src/redis.c

   1 /*
   2  * Copyright (c) 2009-2010, Salvatore Sanfilippo <antirez at gmail dot com>
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions are met:
   7  *
   8  *   * Redistributions of source code must retain the above copyright notice,
   9  *     this list of conditions and the following disclaimer.
  10  *   * Redistributions in binary form must reproduce the above copyright
  11  *     notice, this list of conditions and the following disclaimer in the
  12  *     documentation and/or other materials provided with the distribution.
  13  *   * Neither the name of Redis nor the names of its contributors may be used
  14  *     to endorse or promote products derived from this software without
  15  *     specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  18  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  21  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  22  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  23  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  24  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  25  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  26  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  27  * POSSIBILITY OF SUCH DAMAGE.
  28  */
  29
  30 #include "redis.h"
  31 #include "slowlog.h"
  32 #include "bio.h"
  33
  34 #ifdef HAVE_BACKTRACE
  35 #include <execinfo.h>
  36 #include <ucontext.h>
  37 #endif /* HAVE_BACKTRACE */
  38
  39 #include <time.h>
  40 #include <signal.h>
  41 #include <sys/wait.h>
  42 #include <errno.h>
  43 #include <assert.h>
  44 #include <ctype.h>
  45 #include <stdarg.h>
  46 #include <arpa/inet.h>
  47 #include <sys/stat.h>
  48 #include <fcntl.h>
  49 #include <sys/time.h>
  50 #include <sys/resource.h>
  51 #include <sys/uio.h>
  52 #include <limits.h>
  53 #include <float.h>
  54 #include <math.h>
  55 #include <sys/resource.h>
  56
  57 /* Our shared "common" objects */
  58
  59 struct sharedObjectsStruct shared;
  60
  61 /* Global vars that are actually used as constants. The following double
  62  * values are used for double on-disk serialization, and are initialized
  63  * at runtime to avoid strange compiler optimizations. */
  64
  65 double R_Zero, R_PosInf, R_NegInf, R_Nan;
  66
  67 /*================================= Globals ================================= */
  68
  69 /* Global vars */
  70 struct redisServer server; /* server global state */
  71 struct redisCommand *commandTable;
  72
  73 /* Our command table. Command flags are expressed using strings where every
  74  * character represents a flag. Later the populateCommandTable() function will
  75  * take care of populating the real 'flags' field using this characters.
  76  *
  77  * This is the meaning of the flags:
  78  *
  79  * w: write command (may modify the key space).
  80  * r: read command  (will never modify the key space).
  81  * m: may increase memory usage once called. Don't allow if out of memory.
  82  * a: admin command, like SAVE or SHUTDOWN.
  83  * p: Pub/Sub related command.
  84  * f: force replication of this command, regarless of server.dirty.
  85  * s: command not allowed in scripts.
  86  * r: random command. Command is not deterministic, that is, the same command
  87  *    with the same arguments, with the same key space, may have different
  88  *    results. For instance SPOP and RANDOMKEY are two random commands. */
  89 struct redisCommand redisCommandTable[] = {
  90     {"get",getCommand,2,"r",0,NULL,1,1,1,0,0},
  91     {"set",setCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
  92     {"setnx",setnxCommand,3,"wm",0,noPreloadGetKeys,1,1,1,0,0},
  93     {"setex",setexCommand,4,"wm",0,noPreloadGetKeys,2,2,1,0,0},
  94     {"append",appendCommand,3,"wm",0,NULL,1,1,1,0,0},
  95     {"strlen",strlenCommand,2,"r",0,NULL,1,1,1,0,0},
  96     {"del",delCommand,-2,"w",0,noPreloadGetKeys,1,-1,1,0,0},
  97     {"exists",existsCommand,2,"r",0,NULL,1,1,1,0,0},
  98     {"setbit",setbitCommand,4,"wm",0,NULL,1,1,1,0,0},
  99     {"getbit",getbitCommand,3,"r",0,NULL,1,1,1,0,0},
 100     {"setrange",setrangeCommand,4,"wm",0,NULL,1,1,1,0,0},
 101     {"getrange",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
 102     {"substr",getrangeCommand,4,"r",0,NULL,1,1,1,0,0},
 103     {"incr",incrCommand,2,"wm",0,NULL,1,1,1,0,0},
 104     {"decr",decrCommand,2,"wm",0,NULL,1,1,1,0,0},
 105     {"mget",mgetCommand,-2,"r",0,NULL,1,-1,1,0,0},
 106     {"rpush",rpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
 107     {"lpush",lpushCommand,-3,"wm",0,NULL,1,1,1,0,0},
 108     {"rpushx",rpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
 109     {"lpushx",lpushxCommand,3,"wm",0,NULL,1,1,1,0,0},
 110     {"linsert",linsertCommand,5,"wm",0,NULL,1,1,1,0,0},
 111     {"rpop",rpopCommand,2,"w",0,NULL,1,1,1,0,0},
 112     {"lpop",lpopCommand,2,"w",0,NULL,1,1,1,0,0},
 113     {"brpop",brpopCommand,-3,"w",0,NULL,1,1,1,0,0},
 114     {"brpoplpush",brpoplpushCommand,4,"wm",0,NULL,1,2,1,0,0},
 115     {"blpop",blpopCommand,-3,"w",0,NULL,1,-2,1,0,0},
 116     {"llen",llenCommand,2,"r",0,NULL,1,1,1,0,0},
 117     {"lindex",lindexCommand,3,"r",0,NULL,1,1,1,0,0},
 118     {"lset",lsetCommand,4,"wm",0,NULL,1,1,1,0,0},
 119     {"lrange",lrangeCommand,4,"r",0,NULL,1,1,1,0,0},
 120     {"ltrim",ltrimCommand,4,"w",0,NULL,1,1,1,0,0},
 121     {"lrem",lremCommand,4,"w",0,NULL,1,1,1,0,0},
 122     {"rpoplpush",rpoplpushCommand,3,"wm",0,NULL,1,2,1,0,0},
 123     {"sadd",saddCommand,-3,"wm",0,NULL,1,1,1,0,0},
 124     {"srem",sremCommand,-3,"w",0,NULL,1,1,1,0,0},
 125     {"smove",smoveCommand,4,"w",0,NULL,1,2,1,0,0},
 126     {"sismember",sismemberCommand,3,"r",0,NULL,1,1,1,0,0},
 127     {"scard",scardCommand,2,"r",0,NULL,1,1,1,0,0},
 128     {"spop",spopCommand,2,"wRs",0,NULL,1,1,1,0,0},
 129     {"srandmember",srandmemberCommand,2,"rR",0,NULL,1,1,1,0,0},
 130     {"sinter",sinterCommand,-2,"r",0,NULL,1,-1,1,0,0},
 131     {"sinterstore",sinterstoreCommand,-3,"wm",0,NULL,2,-1,1,0,0},
 132     {"sunion",sunionCommand,-2,"r",0,NULL,1,-1,1,0,0},
 133     {"sunionstore",sunionstoreCommand,-3,"wm",0,NULL,2,-1,1,0,0},
 134     {"sdiff",sdiffCommand,-2,"r",0,NULL,1,-1,1,0,0},
 135     {"sdiffstore",sdiffstoreCommand,-3,"wm",0,NULL,2,-1,1,0,0},
 136     {"smembers",sinterCommand,2,"r",0,NULL,1,1,1,0,0},
 137     {"zadd",zaddCommand,-4,"wm",0,NULL,1,1,1,0,0},
 138     {"zincrby",zincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
 139     {"zrem",zremCommand,-3,"w",0,NULL,1,1,1,0,0},
 140     {"zremrangebyscore",zremrangebyscoreCommand,4,"w",0,NULL,1,1,1,0,0},
 141     {"zremrangebyrank",zremrangebyrankCommand,4,"w",0,NULL,1,1,1,0,0},
 142     {"zunionstore",zunionstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
 143     {"zinterstore",zinterstoreCommand,-4,"wm",0,zunionInterGetKeys,0,0,0,0,0},
 144     {"zrange",zrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
 145     {"zrangebyscore",zrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
 146     {"zrevrangebyscore",zrevrangebyscoreCommand,-4,"r",0,NULL,1,1,1,0,0},
 147     {"zcount",zcountCommand,4,"r",0,NULL,1,1,1,0,0},
 148     {"zrevrange",zrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
 149     {"zcard",zcardCommand,2,"r",0,NULL,1,1,1,0,0},
 150     {"zscore",zscoreCommand,3,"r",0,NULL,1,1,1,0,0},
 151     {"zrank",zrankCommand,3,"r",0,NULL,1,1,1,0,0},
 152     {"zrevrank",zrevrankCommand,3,"r",0,NULL,1,1,1,0,0},
 153     {"hset",hsetCommand,4,"wm",0,NULL,1,1,1,0,0},
 154     {"hsetnx",hsetnxCommand,4,"wm",0,NULL,1,1,1,0,0},
 155     {"hget",hgetCommand,3,"r",0,NULL,1,1,1,0,0},
 156     {"hmset",hmsetCommand,-4,"wm",0,NULL,1,1,1,0,0},
 157     {"hmget",hmgetCommand,-3,"r",0,NULL,1,1,1,0,0},
 158     {"hincrby",hincrbyCommand,4,"wm",0,NULL,1,1,1,0,0},
 159     {"hdel",hdelCommand,-3,"w",0,NULL,1,1,1,0,0},
 160     {"hlen",hlenCommand,2,"r",0,NULL,1,1,1,0,0},
 161     {"hkeys",hkeysCommand,2,"r",0,NULL,1,1,1,0,0},
 162     {"hvals",hvalsCommand,2,"r",0,NULL,1,1,1,0,0},
 163     {"hgetall",hgetallCommand,2,"r",0,NULL,1,1,1,0,0},
 164     {"hexists",hexistsCommand,3,"r",0,NULL,1,1,1,0,0},
 165     {"incrby",incrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
 166     {"decrby",decrbyCommand,3,"wm",0,NULL,1,1,1,0,0},
 167     {"getset",getsetCommand,3,"wm",0,NULL,1,1,1,0,0},
 168     {"mset",msetCommand,-3,"wm",0,NULL,1,-1,2,0,0},
 169     {"msetnx",msetnxCommand,-3,"wm",0,NULL,1,-1,2,0,0},
 170     {"randomkey",randomkeyCommand,1,"rR",0,NULL,0,0,0,0,0},
 171     {"select",selectCommand,2,"r",0,NULL,0,0,0,0,0},
 172     {"move",moveCommand,3,"w",0,NULL,1,1,1,0,0},
 173     {"rename",renameCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
 174     {"renamenx",renamenxCommand,3,"w",0,renameGetKeys,1,2,1,0,0},
 175     {"expire",expireCommand,3,"w",0,NULL,1,1,1,0,0},
 176     {"expireat",expireatCommand,3,"w",0,NULL,1,1,1,0,0},
 177     {"keys",keysCommand,2,"r",0,NULL,0,0,0,0,0},
 178     {"dbsize",dbsizeCommand,1,"r",0,NULL,0,0,0,0,0},
 179     {"auth",authCommand,2,"r",0,NULL,0,0,0,0,0},
 180     {"ping",pingCommand,1,"r",0,NULL,0,0,0,0,0},
 181     {"echo",echoCommand,2,"r",0,NULL,0,0,0,0,0},
 182     {"save",saveCommand,1,"ar",0,NULL,0,0,0,0,0},
 183     {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
 184     {"bgrewriteaof",bgrewriteaofCommand,1,"ar",0,NULL,0,0,0,0,0},
 185     {"shutdown",shutdownCommand,1,"ar",0,NULL,0,0,0,0,0},
 186     {"lastsave",lastsaveCommand,1,"r",0,NULL,0,0,0,0,0},
 187     {"type",typeCommand,2,"r",0,NULL,1,1,1,0,0},
 188     {"multi",multiCommand,1,"rs",0,NULL,0,0,0,0,0},
 189     {"exec",execCommand,1,"wms",0,NULL,0,0,0,0,0},
 190     {"discard",discardCommand,1,"rs",0,NULL,0,0,0,0,0},
 191     {"sync",syncCommand,1,"ars",0,NULL,0,0,0,0,0},
 192     {"flushdb",flushdbCommand,1,"w",0,NULL,0,0,0,0,0},
 193     {"flushall",flushallCommand,1,"w",0,NULL,0,0,0,0,0},
 194     {"sort",sortCommand,-2,"wm",0,NULL,1,1,1,0,0},
 195     {"info",infoCommand,-1,"r",0,NULL,0,0,0,0,0},
 196     {"monitor",monitorCommand,1,"ars",0,NULL,0,0,0,0,0},
 197     {"ttl",ttlCommand,2,"r",0,NULL,1,1,1,0,0},
