pack-check.con commit Merge branch 'sb/doc-unify-bottom' (153a33f)
   1#include "cache.h"
   2#include "pack.h"
   3#include "pack-revindex.h"
   4#include "progress.h"
   5
   6struct idx_entry {
   7        off_t                offset;
   8        const unsigned char *sha1;
   9        unsigned int nr;
  10};
  11
  12static int compare_entries(const void *e1, const void *e2)
  13{
  14        const struct idx_entry *entry1 = e1;
  15        const struct idx_entry *entry2 = e2;
  16        if (entry1->offset < entry2->offset)
  17                return -1;
  18        if (entry1->offset > entry2->offset)
  19                return 1;
  20        return 0;
  21}
  22
  23int check_pack_crc(struct packed_git *p, struct pack_window **w_curs,
  24                   off_t offset, off_t len, unsigned int nr)
  25{
  26        const uint32_t *index_crc;
  27        uint32_t data_crc = crc32(0, NULL, 0);
  28
  29        do {
  30                unsigned long avail;
  31                void *data = use_pack(p, w_curs, offset, &avail);
  32                if (avail > len)
  33                        avail = len;
  34                data_crc = crc32(data_crc, data, avail);
  35                offset += avail;
  36                len -= avail;
  37        } while (len);
  38
  39        index_crc = p->index_data;
  40        index_crc += 2 + 256 + p->num_objects * (20/4) + nr;
  41
  42        return data_crc != ntohl(*index_crc);
  43}
  44
  45static int verify_packfile(struct packed_git *p,
  46                           struct pack_window **w_curs,
  47                           verify_fn fn,
  48                           struct progress *progress, uint32_t base_count)
  49
  50{
  51        off_t index_size = p->index_size;
  52        const unsigned char *index_base = p->index_data;
  53        git_SHA_CTX ctx;
  54        unsigned char sha1[20], *pack_sig;
  55        off_t offset = 0, pack_sig_ofs = 0;
  56        uint32_t nr_objects, i;
  57        int err = 0;
  58        struct idx_entry *entries;
  59
  60        if (!is_pack_valid(p))
  61                return error("packfile %s cannot be accessed", p->pack_name);
  62
  63        git_SHA1_Init(&ctx);
  64        do {
  65                unsigned long remaining;
  66                unsigned char *in = use_pack(p, w_curs, offset, &remaining);
  67                offset += remaining;
  68                if (!pack_sig_ofs)
  69                        pack_sig_ofs = p->pack_size - 20;
  70                if (offset > pack_sig_ofs)
  71                        remaining -= (unsigned int)(offset - pack_sig_ofs);
  72                git_SHA1_Update(&ctx, in, remaining);
  73        } while (offset < pack_sig_ofs);
  74        git_SHA1_Final(sha1, &ctx);
  75        pack_sig = use_pack(p, w_curs, pack_sig_ofs, NULL);
  76        if (hashcmp(sha1, pack_sig))
  77                err = error("%s SHA1 checksum mismatch",
  78                            p->pack_name);
  79        if (hashcmp(index_base + index_size - 40, pack_sig))
  80                err = error("%s SHA1 does not match its index",
  81                            p->pack_name);
  82        unuse_pack(w_curs);
  83
  84        /* Make sure everything reachable from idx is valid.  Since we
  85         * have verified that nr_objects matches between idx and pack,
  86         * we do not do scan-streaming check on the pack file.
  87         */
  88        nr_objects = p->num_objects;
  89        ALLOC_ARRAY(entries, nr_objects + 1);
  90        entries[nr_objects].offset = pack_sig_ofs;
  91        /* first sort entries by pack offset, since unpacking them is more efficient that way */
  92        for (i = 0; i < nr_objects; i++) {
  93                entries[i].sha1 = nth_packed_object_sha1(p, i);
  94                if (!entries[i].sha1)
  95                        die("internal error pack-check nth-packed-object");
  96                entries[i].offset = nth_packed_object_offset(p, i);
  97                entries[i].nr = i;
  98        }
  99        QSORT(entries, nr_objects, compare_entries);
 100
 101        for (i = 0; i < nr_objects; i++) {
 102                void *data;
 103                enum object_type type;
 104                unsigned long size;
 105                off_t curpos;
 106                int data_valid;
 107
 108                if (p->index_version > 1) {
 109                        off_t offset = entries[i].offset;
 110                        off_t len = entries[i+1].offset - offset;
 111                        unsigned int nr = entries[i].nr;
 112                        if (check_pack_crc(p, w_curs, offset, len, nr))
 113                                err = error("index CRC mismatch for object %s "
 114                                            "from %s at offset %"PRIuMAX"",
 115                                            sha1_to_hex(entries[i].sha1),
 116                                            p->pack_name, (uintmax_t)offset);
 117                }
 118
 119                curpos = entries[i].offset;
 120                type = unpack_object_header(p, w_curs, &curpos, &size);
 121                unuse_pack(w_curs);
 122
 123                if (type == OBJ_BLOB && big_file_threshold <= size) {
 124                        /*
 125                         * Let check_sha1_signature() check it with
 126                         * the streaming interface; no point slurping
 127                         * the data in-core only to discard.
 128                         */
 129                        data = NULL;
 130                        data_valid = 0;
 131                } else {
 132                        data = unpack_entry(p, entries[i].offset, &type, &size);
 133                        data_valid = 1;
 134                }
 135
 136                if (data_valid && !data)
 137                        err = error("cannot unpack %s from %s at offset %"PRIuMAX"",
 138                                    sha1_to_hex(entries[i].sha1), p->pack_name,
 139                                    (uintmax_t)entries[i].offset);
 140                else if (check_sha1_signature(entries[i].sha1, data, size, typename(type)))
 141                        err = error("packed %s from %s is corrupt",
 142                                    sha1_to_hex(entries[i].sha1), p->pack_name);
 143                else if (fn) {
 144                        int eaten = 0;
 145                        err |= fn(entries[i].sha1, type, size, data, &eaten);
 146                        if (eaten)
 147                                data = NULL;
 148                }
 149                if (((base_count + i) & 1023) == 0)
 150                        display_progress(progress, base_count + i);
 151                free(data);
 152
 153        }
 154        display_progress(progress, base_count + i);
 155        free(entries);
 156
 157        return err;
 158}
 159
 160int verify_pack_index(struct packed_git *p)
 161{
 162        off_t index_size;
 163        const unsigned char *index_base;
 164        git_SHA_CTX ctx;
 165        unsigned char sha1[20];
 166        int err = 0;
 167
 168        if (open_pack_index(p))
 169                return error("packfile %s index not opened", p->pack_name);
 170        index_size = p->index_size;
 171        index_base = p->index_data;
 172
 173        /* Verify SHA1 sum of the index file */
 174        git_SHA1_Init(&ctx);
 175        git_SHA1_Update(&ctx, index_base, (unsigned int)(index_size - 20));
 176        git_SHA1_Final(sha1, &ctx);
 177        if (hashcmp(sha1, index_base + index_size - 20))
 178                err = error("Packfile index for %s SHA1 mismatch",
 179                            p->pack_name);
 180        return err;
 181}
 182
 183int verify_pack(struct packed_git *p, verify_fn fn,
 184                struct progress *progress, uint32_t base_count)
 185{
 186        int err = 0;
 187        struct pack_window *w_curs = NULL;
 188
 189        err |= verify_pack_index(p);
 190        if (!p->index_data)
 191                return -1;
 192
 193        err |= verify_packfile(p, &w_curs, fn, progress, base_count);
 194        unuse_pack(&w_curs);
 195
 196        return err;
 197}