Merge branch 'ag/rebase-p'
[git.git] / pack-check.c
blobd3a57df34f2d2bf0ef9935a51f171f24d8ed6a72
1 #include "cache.h"
2 #include "repository.h"
3 #include "pack.h"
4 #include "pack-revindex.h"
5 #include "progress.h"
6 #include "packfile.h"
7 #include "object-store.h"
9 struct idx_entry {
10 off_t offset;
11 union idx_entry_object {
12 const unsigned char *hash;
13 struct object_id *oid;
14 } oid;
15 unsigned int nr;
18 static int compare_entries(const void *e1, const void *e2)
20 const struct idx_entry *entry1 = e1;
21 const struct idx_entry *entry2 = e2;
22 if (entry1->offset < entry2->offset)
23 return -1;
24 if (entry1->offset > entry2->offset)
25 return 1;
26 return 0;
29 int check_pack_crc(struct packed_git *p, struct pack_window **w_curs,
30 off_t offset, off_t len, unsigned int nr)
32 const uint32_t *index_crc;
33 uint32_t data_crc = crc32(0, NULL, 0);
35 do {
36 unsigned long avail;
37 void *data = use_pack(p, w_curs, offset, &avail);
38 if (avail > len)
39 avail = len;
40 data_crc = crc32(data_crc, data, avail);
41 offset += avail;
42 len -= avail;
43 } while (len);
45 index_crc = p->index_data;
46 index_crc += 2 + 256 + p->num_objects * (the_hash_algo->rawsz/4) + nr;
48 return data_crc != ntohl(*index_crc);
51 static int verify_packfile(struct packed_git *p,
52 struct pack_window **w_curs,
53 verify_fn fn,
54 struct progress *progress, uint32_t base_count)
57 off_t index_size = p->index_size;
58 const unsigned char *index_base = p->index_data;
59 git_hash_ctx ctx;
60 unsigned char hash[GIT_MAX_RAWSZ], *pack_sig;
61 off_t offset = 0, pack_sig_ofs = 0;
62 uint32_t nr_objects, i;
63 int err = 0;
64 struct idx_entry *entries;
66 if (!is_pack_valid(p))
67 return error("packfile %s cannot be accessed", p->pack_name);
69 the_hash_algo->init_fn(&ctx);
70 do {
71 unsigned long remaining;
72 unsigned char *in = use_pack(p, w_curs, offset, &remaining);
73 offset += remaining;
74 if (!pack_sig_ofs)
75 pack_sig_ofs = p->pack_size - the_hash_algo->rawsz;
76 if (offset > pack_sig_ofs)
77 remaining -= (unsigned int)(offset - pack_sig_ofs);
78 the_hash_algo->update_fn(&ctx, in, remaining);
79 } while (offset < pack_sig_ofs);
80 the_hash_algo->final_fn(hash, &ctx);
81 pack_sig = use_pack(p, w_curs, pack_sig_ofs, NULL);
82 if (hashcmp(hash, pack_sig))
83 err = error("%s pack checksum mismatch",
84 p->pack_name);
85 if (hashcmp(index_base + index_size - the_hash_algo->hexsz, pack_sig))
86 err = error("%s pack checksum does not match its index",
87 p->pack_name);
88 unuse_pack(w_curs);
90 /* Make sure everything reachable from idx is valid. Since we
91 * have verified that nr_objects matches between idx and pack,
92 * we do not do scan-streaming check on the pack file.
94 nr_objects = p->num_objects;
95 ALLOC_ARRAY(entries, nr_objects + 1);
96 entries[nr_objects].offset = pack_sig_ofs;
97 /* first sort entries by pack offset, since unpacking them is more efficient that way */
98 for (i = 0; i < nr_objects; i++) {
99 entries[i].oid.hash = nth_packed_object_sha1(p, i);
100 if (!entries[i].oid.hash)
101 die("internal error pack-check nth-packed-object");
102 entries[i].offset = nth_packed_object_offset(p, i);
103 entries[i].nr = i;
105 QSORT(entries, nr_objects, compare_entries);
107 for (i = 0; i < nr_objects; i++) {
108 void *data;
109 enum object_type type;
110 unsigned long size;
111 off_t curpos;
112 int data_valid;
114 if (p->index_version > 1) {
115 off_t offset = entries[i].offset;
116 off_t len = entries[i+1].offset - offset;
117 unsigned int nr = entries[i].nr;
118 if (check_pack_crc(p, w_curs, offset, len, nr))
119 err = error("index CRC mismatch for object %s "
120 "from %s at offset %"PRIuMAX"",
121 oid_to_hex(entries[i].oid.oid),
122 p->pack_name, (uintmax_t)offset);
125 curpos = entries[i].offset;
126 type = unpack_object_header(p, w_curs, &curpos, &size);
127 unuse_pack(w_curs);
129 if (type == OBJ_BLOB && big_file_threshold <= size) {
131 * Let check_object_signature() check it with
132 * the streaming interface; no point slurping
133 * the data in-core only to discard.
135 data = NULL;
136 data_valid = 0;
137 } else {
138 data = unpack_entry(the_repository, p, entries[i].offset, &type, &size);
139 data_valid = 1;
142 if (data_valid && !data)
143 err = error("cannot unpack %s from %s at offset %"PRIuMAX"",
144 oid_to_hex(entries[i].oid.oid), p->pack_name,
145 (uintmax_t)entries[i].offset);
146 else if (check_object_signature(entries[i].oid.oid, data, size, type_name(type)))
147 err = error("packed %s from %s is corrupt",
148 oid_to_hex(entries[i].oid.oid), p->pack_name);
149 else if (fn) {
150 int eaten = 0;
151 err |= fn(entries[i].oid.oid, type, size, data, &eaten);
152 if (eaten)
153 data = NULL;
155 if (((base_count + i) & 1023) == 0)
156 display_progress(progress, base_count + i);
157 free(data);
160 display_progress(progress, base_count + i);
161 free(entries);
163 return err;
166 int verify_pack_index(struct packed_git *p)
168 off_t index_size;
169 const unsigned char *index_base;
170 git_hash_ctx ctx;
171 unsigned char hash[GIT_MAX_RAWSZ];
172 int err = 0;
174 if (open_pack_index(p))
175 return error("packfile %s index not opened", p->pack_name);
176 index_size = p->index_size;
177 index_base = p->index_data;
179 /* Verify SHA1 sum of the index file */
180 the_hash_algo->init_fn(&ctx);
181 the_hash_algo->update_fn(&ctx, index_base, (unsigned int)(index_size - the_hash_algo->rawsz));
182 the_hash_algo->final_fn(hash, &ctx);
183 if (hashcmp(hash, index_base + index_size - the_hash_algo->rawsz))
184 err = error("Packfile index for %s hash mismatch",
185 p->pack_name);
186 return err;
189 int verify_pack(struct packed_git *p, verify_fn fn,
190 struct progress *progress, uint32_t base_count)
192 int err = 0;
193 struct pack_window *w_curs = NULL;
195 err |= verify_pack_index(p);
196 if (!p->index_data)
197 return -1;
199 err |= verify_packfile(p, &w_curs, fn, progress, base_count);
200 unuse_pack(&w_curs);
202 return err;