 198     {"persist",persistCommand,2,"w",0,NULL,1,1,1,0,0},
 199     {"slaveof",slaveofCommand,3,"aws",0,NULL,0,0,0,0,0},
 200     {"debug",debugCommand,-2,"aw",0,NULL,0,0,0,0,0},
 201     {"config",configCommand,-2,"ar",0,NULL,0,0,0,0,0},
 202     {"subscribe",subscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
 203     {"unsubscribe",unsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
 204     {"psubscribe",psubscribeCommand,-2,"rps",0,NULL,0,0,0,0,0},
 205     {"punsubscribe",punsubscribeCommand,-1,"rps",0,NULL,0,0,0,0,0},
 206     {"publish",publishCommand,3,"rpf",0,NULL,0,0,0,0,0},
 207     {"watch",watchCommand,-2,"rs",0,noPreloadGetKeys,1,-1,1,0,0},
 208     {"unwatch",unwatchCommand,1,"rs",0,NULL,0,0,0,0,0},
 209     {"cluster",clusterCommand,-2,"ar",0,NULL,0,0,0,0,0},
 210     {"restore",restoreCommand,4,"awm",0,NULL,1,1,1,0,0},
 211     {"migrate",migrateCommand,6,"aw",0,NULL,0,0,0,0,0},
 212     {"dump",dumpCommand,2,"ar",0,NULL,0,0,0,0,0},
 213     {"object",objectCommand,-2,"r",0,NULL,0,0,0,0,0},
 214     {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
 215     {"eval",evalCommand,-3,"wms",0,zunionInterGetKeys,0,0,0,0,0},
 216     {"evalsha",evalShaCommand,-3,"wms",0,zunionInterGetKeys,0,0,0,0,0},
 217     {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0}
 218 };
 219
 220 /*============================ Utility functions ============================ */
 221
 222 /* Low level logging. To use only for very big messages, otherwise
 223  * redisLog() is to prefer. */
 224 void redisLogRaw(int level, const char *msg) {
 225     const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING };
 226     const char *c = ".-*#";
 227     time_t now = time(NULL);
 228     FILE *fp;
 229     char buf[64];
 230     int rawmode = (level & REDIS_LOG_RAW);
 231
 232     level &= 0xff; /* clear flags */
 233     if (level < server.verbosity) return;
 234
 235     fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a");
 236     if (!fp) return;
 237
 238     if (rawmode) {
 239         fprintf(fp,"%s",msg);
 240     } else {
 241         strftime(buf,sizeof(buf),"%d %b %H:%M:%S",localtime(&now));
 242         fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg);
 243     }
 244     fflush(fp);
 245
 246     if (server.logfile) fclose(fp);
 247
 248     if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg);
 249 }
 250
 251 /* Like redisLogRaw() but with printf-alike support. This is the funciton that
 252  * is used across the code. The raw version is only used in order to dump
 253  * the INFO output on crash. */
 254 void redisLog(int level, const char *fmt, ...) {
 255     va_list ap;
 256     char msg[REDIS_MAX_LOGMSG_LEN];
 257
 258     if ((level&0xff) < server.verbosity) return;
 259
 260     va_start(ap, fmt);
 261     vsnprintf(msg, sizeof(msg), fmt, ap);
 262     va_end(ap);
 263
 264     redisLogRaw(level,msg);
 265 }
 266
 267 /* Redis generally does not try to recover from out of memory conditions
 268  * when allocating objects or strings, it is not clear if it will be possible
 269  * to report this condition to the client since the networking layer itself
 270  * is based on heap allocation for send buffers, so we simply abort.
 271  * At least the code will be simpler to read... */
 272 void oom(const char *msg) {
 273     redisLog(REDIS_WARNING, "%s: Out of memory\n",msg);
 274     sleep(1);
 275     abort();
 276 }
 277
 278 /* Return the UNIX time in microseconds */
 279 long long ustime(void) {
 280     struct timeval tv;
 281     long long ust;
 282
 283     gettimeofday(&tv, NULL);
 284     ust = ((long long)tv.tv_sec)*1000000;
 285     ust += tv.tv_usec;
 286     return ust;
 287 }
 288
 289 /*====================== Hash table type implementation  ==================== */
 290
 291 /* This is an hash table type that uses the SDS dynamic strings libary as
 292  * keys and radis objects as values (objects can hold SDS strings,
 293  * lists, sets). */
 294
 295 void dictVanillaFree(void *privdata, void *val)
 296 {
 297     DICT_NOTUSED(privdata);
 298     zfree(val);
 299 }
 300
 301 void dictListDestructor(void *privdata, void *val)
 302 {
 303     DICT_NOTUSED(privdata);
 304     listRelease((list*)val);
 305 }
 306
 307 int dictSdsKeyCompare(void *privdata, const void *key1,
 308         const void *key2)
 309 {
 310     int l1,l2;
 311     DICT_NOTUSED(privdata);
 312
 313     l1 = sdslen((sds)key1);
 314     l2 = sdslen((sds)key2);
 315     if (l1 != l2) return 0;
 316     return memcmp(key1, key2, l1) == 0;
 317 }
 318
 319 /* A case insensitive version used for the command lookup table. */
 320 int dictSdsKeyCaseCompare(void *privdata, const void *key1,
 321         const void *key2)
 322 {
 323     DICT_NOTUSED(privdata);
 324
 325     return strcasecmp(key1, key2) == 0;
 326 }
 327
 328 void dictRedisObjectDestructor(void *privdata, void *val)
 329 {
 330     DICT_NOTUSED(privdata);
 331
 332     if (val == NULL) return; /* Values of swapped out keys as set to NULL */
 333     decrRefCount(val);
 334 }
 335
 336 void dictSdsDestructor(void *privdata, void *val)
 337 {
 338     DICT_NOTUSED(privdata);
 339
 340     sdsfree(val);
 341 }
 342
 343 int dictObjKeyCompare(void *privdata, const void *key1,
 344         const void *key2)
 345 {
 346     const robj *o1 = key1, *o2 = key2;
 347     return dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
 348 }
 349
 350 unsigned int dictObjHash(const void *key) {
 351     const robj *o = key;
 352     return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
 353 }
 354
 355 unsigned int dictSdsHash(const void *key) {
 356     return dictGenHashFunction((unsigned char*)key, sdslen((char*)key));
 357 }
 358
 359 unsigned int dictSdsCaseHash(const void *key) {
 360     return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key));
 361 }
 362
 363 int dictEncObjKeyCompare(void *privdata, const void *key1,
 364         const void *key2)
 365 {
 366     robj *o1 = (robj*) key1, *o2 = (robj*) key2;
 367     int cmp;
 368
 369     if (o1->encoding == REDIS_ENCODING_INT &&
 370         o2->encoding == REDIS_ENCODING_INT)
 371             return o1->ptr == o2->ptr;
 372
 373     o1 = getDecodedObject(o1);
 374     o2 = getDecodedObject(o2);
 375     cmp = dictSdsKeyCompare(privdata,o1->ptr,o2->ptr);
 376     decrRefCount(o1);
 377     decrRefCount(o2);
 378     return cmp;
 379 }
 380
 381 unsigned int dictEncObjHash(const void *key) {
 382     robj *o = (robj*) key;
 383
 384     if (o->encoding == REDIS_ENCODING_RAW) {
 385         return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
 386     } else {
 387         if (o->encoding == REDIS_ENCODING_INT) {
 388             char buf[32];
 389             int len;
 390
 391             len = ll2string(buf,32,(long)o->ptr);
 392             return dictGenHashFunction((unsigned char*)buf, len);
 393         } else {
 394             unsigned int hash;
 395
 396             o = getDecodedObject(o);
 397             hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
 398             decrRefCount(o);
 399             return hash;
 400         }
 401     }
 402 }
 403
 404 /* Sets type hash table */
 405 dictType setDictType = {
 406     dictEncObjHash,            /* hash function */
 407     NULL,                      /* key dup */
 408     NULL,                      /* val dup */
 409     dictEncObjKeyCompare,      /* key compare */
 410     dictRedisObjectDestructor, /* key destructor */
 411     NULL                       /* val destructor */
 412 };
 413
 414 /* Sorted sets hash (note: a skiplist is used in addition to the hash table) */
 415 dictType zsetDictType = {
 416     dictEncObjHash,            /* hash function */
 417     NULL,                      /* key dup */
 418     NULL,                      /* val dup */
 419     dictEncObjKeyCompare,      /* key compare */
 420     dictRedisObjectDestructor, /* key destructor */
 421     NULL                       /* val destructor */
 422 };
 423
 424 /* Db->dict, keys are sds strings, vals are Redis objects. */
 425 dictType dbDictType = {
 426     dictSdsHash,                /* hash function */
 427     NULL,                       /* key dup */
 428     NULL,                       /* val dup */
 429     dictSdsKeyCompare,          /* key compare */
 430     dictSdsDestructor,          /* key destructor */
 431     dictRedisObjectDestructor   /* val destructor */
 432 };
 433
 434 /* Db->expires */
 435 dictType keyptrDictType = {
 436     dictSdsHash,               /* hash function */
 437     NULL,                      /* key dup */
 438     NULL,                      /* val dup */
 439     dictSdsKeyCompare,         /* key compare */
 440     NULL,                      /* key destructor */
 441     NULL                       /* val destructor */
 442 };
 443
 444 /* Command table. sds string -> command struct pointer. */
 445 dictType commandTableDictType = {
 446     dictSdsCaseHash,           /* hash function */
 447     NULL,                      /* key dup */
 448     NULL,                      /* val dup */
 449     dictSdsKeyCaseCompare,     /* key compare */
 450     dictSdsDestructor,         /* key destructor */
 451     NULL                       /* val destructor */
 452 };
 453
 454 /* Hash type hash table (note that small hashes are represented with zimpaps) */
 455 dictType hashDictType = {
 456     dictEncObjHash,             /* hash function */
 457     NULL,                       /* key dup */
 458     NULL,                       /* val dup */
 459     dictEncObjKeyCompare,       /* key compare */
 460     dictRedisObjectDestructor,  /* key destructor */
 461     dictRedisObjectDestructor   /* val destructor */
 462 };
 463
 464 /* Keylist hash table type has unencoded redis objects as keys and
 465  * lists as values. It's used for blocking operations (BLPOP) and to
 466  * map swapped keys to a list of clients waiting for this keys to be loaded. */
 467 dictType keylistDictType = {
 468     dictObjHash,                /* hash function */
 469     NULL,                       /* key dup */
 470     NULL,                       /* val dup */
