
585 die("final sha1 did not match");
586 use(the_hash_algo
->rawsz
);
588 /* Write the last part of the buffer to stdout */
590 int ret
= xwrite(1, buffer
+ offset
, len
);