src/db.c

   1 #include "redis.h"
   2
   3 #include <signal.h>
   4
   5 /*-----------------------------------------------------------------------------
   6  * C-level DB API
   7  *----------------------------------------------------------------------------*/
   8
   9 robj *lookupKey(redisDb *db, robj *key) {
  10     dictEntry *de = dictFind(db->dict,key->ptr);
  11     if (de) {
  12         robj *val = dictGetEntryVal(de);
  13
  14         /* Update the access time for the aging algorithm.
  15          * Don't do it if we have a saving child, as this will trigger
  16          * a copy on write madness. */
  17         if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1)
  18             val->lru = server.lruclock;
  19
  20         if (server.vm_enabled) {
  21             if (val->storage == REDIS_VM_MEMORY ||
  22                 val->storage == REDIS_VM_SWAPPING)
  23             {
  24                 /* If we were swapping the object out, cancel the operation */
  25                 if (val->storage == REDIS_VM_SWAPPING)
  26                     vmCancelThreadedIOJob(val);
  27             } else {
  28                 int notify = (val->storage == REDIS_VM_LOADING);
  29
  30                 /* Our value was swapped on disk. Bring it at home. */
  31                 redisAssert(val->type == REDIS_VMPOINTER);
  32                 val = vmLoadObject(val);
  33                 dictGetEntryVal(de) = val;
  34
  35                 /* Clients blocked by the VM subsystem may be waiting for
  36                  * this key... */
  37                 if (notify) handleClientsBlockedOnSwappedKey(db,key);
  38             }
  39         }
  40         server.stat_keyspace_hits++;
  41         return val;
  42     } else {
  43         server.stat_keyspace_misses++;
  44         return NULL;
  45     }
  46 }
  47
  48 robj *lookupKeyRead(redisDb *db, robj *key) {
  49     expireIfNeeded(db,key);
  50     return lookupKey(db,key);
  51 }
  52
  53 robj *lookupKeyWrite(redisDb *db, robj *key) {
  54     expireIfNeeded(db,key);
  55     return lookupKey(db,key);
  56 }
  57
  58 robj *lookupKeyReadOrReply(redisClient *c, robj *key, robj *reply) {
  59     robj *o = lookupKeyRead(c->db, key);
  60     if (!o) addReply(c,reply);
  61     return o;
  62 }
  63
  64 robj *lookupKeyWriteOrReply(redisClient *c, robj *key, robj *reply) {
  65     robj *o = lookupKeyWrite(c->db, key);
  66     if (!o) addReply(c,reply);
  67     return o;
  68 }
  69
  70 /* Add the key to the DB. If the key already exists REDIS_ERR is returned,
  71  * otherwise REDIS_OK is returned, and the caller should increment the
  72  * refcount of 'val'. */
  73 int dbAdd(redisDb *db, robj *key, robj *val) {
  74     /* Perform a lookup before adding the key, as we need to copy the
  75      * key value. */
  76     if (dictFind(db->dict, key->ptr) != NULL) {
  77         return REDIS_ERR;
  78     } else {
  79         sds copy = sdsdup(key->ptr);
  80         dictAdd(db->dict, copy, val);
  81         return REDIS_OK;
  82     }
  83 }
  84
  85 /* If the key does not exist, this is just like dbAdd(). Otherwise
  86  * the value associated to the key is replaced with the new one.
  87  *
  88  * On update (key already existed) 0 is returned. Otherwise 1. */
  89 int dbReplace(redisDb *db, robj *key, robj *val) {
  90     if (dictFind(db->dict,key->ptr) == NULL) {
  91         sds copy = sdsdup(key->ptr);
  92         dictAdd(db->dict, copy, val);
  93         return 1;
  94     } else {
  95         dictReplace(db->dict, key->ptr, val);
  96         return 0;
  97     }
  98 }
  99
 100 int dbExists(redisDb *db, robj *key) {
 101     return dictFind(db->dict,key->ptr) != NULL;
 102 }
 103
 104 /* Return a random key, in form of a Redis object.
