optimize verify-pack a bit
[git/git.git] / pack-check.c
1 #include "cache.h"
2 #include "pack.h"
3 #include "pack-revindex.h"
4
5 struct idx_entry
6 {
7 const unsigned char *sha1;
8 off_t offset;
9 };
10
11 static int compare_entries(const void *e1, const void *e2)
12 {
13 const struct idx_entry *entry1 = e1;
14 const struct idx_entry *entry2 = e2;
15 if (entry1->offset < entry2->offset)
16 return -1;
17 if (entry1->offset > entry2->offset)
18 return 1;
19 return 0;
20 }
21
22 static int verify_packfile(struct packed_git *p,
23 struct pack_window **w_curs)
24 {
25 off_t index_size = p->index_size;
26 const unsigned char *index_base = p->index_data;
27 SHA_CTX ctx;
28 unsigned char sha1[20], *pack_sig;
29 off_t offset = 0, pack_sig_ofs = p->pack_size - 20;
30 uint32_t nr_objects, i;
31 int err = 0;
32 struct idx_entry *entries;
33
34 /* Note that the pack header checks are actually performed by
35 * use_pack when it first opens the pack file. If anything
36 * goes wrong during those checks then the call will die out
37 * immediately.
38 */
39
40 SHA1_Init(&ctx);
41 while (offset < pack_sig_ofs) {
42 unsigned int remaining;
43 unsigned char *in = use_pack(p, w_curs, offset, &remaining);
44 offset += remaining;
45 if (offset > pack_sig_ofs)
46 remaining -= (unsigned int)(offset - pack_sig_ofs);
47 SHA1_Update(&ctx, in, remaining);
48 }
49 SHA1_Final(sha1, &ctx);
50 pack_sig = use_pack(p, w_curs, pack_sig_ofs, NULL);
51 if (hashcmp(sha1, pack_sig))
52 err = error("%s SHA1 checksum mismatch",
53 p->pack_name);
54 if (hashcmp(index_base + index_size - 40, pack_sig))
55 err = error("%s SHA1 does not match its inddex",
56 p->pack_name);
57 unuse_pack(w_curs);
58
59 /* Make sure everything reachable from idx is valid. Since we
60 * have verified that nr_objects matches between idx and pack,
61 * we do not do scan-streaming check on the pack file.
62 */
63 nr_objects = p->num_objects;
64 entries = xmalloc(nr_objects * sizeof(*entries));
65 /* first sort entries by pack offset, since unpacking them is more efficient that way */
66 for (i = 0; i < nr_objects; i++) {
67 entries[i].sha1 = nth_packed_object_sha1(p, i);
68 if (!entries[i].sha1)
69 die("internal error pack-check nth-packed-object");
70 entries[i].offset = nth_packed_object_offset(p, i);
71 }
72 qsort(entries, nr_objects, sizeof(*entries), compare_entries);
73
74 for (i = 0; i < nr_objects; i++) {
75 void *data;
76 enum object_type type;
77 unsigned long size;
78
79 data = unpack_entry(p, entries[i].offset, &type, &size);
80 if (!data) {
81 err = error("cannot unpack %s from %s at offset %"PRIuMAX"",
82 sha1_to_hex(entries[i].sha1), p->pack_name,
83 (uintmax_t)entries[i].offset);
84 break;
85 }
86 if (check_sha1_signature(entries[i].sha1, data, size, typename(type))) {
87 err = error("packed %s from %s is corrupt",
88 sha1_to_hex(entries[i].sha1), p->pack_name);
89 free(data);
90 break;
91 }
92 free(data);
93 }
94 free(entries);
95
96 return err;
97 }
98
99
100 #define MAX_CHAIN 50
101
102 static void show_pack_info(struct packed_git *p)
103 {
104 uint32_t nr_objects, i, chain_histogram[MAX_CHAIN+1];
105
106 nr_objects = p->num_objects;
107 memset(chain_histogram, 0, sizeof(chain_histogram));
108
109 for (i = 0; i < nr_objects; i++) {
110 const unsigned char *sha1;
111 unsigned char base_sha1[20];
112 const char *type;
113 unsigned long size;
114 unsigned long store_size;
115 off_t offset;
116 unsigned int delta_chain_length;
117
118 sha1 = nth_packed_object_sha1(p, i);
119 if (!sha1)
120 die("internal error pack-check nth-packed-object");
121 offset = find_pack_entry_one(sha1, p);
122 if (!offset)
123 die("internal error pack-check find-pack-entry-one");
124
125 type = packed_object_info_detail(p, offset, &size, &store_size,
126 &delta_chain_length,
127 base_sha1);
128 printf("%s ", sha1_to_hex(sha1));
129 if (!delta_chain_length)
130 printf("%-6s %lu %lu %"PRIuMAX"\n",
131 type, size, store_size, (uintmax_t)offset);
132 else {
133 printf("%-6s %lu %lu %"PRIuMAX" %u %s\n",
134 type, size, store_size, (uintmax_t)offset,
135 delta_chain_length, sha1_to_hex(base_sha1));
136 if (delta_chain_length <= MAX_CHAIN)
137 chain_histogram[delta_chain_length]++;
138 else
139 chain_histogram[0]++;
140 }
141 }
142
143 for (i = 0; i <= MAX_CHAIN; i++) {
144 if (!chain_histogram[i])
145 continue;
146 printf("chain length = %d: %d object%s\n", i,
147 chain_histogram[i], chain_histogram[i] > 1 ? "s" : "");
148 }
149 if (chain_histogram[0])
150 printf("chain length > %d: %d object%s\n", MAX_CHAIN,
151 chain_histogram[0], chain_histogram[0] > 1 ? "s" : "");
152 }
153
154 int verify_pack(struct packed_git *p, int verbose)
155 {
156 off_t index_size;
157 const unsigned char *index_base;
158 SHA_CTX ctx;
159 unsigned char sha1[20];
160 int err = 0;
161 struct pack_window *w_curs = NULL;
162
163 if (open_pack_index(p))
164 return error("packfile %s index not opened", p->pack_name);
165 index_size = p->index_size;
166 index_base = p->index_data;
167
168 /* Verify SHA1 sum of the index file */
169 SHA1_Init(&ctx);
170 SHA1_Update(&ctx, index_base, (unsigned int)(index_size - 20));
171 SHA1_Final(sha1, &ctx);
172 if (hashcmp(sha1, index_base + index_size - 20))
173 err = error("Packfile index for %s SHA1 mismatch",
174 p->pack_name);
175
176 /* Verify pack file */
177 err |= verify_packfile(p, &w_curs);
178 unuse_pack(&w_curs);
179
180 if (verbose) {
181 if (err)
182 printf("%s: bad\n", p->pack_name);
183 else {
184 show_pack_info(p);
185 printf("%s: ok\n", p->pack_name);
186 }
187 }
188
189 return err;
190 }