 471     dictObjKeyCompare,          /* key compare */
 472     dictRedisObjectDestructor,  /* key destructor */
 473     dictListDestructor          /* val destructor */
 474 };
 475
 476 /* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to
 477  * clusterNode structures. */
 478 dictType clusterNodesDictType = {
 479     dictSdsHash,                /* hash function */
 480     NULL,                       /* key dup */
 481     NULL,                       /* val dup */
 482     dictSdsKeyCompare,          /* key compare */
 483     dictSdsDestructor,          /* key destructor */
 484     NULL                        /* val destructor */
 485 };
 486
 487 int htNeedsResize(dict *dict) {
 488     long long size, used;
 489
 490     size = dictSlots(dict);
 491     used = dictSize(dict);
 492     return (size && used && size > DICT_HT_INITIAL_SIZE &&
 493             (used*100/size < REDIS_HT_MINFILL));
 494 }
 495
 496 /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
 497  * we resize the hash table to save memory */
 498 void tryResizeHashTables(void) {
 499     int j;
 500
 501     for (j = 0; j < server.dbnum; j++) {
 502         if (htNeedsResize(server.db[j].dict))
 503             dictResize(server.db[j].dict);
 504         if (htNeedsResize(server.db[j].expires))
 505             dictResize(server.db[j].expires);
 506     }
 507 }
 508
 509 /* Our hash table implementation performs rehashing incrementally while
 510  * we write/read from the hash table. Still if the server is idle, the hash
 511  * table will use two tables for a long time. So we try to use 1 millisecond
 512  * of CPU time at every serverCron() loop in order to rehash some key. */
 513 void incrementallyRehash(void) {
 514     int j;
 515
 516     for (j = 0; j < server.dbnum; j++) {
 517         if (dictIsRehashing(server.db[j].dict)) {
 518             dictRehashMilliseconds(server.db[j].dict,1);
 519             break; /* already used our millisecond for this loop... */
 520         }
 521     }
 522 }
 523
 524 /* This function is called once a background process of some kind terminates,
 525  * as we want to avoid resizing the hash tables when there is a child in order
 526  * to play well with copy-on-write (otherwise when a resize happens lots of
 527  * memory pages are copied). The goal of this function is to update the ability
 528  * for dict.c to resize the hash tables accordingly to the fact we have o not
 529  * running childs. */
 530 void updateDictResizePolicy(void) {
 531     if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1)
 532         dictEnableResize();
 533     else
 534         dictDisableResize();
 535 }
 536
 537 /* ======================= Cron: called every 100 ms ======================== */
 538
 539 /* Try to expire a few timed out keys. The algorithm used is adaptive and
 540  * will use few CPU cycles if there are few expiring keys, otherwise
 541  * it will get more aggressive to avoid that too much memory is used by
 542  * keys that can be removed from the keyspace. */
 543 void activeExpireCycle(void) {
 544     int j;
 545
 546     for (j = 0; j < server.dbnum; j++) {
 547         int expired;
 548         redisDb *db = server.db+j;
 549
 550         /* Continue to expire if at the end of the cycle more than 25%
 551          * of the keys were expired. */
 552         do {
 553             long num = dictSize(db->expires);
 554             time_t now = time(NULL);
 555
 556             expired = 0;
 557             if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
 558                 num = REDIS_EXPIRELOOKUPS_PER_CRON;
 559             while (num--) {
 560                 dictEntry *de;
 561                 time_t t;
 562
 563                 if ((de = dictGetRandomKey(db->expires)) == NULL) break;
 564                 t = (time_t) dictGetEntryVal(de);
 565                 if (now > t) {
 566                     sds key = dictGetEntryKey(de);
 567                     robj *keyobj = createStringObject(key,sdslen(key));
 568
 569                     propagateExpire(db,keyobj);
 570                     dbDelete(db,keyobj);
 571                     decrRefCount(keyobj);
 572                     expired++;
 573                     server.stat_expiredkeys++;
 574                 }
 575             }
 576         } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
 577     }
 578 }
 579
 580 void updateLRUClock(void) {
 581     server.lruclock = (time(NULL)/REDIS_LRU_CLOCK_RESOLUTION) &
 582                                                 REDIS_LRU_CLOCK_MAX;
 583 }
 584
 585 int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
 586     int j, loops = server.cronloops;
 587     REDIS_NOTUSED(eventLoop);
 588     REDIS_NOTUSED(id);
 589     REDIS_NOTUSED(clientData);
 590
 591     /* We take a cached value of the unix time in the global state because
 592      * with virtual memory and aging there is to store the current time
 593      * in objects at every object access, and accuracy is not needed.
 594      * To access a global var is faster than calling time(NULL) */
 595     server.unixtime = time(NULL);
 596
 597     /* We have just 22 bits per object for LRU information.
 598      * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
 599      * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
 600      *
 601      * Note that even if this will wrap after 1.5 years it's not a problem,
 602      * everything will still work but just some object will appear younger
 603      * to Redis. But for this to happen a given object should never be touched
 604      * for 1.5 years.
 605      *
 606      * Note that you can change the resolution altering the
 607      * REDIS_LRU_CLOCK_RESOLUTION define.
 608      */
 609     updateLRUClock();
 610
 611     /* Record the max memory used since the server was started. */
 612     if (zmalloc_used_memory() > server.stat_peak_memory)
 613         server.stat_peak_memory = zmalloc_used_memory();
 614
 615     /* We received a SIGTERM, shutting down here in a safe way, as it is
 616      * not ok doing so inside the signal handler. */
 617     if (server.shutdown_asap) {
 618         if (prepareForShutdown() == REDIS_OK) exit(0);
 619         redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information");
 620     }
 621
 622     /* Show some info about non-empty databases */
 623     for (j = 0; j < server.dbnum; j++) {
 624         long long size, used, vkeys;
 625
 626         size = dictSlots(server.db[j].dict);
 627         used = dictSize(server.db[j].dict);
 628         vkeys = dictSize(server.db[j].expires);
 629         if (!(loops % 50) && (used || vkeys)) {
 630             redisLog(REDIS_VERBOSE,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
 631             /* dictPrintStats(server.dict); */
 632         }
 633     }
 634
 635     /* We don't want to resize the hash tables while a bacground saving
 636      * is in progress: the saving child is created using fork() that is
 637      * implemented with a copy-on-write semantic in most modern systems, so
 638      * if we resize the HT while there is the saving child at work actually
 639      * a lot of memory movements in the parent will cause a lot of pages
 640      * copied. */
 641     if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1) {
 642         if (!(loops % 10)) tryResizeHashTables();
 643         if (server.activerehashing) incrementallyRehash();
 644     }
 645
 646     /* Show information about connected clients */
 647     if (!(loops % 50)) {
 648         redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use",
 649             listLength(server.clients)-listLength(server.slaves),
 650             listLength(server.slaves),
 651             zmalloc_used_memory());
 652     }
 653
 654     /* Close connections of timedout clients */
 655     if ((server.maxidletime && !(loops % 100)) || server.bpop_blocked_clients)
 656         closeTimedoutClients();
 657
 658     /* Start a scheduled AOF rewrite if this was requested by the user while
 659      * a BGSAVE was in progress. */
 660     if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1 &&
 661         server.aofrewrite_scheduled)
 662     {
 663         rewriteAppendOnlyFileBackground();
 664     }
 665
 666     /* Check if a background saving or AOF rewrite in progress terminated. */
 667     if (server.bgsavechildpid != -1 || server.bgrewritechildpid != -1) {
 668         int statloc;
 669         pid_t pid;
 670
 671         if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) {
 672             int exitcode = WEXITSTATUS(statloc);
 673             int bysignal = 0;
 674
 675             if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc);
 676
 677             if (pid == server.bgsavechildpid) {
 678                 backgroundSaveDoneHandler(exitcode,bysignal);
 679             } else {
 680                 backgroundRewriteDoneHandler(exitcode,bysignal);
 681             }
 682             updateDictResizePolicy();
 683         }
 684     } else {
 685          time_t now = time(NULL);
 686
 687         /* If there is not a background saving/rewrite in progress check if
 688          * we have to save/rewrite now */
 689          for (j = 0; j < server.saveparamslen; j++) {
 690             struct saveparam *sp = server.saveparams+j;
 691
 692             if (server.dirty >= sp->changes &&
 693                 now-server.lastsave > sp->seconds) {
 694                 redisLog(REDIS_NOTICE,"%d changes in %d seconds. Saving...",
 695                     sp->changes, sp->seconds);
 696                 rdbSaveBackground(server.dbfilename);
 697                 break;
 698             }
 699          }
 700
 701          /* Trigger an AOF rewrite if needed */
 702          if (server.bgsavechildpid == -1 &&
 703              server.bgrewritechildpid == -1 &&
 704              server.auto_aofrewrite_perc &&
 705              server.appendonly_current_size > server.auto_aofrewrite_min_size)
 706          {
 707             long long base = server.auto_aofrewrite_base_size ?