 105  * If there are no keys, NULL is returned.
 106  *
 107  * The function makes sure to return keys not already expired. */
 108 robj *dbRandomKey(redisDb *db) {
 109     struct dictEntry *de;
 110
 111     while(1) {
 112         sds key;
 113         robj *keyobj;
 114
 115         de = dictGetRandomKey(db->dict);
 116         if (de == NULL) return NULL;
 117
 118         key = dictGetEntryKey(de);
 119         keyobj = createStringObject(key,sdslen(key));
 120         if (dictFind(db->expires,key)) {
 121             if (expireIfNeeded(db,keyobj)) {
 122                 decrRefCount(keyobj);
 123                 continue; /* search for another key. This expired. */
 124             }
 125         }
 126         return keyobj;
 127     }
 128 }
 129
 130 /* Delete a key, value, and associated expiration entry if any, from the DB */
 131 int dbDelete(redisDb *db, robj *key) {
 132     /* If VM is enabled make sure to awake waiting clients for this key:
 133      * deleting the key will kill the I/O thread bringing the key from swap
 134      * to memory, so the client will never be notified and unblocked if we
 135      * don't do it now. */
 136     if (server.vm_enabled) handleClientsBlockedOnSwappedKey(db,key);
 137     /* Deleting an entry from the expires dict will not free the sds of
 138      * the key, because it is shared with the main dictionary. */
 139     if (dictSize(db->expires) > 0) dictDelete(db->expires,key->ptr);
 140     return dictDelete(db->dict,key->ptr) == DICT_OK;
 141 }
 142
 143 /* Empty the whole database */
 144 long long emptyDb() {
 145     int j;
 146     long long removed = 0;
 147
 148     for (j = 0; j < server.dbnum; j++) {
 149         removed += dictSize(server.db[j].dict);
 150         dictEmpty(server.db[j].dict);
 151         dictEmpty(server.db[j].expires);
 152     }
 153     return removed;
 154 }
 155
 156 int selectDb(redisClient *c, int id) {
 157     if (id < 0 || id >= server.dbnum)
 158         return REDIS_ERR;
 159     c->db = &server.db[id];
 160     return REDIS_OK;
 161 }
 162
 163 /*-----------------------------------------------------------------------------
 164  * Type agnostic commands operating on the key space
 165  *----------------------------------------------------------------------------*/
 166
 167 void flushdbCommand(redisClient *c) {
 168     server.dirty += dictSize(c->db->dict);
 169     touchWatchedKeysOnFlush(c->db->id);
 170     dictEmpty(c->db->dict);
 171     dictEmpty(c->db->expires);
 172     addReply(c,shared.ok);
 173 }
 174
 175 void flushallCommand(redisClient *c) {
 176     touchWatchedKeysOnFlush(-1);
 177     server.dirty += emptyDb();
 178     addReply(c,shared.ok);
 179     if (server.bgsavechildpid != -1) {
 180         kill(server.bgsavechildpid,SIGKILL);
 181         rdbRemoveTempFile(server.bgsavechildpid);
 182     }
 183     rdbSave(server.dbfilename);
 184     server.dirty++;
 185 }
 186
 187 void delCommand(redisClient *c) {
 188     int deleted = 0, j;
 189
 190     for (j = 1; j < c->argc; j++) {
 191         if (dbDelete(c->db,c->argv[j])) {
 192             touchWatchedKey(c->db,c->argv[j]);
 193             server.dirty++;
 194             deleted++;
 195         }
 196     }
 197     addReplyLongLong(c,deleted);
 198 }
 199
 200 void existsCommand(redisClient *c) {
 201     expireIfNeeded(c->db,c->argv[1]);
 202     if (dbExists(c->db,c->argv[1])) {
 203         addReply(c, shared.cone);
 204     } else {
 205         addReply(c, shared.czero);
 206     }
 207 }
 208
 209 void selectCommand(redisClient *c) {
