lib/tdb2/tdb1_freelist.c

   1  /*
   2    Unix SMB/CIFS implementation.
   3
   4    trivial database library
   5
   6    Copyright (C) Andrew Tridgell              1999-2005
   7    Copyright (C) Paul `Rusty' Russell              2000
   8    Copyright (C) Jeremy Allison                    2000-2003
   9
  10      ** NOTE! The following LGPL license applies to the tdb
  11      ** library. This does NOT imply that all of Samba is released
  12      ** under the LGPL
  13
  14    This library is free software; you can redistribute it and/or
  15    modify it under the terms of the GNU Lesser General Public
  16    License as published by the Free Software Foundation; either
  17    version 3 of the License, or (at your option) any later version.
  18
  19    This library is distributed in the hope that it will be useful,
  20    but WITHOUT ANY WARRANTY; without even the implied warranty of
  21    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  22    Lesser General Public License for more details.
  23
  24    You should have received a copy of the GNU Lesser General Public
  25    License along with this library; if not, see <http://www.gnu.org/licenses/>.
  26 */
  27
  28 #include "tdb1_private.h"
  29
  30 /* read a freelist record and check for simple errors */
  31 int tdb1_rec_free_read(struct tdb_context *tdb, tdb1_off_t off, struct tdb1_record *rec)
  32 {
  33         if (tdb->tdb1.io->tdb1_read(tdb, off, rec, sizeof(*rec),TDB1_DOCONV()) == -1)
  34                 return -1;
  35
  36         if (rec->magic == TDB1_MAGIC) {
  37                 /* this happens when a app is showdown while deleting a record - we should
  38                    not completely fail when this happens */
  39                 tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_WARNING,
  40                            "tdb1_rec_free_read non-free magic 0x%x at offset=%d - fixing\n",
  41                            rec->magic, off);
  42                 rec->magic = TDB1_FREE_MAGIC;
  43                 if (tdb->tdb1.io->tdb1_write(tdb, off, rec, sizeof(*rec)) == -1)
  44                         return -1;
  45         }
  46
  47         if (rec->magic != TDB1_FREE_MAGIC) {
  48                 tdb->last_error = tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
  49                                         "tdb1_rec_free_read bad magic 0x%x at offset=%d\n",
  50                                         rec->magic, off);
  51                 return -1;
  52         }
  53         if (tdb->tdb1.io->tdb1_oob(tdb, rec->next+sizeof(*rec), 0) != 0)
  54                 return -1;
  55         return 0;
  56 }
  57
  58
  59 /* update a record tailer (must hold allocation lock) */
  60 static int update_tailer(struct tdb_context *tdb, tdb1_off_t offset,
  61                          const struct tdb1_record *rec)
  62 {
  63         tdb1_off_t totalsize;
  64
  65         /* Offset of tailer from record header */
  66         totalsize = sizeof(*rec) + rec->rec_len;
  67         return tdb1_ofs_write(tdb, offset + totalsize - sizeof(tdb1_off_t),
  68                          &totalsize);
  69 }
  70
  71 /* Add an element into the freelist. Merge adjacent records if
  72    necessary. */
  73 int tdb1_free(struct tdb_context *tdb, tdb1_off_t offset, struct tdb1_record *rec)
  74 {
  75         /* Allocation and tailer lock */
  76         if (tdb1_lock(tdb, -1, F_WRLCK) != 0)
  77                 return -1;
  78
  79         /* set an initial tailer, so if we fail we don't leave a bogus record */
  80         if (update_tailer(tdb, offset, rec) != 0) {
  81                 tdb_logerr(tdb, tdb->last_error, TDB_LOG_ERROR,
  82                            "tdb_free: update_tailer failed!\n");
  83                 goto fail;
  84         }
  85
  86         tdb->stats.alloc_coalesce_tried++;
  87         /* Look left */
  88         if (offset - sizeof(tdb1_off_t) > TDB1_DATA_START(tdb->tdb1.header.hash_size)) {
  89                 tdb1_off_t left = offset - sizeof(tdb1_off_t);
  90                 struct tdb1_record l;
  91                 tdb1_off_t leftsize;
  92
  93                 /* Read in tailer and jump back to header */
  94                 if (tdb1_ofs_read(tdb, left, &leftsize) == -1) {
  95                         tdb_logerr(tdb, tdb->last_error, TDB_LOG_ERROR,
  96                                    "tdb1_free: left offset read failed at %u", left);
  97                         goto update;
  98                 }
  99
 100                 /* it could be uninitialised data */