 708                             server.auto_aofrewrite_base_size : 1;
 709             long long growth = (server.appendonly_current_size*100/base) - 100;
 710             if (growth >= server.auto_aofrewrite_perc) {
 711                 redisLog(REDIS_NOTICE,"Starting automatic rewriting of AOF on %lld%% growth",growth);
 712                 rewriteAppendOnlyFileBackground();
 713             }
 714          }
 715     }
 716
 717
 718     /* If we postponed an AOF buffer flush, let's try to do it every time the
 719      * cron function is called. */
 720     if (server.aof_flush_postponed_start) flushAppendOnlyFile(0);
 721
 722     /* Expire a few keys per cycle, only if this is a master.
 723      * On slaves we wait for DEL operations synthesized by the master
 724      * in order to guarantee a strict consistency. */
 725     if (server.masterhost == NULL) activeExpireCycle();
 726
 727     /* Replication cron function -- used to reconnect to master and
 728      * to detect transfer failures. */
 729     if (!(loops % 10)) replicationCron();
 730
 731     /* Run other sub-systems specific cron jobs */
 732     if (server.cluster_enabled && !(loops % 10)) clusterCron();
 733
 734     server.cronloops++;
 735     return 100;
 736 }
 737
 738 /* This function gets called every time Redis is entering the
 739  * main loop of the event driven library, that is, before to sleep
 740  * for ready file descriptors. */
 741 void beforeSleep(struct aeEventLoop *eventLoop) {
 742     REDIS_NOTUSED(eventLoop);
 743     listNode *ln;
 744     redisClient *c;
 745
 746     /* Try to process pending commands for clients that were just unblocked. */
 747     while (listLength(server.unblocked_clients)) {
 748         ln = listFirst(server.unblocked_clients);
 749         redisAssert(ln != NULL);
 750         c = ln->value;
 751         listDelNode(server.unblocked_clients,ln);
 752         c->flags &= ~REDIS_UNBLOCKED;
 753
 754         /* Process remaining data in the input buffer. */
 755         if (c->querybuf && sdslen(c->querybuf) > 0)
 756             processInputBuffer(c);
 757     }
 758
 759     /* Write the AOF buffer on disk */
 760     flushAppendOnlyFile(0);
 761 }
 762
 763 /* =========================== Server initialization ======================== */
 764
 765 void createSharedObjects(void) {
 766     int j;
 767
 768     shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n"));
 769     shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n"));
 770     shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n"));
 771     shared.emptybulk = createObject(REDIS_STRING,sdsnew("$0\r\n\r\n"));
 772     shared.czero = createObject(REDIS_STRING,sdsnew(":0\r\n"));
 773     shared.cone = createObject(REDIS_STRING,sdsnew(":1\r\n"));
 774     shared.cnegone = createObject(REDIS_STRING,sdsnew(":-1\r\n"));
 775     shared.nullbulk = createObject(REDIS_STRING,sdsnew("$-1\r\n"));
 776     shared.nullmultibulk = createObject(REDIS_STRING,sdsnew("*-1\r\n"));
 777     shared.emptymultibulk = createObject(REDIS_STRING,sdsnew("*0\r\n"));
 778     shared.pong = createObject(REDIS_STRING,sdsnew("+PONG\r\n"));
 779     shared.queued = createObject(REDIS_STRING,sdsnew("+QUEUED\r\n"));
 780     shared.wrongtypeerr = createObject(REDIS_STRING,sdsnew(
 781         "-ERR Operation against a key holding the wrong kind of value\r\n"));
 782     shared.nokeyerr = createObject(REDIS_STRING,sdsnew(
 783         "-ERR no such key\r\n"));
 784     shared.syntaxerr = createObject(REDIS_STRING,sdsnew(
 785         "-ERR syntax error\r\n"));
 786     shared.sameobjecterr = createObject(REDIS_STRING,sdsnew(
 787         "-ERR source and destination objects are the same\r\n"));
 788     shared.outofrangeerr = createObject(REDIS_STRING,sdsnew(
 789         "-ERR index out of range\r\n"));
 790     shared.noscripterr = createObject(REDIS_STRING,sdsnew(
 791         "-NOSCRIPT No matching script. Please use EVAL.\r\n"));
 792     shared.loadingerr = createObject(REDIS_STRING,sdsnew(
 793         "-LOADING Redis is loading the dataset in memory\r\n"));
 794     shared.space = createObject(REDIS_STRING,sdsnew(" "));
 795     shared.colon = createObject(REDIS_STRING,sdsnew(":"));
 796     shared.plus = createObject(REDIS_STRING,sdsnew("+"));
 797     shared.select0 = createStringObject("select 0\r\n",10);
 798     shared.select1 = createStringObject("select 1\r\n",10);
 799     shared.select2 = createStringObject("select 2\r\n",10);
 800     shared.select3 = createStringObject("select 3\r\n",10);
 801     shared.select4 = createStringObject("select 4\r\n",10);
 802     shared.select5 = createStringObject("select 5\r\n",10);
 803     shared.select6 = createStringObject("select 6\r\n",10);
 804     shared.select7 = createStringObject("select 7\r\n",10);
 805     shared.select8 = createStringObject("select 8\r\n",10);
 806     shared.select9 = createStringObject("select 9\r\n",10);
 807     shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13);
 808     shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14);
 809     shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15);
 810     shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18);
 811     shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17);
 812     shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19);
 813     shared.mbulk3 = createStringObject("*3\r\n",4);
 814     shared.mbulk4 = createStringObject("*4\r\n",4);
 815     for (j = 0; j < REDIS_SHARED_INTEGERS; j++) {
 816         shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j);
 817         shared.integers[j]->encoding = REDIS_ENCODING_INT;
 818     }
 819 }
 820
 821 void initServerConfig() {
 822     server.port = REDIS_SERVERPORT;
 823     server.bindaddr = NULL;
 824     server.unixsocket = NULL;
 825     server.ipfd = -1;
 826     server.sofd = -1;
 827     server.dbnum = REDIS_DEFAULT_DBNUM;
 828     server.verbosity = REDIS_VERBOSE;
 829     server.maxidletime = REDIS_MAXIDLETIME;
 830     server.saveparams = NULL;
 831     server.loading = 0;
 832     server.logfile = NULL; /* NULL = log on standard output */
 833     server.syslog_enabled = 0;
 834     server.syslog_ident = zstrdup("redis");
 835     server.syslog_facility = LOG_LOCAL0;
 836     server.daemonize = 0;
 837     server.appendonly = 0;
 838     server.appendfsync = APPENDFSYNC_EVERYSEC;
 839     server.no_appendfsync_on_rewrite = 0;
 840     server.auto_aofrewrite_perc = REDIS_AUTO_AOFREWRITE_PERC;
 841     server.auto_aofrewrite_min_size = REDIS_AUTO_AOFREWRITE_MIN_SIZE;
 842     server.auto_aofrewrite_base_size = 0;
 843     server.aofrewrite_scheduled = 0;
 844     server.lastfsync = time(NULL);
 845     server.appendfd = -1;
 846     server.appendseldb = -1; /* Make sure the first time will not match */
 847     server.aof_flush_postponed_start = 0;
 848     server.pidfile = zstrdup("/var/run/redis.pid");
 849     server.dbfilename = zstrdup("dump.rdb");
 850     server.appendfilename = zstrdup("appendonly.aof");
 851     server.requirepass = NULL;
 852     server.rdbcompression = 1;
 853     server.activerehashing = 1;
 854     server.maxclients = 0;
 855     server.bpop_blocked_clients = 0;
 856     server.maxmemory = 0;
 857     server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
 858     server.maxmemory_samples = 3;
 859     server.hash_max_zipmap_entries = REDIS_HASH_MAX_ZIPMAP_ENTRIES;
 860     server.hash_max_zipmap_value = REDIS_HASH_MAX_ZIPMAP_VALUE;
 861     server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
 862     server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
 863     server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
 864     server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES;
 865     server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE;
 866     server.shutdown_asap = 0;
 867     server.cluster_enabled = 0;
 868     server.cluster.configfile = zstrdup("nodes.conf");
 869     server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
 870
 871     updateLRUClock();
 872     resetServerSaveParams();
 873
 874     appendServerSaveParams(60*60,1);  /* save after 1 hour and 1 change */
 875     appendServerSaveParams(300,100);  /* save after 5 minutes and 100 changes */
 876     appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
 877     /* Replication related */
 878     server.isslave = 0;
 879     server.masterauth = NULL;
 880     server.masterhost = NULL;
 881     server.masterport = 6379;
 882     server.master = NULL;
 883     server.replstate = REDIS_REPL_NONE;
 884     server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT;
 885     server.repl_serve_stale_data = 1;
 886     server.repl_down_since = -1;
 887
 888     /* Double constants initialization */
 889     R_Zero = 0.0;
 890     R_PosInf = 1.0/R_Zero;
 891     R_NegInf = -1.0/R_Zero;
 892     R_Nan = R_Zero/R_Zero;
 893
 894     /* Command table -- we intiialize it here as it is part of the
 895      * initial configuration, since command names may be changed via
 896      * redis.conf using the rename-command directive. */
 897     server.commands = dictCreate(&commandTableDictType,NULL);
 898     populateCommandTable();
 899     server.delCommand = lookupCommandByCString("del");
 900     server.multiCommand = lookupCommandByCString("multi");
 901
 902     /* Slow log */
 903     server.slowlog_log_slower_than = REDIS_SLOWLOG_LOG_SLOWER_THAN;
