Merge branch 'ds/commit-graph'
[git] / pack-check.c
1 #include "cache.h"
2 #include "pack.h"
3 #include "pack-revindex.h"
4 #include "progress.h"
5 #include "packfile.h"
6 #include "object-store.h"
7
8 struct idx_entry {
9         off_t                offset;
10         union idx_entry_object {
11                 const unsigned char *hash;
12                 struct object_id *oid;
13         } oid;
14         unsigned int nr;
15 };
16
17 static int compare_entries(const void *e1, const void *e2)
18 {
19         const struct idx_entry *entry1 = e1;
20         const struct idx_entry *entry2 = e2;
21         if (entry1->offset < entry2->offset)
22                 return -1;
23         if (entry1->offset > entry2->offset)
24                 return 1;
25         return 0;
26 }
27
28 int check_pack_crc(struct packed_git *p, struct pack_window **w_curs,
29                    off_t offset, off_t len, unsigned int nr)
30 {
31         const uint32_t *index_crc;
32         uint32_t data_crc = crc32(0, NULL, 0);
33
34         do {
35                 unsigned long avail;
36                 void *data = use_pack(p, w_curs, offset, &avail);
37                 if (avail > len)
38                         avail = len;
39                 data_crc = crc32(data_crc, data, avail);
40                 offset += avail;
41                 len -= avail;
42         } while (len);
43
44         index_crc = p->index_data;
45         index_crc += 2 + 256 + p->num_objects * (the_hash_algo->rawsz/4) + nr;
46
47         return data_crc != ntohl(*index_crc);
48 }
49
50 static int verify_packfile(struct packed_git *p,
51                            struct pack_window **w_curs,
52                            verify_fn fn,
53                            struct progress *progress, uint32_t base_count)
54
55 {
56         off_t index_size = p->index_size;
57         const unsigned char *index_base = p->index_data;
58         git_hash_ctx ctx;
59         unsigned char hash[GIT_MAX_RAWSZ], *pack_sig;
60         off_t offset = 0, pack_sig_ofs = 0;
61         uint32_t nr_objects, i;
62         int err = 0;
63         struct idx_entry *entries;
64
65         if (!is_pack_valid(p))
66                 return error("packfile %s cannot be accessed", p->pack_name);
67
68         the_hash_algo->init_fn(&ctx);
69         do {
70                 unsigned long remaining;
71                 unsigned char *in = use_pack(p, w_curs, offset, &remaining);
72                 offset += remaining;
73                 if (!pack_sig_ofs)
74                         pack_sig_ofs = p->pack_size - the_hash_algo->rawsz;
75                 if (offset > pack_sig_ofs)
76                         remaining -= (unsigned int)(offset - pack_sig_ofs);
77                 the_hash_algo->update_fn(&ctx, in, remaining);
78         } while (offset < pack_sig_ofs);
79         the_hash_algo->final_fn(hash, &ctx);
80         pack_sig = use_pack(p, w_curs, pack_sig_ofs, NULL);
81         if (hashcmp(hash, pack_sig))
82                 err = error("%s pack checksum mismatch",
83                             p->pack_name);
84         if (hashcmp(index_base + index_size - the_hash_algo->hexsz, pack_sig))
85                 err = error("%s pack checksum does not match its index",
86                             p->pack_name);
87         unuse_pack(w_curs);
88
89         /* Make sure everything reachable from idx is valid.  Since we
90          * have verified that nr_objects matches between idx and pack,
91          * we do not do scan-streaming check on the pack file.
92          */
93         nr_objects = p->num_objects;
94         ALLOC_ARRAY(entries, nr_objects + 1);
95         entries[nr_objects].offset = pack_sig_ofs;
96         /* first sort entries by pack offset, since unpacking them is more efficient that way */
97         for (i = 0; i < nr_objects; i++) {
98                 entries[i].oid.hash = nth_packed_object_sha1(p, i);
99                 if (!entries[i].oid.hash)
100                         die("internal error pack-check nth-packed-object");
101                 entries[i].offset = nth_packed_object_offset(p, i);
102                 entries[i].nr = i;
103         }
104         QSORT(entries, nr_objects, compare_entries);
105
106         for (i = 0; i < nr_objects; i++) {
107                 void *data;
108                 enum object_type type;
109                 unsigned long size;
110                 off_t curpos;
111                 int data_valid;
112
113                 if (p->index_version > 1) {
114                         off_t offset = entries[i].offset;
115                         off_t len = entries[i+1].offset - offset;
116                         unsigned int nr = entries[i].nr;
117                         if (check_pack_crc(p, w_curs, offset, len, nr))
118                                 err = error("index CRC mismatch for object %s "
119                                             "from %s at offset %"PRIuMAX"",
120                                             oid_to_hex(entries[i].oid.oid),
121                                             p->pack_name, (uintmax_t)offset);
122                 }
123
124                 curpos = entries[i].offset;
125                 type = unpack_object_header(p, w_curs, &curpos, &size);
126                 unuse_pack(w_curs);
127
128                 if (type == OBJ_BLOB && big_file_threshold <= size) {
129                         /*
130                          * Let check_object_signature() check it with
131                          * the streaming interface; no point slurping
132                          * the data in-core only to discard.
133                          */
134                         data = NULL;
135                         data_valid = 0;
136                 } else {
137                         data = unpack_entry(p, entries[i].offset, &type, &size);
138                         data_valid = 1;
139                 }
140
141                 if (data_valid && !data)
142                         err = error("cannot unpack %s from %s at offset %"PRIuMAX"",
143                                     oid_to_hex(entries[i].oid.oid), p->pack_name,
144                                     (uintmax_t)entries[i].offset);
145                 else if (check_object_signature(entries[i].oid.oid, data, size, type_name(type)))
146                         err = error("packed %s from %s is corrupt",
147                                     oid_to_hex(entries[i].oid.oid), p->pack_name);
148                 else if (fn) {
149                         int eaten = 0;
150                         err |= fn(entries[i].oid.oid, type, size, data, &eaten);
151                         if (eaten)
152                                 data = NULL;
153                 }
154                 if (((base_count + i) & 1023) == 0)
155                         display_progress(progress, base_count + i);
156                 free(data);
157
158         }
159         display_progress(progress, base_count + i);
160         free(entries);
161
162         return err;
163 }
164
165 int verify_pack_index(struct packed_git *p)
166 {
167         off_t index_size;
168         const unsigned char *index_base;
169         git_hash_ctx ctx;
170         unsigned char hash[GIT_MAX_RAWSZ];
171         int err = 0;
172
173         if (open_pack_index(p))
174                 return error("packfile %s index not opened", p->pack_name);
175         index_size = p->index_size;
176         index_base = p->index_data;
177
178         /* Verify SHA1 sum of the index file */
179         the_hash_algo->init_fn(&ctx);
180         the_hash_algo->update_fn(&ctx, index_base, (unsigned int)(index_size - the_hash_algo->rawsz));
181         the_hash_algo->final_fn(hash, &ctx);
182         if (hashcmp(hash, index_base + index_size - the_hash_algo->rawsz))
183                 err = error("Packfile index for %s hash mismatch",
184                             p->pack_name);
185         return err;
186 }
187
188 int verify_pack(struct packed_git *p, verify_fn fn,
189                 struct progress *progress, uint32_t base_count)
190 {
191         int err = 0;
192         struct pack_window *w_curs = NULL;
193
194         err |= verify_pack_index(p);
195         if (!p->index_data)
196                 return -1;
197
198         err |= verify_packfile(p, &w_curs, fn, progress, base_count);
199         unuse_pack(&w_curs);
200
201         return err;
202 }