freshlybakedca.ke
Git fork
1#define USE_THE_REPOSITORY_VARIABLE
2#define DISABLE_SIGN_COMPARE_WARNINGS
3
4#include "builtin.h"
5#include "config.h"
6#include "environment.h"
7#include "gettext.h"
8#include "git-zlib.h"
9#include "hex.h"
10#include "object-file.h"
11#include "odb.h"
12#include "object.h"
13#include "delta.h"
14#include "pack.h"
15#include "blob.h"
16#include "replace-object.h"
17#include "strbuf.h"
18#include "progress.h"
19#include "decorate.h"
20#include "fsck.h"
21#include "packfile.h"
22
23static int dry_run, quiet, recover, has_errors, strict;
24static const char unpack_usage[] = "git unpack-objects [-n] [-q] [-r] [--strict]";
25
26/* We always read in 4kB chunks. */
27static unsigned char buffer[4096];
28static unsigned int offset, len;
29static off_t consumed_bytes;
30static off_t max_input_size;
31static struct git_hash_ctx ctx;
32static struct fsck_options fsck_options = FSCK_OPTIONS_STRICT;
33static struct progress *progress;
34
35/*
36 * When running under --strict mode, objects whose reachability are
37 * suspect are kept in core without getting written in the object
38 * store.
39 */
40struct obj_buffer {
41 char *buffer;
42 unsigned long size;
43};
44
45static struct decoration obj_decorate;
46
47static struct obj_buffer *lookup_object_buffer(struct object *base)
48{
49 return lookup_decoration(&obj_decorate, base);
50}
51
52static void add_object_buffer(struct object *object, char *buffer, unsigned long size)
53{
54 struct obj_buffer *obj;
55 CALLOC_ARRAY(obj, 1);
56 obj->buffer = buffer;
57 obj->size = size;
58 if (add_decoration(&obj_decorate, object, obj))
59 die("object %s tried to add buffer twice!", oid_to_hex(&object->oid));
60}
61
62/*
63 * Make sure at least "min" bytes are available in the buffer, and
64 * return the pointer to the buffer.
65 */
66static void *fill(int min)
67{
68 if (min <= len)
69 return buffer + offset;
70 if (min > sizeof(buffer))
71 die("cannot fill %d bytes", min);
72 if (offset) {
73 git_hash_update(&ctx, buffer, offset);
74 memmove(buffer, buffer + offset, len);
75 offset = 0;
76 }
77 do {
78 ssize_t ret = xread(0, buffer + len, sizeof(buffer) - len);
79 if (ret <= 0) {
80 if (!ret)
81 die("early EOF");
82 die_errno("read error on input");
83 }
84 len += ret;
85 } while (len < min);
86 return buffer;
87}
88
89static void use(int bytes)
90{
91 if (bytes > len)
92 die("used more bytes than were available");
93 len -= bytes;
94 offset += bytes;
95
96 /* make sure off_t is sufficiently large not to wrap */
97 if (signed_add_overflows(consumed_bytes, bytes))
98 die("pack too large for current definition of off_t");
99 consumed_bytes += bytes;
100 if (max_input_size && consumed_bytes > max_input_size)
101 die(_("pack exceeds maximum allowed size"));
102 display_throughput(progress, consumed_bytes);
103}
104
105/*
106 * Decompress zstream from the standard input into a newly
107 * allocated buffer of specified size and return the buffer.
108 * The caller is responsible to free the returned buffer.
109 *
110 * But for dry_run mode, "get_data()" is only used to check the
111 * integrity of data, and the returned buffer is not used at all.
112 * Therefore, in dry_run mode, "get_data()" will release the small
113 * allocated buffer which is reused to hold temporary zstream output
114 * and return NULL instead of returning garbage data.
115 */
116static void *get_data(unsigned long size)
117{
118 git_zstream stream;
119 unsigned long bufsize = dry_run && size > 8192 ? 8192 : size;
120 void *buf = xmallocz(bufsize);
121
122 memset(&stream, 0, sizeof(stream));
123
124 stream.next_out = buf;
125 stream.avail_out = bufsize;
126 stream.next_in = fill(1);
127 stream.avail_in = len;
128 git_inflate_init(&stream);
129
130 for (;;) {
131 int ret = git_inflate(&stream, 0);
132 use(len - stream.avail_in);
133 if (stream.total_out == size && ret == Z_STREAM_END)
134 break;
135 if (ret != Z_OK) {
136 error("inflate returned %d", ret);
137 FREE_AND_NULL(buf);
138 if (!recover)
139 exit(1);
140 has_errors = 1;
141 break;
142 }
143 stream.next_in = fill(1);
144 stream.avail_in = len;
145 if (dry_run) {
146 /* reuse the buffer in dry_run mode */
147 stream.next_out = buf;
148 stream.avail_out = bufsize > size - stream.total_out ?
