lib/tdb2/tdb1_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 "tdb1_private.h"
  19
  20 #define SUMMARY_FORMAT1 \
  21         "Size of file/data: %u/%zu\n" \
  22         "Number of records: %zu\n" \
  23         "Smallest/average/largest keys: %zu/%zu/%zu\n" \
  24         "Smallest/average/largest data: %zu/%zu/%zu\n" \
  25         "Smallest/average/largest padding: %zu/%zu/%zu\n" \
  26         "Number of dead records: %zu\n" \
  27         "Smallest/average/largest dead records: %zu/%zu/%zu\n" \
  28         "Number of free records: %zu\n" \
  29         "Smallest/average/largest free records: %zu/%zu/%zu\n" \
  30         "Number of hash chains: %zu\n" \
  31         "Smallest/average/largest hash chains: %zu/%zu/%zu\n" \
  32         "Number of uncoalesced records: %zu\n" \
  33         "Smallest/average/largest uncoalesced runs: %zu/%zu/%zu\n" \
  34         "Percentage keys/data/padding/free/dead/rechdrs&tailers/hashes: %.0f/%.0f/%.0f/%.0f/%.0f/%.0f/%.0f\n"
  35
  36 /* We don't use tally module, to keep upstream happy. */
  37 struct tally {
  38         size_t min, max, total;
  39         size_t num;
  40 };
  41
  42 static void tally1_init(struct tally *tally)
  43 {
  44         tally->total = 0;
  45         tally->num = 0;
  46         tally->min = tally->max = 0;
  47 }
  48
  49 static void tally1_add(struct tally *tally, size_t len)
  50 {
  51         if (tally->num == 0)
  52                 tally->max = tally->min = len;
  53         else if (len > tally->max)
  54                 tally->max = len;
  55         else if (len < tally->min)
  56                 tally->min = len;
  57         tally->num++;
  58         tally->total += len;
  59 }
  60
  61 static size_t tally1_mean(const struct tally *tally)
  62 {
  63         if (!tally->num)
  64                 return 0;
  65         return tally->total / tally->num;
  66 }
  67
  68 static size_t get_hash_length(struct tdb_context *tdb, unsigned int i)
  69 {
  70         tdb1_off_t rec_ptr;
  71         size_t count = 0;
  72
  73         if (tdb1_ofs_read(tdb, TDB1_HASH_TOP(i), &rec_ptr) == -1)
  74                 return 0;
  75
  76         /* keep looking until we find the right record */
  77         while (rec_ptr) {
  78                 struct tdb1_record r;
  79                 ++count;
  80                 if (tdb1_rec_read(tdb, rec_ptr, &r) == -1)
  81                         return 0;
  82                 rec_ptr = r.next;
  83         }
  84         return count;
  85 }
  86
  87 char *tdb1_summary(struct tdb_context *tdb)
  88 {
  89         tdb1_off_t off, rec_off;
  90         struct tally freet, keys, data, dead, extra, hash, uncoal;
  91         struct tdb1_record rec;
  92         char *ret = NULL;
  93         bool locked;
  94         size_t len, unc = 0;
  95         struct tdb1_record recovery;
  96
  97         /* We may have a write lock already, so don't lock. */
  98         if (tdb->file->allrecord_lock.count != 0) {
  99                 locked = false;
 100         } else {
 101                 if (tdb_lockall_read(tdb) != TDB_SUCCESS)
 102                         return NULL;
 103                 locked = true;
 104         }
 105
 106         if (tdb1_recovery_area(tdb, tdb->tdb1.io, &rec_off, &recovery) != 0) {
 107                 goto unlock;
 108         }
 109
 110         tally1_init(&freet);
 111         tally1_init(&keys);
 112         tally1_init(&data);
 113         tally1_init(&dead);
 114         tally1_init(&extra);
 115         tally1_init(&hash);
 116         tally1_init(&uncoal);
 117
 118         for (off = TDB1_DATA_START(tdb->tdb1.header.hash_size);
 119              off < tdb->file->map_size - 1;
 120              off += sizeof(rec) + rec.rec_len) {
 121                 if (tdb->tdb1.io->tdb1_read(tdb, off, &rec, sizeof(rec),
 122                                            TDB1_DOCONV()) == -1)
