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: %u/%zu\n" \
  22         "Number of records: %zu\n" \
  23         "Incompatible hash: %s\n" \
  24         "Smallest/average/largest keys: %zu/%zu/%zu\n" \
  25         "Smallest/average/largest data: %zu/%zu/%zu\n" \
  26         "Smallest/average/largest padding: %zu/%zu/%zu\n" \
  27         "Number of dead records: %zu\n" \
  28         "Smallest/average/largest dead records: %zu/%zu/%zu\n" \
  29         "Number of free records: %zu\n" \
  30         "Smallest/average/largest free records: %zu/%zu/%zu\n" \
  31         "Number of hash chains: %zu\n" \
  32         "Smallest/average/largest hash chains: %zu/%zu/%zu\n" \
  33         "Number of uncoalesced records: %zu\n" \
  34         "Smallest/average/largest uncoalesced runs: %zu/%zu/%zu\n" \
  35         "Percentage keys/data/padding/free/dead/rechdrs&tailers/hashes: %.0f/%.0f/%.0f/%.0f/%.0f/%.0f/%.0f\n"
  36
  37 /* We don't use tally module, to keep upstream happy. */
  38 struct tally {
  39         size_t min, max, total;
  40         size_t num;
  41 };
  42
  43 static void tally_init(struct tally *tally)
  44 {
  45         tally->total = 0;
  46         tally->num = 0;
  47         tally->min = tally->max = 0;
  48 }
  49
  50 static void tally_add(struct tally *tally, size_t len)
  51 {
  52         if (tally->num == 0)
  53                 tally->max = tally->min = len;
  54         else if (len > tally->max)
  55                 tally->max = len;
  56         else if (len < tally->min)
  57                 tally->min = len;
  58         tally->num++;
  59         tally->total += len;
  60 }
  61
  62 static size_t tally_mean(const struct tally *tally)
  63 {
  64         if (!tally->num)
  65                 return 0;
  66         return tally->total / tally->num;
  67 }
  68
  69 static size_t get_hash_length(struct tdb_context *tdb, unsigned int i)
  70 {
  71         tdb_off_t rec_ptr;
  72         size_t count = 0;
  73
  74         if (tdb_ofs_read(tdb, TDB_HASH_TOP(i), &rec_ptr) == -1)
  75                 return 0;
  76
  77         /* keep looking until we find the right record */
  78         while (rec_ptr) {
  79                 struct tdb_record r;
  80                 ++count;
  81                 if (tdb_rec_read(tdb, rec_ptr, &r) == -1)
  82                         return 0;
  83                 rec_ptr = r.next;
  84         }
  85         return count;
  86 }
  87
  88 _PUBLIC_ char *tdb_summary(struct tdb_context *tdb)
  89 {
  90         tdb_off_t off, rec_off;
  91         struct tally freet, keys, data, dead, extra, hashval, uncoal;
  92         struct tdb_record rec;
  93         char *ret = NULL;
  94         bool locked;
  95         size_t len, unc = 0;
  96         struct tdb_record recovery;
  97
  98         /* Read-only databases use no locking at all: it's best-effort.
  99          * We may have a write lock already, so skip that case too. */
 100         if (tdb->read_only || tdb->allrecord_lock.count != 0) {
 101                 locked = false;
 102         } else {
 103                 if (tdb_lockall_read(tdb) == -1)
 104                         return NULL;
 105                 locked = true;
 106         }
 107
 108         if (tdb_recovery_area(tdb, tdb->methods, &rec_off, &recovery) != 0) {
 109                 goto unlock;
 110         }
 111
 112         tally_init(&freet);
 113         tally_init(&keys);
 114         tally_init(&data);
 115         tally_init(&dead);
 116         tally_init(&extra);
 117         tally_init(&hashval);
 118         tally_init(&uncoal);
 119
 120         for (off = TDB_DATA_START(tdb->hash_size);
 121              off < tdb->map_size - 1;
 122              off += sizeof(rec) + rec.rec_len) {
 123                 if (tdb->methods->tdb_read(tdb, off, &rec, sizeof(rec),
 124                                            DOCONV()) == -1)
 125                         goto unlock;
 126                 switch (rec.magic) {
 127                 case TDB_MAGIC:
 128                         tally_add(&keys, rec.key_len);
 129                         tally_add(&data, rec.data_len);
 130                         tally_add(&extra, rec.rec_len - (rec.key_len
 131                                                          + rec.data_len));
 132                         if (unc > 1)
 133                                 tally_add(&uncoal, unc - 1);
 134                         unc = 0;
 135                         break;
 136                 case TDB_FREE_MAGIC:
 137                         tally_add(&freet, rec.rec_len);
 138                         unc++;
 139                         break;
 140                 /* If we crash after ftruncate, we can get zeroes or fill. */
 141                 case TDB_RECOVERY_INVALID_MAGIC:
 142                 case 0x42424242:
 143                         unc++;
 144                         /* If it's a valid recovery, we can trust rec_len. */
 145                         if (off != rec_off) {
 146                                 rec.rec_len = tdb_dead_space(tdb, off)
 147                                         - sizeof(rec);
 148                         }
 149                         /* Fall through */
 150                 case TDB_DEAD_MAGIC:
 151                         tally_add(&dead, rec.rec_len);
 152                         break;
 153                 default:
 154                         TDB_LOG((tdb, TDB_DEBUG_ERROR,
 155                                  "Unexpected record magic 0x%x at offset %u\n",
 156                                  rec.magic, off));
 157                         goto unlock;
 158                 }
 159         }
 160         if (unc > 1)
 161                 tally_add(&uncoal, unc - 1);
 162
 163         for (off = 0; off < tdb->hash_size; off++)
 164                 tally_add(&hashval, get_hash_length(tdb, off));
 165
 166         /* 20 is max length of a %zu. */
 167         len = strlen(SUMMARY_FORMAT) + 35*20 + 1;
 168         ret = (char *)malloc(len);
 169         if (!ret)
 170                 goto unlock;
 171
 172         snprintf(ret, len, SUMMARY_FORMAT,
 173                  tdb->map_size, keys.total+data.total,
 174                  keys.num,
 175                  (tdb->hash_fn == tdb_jenkins_hash)?"yes":"no",
 176                  keys.min, tally_mean(&keys), keys.max,
 177                  data.min, tally_mean(&data), data.max,
 178                  extra.min, tally_mean(&extra), extra.max,
 179                  dead.num,
 180                  dead.min, tally_mean(&dead), dead.max,
 181                  freet.num,
 182                  freet.min, tally_mean(&freet), freet.max,
 183                  hashval.num,
 184                  hashval.min, tally_mean(&hashval), hashval.max,
 185                  uncoal.total,
 186                  uncoal.min, tally_mean(&uncoal), uncoal.max,
 187                  keys.total * 100.0 / tdb->map_size,
 188                  data.total * 100.0 / tdb->map_size,
 189                  extra.total * 100.0 / tdb->map_size,
 190                  freet.total * 100.0 / tdb->map_size,
 191                  dead.total * 100.0 / tdb->map_size,
 192                  (keys.num + freet.num + dead.num)
 193                  * (sizeof(struct tdb_record) + sizeof(uint32_t))
 194                  * 100.0 / tdb->map_size,
 195                  tdb->hash_size * sizeof(tdb_off_t)
 196                  * 100.0 / tdb->map_size);
 197
 198 unlock:
 199         if (locked) {
 200                 tdb_unlockall_read(tdb);
 201         }
 202         return ret;
 203 }