 904     server.slowlog_max_len = REDIS_SLOWLOG_MAX_LEN;
 905 }
 906
 907 void initServer() {
 908     int j;
 909
 910     signal(SIGHUP, SIG_IGN);
 911     signal(SIGPIPE, SIG_IGN);
 912     setupSignalHandlers();
 913
 914     if (server.syslog_enabled) {
 915         openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT,
 916             server.syslog_facility);
 917     }
 918
 919     server.clients = listCreate();
 920     server.slaves = listCreate();
 921     server.monitors = listCreate();
 922     server.unblocked_clients = listCreate();
 923
 924     createSharedObjects();
 925     server.el = aeCreateEventLoop();
 926     server.db = zmalloc(sizeof(redisDb)*server.dbnum);
 927
 928     if (server.port != 0) {
 929         server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr);
 930         if (server.ipfd == ANET_ERR) {
 931             redisLog(REDIS_WARNING, "Opening port %d: %s",
 932                 server.port, server.neterr);
 933             exit(1);
 934         }
 935     }
 936     if (server.unixsocket != NULL) {
 937         unlink(server.unixsocket); /* don't care if this fails */
 938         server.sofd = anetUnixServer(server.neterr,server.unixsocket);
 939         if (server.sofd == ANET_ERR) {
 940             redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr);
 941             exit(1);
 942         }
 943     }
 944     if (server.ipfd < 0 && server.sofd < 0) {
 945         redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting.");
 946         exit(1);
 947     }
 948     for (j = 0; j < server.dbnum; j++) {
 949         server.db[j].dict = dictCreate(&dbDictType,NULL);
 950         server.db[j].expires = dictCreate(&keyptrDictType,NULL);
 951         server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL);
 952         server.db[j].watched_keys = dictCreate(&keylistDictType,NULL);
 953         server.db[j].id = j;
 954     }
 955     server.pubsub_channels = dictCreate(&keylistDictType,NULL);
 956     server.pubsub_patterns = listCreate();
 957     listSetFreeMethod(server.pubsub_patterns,freePubsubPattern);
 958     listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern);
 959     server.cronloops = 0;
 960     server.bgsavechildpid = -1;
 961     server.bgrewritechildpid = -1;
 962     server.bgrewritebuf = sdsempty();
 963     server.aofbuf = sdsempty();
 964     server.lastsave = time(NULL);
 965     server.dirty = 0;
 966     server.stat_numcommands = 0;
 967     server.stat_numconnections = 0;
 968     server.stat_expiredkeys = 0;
 969     server.stat_evictedkeys = 0;
 970     server.stat_starttime = time(NULL);
 971     server.stat_keyspace_misses = 0;
 972     server.stat_keyspace_hits = 0;
 973     server.stat_peak_memory = 0;
 974     server.stat_fork_time = 0;
 975     server.unixtime = time(NULL);
 976     aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
 977     if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE,
 978         acceptTcpHandler,NULL) == AE_ERR) oom("creating file event");
 979     if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE,
 980         acceptUnixHandler,NULL) == AE_ERR) oom("creating file event");
 981
 982     if (server.appendonly) {
 983         server.appendfd = open(server.appendfilename,O_WRONLY|O_APPEND|O_CREAT,0644);
 984         if (server.appendfd == -1) {
 985             redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
 986                 strerror(errno));
 987             exit(1);
 988         }
 989     }
 990
 991     if (server.cluster_enabled) clusterInit();
 992     scriptingInit();
 993     slowlogInit();
 994     bioInit();
 995     srand(time(NULL)^getpid());
 996 }
 997
 998 /* Populates the Redis Command Table starting from the hard coded list
 999  * we have on top of redis.c file. */
1000 void populateCommandTable(void) {
1001     int j;
1002     int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1003
1004     for (j = 0; j < numcommands; j++) {
1005         struct redisCommand *c = redisCommandTable+j;
1006         char *f = c->sflags;
1007         int retval;
1008
1009         while(*f != '\0') {
1010             switch(*f) {
1011             case 'w': c->flags |= REDIS_CMD_WRITE; break;
1012             case 'r': c->flags |= REDIS_CMD_READONLY; break;
1013             case 'm': c->flags |= REDIS_CMD_DENYOOM; break;
1014             case 'a': c->flags |= REDIS_CMD_ADMIN; break;
1015             case 'p': c->flags |= REDIS_CMD_PUBSUB; break;
1016             case 'f': c->flags |= REDIS_CMD_FORCE_REPLICATION; break;
1017             case 's': c->flags |= REDIS_CMD_NOSCRIPT; break;
1018             case 'R': c->flags |= REDIS_CMD_RANDOM; break;
1019             default: redisPanic("Unsupported command flag"); break;
1020             }
1021             f++;
1022         }
1023
1024         retval = dictAdd(server.commands, sdsnew(c->name), c);
1025         assert(retval == DICT_OK);
1026     }
1027 }
1028
1029 void resetCommandTableStats(void) {
1030     int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1031     int j;
1032
1033     for (j = 0; j < numcommands; j++) {
1034         struct redisCommand *c = redisCommandTable+j;
1035
1036         c->microseconds = 0;
1037         c->calls = 0;
1038     }
1039 }
1040
1041 /* ====================== Commands lookup and execution ===================== */
1042
1043 struct redisCommand *lookupCommand(sds name) {
1044     return dictFetchValue(server.commands, name);
1045 }
1046
1047 struct redisCommand *lookupCommandByCString(char *s) {
1048     struct redisCommand *cmd;
1049     sds name = sdsnew(s);
1050
1051     cmd = dictFetchValue(server.commands, name);
1052     sdsfree(name);
1053     return cmd;
1054 }
1055
1056 /* Call() is the core of Redis execution of a command */
1057 void call(redisClient *c) {
1058     long long dirty, start = ustime(), duration;
1059
1060     dirty = server.dirty;
1061     c->cmd->proc(c);
1062     dirty = server.dirty-dirty;
1063     duration = ustime()-start;
1064     c->cmd->microseconds += duration;
1065     slowlogPushEntryIfNeeded(c->argv,c->argc,duration);
1066     c->cmd->calls++;
1067
1068     if (server.appendonly && dirty > 0)
1069         feedAppendOnlyFile(c->cmd,c->db->id,c->argv,c->argc);
1070     if ((dirty > 0 || c->cmd->flags & REDIS_CMD_FORCE_REPLICATION) &&
1071         listLength(server.slaves))
1072         replicationFeedSlaves(server.slaves,c->db->id,c->argv,c->argc);
1073     if (listLength(server.monitors))
1074         replicationFeedMonitors(server.monitors,c->db->id,c->argv,c->argc);
1075     server.stat_numcommands++;
1076 }
1077
1078 /* If this function gets called we already read a whole
1079  * command, argments are in the client argv/argc fields.
1080  * processCommand() execute the command or prepare the
1081  * server for a bulk read from the client.
1082  *
1083  * If 1 is returned the client is still alive and valid and
1084  * and other operations can be performed by the caller. Otherwise
1085  * if 0 is returned the client was destroied (i.e. after QUIT). */
1086 int processCommand(redisClient *c) {
1087     /* The QUIT command is handled separately. Normal command procs will
1088      * go through checking for replication and QUIT will cause trouble
1089      * when FORCE_REPLICATION is enabled and would be implemented in
1090      * a regular command proc. */
1091     if (!strcasecmp(c->argv[0]->ptr,"quit")) {
1092         addReply(c,shared.ok);
1093         c->flags |= REDIS_CLOSE_AFTER_REPLY;
1094         return REDIS_ERR;
1095     }
1096
1097     /* Now lookup the command and check ASAP about trivial error conditions
1098      * such as wrong arity, bad command name and so forth. */
1099     c->cmd = lookupCommand(c->argv[0]->ptr);
1100     if (!c->cmd) {
1101         addReplyErrorFormat(c,"unknown command '%s'",
1102             (char*)c->argv[0]->ptr);
1103         return REDIS_OK;
1104     } else if ((c->cmd->arity > 0 && c->cmd->arity != c->argc) ||
1105                (c->argc < -c->cmd->arity)) {
1106         addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
1107             c->cmd->name);
1108         return REDIS_OK;
1109     }
1110
1111     /* Check if the user is authenticated */
1112     if (server.requirepass && !c->authenticated && c->cmd->proc != authCommand)
1113     {
1114         addReplyError(c,"operation not permitted");
1115         return REDIS_OK;
1116     }
1117
1118     /* If cluster is enabled, redirect here */
1119     if (server.cluster_enabled &&
1120                 !(c->cmd->getkeys_proc == NULL && c->cmd->firstkey == 0)) {
1121         int hashslot;
1122
1123         if (server.cluster.state != REDIS_CLUSTER_OK) {
1124             addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information");
1125             return REDIS_OK;
1126         } else {
1127             int ask;
1128             clusterNode *n = getNodeByQuery(c,c->cmd,c->argv,c->argc,&hashslot,&ask);
1129             if (n == NULL) {
1130                 addReplyError(c,"Multi keys request invalid in cluster");
1131                 return REDIS_OK;
1132             } else if (n != server.cluster.myself) {
1133                 addReplySds(c,sdscatprintf(sdsempty(),
1134                     "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED",
1135                     hashslot,n->ip,n->port));
1136                 return REDIS_OK;
1137             }
1138         }
1139     }
1140
1141     /* Handle the maxmemory directive.