149 size - stream.total_out :
150 bufsize;
151 }
152 }
153 git_inflate_end(&stream);
154 if (dry_run)
155 FREE_AND_NULL(buf);
156 return buf;
157}
158
159struct delta_info {
160 struct object_id base_oid;
161 unsigned nr;
162 off_t base_offset;
163 unsigned long size;
164 void *delta;
165 struct delta_info *next;
166};
167
168static struct delta_info *delta_list;
169
170static void add_delta_to_list(unsigned nr, const struct object_id *base_oid,
171 off_t base_offset,
172 void *delta, unsigned long size)
173{
174 struct delta_info *info = xmalloc(sizeof(*info));
175
176 oidcpy(&info->base_oid, base_oid);
177 info->base_offset = base_offset;
178 info->size = size;
179 info->delta = delta;
180 info->nr = nr;
181 info->next = delta_list;
182 delta_list = info;
183}
184
185struct obj_info {
186 off_t offset;
187 struct object_id oid;
188 struct object *obj;
189};
190
191/* Remember to update object flag allocation in object.h */
192#define FLAG_OPEN (1u<<20)
193#define FLAG_WRITTEN (1u<<21)
194
195static struct obj_info *obj_list;
196static unsigned nr_objects;
197
198/*
199 * Called only from check_object() after it verified this object
200 * is Ok.
201 */
202static void write_cached_object(struct object *obj, struct obj_buffer *obj_buf)
203{
204 struct object_id oid;
205
206 if (odb_write_object(the_repository->objects, obj_buf->buffer, obj_buf->size,
207 obj->type, &oid) < 0)
208 die("failed to write object %s", oid_to_hex(&obj->oid));
209 obj->flags |= FLAG_WRITTEN;
210}
211
212/*
213 * At the very end of the processing, write_rest() scans the objects
214 * that have reachability requirements and calls this function.
215 * Verify its reachability and validity recursively and write it out.
216 */
217static int check_object(struct object *obj, enum object_type type,
218 void *data UNUSED,
219 struct fsck_options *options UNUSED)
220{
221 struct obj_buffer *obj_buf;
222
223 if (!obj)
224 return 1;
225
226 if (obj->flags & FLAG_WRITTEN)
227 return 0;
228
229 if (type != OBJ_ANY && obj->type != type)
230 die("object type mismatch");
231
232 if (!(obj->flags & FLAG_OPEN)) {
233 unsigned long size;
234 int type = odb_read_object_info(the_repository->objects, &obj->oid, &size);
235 if (type != obj->type || type <= 0)
236 die("object of unexpected type");
237 obj->flags |= FLAG_WRITTEN;
238 return 0;
239 }
240
241 obj_buf = lookup_object_buffer(obj);
242 if (!obj_buf)
243 die("Whoops! Cannot find object '%s'", oid_to_hex(&obj->oid));
244 if (fsck_object(obj, obj_buf->buffer, obj_buf->size, &fsck_options))
245 die("fsck error in packed object");
246 fsck_options.walk = check_object;
247 if (fsck_walk(obj, NULL, &fsck_options))
248 die("Error on reachable objects of %s", oid_to_hex(&obj->oid));
249 write_cached_object(obj, obj_buf);
250 return 0;
251}
252
253static void write_rest(void)
254{
255 unsigned i;
256 for (i = 0; i < nr_objects; i++) {
257 if (obj_list[i].obj)
258 check_object(obj_list[i].obj, OBJ_ANY, NULL, NULL);
259 }
260}
261
262static void added_object(unsigned nr, enum object_type type,
263 void *data, unsigned long size);
264
265/*
266 * Write out nr-th object from the list, now we know the contents
267 * of it. Under --strict, this buffers structured objects in-core,
268 * to be checked at the end.
