midx: write object offsets
The final pair of chunks for the multi-pack-index file stores the object offsets. We default to using 32-bit offsets as in the pack-index version 1 format, but if there exists an offset larger than 32-bits, we use a trick similar to the pack-index version 2 format by storing all offsets at least 2^31 in a 64-bit table; we use the 32-bit table to point into that 64-bit table as necessary. We only store these 64-bit offsets if necessary, so create a test that manipulates a version 2 pack-index to fake a large offset. This allows us to test that the large offset table is created, but the data does not match the actual packfile offsets. The multi-pack-index offset does match the (corrupted) pack-index offset, so a future feature will compare these offsets during a 'verify' step. Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Derrick Stolee committed
Jul 12, 2018 at 15:39 UTC
662148c4356620f6567b0a689602079339ebadb7
5 files changed
+155
-15
Documentation/technical/pack-format.txt
+14
-1
@@ -311,7 +311,20 @@ CHUNK DATA:
311
The OIDs for all objects in the MIDX are stored in lexicographic
312
order in this chunk.
313
314
- (This section intentionally left incomplete.)
314
+ Object Offsets (ID: {'O', 'O', 'F', 'F'})
315
+ Stores two 4-byte values for every object.
316
+ 1: The pack-int-id for the pack storing this object.
317
+ 2: The offset within the pack.
318
+ If all offsets are less than 2^31, then the large offset chunk
319
+ will not exist and offsets are stored as in IDX v1.
320
+ If there is at least one offset value larger than 2^32-1, then
321
+ the large offset chunk must exist. If the large offset chunk
322
+ exists and the 31st bit is on, then removing that bit reveals
323
+ the row in the large offsets containing the 8-byte offset of
324
+ this object.
325
+
326
+ [Optional] Object Large Offsets (ID: {'L', 'O', 'F', 'F'})
327
+ 8-byte offsets into large packfiles.
328
329
TRAILER:
330
midx.c
+96
-4
@@ -18,13 +18,18 @@
18
#define MIDX_HASH_LEN 20
19
#define MIDX_MIN_SIZE (MIDX_HEADER_SIZE + MIDX_HASH_LEN)
20
21
-#define MIDX_MAX_CHUNKS 3
21
+#define MIDX_MAX_CHUNKS 5
22
#define MIDX_CHUNK_ALIGNMENT 4
23
#define MIDX_CHUNKID_PACKNAMES 0x504e414d /* "PNAM" */
24
#define MIDX_CHUNKID_OIDFANOUT 0x4f494446 /* "OIDF" */
25
#define MIDX_CHUNKID_OIDLOOKUP 0x4f49444c /* "OIDL" */
26
+#define MIDX_CHUNKID_OBJECTOFFSETS 0x4f4f4646 /* "OOFF" */
27
+#define MIDX_CHUNKID_LARGEOFFSETS 0x4c4f4646 /* "LOFF" */
28
#define MIDX_CHUNKLOOKUP_WIDTH (sizeof(uint32_t) + sizeof(uint64_t))
29
#define MIDX_CHUNK_FANOUT_SIZE (sizeof(uint32_t) * 256)
30
+#define MIDX_CHUNK_OFFSET_WIDTH (2 * sizeof(uint32_t))
31
+#define MIDX_CHUNK_LARGE_OFFSET_WIDTH (sizeof(uint64_t))
32
+#define MIDX_LARGE_OFFSET_NEEDED 0x80000000
33
34
static char *get_midx_filename(const char *object_dir)
35
{
@@ -112,6 +117,14 @@ struct multi_pack_index *load_multi_pack_index(const char *object_dir)
117
m->chunk_oid_lookup = m->data + chunk_offset;
118
break;
119
120