 101                 if (leftsize == 0 || leftsize == TDB1_PAD_U32) {
 102                         goto update;
 103                 }
 104
 105                 left = offset - leftsize;
 106
 107                 if (leftsize > offset ||
 108                     left < TDB1_DATA_START(tdb->tdb1.header.hash_size)) {
 109                         goto update;
 110                 }
 111
 112                 /* Now read in the left record */
 113                 if (tdb->tdb1.io->tdb1_read(tdb, left, &l, sizeof(l), TDB1_DOCONV()) == -1) {
 114                         tdb_logerr(tdb, tdb->last_error, TDB_LOG_ERROR,
 115                                    "tdb1_free: left read failed at %u (%u)", left, leftsize);
 116                         goto update;
 117                 }
 118
 119                 /* If it's free, expand to include it. */
 120                 if (l.magic == TDB1_FREE_MAGIC) {
 121                         /* we now merge the new record into the left record, rather than the other
 122                            way around. This makes the operation O(1) instead of O(n). This change
 123                            prevents traverse from being O(n^2) after a lot of deletes */
 124                         l.rec_len += sizeof(*rec) + rec->rec_len;
 125                         if (tdb1_rec_write(tdb, left, &l) == -1) {
 126                                 tdb_logerr(tdb, tdb->last_error, TDB_LOG_ERROR,
 127                                            "tdb1_free: update_left failed at %u", left);
 128                                 goto fail;
 129                         }
 130                         if (update_tailer(tdb, left, &l) == -1) {
 131                                 tdb_logerr(tdb, tdb->last_error, TDB_LOG_ERROR,
 132                                            "tdb1_free: update_tailer failed at %u", offset);
 133                                 goto fail;
 134                         }
 135                         tdb->stats.alloc_coalesce_succeeded++;
 136                         tdb->stats.alloc_coalesce_num_merged++;
 137                         tdb->stats.frees++;
 138                         tdb1_unlock(tdb, -1, F_WRLCK);
 139                         return 0;
 140                 }
 141         }
 142
 143 update:
 144
 145         /* Now, prepend to free list */
 146         rec->magic = TDB1_FREE_MAGIC;
 147
 148         if (tdb1_ofs_read(tdb, TDB1_FREELIST_TOP, &rec->next) == -1 ||
 149             tdb1_rec_write(tdb, offset, rec) == -1 ||
 150             tdb1_ofs_write(tdb, TDB1_FREELIST_TOP, &offset) == -1) {
 151                 tdb_logerr(tdb, tdb->last_error, TDB_LOG_ERROR,
 152                            "tdb1_free record write failed at offset=%d",
 153                            offset);
 154                 goto fail;
 155         }
 156
 157         /* And we're done. */
 158         tdb->stats.frees++;
 159         tdb1_unlock(tdb, -1, F_WRLCK);
 160         return 0;
 161
 162  fail:
 163         tdb1_unlock(tdb, -1, F_WRLCK);
 164         return -1;
 165 }
 166
 167
 168
 169 /*
 170    the core of tdb1_allocate - called when we have decided which
 171    free list entry to use
 172
 173    Note that we try to allocate by grabbing data from the end of an existing record,
 174    not the beginning. This is so the left merge in a free is more likely to be
 175    able to free up the record without fragmentation
 176  */
 177 static tdb1_off_t tdb1_allocate_ofs(struct tdb_context *tdb,
 178                                   tdb1_len_t length, tdb1_off_t rec_ptr,
 179                                   struct tdb1_record *rec, tdb1_off_t last_ptr)
 180 {
 181 #define MIN_REC_SIZE (sizeof(struct tdb1_record) + sizeof(tdb1_off_t) + 8)
 182
 183         if (rec->rec_len < length + MIN_REC_SIZE) {
 184                 /* we have to grab the whole record */
 185
 186                 /* unlink it from the previous record */
 187                 if (tdb1_ofs_write(tdb, last_ptr, &rec->next) == -1) {
 188                         return 0;
 189                 }
 190
 191                 /* mark it not free */
 192                 rec->magic = TDB1_MAGIC;
 193                 if (tdb1_rec_write(tdb, rec_ptr, rec) == -1) {
 194                         return 0;
 195                 }
 196                 tdb->stats.allocs++;
 197                 return rec_ptr;
 198         }
 199
 200         /* we're going to just shorten the existing record */
 201         rec->rec_len -= (length + sizeof(*rec));