269 */
270static void write_object(unsigned nr, enum object_type type,
271 void *buf, unsigned long size)
272{
273 if (!strict) {
274 if (odb_write_object(the_repository->objects, buf, size, type,
275 &obj_list[nr].oid) < 0)
276 die("failed to write object");
277 added_object(nr, type, buf, size);
278 free(buf);
279 obj_list[nr].obj = NULL;
280 } else if (type == OBJ_BLOB) {
281 struct blob *blob;
282 if (odb_write_object(the_repository->objects, buf, size, type,
283 &obj_list[nr].oid) < 0)
284 die("failed to write object");
285 added_object(nr, type, buf, size);
286 free(buf);
287
288 blob = lookup_blob(the_repository, &obj_list[nr].oid);
289 if (blob)
290 blob->object.flags |= FLAG_WRITTEN;
291 else
292 die("invalid blob object");
293 obj_list[nr].obj = NULL;
294 } else {
295 struct object *obj;
296 int eaten;
297 hash_object_file(the_hash_algo, buf, size, type,
298 &obj_list[nr].oid);
299 added_object(nr, type, buf, size);
300 obj = parse_object_buffer(the_repository, &obj_list[nr].oid,
301 type, size, buf,
302 &eaten);
303 if (!obj)
304 die("invalid %s", type_name(type));
305 add_object_buffer(obj, buf, size);
306 obj->flags |= FLAG_OPEN;
307 obj_list[nr].obj = obj;
308 }
309}
310
311static void resolve_delta(unsigned nr, enum object_type type,
312 void *base, unsigned long base_size,
313 void *delta, unsigned long delta_size)
314{
315 void *result;
316 unsigned long result_size;
317
318 result = patch_delta(base, base_size,
319 delta, delta_size,
320 &result_size);
321 if (!result)
322 die("failed to apply delta");
323 free(delta);
324 write_object(nr, type, result, result_size);
325}
326
327/*
328 * We now know the contents of an object (which is nr-th in the pack);
329 * resolve all the deltified objects that are based on it.
330 */
331static void added_object(unsigned nr, enum object_type type,
332 void *data, unsigned long size)
333{
334 struct delta_info **p = &delta_list;
335 struct delta_info *info;
336
337 while ((info = *p) != NULL) {
338 if (oideq(&info->base_oid, &obj_list[nr].oid) ||
339 info->base_offset == obj_list[nr].offset) {
340 *p = info->next;
341 p = &delta_list;
342 resolve_delta(info->nr, type, data, size,
343 info->delta, info->size);
344 free(info);
345 continue;
346 }
347 p = &info->next;
348 }
349}
350
351static void unpack_non_delta_entry(enum object_type type, unsigned long size,
352 unsigned nr)
353{
354 void *buf = get_data(size);
355
356 if (buf)
357 write_object(nr, type, buf, size);
358}
359
360struct input_zstream_data {
361 git_zstream *zstream;
362 unsigned char buf[8192];
363 int status;
364};
365
366static const void *feed_input_zstream(struct input_stream *in_stream,
367 unsigned long *readlen)
368{
369 struct input_zstream_data *data = in_stream->data;
370 git_zstream *zstream = data->zstream;
371 void *in = fill(1);
372
373 if (in_stream->is_finished) {
374 *readlen = 0;
375 return NULL;
376 }
377
378 zstream->next_out = data->buf;
379 zstream->avail_out = sizeof(data->buf);
380 zstream->next_in = in;
381 zstream->avail_in = len;
382
383 data->status = git_inflate(zstream, 0);
384
385 in_stream->is_finished = data->status != Z_OK;
386 use(len - zstream->avail_in);
387 *readlen = sizeof(data->buf) - zstream->avail_out;
388
389 return data->buf;
390}
391
392static void stream_blob(unsigned long size, unsigned nr)