+ case MIDX_CHUNKID_OBJECTOFFSETS:
121
+ m->chunk_object_offsets = m->data + chunk_offset;
122
+ break;
123
+
124
+ case MIDX_CHUNKID_LARGEOFFSETS:
125
+ m->chunk_large_offsets = m->data + chunk_offset;
126
+ break;
127
+
128
case 0:
129
die(_("terminating multi-pack-index chunk id appears earlier than expected"));
130
break;
@@ -131,6 +144,8 @@ struct multi_pack_index *load_multi_pack_index(const char *object_dir)
144
die(_("multi-pack-index missing required OID fanout chunk"));
145
if (!m->chunk_oid_lookup)
146
die(_("multi-pack-index missing required OID lookup chunk"));
147
+ if (!m->chunk_object_offsets)
148
+ die(_("multi-pack-index missing required object offsets chunk"));
149
150
m->num_objects = ntohl(m->chunk_oid_fanout[255]);
151
@@ -454,6 +469,56 @@ static size_t write_midx_oid_lookup(struct hashfile *f, unsigned char hash_len,
469
return written;
470
}
471
472
+static size_t write_midx_object_offsets(struct hashfile *f, int large_offset_needed,
473
+ struct pack_midx_entry *objects, uint32_t nr_objects)
474
+{
475
+ struct pack_midx_entry *list = objects;
476
+ uint32_t i, nr_large_offset = 0;
477
+ size_t written = 0;
478
+
479
+ for (i = 0; i < nr_objects; i++) {
480
+ struct pack_midx_entry *obj = list++;
481
+
482
+ hashwrite_be32(f, obj->pack_int_id);
483
+
484
+ if (large_offset_needed && obj->offset >> 31)
485
+ hashwrite_be32(f, MIDX_LARGE_OFFSET_NEEDED | nr_large_offset++);
486
+ else if (!large_offset_needed && obj->offset >> 32)
487
+ BUG("object %s requires a large offset (%"PRIx64") but the MIDX is not writing large offsets!",
488
+ oid_to_hex(&obj->oid),
489
+ obj->offset);
490
+ else
491
+ hashwrite_be32(f, (uint32_t)obj->offset);
492
+
493
+ written += MIDX_CHUNK_OFFSET_WIDTH;
494
+ }
495
+
496
+ return written;
497
+}
498
+
499
+static size_t write_midx_large_offsets(struct hashfile *f, uint32_t nr_large_offset,
500
+ struct pack_midx_entry *objects, uint32_t nr_objects)
501
+{
502
+ struct pack_midx_entry *list = objects;
503
+ size_t written = 0;
504
+
505
+ while (nr_large_offset) {
506
+ struct pack_midx_entry *obj = list++;
507
+ uint64_t offset = obj->offset;
508
+
509
+ if (!(offset >> 31))
510
+ continue;
511
+
512
+ hashwrite_be32(f, offset >> 32);
513
+ hashwrite_be32(f, offset & 0xffffffffUL);
514
+ written += 2 * sizeof(uint32_t);
515
+
516
+ nr_large_offset--;
517
+ }
518
+
519
+ return written;
520
+}
521
+
522
int write_midx_file(const char *object_dir)
523
{
524
unsigned char cur_chunk, num_chunks = 0;
@@ -466,8 +531,9 @@ int write_midx_file(const char *object_dir)
531
uint64_t written = 0;
532
uint32_t chunk_ids[MIDX_MAX_CHUNKS + 1];
533
uint64_t chunk_offsets[MIDX_MAX_CHUNKS + 1];
469
- uint32_t nr_entries;
534
+ uint32_t nr_entries, num_large_offsets = 0;
535
struct pack_midx_entry *entries = NULL;
536
+ int large_offsets_needed = 0;
537
538
midx_name = get_midx_filename(object_dir);
539
if (safe_create_leading_directories(midx_name)) {
@@ -494,13 +560,19 @@ int write_midx_file(const char *object_dir)