 202         if (tdb1_rec_write(tdb, rec_ptr, rec) == -1) {
 203                 return 0;
 204         }
 205         if (update_tailer(tdb, rec_ptr, rec) == -1) {
 206                 return 0;
 207         }
 208
 209         /* and setup the new record */
 210         rec_ptr += sizeof(*rec) + rec->rec_len;
 211
 212         memset(rec, '\0', sizeof(*rec));
 213         rec->rec_len = length;
 214         rec->magic = TDB1_MAGIC;
 215
 216         if (tdb1_rec_write(tdb, rec_ptr, rec) == -1) {
 217                 return 0;
 218         }
 219
 220         if (update_tailer(tdb, rec_ptr, rec) == -1) {
 221                 return 0;
 222         }
 223
 224         tdb->stats.allocs++;
 225         tdb->stats.alloc_leftover++;
 226         return rec_ptr;
 227 }
 228
 229 /* allocate some space from the free list. The offset returned points
 230    to a unconnected tdb1_record within the database with room for at
 231    least length bytes of total data
 232
 233    0 is returned if the space could not be allocated
 234  */
 235 tdb1_off_t tdb1_allocate(struct tdb_context *tdb, tdb1_len_t length, struct tdb1_record *rec)
 236 {
 237         tdb1_off_t rec_ptr, last_ptr, newrec_ptr;
 238         struct {
 239                 tdb1_off_t rec_ptr, last_ptr;
 240                 tdb1_len_t rec_len;
 241         } bestfit;
 242         float multiplier = 1.0;
 243
 244         if (tdb1_lock(tdb, -1, F_WRLCK) == -1)
 245                 return 0;
 246
 247         /* over-allocate to reduce fragmentation */
 248         length *= 1.25;
 249
 250         /* Extra bytes required for tailer */
 251         length += sizeof(tdb1_off_t);
 252         length = TDB1_ALIGN(length, TDB1_ALIGNMENT);
 253
 254  again:
 255         last_ptr = TDB1_FREELIST_TOP;
 256
 257         /* read in the freelist top */
 258         if (tdb1_ofs_read(tdb, TDB1_FREELIST_TOP, &rec_ptr) == -1)
 259                 goto fail;
 260
 261         bestfit.rec_ptr = 0;
 262         bestfit.last_ptr = 0;
 263         bestfit.rec_len = 0;
 264
 265         /*
 266            this is a best fit allocation strategy. Originally we used
 267            a first fit strategy, but it suffered from massive fragmentation
 268            issues when faced with a slowly increasing record size.
 269          */
 270         while (rec_ptr) {
 271                 if (tdb1_rec_free_read(tdb, rec_ptr, rec) == -1) {
 272                         goto fail;
 273                 }
 274
 275                 if (rec->rec_len >= length) {
 276                         if (bestfit.rec_ptr == 0 ||
 277                             rec->rec_len < bestfit.rec_len) {
 278                                 bestfit.rec_len = rec->rec_len;
 279                                 bestfit.rec_ptr = rec_ptr;
 280                                 bestfit.last_ptr = last_ptr;
 281                         }
 282                 }
 283
 284                 /* move to the next record */
 285                 last_ptr = rec_ptr;
 286                 rec_ptr = rec->next;
 287
 288                 /* if we've found a record that is big enough, then
 289                    stop searching if its also not too big. The
 290                    definition of 'too big' changes as we scan
 291                    through */
 292                 if (bestfit.rec_len > 0 &&
 293                     bestfit.rec_len < length * multiplier) {
 294                         break;
 295                 }
 296
 297                 /* this multiplier means we only extremely rarely
 298                    search more than 50 or so records. At 50 records we
 299                    accept records up to 11 times larger than what we
 300                    want */
 301                 multiplier *= 1.05;
 302         }
 303
 304         if (bestfit.rec_ptr != 0) {
 305                 if (tdb1_rec_free_read(tdb, bestfit.rec_ptr, rec) == -1) {
 306                         goto fail;
 307                 }
 308
 309                 newrec_ptr = tdb1_allocate_ofs(tdb, length, bestfit.rec_ptr,
 310                                               rec, bestfit.last_ptr);
 311                 tdb1_unlock(tdb, -1, F_WRLCK);
 312                 return newrec_ptr;
 313         }
 314
 315         /* we didn't find enough space. See if we can expand the
 316            database and if we can then try again */
 317         if (tdb1_expand(tdb, length + sizeof(*rec)) == 0)
 318                 goto again;
 319  fail:
 320         tdb1_unlock(tdb, -1, F_WRLCK);
 321         return 0;
 322 }