pack-check.c

   1 #include "cache.h"
   2 #include "pack.h"
   3 #include "pack-revindex.h"
   4 #include "progress.h"
   5
   6 struct idx_entry {
   7         off_t                offset;
   8         union idx_entry_object {
   9                 const unsigned char *hash;
  10                 struct object_id *oid;
  11         } oid;
  12         unsigned int nr;
  13 };
  14
  15 static int compare_entries(const void *e1, const void *e2)
  16 {
  17         const struct idx_entry *entry1 = e1;
  18         const struct idx_entry *entry2 = e2;
  19         if (entry1->offset < entry2->offset)
  20                 return -1;
  21         if (entry1->offset > entry2->offset)
  22                 return 1;
  23         return 0;
  24 }
  25
  26 int check_pack_crc(struct packed_git *p, struct pack_window **w_curs,
  27                    off_t offset, off_t len, unsigned int nr)
  28 {
  29         const uint32_t *index_crc;
  30         uint32_t data_crc = crc32(0, NULL, 0);
  31
  32         do {
  33                 unsigned long avail;
  34                 void *data = use_pack(p, w_curs, offset, &avail);
  35                 if (avail > len)
  36                         avail = len;
  37                 data_crc = crc32(data_crc, data, avail);
  38                 offset += avail;
  39                 len -= avail;
  40         } while (len);
  41
  42         index_crc = p->index_data;
  43         index_crc += 2 + 256 + p->num_objects * (20/4) + nr;
  44
  45         return data_crc != ntohl(*index_crc);
  46 }
  47
  48 static int verify_packfile(struct packed_git *p,
  49                            struct pack_window **w_curs,
  50                            verify_fn fn,
  51                            struct progress *progress, uint32_t base_count)
  52
  53 {
  54         off_t index_size = p->index_size;
  55         const unsigned char *index_base = p->index_data;
  56         git_SHA_CTX ctx;
  57         unsigned char hash[GIT_MAX_RAWSZ], *pack_sig;
  58         off_t offset = 0, pack_sig_ofs = 0;
  59         uint32_t nr_objects, i;
  60         int err = 0;
  61         struct idx_entry *entries;
  62
  63         if (!is_pack_valid(p))
  64                 return error("packfile %s cannot be accessed", p->pack_name);
  65
  66         git_SHA1_Init(&ctx);
  67         do {
  68                 unsigned long remaining;
  69                 unsigned char *in = use_pack(p, w_curs, offset, &remaining);
  70                 offset += remaining;
  71                 if (!pack_sig_ofs)
  72                         pack_sig_ofs = p->pack_size - 20;
  73                 if (offset > pack_sig_ofs)
  74                         remaining -= (unsigned int)(offset - pack_sig_ofs);
  75                 git_SHA1_Update(&ctx, in, remaining);
  76         } while (offset < pack_sig_ofs);
  77         git_SHA1_Final(hash, &ctx);
  78         pack_sig = use_pack(p, w_curs, pack_sig_ofs, NULL);
  79         if (hashcmp(hash, pack_sig))
  80                 err = error("%s SHA1 checksum mismatch",
  81                             p->pack_name);
  82         if (hashcmp(index_base + index_size - 40, pack_sig))
  83                 err = error("%s SHA1 does not match its index",
  84                             p->pack_name);
  85         unuse_pack(w_curs);
  86
  87         /* Make sure everything reachable from idx is valid.  Since we
  88          * have verified that nr_objects matches between idx and pack,
  89          * we do not do scan-streaming check on the pack file.
