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