 210     int id = atoi(c->argv[1]->ptr);
 211
 212     if (selectDb(c,id) == REDIS_ERR) {
 213         addReplyError(c,"invalid DB index");
 214     } else {
 215         addReply(c,shared.ok);
 216     }
 217 }
 218
 219 void randomkeyCommand(redisClient *c) {
 220     robj *key;
 221
 222     if ((key = dbRandomKey(c->db)) == NULL) {
 223         addReply(c,shared.nullbulk);
 224         return;
 225     }
 226
 227     addReplyBulk(c,key);
 228     decrRefCount(key);
 229 }
 230
 231 void keysCommand(redisClient *c) {
 232     dictIterator *di;
 233     dictEntry *de;
 234     sds pattern = c->argv[1]->ptr;
 235     int plen = sdslen(pattern), allkeys;
 236     unsigned long numkeys = 0;
 237     void *replylen = addDeferredMultiBulkLength(c);
 238
 239     di = dictGetIterator(c->db->dict);
 240     allkeys = (pattern[0] == '*' && pattern[1] == '\0');
 241     while((de = dictNext(di)) != NULL) {
 242         sds key = dictGetEntryKey(de);
 243         robj *keyobj;
 244
 245         if (allkeys || stringmatchlen(pattern,plen,key,sdslen(key),0)) {
 246             keyobj = createStringObject(key,sdslen(key));
 247             if (expireIfNeeded(c->db,keyobj) == 0) {
 248                 addReplyBulk(c,keyobj);
 249                 numkeys++;
 250             }
 251             decrRefCount(keyobj);
 252         }
 253     }
 254     dictReleaseIterator(di);
 255     setDeferredMultiBulkLength(c,replylen,numkeys);
 256 }
 257
 258 void dbsizeCommand(redisClient *c) {
 259     addReplyLongLong(c,dictSize(c->db->dict));
 260 }
 261
 262 void lastsaveCommand(redisClient *c) {
 263     addReplyLongLong(c,server.lastsave);
 264 }
 265
 266 void typeCommand(redisClient *c) {
 267     robj *o;
 268     char *type;
 269
 270     o = lookupKeyRead(c->db,c->argv[1]);
 271     if (o == NULL) {
 272         type = "none";
 273     } else {
 274         switch(o->type) {
 275         case REDIS_STRING: type = "string"; break;
 276         case REDIS_LIST: type = "list"; break;
 277         case REDIS_SET: type = "set"; break;
 278         case REDIS_ZSET: type = "zset"; break;
 279         case REDIS_HASH: type = "hash"; break;
 280         default: type = "unknown"; break;
 281         }
 282     }
 283     addReplyStatus(c,type);
 284 }
 285
 286 void saveCommand(redisClient *c) {
 287     if (server.bgsavechildpid != -1) {
 288         addReplyError(c,"Background save already in progress");
 289         return;
 290     }
 291     if (rdbSave(server.dbfilename) == REDIS_OK) {
 292         addReply(c,shared.ok);
 293     } else {
 294         addReply(c,shared.err);
 295     }
 296 }
 297
 298 void bgsaveCommand(redisClient *c) {
 299     if (server.bgsavechildpid != -1) {
 300         addReplyError(c,"Background save already in progress");
 301         return;
 302     }
 303     if (rdbSaveBackground(server.dbfilename) == REDIS_OK) {
 304         addReplyStatus(c,"Background saving started");
 305     } else {
 306         addReply(c,shared.err);
 307     }
 308 }
 309
 310 void shutdownCommand(redisClient *c) {
 311     if (prepareForShutdown() == REDIS_OK)
 312         exit(0);
 313     addReplyError(c,"Errors trying to SHUTDOWN. Check logs.");
 314 }
 315
 316 void renameGenericCommand(redisClient *c, int nx) {
 317     robj *o;
 318
 319     /* To use the same key as src and dst is probably an error */
 320     if (sdscmp(c->argv[1]->ptr,c->argv[2]->ptr) == 0) {
 321         addReply(c,shared.sameobjecterr);
 322         return;
 323     }
 324
 325     if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr)) == NULL)
 326         return;