560
sort_packs_by_name(packs.names, packs.nr, pack_perm);
561
562
entries = get_sorted_entries(packs.list, pack_perm, packs.nr, &nr_entries);
563
+ for (i = 0; i < nr_entries; i++) {
564
+ if (entries[i].offset > 0x7fffffff)
565
+ num_large_offsets++;
566
+ if (entries[i].offset > 0xffffffff)
567
+ large_offsets_needed = 1;
568
+ }
569
570
hold_lock_file_for_update(&lk, midx_name, LOCK_DIE_ON_ERROR);
571
f = hashfd(lk.tempfile->fd, lk.tempfile->filename.buf);
572
FREE_AND_NULL(midx_name);
573
574
cur_chunk = 0;
503
- num_chunks = 3;
575
+ num_chunks = large_offsets_needed ? 5 : 4;
576
577
written = write_midx_header(f, num_chunks, packs.nr);
578
@@ -516,9 +588,21 @@ int write_midx_file(const char *object_dir)
588
chunk_offsets[cur_chunk] = chunk_offsets[cur_chunk - 1] + MIDX_CHUNK_FANOUT_SIZE;
589
590
cur_chunk++;
519
- chunk_ids[cur_chunk] = 0;
591
+ chunk_ids[cur_chunk] = MIDX_CHUNKID_OBJECTOFFSETS;
592
chunk_offsets[cur_chunk] = chunk_offsets[cur_chunk - 1] + nr_entries * MIDX_HASH_LEN;
593
594
+ cur_chunk++;
595
+ chunk_offsets[cur_chunk] = chunk_offsets[cur_chunk - 1] + nr_entries * MIDX_CHUNK_OFFSET_WIDTH;
596
+ if (large_offsets_needed) {
597
+ chunk_ids[cur_chunk] = MIDX_CHUNKID_LARGEOFFSETS;
598
+
599
+ cur_chunk++;
600
+ chunk_offsets[cur_chunk] = chunk_offsets[cur_chunk - 1] +
601
+ num_large_offsets * MIDX_CHUNK_LARGE_OFFSET_WIDTH;
602
+ }
603
+
604
+ chunk_ids[cur_chunk] = 0;
605
+
606
for (i = 0; i <= num_chunks; i++) {
607
if (i && chunk_offsets[i] < chunk_offsets[i - 1])
608
BUG("incorrect chunk offsets: %"PRIu64" before %"PRIu64,
@@ -556,6 +640,14 @@ int write_midx_file(const char *object_dir)
640
written += write_midx_oid_lookup(f, MIDX_HASH_LEN, entries, nr_entries);
641
break;
642
643
+ case MIDX_CHUNKID_OBJECTOFFSETS:
644
+ written += write_midx_object_offsets(f, large_offsets_needed, entries, nr_entries);
645
+ break;
646
+
647
+ case MIDX_CHUNKID_LARGEOFFSETS:
648
+ written += write_midx_large_offsets(f, num_large_offsets, entries, nr_entries);
649
+ break;
650
+
651
default:
652
BUG("trying to write unknown chunk id %"PRIx32,
653
chunk_ids[i]);
midx.h
+2
@@ -17,6 +17,8 @@ struct multi_pack_index {
17
const unsigned char *chunk_pack_names;
18
const uint32_t *chunk_oid_fanout;
19
const unsigned char *chunk_oid_lookup;
20
+ const unsigned char *chunk_object_offsets;
21
+ const unsigned char *chunk_large_offsets;
22
23
const char **pack_names;
24
char object_dir[FLEX_ARRAY];
t/helper/test-read-midx.c
+4
@@ -26,6 +26,10 @@ static int read_midx_file(const char *object_dir)
26
printf(" oid-fanout");
27
if (m->chunk_oid_lookup)
28
printf(" oid-lookup");
29
+ if (m->chunk_object_offsets)
30
+ printf(" object-offsets");
31
+ if (m->chunk_large_offsets)
32
+ printf(" large-offsets");
33
34
printf("\nnum_objects: %d\n", m->num_objects);
35
t/t5319-multi-pack-index.sh
+39
-10
@@ -6,27 +6,30 @@ test_description='multi-pack-indexes'
6
midx_read_expect () {
7
NUM_PACKS=$1
8
NUM_OBJECTS=$2
9
+ NUM_CHUNKS=$3
10
+ OBJECT_DIR=$4
11