  90          */
  91         nr_objects = p->num_objects;
  92         ALLOC_ARRAY(entries, nr_objects + 1);
  93         entries[nr_objects].offset = pack_sig_ofs;
  94         /* first sort entries by pack offset, since unpacking them is more efficient that way */
  95         for (i = 0; i < nr_objects; i++) {
  96                 entries[i].oid.hash = nth_packed_object_sha1(p, i);
  97                 if (!entries[i].oid.hash)
  98                         die("internal error pack-check nth-packed-object");
  99                 entries[i].offset = nth_packed_object_offset(p, i);
 100                 entries[i].nr = i;
 101         }
 102         QSORT(entries, nr_objects, compare_entries);
 103
 104         for (i = 0; i < nr_objects; i++) {
 105                 void *data;
 106                 enum object_type type;
 107                 unsigned long size;
 108                 off_t curpos;
 109                 int data_valid;
 110
 111                 if (p->index_version > 1) {
 112                         off_t offset = entries[i].offset;
 113                         off_t len = entries[i+1].offset - offset;
 114                         unsigned int nr = entries[i].nr;
 115                         if (check_pack_crc(p, w_curs, offset, len, nr))
 116                                 err = error("index CRC mismatch for object %s "
 117                                             "from %s at offset %"PRIuMAX"",
 118                                             oid_to_hex(entries[i].oid.oid),
 119                                             p->pack_name, (uintmax_t)offset);
 120                 }
 121
 122                 curpos = entries[i].offset;
 123                 type = unpack_object_header(p, w_curs, &curpos, &size);
 124                 unuse_pack(w_curs);
 125
 126                 if (type == OBJ_BLOB && big_file_threshold <= size) {
 127                         /*
 128                          * Let check_sha1_signature() check it with
 129                          * the streaming interface; no point slurping
 130                          * the data in-core only to discard.
 131                          */
 132                         data = NULL;
 133                         data_valid = 0;
 134                 } else {
 135                         data = unpack_entry(p, entries[i].offset, &type, &size);
 136                         data_valid = 1;
 137                 }
 138
 139                 if (data_valid && !data)
 140                         err = error("cannot unpack %s from %s at offset %"PRIuMAX"",
 141                                     oid_to_hex(entries[i].oid.oid), p->pack_name,
 142                                     (uintmax_t)entries[i].offset);
 143                 else if (check_sha1_signature(entries[i].oid.hash, data, size, typename(type)))
 144                         err = error("packed %s from %s is corrupt",
 145                                     oid_to_hex(entries[i].oid.oid), p->pack_name);
 146                 else if (fn) {
 147                         int eaten = 0;
 148                         err |= fn(entries[i].oid.oid, type, size, data, &eaten);
 149                         if (eaten)
 150                                 data = NULL;
 151                 }
 152                 if (((base_count + i) & 1023) == 0)
 153                         display_progress(progress, base_count + i);
 154                 free(data);
 155
 156         }
 157         display_progress(progress, base_count + i);
 158         free(entries);
 159
 160         return err;
 161 }
 162
 163 int verify_pack_index(struct packed_git *p)
 164 {
 165         off_t index_size;
 166         const unsigned char *index_base;
 167         git_SHA_CTX ctx;
 168         unsigned char sha1[20];
 169         int err = 0;
 170
 171         if (open_pack_index(p))
 172                 return error("packfile %s index not opened", p->pack_name);
 173         index_size = p->index_size;
 174         index_base = p->index_data;
 175
 176         /* Verify SHA1 sum of the index file */
 177         git_SHA1_Init(&ctx);
 178         git_SHA1_Update(&ctx, index_base, (unsigned int)(index_size - 20));
 179         git_SHA1_Final(sha1, &ctx);
 180         if (hashcmp(sha1, index_base + index_size - 20))
 181                 err = error("Packfile index for %s SHA1 mismatch",
 182                             p->pack_name);
 183         return err;
 184 }
 185
 186 int verify_pack(struct packed_git *p, verify_fn fn,
 187                 struct progress *progress, uint32_t base_count)
 188 {
 189         int err = 0;
 190         struct pack_window *w_curs = NULL;
 191
 192         err |= verify_pack_index(p);
 193         if (!p->index_data)
 194                 return -1;
 195
 196         err |= verify_packfile(p, &w_curs, fn, progress, base_count);
 197         unuse_pack(&w_curs);
 198
 199         return err;
 200 }