pack-check.con commit archive: refactor file extension format-guessing (08716b3)
   1#include "cache.h"
   2#include "pack.h"
   3#include "pack-revindex.h"
   4
   5struct idx_entry {
   6        off_t                offset;
   7        const unsigned char *sha1;
   8        unsigned int nr;
   9};
  10
  11static 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
  22int check_pack_crc(struct packed_git *p, struct pack_window **w_curs,
  23                   off_t offset, off_t len, unsigned int nr)
  24{
  25        const uint32_t *index_crc;
  26        uint32_t data_crc = crc32(0, NULL, 0);
  27
  28        do {
  29                unsigned int avail;
  30                void *data = use_pack(p, w_curs, offset, &avail);
  31                if (avail > len)
  32                        avail = len;
  33                data_crc = crc32(data_crc, data, avail);
  34                offset += avail;
  35                len -= avail;
  36        } while (len);
  37
  38        index_crc = p->index_data;
  39        index_crc += 2 + 256 + p->num_objects * (20/4) + nr;
  40
  41        return data_crc != ntohl(*index_crc);
  42}
  43
  44static int verify_packfile(struct packed_git *p,
  45                struct pack_window **w_curs)
  46{
  47        off_t index_size = p->index_size;
  48        const unsigned char *index_base = p->index_data;
  49        git_SHA_CTX ctx;
  50        unsigned char sha1[20], *pack_sig;
  51        off_t offset = 0, pack_sig_ofs = 0;
  52        uint32_t nr_objects, i;
  53        int err = 0;
  54        struct idx_entry *entries;
  55
  56        /* Note that the pack header checks are actually performed by
  57         * use_pack when it first opens the pack file.  If anything
  58         * goes wrong during those checks then the call will die out
  59         * immediately.
  60         */
  61
  62        git_SHA1_Init(&ctx);
  63        do {
  64                unsigned int remaining;
  65                unsigned char *in = use_pack(p, w_curs, offset, &remaining);
  66                offset += remaining;
  67                if (!pack_sig_ofs)
  68                        pack_sig_ofs = p->pack_size - 20;
  69                if (offset > pack_sig_ofs)
  70                        remaining -= (unsigned int)(offset - pack_sig_ofs);
  71                git_SHA1_Update(&ctx, in, remaining);
  72        } while (offset < pack_sig_ofs);
  73        git_SHA1_Final(sha1, &ctx);
  74        pack_sig = use_pack(p, w_curs, pack_sig_ofs, NULL);
  75        if (hashcmp(sha1, pack_sig))
  76                err = error("%s SHA1 checksum mismatch",
  77                            p->pack_name);
  78        if (hashcmp(index_base + index_size - 40, pack_sig))
  79                err = error("%s SHA1 does not match its index",
  80                            p->pack_name);
  81        unuse_pack(w_curs);
  82
  83        /* Make sure everything reachable from idx is valid.  Since we
  84         * have verified that nr_objects matches between idx and pack,
  85         * we do not do scan-streaming check on the pack file.
  86         */
  87        nr_objects = p->num_objects;
  88        entries = xmalloc((nr_objects + 1) * sizeof(*entries));
  89        entries[nr_objects].offset = pack_sig_ofs;
  90        /* first sort entries by pack offset, since unpacking them is more efficient that way */
  91        for (i = 0; i < nr_objects; i++) {
  92                entries[i].sha1 = nth_packed_object_sha1(p, i);
  93                if (!entries[i].sha1)
  94                        die("internal error pack-check nth-packed-object");
  95                entries[i].offset = nth_packed_object_offset(p, i);
  96                entries[i].nr = i;
  97        }
  98        qsort(entries, nr_objects, sizeof(*entries), compare_entries);
  99
 100        for (i = 0; i < nr_objects; i++) {
 101                void *data;
 102                enum object_type type;
 103                unsigned long size;
 104
 105                if (p->index_version > 1) {
 106                        off_t offset = entries[i].offset;
 107                        off_t len = entries[i+1].offset - offset;
 108                        unsigned int nr = entries[i].nr;
 109                        if (check_pack_crc(p, w_curs, offset, len, nr))
 110                                err = error("index CRC mismatch for object %s "
 111                                            "from %s at offset %"PRIuMAX"",
 112                                            sha1_to_hex(entries[i].sha1),
 113                                            p->pack_name, (uintmax_t)offset);
 114                }
 115                data = unpack_entry(p, entries[i].offset, &type, &size);
 116                if (!data) {
 117                        err = error("cannot unpack %s from %s at offset %"PRIuMAX"",
 118                                    sha1_to_hex(entries[i].sha1), p->pack_name,
 119                                    (uintmax_t)entries[i].offset);
 120                        break;
 121                }
 122                if (check_sha1_signature(entries[i].sha1, data, size, typename(type))) {
 123                        err = error("packed %s from %s is corrupt",
 124                                    sha1_to_hex(entries[i].sha1), p->pack_name);
 125                        free(data);
 126                        break;
 127                }
 128                free(data);
 129        }
 130        free(entries);
 131
 132        return err;
 133}
 134
 135int verify_pack_index(struct packed_git *p)
 136{
 137        off_t index_size;
 138        const unsigned char *index_base;
 139        git_SHA_CTX ctx;
 140        unsigned char sha1[20];
 141        int err = 0;
 142
 143        if (open_pack_index(p))
 144                return error("packfile %s index not opened", p->pack_name);
 145        index_size = p->index_size;
 146        index_base = p->index_data;
 147
 148        /* Verify SHA1 sum of the index file */
 149        git_SHA1_Init(&ctx);
 150        git_SHA1_Update(&ctx, index_base, (unsigned int)(index_size - 20));
 151        git_SHA1_Final(sha1, &ctx);
 152        if (hashcmp(sha1, index_base + index_size - 20))
 153                err = error("Packfile index for %s SHA1 mismatch",
 154                            p->pack_name);
 155        return err;
 156}
 157
 158int verify_pack(struct packed_git *p)
 159{
 160        int err = 0;
 161        struct pack_window *w_curs = NULL;
 162
 163        err |= verify_pack_index(p);
 164        if (!p->index_data)
 165                return -1;
 166
 167        err |= verify_packfile(p, &w_curs);
 168        unuse_pack(&w_curs);
 169
 170        return err;
 171}