lib/tdb/common/summary.c

   1  /*
   2    Trivial Database: human-readable summary code
   3    Copyright (C) Rusty Russell 2010
   4
   5    This library is free software; you can redistribute it and/or
   6    modify it under the terms of the GNU Lesser General Public
   7    License as published by the Free Software Foundation; either
   8    version 3 of the License, or (at your option) any later version.
   9
  10    This library is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13    Lesser General Public License for more details.
  14
  15    You should have received a copy of the GNU Lesser General Public
  16    License along with this library; if not, see <http://www.gnu.org/licenses/>.
  17 */
  18 #include "tdb_private.h"
  19
  20 #define SUMMARY_FORMAT \
  21         "Size of file/data: %llu/%zu\n" \
  22         "Header offset/logical size: %zu/%zu\n" \
  23         "Number of records: %zu\n" \
  24         "Incompatible hash: %s\n" \
  25         "Active/supported feature flags: 0x%08x/0x%08x\n" \
  26         "Smallest/average/largest keys: %zu/%zu/%zu\n" \
  27         "Smallest/average/largest data: %zu/%zu/%zu\n" \
  28         "Smallest/average/largest padding: %zu/%zu/%zu\n" \
  29         "Number of dead records: %zu\n" \
  30         "Smallest/average/largest dead records: %zu/%zu/%zu\n" \
  31         "Number of free records: %zu\n" \
  32         "Smallest/average/largest free records: %zu/%zu/%zu\n" \
  33         "Number of hash chains: %zu\n" \
  34         "Smallest/average/largest hash chains: %zu/%zu/%zu\n" \
  35         "Number of uncoalesced records: %zu\n" \
  36         "Smallest/average/largest uncoalesced runs: %zu/%zu/%zu\n" \
  37         "Percentage keys/data/padding/free/dead/rechdrs&tailers/hashes: %.0f/%.0f/%.0f/%.0f/%.0f/%.0f/%.0f\n"
  38
  39 /* We don't use tally module, to keep upstream happy. */
  40 struct tally {
  41         size_t min, max, total;
  42         size_t num;
  43 };
  44
  45 static void tally_init(struct tally *tally)
  46 {
  47         tally->total = 0;
  48         tally->num = 0;
  49         tally->min = tally->max = 0;
  50 }
  51
  52 static void tally_add(struct tally *tally, size_t len)
  53 {
  54         if (tally->num == 0)
  55                 tally->max = tally->min = len;
  56         else if (len > tally->max)
  57                 tally->max = len;
  58         else if (len < tally->min)
  59                 tally->min = len;
  60         tally->num++;
  61         tally->total += len;
  62 }
  63
  64 static size_t tally_mean(const struct tally *tally)
  65 {
  66         if (!tally->num)
  67                 return 0;
  68         return tally->total / tally->num;
  69 }
  70
  71 static size_t get_hash_length(struct tdb_context *tdb, unsigned int i)
  72 {
  73         tdb_off_t rec_ptr;
  74         size_t count = 0;
  75
  76         if (tdb_ofs_read(tdb, TDB_HASH_TOP(i), &rec_ptr) == -1)
  77                 return 0;
  78
  79         /* keep looking until we find the right record */
  80         while (rec_ptr) {
  81                 struct tdb_record r;
  82                 ++count;
  83                 if (tdb_rec_read(tdb, rec_ptr, &r) == -1)
  84                         return 0;
  85                 rec_ptr = r.next;
  86         }
  87         return count;
  88 }
  89
  90 _PUBLIC_ char *tdb_summary(struct tdb_context *tdb)
  91 {
  92         off_t file_size;
  93         tdb_off_t off, rec_off;
  94         struct tally freet, keys, data, dead, extra, hashval, uncoal;
  95         struct tdb_record rec;
  96         char *ret = NULL;
  97         bool locked;
  98         size_t unc = 0;
  99         int len;
 100         struct tdb_record recovery;
 101
 102         /* Read-only databases use no locking at all: it's best-effort.