393{
394 git_zstream zstream = { 0 };
395 struct input_zstream_data data = { 0 };
396 struct input_stream in_stream = {
397 .read = feed_input_zstream,
398 .data = &data,
399 };
400 struct obj_info *info = &obj_list[nr];
401
402 data.zstream = &zstream;
403 git_inflate_init(&zstream);
404
405 if (stream_loose_object(the_repository->objects->sources,
406 &in_stream, size, &info->oid))
407 die(_("failed to write object in stream"));
408
409 if (data.status != Z_STREAM_END)
410 die(_("inflate returned (%d)"), data.status);
411 git_inflate_end(&zstream);
412
413 if (strict) {
414 struct blob *blob = lookup_blob(the_repository, &info->oid);
415
416 if (!blob)
417 die(_("invalid blob object from stream"));
418 blob->object.flags |= FLAG_WRITTEN;
419 }
420 info->obj = NULL;
421}
422
423static int resolve_against_held(unsigned nr, const struct object_id *base,
424 void *delta_data, unsigned long delta_size)
425{
426 struct object *obj;
427 struct obj_buffer *obj_buffer;
428 obj = lookup_object(the_repository, base);
429 if (!obj)
430 return 0;
431 obj_buffer = lookup_object_buffer(obj);
432 if (!obj_buffer)
433 return 0;
434 resolve_delta(nr, obj->type, obj_buffer->buffer,
435 obj_buffer->size, delta_data, delta_size);
436 return 1;
437}
438
439static void unpack_delta_entry(enum object_type type, unsigned long delta_size,
440 unsigned nr)
441{
442 void *delta_data, *base;
443 unsigned long base_size;
444 struct object_id base_oid;
445
446 if (type == OBJ_REF_DELTA) {
447 oidread(&base_oid, fill(the_hash_algo->rawsz), the_repository->hash_algo);
448 use(the_hash_algo->rawsz);
449 delta_data = get_data(delta_size);
450 if (!delta_data)
451 return;
452 if (odb_has_object(the_repository->objects, &base_oid,
453 HAS_OBJECT_RECHECK_PACKED | HAS_OBJECT_FETCH_PROMISOR))
454 ; /* Ok we have this one */
455 else if (resolve_against_held(nr, &base_oid,
456 delta_data, delta_size))
457 return; /* we are done */
458 else {
459 /* cannot resolve yet --- queue it */
460 oidclr(&obj_list[nr].oid, the_repository->hash_algo);
461 add_delta_to_list(nr, &base_oid, 0, delta_data, delta_size);
462 return;
463 }
464 } else {
465 unsigned base_found = 0;
466 unsigned char *pack, c;
467 off_t base_offset;
468 unsigned lo, mid, hi;
469
470 pack = fill(1);
471 c = *pack;
472 use(1);
473 base_offset = c & 127;
474 while (c & 128) {
475 base_offset += 1;
476 if (!base_offset || MSB(base_offset, 7))
477 die("offset value overflow for delta base object");
478 pack = fill(1);
479 c = *pack;
480 use(1);
481 base_offset = (base_offset << 7) + (c & 127);
482 }
483 base_offset = obj_list[nr].offset - base_offset;
484 if (base_offset <= 0 || base_offset >= obj_list[nr].offset)
485 die("offset value out of bound for delta base object");
486
487 delta_data = get_data(delta_size);
488 if (!delta_data)
489 return;
490 lo = 0;
491 hi = nr;
492 while (lo < hi) {
493 mid = lo + (hi - lo) / 2;
494 if (base_offset < obj_list[mid].offset) {
495 hi = mid;
496 } else if (base_offset > obj_list[mid].offset) {
497 lo = mid + 1;
498 } else {
499 oidcpy(&base_oid, &obj_list[mid].oid);
500 base_found = !is_null_oid(&base_oid);
501 break;
502 }
503 }
504 if (!base_found) {
505 /*
506 * The delta base object is itself a delta that
507 * has not been resolved yet.