+ EXTRA_CHUNKS="$5"
12
{
13
cat <<-EOF &&
11
- header: 4d494458 1 3 $NUM_PACKS
12
- chunks: pack-names oid-fanout oid-lookup
14
+ header: 4d494458 1 $NUM_CHUNKS $NUM_PACKS
15
+ chunks: pack-names oid-fanout oid-lookup object-offsets$EXTRA_CHUNKS
16
num_objects: $NUM_OBJECTS
17
packs:
18
EOF
19
if test $NUM_PACKS -ge 1
20
then
18
- ls pack/ | grep idx | sort
21
+ ls $OBJECT_DIR/pack/ | grep idx | sort
22
fi &&
20
- printf "object-dir: .\n"
23
+ printf "object-dir: $OBJECT_DIR\n"
24
} >expect &&
22
- test-tool read-midx . >actual &&
25
+ test-tool read-midx $OBJECT_DIR >actual &&
26
test_cmp expect actual
27
}
28
29
test_expect_success 'write midx with no packs' '
30
test_when_finished rm -f pack/multi-pack-index &&
31
git multi-pack-index --object-dir=. write &&
29
- midx_read_expect 0 0
32
+ midx_read_expect 0 0 4 .
33
'
34
35
generate_objects () {
@@ -76,13 +79,13 @@ test_expect_success 'write midx with one v1 pack' '
79
pack=$(git pack-objects --index-version=1 pack/test <obj-list) &&
80
test_when_finished rm pack/test-$pack.pack pack/test-$pack.idx pack/multi-pack-index &&
81
git multi-pack-index --object-dir=. write &&
79
- midx_read_expect 1 18
82
+ midx_read_expect 1 18 4 .
83
'
84
85
test_expect_success 'write midx with one v2 pack' '
86
git pack-objects --index-version=2,0x40 pack/test <obj-list &&
87
git multi-pack-index --object-dir=. write &&
85
- midx_read_expect 1 18
88
+ midx_read_expect 1 18 4 .
89
'
90
91
test_expect_success 'add more objects' '
@@ -96,7 +99,7 @@ test_expect_success 'add more objects' '
99
test_expect_success 'write midx with two packs' '
100
git pack-objects --index-version=1 pack/test-2 <obj-list &&
101
git multi-pack-index --object-dir=. write &&
99
- midx_read_expect 2 34
102
+ midx_read_expect 2 34 4 .
103
'
104
105
test_expect_success 'add more packs' '
@@ -110,7 +113,33 @@ test_expect_success 'add more packs' '
113
114
test_expect_success 'write midx with twelve packs' '
115
git multi-pack-index --object-dir=. write &&
113
- midx_read_expect 12 74
116
+ midx_read_expect 12 74 4 .
117
+'
118
+
119
+# usage: corrupt_data <file> <pos> [<data>]
120
+corrupt_data () {
121
+ file=$1
122
+ pos=$2
123
+ data="${3:-\0}"
124
+ printf "$data" | dd of="$file" bs=1 seek="$pos" conv=notrunc
125
+}
126
+
127
+# Force 64-bit offsets by manipulating the idx file.
128
+# This makes the IDX file _incorrect_ so be careful to clean up after!
129
+test_expect_success 'force some 64-bit offsets with pack-objects' '
130
+ mkdir objects64 &&
131
+ mkdir objects64/pack &&
132
+ for i in $(test_seq 1 11)
133
+ do
134
+ generate_objects 11
135
+ done &&
136
+ commit_and_list_objects &&
137
+ pack64=$(git pack-objects --index-version=2,0x40 objects64/pack/test-64 <obj-list) &&
138
+ idx64=objects64/pack/test-64-$pack64.idx &&
139
+ chmod u+w $idx64 &&
140
+ corrupt_data $idx64 2999 "\02" &&
141
+ midx64=$(git multi-pack-index --object-dir=objects64 write) &&
142
+ midx_read_expect 1 63 5 objects64 " large-offsets"
143
'
144
145
test_done