 103          * We may have a write lock already, so skip that case too. */
 104         if (tdb->read_only || tdb->allrecord_lock.count != 0) {
 105                 locked = false;
 106         } else {
 107                 if (tdb_lockall_read(tdb) == -1)
 108                         return NULL;
 109                 locked = true;
 110         }
 111
 112         if (tdb_recovery_area(tdb, tdb->methods, &rec_off, &recovery) != 0) {
 113                 goto unlock;
 114         }
 115
 116         tally_init(&freet);
 117         tally_init(&keys);
 118         tally_init(&data);
 119         tally_init(&dead);
 120         tally_init(&extra);
 121         tally_init(&hashval);
 122         tally_init(&uncoal);
 123
 124         for (off = TDB_DATA_START(tdb->hash_size);
 125              off < tdb->map_size - 1;
 126              off += sizeof(rec) + rec.rec_len) {
 127                 if (tdb->methods->tdb_read(tdb, off, &rec, sizeof(rec),
 128                                            DOCONV()) == -1)
 129                         goto unlock;
 130                 switch (rec.magic) {
 131                 case TDB_MAGIC:
 132                         tally_add(&keys, rec.key_len);
 133                         tally_add(&data, rec.data_len);
 134                         tally_add(&extra, rec.rec_len - (rec.key_len
 135                                                          + rec.data_len));
 136                         if (unc > 1)
 137                                 tally_add(&uncoal, unc - 1);
 138                         unc = 0;
 139                         break;
 140                 case TDB_FREE_MAGIC:
 141                         tally_add(&freet, rec.rec_len);
 142                         unc++;
 143                         break;
 144                 /* If we crash after ftruncate, we can get zeroes or fill. */
 145                 case TDB_RECOVERY_INVALID_MAGIC:
 146                 case 0x42424242:
 147                         unc++;
 148                         /* If it's a valid recovery, we can trust rec_len. */
 149                         if (off != rec_off) {
 150                                 rec.rec_len = tdb_dead_space(tdb, off)
 151                                         - sizeof(rec);
 152                         }
 153                         /* Fall through */
 154                 case TDB_DEAD_MAGIC:
 155                         tally_add(&dead, rec.rec_len);
 156                         break;
 157                 default:
 158                         TDB_LOG((tdb, TDB_DEBUG_ERROR,
 159                                  "Unexpected record magic 0x%x at offset %u\n",
 160                                  rec.magic, off));
 161                         goto unlock;
 162                 }
 163         }
 164         if (unc > 1)
 165                 tally_add(&uncoal, unc - 1);
 166
 167         for (off = 0; off < tdb->hash_size; off++)
 168                 tally_add(&hashval, get_hash_length(tdb, off));
 169
 170         file_size = tdb->hdr_ofs + tdb->map_size;
 171
 172         len = asprintf(&ret, SUMMARY_FORMAT,
 173                  (unsigned long long)file_size, keys.total+data.total,
 174                  (size_t)tdb->hdr_ofs, (size_t)tdb->map_size,
 175                  keys.num,
 176                  (tdb->hash_fn == tdb_jenkins_hash)?"yes":"no",
 177                  (unsigned)tdb->feature_flags, TDB_SUPPORTED_FEATURE_FLAGS,
 178                  keys.min, tally_mean(&keys), keys.max,
 179                  data.min, tally_mean(&data), data.max,
 180                  extra.min, tally_mean(&extra), extra.max,
 181                  dead.num,
 182                  dead.min, tally_mean(&dead), dead.max,
 183                  freet.num,
 184                  freet.min, tally_mean(&freet), freet.max,
 185                  hashval.num,
 186                  hashval.min, tally_mean(&hashval), hashval.max,
 187                  uncoal.total,
 188                  uncoal.min, tally_mean(&uncoal), uncoal.max,
 189                  keys.total * 100.0 / file_size,
 190                  data.total * 100.0 / file_size,
 191                  extra.total * 100.0 / file_size,
 192                  freet.total * 100.0 / file_size,
 193                  dead.total * 100.0 / file_size,
 194                  (keys.num + freet.num + dead.num)
 195                  * (sizeof(struct tdb_record) + sizeof(uint32_t))
 196                  * 100.0 / file_size,
 197                  tdb->hash_size * sizeof(tdb_off_t)
 198                  * 100.0 / file_size);
 199         if (len == -1) {
 200                 goto unlock;
 201         }
 202
 203 unlock:
 204         if (locked) {
 205                 tdb_unlockall_read(tdb);
 206         }
 207         return ret;
 208 }