508 */
509 oidclr(&obj_list[nr].oid, the_repository->hash_algo);
510 add_delta_to_list(nr, null_oid(the_hash_algo), base_offset,
511 delta_data, delta_size);
512 return;
513 }
514 }
515
516 if (resolve_against_held(nr, &base_oid, delta_data, delta_size))
517 return;
518
519 base = odb_read_object(the_repository->objects, &base_oid,
520 &type, &base_size);
521 if (!base) {
522 error("failed to read delta-pack base object %s",
523 oid_to_hex(&base_oid));
524 if (!recover)
525 exit(1);
526 has_errors = 1;
527 return;
528 }
529 resolve_delta(nr, type, base, base_size, delta_data, delta_size);
530 free(base);
531}
532
533static void unpack_one(unsigned nr)
534{
535 unsigned shift;
536 unsigned char *pack;
537 unsigned long size, c;
538 enum object_type type;
539
540 obj_list[nr].offset = consumed_bytes;
541
542 pack = fill(1);
543 c = *pack;
544 use(1);
545 type = (c >> 4) & 7;
546 size = (c & 15);
547 shift = 4;
548 while (c & 0x80) {
549 pack = fill(1);
550 c = *pack;
551 use(1);
552 size += (c & 0x7f) << shift;
553 shift += 7;
554 }
555
556 switch (type) {
557 case OBJ_BLOB:
558 if (!dry_run &&
559 size > repo_settings_get_big_file_threshold(the_repository)) {
560 stream_blob(size, nr);
561 return;
562 }
563 /* fallthrough */
564 case OBJ_COMMIT:
565 case OBJ_TREE:
566 case OBJ_TAG:
567 unpack_non_delta_entry(type, size, nr);
568 return;
569 case OBJ_REF_DELTA:
570 case OBJ_OFS_DELTA:
571 unpack_delta_entry(type, size, nr);
572 return;
573 default:
574 error("bad object type %d", type);
575 has_errors = 1;
576 if (recover)
577 return;
578 exit(1);
579 }
580}
581
582static void unpack_all(void)
583{
584 int i;
585 unsigned char *hdr = fill(sizeof(struct pack_header));
586 struct odb_transaction *transaction;
587
588 if (get_be32(hdr) != PACK_SIGNATURE)
589 die("bad pack file");
590 hdr += 4;
591 if (!pack_version_ok_native(get_be32(hdr)))
592 die("unknown pack file version %"PRIu32,
593 get_be32(hdr));
594 hdr += 4;
595 nr_objects = get_be32(hdr);
596 use(sizeof(struct pack_header));
597
598 if (!quiet)
599 progress = start_progress(the_repository,
600 _("Unpacking objects"), nr_objects);
601 CALLOC_ARRAY(obj_list, nr_objects);
602 transaction = odb_transaction_begin(the_repository->objects);
603 for (i = 0; i < nr_objects; i++) {
604 unpack_one(i);
605 display_progress(progress, i + 1);
606 }
607 odb_transaction_commit(transaction);
608 stop_progress(&progress);
609
610 if (delta_list)
611 die("unresolved deltas left after unpacking");
612}
613
614int cmd_unpack_objects(int argc,
615 const char **argv,
616 const char *prefix UNUSED,
617 struct repository *repo UNUSED)
618{
619 int i;
620 struct object_id oid;
621 struct git_hash_ctx tmp_ctx;
622
623 disable_replace_refs();
624
625 repo_config(the_repository, git_default_config, NULL);
626
627 quiet = !isatty(2);
628
629 show_usage_if_asked(argc, argv, unpack_usage);
630
631 for (i = 1 ; i < argc; i++) {
632 const char *arg = argv[i];
633
634 if (*arg == '-') {
635 if (!strcmp(arg, "-n")) {
636 dry_run = 1;
637 continue;
638 }
639 if (!strcmp(arg, "-q")) {
640 quiet = 1;
641 continue;
642 }
643 if (!strcmp(arg, "-r")) {
644 recover = 1;
645 continue;
646 }
647 if (!strcmp(arg, "--strict")) {
648 strict = 1;
649 continue;
650 }
651 if (skip_prefix(arg, "--strict=", &arg)) {
652 strict = 1;
653 fsck_set_msg_types(&fsck_options, arg);
654 continue;
655 }
656 if (skip_prefix(arg, "--pack_header=", &arg)) {
657 if (parse_pack_header_option(arg,
658 buffer, &len) < 0)
659 die(_("bad --pack_header: %s"), arg);
660 continue;
661 }
662 if (skip_prefix(arg, "--max-input-size=", &arg)) {
663 max_input_size = strtoumax(arg, NULL, 10);
664 continue;
665 }
666 usage(unpack_usage);
667 }
668
669 /* We don't take any non-flag arguments now.. Maybe some day */
670 usage(unpack_usage);
671 }
672 the_hash_algo->init_fn(&ctx);
673 unpack_all();
674 git_hash_update(&ctx, buffer, offset);
675 the_hash_algo->init_fn(&tmp_ctx);
676 git_hash_clone(&tmp_ctx, &ctx);
677 git_hash_final_oid(&oid, &tmp_ctx);
678 if (strict) {
679 write_rest();
680 if (fsck_finish(&fsck_options))
681 die(_("fsck error in pack objects"));
682 }
683 if (!hasheq(fill(the_hash_algo->rawsz), oid.hash,
684 the_repository->hash_algo))
685 die("final sha1 did not match");
686 use(the_hash_algo->rawsz);
687
688 /* Write the last part of the buffer to stdout */
689 write_in_full(1, buffer + offset, len);
690
691 /* All done */
692 return has_errors;
693}