1142      *
1143      * First we try to free some memory if possible (if there are volatile
1144      * keys in the dataset). If there are not the only thing we can do
1145      * is returning an error. */
1146     if (server.maxmemory) freeMemoryIfNeeded();
1147     if (server.maxmemory && (c->cmd->flags & REDIS_CMD_DENYOOM) &&
1148         zmalloc_used_memory() > server.maxmemory)
1149     {
1150         addReplyError(c,"command not allowed when used memory > 'maxmemory'");
1151         return REDIS_OK;
1152     }
1153
1154     /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
1155     if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
1156         &&
1157         c->cmd->proc != subscribeCommand &&
1158         c->cmd->proc != unsubscribeCommand &&
1159         c->cmd->proc != psubscribeCommand &&
1160         c->cmd->proc != punsubscribeCommand) {
1161         addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
1162         return REDIS_OK;
1163     }
1164
1165     /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
1166      * we are a slave with a broken link with master. */
1167     if (server.masterhost && server.replstate != REDIS_REPL_CONNECTED &&
1168         server.repl_serve_stale_data == 0 &&
1169         c->cmd->proc != infoCommand && c->cmd->proc != slaveofCommand)
1170     {
1171         addReplyError(c,
1172             "link with MASTER is down and slave-serve-stale-data is set to no");
1173         return REDIS_OK;
1174     }
1175
1176     /* Loading DB? Return an error if the command is not INFO */
1177     if (server.loading && c->cmd->proc != infoCommand) {
1178         addReply(c, shared.loadingerr);
1179         return REDIS_OK;
1180     }
1181
1182     /* Exec the command */
1183     if (c->flags & REDIS_MULTI &&
1184         c->cmd->proc != execCommand && c->cmd->proc != discardCommand &&
1185         c->cmd->proc != multiCommand && c->cmd->proc != watchCommand)
1186     {
1187         queueMultiCommand(c);
1188         addReply(c,shared.queued);
1189     } else {
1190         call(c);
1191     }
1192     return REDIS_OK;
1193 }
1194
1195 /*================================== Shutdown =============================== */
1196
1197 int prepareForShutdown() {
1198     redisLog(REDIS_WARNING,"User requested shutdown...");
1199     /* Kill the saving child if there is a background saving in progress.
1200        We want to avoid race conditions, for instance our saving child may
1201        overwrite the synchronous saving did by SHUTDOWN. */
1202     if (server.bgsavechildpid != -1) {
1203         redisLog(REDIS_WARNING,"There is a child saving an .rdb. Killing it!");
1204         kill(server.bgsavechildpid,SIGKILL);
1205         rdbRemoveTempFile(server.bgsavechildpid);
1206     }
1207     if (server.appendonly) {
1208         /* Kill the AOF saving child as the AOF we already have may be longer
1209          * but contains the full dataset anyway. */
1210         if (server.bgrewritechildpid != -1) {
1211             redisLog(REDIS_WARNING,
1212                 "There is a child rewriting the AOF. Killing it!");
1213             kill(server.bgrewritechildpid,SIGKILL);
1214         }
1215         /* Append only file: fsync() the AOF and exit */
1216         redisLog(REDIS_NOTICE,"Calling fsync() on the AOF file.");
1217         aof_fsync(server.appendfd);
1218     }
1219     if (server.saveparamslen > 0) {
1220         redisLog(REDIS_NOTICE,"Saving the final RDB snapshot before exiting.");
1221         /* Snapshotting. Perform a SYNC SAVE and exit */
1222         if (rdbSave(server.dbfilename) != REDIS_OK) {
1223             /* Ooops.. error saving! The best we can do is to continue
1224              * operating. Note that if there was a background saving process,
1225              * in the next cron() Redis will be notified that the background
1226              * saving aborted, handling special stuff like slaves pending for
1227              * synchronization... */
1228             redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit.");
1229             return REDIS_ERR;
1230         }
1231     }
1232     if (server.daemonize) {
1233         redisLog(REDIS_NOTICE,"Removing the pid file.");
1234         unlink(server.pidfile);
1235     }
1236     /* Close the listening sockets. Apparently this allows faster restarts. */
1237     if (server.ipfd != -1) close(server.ipfd);
1238     if (server.sofd != -1) close(server.sofd);
1239
1240     redisLog(REDIS_WARNING,"Redis is now ready to exit, bye bye...");
1241     return REDIS_OK;
1242 }
1243
1244 /*================================== Commands =============================== */
1245
1246 void authCommand(redisClient *c) {
1247     if (!server.requirepass || !strcmp(c->argv[1]->ptr, server.requirepass)) {
1248       c->authenticated = 1;
1249       addReply(c,shared.ok);
1250     } else {
1251       c->authenticated = 0;
1252       addReplyError(c,"invalid password");
1253     }
1254 }
1255
1256 void pingCommand(redisClient *c) {
1257     addReply(c,shared.pong);
1258 }
1259
1260 void echoCommand(redisClient *c) {
1261     addReplyBulk(c,c->argv[1]);
1262 }
1263
1264 /* Convert an amount of bytes into a human readable string in the form
1265  * of 100B, 2G, 100M, 4K, and so forth. */
1266 void bytesToHuman(char *s, unsigned long long n) {
1267     double d;
1268
1269     if (n < 1024) {
1270         /* Bytes */
1271         sprintf(s,"%lluB",n);
1272         return;
1273     } else if (n < (1024*1024)) {
1274         d = (double)n/(1024);
1275         sprintf(s,"%.2fK",d);
1276     } else if (n < (1024LL*1024*1024)) {
1277         d = (double)n/(1024*1024);
1278         sprintf(s,"%.2fM",d);
1279     } else if (n < (1024LL*1024*1024*1024)) {
1280         d = (double)n/(1024LL*1024*1024);
1281         sprintf(s,"%.2fG",d);
1282     }
1283 }
1284
1285 /* Create the string returned by the INFO command. This is decoupled
1286  * by the INFO command itself as we need to report the same information
1287  * on memory corruption problems. */
1288 sds genRedisInfoString(char *section) {
1289     sds info = sdsempty();
1290     time_t uptime = time(NULL)-server.stat_starttime;
1291     int j, numcommands;
1292     struct rusage self_ru, c_ru;
1293     unsigned long lol, bib;
1294     int allsections = 0, defsections = 0;
1295     int sections = 0;
1296
1297     if (section) {
1298         allsections = strcasecmp(section,"all") == 0;
1299         defsections = strcasecmp(section,"default") == 0;
1300     }
1301
1302     getrusage(RUSAGE_SELF, &self_ru);
1303     getrusage(RUSAGE_CHILDREN, &c_ru);
1304     getClientsMaxBuffers(&lol,&bib);
1305
1306     /* Server */
1307     if (allsections || defsections || !strcasecmp(section,"server")) {
1308         if (sections++) info = sdscat(info,"\r\n");
1309         info = sdscatprintf(info,
1310             "# Server\r\n"
1311             "redis_version:%s\r\n"
1312             "redis_git_sha1:%s\r\n"
1313             "redis_git_dirty:%d\r\n"
1314             "arch_bits:%s\r\n"
1315             "multiplexing_api:%s\r\n"
1316             "process_id:%ld\r\n"
1317             "tcp_port:%d\r\n"
1318             "uptime_in_seconds:%ld\r\n"
1319             "uptime_in_days:%ld\r\n"
1320             "lru_clock:%ld\r\n",
1321             REDIS_VERSION,
1322             redisGitSHA1(),
1323             strtol(redisGitDirty(),NULL,10) > 0,
1324             (sizeof(long) == 8) ? "64" : "32",
1325             aeGetApiName(),
1326             (long) getpid(),
1327             server.port,
1328             uptime,
1329             uptime/(3600*24),
1330             (unsigned long) server.lruclock);
1331     }
1332
1333     /* Clients */
1334     if (allsections || defsections || !strcasecmp(section,"clients")) {
1335         if (sections++) info = sdscat(info,"\r\n");
1336         info = sdscatprintf(info,
1337             "# Clients\r\n"
1338             "connected_clients:%d\r\n"
1339             "client_longest_output_list:%lu\r\n"
1340             "client_biggest_input_buf:%lu\r\n"
1341             "blocked_clients:%d\r\n",
1342             listLength(server.clients)-listLength(server.slaves),
1343             lol, bib,
1344             server.bpop_blocked_clients);
1345     }
1346
1347     /* Memory */
1348     if (allsections || defsections || !strcasecmp(section,"memory")) {
1349         char hmem[64];
1350         char peak_hmem[64];
1351
1352         bytesToHuman(hmem,zmalloc_used_memory());
1353         bytesToHuman(peak_hmem,server.stat_peak_memory);
1354         if (sections++) info = sdscat(info,"\r\n");
1355         info = sdscatprintf(info,
1356             "# Memory\r\n"
1357             "used_memory:%zu\r\n"
1358             "used_memory_human:%s\r\n"
1359             "used_memory_rss:%zu\r\n"
1360             "used_memory_peak:%zu\r\n"
1361             "used_memory_peak_human:%s\r\n"
1362             "used_memory_lua:%lld\r\n"
1363             "mem_fragmentation_ratio:%.2f\r\n"
1364             "mem_allocator:%s\r\n",
1365             zmalloc_used_memory(),
1366             hmem,
1367             zmalloc_get_rss(),
1368             server.stat_peak_memory,
1369             peak_hmem,
1370             ((long long)lua_gc(server.lua,LUA_GCCOUNT,0))*1024LL,
1371             zmalloc_get_fragmentation_ratio(),
