Merge branch 'maint'
[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 = find_pack_entry_one(entries[i].sha1, p);
71                 if (!entries[i].offset)
72                         die("internal error pack-check find-pack-entry-one");
73         }
74         qsort(entries, nr_objects, sizeof(*entries), compare_entries);
75
76         for (i = 0; i < nr_objects; i++) {
77                 void *data;
78                 enum object_type type;
79                 unsigned long size;
80
81                 data = unpack_entry(p, entries[i].offset, &type, &size);
82                 if (!data) {
83                         err = error("cannot unpack %s from %s at offset %"PRIuMAX"",
84                                     sha1_to_hex(entries[i].sha1), p->pack_name,
85                                     (uintmax_t)entries[i].offset);
86                         break;
87                 }
88                 if (check_sha1_signature(entries[i].sha1, data, size, typename(type))) {
89                         err = error("packed %s from %s is corrupt",
90                                     sha1_to_hex(entries[i].sha1), p->pack_name);
91                         free(data);
92                         break;
93                 }
94                 free(data);
95         }
96         free(entries);
97
98         return err;
99 }
100
101
102 #define MAX_CHAIN 50
103
104 static void show_pack_info(struct packed_git *p)
105 {
106         uint32_t nr_objects, i, chain_histogram[MAX_CHAIN+1];
107
108         nr_objects = p->num_objects;
109         memset(chain_histogram, 0, sizeof(chain_histogram));
110         init_pack_revindex();
111
112         for (i = 0; i < nr_objects; i++) {
113                 const unsigned char *sha1;
114                 unsigned char base_sha1[20];
115                 const char *type;
116                 unsigned long size;
117                 unsigned long store_size;
118                 off_t offset;
119                 unsigned int delta_chain_length;
120
121                 sha1 = nth_packed_object_sha1(p, i);
122                 if (!sha1)
123                         die("internal error pack-check nth-packed-object");
124                 offset = find_pack_entry_one(sha1, p);
125                 if (!offset)
126                         die("internal error pack-check find-pack-entry-one");
127
128                 type = packed_object_info_detail(p, offset, &size, &store_size,
129                                                  &delta_chain_length,
130                                                  base_sha1);
131                 printf("%s ", sha1_to_hex(sha1));
132                 if (!delta_chain_length)
133                         printf("%-6s %lu %lu %"PRIuMAX"\n",
134                                type, size, store_size, (uintmax_t)offset);
135                 else {
136                         printf("%-6s %lu %lu %"PRIuMAX" %u %s\n",
137                                type, size, store_size, (uintmax_t)offset,
138                                delta_chain_length, sha1_to_hex(base_sha1));
139                         if (delta_chain_length <= MAX_CHAIN)
140                                 chain_histogram[delta_chain_length]++;
141                         else
142                                 chain_histogram[0]++;
143                 }
144         }
145
146         for (i = 0; i <= MAX_CHAIN; i++) {
147                 if (!chain_histogram[i])
148                         continue;
149                 printf("chain length = %d: %d object%s\n", i,
150                        chain_histogram[i], chain_histogram[i] > 1 ? "s" : "");
151         }
152         if (chain_histogram[0])
153                 printf("chain length > %d: %d object%s\n", MAX_CHAIN,
154                        chain_histogram[0], chain_histogram[0] > 1 ? "s" : "");
155 }
156
157 int verify_pack(struct packed_git *p, int verbose)
158 {
159         off_t index_size;
160         const unsigned char *index_base;
161         SHA_CTX ctx;
162         unsigned char sha1[20];
163         int err = 0;
164         struct pack_window *w_curs = NULL;
165
166         if (open_pack_index(p))
167                 return error("packfile %s index not opened", p->pack_name);
168         index_size = p->index_size;
169         index_base = p->index_data;
170
171         /* Verify SHA1 sum of the index file */
172         SHA1_Init(&ctx);
173         SHA1_Update(&ctx, index_base, (unsigned int)(index_size - 20));
174         SHA1_Final(sha1, &ctx);
175         if (hashcmp(sha1, index_base + index_size - 20))
176                 err = error("Packfile index for %s SHA1 mismatch",
177                             p->pack_name);
178
179         /* Verify pack file */
180         err |= verify_packfile(p, &w_curs);
181         unuse_pack(&w_curs);
182
183         if (verbose) {
184                 if (err)
185                         printf("%s: bad\n", p->pack_name);
186                 else {
187                         show_pack_info(p);
188                         printf("%s: ok\n", p->pack_name);
189                 }
190         }
191
192         return err;
193 }