 327
 328     incrRefCount(o);
 329     if (dbAdd(c->db,c->argv[2],o) == REDIS_ERR) {
 330         if (nx) {
 331             decrRefCount(o);
 332             addReply(c,shared.czero);
 333             return;
 334         }
 335         dbReplace(c->db,c->argv[2],o);
 336     }
 337     dbDelete(c->db,c->argv[1]);
 338     touchWatchedKey(c->db,c->argv[1]);
 339     touchWatchedKey(c->db,c->argv[2]);
 340     server.dirty++;
 341     addReply(c,nx ? shared.cone : shared.ok);
 342 }
 343
 344 void renameCommand(redisClient *c) {
 345     renameGenericCommand(c,0);
 346 }
 347
 348 void renamenxCommand(redisClient *c) {
 349     renameGenericCommand(c,1);
 350 }
 351
 352 void moveCommand(redisClient *c) {
 353     robj *o;
 354     redisDb *src, *dst;
 355     int srcid;
 356
 357     /* Obtain source and target DB pointers */
 358     src = c->db;
 359     srcid = c->db->id;
 360     if (selectDb(c,atoi(c->argv[2]->ptr)) == REDIS_ERR) {
 361         addReply(c,shared.outofrangeerr);
 362         return;
 363     }
 364     dst = c->db;
 365     selectDb(c,srcid); /* Back to the source DB */
 366
 367     /* If the user is moving using as target the same
 368      * DB as the source DB it is probably an error. */
 369     if (src == dst) {
 370         addReply(c,shared.sameobjecterr);
 371         return;
 372     }
 373
 374     /* Check if the element exists and get a reference */
 375     o = lookupKeyWrite(c->db,c->argv[1]);
 376     if (!o) {
 377         addReply(c,shared.czero);
 378         return;
 379     }
 380
 381     /* Try to add the element to the target DB */
 382     if (dbAdd(dst,c->argv[1],o) == REDIS_ERR) {
 383         addReply(c,shared.czero);
 384         return;
 385     }
 386     incrRefCount(o);
 387
 388     /* OK! key moved, free the entry in the source DB */
 389     dbDelete(src,c->argv[1]);
 390     server.dirty++;
 391     addReply(c,shared.cone);
 392 }
 393
 394 /*-----------------------------------------------------------------------------
 395  * Expires API
 396  *----------------------------------------------------------------------------*/
 397
 398 int removeExpire(redisDb *db, robj *key) {
 399     /* An expire may only be removed if there is a corresponding entry in the
 400      * main dict. Otherwise, the key will never be freed. */
 401     redisAssert(dictFind(db->dict,key->ptr) != NULL);
 402     return dictDelete(db->expires,key->ptr) == DICT_OK;
 403 }
 404
 405 void setExpire(redisDb *db, robj *key, time_t when) {
 406     dictEntry *de;
 407
 408     /* Reuse the sds from the main dict in the expire dict */
 409     de = dictFind(db->dict,key->ptr);
 410     redisAssert(de != NULL);
 411     dictReplace(db->expires,dictGetEntryKey(de),(void*)when);
 412 }
 413
 414 /* Return the expire time of the specified key, or -1 if no expire
 415  * is associated with this key (i.e. the key is non volatile) */
 416 time_t getExpire(redisDb *db, robj *key) {
 417     dictEntry *de;
 418
 419     /* No expire? return ASAP */
 420     if (dictSize(db->expires) == 0 ||
 421        (de = dictFind(db->expires,key->ptr)) == NULL) return -1;
 422
 423     /* The entry was found in the expire dict, this means it should also
 424      * be present in the main dict (safety check). */
 425     redisAssert(dictFind(db->dict,key->ptr) != NULL);
 426     return (time_t) dictGetEntryVal(de);
 427 }
 428
 429 /* Propagate expires into slaves and the AOF file.
 430  * When a key expires in the master, a DEL operation for this key is sent
 431  * to all the slaves and the AOF file if enabled.