1372             ZMALLOC_LIB
1373             );
1374     }
1375
1376     /* Persistence */
1377     if (allsections || defsections || !strcasecmp(section,"persistence")) {
1378         if (sections++) info = sdscat(info,"\r\n");
1379         info = sdscatprintf(info,
1380             "# Persistence\r\n"
1381             "loading:%d\r\n"
1382             "aof_enabled:%d\r\n"
1383             "changes_since_last_save:%lld\r\n"
1384             "bgsave_in_progress:%d\r\n"
1385             "last_save_time:%ld\r\n"
1386             "bgrewriteaof_in_progress:%d\r\n",
1387             server.loading,
1388             server.appendonly,
1389             server.dirty,
1390             server.bgsavechildpid != -1,
1391             server.lastsave,
1392             server.bgrewritechildpid != -1);
1393
1394         if (server.appendonly) {
1395             info = sdscatprintf(info,
1396                 "aof_current_size:%lld\r\n"
1397                 "aof_base_size:%lld\r\n"
1398                 "aof_pending_rewrite:%d\r\n",
1399                 (long long) server.appendonly_current_size,
1400                 (long long) server.auto_aofrewrite_base_size,
1401                 server.aofrewrite_scheduled);
1402         }
1403
1404         if (server.loading) {
1405             double perc;
1406             time_t eta, elapsed;
1407             off_t remaining_bytes = server.loading_total_bytes-
1408                                     server.loading_loaded_bytes;
1409
1410             perc = ((double)server.loading_loaded_bytes /
1411                    server.loading_total_bytes) * 100;
1412
1413             elapsed = time(NULL)-server.loading_start_time;
1414             if (elapsed == 0) {
1415                 eta = 1; /* A fake 1 second figure if we don't have
1416                             enough info */
1417             } else {
1418                 eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes;
1419             }
1420
1421             info = sdscatprintf(info,
1422                 "loading_start_time:%ld\r\n"
1423                 "loading_total_bytes:%llu\r\n"
1424                 "loading_loaded_bytes:%llu\r\n"
1425                 "loading_loaded_perc:%.2f\r\n"
1426                 "loading_eta_seconds:%ld\r\n"
1427                 ,(unsigned long) server.loading_start_time,
1428                 (unsigned long long) server.loading_total_bytes,
1429                 (unsigned long long) server.loading_loaded_bytes,
1430                 perc,
1431                 eta
1432             );
1433         }
1434     }
1435
1436     /* Stats */
1437     if (allsections || defsections || !strcasecmp(section,"stats")) {
1438         if (sections++) info = sdscat(info,"\r\n");
1439         info = sdscatprintf(info,
1440             "# Stats\r\n"
1441             "total_connections_received:%lld\r\n"
1442             "total_commands_processed:%lld\r\n"
1443             "expired_keys:%lld\r\n"
1444             "evicted_keys:%lld\r\n"
1445             "keyspace_hits:%lld\r\n"
1446             "keyspace_misses:%lld\r\n"
1447             "pubsub_channels:%ld\r\n"
1448             "pubsub_patterns:%u\r\n"
1449             "latest_fork_usec:%lld\r\n",
1450             server.stat_numconnections,
1451             server.stat_numcommands,
1452             server.stat_expiredkeys,
1453             server.stat_evictedkeys,
1454             server.stat_keyspace_hits,
1455             server.stat_keyspace_misses,
1456             dictSize(server.pubsub_channels),
1457             listLength(server.pubsub_patterns),
1458             server.stat_fork_time);
1459     }
1460
1461     /* Replication */
1462     if (allsections || defsections || !strcasecmp(section,"replication")) {
1463         if (sections++) info = sdscat(info,"\r\n");
1464         info = sdscatprintf(info,
1465             "# Replication\r\n"
1466             "role:%s\r\n",
1467             server.masterhost == NULL ? "master" : "slave");
1468         if (server.masterhost) {
1469             info = sdscatprintf(info,
1470                 "master_host:%s\r\n"
1471                 "master_port:%d\r\n"
1472                 "master_link_status:%s\r\n"
1473                 "master_last_io_seconds_ago:%d\r\n"
1474                 "master_sync_in_progress:%d\r\n"
1475                 ,server.masterhost,
1476                 server.masterport,
1477                 (server.replstate == REDIS_REPL_CONNECTED) ?
1478                     "up" : "down",
1479                 server.master ?
1480                 ((int)(time(NULL)-server.master->lastinteraction)) : -1,
1481                 server.replstate == REDIS_REPL_TRANSFER
1482             );
1483
1484             if (server.replstate == REDIS_REPL_TRANSFER) {
1485                 info = sdscatprintf(info,
1486                     "master_sync_left_bytes:%ld\r\n"
1487                     "master_sync_last_io_seconds_ago:%d\r\n"
1488                     ,(long)server.repl_transfer_left,
1489                     (int)(time(NULL)-server.repl_transfer_lastio)
1490                 );
1491             }
1492
1493             if (server.replstate != REDIS_REPL_CONNECTED) {
1494                 info = sdscatprintf(info,
1495                     "master_link_down_since_seconds:%ld\r\n",
1496                     (long)time(NULL)-server.repl_down_since);
1497             }
1498         }
1499         info = sdscatprintf(info,
1500             "connected_slaves:%d\r\n",
1501             listLength(server.slaves));
1502     }
1503
1504     /* CPU */
1505     if (allsections || defsections || !strcasecmp(section,"cpu")) {
1506         if (sections++) info = sdscat(info,"\r\n");
1507         info = sdscatprintf(info,
1508         "# CPU\r\n"
1509         "used_cpu_sys:%.2f\r\n"
1510         "used_cpu_user:%.2f\r\n"
1511         "used_cpu_sys_children:%.2f\r\n"
1512         "used_cpu_user_children:%.2f\r\n",
1513         (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
1514         (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
1515         (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000,
1516         (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000);
1517     }
1518
1519     /* cmdtime */
1520     if (allsections || !strcasecmp(section,"commandstats")) {
1521         if (sections++) info = sdscat(info,"\r\n");
1522         info = sdscatprintf(info, "# Commandstats\r\n");
1523         numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand);
1524         for (j = 0; j < numcommands; j++) {
1525             struct redisCommand *c = redisCommandTable+j;
1526
1527             if (!c->calls) continue;
1528             info = sdscatprintf(info,
1529                 "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n",
1530                 c->name, c->calls, c->microseconds,
1531                 (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls));
1532         }
1533     }
1534
1535     /* Clusetr */
1536     if (allsections || defsections || !strcasecmp(section,"cluster")) {
1537         if (sections++) info = sdscat(info,"\r\n");
1538         info = sdscatprintf(info,
1539         "# Cluster\r\n"
1540         "cluster_enabled:%d\r\n",
1541         server.cluster_enabled);
1542     }
1543
1544     /* Key space */
1545     if (allsections || defsections || !strcasecmp(section,"keyspace")) {
1546         if (sections++) info = sdscat(info,"\r\n");
1547         info = sdscatprintf(info, "# Keyspace\r\n");
1548         for (j = 0; j < server.dbnum; j++) {
1549             long long keys, vkeys;
1550
1551             keys = dictSize(server.db[j].dict);
1552             vkeys = dictSize(server.db[j].expires);
1553             if (keys || vkeys) {
1554                 info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n",
1555                     j, keys, vkeys);
1556             }
1557         }
1558     }
1559     return info;
1560 }
1561
1562 void infoCommand(redisClient *c) {
1563     char *section = c->argc == 2 ? c->argv[1]->ptr : "default";
1564
1565     if (c->argc > 2) {
1566         addReply(c,shared.syntaxerr);
1567         return;
1568     }
1569     sds info = genRedisInfoString(section);
1570     addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",
1571         (unsigned long)sdslen(info)));
1572     addReplySds(c,info);
1573     addReply(c,shared.crlf);
1574 }
1575
1576 void monitorCommand(redisClient *c) {
1577     /* ignore MONITOR if aleady slave or in monitor mode */
1578     if (c->flags & REDIS_SLAVE) return;
1579
1580     c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
1581     c->slaveseldb = 0;
1582     listAddNodeTail(server.monitors,c);
1583     addReply(c,shared.ok);
1584 }
1585
1586 /* ============================ Maxmemory directive  ======================== */
1587
1588 /* This function gets called when 'maxmemory' is set on the config file to limit
1589  * the max memory used by the server, and we are out of memory.
1590  * This function will try to, in order:
1591  *
1592  * - Free objects from the free list
1593  * - Try to remove keys with an EXPIRE set
1594  *
1595  * It is not possible to free enough memory to reach used-memory < maxmemory
1596  * the server will start refusing commands that will enlarge even more the
1597  * memory usage.