 123                         goto unlock;
 124                 switch (rec.magic) {
 125                 case TDB1_MAGIC:
 126                         tally1_add(&keys, rec.key_len);
 127                         tally1_add(&data, rec.data_len);
 128                         tally1_add(&extra, rec.rec_len - (rec.key_len
 129                                                          + rec.data_len));
 130                         if (unc > 1)
 131                                 tally1_add(&uncoal, unc - 1);
 132                         unc = 0;
 133                         break;
 134                 case TDB1_FREE_MAGIC:
 135                         tally1_add(&freet, rec.rec_len);
 136                         unc++;
 137                         break;
 138                 /* If we crash after ftruncate, we can get zeroes or fill. */
 139                 case TDB1_RECOVERY_INVALID_MAGIC:
 140                 case 0x42424242:
 141                         unc++;
 142                         /* If it's a valid recovery, we can trust rec_len. */
 143                         if (off != rec_off) {
 144                                 rec.rec_len = tdb1_dead_space(tdb, off)
 145                                         - sizeof(rec);
 146                         }
 147                         /* Fall through */
 148                 case TDB1_DEAD_MAGIC:
 149                         tally1_add(&dead, rec.rec_len);
 150                         break;
 151                 default:
 152                         tdb->last_error = tdb_logerr(tdb, TDB_ERR_CORRUPT,
 153                                                 TDB_LOG_ERROR,
 154                                                 "Unexpected record magic 0x%x"
 155                                                 " at offset %d",
 156                                                 rec.magic, off);
 157                         goto unlock;
 158                 }
 159         }
 160         if (unc > 1)
 161                 tally1_add(&uncoal, unc - 1);
 162
 163         for (off = 0; off < tdb->tdb1.header.hash_size; off++)
 164                 tally1_add(&hash, get_hash_length(tdb, off));
 165
 166         /* 20 is max length of a %zu. */
 167         len = strlen(SUMMARY_FORMAT1) + 35*20 + 1;
 168         ret = (char *)malloc(len);
 169         if (!ret)
 170                 goto unlock;
 171
 172         snprintf(ret, len, SUMMARY_FORMAT1,
 173                  (tdb1_len_t)tdb->file->map_size, keys.total+data.total,
 174                  keys.num,
 175                  keys.min, tally1_mean(&keys), keys.max,
 176                  data.min, tally1_mean(&data), data.max,
 177                  extra.min, tally1_mean(&extra), extra.max,
 178                  dead.num,
 179                  dead.min, tally1_mean(&dead), dead.max,
 180                  freet.num,
 181                  freet.min, tally1_mean(&freet), freet.max,
 182                  hash.num,
 183                  hash.min, tally1_mean(&hash), hash.max,
 184                  uncoal.total,
 185                  uncoal.min, tally1_mean(&uncoal), uncoal.max,
 186                  keys.total * 100.0 / tdb->file->map_size,
 187                  data.total * 100.0 / tdb->file->map_size,
 188                  extra.total * 100.0 / tdb->file->map_size,
 189                  freet.total * 100.0 / tdb->file->map_size,
 190                  dead.total * 100.0 / tdb->file->map_size,
 191                  (keys.num + freet.num + dead.num)
 192                  * (sizeof(struct tdb1_record) + sizeof(uint32_t))
 193                  * 100.0 / tdb->file->map_size,
 194                  tdb->tdb1.header.hash_size * sizeof(tdb1_off_t)
 195                  * 100.0 / (tdb1_len_t)tdb->file->map_size);
 196
 197 unlock:
 198         if (locked) {
 199                 tdb_unlockall_read(tdb);
 200         }
 201         return ret;
 202 }