 432  *
 433  * This way the key expiry is centralized in one place, and since both
 434  * AOF and the master->slave link guarantee operation ordering, everything
 435  * will be consistent even if we allow write operations against expiring
 436  * keys. */
 437 void propagateExpire(redisDb *db, robj *key) {
 438     struct redisCommand *cmd;
 439     robj *argv[2];
 440
 441     cmd = lookupCommand("del");
 442     argv[0] = createStringObject("DEL",3);
 443     argv[1] = key;
 444     incrRefCount(key);
 445
 446     if (server.appendonly)
 447         feedAppendOnlyFile(cmd,db->id,argv,2);
 448     if (listLength(server.slaves))
 449         replicationFeedSlaves(server.slaves,db->id,argv,2);
 450
 451     decrRefCount(argv[0]);
 452     decrRefCount(argv[1]);
 453 }
 454
 455 int expireIfNeeded(redisDb *db, robj *key) {
 456     time_t when = getExpire(db,key);
 457
 458     /* If we are running in the context of a slave, return ASAP:
 459      * the slave key expiration is controlled by the master that will
 460      * send us synthesized DEL operations for expired keys.
 461      *
 462      * Still we try to return the right information to the caller,
 463      * that is, 0 if we think the key should be still valid, 1 if
 464      * we think the key is expired at this time. */
 465     if (server.masterhost != NULL) {
 466         return time(NULL) > when;
 467     }
 468
 469     if (when < 0) return 0;
 470
 471     /* Return when this key has not expired */
 472     if (time(NULL) <= when) return 0;
 473
 474     /* Delete the key */
 475     server.stat_expiredkeys++;
 476     propagateExpire(db,key);
 477     return dbDelete(db,key);
 478 }
 479
 480 /*-----------------------------------------------------------------------------
 481  * Expires Commands
 482  *----------------------------------------------------------------------------*/
 483
 484 void expireGenericCommand(redisClient *c, robj *key, robj *param, long offset) {
 485     dictEntry *de;
 486     long seconds;
 487
 488     if (getLongFromObjectOrReply(c, param, &seconds, NULL) != REDIS_OK) return;
 489
 490     seconds -= offset;
 491
 492     de = dictFind(c->db->dict,key->ptr);
 493     if (de == NULL) {
 494         addReply(c,shared.czero);
 495         return;
 496     }
 497     if (seconds <= 0) {
 498         if (dbDelete(c->db,key)) server.dirty++;
 499         addReply(c, shared.cone);
 500         touchWatchedKey(c->db,key);
 501         return;
 502     } else {
 503         time_t when = time(NULL)+seconds;
 504         setExpire(c->db,key,when);
 505         addReply(c,shared.cone);
 506         touchWatchedKey(c->db,key);
 507         server.dirty++;
 508         return;
 509     }
 510 }
 511
 512 void expireCommand(redisClient *c) {
 513     expireGenericCommand(c,c->argv[1],c->argv[2],0);
 514 }
 515
 516 void expireatCommand(redisClient *c) {
 517     expireGenericCommand(c,c->argv[1],c->argv[2],time(NULL));
 518 }
 519
 520 void ttlCommand(redisClient *c) {
 521     time_t expire, ttl = -1;
 522
 523     expire = getExpire(c->db,c->argv[1]);
 524     if (expire != -1) {
 525         ttl = (expire-time(NULL));
 526         if (ttl < 0) ttl = -1;
 527     }
 528     addReplyLongLong(c,(long long)ttl);
 529 }
 530
 531 void persistCommand(redisClient *c) {
 532     dictEntry *de;
 533
 534     de = dictFind(c->db->dict,c->argv[1]->ptr);
 535     if (de == NULL) {
 536         addReply(c,shared.czero);
 537     } else {
 538         if (removeExpire(c->db,c->argv[1])) {
 539             addReply(c,shared.cone);
 540             server.dirty++;
 541         } else {
 542             addReply(c,shared.czero);
 543         }
 544     }
 545 }