1598  */
1599 void freeMemoryIfNeeded(void) {
1600     /* Remove keys accordingly to the active policy as long as we are
1601      * over the memory limit. */
1602     if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION) return;
1603
1604     while (server.maxmemory && zmalloc_used_memory() > server.maxmemory) {
1605         int j, k, freed = 0;
1606
1607         for (j = 0; j < server.dbnum; j++) {
1608             long bestval = 0; /* just to prevent warning */
1609             sds bestkey = NULL;
1610             struct dictEntry *de;
1611             redisDb *db = server.db+j;
1612             dict *dict;
1613
1614             if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
1615                 server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
1616             {
1617                 dict = server.db[j].dict;
1618             } else {
1619                 dict = server.db[j].expires;
1620             }
1621             if (dictSize(dict) == 0) continue;
1622
1623             /* volatile-random and allkeys-random policy */
1624             if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
1625                 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
1626             {
1627                 de = dictGetRandomKey(dict);
1628                 bestkey = dictGetEntryKey(de);
1629             }
1630
1631             /* volatile-lru and allkeys-lru policy */
1632             else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
1633                 server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
1634             {
1635                 for (k = 0; k < server.maxmemory_samples; k++) {
1636                     sds thiskey;
1637                     long thisval;
1638                     robj *o;
1639
1640                     de = dictGetRandomKey(dict);
1641                     thiskey = dictGetEntryKey(de);
1642                     /* When policy is volatile-lru we need an additonal lookup
1643                      * to locate the real key, as dict is set to db->expires. */
1644                     if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
1645                         de = dictFind(db->dict, thiskey);
1646                     o = dictGetEntryVal(de);
1647                     thisval = estimateObjectIdleTime(o);
1648
1649                     /* Higher idle time is better candidate for deletion */
1650                     if (bestkey == NULL || thisval > bestval) {
1651                         bestkey = thiskey;
1652                         bestval = thisval;
1653                     }
1654                 }
1655             }
1656
1657             /* volatile-ttl */
1658             else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
1659                 for (k = 0; k < server.maxmemory_samples; k++) {
1660                     sds thiskey;
1661                     long thisval;
1662
1663                     de = dictGetRandomKey(dict);
1664                     thiskey = dictGetEntryKey(de);
1665                     thisval = (long) dictGetEntryVal(de);
1666
1667                     /* Expire sooner (minor expire unix timestamp) is better
1668                      * candidate for deletion */
1669                     if (bestkey == NULL || thisval < bestval) {
1670                         bestkey = thiskey;
1671                         bestval = thisval;
1672                     }
1673                 }
1674             }
1675
1676             /* Finally remove the selected key. */
1677             if (bestkey) {
1678                 robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
1679                 propagateExpire(db,keyobj);
1680                 dbDelete(db,keyobj);
1681                 server.stat_evictedkeys++;
1682                 decrRefCount(keyobj);
1683                 freed++;
1684             }
1685         }
1686         if (!freed) return; /* nothing to free... */
1687     }
1688 }
1689
1690 /* =================================== Main! ================================ */
1691
1692 #ifdef __linux__
1693 int linuxOvercommitMemoryValue(void) {
1694     FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
1695     char buf[64];
1696
1697     if (!fp) return -1;
1698     if (fgets(buf,64,fp) == NULL) {
1699         fclose(fp);
1700         return -1;
1701     }
1702     fclose(fp);
1703
1704     return atoi(buf);
1705 }
1706
1707 void linuxOvercommitMemoryWarning(void) {
1708     if (linuxOvercommitMemoryValue() == 0) {
1709         redisLog(REDIS_WARNING,"WARNING overcommit_memory is set to 0! Background save may fail under low memory condition. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect.");
1710     }
1711 }
1712 #endif /* __linux__ */
1713
1714 void createPidFile(void) {
1715     /* Try to write the pid file in a best-effort way. */
1716     FILE *fp = fopen(server.pidfile,"w");
1717     if (fp) {
1718         fprintf(fp,"%d\n",(int)getpid());
1719         fclose(fp);
1720     }
1721 }
1722
1723 void daemonize(void) {
1724     int fd;
1725
1726     if (fork() != 0) exit(0); /* parent exits */
1727     setsid(); /* create a new session */
1728
1729     /* Every output goes to /dev/null. If Redis is daemonized but
1730      * the 'logfile' is set to 'stdout' in the configuration file
1731      * it will not log at all. */
1732     if ((fd = open("/dev/null", O_RDWR, 0)) != -1) {
1733         dup2(fd, STDIN_FILENO);
1734         dup2(fd, STDOUT_FILENO);
1735         dup2(fd, STDERR_FILENO);
1736         if (fd > STDERR_FILENO) close(fd);
1737     }
1738 }
1739
1740 void version() {
1741     printf("Redis server version %s (%s:%d)\n", REDIS_VERSION,
1742         redisGitSHA1(), atoi(redisGitDirty()) > 0);
1743     exit(0);
1744 }
1745
1746 void usage() {
1747     fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf]\n");
1748     fprintf(stderr,"       ./redis-server - (read config from stdin)\n");
1749     exit(1);
1750 }
1751
1752 void redisAsciiArt(void) {
1753 #include "asciilogo.h"
1754     char *buf = zmalloc(1024*16);
1755
1756     snprintf(buf,1024*16,ascii_logo,
1757         REDIS_VERSION,
1758         redisGitSHA1(),
1759         strtol(redisGitDirty(),NULL,10) > 0,
1760         (sizeof(long) == 8) ? "64" : "32",
1761         server.cluster_enabled ? "cluster" : "stand alone",
1762         server.port,
1763         (long) getpid()
1764     );
1765     redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf);
1766     zfree(buf);
1767 }
1768
1769 int main(int argc, char **argv) {
1770     long long start;
1771
1772     zmalloc_enable_thread_safeness();
1773     initServerConfig();
1774     if (argc == 2) {
1775         if (strcmp(argv[1], "-v") == 0 ||
1776             strcmp(argv[1], "--version") == 0) version();
1777         if (strcmp(argv[1], "--help") == 0) usage();
1778         resetServerSaveParams();
1779         loadServerConfig(argv[1]);
1780     } else if ((argc > 2)) {
1781         usage();
1782     } else {
1783         redisLog(REDIS_WARNING,"Warning: no config file specified, using the default config. In order to specify a config file use 'redis-server /path/to/redis.conf'");
1784     }
1785     if (server.daemonize) daemonize();
1786     initServer();
1787     if (server.daemonize) createPidFile();
1788     redisAsciiArt();
1789     redisLog(REDIS_NOTICE,"Server started, Redis version " REDIS_VERSION);
1790 #ifdef __linux__
1791     linuxOvercommitMemoryWarning();
1792 #endif
1793     start = ustime();
1794     if (server.appendonly) {
1795         if (loadAppendOnlyFile(server.appendfilename) == REDIS_OK)
1796             redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000);
1797     } else {
1798         if (rdbLoad(server.dbfilename) == REDIS_OK)
1799             redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",(float)(ustime()-start)/1000000);
1800     }
1801     if (server.ipfd > 0)
1802         redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
1803     if (server.sofd > 0)
1804         redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket);
1805     aeSetBeforeSleepProc(server.el,beforeSleep);
1806     aeMain(server.el);
1807     aeDeleteEventLoop(server.el);
1808     return 0;
1809 }
1810
1811 #ifdef HAVE_BACKTRACE
1812 static void *getMcontextEip(ucontext_t *uc) {
1813 #if defined(__FreeBSD__)
1814     return (void*) uc->uc_mcontext.mc_eip;
1815 #elif defined(__dietlibc__)
1816     return (void*) uc->uc_mcontext.eip;
1817 #elif defined(__APPLE__) && !defined(MAC_OS_X_VERSION_10_6)
1818   #if __x86_64__
1819     return (void*) uc->uc_mcontext->__ss.__rip;
1820   #elif __i386__
1821     return (void*) uc->uc_mcontext->__ss.__eip;
1822   #else
1823     return (void*) uc->uc_mcontext->__ss.__srr0;
1824   #endif
1825 #elif defined(__APPLE__) && defined(MAC_OS_X_VERSION_10_6)
1826   #if defined(_STRUCT_X86_THREAD_STATE64) && !defined(__i386__)
1827     return (void*) uc->uc_mcontext->__ss.__rip;
1828   #else
1829     return (void*) uc->uc_mcontext->__ss.__eip;
1830   #endif
1831 #elif defined(__i386__)
1832     return (void*) uc->uc_mcontext.gregs[14]; /* Linux 32 */
1833 #elif defined(__X86_64__) || defined(__x86_64__)
1834     return (void*) uc->uc_mcontext.gregs[16]; /* Linux 64 */
1835 #elif defined(__ia64__) /* Linux IA64 */
1836     return (void*) uc->uc_mcontext.sc_ip;
1837 #else
1838     return NULL;
1839 #endif
1840 }
1841
1842 static void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
1843     void *trace[100];
1844     char **messages = NULL;
1845     int i, trace_size = 0;
1846     ucontext_t *uc = (ucontext_t*) secret;
1847     sds infostring;
1848     struct sigaction act;
1849     REDIS_NOTUSED(info);
1850
1851     redisLog(REDIS_WARNING,
1852         "======= Ooops! Redis %s got signal: -%d- =======", REDIS_VERSION, sig);
1853     infostring = genRedisInfoString("all");
1854     redisLogRaw(REDIS_WARNING, infostring);
1855     /* It's not safe to sdsfree() the returned string under memory
1856      * corruption conditions. Let it leak as we are going to abort */
1857
1858     trace_size = backtrace(trace, 100);
1859     /* overwrite sigaction with caller's address */
1860     if (getMcontextEip(uc) != NULL) {
1861         trace[1] = getMcontextEip(uc);
1862     }
1863     messages = backtrace_symbols(trace, trace_size);
1864
1865     for (i=1; i<trace_size; ++i)
1866         redisLog(REDIS_WARNING,"%s", messages[i]);
1867
1868     /* free(messages); Don't call free() with possibly corrupted memory. */
1869     if (server.daemonize) unlink(server.pidfile);
1870
1871     /* Make sure we exit with the right signal at the end. So for instance
1872      * the core will be dumped if enabled. */
1873     sigemptyset (&act.sa_mask);
1874     /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction
1875      * is used. Otherwise, sa_handler is used */
1876     act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
1877     act.sa_handler = SIG_DFL;
1878     sigaction (sig, &act, NULL);
1879     kill(getpid(),sig);
1880 }
1881 #endif /* HAVE_BACKTRACE */
1882
1883 static void sigtermHandler(int sig) {
1884     REDIS_NOTUSED(sig);
1885
1886     redisLog(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
1887     server.shutdown_asap = 1;
1888 }
1889
1890 void setupSignalHandlers(void) {
1891     struct sigaction act;
1892
1893     /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used.
1894      * Otherwise, sa_handler is used. */
1895     sigemptyset(&act.sa_mask);
1896     act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND;
1897     act.sa_handler = sigtermHandler;
1898     sigaction(SIGTERM, &act, NULL);
1899
1900 #ifdef HAVE_BACKTRACE
1901     sigemptyset(&act.sa_mask);
1902     act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
1903     act.sa_sigaction = sigsegvHandler;
1904     sigaction(SIGSEGV, &act, NULL);
1905     sigaction(SIGBUS, &act, NULL);
1906     sigaction(SIGFPE, &act, NULL);
1907     sigaction(SIGILL, &act, NULL);
1908 #endif
1909     return;
1910 }
1911
1912 /* The End */